PACIFIC SOUTHWEST FOREST SERVICE . 7 nnm U.S.DEPARTMENT· AGR Icf)6{ld 1;,,_ P.O. BOX 245, , CALIFORNIA 94 701

orest and Range xperiment Station

USDA FOREST SERVICE RESEARCH PAPER PSW-113 /1976 \

EVALUATION OF SKYLAB (EREP) DATA

FOR FOREST AND RANGELAND SURVEYS

Technical Coordinator

ROBERT C. ALDRICH

USDA FOREST SERVICE RESEARCH PAPER PSW-11 3

Published by

Pacific Southwest Forest and Range Experiment Station Berkeley, Calilornia 947D1

in cooperation with

Rocky Mountain Forest and Range Experiment Station Fort Collins, Colorado 80521

Forest Service, U.S. Department of Agriculture

1976

Technical Coord inator

ROBERT C. ALDRICH, who heads the Silnion's R~mote Sensing Research WOIk Unll. joined the Berkeley staff in 196 5. He h n graduate or Ihe State College of Forest ry, Syracuse. New York (bachelor of science, 1944; master of fo res try. 1948),

The Aun.()f,

ROBERT W. DANA, a physicist, has been a member of the Un it si .we 1969. He earned a bachelor's degree in physks (1963) Jnd master's degree in geo physics (1969) at the Unjversily of Washinglon, RICHARD S, DRISCOLL is in charge of the Range Inven tory and Evaluation Research Work Unit al the Rocky MOlln ta in Forest and Rango:l E'perimo:lnt Station, For t Collins. He studied a\ Colorado Agri­cultural and Mechanlc:tl College (baccalaureate in ran~e management , 195)), Colorado Sta te Univershy (master's degree in range manas-ement, 1957), and Oregon State Universi ty (docto rate lJ1 range ecology, 1962). RICHARD E. FRANCIS, an aS$Ociate r ang~ scient ist, is also auigned to range inventory research al the Rocky MOllntain Station. He earned a bachelor 's degree (1967) in wildli fe management at Humboldt State College, California. WAL.L.ACE J. GREENTR£E also attended Humbold t State College (bachelor of science degree in forest man· agement, 19(1). A for~Stry technician, he has been with the Pacitlc Sou thwest Stat ion's Remote Sensing Research Work Unit since 1967. Unit mathematical sta tist ician since 1966. NANCY X. NORI CK hold5 bachelo r" l I964) and master's (1 969) degrees in stat ist ics fl am the University of California, Ikrkeley. EDWIN H. ROBERTS, a rC5elmh fores ter with the Unit, is 3 fOlestry graduat~ of the Uni­versity of California. Berkeley tbachelor's, 1965 . and master's degrees, 1%9}. He joined (he Sta tion slaff in 1974. THOMAS H. WAITE, fo tmerly a Jore.~ try te.:hni· cian with the Unit , is alw a graduate of the Un iv~njl Y of California, Berkeley (bachelor of sdenco:l in forest ry, 1970) . He is now employed by Bechtel Corpo· ration, San Francisco, California. FREDER ICX P. WEBER, formerly with the Unit, is now remote sensing coordinuor, Forest Economics arod Marketing Re· search st3f1", Forest Service, Washington. D.C. lie was educated a1 the Universit y of Minnesota (baccalaureate in forest ry, 1960) and the University or Michigan (master's degree in fo restry, 1965 ; doctorate in foreSlly. 1969).

NOTE: The Remote Sensin8 Research WOlk Unit , stationed at Berkeley at the time this publication was pJepared , became in July 1976 a part o f the Resources Evaluation Techniques Prosram at FOi t CoUins, Colorado, The Program, under the direction of Richard S, Driscoll as program ffi:lnager , was cstablished to aid in Forest Service compliance With lhe Fores t and Rangeland Renewab l ~ Resources Planning AC1 of 1974, whidl requires inventory and evaluation of lJ1e forest, ran8e1and, and other renewable resources of the United States every 10 YU rs.

Co~r: An enla rged porlion (! IX) of Q false-color composite made from Sky lab S I90A mulliband photovaphs (green. red, and near infri red ronds) taken on Sep tember 11, 19H. Ctark HllI Resery()ir and the Savannah River are s11 0wn in the center. Scale is approximately 1:250,000,

il

FOREWORD

The launchi ng of Skylab in 1973 was linked with the development of large manned space workshops which may orbit the earth in the 1980's. One of these workshol's, the reusable space shuttle. would have the exci ting prospect of responding to critical renewable-resource problems wilh near-real-tlme information. Forest and range investi­gators, recovering from their disappointment with LANDSAT (ERTS-l) data, turned to Skyl ab with expectations of improved resohllioll and greater sensor sensitivity. Some of their expectat ions were rewarded, but as this report shows, reliability of the sensor systems must be improved before a space shuttl e can respond to many forest and fange problems.

This report cvaJuales the usefulness of the Skylab Earth Resources Experiment Package (EREP) data in identify ing forest, rangeland , nonforest, water, and forest stress as a fi rs t level of resource informat ion. The experiments described here wert pe rformed under Contract No. T41068 (March 7, 1973 to December 7, 1975) between the U.S. National AeronllUtics and Space Administration/Johnson Space Center (NASAjJSC) and the U.S. Depar tment of Agriculture. The research was conducted by professional staff members of the Remote Sensing Research Work Unit , Pacific Southwest Forest and Range Experiment Sta tion (PSW), Berkeley , California, and the Range Inventory and Evalualion Research Work Unit , Rocky Mountain Forest and Range Experim en t Station (RM), Fort Collins, Colorado.

Robert C. Heller was originally identified as principal investigator and Robert C. Aldrich, Richard S. Driscoll, and Frederick P. Weber were identified as coin\o'estigators. Robert C. Aldrich was m~de principal investigator upon the retirement of Mr. Heller in August 1974 . Technical 1I1·mitofs for NASA were Rybocn Kirby and Clayton D. Forbes. This Research Paper is based on the final report submitted to NASA in fulfill men t of the cont rac t.

Various portions of the Sky lab data produc ts were processed and analyzed at Ihe Pacific Southwest Station and the University of Kansas Space TeclulOlogy Center. Lawrence, Kansas.

Although this Research Paper points oul some limitations of Skylab sensors, these weaknesses probably will be resolved before the reusable space sh uttle is lau nclled in the 1980's. Meanwhile, the Forest Service will continlle to develop its capabilities to use remote senSing data to IIIventory and manage forest and relat ed resources.

ROBERT W. HARRIS , Director Pacific Southwest Forest and Range Experiment Station

ACKNOWLEDGMENTS

We gratefully acknowl edge the assist3 nce of Professor Robert C. Heller, College of Forestry, Wildlife, 3fJd Ronge Sciences, University of Idaho, who coordinated the original proposal, and untU his retirement from the Forest Service in August 1974, gave valuable technical direction to tlus research.

Members of the staffs of the two cooperating Experi.ment Stations contributed substantially to the study. Richard J. Myhre. scientific photographer, Pacific Southwest Station, was responsibl e for all photographic work required by the investigations, and prepared the illustrations included in this report.

The S 192 data analysis reported in the Forest Stress Detection section was perfonned under subcon tract by Robert HaraUck of the Remote Sensing Laboratory, University of Kansas Space Technology Center, Lawrence, Kansas, with the assistance of Gary Minden, graduale student, Department of Electrical Engineering.

Clayton D. Forbes, technical monitor, Johnson Space Cenler, was responsible for technical advice, and for expedit ing shipme nt s of Sky lab imagery and computer­compatible tapes.

Robert Mattson, forester, Black Hills National Forest, Deadwood, S.D., assisted in maintaining the field data collection system.

CONTENTS

Page

Foreword ... .........• ......• .. • .... ... ... . •.. . . . . . . .. . ..... iii

Acknowledgments ..... . •. ..... ••. .... • • ...... • .. .... • ......... iv Summary

Glossary . .. • •• ....• ••. ..• • •• . . ....• • •...• •• • . .. . . . 4

I n I rod oct ion

Objectives

. ........ . . ......•...... •. ..... .•• •... . •• ....... . 5

...... 5

Study AIeas

Data and Techniques ........ ....... ..... ........ .. . ••... . . , .. Forest Inventory : Forest Resource Evaluation. Sampling Design,

6

7

and Automated Land Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 9 by Robert C. Aldrich, Edwin H. Roberts, Wallace 1, Green/fee,

Nancy X. Norick, and Thomas H. Waite

Study Area ............ ............ ...... . ... ............ 10

Qassiftcation System ....... . •........... ... . . . . . .......... .. JO Skylab Data ....... . .. . ..•• • • .... .•• . •. . .. •. .. .. •• •• . __ . _. 10 Ground Truth ................... ..• .....•...... .•• ... _. . . . 14

Procedures .. . . . . . . . . . . . . . . . . . . . . • . . . . . . . • . . . . . . • . . . . . . . . . 15

Forest Resource Evaluation .. ... .. ..• .• . . . .• .. • . . .. •.. .. .... 15

Forest Sampling Designs ..... ... . . .••... .. . • .. . • . .•......... 19

Automated Land Classification 22 Results and Discussion ... ......... . . .... ......... . . ...• . ..... 23

Forest Resource Evaluation ....... .• ... .. . . . ..• .. . . ... .. .... 24

Forest Sampling Designs .... .......• . .. ... ..... _ . • • . . . . . . . .. 27

Automated Land Classification ..... .. • . • . . .. . .. • .. ... • . . . .... 31 Applicaltons . . . . . . . . . . . . . . . . . . . . . • . • . . . . • . . . . . . . . • . . . . . . .. 32

Classification Systems Alternatives ....................... _ . . . .. 33

Cost Comparisons . . . .. . _ . .... _ ............ ..... . .... _ ... . 33 Range Inventory: CLassifica tion and Mapping of Plant Communities ........ .. .... 3S

by Richard E. Francis and Richard S. Driscoll

Study Area ............ ............ . _ .......... . _ . . . . . . .. 35

Classification Syst em ....... ........... .. . . ...... ............ 36

Skylab and Support Data ..... . .• . _ . . • . . • . . . . . . . . . . . . . . . . . . . .. 37

Ground Truth . ....... .... . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . . .. 37

Procedures .............. ............. .........•.. ... •.. '. 38

Plant Conununity ClassiHcation by Photointerpretation . . • • . . . . . . . .. 38

Plant ConunWl ity Mapping . .. . _ ....... . • .•.•...... 39 CuhuraJ FeJtl,lre Mapping ... ........... . .......... . .... • ... _ 39

Foliar Cover Estimation ................. " ... .. .• •. . , ... ... 39

Pianl Comm unity Classification by Microdensit ometer .... , . ••..... , .. 39

Results and Discussion ......... ... ............ , .......... . 40 Plant Community Classification by Photoint erprelation ..... . ••.... 40

Plant Community Mapping ....... , ........... ,... ... ... . ... 44

Cul tural Feature Mapping ...................... . . . . • . . . . . . .. 47

Fo liaT Cove r Estimation .. ,................................ 47

Plant Community Classification by Microdensito me ter ....•. . •...•. ... 50

Applications ... . ' ..... . ..... ..... ....................... , .. 52

Forest Stress Detection : Ponderosa Pine Mortality frolll

Mo untain Pine Beetle

by Frederick P. Weber

Study Area . . . .................. H Classifica tion System ........• .. . ...•....... ,.,........... 57

~~~................ . ..... g

Ground Truth .... ......... . • .. •. . ...... .. •......•...... 58

Procedures ... . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . . . • . . . . . . . .. 59

Photointerpretation ............... • .....• • . . . . . . • . . • . . . . .. 59

MultispeClral Scanner Data Analysis ....•.•• . . . . •.. • . • . .. • ...... 59

Results and Discussio n ............... • ......• • .•. . . . ......... 60 Photoinlerpretalion .............. .. •.. ... •.......... ...... 60 MultispeCLraI Scanner Data Analysis .................. . ...• . .... 62

Applica tions .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . .. 62

Measurement of Forest Terrain Reflectance: Determination of

Solar and At mospheric EITec is o n Satellite imageI}' ...................... 64

Study Area

Instrumentation

by Robert W. 00/10

.... ..... ................ . M 65

Radiometer .. ..... . . . ... •• •• .. . . • .. ......... . ..... . ... . 66

(rradiance Meter ...... .... • .....•...... .••. ..... ........ . 66

Data Recorder . .......... •• ..... • .......•.......•.. ,., .. 67

Video Eq uipment ..... •.. .•.. .•. .• • .• ••. •••. .... ......... 67 Filter Sets ..... .. ...... . .• ......•......••.............. 67

Skylab Data .......................... .. .•• .... .. . .. . ..... 68

Procedures for Oata Analyses 68

rurcraft Radiance and lrradiance Data ...• ,.,.. ..... . .... .... .. 68

Skylab Photographic Data ............. • .... , , • , . . . . . . . .. 69

Results and Discussion ... ....... , ...... , ..•... ,.. . .......... 70

Applications

Utcrature Cited

......... . ..•• •• •••• • • • • • •••• •• ••••••• • ... , 72

. . . . . . . . . . . • . • . . . . . • . . • . . • • • • . . . . • . • • . . . . . .. 73

SUMMARY

Aldrich , Robert C., technical coordinator 1976. Evaluation of Skylab (BREP) data for forest and rangeland

surveys. USDA Forest Serv. Res. Paper PSW-II J , 74 p. Pacific Southwest ForeSland Range Exp. Stn., Be rkeley. CaJif.

Oxjo,d c U629.191+5S7.7t44) Relrievai Telms: Skylab; Earth Resources Experiment Packag4!; photo.. interpretation ; microdensilomeLric analysis; remote sensors; forest classification; range invento ry. plant communities; forest stress

Skylab pho tographic and Illultispectral scanner data products were sWdied during a period of 21,6 years 10 lest potential applica tions for fores t and range surveys. Fou r se parate studies tes te d classifi· cation of land use :md fo res t and range vegetation, detec tion of st ress on fores t vege tallon, and measure· ment of soiar and atmospheric effects in sa te llite imagery. Sites selected for these stlld ics were near Augusta , Georgia; Ma nitou, Colorado; the Black Hills of South Dakota; and Redding. Cali fo rnia. The Mani· tou and Black Hills sites were used in previous remote senSing studies of aircraft photography and LAND· SAT-I (ERTS-I) multispec tral scanner data.

Skylab data were supplied \,ly NASA as photo­graphic transparencies or computer·compatible tapes for analysis by photointerpretative and compu ter· assisted techniques. Photographic products included duplicates of SI90B terrain mapping camera ex po· sures on color or color infrared (C IR) films. Also in­cluded were color and CIR dupl!ca te transparencies and fOUT black·and -white du plicate transparencies from Ihe Sl90A multispec tral camera. The' fo ur black-and-white bands rep resen ti ng four spectral reo gions of the visible spectru m were combined into fal se-color composites. C'olllpu ler-compalible tapes included unstraightened conical scall data record ed from the 13-channel multispec tral scanner (S I92). Different instru ment s and methods we re used at each site to satisfy separate ex perimen tal requirements, and problems peculia r to each study were encoun· teredo The approaches to the experimental objectives also varied considerably 3t each site .

In general, the Skylab photographic data were found usefu l at two resource-oriented sites for broad cl3ssifica lion. Land use classes. such as forest and nonforest, and range vegetation classes at the Region level (DedduOIlS, Coniferous, 3nd Grassland) were dis tinguished with acceplable 3ccuracy when checked against ground Iru th. t-obps produced from digitized oplical 111m densities, me Asu red on colo l film , were acceptable for forest classes but unaccep table for

non forest. Regardless of in te rpretation technique, Level II nonforesl classes could not be accurately ide ntified on colOJ film, and range plant communi lies at the Se rie s !evel could not be classified with con­sist ency on any film/season combination.

Forest stress, in the form of mountain pine beetle­killed po nderosa pine , was detected only on Skylab's terrain mapping camera photography. No individual trees and only infestations over 26 meters (85 feet) in the longest dimension could be detected. Mountain pine beetle infestations could not be detected by analysis of SI92 multispectral scanner data. 80th photographic data and multispectral data were ac· quired in June-a period of low insect activity and a period when little discolored foliage remains on the trees killed by the previous year's bark beetle popu­lation. Had data been available for September, a period of high insect aclivlly and increaSing dead-tree discoloration, results might have been more encour­aging.

Additional conclusions drawn from the studies: 1. Systematlc or random-sampling designs can be

ove rlaid upon digitized photographic data by com­put er and classified into broad forest-non forest class­es for estimating area proportions. Acceptable results will depend on improving both the classification system and the dassincation procedure.

2. Enlargements of Sky lab te rrain camera photo. graphs (I : 125,0(0) ca n be used with. convenlional photointerpretalioll techniques to estimate propor­tions or broad land cover classes with in la rge political or administrative boundaries in tw().slage sampling designs.

3. Skylab and high-ahitude ae rial photographs can be used 10 Illap areal extent or Conirer and GI3ssiand classes with great er than 90 perce nt accuracy.

4. Paved and gravel roads. utililY corridors con· st ructed within the past 10 years, large mining exca· vations, and clusters of buildings can be mapped on enlarged Sky lab photoguphs.

S. Raruaoce from Skylab SJ90B and LANDSAT sensors was linearly correlated with wide· band terUln rellectance. Coefficients of the regression will be use· ful as linear conversion coefficients for extending spectral signat ures in computer-aided c1ass:i fi cation using satellite imagery.

Summaries of the individual studies composing this report follow. If Skylab·quality data should be· came available on a recurring basis, further investi· gation is needed to support, clarify , and extend these results.

Forest lnventory-Three independent studies were made of the potential application of Skylab photo· graphic dala in (1) forest resource evaluation , (2) sampling designs for computer application, and (3) automated land classification and mapping. Multi· spe<:traJ scanner data were not analyzed because a change in the test site location was required late in the lime period allowed .

Sixteen land use and forest classes, at three levels, were originally defined for both human and com· puler·assisted classification. Two types of photo. graph.ic data were used in the tests- muJtiband false· color composites of Sl90A data for September 12, 1973, and S I90B color for November 30, 1973. These photographs were enlarged to a scale of I: 125 ,000, J :250,000, and 1 :500,000 for interpreta­tion. Ground truth for forest resource evaluation was provided by a 1971 forest survey. High·altitude CIR photographs ( I : 120,000) and grou nd checks were used to map one county into two forest and two nonforest classes for a sampling design study. To eval· uate computer-assisted classification, two study blocks were mapped into 31 Levt'i III and Level IV land use classes on high·altitude C(R photographs, and corrected using ground observations.

In one forest resource evaluat ion, land use propor. tions were estimated for a four·county area. The esti· mate of fo rest area was within 2 percent of the 197 1 Forest Survey figure for the four counties. By indio vidual counties, the estimates were within 2 percent for three of the fOllr counties. With lhe exception of pasture and idle land , the four-county estimates we re aU close to ±1 percent of the 1971 Forest Survey estimates. A quasi..operalional application test using regression techniques in one county estimated fores t area 3 percent above the Forest Survey estimate. The sampling eHor was ±3.53 percen t.

Using computer·simulation techniques. sampling designs were tested with a digitil ed ground truth map and digitized Sky lab photographic data for one coun­ty. The variance in forest area resulting from sys· tematic sampling was always smaller than that from

2

simple random sampling. When a digitized type·map classified from Skylab S 1908 microdensitometer data was used in a postsampling stratification strategy, the variance of the forest area estimate was smaller than that from systematic sampling alone-but only when the sample intensity was low.

Computer-assisted analysis of microdensitometer scans made on Skylab S I90B color photographs separated forest land from nonfores! land with an accuracy of approximately 95 percent. Pine and hard­woods could Ix separated with an accuracy of ap­proximate ly 70 percent

Forest inventory studies in the Augusta, Georgia, sit e showed that Skylab S I908 photographs provide a good base for Level I forest classification. Although conventional pholointerpretatioll can provide accept­able area estimates for some nonforest classes, idle land, pasture, and water were difficult to separ:He on normal-color film.

Range Inventory- Classification of range plant communities was attempted at two levels of the ECO· CLASS system- three Region and eight Series classes. Skylab photographs from the S I90A muJtiband camera and the 5190B terrain mapping camera (June and August 1973), high·altitude aircraft photographs (June and August 1973), and Forest Service-acquired large·scale photography were used in the lests. Both visual and microdensitome ter techniques were tested.

Procedures were developed for sampling plant communities for use in photointerpretation training and tests. Procedures were also developed to map cultural features from 5kylab photographs. In a sepa· rat e study, foliar cover estimates made on Iarge·scale color photographs were compared with measurements made on ground transects.

Interpret ers classified Grassland and Conifer Re· gion classes with a mean accuracy of 98 percent or greater on both Skylab and support aircraft photo· graphy , regardless of dale or fUm type. Fo r the De· ciduous (Aspen) class, 3ccuracy was 80 percent or greater on the August CIR aircraft photogra phs, but was not acceptable on Skylab photographs. Coni· ferous Series class accuracies were dependent on dale and film type, blll inconsistently so. Accuracies were greater at small scales, probably because mix.ed tree species formed homogeneous units with a dominant species Signature and a lower resolution. Within the grassland Series, Shorlgrass was classified with an ac­curacy of 95 percent or greater on both Skyl ab and aircraft photographs, regardless of date or film type. For Wet Meadow, accuracy was greater than 90 per­cent on both June aod August aircraft photographs, regardless of fibn type or scale. and was also accept·

able on both color and CfR Skylab photographs laken in August. Mountain Bunchgl3ss was not ac­curately classified on Skylab photographs, but on the August aircraft photographs the classifJCation was acceptable regardless of film type or scale. Topo­graphic slope and aspect, mountain shadows, eco­tones, season, and class-mixing affected classification of plant communities.

In microdensitometer point-sampling, significant differences in mean optical densities at the 95 percent probability levd were a measure of separability. At the Region level, Conifer, Deciduous (Aspen), and Grassland classes were se parable. Howeve r. the Decid­uous class could be separated from the other classes with significant differences only on color film. Pon­derosa Pine was the only con iferous Series class that was separable from the o ther three conifers, rega rd· less of date or film type. Spruce/ Fit and Lodgepole Pine were not separable at any date or on any scale or fUm type. Douglas-fir was separable from the other duee conifers on both the June CJR and August color SI90A Skylab photographs. Grassland classifications at the Series level varied in acceptability. However, Shortgrass, Mountain Bunchgl1lss, and Wet Meadow were separable on August S1 90A color pho tographs. Optical demity was more depen dent on community mixing than all the growth stage of the plants at the lime (seaso n).

Both Skylab and aircraft photographs were useful to rn.:lP the areal extent of Conifer and Grassland, but the Deciduous (Aspen) class could no t be mapped with acceptable accu racy. Except for Wei Mead ow, Series level classes could be mapped with acceptable accuracy only if class·complexes were formed : Pan· derosa Pine/Douglas· lir, Lodgepole Pine/Spruce/Fir, and ShortgrassjMountain Bunchgrass.

Paved 3Jld gravel roads, ut ili ty corridors con· structed within the las t 10 yean, larger mining exca· valions, and clusters of buildings could be mapped on Sky lab photograph enlargements. On the other hand, I : 100,000 scale aircraft photographs were needed to map dirt roads, minor earth excavat ions, utility corri­dors older than to years, and individual buildings. Foliar cover and plant litter measured on large-scale CIR photographs of non-d iverse grasslands were re­lated to ground measurements with a correlation co­efficient of 0 .75. This is considered acceptable for range surveys. The relationship for foliar cover of shrubs was acceptable on ly on diverse grasslands.

Forest Stress Detection-An evaluation of Sky lab data in the Black Hills showed that mountain pine beetle infestations in ponderosa pine could not be identified on any S I90A multiband camen system

3

photographic product. AU positive identifications of bark beetle infestations were made on color photo­graphs taken by the SI90B terrain mapping camera. To be detected, infesta tions had to exceed 26 m (85 ft) in the 101lgest dimension. On one Site, only infesta­tions over 50 m ( 164 ft) in size could be detected. Infestations over 100 m (328 ft) in the longest dimen­sion were located with 100 percent accuracy. The optimum viewing scale with a stereomicroscope was about I :75,000. Best results were obtained with view· ing on a high-quality, variable high·intensity light table; however, stereoscopic viewing was preferred and usually resulted ill fewer commission errors than monocular viewing. lnterpreta tion on a rear·projec­tion viewer with high magnification was judged to be inferior to microscopic viewing on a light table.

Because late August through mid-September is the best period for detection of trees killed by mountain pine beetJe, the June Skylab image ry used in this analySiS was poorly timed. All dead ponderosa pine in the si te were killed during the previous year and had lost most of their discolored foliage before the Skylab pass. Si.nce the distinct red-orange color of dead tree foliage is used for recognition, many infestat ions were missed. Also, the early morning low angle of the sun at the time of the Skyiab missions made interpre­tation difficult. In the steep terrain of the Black Hills, west- and north· fac ing slopes were in shad ow.

Ponderosa pine trees killed by the mountain pine beetle were not detected by computer processing of 13-channel multispectral scanne r (MSS) data (5 192, June 9, 1973). Only five bands of the MSS were usa­ble, and misregistration of the data se riously de· tracted from the analysis results. Attempts to correct the registration improved ciassifiC3 tion somewhat in several instances, but beetle-killed trees were not identified.

The analysis for stress detection in this report was restric ted by circumstances beyond Ihe control of the invesl igalOfs. A Skylab ea rth resources pass was re­quested and scheduled during the desirable period (September 18, 1973) and correlative data needed in the analysiS were coUected by ground·based instru­ments on that date . However, the Skylab earth re­sources sensors were turned off unexpectedly during the pass. The results of the analysis in this report are therefore inconclusive. and further investiga tion is re­quired to determine if forest stress can be detected on Skylab-qualit y remOle senSing data.

Measurement of Forest Terrain Reflectance-Data on terrain radiance and solar irrudiance, gathered on o r near the Earth's surface, could be valuable for earth resource investigatioos. Calibrated satellite data

might be more easily inter pret ed and more fu lly used if investigators could estimate temporal and loea· lional variations in the effects of solar and atmos­pheric conditions on the sa tellite image ry.

An effort was made to use measurements of air· bOllle radiance and irradiance to comput e terrain reo Ilectance values. These values (when correla ted with satellite radiance of the same terrain element s) yield a first-order measure of solar and atmospheric propel'· ties at the time of a satellite overpass. Trea ting sa tel­lite radiance as the dependent variable, the correla­tion procedure produces an additive coefficient which is the path radiance, and a multiplicative coefficient representing the product of total irradiance and beam transmittance.

The airborne system for reflectance measurements consisted of an upward.pointing irradiance meter, a downward-pointing radiometer, a silicon vidicon camera (for support imagery) and associated data · recording instruments. Spectral matching to the band­widths of the Skylab S190A and LANDSAT· I MSS sensors was achieved for all three airborne sensors.

The SI 90A photographs were scanned by a digital

microdensitometer. Programs were written to convert microoensi ty values to diffuse density and subse­quently to effective 111m ex posure. Finally, Ihe satel· lite rad iance was computed from rum exposure values, using a camera radiance equation. Techniques, as outlined, we re develope d (or use o f the film sam­ples and the sensitometric package (provided by NASA) necessa ry (or these conve rsions.

The analysis of one se t of Sky lab photographs and one se t of LANDSAT· [ images resulted in a high linear corre lation between sa tellite ra diance and reo flectance. The derived palh rad iance values are in agreement with other published values.

The reflectance measurement technique presents tw o poss ib le a dval1tages over o ther empirical methods. One is that the results are derived in terms of the satellite scale of units, without concern for the calibration accuracy of ground·based or airborne radiometers. The other is that during stable periods in terrain reflectance propertieS", the reflectance mea­surement s need not be made on the same day as the satellite overpass.

GLOSSARY

Band: One of the wavelength bands of the electro­magnetic spectrum sensed by a mullispecual scan· ner (MSS) or passed by a bond-pass filter and recorded on photographic filOl.

Band Pair: Any two defined wavelength bands of Ihe electrOnlagnetic spectrum used in signalUre analysis.

Band-pass Filter: An optical mter that allows only defined portions of Ihe electromagnetic specuum to pass to the sensor surface.

Bias: The difference between the expected value of a statistic over aU possible samples and Ille true popul~tiOil value of th:lt sta tistic.

Color Composite: A fal se-coJor reronstruct ion of multiband photographs created from hVo or more filtered photographic bands. The four filt ered bands on Skylab (SI90A) wele O.Hl.6 f.lm (nation 1), and 0.8-0.9 f.lm (station 2).

Computer-Compltible Ta!)fl (CCT): A reeonmllction of data in magnetic tape form suilabl l/l for COin­piller analysis. In this sludy, CCT's o f Skylab M5S da ta and of digitized r hotogl3phic ot'tiCli t densi(ies were used.

Confusion Matrix: A tabular presentation of ctassifica· lion da ta showing Ihe propollion of actu:l.l vegeta· tion types Ihat were d anifie.d as each of the prN k le<! types.

Dilitat Element : A singte picture element of digital image density recorded on compUier-compatiblc lape by a micrOOensilomelet. The siu of the element V3ries with Ihe microoensilOlneter aperture.

4

itratiiance: The amount of light measured on a sur­face. In physics, the radiant nux density on a given swface. Usually expressed in watts per square meIC!,

MJcrodensitometer (MDT): An instrument used to measure the optical density of an image on a photographic t ransparency, using a calibrated light source,

Multispectral Scanner (MSS): For Skylab, an elec­tlonic optical li ne.scanning device (5192) that collects reflected and emitted radiation in 13 spec­tral inlervals (bands) of the visible, near-infrared, and Illt Unal·in frared regions of the elcctromagnetk speCtrum, The 5192 has a conical tin e ilean which meant line &e.a n data had to be straightened for computel analysis.

PostS3mpJing Stratification (PSS): SIr3tifJed sampling in which lhe 511&18 usignments are unknown Of are not used at the lime of sample stl~Cion .

Radi~flCe : The brightness of an object as Ken fTOm a 1emote observat ion point. In physkll, it j ~ a measUie of the power radiating flOm a unil area (If a source Ihrough a unit s<>li" angle. Typical units (I f radiance are wattsfmeter'·sleradian.

Sampling Fraction: Percenl of unitl $Impled t(l the t01a1 nlunber of IIn;1$ in Ihe populaliQn .

UTM: Universal Transverse MerCator map proj« tion. Zoom Tl'1Insfer Scope (21'S): An optkal instrument

ror transferring data fl om a small-scate pho tograph 10 a lalger $(..lIe pho lograph or map. The scate change tanse is flom I X to 13X. (Manufactured by Bausch and Lomb Oplical Company.)

INTRODUCTION

H igh-resolulioo sensors such as lhose on board Skylab may play an important p:ut in forest and

rangeland surveys in the future. Although conven­tional aerial photographs have been an aid in resource surveys for several decades, recent developments have moved toward more sophisticated photographic and nonphOiogJaphic remote sensorS and computer­assisted dat a analysis. The new teChnology is valuable for several reasons: ( 1) costs of acquiring resource data aTe rapidly increasing, (2) more reSOUfce data is required at shorter intervals to measure rapid changes in land use Ihat affect the environment , (3) urban and recreational uses o f land are encroaching upon avail­able resources and ahernate sources must be planned and provided for, and (4) there is a continuing need for up.T o·date resource information for day·to-day land management decisions and program planning.

There were Three separate Sky lab missions be· tween May 1973 and February 1974, These missio ns were deSignated SL·2 , SL-3 . and SL4. For each nlission a three·man crew was launched in a space vehicle to dock with the Sky iab Workshop (SWS) already in o rbit: SL·2, launched on May 15, 1973, continued for 28 days; SL.3, launched on July 13, 1973, continued 74 days; and the S lA miSSi on , launched on November 14 , 1973, continued 84 days. The Skylab experimen1 was completed on February S, 1974. During each mission, the Skylab Workshop was in o rbit approximately 235 nautical miles (435 km) above the earth . Earth resources co verage was confi ned by this o rbit to Ihe Earth's surface lying between the Equato r and 50 degrees north latitude and between t he Equator and SO degrees south latitude. Ou ring the missions, crew members were instructed to carry out certain experiments each day. Experiments included human physiology , astronomy, space technology , and earth resources. The Earth Resources Experimental Package (EREP) occupied only a small pOTtion of 1he astronauts' time and had a lowt'r priorit y than mos t of the studies.

Sky lab's EREP is o nly one o f several projects in NASA' s Earth Resources Progra m, which began in

5

1965 with NASA support of studies using photo· graphic and multispectral scanner (MSS) data ob· tained from aircraft flights . These eady studies we re made to define earth resource parameters and sensor requirements for projected space experiments, such as a se ries of LANDSAT (formerly ERTS) unmanned satellites, and Skylab and the Space Shuttle , which are manned satellites. The Remote Sensing Research Unil at Pacific Southwest Forest and Range Experi· ment Station has been involved in the Earth Re· sources Program since ' 965 , and the Rocky Mountain Forest and Range Experiment Station bec3me in­volved in 1969 . First, aircraft data were studied (Personnel of the Remote Sensing Research Work Unit 1972 ; Driscoll and Francis 1972), and then in 1973, LANDSAT· I (ERTS) MSS data were evaluated (Hell er 1975).

The conclusio n rrom the LANDSAT study was that with improved spectral and spatia1 resolution, satellite imagery could provide the first·level informa· tion required in ex tensive forest inventory sampling strategies. Skylab data provided an opportunity 10 investiga te and substantiate the conclusion, using machine'3ssisted classification procedures in addition to those depe ndent on human skill and judgmenl.

Objectives The original objectives o f this investigation as

outlined in the NASA contract proposal were these: I . 10 test the hypothesis that Skylab d31a will

pennit identification of fo rest, rangeland nonfo rest, water resources, and forest stress.

2. To determine if info rmation in satellite imagery, coupled with information in ailcrafl photog· raphy and ground examinations, will increase effi· ciency of surveys of forest·related resources.

3. To compare the accuracy and cost efrective· ness of vario us types o f data and compare direct visual classification techniques with computer·assisted classification to separate and identify forest and rangeland resources.

4. To develop and lest a praclicaJ method of correcting satellile radiance data for solar and atmos­pheric effe<:1 ill computer-assis1ed classification.

At each of Ihe four study sites, modificai.ions in specific obje<:1ives were necessary because operational and te<:hnical problems developed. For instaoce, when Skylab coverage for the Atlanta study site was found to be unavailable, the forest inventory study was move<! to an area north of Augusta, Georgia. This mOve was made in March 1974 after the Skylab experiment had been completed, and so no support­ing aircraft flighlS or ground truth for the actual times of the Skylab passes could be obtained. The multiseasonal data evaluatIons were therefore elimi­nated from the slUdy. Omission of proper mters for the multispectral camera during the November 30, 1973, Skylab pass over the Augusta site prevented our combining these data for analysis. In the Black Hills, a Skylab pass scheduled for September 18, 1973, was canceled at the last moment. Although biophysical data were collected on the ground during the period scheduled for Ihe pass, there were no Skylab dala in the proper time frame to analyze for stress detec tion.

Originally the analysiS at all sites was to include all data types and al l interpretalion techniques. Insuf· flcient data and time constraint s, however. required thai ( I) human pholointerpretation' techniques be used al all siles, (2) microdensitometric analysis of pholographic data be used on Ihe Augusta and Mani· tou siles only. and (3) an analysis of the multispect ral scanner data be performed by a suboontractor for the Black Hills site. These modifications made cost oom· parisons between aU methods impossible.

Study Areas

Experiments were conducted in four widely sepa­rated locations across the coun try (fig. 1). We had originally planned to use the same sites as those used in evaluations of aircraft imagery and LANDSAT-l (ERTS-I) data : AUanta. Georgia, for forest inven· tory ; the Black Hills in South Dakota for forest stress; and Manitou, Colorado, for range inventory. By retaining these Sites, we could use in the analysis firsthand knowledge of conditions accumulated over a period of years, and make comparisons between systems. However, problems of data acquisition for the Atlanta site made selecling of two new sites necessary to carry OUI the SlUdies originally proposed. The study 10 develop te<:hniques for measuring and correcting for a tmospheric interference, originally located at Atlanta, was moved to a new site nea r

6

Figure I - The stuliy lIeas indicated on this m.p were used for different phasts of the Skylab dua eVll.1ualion: Augusta. Georgia-foltst inventory; Manitou, Colorado- unjeland Inventory; Black Hills, SOllth Dakota-forest stress (rom mountain pille beetle; Redding, Califomia-sohu and atmos­pheric effects on dUa evaluation.

Redding, California. TIle sile change was made early enough so that both Sky lab and aircraft underflights could be scheduled on the same day. The forest inventory study was moved to an area near Augusta, Georgia, and the study was redeSigned . New ground tmth was coUected I yea r following the best SlA pass (November 30, 1973), and comparisons between Skylab, aircraft, and ground da ta were made from this new beginning.

The forest inventory site JUSt north of Augusta. Georgia, is representative of a large portion of the Southeastern United States where a high level of forest management is taking place and rapid changes are occu rring. Forests here occupy 75 pe rcent of the land area and are found in large contiguous bodies as well as in small woodlots intermingled with nonfores! land. Pulpwood, wildlife. and recrea tion are three major uses of the forest land in Ihis area. With a major lake , varied forest practices, and many forest and field borderlines, this area is a challenge to the photointerpreter.

Rangelands such as the Manitou, Colorado. site a.re important national resources and need to be inven· toried, protected. and managed. They are becoming more valuable as our food and fiber supplies become more critical. Classification of the predominant vege· tation according to its retation to other plants and animals and its potential for development is valuable. The Manitou site is appropriate for determination of the classification level at wh ich Sky lab data can be used to accurately assess range vegetation types.

In the Black Hills of South Dakota, the third site, a severe outbreak of mount ain pine beetle (Dendroc· tonus pollderosae Hopk.) has killed several hundred

thousand ponderosa pine Irees (P;IIUS ponderosa Laws.) over the past 10 years. Early detection of the dying pines. which discolor to a yellow aJld yellow­red hue, wouJd assis t forest managers in assessing the severity of the outbreak and in planning control and salvage operations- particularly if the discoloration could be detected accurately and quickly from either manned o r unmanned satellites.

The Whiskey town Reservoir-Redding, Califomia site was selected as an alternate site because it was on a Skylab ground track and convenient fOf low­allitude renectance measurements. This site consisted of a Slrip along one east -to-west flight lin e from the reservoir to Ihe Redding Airport. Measurements of reflectance by land use and vegetation classes were rather limited because of a partial cloud cover, Only mixed oak and digger pine (Pinus sabiniana Dougl.), ponderosa pine, brush species, and water were avail­able for reflectance measurements. The topograph.y ranged from rolling hilts 10 moderately steep foothill slopes and deep gullies. A light, variable cirrus over­cast was presem during both the aircraft and Skylab passes.

Dat a and Techniques Six EREP sensors were on board the Skylab

Workshop. These consisted of a multispectral camera, Si90A, with six high-precision matched lenses, an earth terrain camera, S190B, an infrared spect romete r (SI91), a H-channel multispectral scanne r (MSS), S192, a microwave radiomeler/scatterometer and altimeter, S193, and an L-Band radiometer, S194. Only SI90A, S 190B, and SI92 data were evaluated in the studies reported here. We o riginally intended to use SI91 infrared spectrometer data to evaJuat e the effects of the atmosphere; however, owing to a Isck of coverage that could be related to the ground truth measurements, Otis part of the experiment was elimi­nated.

The S I90A data were provided by NASA/JSC in the fonn of "2~ ·inch square film transpuencies, The original scale o f these dala was approximately I :2,SOO,IX>O, but they were enlarged to suit the needs of the individual studies. Special filters used on the camera lenses separated the visible and reileeted infrared spect rum into bands for muJtispectral anal · ysis. For each EREP pass over a test site, a set of six pictures was received. The bands, expressed as wave· length in micrometers, for each film type were these: color infrared (e IR): 0.5 to 0.88; high-resolution color: 0.4 to 0 .7; black·and-white panchromatic mm: 0.5 10 0 .6 and 0 .6 to 0 .7; and infrared sensilive mm: 0 .7 (00.8 and 0 .8 to 0.9 .

7

For each site covered by an EREP pass, a set of contact duplicate transpa rencies (5 by 5 inches) was received for all Sl9GB coverage. The original scale of these photographs was approximately 1 :900,000 but for special purposes they were photographically en­larged to 1:125,000. These photos were taken on either normal color or CIR nIm; the decision de· pended on the requirements o f the maj ority o f inves­tigators who had requested coverage on each pass, and in some instances we therefore received .)onnal colo r whe n CIR was pre ferred.

Computer-compatible tapes (CITs) with S I92 multispectral scanner data were received on1y for the Black Hills test site. These tapes were used by a subcontractor to claSSify stressed ponderosa pine under attack by the mountain pine beetle, The tapes, with 13 channels of MSS data, covered six discrete bands in the visible spectrum , six discrete bands in the reflected infrared, and one single thermal infrared band from 10.2 to 12.5 micrometers.

Techniques and instruments used by the invesli­gato rs to 3Jlalyze Skylab photographic data varied from one site to the nex!. In both the forest and range inventory investigations, a Bausch and Lomb Zoom Transfer Scope (ZTSl was used for mapping and dual image correlations. In the Black Hills investi· gation, a Bausch and Lomb 240 zoom stereo micro· scope was used to test a wide range of image mag· nifications for stress detection- both stereoscopically and monocularly. The investigator here also used a Variscan rear-plOjection viewer to interpret images at magnifacalions up to 29.5 times. On the forest lJlVen­tory Site, conventional photo interpretation was car­ried out with both an Old Delft scanning stereoscope and a lamp magnifier. A Photo Data Systems (PDS) automatic scanning microdensitometer and process computer were aJso used to scan and record optical density on one 519GB color photograph for com· puter-assisted classit1cation. At the range inventory site a General Anilh\e and Film Corp. (GAF) micro­densitometer was used to relate fllm density 10 plant communities. Similarly. a point-sampling technique was used al the Manitou site to classify plant com­munities by conventional interpretation. Interpreta­tions were verified on ground-truth maps prepared from high-altitude color infrared photography and ground checks. At the Augusta Site, existing forest inventory photo samples and ground subsamples pro­vided a basis for land use and forest -type evaluations.

, Ttade names and commeJciat enterprises (>1 pJOOUcU ale mentioned solei), rOI necusat)' infonnalion. No endolSement by the U.S. Depac tmell1 o f Agl iculture is implied.

The investigator at the Black Hills site systematically scanned each sub·block study site to detect bark­beetle-killed trees- the trees were counted, infested spots we re mapped, verified o n aerial observations.

and the Skylab interpretat ion pho tography or by ae rial

During the studies described here , large quan ti ties of data were gathered over a long period of time. Many of the techniques used were developed and modified in light of required changes in the analysis plan and from

8

experience. In this report , only enough detail has been included to help the reader understand what was done and evaluate the results. The report is intended pri· marily to aid those who may be using manned earth resources satellite data in forestry applicat ions in the future. Some additional de13iJed informatio n is avail · able o n request, as noted in later sections of this report. Requests should be ad dressed to Director, Pacific Southwest Forest and Range Experiment Station, P.O. Box 245 , Berkeley, California 94701.

FOREST INVENTORY

Forest Resource Evaluation, Samp ling Design, and Automated Lend Classification

Robert C. Aldri ch Edwin H. Roberts Wallace J . Greentree

Nancy X. Narick

The Skylab studies conducted near Augusta, Georgia, are part of 3 con tinuing research program to improve extensive forest inventory techniques. Inten­sive inventory techniques relate to individual forest stands as small as 2 hectares (5 acres), whereas the research reported here deals with the measurement of resource conditions within broad natural, administra­tive, and political boundaries using sampling proce­dures. Additional information about specific loca­Hons within the survey boundaries is useful bUI js not usually required.

Three independent studies were carried OUI by the investiga tors to evaluate Skylab photographic data for national forest resource sorveys:

I. An evaluat ion of Skylab S I90A (muJtispe<:traJ camera) and S I908 (earth terrain camera) photo­graphic data for classifying fo rest and related land 1I~. This study includes a quasi·operational one· county forest evaluation.

2. A study to det ermine an effective method for sampling digitized remote sensing data. This study includes a one-county evaluation of sampling designs and sampling intensity, using optical densities from an SI90B color ph otograph.

3. A study of microdensitometer techniques for classifying forest and related land use on Sky lab S190B photography. Two IO,OOO·me ter-square (6.22. mile-square) blocks are used as experimental areas to map Llvel II and Level It[ land use classes.

Synoptic vertical photographs taken from space platforms are of considerable in terest to forest inven· tory specialists. In March 1969, foresters found that color infrared photographs taken by AlXlllo 9 astro­nauts could be used to stratify fores t land into broad classes (Langley and o thers 1969; Ald rich 197 1) . In a mullistage forest inventory in the Mississippi Valley. this lirst· level stratification increased sampling effi· ciency by over 58 percent. Interest in muJtistage sampling strategies USing satellite imagery coupled with aircraft photography and ground observations has increased as a result (Draeger and o thers 197 1; Hildebrandt 1973; Kuusela and Poso 1972 ; Nichols and others 1974).

9

Thomas H. Waite

Since the early 1940's, land use classification has been the first stage in ex tensive nationwide forest resource inventories. This classi fication is usually done by photointerpreters on available medium-scale panchromatic aerial photographs. Many times these photographs are 5 to 10 years o ld at the time they are used. Since the primary purpose o f land use classifica­tion is to deteonine an accurate forest area base for expanding forest resou rce statistics, changes in land use since the photographs were taken can be a serious problem. If the foren uea base is not accurale, data from ground subsamples expanded by a forest area expansion fac tor can be inaccurate. Unless up-to-date photography or other remote sensing imagery is available on a wide area basis, adequate measurement of the changes in the fo rest area base is very diflicult .

In 1969, a research program was begun by the Pacific Southwest Station's Remote Sensi ng Work Unit to study high-altitude aircraft photography and satelli te imagery for land use and forest classification as a lirst level of information in resource surveys. These studies were conducted under the Earth Re· sources Survey Programs sponsored by NASA. A study near Atlanta. Georgia. using 1 :400,000 and 1:120,000 erR aerial photography, showed that for­est land could be identilied correctly over 96 percent of the time regardless of scale (Heller and o thers 1973). During this study, techniques for classi fying land use by optical mm density were investigated with only limited success. Multispectral scanner (MSS) data from airc raft flight s ove r lwo 4049·ltectare (lO,OO().acre) study blocks were also analyz.ed. Al­though land use classillcat ions were reasonably accu· rate, distortions in the processed data were a limiting factor (Weber and othe rs 1973). In 1973, LAND­SAT·I (ERTS- I) MSS data for the Atlanta test site were studied by bOlh conventional photointerpreta· tion techniques and by computer-assisted classilica­tion procedures (Aldrich and o thers 1975). Inter­preters could coerectly classify Level I information (forest , nonforest, water) over 96 percent of the time on false-color phot o composites. Comput er·assisted classilication using four bands of scanner data was 94

to 96 percent correct for the Level I classes_ Neither human nor machine classification couJd separate Level U infonnation with a high degree of accuracy.

Skylab (EREP) data with ground resolutions of from 10 to 30 meters offered an opportunity 10 frnd out whether higher resolution dala would improve the accuracy of Level I and Level IIlnformation for National Forest resource surveys. We also needed 10

kJlow if Levell information could be oblained from Skylab photographs by microdensitometry and com­puter·assisted classification and sampling procedures.

Study Area

The study site near Augusta, Georgia, (fig. 2) lies in the Piedmont physical division and is part of Ceorgia Forest Survey Unit 4.2 The area Is lypical of a large part of the Southern United States. with both broad contiguous bodies of forest land and m1a11 fann woodlots mingled with non forest land use. Foresl land is 75 percent of the total land area. Major forest types are loblolly pine (Pinus taeda L.), oak-pine, oak.htckory, and oak.gum cypress. Topography is gently rolling to hilly, with many narrow stream valleys. Principal land uses are forest. grasslaJtd (pas­ture), urban, and water. Agriculture is not a major land use , though scattered grain and row crops 8re found throughout the area. The Clark Hill Reservoir fonns the eastern bound3I)' and is the site of recrea· tion homes; it is a base for hunting, fishing, and other recreational uses. Major forest disturbances are caused by forest management practices and by development of recreation home sites.

Classification System

The diversity of techniques used in these studies called for a variety of classification systems. Although somewhat different in nomenclature, these systems have the same objective: to measure the forest area within an error of ±3 percent per million acres (404,858 hectares).

The classification system used in the forest re­source evaluation study included 10 individual land use classes (table 1). Not all of these classes. however,

I A sial e-wide rorest resource evalualion is made every 8 10 to years by Ihe Forest Resource Research Unit at Ihe Southeastern Fore$! txpeJiment Stali()n, Asheville, N.C. Section 9 or the McS~ney-McNary Forest Research Act or 1928, a$ IImended, and Ihe Forest and R.3ngelaod Renewable Resources Pl:lnning Act or 1974 authorize these forest resource evaluat ions.

10

were found within the four-county site. In some portions of the study the five agricuJtural classes were grouped as "other miscellaneous" to conform to the system used by the Forest Survey in their first·level stratification on aerial photographs.

The forest sampling design study in McDuffie County used a classification hierarchy of three Level J and two Level Jl classes. Forest land, nonforeslland , and water were mapped at Level Ito build a base fOl tesling computer sampling designs. Two Level II classes- pine and deciduous forest-were also de· lineated to see if they could be separated and stratified on Skylab data digitized by microdensi­tometer.

To test the value of computer classiHcation al· gorithms developed for LANDSAT·l (ERTS-I), a classification system of four levels was developed for Skylab (table 2). The land area in two 10,OOO-meter' square (6.22·mile-square) blocks was mapped to a O.4-hectare (i·acre) minimum using this system. On five randomly selected IOOO'meter (O.62·mile) sub· blocks within the test blocks, forest land was classi· fied by three stand-size classes to help explain varia­lions in the Skylab mm densities.

Skylab Data

Restrictions imposed by short missions, seasonal data requirements, and the scientists' lack of control over scheduling reduced the probability of obtaining a clear photographic day for Skylab data collection. For this reason, remotely sensed data were not ob­tained fOl the inventory test sHe near Atlanta . Georgia , where earlier studies had been made. To complete the proposed research using the mosl suit­able Skylab data products, an alternate si te was selected after the Sky lab experiment had been com· pleted and all data tabulated. By that time, SL-3 and SL-4 data for the new site were aheady 4 to 6 months old and it was no longer possible 10 gather any time­dependen t ground truth .

The Skylab data used in these studies include only SI90A and SI908 photography (fig. 2), taken on September 12 (SL-3 ; Pass 36 , Track 43) and November 30, 1973 (SlA; Pass 54. Track 19).

The panchromatic and infrared SL-3 SI90A multi· band photography was combined and enhanced on an 12 S (Stanford Technology Corp.) additive color viewer, and a color lnternegalive was. made of the combined image following the technique described by Myhre for combining LANDSAT-I (ERTS.I) fUm chips (Aldrich and others 197$). From the inter-

Sky lab 5190A multi band composite, Sept. 12, 1973

5kylab 51908 color photograph, No .... 30,1973

11

Augusta site map

FisuJe 1 - The Augu!ta !ite, wed for evaluatina Skylab d.ata In forest jpventory. indudes fout (:ounCie,_ Two inCentive study Ilte, l2 I.I'Id 4) wen ~ for compucer·.uisted mapping with diji!iud phocOi"phic (ilm densiti«. TIle Sl(ylab S190A mliitiband (ompo.!ite for September 12, 1973,and the Skylab SI908 color ph<JtOJrlPh for Novembt1: JO, 1973, wac llsed In the data a.ru.lysis. Photo­gnp hie sute is approximuely I :8S0,OOO.

+1

Forest L:lIld

Croplal\d

Jdle f3/m land

Improved pasture

Grassland

Other agriculture

Marsh ~nd 5warnpbnd

Urban and other areas

Cens.us waler

Noncensus w~ler

'u.S. Forest Service 1974.

Table I-A land·!;u Ilierarchy' for forest resource e~aluation

Dtrinilion

Areas 0.4 hectar\' ( I acre) or larger, capable of supponing more lhan 10 perctcnt cover by fortsl Irees and no t de~eloped for non forest use.

Land currently being utilized 10 produce agricultural crops that are harvested oJ irectly and not indirectly as pUlurc forage consumed by livestock.

Formcr cropland. orchards, implovll(( pasture, and farm sites not tended Wilhi" llle past 2 years and presently Jess than 10 percent stocked with Irees.

Land currently being improved for 8):u:ing by cultivation, seeding, iuigation, or clearing of brush and trees.

Land other than improved p~sture on which the primary natural cover is grass and forbs.

AI! other farm land no\ used for crON, idle, or pasture. Includes Cannstead, buildings, and service areas.

Land temporarily (){ partially CO'Ie red by Wafer, and poorly drained land capable o f supporting more than 10 percent cover or swamp vegetat ion (m:lJsh grasses, cal tails, etc. ). Dot's not include splU~e bogs, cypress laoo, or olher hyllr ic fO rtsl sites.

Areas within lepl boundaries of cit ies 3nd towns; subu,b~ n areas devtla..,ed for residential, industrial, or recr~tional purposes; schoolyartb ; cemete,ies; roads; railroads: airporU: beaches; pOlverllnes and other rish ts-of·way ; o r o ther nonforest land not included in any o ther specinc land class.

Streams, sloughs, esluaries, and canal. more than 0.2 kilometers (l/8 statute mile) in width: and lakes, res.crvoirs. and ponds more than 16.2 hectares (40 acres) in a!e~.

Streams, sloughs, estuaries, and canals less than 0.2 kilometer 0 /8 statute mile) in widlh; and lakes, reservoirs, and ponds less than 16.2 hec tares (40 acres) in area. Minimum width of streams, etc., and minimum diameter of lakes, etc., are 9.1 meters (30 fe et).

'U.S. Fortst Service 1968. Definition of Terms, p. 0-1 throusfl D·9.

nega tive, I :500,000 and 1:250,000 enl:ngements were made on color transparency film. These ftlms were used in combination wilh 5 1908 color trans· parencies fo r forest and land use classification; how· ever, the lower resolution and graininess of Ihe S I90A composite caused eye fatigue, and th e effect was to increase interpretation time. Although the ~ctral information was helpful in making land use and forest-type deciSions, we concluded thaI high· resolution CIR film from the S190B earth lerrain camera would have been much more useful. We did not use the S l90A OR or color films in our analysis because, after a careful review of lhe malerials, we feit that they contained only redundant information and at a lower resolution than the SI908 imagery.

The S I90A mulliband photography for SLA (November 30, 1973) could not be used in the anal·

12

ysis because Ihe flIms had been exposed without filters.)

Of Ihe two sets of color photographs produced by the $ 1908 terrain camera, only the SL4 (November 30, 1973) could be effectively use<! . The SL·3 photo· graphs (September 12, J973) were low in conlrast, showed a general haze condition, and did not include overlap for stereoscopic viewing. The SL4 photo­graphs, on the other hand, were taken on a clear day, contained reasonably good contrast between i3nd uses and forest conditions, and had 60 percent over· lap for stereoscopic viewing. We found this photog·

I National Aeronautics and Space Admini$lra tion. 1974. Sky/at! program, !il'1/SQI' perfo,mallCf! lepor', 1'01. I (S I90A), MSC·OSS28. NASA L. 8. Johnsoo Spacc Cenler, p. 3-6d.

Table 2-.4 flmd use cia~~ifirtllion hiQarchy for r{'mou sensing and ground informa/iOI/ sourfel compatible wilh furrenr n.o/ionwlde forNt rCKJurce el'tl/Utllion objectiv/!S. Color definitions are baM! on high-allitude cclor infrared photography (lIId simulated color infrared fompolite! of LANDSA T dala.

Oassirie:l1ion

FOREST LAND II Conifer

II Pine Pine-hardwood IV Seedling and sapling

Poles Sawtimber

II Deciduous hardwood 111 Upland hllldwood

Bottomland hardwood IV Seedling and sapling

Poles Sawtimbel

I NONFOREST l AND II Grassland

m Undisturbed grass Disturbed SJass Dead grass (annual) Ne .... implOved grass

II Cropland III Immature grain

Immature ClOp Mature crop Harvested crop Orchard Fanmteads

II Baie Soil 111 Plowed field s

Erosion Urban (site preparations) Rock ou tcrop

II Wild ~egetation 111 Idle land

Abandoned land Tlansilional Kud~u

Marshlal'ld Alder swamp

II Urban J1l Tran$pOrtation &

utilities Home developments Recrf'.1tion

WATER 11 Waler

1II Clear lakes & pond$ TUfbkllakes & ponds RiveB & $treams

Color definitions (based on Munsell 1920-60, NBS 1955)

Density of conifer stands, number of haldwoods mixed in stand , and stand size influen~ colo. ~a1ue and chloma. Dense stands are darker with less chrcm~. In the fall, befOle ad~;mced hardWOOd colo ra lion and leaf faU, conifer stands appeal dark purplish red. Separation between coni fer and hardwood classes is less. distUrocl in fall than in winter or early spring. Where hardwoods and coniferS lIle mixed ill stands, hardwood tolor predominates, and stand is usually class.ifieu as hardwood. In spring before haJdwDOds are foliated. conifefS appear moderate to dark pUIj>Jish ted. Seedlings and ~plings on prepared si tes appear lighter than poles and matu re sawtimber with closed canopies.

Stands appear moderate guyi$h purplish red in fall and pale pUlple to moderale purplish red in spring. In fall, upland hardwoods cannot be distinguished from b-otlomland hardwoods. In spring, bdore foli~ lion, upland hardwoods appear pale purple 10 light graybh purplish red . Bottomland haldwoods ate generally a moderate purplish ,ed . Stand me class (tex ture), density of crown canopy , and ground cover influenCt:t color value, density, and chroma bu t to a lesser eXltnl than in conifer stands.

Grassland appe3JS deep pink in both fal l and spring; sometimes mistakn for lmmatule cropland in spring.

Matule crops in fall appear bluish 8'3Y to grayish blue. In spring, immature CfOP S

appear deep pink and may be mistaken for 8ras:>Jand.

In fall and 1pring bare soil ap~T$ cream colored on LANDSAT imagery. There is no distinction between plowed agricultural fie lds and sites prepared for new com mercial developments. Generally in spring most areas of oare soil are newly plowed fields ei ther recently or soon-Io-be planted.

In fal l, artas unge fronl grayish purple o f idle land to glayish pUlpllsh red of abandoned land to deep pink of Kudzu vinc. Marsh and alder swamps a.le a moderate purple becaue of wet background. In spring, idle land becomes H~t grayish red to dark pink becau!.t of influx of ncw infJared-reflectant vegetation. Abandoned·transitional l~nd (Ieverling to fores]), o n the other hand . is 8rayish purplish red and malsh and alder swamps are grayish viole t. Deciduoos Kudtu vine. purplish gray in the spring, easily ~arates itself from all othe, vegetation when {aJi and sprin8 !mates are viewed together.

Areas ale light blue in the fall and very pale blue in the spring. UnfortunatelY, because of low resolutioo of LANDSAT data, secondalY roads, minor load •. and most utility lines are not lesolved.

Wate, is dark greenish blue ill fall and light greenish blue in spring. Farm ponds of le.u than .4 hectare (I acre. C1In be seen on LANDSAT Images if there is sufficient cont,nl with background .

13

raphy most useful for forest inventory purposes be· cause (I) a single photograph includes 8 to 10 coun· ties, (2) a photograph can be enlarged to 1: 125,000 without loss of information, and (3) a scaled map overlay of county boundaries can be used on the photograph without serious photographic distortion problems.

Cenerally speaking, forest land can be easily sepa· rated from nonforest land on SI90B color photo­graphs. There afe some confii.;ts, however, where shadows, small bodies of water, streams, and idle land appea r. Where fOlest land bOlders on "on rarest land, shadows cast by timber slands on west- and north­west-faci ng edges blend with the forest and can cause misclassil1cation bias in favOl of forest land . Small bodies of water within (orested areas are not easily separated from the surrounding (oresl - particularly pine foresl-because. to the eye, both features have the same density and hue. Although small bodies of water within hardwood stands show more cont rast with surrounding features, (hey too can be misinter­preted as pine . Idle land in many instances appears very similar to abandoned land stocked with hard· wood and pine saplings. The difference is very subtle and can cause a bias in favor of forest land.

Pine land is easily separated from the deciduous hardwoods in November. The separation is made by density (tone) and hue to a limited extent. Pine is much darker than the leafless deciduous forest , and the ground cover under the deciduous forest is a gray· green. Because of a limited range in color hues, pas­ture, cropland, idle land, and wild range are difficult 10 separate. Graininess of the mm gives the impres­sion of texture that interpreters look for in idle, aban­doned, or transitional land use types. Unfortunately, pasture land with this artificial texture appears very much like idle land. This confusion can be a problem in bleaking down land use beyond broad classes.

All major roads and secondary roads are clearly visible , as are utility corridors. However, woods roads are difficult to see and secondary roads are not re­solved where they pass through areas of little con­trast- agriculturalland , primarily.

Ground Truth

Ground truth for the forest inventory studies came flOm a number of sources. High-altitude aircraft (RB· 57) color infrared (eIR) photography, Forest Survey ground sample data, and Oeld checks were uStd in­dependently, Of together, as a basis for evaluating Skylab data int erpretaLion . Because we lacked ground observations and aerial photography for these studies taken at (he (ime of the Skylab pass, November 30,

14

1973, we had to rely very heavily on the RB-57 high· altitude CIR photography taken on April 25, 1974. The quality of r..hese photographs was excellent.

In the forest resource evaluation study we used 16·point photo cluster classifications made in 1971 on Agricultural Slabililation and Conservation Service (ASCS) panchromatic photographs, and a subsample of these clusters on the ground. These photographic and ground classifications included fi rst- and second-level informat ion for land use and forest stratification. Because these data were 2 years old at the time of this study, subsamples of locations were checked on the April 1974 CIR phOiographs, which were taken within 5 months 3f1er the Sky lab pass. With onJy minor differences in agricul tural use, the ASCS photographs renec ted land use conditiOlls at the time of the Skylab pass and showed where changes had occurred since the ground data we re cotle<: led.

The ground truth for the forest sampling design study was derived from the April 1974 CIR photog­raphy and a ground ch~k in December 1974 (hat included nearly 20 percent of the cou nty.

Initial ground truth for the automated land classi­fication study was ob tained by using the April 1974 CIR phot ography . Photographs were selected from the coverage ror two JO,OOO·meter-square (6.2-mile­square) sample blocks to be used in developing and testing digitized Sky lab film densities for land classi­fication . These two blocks were located by random selection fr om the tot al number of 1O,OOO.meter Uni­versal Transverse Mercator (UTM) intersections with· in the four·county test site.

Cround truth maps were made for the two sample blocks in the following way:

I. The photo for each sample block was mounted 011 a Zoom Transfer Scope and enlarged five limes to match a USGS I :24.000 quadrangle sheet used as control (jig. J). A lOoo·meter (O.62-mile) grid tem­plate (lO·by-tO grid) was placed over Ihe outlined block in the cont rol map 10 facilitate mapping.

2. Four forest and 27 nonforest classes (Level III) were delineated within each grid ceU.

3. In five randomly selected cells, forest land was further subdivided into three stand·sile classes: (I ) stedtings and saplings, (2) potes, and (3) sawtimber. Recognition of the classes was based on a combina­tion of crown closure, crown size, amount of bare soil , and arrangement of vegetation. These rermed delineations were needed 10 explain . discrepa ncies in automated classilication.

In early December 1974, we visited the test site to check map classifications and to observe conditions

-'

F \\ ~ c~:...

Figure 3-Zoorn Transfer Scope (/efl) was used to iruerprel and trlflSjX)se fore~t aoo {elated Ja"d use classes from 1: 120,000 CIR photograptls to II. map control hase. This instrument was also used to verify SkyJab interpretatiollS on I: 120,000 scale CIR photographs. An Old Delft scanning stereoscope (right) was used at 4X power to interpret overl3pping t;500,000 scale enlarsements of SI90B color photographs wilh stereoscopic effect.

that could affect land use interpretation. To aid in this effort, enlargements of the CIR aerial photo­graphs were made on which land use. forest condi­tions, and other observations could be retorded di­rectly in the field. Several photographic examples of local forest management practices that we re found to affect image inlerprel ation are shown in figllre 4. Ex­amples of land use on Skylab and high·altilllde erR were used in photointerpretation aids (jig. 5).

Finally. after the ground checks had been com· pleted, the gwund truth maps were adjusted and cor· rected where necessary. The maps were photographed and both prints and film overlays made to the scale of digital maps produced by the comput er-assisted classi­fication procedures. The map for block 4 was used 10 develop training sets for the computer classification algoritiun and the map for block 2 was used to test the system once it was developed.

Procedures Procedures used for the forest resource evaluation

study differed from those for the sampling deslgn study because the objectives we re different. One

IS

study was geared to the current photo procedures used by the Forest Survey in the Southeastern United States (U.S. Forest Serv_ 1968, pt. I, p. 1.6). The other was designed to develop the test computer tech­niques for sam pling digitized remote sensing data re­corded on computer-com patible tapeS.

Forest Resource Evaluation

To evaJuate Skylab photographic imagery as a source of firs t-level information. we adopted the Forest Survey procedure followed in the 1971 inven­tory of the State of Georgia. This procedure included computing the regression coefficients for th e relation­ship between photointerpretation results and ground truth for land use classes. First, an interpreter exam· ined points in a design printed on 1 :20,000 scale pan· chromatic aerial photographs purchased from the AgricultUral Stabilization and Conservation SelVice (ASeS). The design consisted of 25 clusters of 16 points (4 by 4) in a systematic pattern on each photo­graph. Each point in a cluster represented a circular O.4-hectare (I-acre) plot and was classified into one of five classes-(I) forest land , (2) urban and other.

Figure 4_ Poresl man.qement and other land use practices cau~ YlUiations in imllj;e pattenlS on Sltylab photographs. Examples are shown in the Skylab photo (left) and a correspoooing CJR aerial photo (right), at a scale of 1: 120,000. Intensive management of JobloUy pine involves pure, even-agoo sta nds. The sta nd at (0) lias an average heIght of J9.8 meterll (65 ft) and d.h.h. of 22.9 centimeters (9 in). These 8ttnds are the dUKest images in the photograph with the exception of water and shadOWs. The cleJ.1Cut are. at (b) $hOWl that debris hu 'been win<ifowed and burned, IlJ1d the soil ha:rrowed fOl machine planting. Pioneering welllU and har(lwood reproduction have not yet invaded the area8 between (Oil's. In (c) the pioneering apeeies have filled in spaces between pine seedlings. Sometimts logging debris is still yisible in wind.cOws Cd) because Slplin,:flto pin6 lit not large tnouJh to eomplddy cOYer the ground. Thinning operations bef~ stand mdurity remove NCry other: row and appel1 as in (e).

(3) census water, (4) oonccnsus water, and (5) other misceUaneous land uses (a grouping of five agricul· tural classes). lllis sample was designed to estimale forest area withifl a sampling eHor of 1:3 percent per million acres (404,858 hectares).

After interpretation was completed. a subsample of photo clusters was selected for ground examina­tion at the time of the inventory. Each point in the clusters was located on the ground with the aid of thc

16

photographs, and the individual land IJse class was detennined _according to the Forest Swvey hierarchi· cal classifIcaljon (table I). Because the photographs, laken in 1967, were 5 years old at the time of the inventory, the subsample provided in(onnation that reflected changes.

From the photograph and corresponding ground· cluster classifications, the regression coefficients were computed for each land use class in the five-class

Skylab S190A mulliband composite, Sept . 12. 1973

Skylab S1908 color photograph, Nov. 30, 1973

II

+ ,. .....

+ •.

.± jo. + J;.:f • . -Augusta site map

Fiaure 2-The Augusta site, uaed for evlluttilll Skylab dati in forest inve-ntory, iuctudes fOUf counties. Two intensive study sites (2 and 4) Wele uS«! for computer-assisted mapping with diaitlud photographic film densities. The SkyLab S I90A multiband composite for September 12, 1973,'1'd the SkyLab S190S color photograph fOr Novcmbet 30, 1973 , wen used in the data an. lyti$. Ph<lto­Jrlphic reale is approximately 1:850,000.

Classification

Forest Lmd

Cropland

Idle farmland

Im proved pas ture

Gr~ssland

Olher agrkullure

Maish and swamp\ond

Urbon and olher afe~s

Census waler

NOI'ICeRSUS waler

iU.S. Forest Setvil:e 1974.

T!ll;>le I _A faud'lIse /tierardIY ' [ or {orf!$1 r esOllrce eV(1fUO(Km

Definit ion

Arus 0.4 hec tare (I ac re) or larger, capable of supporting more Ihan 10 peH;enl cove r by foresl h ees and no t deve loped for nOnfOreli l use.

Land cu rrently heiot utiliud to produce agricultural crops that are harve:oted dir t'(; tly and not indheC"t ly as pasture forage consumed I;>y lives tock.

FOlmer cropland, o rchards, improved pasture, and farm sites not tended within the past 2 years and presently less thall 10 percent slod.ed with IIees.

und cunem ly being imploved fQ r glazin, by cultivat ion. seeding, i rrigation, QI ch~aring Qf brush and trees.

Land other than improved pasture on which the primary na tu ral cover is grass ~nd forbs.

All Olher farmland no t u$Cd fQI crops. idle, QI paslure. Inc ludes farmslead, bui ldings. and service ~ reas..

Land temporarily or partially cOl'ered by water, and poorly drained land capable of supporting m Ot e Ihon 10 percent cov~r of swamp vegetat ion (marsh gusses. cat tails, e lc.). Does not include spruCe IJ O(l!S , cypress land, or other hydric fOTl'st sites.

Areas wilhin legal boundar ies Q[ ci lies and towns; suburban areas deve loped (Qt res.idenliaJ. indllsuial , or tec~tion aJ purp-oses; schoolyards; cemete ries; roods; raihoods: ai rpo tl s; beaches: p owert ines and other righ ts-of·way; or Qlher nonforest land 001 included in any Qthet specifIC la nd class.

Stre~mi , sloughs., estuaries, and canals m ore tttan 0.2 kilometers (l/8 statute mile) in width; and laKes. reservoirs . and ponds mOTe than 16 . 2 hec laTes (40 ac res) in area.

Stlt amS, sloughs, estuaries, Jnd canals less than 0.2 kilQmttel ( 1/ 8 statu te mile) in Wid th ; aDd lakes. reservoi rs , aDd ponds less than \ 6.2 lIeclares (40 oteres) in area. Minimum widlh of streams, e tc., and minimum diameter of Jakes, etc., are 9.1 meters (30 feet).

'U.S. FOTesl Service 1968, Defini tion of Terms, p. D·I IlI rou&-h J)..9.

negative, I :500.000 and I :250,000 enlargements were made ori color transparency film. These mms we re used ill combination with 5 190B color trans­parencies for fo resl and land use claSSification ; how· ever, the lower resolution and graininess of the SI90A composite caused eye fati gue , and the effect was to increase interpretation time. Although the spect ral information was helpful in making land use and foresH ype decisio ns, we concluded Ihat high. resolutio n CIR film from the S I90B earth terrain C3J1lera would have been much more useful. We diu not use the SI90A CIR or col or fi lms in our analysiS because, after a ca reful review of the materials, we felt that they conta ined only redunda nt infonnatio n and al a lower resolutio n than the S I90B imagery.

The S I90A rnultiband photography for SL-4 (November 3D, 1973) could not be used in the anal·

12

ysis because the films had been exposed without filters.)

Of the two sets o f color photographs prod uced by the S I908 tercain came ra , onJy the SL-4 (November 30, 1973) could be effectively used. The SL-3 photo· graphs (September 12, 1973) were low ;n contrast, showed a general haze condition, and did not include overlap for stereoscopic view ing. The SL4 photo­graphs, on the o ther hand , were taken o n l clear day. contained reasonably good contrast be tween land uses and forest conditions, an d had 60 percent over· lap for stereoscopic viewing. We found this photog·

l Nation:l.I Arlonau!ics and Space AdmininlHion. 1974. Skylab prorram, tensor performance rtpofl. 110/. J (S I90A), MSC-05528. NASA L B. Johnson SpJce Center, p. 3-6d.

Table 2-A lund USt dttui{!aJlion hierarchy fm remote sensing and ground in{ormation U)Urce$ compatible with cwrellt nationwide foral refoW'ce el'tlluiltion objecti~es. CoIm definitions are IHlsed 011 high·altitude co/or infrared photography and simulated cofOl" '''/rared comporius o[ LANDSA T dora.

FOREST LAND II Conifu

II Pine Pine-haruwood 1'1 Seedling and saplinS

Poles Sawtimber

II Deciduou$ hardwood III Upland hardwood

Bottomland hardwood IV Seedling and sapling

Poles Sawtimber

[ NONFOREST LAND II Grassland

lIJ Undisturbed grass Disturbed "ass Dead gra~ (annual) New improved IraM

II Cropland HI Immature &rain

ImmatufC ClOP Mature crop Harvested crop Orchard Farmsteads

II Bare Soli III Plowed r.elds

Erosion Urban (site preparatIons) Rock outcrop

II Wild vcgetatlon III IdJf land

Abando~ land Transitional Kudtu Marshland Aider swamp

II Urban III Transporiation &

utilities Home developments Recreation

WATER II Water

III aear lakes & ponds Turbid lakes & pond s RiveH &: strcams

Colo r definitions (based on Munsell 1920-60. NBS 1955)

Densily of conifer stands, number of h3Jdwoods mixed in stand , and stand size Influence colOI value and chroma. Dense stands are darker with less chroma. In the fall, before advanced hardwood coloration and kaf [aU, conifer stands appear dark purplish led. Separation between conifer and hardwood classes is less dist iincr in fall than in winte r or early spring. Where har(\W()o.:1$ and conifers art mixed in nandi. hardwood color predominates, and stand is usually classified as hardwood . In Spring before hardwoods are foliated , conifers appear moderl te to dark purplish red. Seedlings anu saplings on prepared si les appear lighter than poles and mature sawtimbc.r with closed canopies.

Stands appear moderate grayish purplish red in fall and pale purple to moderate purplish red in spring. In fail, upland hardwoo(]s CMnOI be distinguished from bottomland hardwoods. In spring, before foliation. upbnd hardwoods appear pale purple to light grayish purplish red. Bottomland hardwoods aTe generally a modeute purplish red. Stand size class (ruture), density of crown canopy, and ground cover influencee color value, density, and chroma bUI 10 a lesser extent than in conifer stands.

Grassland appears deep pink in both fall and $pIing; sometimes mistaken for immature cropland in spring.

Mature crops in fall appear bluish gray to grayi$h blue. In spring, immaUlfe crops appear deep pink and may be mist:J..icen for grassland.

In fall and spring bare soil appears cream colored on LANDSAT imagery. There is no distinCt ion between plowed agricultural field s and sites prepa.red for new commerciDI developments. General ly in spring mOSI areas of bare soil are newly plowed fields either cecently or soon- to-bc planted.

In fall , areas range from grayish purple of idle land to /illayish pll rpli~h red of abandoned bnd to deep pint of Kudtu vine. MaISh and !Idel swamps aTe a modera t ~ purple because of weI background. In spling, idle land becomes Ught grayish lcod 10 dark pink be<:au$e of influx of new inrra:red-refle~lant vegetation. Abandoned-llansidonal land (reverting 10 fOTest) , o n the olher hand, is grayish purplish red and marsh and alder ~wamps arc gray ish viole t. Deciduous Kudzu vine: , purplish gray in the spring. easily separ8lU itself from all olhel vege tation when fail and spring images are vicwed toge(htJ.

Areas lUe tight blue in the faU and very paie' blue in the spring. Unfortunately, because of low resolution of LANDSAT data, secondary roads, minor roads, and most utility lines a.e not resolved.

Water is dirk greenish blue in fall and light greenish blue in Wring. Farm ]JOnds of less than .4 he("tare (I acre) can be seen on LANDSAT images if there is surnclent contrast with background.

J3

raphy most useful for forest inventory purposes be­cause (I) a single photograph includes 8 to 10 coun­ties, (2) a photograph can be enlarged to I: 125,000 without loss of information, and (3) a scaled map ove rlay of county boundaries can be used all the phologl'aph without serious photographic distortion problems.

Generally speaking. forest land can be easily sepa­rated from nOllforest land on SI90B color photo­graphs. There are some coomcts, however, where shadows, small bodies of wa ter , streams. and idle land appear. Where forest land borders on non forest land , shadows cast by limber stands on west- and north­west-facing edges blend with the forest and can cause mi sclassification bias in favor of forest land. Small bodies of water within forested areas are not easily separated from the surrounding forest-particularly pine forest - because. to the eye, both features have the same density and hue. Although small bodies of water within hardwood stands show more contrast with surrounding fealures, they too can be misinter­preted as pine. Idle land in many instances appears very similar to abandoned land stocked with hard­wood and pine saplings. The difference is very subtle and can cause a bias in favor of forest land.

Pine land is easily separated from the deciduous hardwoods in November. The separation is made by density (tone) and hue to a limited extent. Pine is much darker than the leafless deciduous forest, and the ground cover under the deciduous forest is a gray­green. Because of a limited range in color hues, pas­ture. cropland. idle land, and wild range are difficult to separate. Graininess of the mm gives the impres­sion of tex ture that interpreters look f01 in idle, aban­doned. or transitionilland use types. Unfortunately. pasture land with this artificial texture appears very nluch like idJe land . This confusion can be a problem in breaking down land use beyond broad classes.

AJI major roads and secondary roads are clearly visible, as are utility corridors. However, woods roads are dimcuh to see and secondary roads are not re­solved where they pass through areas of little con· trast - agriculturalland , primarily.

Ground Truth Ground truth for the forest inventory studies came

from a number of sources. High.altitude aircraft (RB-57) color infrared (C1R) photography, Forest Survey ground sample data, and field checks were used in­dependently, or together, as a basis for evaluating Skylab data interpretation. Because we lacked ground observations and aerial photography for these studies taken at the time of the Sky lab pass, November 30,

14

1973, we nad to rely very heavily on the RB·S7 high­altitude erR photography taken on April 25, 1974. The quality of these photographs was ex.cellent.

In the forest resource evaluation study we used 16-point photo cluster classifications made in 197 1 on Agricultural Slllbiliz.ation and Conservation Service (ASeS) panchromatic photographs, and a subsarnple of these clusters on the ground. These photographic and ground classifications included first - arid second-level information for land use and forest slrat il1cation. Because these data we re 2 years old al the Ilme of this study, subsamples of locations were checked Oil the April 1974 CIR photographs, wJtich were taken within 5 month s after the Skylab pass. Wilh only minor differences in agricultural use, the ASeS photographs renee ted land use conditions at the time of the Skylab pass and showed where changes had occurred since the ground data were collected.

The ground truth for lhe forest sampling design study was derived from the April 1974 eIR photog­raphy and a ground check in December 1974 that included nearly 20 percent of the county.

Initial ground truth for the automated land classi­fication study was obtained by using the April 1974 CIR photography. Photographs were selected from the coverage for two 10,OOO.meter.square (6.2-mile­square) sample blocks to be used in developing and testing digitized Skyiab film densities for land classi­fication. These two blocks were located by random selection from the total nwnber of 10,OOO-rneter Uni­versal Transverse Mercator (UTM) intersections with­in the four·county test site.

Ground truth maps were made for the two sample blocks in the following way:

I . The photo for each sample block was mounted on a Zoom Transfer Scope and enlarged fi ve times to match a USGS I :24 ,000 quadrangle sheel used as conlrol (frg. 3). A 10000meter (O.62·mile) grid tern· plate (IO-by-IO grid) was placed over the outlined block in !.he control map to facilitate mapping.

2. Four forest and 27 non forest classes (Level Ill) were delineated wllhin each grid cell.

3. In five randomly seletted cells, forest land was further subdivided into three stand-size classes: (1) seedlings and saplings, (2) poles, and (3) sawtimber. Recognition of the classes was b3sed on a combina­tion of crown closure , crown size, amount of bare soil, and arrangement of vegetation. These refmed delineations were needed to explain . discrepancies in automated classification.

In early December 1974, we visited the test site to check map classifications and to observe conditions

· ........ l •

.. .... . A . \\

J

Figw:e 3-Zoom Transfer S~ (It/I) was used to interpret alld transpose forest and ,elatoo laRd use cluses from 1: 120,000 CIR photo81aptu to a map control base. This instrument WJ$ lIso used 10 verify Skylab interpretations 011 1:120,000 scale CIR photographs. An Old Delft seaMing stereoscope (rig/II) was u~ed at 4X powe, to interpret overlaPl'ill8 1:500,000 scale enlargemenU of 51908 color phologmphs with stereoscopic effect.

that could affect land use interpre tation. To aid in this effort , enlargements of the CIR aerial photo­graphs were made on which land use, forest condi· IiOIlS, and Olher observations could be recorded dj· rectly in Lhe fi eld . Several pho tographic example5 of local forest management practices that were found to affect image interpretation are shown in figllre 4. Ex­amples of land use on Skylab and high-altitude CIR were used in photointerpretation aids (Jig. 5).

Finally . after the ground checks had been com­pleted, the ground truth maps were adjusted and cor­rected where necessary. The maps were photographed 3Jld both prints and film overlays made to the scale of digital maps produced by the compute r-assisted classt­ficalion procedures. The map for block 4 was used to

develop training sets for the computer classification :llgori thm and (he map for block '2 was used to test the system once it was developed.

Procedures Proced ures used for the forest resource evaluation

study differed from those for the sampling design study because the objectives were different. One

15

study was gea red to the current photo procedures used by the Forest Su rvey in the Southeastern United States (U .S. Forest Serv. 1968, pI. I , p. 1-6). TIle other was designed to develop the test computer tech­niques for sampling digitized remote sensing data reo corded on computer-compatible tapes.

Forest Resource Evaluation

To evaluate Sky lab photographic imagery as a source of first·level information. we adopt ed the Forest Survey procedu re followed in Ihe 1971 inven· tory of the Sute of Georgia. This procedure included computing the regression coefncie nts for the relation­ship between photoinlerprel3tion results and ground truth for land use classes. First, an inte rpreter exam· ined points in a design primed on I :20 ,000 scale pan· chromatic aerial photographs purchased from the Agricultural Stabilization and Conservation Service (ASCS). Th e design consisted of 25 clusters of J 6 points (4 by 4) in a systematic pattern on each photo­graph. Each point in a cluster represented a circular O.4-hectare (I·acre) plol and was classified into one of five c1asses- (J) forest land, (2) urban and other,

FiJure 4-Forest mamgement and other land I/se practices cal/sed. variations in una&e patte{ns on Skylab photographs. ExampJes are shown in the SkyJab photo (ft fl) and a couespondJna CIK aerial photo (righl), at a scale of 1:120,000. Intenmve mlnagement of loblolly pine involVe! pure, oven·aged stands.. The stand at (a) hu an average height of 19.8 meters (65 ft) and d.b.h. of 22.9 centimeters (9 in). These stands tee the darkest images i.n the photograph with the ex~pdon of water and slladowi. The cleUCut area at (b) shows that debris has-been windrowed and burned, and the $On }Jarrowed fOT mlclliM planting. Pioneering weeds and hardwood reproduction have not yet invaded the areas between rowa. [n (c) the pioneering species have filled in spaces between pine seedlings. Sometimes logging debris is "ill visible in wjnd!ows (d) becawre sapl~·size pinJs are not large enough to completely cover the JrOUnd. TlIlnnif1g operations before stand maturity lCffiOve every other row and appear all in (1").

(3) census water, (4) noncensus water, and (5) other miscellaneous land uses (a grouping of five agricul­IUral classes). This sample was designed to estimate forest area within a sampling error of ±3 percent per million acres (404.858 hectares).

After interpretation was completed. a subs.ample of photo clusters was selected for ground examina­tion at the time of the inventory. Each point in the clusters. was located on the ground wiih the aid of the

16

photographs, and the individual land use class was determined .according to the Forest Survey hierarchi­cal classiftCation (table /). Because the photographs> taken in 1967. were 5 years old at the time of the inventory . the subsample provided information that reflected changes.

From the photograph and correspOnding ground· cluster classifications, the regression coefficients were computed for each land lise class in the five-class

FJiure $-A Skylsb .photoplph (!tll ) and I 1:120,000 CIR photOifllph (righ l ) wen used jJl l llte~pr t t.ltion

aid. to illustnlt the I«est and rtlatoe! land use classe;;: (a) plJ\C fOf~t, (b) hard .... ood fo",st, (e) pssland, (d)

cropland, (~) bile soil. if) wild 'fegetation, (g) urban, and (h) ""l ie ••

grouping. The proportions of points in the clusters were used as continuous variables in the computation.

Our evaluation of Skylab photographic imagery followed the Porest Survey pr~dure clo~ly. All photo subsample clusters located on 1 :20,000 ASCS pbotograplu hAd to be located on Skylab phot~ graphs and tile points classified by mdividual laoo lISe. This was accomplished using an Old Delft stere~ scope (fig J). a scaled cluster template, and a photo illuminator.

Three separate evaluations were made. In the first , Uncoln County was u.sed to test three combinations of Skylab photographic dau to find the one most suitable for the remairung evalu.ations :

Combinorion 1: 1:2S0,000 enlargement of an

17

SI90B color photograph (November 30, 1973) made at PSW, and a I :2SO,OOO enlargement of an Sl90A /alse-color combination (September 12, 1973) made at PSW. In terpretation was without stereoscopic effect.

Combination 2: 1:500,000 enlargement of an SI908 color photograpn (November 30 . 1973) made by NASA/JSC, and a 1:500,000 enJargement o f an SJ90A faJ~<olor combination (September L2, L973) made at PSW. Interpretation was without stereo­scopic effect.

Combillal;orJ 3: I :SOO,QCK) tnlargement of an SI908 colo r photograph (November 30, 1973) made by NASAIISC, and an overlappillg enJargement of an SI90B color photograph (November 30, 1973) made

by NASApSC. Interpretat ion was with stereoscopic effect.

Forty 16-point photo clusters Ihal had been sub­sampled on the ground in the 1971 forest inventory were located and interpret ed, and the individual land use at each point was recorded_ For the center point only, in each cluster, an additional record was made of stand origin, physiographic si te, fores t type, and disturbance class. This center point corresponds to the plol center used on the ground to record tree and stand conditions. A record of interpretation time was kept for each set of data. The data were then tabu­lated and processed by compute r 10 de termine differ· ences between the three se ts and to learn which com­pared most favorably witll the ground truth.

A second evaluation of Skylab photographic imagery was made fo r the four·county area , using only combination 3, the NASNJSC 1:500,000 scale colo r enl3Igements. We chose tltis only partly because it gave acceptable results on the I1rsl evalu:uion. In addition , we found that with this combina tion, the interpretation was faster, and the interpreter felt more confident of his cla ssil1cations because the stereoscopic effect was helpful in making borderline decisions and the images were sharper.

In this second evaluation we located and inter­preted all 210 16-point clusters in the four-county area by procedures identical to those used for the first evaluation. Classil1cations made on ASCS photography, on Skylab imagery, and on the ground were analyzed by both grouped and individual land use classes.

In the third and final evaluation, Uncoln County was used as a case study to tesl the hypothesis that Skylab photographic data could be used in the Forest Survey procedure to me3sure forest area within sped­fied accuracy limits. f o r this evaluation, a I : 125,000 en1argement was made of one Skylab 5 190B photo­graph tha i covered the enlire lest site. A grid overlay of 16-point d usters was conslructed to sample the photograph af an intensity very simil ar to Ihal used by the Forest Survey Unit. The boundary of Uncoln COlmty was drawn from a I :250,000 USGS map sheet and an overlay was made to the scale of the photo enlargement. Interpretation was completed us­ing a lamp magnifier. The grouped land use class was recorded for each cluster point. Forest type and dis­turbances were recorded for the cluster center. Time to complete interpretation was recorded for cost ef­fectiveness analysis.

Using the proportions in each cluster as a con­tinuous variable, computations we re made for each grouped land use class in the photo sample according

18

to procedures described in the Forest Survey Manual for the Southeast. Tenns used in the computations were

x = proportion of land use in a photo (ASeS) subsample cluster

y =: proportion of land use in a ground or Skylab subsampJe cluster

P .. proportion of land use in the photo sample cluster

P = adjusted land use proportion in county a = regreSSion constant b = regreSSion slope coefficient n = number of clusters in subsample m = total number of pho to clusters N = total number of sampling units in the popula­

tion L = adjusted area in land use class A '" total area being sampled

, nd SSy = corrected sums of squares of y

= Ey2 _ (Eyf n

SSx = corrected sums of Squares of x = Ex2 _ (:Bx/

n

SP xy = corrected sums of squares of cross products

= Exy _ ("x)("y) n

Sy 2 = variance of y

SS, --

n-l

Computations were made for I. The mean proportions ror individual land use

classes using ground or Skyiab data

Y; = _""r::-,,' "(Y,,;,,,) n

in which Yij == proportion o r the jlh cluster in land use

classj Yj '" mean proportion in land use class j

2. The mean proportions for grouped land use classes using ASeS, Skylab, and ground da ta

in which:

:B~,q (Yjk) < -'-'="""""­

n

Yik '" proportion of the jth clustedn youped land use class k

~h = mean proportion in grouped land use k

3. Regression constant (a) and slope coefficient (b) as follows: for (1) the relationship between land use proportions on ASCS photographs and the corresponding proportions on the ground, (2) the relationship between land use proportions on the ground and the corresponding proportions on Skyiab photography, and (3) the relationship between land use proportions on ASCS photographs and the corre­sponding proportions on Skylab photographs. The equation for adjusting land use proportions tOok the form

P=a+b(p)

4. The squared standard deviation from the regres­sion for each method

(SPXy)l SS, - '--'''"­

SS, Sy.x' = ___ ,,-"_

"-2

5. The adjusted area class for each method

for each combined land use

L = (PXA)

6. The standard error of the- adjusted forest area for each method expressed as a percent

"=j , [~ + (j>-X)'J~I- ~J +l' [1-mN-~ P y.x n SS m m , (" ) SE = fff (100) at the 67-percent level of confidence

Forest Sampling Designs

McDuffie County, in the Augusta site, was used to test the value of using Skylab photographic products in sampling strategies for obtaining Level I and some Level II land use statistics. These statistics are pres­ently obtained for extensive forest inventories from aircraft photography and ground observations.

To provide a data base for evaluating several sampling strategies, a land use map (fig. 6) of the county was made by interpretation of 1:120,000 scale CIR photography from RB-57 Mission 274. All Level I land use classes (forest, nonforest, and water) and two Level II classes (pine and hardwood) were mapped at a scale of 1 :50,000 from ZTS interpreta­tion. A control base, enJarged from the 1 :250,000 scale USGS Athens, Georgia, quadrangle, was placed on the ZTS mapping table (fig. 3), and class bounda·

19

fies interpreted from the ClR photo on the easel were drawn on clear acetate overlaid on the map base. This method eliminated cumulative pOSitional errors.

The photography used allowed rather easy, though time-consuming, interpretation into the four land use classes (pine, hardwood, nonforest, water). A ground check which covered nearly 20 percent of the county showed that the only potential classification problem was the distinction between idle farmland and transi­tional land reverting to forest. The decision on the ground is made on the basis of the density of natural stocking with commercial seedling species. Lands thus classified as forest are often plowed the following sea· son and ret\lrned to farmland. Changes back and forth between forest and farmland can be quite rapid in this part of the country_

Besides the 20 percent sampling checked on the ground, 6 percent of the county was intensively mapped on the ground directly onto I: 10,000 scale photo prints, which were used to check the accuracy of the land use classification and correct the map produced from interpretation of high-altitude CIR photography _ The resulting land use map is thus an accurate representation of the actual land use on the ground within the limits of the classification system used_

The land use map was used as a data base to try out different sampling strategies and to compare the results of computer classification from microdensi­tometer scans of a Skylab color transparency. To do this, however, we first had to digitize the map with a Bendix Data Grid. This was done in three steps: (1) a dot grid of 16 dots per square inch, representing ap­proximately 50.3-meter (l65-foot) spacing on the ground, was laid over the completed base map, (2) the map and grid overlay were oriented with the digitizer table, and (3) the digitizer Cursor was placed over each grid point and its mapped land use type was keypunched onto magnetic tape. Once the ground truth map was completely digitized and on magnetic tape, each grid point could be relocated on the tape by its ]jne number and position in the line, using a tape-controlled plotting device (jig. 7) and a facsimile map plotted in color. This map could then be used to check against the Original hand-drawn map; one data element in the plotted map represents 0.6 hectares (1.5 acres) on lhe ground.

When the ground truth had been digitized, the next step was to stratify land cover types in McDuffie County on Skylab imagery with a scanning micro­densitometer. To determine if stralification would improve the efficiency of sampling, we first tried out a very simple type of land classification for a small

PINE

D 'HAROWOOD

10 NON'OREST

I.

MCDUfFIE C0UNTV. GA.

Fiaure 6-Land use maps of McDuffie County, Geo'1ia, wue produced by hand IrId by machine. The map shown in part at ri&)l t was produced by conventional interpretation of 1: 120,000 CIR photos.. Ceometric control was obtained by mappi"l onlO I U.s. Geological Survey base map. (Land use claSRS m.ppcd are I -I , pine; 1-2, hlIdwood; 2, ro~sl; and 3, wate(.) The colored map was produced by digilili"8, at a resolutiotl of 0.6 h. (1.5 aCrcs), the h:uld'm1pped data, ami replotting with a OOmputcr-cGntrolled plon e, . The resolut ion of this map matches thlt of The microdensitomete, data.

20

:: ,,-;: . I -..,--~-

. , - ~.

Figure 7- A Photo Data Systems automatic sca n­nina microdensitometer (above) and companion model 2300 process Compuler were used to ItN and reeord film densilie5 on ccr tapes. The £ Ie(: tro roic Associates, Jnc .. tape-(\riven plotter (be· low) was used to plot color-t:oded cbSliifteation maps.

portion of the county, USing density slicing. Using a 9· by 9·inch ( I :500,000) 3190B color transparency from November 30, 1973 (mission SL4), we pro­grammed the microdensilometer to scan the sele<:ted area with a red filter (Wratten 92), and record film density at Ihe same intervals and in the same rnatrLx

used 10 digitae the hand-drawn ground truth map ; that Is, in an array of 278 (x) by 539 (y), or 149,842 data elements. Next, using density ranges we had es-

21

tablished on sample areas before the programmed scanning began, we classified each d:lta point into one of the four land use lypes. The counterpart of each of these points could then be located on the data tape for Ihe digitized land use map to delennine the ground classification.

A facsimile gray-scale map was pJOduced on a line­printer from the red-fIllered microdensilomelel stan of the Skylab 5190B colo r image. Alphabetic charac­ters were used to represent rour optical-density ranges (density slices). These maps are st retched in the y direction because the spacing is greater between rows o r characters than between characters ::liong a row. Using this same procedure with digitized data from the ground truth map, we produced a geometlically similar Hne-printer facsimile for comparison. The two facsimile maps were compared and the regist ration appeared to be good. Based on this resull, we went on to try a more powerful classifier using lhree spec t/al bands of microdensitometer data.

A more refined method of classification is linear discriminant analysis, using maximum likelihood and Gaussian assumptions. Training sets were located in the microoensitometer scan data mat rices for each of the land use types identified on the ground truth map. Ojllical densities fr om the red , green, and blue microdellsitometer scans, and generated sialist ics, were used 10 develop linear discriminant (unClions 10

classify the entire county InlO land use types from the Skylab data. A facsimile type-map was generated on a line printer using this classification procedure, This type-map is compared with the line-printer ground truth map (jig. 8). The land use map is strati­fied according to the resulting claSSification, and sampling effectiveness is compared with random sampling and with systematic sampling.

Having the ground truth map digitized as a regu­larly spaced grid of data points at 80.S-meter (264-feet) spacing allowed a comparison to be made among &imple random sampling (SRS), systematic sampling (SYS), and systematic sampling with posts3ITlpling strat ificat ion on Skylab data (PSS)-all at 16 different sampling intensities and on a " real" population. Cha r­acteristics o f the systematic grid sampling at the d if­ferent intensities are given in table 3. A computer program simulated placing of the grids over the ground truth map at all possible placements fo r each grid spacing. The percent sampling intensity and ground area per sample point in the table are based on the mean number of sample point s falling in the mapped area for each grid spacing.

For each grid spacing the expected values of forest and nonforest proportions and the variance of the

• N Pine

~ Hardwood

C Nonforest

Water

22

Pigwe 8-The land use maps sfIown hue C()~r the nonhwe$tem portion of McDuffle County, Georgi.., and were used in the tn t of sampling designs.. The "ground truth" map (obo~e) was produced from data digitized from the hand drawn laod Utr map (fig. 6), The other map (below) was produced from the classification of mlcmdensi­tomeler scans on In S1908 color tn..nSfllLfency. Many rIOpforest ueu, moslly roads Of load seg­ments below Ole minimum mappin& au for sepuation on lTOund truth map, au shown on the miexociensitomcte:r lUll map.

estimate were computed for aU possible grid place­ments. From these stalistics the sample bias and error of the estimate were measured.

Automated Land Classification

A Photo Data Systems (POS) microdensitometer and companion Model 2300 pcocess computer (fig. 1), a nearest neighbor classification algorithm devel­o ped rOT LANDSAT daiS (AJdrich and others 1975), and a CDC 7600 computer at the University of CaJi­fornia Lawrence Berkeley Laboratory were used to classify land use in two IO,OOO-meler-square (6.22-mile-square) study blocks. (A more complete descrip­tion of the procedure is available on request-see the Introduction.) The study blocks we re located on one SI..-ylab S I90B color u3nsparency {November

Table 3- ClwrtlClerlslics of grid sizes uud to SIlmple data ITUItrix for lund ute I1IllP, McDuffie County, Georgig

Grid N .... mber spacing ,r Groooo Sampll rc Ground area (dala sample spact"' fuclion "" elements) poinu sample poinl

Meters Feet Pert;tnt Hectares A crel

50 " 4,020 13,189 0.04 1,567 3,813 42 " 3,317 11 ,078 .06 1,106 2,733 36 82 2,894 9,496 .08 813 2 .... 25 171 2,010 6,.594 .16 392 96' 21 '" 1,688 5,.539 .J) 171 684 18 329 1,447 4 ,148 .31 203 SOl .. 544 1,126 3,693 .S1 123 304 12 741 965 3.1 65 .10 90 m 10 1,066 804 2,638 ),00 63 155 9 1,3 17 724 2,374 l.24 51 126 1 2, 116 563 1,846 l .O4 31 16 6 2,962 482 1,.583 2.78 23 56 5 4.266 402 1,3 19 4.00 16 39 4 6,665 322 1,055 6,iS 10 2S 3 11 ,849 241 191 11.11 6 4 2 26,660 161 528 25.00 3 6

30, 1973). One block was used to develop training se ts and forest and nonfores! classifiers; the other block was used to lest the classifiers on a site some distance (rom where the classificrs were developed. The study blocks were purposely selected 10 salisfy two criteria : ( I) each h.ad to have a b road represen ta· iion of land use types with a significant amount of forest and nonforest land and (2) the Skylab ph.oto coverage for each site had to be complete . Bound aries of the two study blocks were delineated on ace late templates. The template was precisely oriented with the photos and fiduciaJ marks; then major h ighways and ot her identi fi able (ealures were outlined with a fine draflwg pen . These templates were used on the microdensitometer to orient the X, Y scans with the block boundaries but were removed before scanning began.

Block 4 was used to develop training sets and classifiers for 16 Level II and Level III forcst and nonforest classes to be recognized in the study (table 2). Four scallfling apertures (28 J.llll, 37 J.llTl, S6 pm, and 74 pm) were used , and the microdensitometer was programmed to read den~ties from the phot~ graph at an interval approximating the aperture

diameter. This procedure avo ided excessive data in· tegration and redundancy: that is, Ihe smallest area o r interest fo r our classification purposes was 0.4 hectare ( \ acre), and the range of aperture settings sampled the J :900,000 scale photograph in units of effective ground area ranging from approximately 0.05 hectare (0. 12 acre) at Ihe 28 pm aperture to 0.35 hectare (0.86 ac re) at the 74 pm aperture. The 37 ~ aperture sampled approximately 0 .09 hectare (0.20 acre).

The mic rodensltom eter was calibrated be fore each run after the proper aperture selling and optical filter were in position . Afler each filter ru n. the mtef was changed and the mt(:rodensitometer recalibrated . A fUll for the complete block was made for each aper· ture $Clling and for fou r mler settings-blue (Wratten· 94), Green (Wrauen-93) , red (Wratten·92), and dear (no ft.lt er)-before continuing to the next aperture. In addition, a step tablet furnished with Itte SLA data package was scanned berore the scanning began and again after the scanning was completed . This made it possible to convert microde nsitoOieter optical density readings to NASNJSC·PTD diffuse densities. The en· tire procedUre was repeat ed for each of th.e four aperture scanS.

Origin ally. we intended to evaluate aperture size by producing an eight·class land use map from each aperture sea!l. Time lim itations would nol allow this, however. and we had 10 modiry the plan. Instead, we

23

produced an eight·level gray-scale Illap for each aper· ture scan . From examination o r the gray-scale maps, it was apparent thai the 37 pm aperture resulted in a ground resolutio n that besl suited the requirements o f this study. Classifiers were then developed for the 161and use: cI:asses and applied to the diffuse densities for block 4 using a nearest neighbor classification procedure .

A color-coded land use map was produced on an EA I tape-driven plotter for block 4 (Jrg. 7). The classifiers developed for block 4 were then extended to the data for block 2 and a ploHer map was pro­duced to detennlne the operational capabili ty of the classification system .

Accuracy of computer classification was checked in two ways. The first method computed the propor­tional areas in each land use class and compared them with proport ions detennined from the ground truth map. A second method checked the positional accu­racy . That is, the plotter map was overlaid with a systematic grid of 480 points for which the classifica· tion on the ground truth map was known. The classi· fication was reco rded and compared with its counter· part on the ground . Plotter maps for both blocks were checked in this way and the number o f correct calls determi ned and analyzed . Po ints falling in five randomly selected I()()().meter cells with detailed classifications were checked t o explain any discrepan· cies.

Results and Discussion

Result s of the forest invent ory studies-resou rce evalua tion, sampling designs, and automated dassifl· cation and mapping-varied , and must be compared with caut io n. 8ecause each of Ihe studies was carried ou t independen tly to meet a sepa rate objective, study co nd itions differed . A recu rring source of difficulty, however, was th e problem of relating the established classifications used in forest invenlory to the ac tual­ities of remo te senSing data . Such data reveaJs land cover at a particular time, but does not necessarily indicate th e land manager 's intention ro r later use of the land . Ground ' surveys may determine, for ex· ample, that areas lying bare in early spring arc in fact designed for agricultural o r fo rest planting, whereas remote senSing can onty indica te that the land is bare. A restructuring o r land classifications 10 recognize what is present on the ground , rather than thc in· tended uSe of the land , would simplify evaluation of the accu racy o f remote senSing data fo r land classi­fication.

Forest Resource Evaluation

The choice of a Skylab data combination for con· ventional visuaJ interpretation had Jess effect on accu· racy than expected. We reached Ihis conclusion by comparing omission and commission errors for land usc classification in one county. using three data combinations (see ProcedUJes). Some omission and commission en ors arc caused by misinterpretation of the photo images. In many instances, however, the errors are the result of chance judgments. For ex­ampl e, a pOint falling on a boundary line between forest and cropland may be called "forest" by the interpreter but "cropland" by the field man . Also, the fi eld man may make judgment errors, but they are usually Ignored. These facts should be cOllsidered in comparing results of the Skylab interpretation with grou nd truth .

As desc ribed in the section on Procedures, combi· nation 3, used to interpret land use in a four -cowlly survey . consisted of overlapping eniargemeills of S I908 color photos produced by NASA, viewed stereoscopically. Scale was 1: 500,000 .

Four·County Survey-A comparison of individual land use proportions resulting from the four-county survey shows close correlations between Skylab and ground truth for some classes (table 4) . Only in pasture and idle land are discrepancies much greater than ± j percent. The differences of -6.61 pe rcent for pasture and + 4.34 percent for idle land fo r the most part represent errors in classification-eithe r on the ground or on tne Skylab imagery-atlribucable to several ca\Jses. Some represent borderline judgments where a point faDs on or near a field boundary; others

reflect seasonal differences between the date of the imagery and the date th e field crews located the points, and still others result from changes occurring after the ground dat a were recorded .

If c ropland , pasture, o ther agricuhure, and idle land classes are combined as "other miscellaneous," tJle difference is less appatent- O.l 801 on theg.round as compared to 0.1595 on the Skylab data (table 4). Because a major purpose of land use classification in forest resource surveys is to measure the forest area base, some inaccuracies in agricultural land use would not be a drawback- the estimate of forest land would be well within the accuracy limit of ±3 percent per million acres of commercial forest land. The data do indicate, however, th at either "other miscellaneous" is slightly underestim:lIed or there is a real increase in forest and urban area. This problem is discussed further in a later section of this report.

Increases in urban development during the 2·yea r span between ground and Skylab data acquisition can account for the greater proportion of urban aIca estimated from SkyJab . Nonce nsus water measured on Skylab normal-color mm is underestimated be­cause small bodies of water cannot be distinguished when background contrast is low. This limitation would be overcome using CIR Him.

Land use proportions listed by individual counties in table 4 show much greater variation between Sky­lab and ground data th an the grouped county da1 a. The smaller sam ple siu and the peculiarities of sampl. ing arc responsible. Norm ally, jf area statistics by county are required , they should be based upon a much larger $3I11ple.

Table 4 - Summary of indMduolland use dO$! proporlicms as de/ermined by ground exomination (July 19 71 J and by inrerprelll' 11011 of Skyfab photograplJy (November 1973) for 101U Georgia collnties

Proportion of land in U$e class Land use

Columbia Co. (nc 49) Lincoln Co. (n=40) McDuffie Co. (n::44) WUl\es Co. (n=17) All count ies (n:2 10) clas5

Ground I Skylnb Gloulld I Sl\ylab Ground I Sky lab Ground I Skylab Ground I Skylab

Forest 0.6811 0.7283 0.6984 0.68 I 2 0.6932 0.7088 0.7597 0.7597 0.71 58 0. 7268 Cropla nd .0293 .0268 .0000 .01 87 .0909 .0639 .0130 .038 1 .0307 .0312 PUlure .0861 .0 166 .1062 .O)(}:I .0581 .0 t7t . t 542 .08 11 . 1091 .043 t Idle .0294 .0367 .0 11 0 .0578 .0) 56 .0852 .0)90 .0682 .0307 .0741 Olhe.

:.gricullure .0128 .0013 .0094 .0062 .02 1) .007 t .0008 .0051 .0095 .005 1 Olher

misce llaneous' .15 82 . 13 14 . 1266 .1I4t .2060 .1733 .2010 .1931 .1 801 . 1595

Urban .0816 .0689 .03 13 .0625 .0611 .0824 .0227 .0422 .046 1 .06 01 Census waler .0548 .0574 .1359 .1391 .034] .0270 .004t .0049 .0452 .04 13 Noncensus water .0242 .0140 .0018 .0031 .0]56 .OOBS .006S .0000 .0218 .005 7

.~~--

'Cropland, pasture, idle, and other agricul ture c\:;J.sses, grouped.

24

In this four-county survey of land use cJassHica· tion, we also attempted to classify certrun forest conditions usefuJ in resource evaluations: stand ori· gin, physiographic site, forest type. and treatment class. QassiJication of slolld origin into two cate· gories, nalOral and aruficial , was unsuccessful. Failure can be attributed to the ground resolution (20 to 30 m or 66 to 98 fl) of the color film. The effect of the ground resolution was to make all stands within types (pine type parlicularly) look aUke.

Physiographic site was classified for 148 forest plots, but the restricted geographical area within the study site placed 90 percent of the plots in one category-rolling upland. Of these plots, 97 percent wece correctly classified. No attempt was made to analyze the remaining data because the observations were too few to be relevanl to normal operaling conditions. It was observed, however, that stereo­scopic coverage is essen tial to successful classification of physiographic site.

Forat types in the four-county area were classi­fied correctly o n Skylab S 1900 color only 50 percent of the time. Accuracies ranged from a low of 16 percent for mixed pine-hardwood to a high of 59 percent for pine. The distribution of 161 plots by Skylab and ground types is shown in table 5.

Two conclusions can be drawn from this test. First, mixed pine·hardwood cannot be separated from pine and hardwood with surticient ac<:uracy to be usefuJ in resource surveys. The problem here is [hat a pine-hardwood classification is based on the decision rule, "Does pine comprise more than 25 percent but less than 50 percent of the stocking?" This rule can be .tested by stem counts on the grollnd. On photo· graphs, however, the decision must be based on crown counts, which cannol be accomplished at the resolution of the imagery used here . Second, upland and bottomland hardwood cannot be separated on normal·color film excep t in the most obvious situa· tions. Experience with CIR HIm, howeve r, indicates that high moisture contenl in bottomland siles can be detected during late fall, winter, and early spring seasons, with higher classification accuracies as a re­sult .

On SI...-ylab 81908 color, fore st types must be combined to maximize their usefulness. For instance, when pine and pine·hardwood plots are combined in a single pine class, the accuracy of classification in· creases to 73 percent-a much more r~spectable

figure. Combining upland and bottomland hardwoods into a single hardwood class improves accuracy to 69 percen!.

Treatment classes, redefined as disturbance classes

2S

for remote sensing, were classified for 40 plots in Lincoln County. To compare relative accuracy at three levels of remote sensing, classifications were made on LANDSAT· I , Skylab S I90B , and I : 120 .000 CIR photographs. The results are shown in table 6.

As would be expected, detection varies according to the ground resolu tion of the sensors and the level of disturbance. The greater disturbances, such as harvesting with artificial regeneration, and clearing or site preparati on, are detected on LANDSAT· 1 data. However, only 63 percent of the plots were correctly classified on LANDSAT imagery, against 80 percent on both Skylab and the RB·5? imagery. One explana· tion for failure to detect commercial thinnings and the more subtle disturbances might be the I ~· to 30 year time lapse between the ground observation and the imagery .

One-County Test of Forest Area Estimation-An estimate of the proport ion of forest area in Uncoln COWity was made 011 a I: J 25,000 color enlargement of SkyJab 8190B (November 30,1973) using conven· tional interpret ation techniques. As described earlier (procedures), an overlay with 16-point sample clusters was used for an intensive sampling. The re­sulting pcoportion-the number o f fo rest points divided by the total - was adjusted using regreSSion coefficients derived from ground examination of a subsample of the 16·point clusters. The County mean adjusted forest land proportion estimate was com­pared with existing Forest Survey estimates, which had been detennined from ASCS aerial photographs and Similarly adjusted . Comparisons were also made between unadjusted grouped land use proportions for Skylab and ASCS. Skylab, RBS7 CIR photographs, and the ground . These comparisons help to evaluate the fuU potential of the Skylab photography.

The duster overlay used on Skylab pho tos for Lincoln County sampled at a slightly higher intensity than the existing Forest Survey design. The Skylab overlay covered the entire county area. whereas the Survey design used every other photograph in alter­nate rught lines. Our design eliminated the chance of oversampling some classes and undersampling others. The results (table 1) show that, according to Skylab, the unadjusted forest proportion is almost 3 percent higher than that obtained from 4·year-old ASeS ph otographs; also, [he "other miscellaneous" class has decreased , urban has increased , and both census water and noncensus water are reduced . These couJd be real differences detected by the more recent Skylab data and the improved dislribution of the photo sampl e. Unfortu nately , there is no way to check this because of tJle time lapse since the data were collected.

Tabk 5-Distributiol! of forest plots by Skylab' and ground jorest type classes in a four-county area, Georgia (correct calls underscored)

Table 6-Accuracy of delecting disturbances by interpreta­tion of data from four sources, Lincoln County, Georgia

5kylab class Pine

Pine 54 Pine-

hardwood 20 Upland

hardwood 13 Bottomland

hardwood Nonforest

1 3

Tot.il, ground ,91

Ground class

Botlom­Pine Upland land hard- hard- hard-wood wood wood

7

4

12

2 o

24

,

i3

4 1

24

4 1

8

Non- Total, forest Sky lab

3

2

2

o 6

i3

70

28

41

11 II

161

Disturbance class

No disturbance Prescribed burn Oearing, release, or

intermediate cutting High grading Commercial thinning Harvesting with

artificial regeneration Oearing of site

preparation Natural regeneration

on nonforest land Artificial regeneration

on nonforest land

Plots correctly classified

Ground ] RB-57) 5kYlab]LANDSAT

22 1

, I 7

21 o

4 o 3

1 1

21 o

4

2

19 o

o

'Stereoscopic interpretation of ovetlapping 5190B photo­graphs.

All classes 40 32 {lOO%) (80%)

32 (80%)

2S (63%)

Table 7 -Mean proportions for grouped land use clasSes, computed from an intensive sampling on Skylab and aerial photography, and from a small subsample on Sky/ab, atrial photography, and ground observation, and used to compute adiusted values (Lincoln Counly, Georgia)

Total Other Census Non-

Data sources clusters Forest

misc. Urban census

water water

Intensive sample, Proportion oj" area

unadjusted: ASCS (J :20,000)' 536 0.6472 0.1351 0.0169 0.1725 0.0283 Skylab (1 :500,000) 667 .6747 .1201 .0260 .1636 .0150 Difference +.0275 -.0150 +.0091 -.0089 -.0127

Subsample, unadjusted: ASCS 40 .7094 .1313 .0156 .1375 .0062 Sky lab 40 .6812 .1151 .1)625 .1391 .0031 RB-57 (l: 120,000)' 40 .6984 .1266 .0313 .1359 .0078 Ground 40 .7141 .1250 .0187 .J 391 .0031

Il1tensive sample, adjusted:) ASCS 536 .63753 5kylab 667 .69229 Difference +.05476

'Panchromatic aerial photography, by Agricultural Stabilization and Conservation Service. 'High_altitude color infrared aerial photography by NASA. 3 Adjusted using regression equations developed from small subsample. Adjusted forest area and standard error of estimate are 42,123 ha ±3.17 percent for ASCS and 45,741 ha ±3.53 percent for Skylab.

26

For the 40-cluster subsample examined on Skylab and ASCS photographs and on the ground to classify land use in lincoln Coun ly by grouped classes, the unadjusted forest proportion est imate from Skylab data was almost 2 percent lower than Ihe ground check and 3 percent lower than Ihe A$CS photo est imate (tab le 7). The latter is a complete reversal from the estimate made from the more intensive sample and reflects differences in the two samples. The subsampJe apparently fell in areas under develop­ment for homes at a greater intensity than the larger sample. The largel proportion of land area estimated in the urban class on bOlh the Skylab photos and the ground seems to indica te this. Interpretation of for­ested areas within residential developments as urb:m resulted in an over-estimate of urban and an under­estimate of forest land. Errors of this sort can be avoided by improving the criteria for interpreting urban bnd.

The 40 subsample clusters were examined on I: 120,000 scale CIR pho tographs to verify the mean Skylab proportions (lable 1). Here forest proportion was found 10 be more than 3 percent higher and urban 4 percent lower than on the Skylab imageiY. This result seems to support the conclusion that ur­ban developments within forested areas are overesti­mated on Skylab. Because the ground estimate of urban is higher than either Sky lab or ASCS, the CIR

estimate can be assumed 10 be low; the true propor· tion of urban area is presumably nearer 3 percent.

The adjusted Sky lab estUnates for forest area, 3600 hectares (8892 acres), were 8 .6 percent hjgher than the ASCS estimales. [)espile lhis, the standard error o f the adjusted forest area estim ate using Skylab data was ±3.53 percent- slightly higher than the ad· justed estimate on ASCS photography. The difference in the estimates can be auribuled 10 differences in sampling intensity, sample distribulion, and mis· classification of urban areas as forest land.

Forest Sampling Designs

The area of commercial forest land summarized from the digitized land use map for McDuffie County (hereafter called the ground truth map) and the area of commercial forest from the J 97 1 Forest Survey agree closely (69.6 percent map, 69 percent Forest Survey). The Forest Survey gives no information, howeve r. on the location of comme rcial forest and o ther land use types; Iherefore, no cros:s-checks call be made with the ground truth map. The area of noncensus water (defined in table 1) in the county, according to the survey, is less than 2 percent. The mapped area of noncensus water differs from the Forest Survey figure primarily because much of the water was excluded from the land use map where the

Table 8 - Bias oJld I'tlriQl/(~f' for estimotif of 10rtSI proportions mode by three somp/illg mt!lhodf (SRS, Simple rOIl</OIll $4l/1pfillg; SYS. sy$/emolh; sompfing: olld SYS·PSS. sySfcmlllie sqmp!il!g willi post$l1mpfing s/ralifiwfiOJIJ lOT 16 $OII/pfjllg [roc/ions McDuffie COIIIII)' CeOt'gio

Grid Bias of Variance of

spudnll S~mpling forest proportion forest proportion

(data fr3ction clements SRS I SYS I SYS·PSS SRS I SYS I SYS,PSS

PcrcclII

SO 0.04 0 0.0006 0.0008 0.00495 0.004 11 0.00334 42 .06 0 .0000 .OOOS .00350 .00278 .00241 36 .08 0 .0002 .0007 .OO2S7 .0023 1 .0018 1 25 .16 0 .0001 .0003 .00t24 .00075 .00074 21 .23 0 .0000 .0000 .00087 .00057 .00052 IS .31 0 .0000 .0000 .00064 .00035 .00032 I' .S I 0 .0000 .0000 .00039 .000),5 .00018 12 .7. 0 .0000 .0001 .00029 .00014 .00015 10 1,00 0 .0000 .0000 .00020 .00011 .00010 9 1.24 0 .0000 .0000 .00016 .00007 .00008 7 2.04 0 .0000 .0000 .00010 .00004 .oooos 6 2.78 0 .0000 .0000 .00007 .00003 .0Il004 S 4.00 0 .0000 .0000 .00005 .00002 .00002 , 6.25 0 .0000 .0000 .00003 .00001 .00001 J 1 J.\ I 0 .0000 .0000 .00002 .00001 .00001 2 25 ,00 0 .0000 .0000 .00001 .00000 .00001

27

Ground truth map, Augusta, Ga., block 4; 1: 120,000 CI A photo by NASA, April 1974

1 Pint! 12 Harvested ClOp 22 Kudzu 2 Plne-nardwood 13 Orchard 23 Marshland 3 Upland hardwood 14 Farmstead 24 Aid&!" swamp 4 8on omland ha rdwood 15 Plowed field 26 TranlPOrtolion and 5 Ufldlsturbed wass 16 Erosion utillti" 6 Distwbe<l grass 17 Si te prtl)¥ation 26 Hom. developments 7 Dead grass lurbanl 27 Commercia l dewlopments 8 New imployed grass 18 Rock 28 Recrealm 9 'mmatura groin 19 Idle land "29 Clear lakes 8/'Id ponds

10 Immal!H' row crop 20 Abandoned bnd 30 Turbid laIt." . nd ponds 11 Mllur. crop 21 T~nsi l iol\ll 31 RMR Ind streams

figure 9 - Ground truth. a Skyilh phQl0, and • map mad. {rom th. Sky'ab d.l. by computer-assisted methods can be oompl1ed pomt-to-poinl here. The compu ter map (opposlle) W.II madc \151", difftue densities converted from optical densities measured. on the 51908 phQlograph. The phQ'OSJ1Iph wu II:CIJIned with I miCJQden5;lometer IJ'Id the optk::aJ densities werc dlJltiztd and recorded on tape_ Tht map was then produced by the computer classification system developed . t Pacific Southwest Sution, using ncuHt· rteighbor the01Y and an off-Une tlpe-driven plotter- with eight (:Glored marJr::in& ptlil.

28

Skylab S190B Nov. 1973

Computer map

• Pine C • Hardwood • • Cutover • II Grassland • [: " Crops

Bare s.oil N

Wildlands

Urban

Water

29

northern boundary of the cou ney crosses Clark Hill Reser;oir, and could not be accurately locat ed. Non­census wal er, therefore, is included with all other !\Onforest for stratifica!ion purposes in Utis study.

Linear discrimination functions used to stratify three Level I and two Level II land use types could not separate pine and hardwood with sufficient accu­racy for invenlory purposes. Thus, all forest land was aggregated into a single forest type for further anal­ysis. The confusion matrix for the training set shows only 5.3 percent of the 10lal forest data elements (506) called nonfofest and only 17.7 percent of the total nonfores, data elements (61 5) called forest. For the entire populalion, 19.5 percent of the total for est data elements (74,290) were called nonforest ; 30 percent of the total nonforesl data elements (32,351) were called forest. These classifications were used to test systematic sampling with postsampling stratifica­tion in the following comparison of sampling methods.

Systematic sampling and simple random sampling for forest proportions are compared in fable 8. The maximum bias was 0.0006 for the smallest sampling fraction (0.04 percent): for sampling fractions (0 .23 percent and greater, !.he bias was less than 0.00005. The variance for SYS was smaller than for SRS in every case. The advantage increased as the sampling fraction increased.

The systematic sample was repeated , but this lime the type·map classified from microdensitometer scans of Sky\ab imagery was used for pomampling strati· fi cation. The biases again were negligible, although larger than for SYS. The variance using SYS and PSS was smaller than for SYS alone for !.he smallest sampling fractions. At a sampling fraction of 0.04 percent the variance is 19 percent lower than SYS and SO percent lower than SRS. The advantage of using poslSampling stratifica tion with SYSlematic sampling quiddy decreases as the sample size is increased.

The study has shown that forest area proportion is mOle efficiently detemtined using systematic sampl­ing rather than simple random sampling when applied to an area with forest distribution as in McDuffie County, Georgia . The precision of the estimate is further increased by using postsampling stratification from the computer damfication of Skylab S190B microdensitometer data. However, there is no advan­tage when the sampling fraction is larger than 0. \6 percent. If it had beell possible to direct more effoR to developing beller discriminant functions for the supervised Skyiab data classification, an advantage might have been shown for lacger sampling fractions.

Ground truth map, Auguna, Ga., block 2; 1: 120,000 CIR photo by NASA, April 1974

1 Pine 12 Harvested crop 22 Kudzu 2 Pine-hardwood 130rch_d 23 Marshland 3 Upland hardwood 14 Farmuead 24A~swamp

4 BoUomland hardw¢od 15 Plowed field 25 Transponalion and 5 Uncli$turbed gtass 16 Erosion ulil,1i1l5 6 O,$1utbed 9'35$ 11 Site preparation 26 Home developmenTS 7 [nadgrass (urban) 27 Commenial developments S New improved ",ass JS Rock 28 Recrealion 9 Immature grein 19 Idle 13nd 29 Clear lakes and ponds

10 1I1'1IT\IIII.,o,e row crop 20 Abilndo..ed land 3Q TlKbid lakes and poncIs 11 Mature crop 21 Trllnsitklnll l 31 RWers and streams

Fiaurc 10_ Thc computer map (oPpolite) WIS made by the same method used for the map shown infigwe 9. Specml signatwes devdoped. fOr block 4 were used in this classifkdkm without modification. Point·by­point evaluations can be nude with the sround truth map (above).

30

Sky tab 1908 Nov. 1973

Computer map

• Pine r '---' • Hardwood • • Cutover • iii Grassland • [2 CroPt

Sare soil

Wildlands

U""" Water

The claSSification effon pointed oullhe difficulty of trying to claSSify photo data into categories which reOect the illtent of land use rather than actual ground conditions. With the inevitable revislon of resource classifica tion ~stems into a common and aggregative type of inventory system it should be possible to use unsupervised classification more effec­tively to provide a useful data stage for future sampl­ing systems.

Automated land Classification

Optical fllm densities measured on nonnal-color ftlm (S I908), and analyzed by the nearesl-neighbor classification algorithm. showed promise for computer·assisted land classification. Only broad Level r land classes and Level 11 forest types, how­ever, could be separated with reasonable accuracy (table 9). For instance. in bLock 4. pine area was classified 17 percent lower and the hardwood area 40 percent higher than ground measurements. The latter difference seems high, but considered as a percentage of the total area it is onJy about 6.S percent (22.82 percent vs. J6.28 percent)_ In block 2, pine area was classified 19 percent lower and hardwood 14 percent h igher than ground truth, even though block 2 was classified using training stU derived from block 4-61 kilmoeters (38 miles) away . The total forest area in block 4 was 6 percent higher th.m ground truth . Pine was always underestimated and hardwood was always overestimated.

Total nonforest area, like to tal forest area. was estimated within reasonable accuracy- block 4 within II percent and block 2 within 7 percent-and both were underestimated. Area estimates by Level n non­forest classes varied considuably from the ground truth and can be explained by seasonal differences between the Skylab data (late November) and the ground truth (late spring). nlis problem will always arise in machine classification unless land classifica­tions are restructured to recognize what is present on the ground rather than an interpretation of the inten-tions of local land managers. The best time of year to measure the landowner's intended use is ill the late

N sprins. when die fields are plowed or planted and can be separated from grassland and idle land.

31

Areas o f water and urban land estim ated in both blocks 4 a.nd 2 in some instances appear to be qu.ite accurate . This is misleading, however. because major bodies of water in both blocks were called bare soil, grassland, and crops. Water was misclassified because high levels of sedimentation created false signatures. Urban areas were misclassified as grassland, bare soil.

and crops, because on color mm the spectral signa­tures are very similar. Roads and utility rights-of-way, passing through forested areas, can be detected by the spatial alignment of digital elements but not by their spectral characteristics.

The computer maps shown in figures 9 and 10 can be visually compared with both the Skylab SI90B photograph and the ground truth map. The agree­ment between Level I forest areas and non forest areas, and between Level II forest types, is quite apparent. A point-by-point comparison was made be­tween the ground truth map and the computer map. We found that 92 and 93 percent of forest points were correctly classified on the computer map for blocks 4 and 2, respectively, table 9). Of the points in pine type, 70 percent (block 4) and 83 percent (block 2) were correct. Points falling in hardwood were correct 68 and 74 percent of the time for blocks 4 and 2, respectively. Although nonforest points were found accurate 74 percent (block 4) and 85 percent (block 2) of the time, the accuracy for individual Level III classes was extremely poor.

The results show that forest area can be stratified on Skylab-quality photographs (20- to 30-meter reso­lution) with an accuracy of approximately 96 per­cent. In other words, the forest area can be mapped and an estimate of the area made within the limits that can be met using aerial photographs. We also found this to be true in our LANDSAT-l experiment (Aldrich and others 1975). Geometric distortions in the LANDSAT image, however, did not allow more than 50 percent accuracy in point-by·point compari­sons.

Applications Under certain assumptions, Skylab-quality photo·

graphs will provide a data base for Levell and Level II land use and broad forest-type stratifications (as defined in this report) by both human and machine­assisted procedures. The assumptions are that

1. High resolution color infrared photography is available.

2. Radiometric measurements by supporting air­craft flights are available within a few days of a satellite pass.

3. A high-speed scanning microdensitometer is available for recording optical ftlm density on com­puter-compatible tape.

4. An off-line minicomputer and interactive dis­play (CRT) are available for preprocessing remotely sensed data for input to a large nondedicated high­speed computer for classification mapping or

32

sampling. If a minicomputer is not provided, an inter­active display-on line with a dedicated large, high­speed computer-must be available.

5. A classification system is designed wnich is based upon land cover rather than land use intent.

In this investigation, nonnal-color photography taken from orbital altitudes did not allow complete land use classification even at Level I. Here the failure of both human and automated classification proce­dures to resolve differences between water, grassland, bare soil, and idle land was important. Examples of Skylab CIR photographs for nearby areas indicated, however, that the infrared reflectance and absorption differences for these classes will provide excellent separation. Thus, if CJR ftIm is available, both human

Table 9-Accuracy, based on ground /ruth, of area estimates and sample point classification fOT PSW computer mapping of two te~t blocks from Skylab S190B color photography, November 30,1973

Block and land-Area, !uea,

Nllmber of Classification use class

ground' Skylab sample accuracy,

(Levels I &- II) points Skylab

H, Percellr Percellf Percent

Block 4 6055 60.55 64.41 290 92 Dlock 4

I Forest 6055 60.55 64.41 290 92 II Pine 4427 44.27 36.62 217 70

Hardwood 1628 16.28 22.82 73 68 Cutover -, - 4.97 0 -

I Nonforest 3616 36.16 32.14 180 74 II Grassland 1548 15.48 9.51 78 53

Crops 61 0.61 6.57 2 0 Bare soil 881 8.81 6.90 52 37 Wild vegetation 441 4.41 4.96 26 27 Urban 685 6.85 4.20 22 14

I Water 329 3.29 3.45 10 20

Block 2 I Forest 5151 51.51 53.48 271 93

II Pine 3327 33.27 26.95 150 83 Hardwood 1824 18.24 20.84 121 74 Cutover -, - 5.69 0 -

I Nonforest 4721 47.21 44.04 204 85 II Grassland 1473 14.73 8.25 57 54

Crops lSI 1.51 5.00 23 35 Bare soil 700 7.00 22.92 33 67 Wild vq;etation 805 8.05 4.41 26 42 Urban 1592 15.92 3.46 65 17

I Water 128 1.28 2.48 5 80

I Areas were detennined by dot COllnt Oil grollnd truth maps at an intensity of approximately one dot per 0.4 hectare (1 acre). lCutover land was not classified separately on the grOllnd trllth map: cutover forest land was placed in the appropIiate forest type.

and computer classification of land use and forest type should be more effective.

In operational applications of machine-assisted classification of Skylab-quality photographs, spectral differences within and between photographs, and be­tween seasonal (temporal) coverage, caused by changes in solar angle and atmospheric interference, will be limiting factors (see "Measurement of Forest Tenain Reflectance" ). 1f, however, satell ite phot o­graphic coverage is either p receded by , o r followed by, radiometric measurements from a low.flying air­craft, then correction factors can be computed and applied to the data. This would permit exte nsion of ground truth and reduce cOSIS. The radiometric data must be collected within a few days of an overpass under similar atmospheric conditions.

AHhough mapping of all the forest and rangeland (610 million hectares or 1.5 billion acres) in the United Slates may some day be desirable, limitations in computers and com puler storage make this some· what unlikely for extensive invenlories at the present lime. Instead, it is much more reasonable to think of sampling applications. For example, titis repo rt demonstrates that a systematic sample grid can be overlaid upon digitized land use map data by com· puter to estimate forest area within a county-the variance was always lower than simple random sampl­ing. Using Sky lab no rmal-colo r mm to classify forest and nonforest land in one counly (lOO percent) re­sulted in an accuracy of 80 percent for forest land with a 30 percent commission error. Forest cover types and olher land cover types may be estimated by sampling digitized data from future satellite coverage if CIR film is made available and if a classification system on the land cover rather than intended use is designed .

Classification System Alternatives

TIle Forest Survey delillitiOIl of forest land iIIus· trates the problems and inefficiencies of uSing a data· gathering system which views present land cover in order to claSSify use based on the intent of the manager . Using this definilion , a ciearclll area pre­pared for planting would be considered forest al­though no trees were present. Computer-aided classi­fication would probably claSSify the alea as bare soil . In some parts of the country a fo rest opening ofbare soil might be assumed to be a cJearcut to be planted, but in the Southeast the area Is just as likely to be agricultural land.

There is much greater probability of !Ugh accuracy if classification is made in terms of present land

33

cove •. The spectral signatures of c1assWtcation based on present land cover have inherently less variability than those based on intended use, which may include diverse land-cover categories. It is true, however, that a system based solely on land cover. although appeal­ing from the standpoint of identification from remote sensing data, may not be amenable to aggregation and disaggregation into the classes desired by the dala users.

All alternative classification syslem might be one that makes use of information directly attainabl e from satellite data and calls fo r a separale subsample of those classes no t well characteriud by the land cover. Before the system could be implemented, how­ever, it would be necessary to know the land cover types that can be recognized in a given region. To date, mOSI studies have concentrated on detennining the accuracies attainable when classifying land int o predetermined categories. Results of these studies have often been discou raging because 3. satisfac[ory classification accurac» could not be anained for the predetermined ca tegories. At the same time, the pos­sibilities for further subdivision within easily identi· fied categories went unexplo red. We feel that opera­tional applications of machine-assisted classincalion

Table t o-Compar.lIi~t to!t of/alJd IIU and fOf'eJt srrati· jiOllion 0(1 Skywb photographs and 011 1:10,000 ASCS photographs

Cost Item and basis

ASCS I SJ(ylab

Photo handling

Ordering,organil ing $ 14 0.00 $125.00 and labeling pho to-gr3. pll$ and uans­felling pIOIS-$5 per lloor.

Photography J974 ASeS price 366.00 100.00 cost list; 00$1 of one

51908 colol intet · neg.l tive al,d one 1: J 25,000 color print.

I'tlol~ Land use : ASCS 36 .35 50.90 inlerpre- at 5 clusters per tation minute, SkyJab

at 3 e!us ters per minu te- $S per hour; forest type classification: ASCS at 40 plots per hour- SS pel hour.

Total $542.35 $215.90

will depend on additi onal research to determine the l::!nd cover categories which ate, and those which are not , spectrally separable with acceptable accuracy.

Cost Comparisons

The primary advantage of Skylab-qu ality photog­raphy in forest resource surveys is Ihe broad areal coverage within a single frame. Ln Ihe four coumies used in IlLis experiment , IS3 aerial photographs (I :20 ,000) were required to cover an estimated 80 percent of the tOI3J area. This was single pho to­graphic cove rage witham the advant ages of stereo­scopic overlap. A Skylab S I908 photograph. on Lhe other hand , can cover thest fOllr counties and two to four additional counties as well . Complete county coverage ofrers better distribution of photo samples and reduces data handling and photo acquisition costs- assuming th at only printing and processing costs are considered.

If all other survey costs are assumed equal, the costs of Level I and Level II land use and forest stratification would be 49 percent lower on Skyiab phot ographs than on conventional 1:20,000 scale photographs (table 10). The major difference be­tween the two methods is the cost of photographs. The small scale of Skylab photos and the use of normal-color film meant ihat more time was required to make interpretative decisiotlS on Skylab. However, if color infrared ~hotogtaphs were available on a regular recurring basjs, time would be somewhat re­duced, and the advantages of up·to-date infonnation would far outweigh any addi tional interpretation cost .

Because of variation! in system effi ciency and ef­fectiveness, it is difficult to compare conventional interpretation and computer-a$Sisted classifica tion and mapping techniques. Numerous computer classifi­ca tion procedures 3re now available for resource anal­ysis, but (here have becn no comparative tests. Com­puters and ao;essories vary widely. There are non· dedicated computers accessed through terminals on a lime-sharing basis, dedicated computers with on-line interactive displays, and complex systcms including off-line minicomputers and interactive displays to preprocess data for classifica tion and mapping on large comput ers. An analysis of the efficiency of one system over another is lacking_

In this study a system was used that included a

34

Table II-Comparison of cost of producing dass/fiCln/on //lap for 40,000 lIecrare (98,800·acrt) area u$ing convell tlOlial mappillg alld computer·assisted Uchniquts on CIR phOtography

-Time used Cost

Method and task description Man~ l Machine- Man;, 1 Machine hOlliS time hours' time

_ doUart -

Conventional interpretatioll

Map and tnnsfel class boundaries from I : 120,000 CIR 10 t :24,000 map base tOO.OO • 6 17 •

Compu ter.gssisted classificotion

Micwdensi tomellY 2.00 4 h 12 ". Compu ter classifICa tion: HIST -rewrites law den!i ty

tape and intedeaves scans 0.50 356 $ 3 ' 41 IGREY -grey~cale printout ;

area tocation. tnining sel selection 16.00 560 s 99 ' 5.

TAPEJOB-rewtitcs tape h om IHIST in BCD 0.25 700 , 2 ' 63

PROTO-produccs sums of data used in TYPEPIX 2.00 6, 12 "

TYPEPIX-ploduces classification map (4 parts) on EAI plotter 4.00 389 s 25 ' 245

Total 24.75 4.56 h 153

, GS-9, $6.17 pCI hour rounded to the nearest do llar. 'Commercial nte, $60 per houl . I Univac I J 08; $0.09 per se~ond.

64 1

'CDC 7600; charged for 34 acco unting units (AUS) at $0.07 per AUS. Averages out al $O.3S per second. s CDC 7600; average fOI 4 HUlS" 2 ]8 AUS al SO.07/AUS. Avefag;~ $0.63 per second.

nearest-neighbor classifica tion algorithm and two time-sharing computers. Withou t an interactive dis­play, two or more iterations of lhe classincations were required to improve accuracy. This is more time consuming than an interac live display and less effi· cient. The cost breakdown in fable II compares our PSW computer-assisted technique with conventional mapping and transfe r procedures on a ZTS. Although the computer technique was more expensive ($ 177 more), it requited only one·fourth the man-hours. Thus, when manpower and time 3re short, the machine-assisted system would be beneficial desp ite the higher cost.

RANGE INVENTORY

Classification and Mapping of Plant Communities

Richard E. Fr3ncis

The primary objective of the research at the Manitou site was to determine at what classification level Skylab photographic products, or their equivalent, can be used for plant community classifi· cation in a central Colorado mountainous area. Secondary objectives were ( I) to determine the kind of aircraft support photography needed to extend these classifications to other levels in the hierarchy , (2) to determine how both Skyiab and aircraft sup­port photography can be used 10 detect and identify cullura! features of a mountainous landscape that could affect resource management alternalives, and (3) to make quantitative estimates of certain plant community characteristics from large-scale aircraft photography .

This is the latest in a series of studies that began in 1967 to evaluate remote sensing technology for range inventories. The Rocky Mountain Forest and Range Experiment Station, l~ated in the center of a variety of range plant community systems, was a proper foc al point for this research,

Initial stud ies conducted in 1967 invest igated large-scale 70-mm color and color infrared (C IR) photographs ( \ :600 to I :4600) fo r identifying plant species (Driscoll 1969)_ The conclusion from this work was that major rangeland species can be identi· fied most consistently 00 1 :600 and I : 1200 scale CIR photographs.

In 1968 , the Rocky Mountain Station entered into a 4-year study under the NASA Earth Resources Aircraft Program to evaluate multiband photography and muhispectral sca nner imagery taken from both aircraft and spacecraft. Resull s indicated that CIR photographs taken from space (Apollo 9, IOO·meter ground resolution) can be used to map gene ral plant community systems (Driscoll and Francis 1972). Larger scale ( I :20 .000 to I :80,000) airc raft phot­ographs with J. to lO'meter ground resolution were required. however, to determioe the areal extent of the individual habitat types. Still la rge r scale aerial phot ographs (> 1 :2400) were needed to analyze com­munity components. lnilial efforts to relue optical film denSity me:JSlued on small-scaJe CIR photo­graphs to plant communilies showed promise.

35

Richard S. Driscoll

In 1972 , the Rocky Mountain Station and its range invent ory research team initiated a study to assess the value of low-resolution (65- to loo-meter ground resolution) LANDSAT-J CERTS-l) multi· spec tral scanner data for classifying plant communi· ties (Driscoll and Francis 1975). This study showed that LANDSAT imagery acquired in August was best for the first level of classification. Aspen can be reliably separated from conifer only during this time of year (92 per~nt accuracy). Conifer and grassland, on th.e o ther hand , can be classified on June-to­September LANDSAT and high-flight aerial photos with 95 to 99 percent accuracy. Regardless of season. interpreters could not classify to the second level with acceptable results. An analySiS of optical film density measured on LANDSAT photographic color composites showed highly significant differences between aU first-level vegetation classes. Thus, auto­matic scanning of LANDSAT photographic data and computer-assisted classification of first-level vegeta­tion classes seemed poSSible.

Sky lab ph o tographs used in this st udy have better ground resolution (S 190A, 30 to 120 meters ; SI90B. 20 to 30 meters) Ulan either Apollo 9 or LANDSAT images. An evaluation of this improved dala wa<>

importan t for plwning future earl h resou rces satelli te programs (or fo rest-rangeland assessments.

Study Area nle study area lies betwee n 38°30' and 39°30'

north lat itude and 104°40' and 106° 10' wesllOflgi­tlIde and covers approximately 14,000 Jcm 2 (5400 sq mi ; fig. I I). Included within the area is the NASAl Manitou Test Sile, No. 242. This nonurban, non· agricultural area in' centra1 Colorado is characte rized by exlreme diversity in planl community systems and ex treme variations in topography. The vege tation in the area consists of a variety of forests and grasslands. The forests, ranging from approximately 1900 m (6232 ft) above me an sea level 10 tree line at apprOXimately 3500 m ( It ,480 ft) , include: (I) ponderosa pine, (2) Douglas-fir (PS(!UdOfsllga mCI/­ziesiI' var. glauca IBeissn.) Funco), (3) lodgepole pine

, , , , /

: SOUTH

"

4

, , ,

'0

1,100,10" E'~"m.ntol Fo,~"

D, ;" de

p ,~", P~o~'"

o C"pple C'ee~

'0 '0

Figure II-The mounlaillous central Colorado study area includes two intensive test sites, South Park and the Manitou Experimental Forest. Colo­rado Springs is near the southeast corner of the ~~,

(Pinus con/ana var. Iatijolia Engelm), and (4) spruce /fir (Abies lasiocarpa [Hook.] Nutt.) Tntemlingled throughout the area are deciduous forests of quaking aspen (Populus tremuloides MicllX.). Mountain bunchgrass parks, in wltich Arizona fescue (Fcstuea arizonica Vasey) and mountain muhly (Muhlenbergia montana [Nutt.] Hitchc.) wefe the dominant grasses, occur in the lower elevation areas and are principaliy associated with the ponderosa pine forests. These dominants are replaced by other species of fescue at higher elevations. In many instances the gradation

36

from forest to grassland is subtle, and it was difficult even at ground level to establish the line of demarca­tion between the two systems (jig. 12).

Within the central part of the siudy area is South Park, a large, nearly treeless area. The vegetation of South Park is generally of low stature in which blue grama (Boute/oua gracilis [H.B.K.J Lag.), slirnstem muhly (Muhlenbergia Ji/iculmis Vasey), and several low-growing shrubs and forbs are the most promi­nent. These and associated species provide the aspect of a shortgrass prairie. Around the fringes of the Park, and in some places within the Park where herbaceous communities interface with the forests, mountain bunchgrass communities become prominent.

Wet meadow and stream bank communities are especially well developed in South Park and occur as occasional narrow strips throughout the entire area. In association with the meadows there are shrubby communities dominated by species of willow (Salix L.) and shrubby cinquefoil (PotentilIa [mUcosa L.).

The main portion of the study area varies in eleva­tion from 2100 to 4300 m (6888 to 14,1C» ft) above mean sea level. Elevation variations are dramatic-as much as 400 m per kilometer (2000 ft per mile) in many places. The average elevation of South Park is 2750 m (9020 ft) above mean sea level. (A more complete description of the study can be found in the report of the ERTS studies-Driscoll and Francis 1975.)

Classification System

The hierarchical vegetation classification scheme used to evaluate the effectiveness of the Sky lab and supporting aircraft data was ECOCLASS (pfister and Corliss 1973), which was established according to

Figure 12-111e gradation and mixing between and within the fOlest and grass­land vegetation system is evident on the ground. The forest system is a mixture of Aspen and Spruce/Fir; Mountain Bunchgwss (on slop~) and a high eleva­tion Wet Meadow (nl1ey floor) are the grasslaItd systems.

ecological principles o f polyclimax conce pts (Dauben· mire 1952). This system is in curren t use by the Fo rest Service to classify plant communities for I,and man agement planning, and is in accord with Ih.at established by the 1nte rnational Biological Program for classifying terrestr ial communities (peterken 1970).

The system defines live categories proceeding from the most general 10 Ihe mOSI specific, as foUows:

V. Formation- The most general class o f vegeta­tion, characterized by general appearance: Grassland, Coni fe rous Forest . Dedduous Forest, e tc. The basis of this calegory is continental in scope. i.e ., all of the Uni ted Slates. and is controlled by continental cli· matic differe nces.

IV. Region- Subdivisions of the Fonnation , asso­cia ted regionally and therefore de tcnnined by sub· clhnates within con tinental climates: Montane Grass· land , Temperate Mesophytic Conifero us fo rest , Al­pine Grassland ,. e tc.

Ill . Series- A group of vege tation systems within the Region category, wi th a common dominant cli­max species: Ponderosa Pine Forest, Fescue Grass­land , Herbaceous Meadow, e tc.

U. Habitat Type-Units within a Series , each with relatively pure internal bio tic and abiot ic structu re: Ponderosa Pine-Arizona Fescue habi tat type. Ari­zona Fescue-Moun tain Muhly Habita t Type, etc. These are 'he elemental IIl1ils of the ciassi/icario" scheme upon which primary management is based. These unils are frequently related to climax vegeta· (ion, or to vegeta lion held in a relat ively stable state of high succession by proper m:magement.

I. Community Type- A system thai appears rela· tively stable under managemtnt and may be equiva· lent to the Habital Type . Usually the biotic oompo­nen ls 3re dissimilar, but abio tic components are anal­ogous to Habi tat Type.

The en tire area is considered to be within one general physiographic province, the Central Rocky Mount ains, in ..... hich three Fonnation clilssesexisted: Grasslands. Con ifero us Forests, and Deciduous Forests. With in lhese Formation classes, three Region and eighi Series cla~ were defmed for this study:

IV . Region III · St.rlrs

L Coniferous Foresl 1. Ponderosa Pine 2. Lodgepole Pint 3. Douglas-flr .s . Spluce{Fir

2. l)e("iduous Forest I . Aspen

3. Gussbnd I. Shortgtass 2. Mounuin Bunchsrass 1 . Wei Meado ....

37

Skylab and Support Data

Only S I90A and S I908 pho tographic data for June II. 1973 (Sl -2 , Pass 8, Track 48) and August 4, 1973 (SI..-3 , Pass 13. Track 48) were used in tttis anaJysis. Th e June mission oovered 85 percent ofthe area ; the August missio n covered JOO percent on the S I90A and 90 percent on the S I908. Of the S I90A multispectral camera system products (scale I :3,000,000). only color and CIR 7Q·mm photo­graphs could be used. The Sl90A material was avail· able for both J une and August but the overall image quali ty was o nly fai r. Photographs from SI90B earth terrain camera system ( t :950,000) were available only from the August mission. The overall quality o f these pho tographs was very good .

OUf initial proposal also called fo r the aJlalysis o f dat a from both the S I9 1 infrared speetrometer and the SI92 muhispect ral scanner systems. The image ry from both of these systems was excl uded from the study bccause delays in obtailling the rec tified da ta made analysis impossible in the time available. The sampling design ..... e used (or testing .photographic p rod ucts was based , however . on the expected resolu· tion o f the S I 91 system (435 m,I 427 fI).

Two aircraft missions- June 22 , 1973 and August 9, 1913- were flown by the NASA aircraft support program to assist the interpretation of the Skylab products for the defined objectives and provided both color and CLR photographic p roducts at three scales ( 1:50 ,000, 1:100,000, and 1:400,000). The quality o f the two larger scales was very good, and coverage ranged from 85 10 100 percent. TI\is photog­raphy was timed to represent plant phenological COll '

ditions on the dates of the Skylab passes. In addition , bo th color and C{R large-scale 7()..mm sampl ing photography missions were nown by a U.S. Forest Service aircraft at the scaJe of I :600. on August 24, 1973- wit hin 3 weeks of the Sl-3 ph.o tographs. Only the CIR was analyzed to estimate speci fic quantita­tiye grassland plant communit y parameters. Film quality was poor because of severe reliculatton.

Ground Truth

Base maps made.f rom existing Fo rest Service range and fo rest vegetation type·maps were used to verify vegetation mapS prepared from Skylab and unde r· fU ghl pho tographs . To prepare a base vegetation map from Forest Service type·maps, Series level (ECOCLASS) class boundaries were traced directly on the base. Because the Forest Service type·maps were funclionally developed. ho ..... ever, they had to be reorgrulized o n an ecological basis. This meant com·

bining and integrating some of the existing type­groups to meet our objectives . As an example, on the forest type-map, Aspen was separated into commer­cial and noncommerci al classes, but for ou r purposes, the two classes were combined.

The revised vegetation maps were checked on the ground with a 10 percent sample of th e 569 samples (each 500 m1 or 1640 sq ftl used in the photointer­plelation study. Less Ihan 5 percent of the ground samples were found to be misclassified.

Foliar estimates made on large-scale color photo­graphs were verified using ground data obtained from two grassland sites within South Park- Eleven Mile and Amero. Five plots were established at lhe Eleven Mile sile and four plots at the Antero site . Each plot was 10 meters (32.8 feet) square . Within each plot. four IO-meter transects we re established and marked with white wooden su rveyor stakes laid nat on the ground. The transect locations were marked before the ph.oto mission so that they could be located in the aerial pholography .

Estimates of plant species foliar cover, litter, and bare soil were ob lained using a line intercept tech­nique (Canfield 1942), modified to use IO-meter tran­sects rather than the 1000foot transects specified.

Procedures Procedures used to meet the study objectives

included visual interpret3tion of photos with high­powered magllifying stereoscopes, mapping with a photo image transfer device, and classification based on machine·measured optical film denSity.

Plant Community Classification by Photointerpretation

Vegetation type-maps, topographic maps, ground reconnaissance , and aeria] photographs were used to selec t and loca te sample cells for interpretation of bOlh the Skylab and support atrial photographs. The sample cells were initially selected and plotted on the maps to represent an area approximately 500 melers (1640 fee t) square. The size of the data cell was determined by (I) l1te expected resolution of the Skylab SI91 EREP products (originally part of the analysis bullater excl uded-see Introduction),(2) the expected positional errors in both the satellite and data collections systems . (3) the probability of plot­ting errors in transferring sample cells from maps to the Skylab and support ing aircraft photographs, and (4) the need to minimize edge effects of cell-wall lines. For training and testing pho\ointerpretation, at

38

least 20 cells were selected in each vegetation class, for a total of 569 cells class as follows:

Region and Seri~s class:

Num bcro[ cells

Coniferous 44 6 Ponderosa Pine 1 SO Lodgepote Pine 62 Dougl:J.s- l'i r 94 Spruce/ Fir 140

Deciduous SO Aspen SO

Grassl:Jnd 73 Shorlgrus 29 Mounl3in Bunchgrass 24 Wei Mcadow 20

TIle number of sample cells in each film , scale, or season test may have varied from these tOtals because photographic coverage was inadequate.

Transparent overlays were constructed for the Sky lab photographs to show the training and test celt locations for the study area. The Unive rsal Transverse Mercator coordinates representing the location of each cell were precisely plotted to a scale of 1:100,000 on the overlay. Obvious landmarks were also plotted to assist in pOSitioning, the overlays on the Skylab frames. These overlays were then photo­gr3phically reduced on 0.004 mil clear positive mm to the appropriate scale.

Interpret31ion covered the SL-2 and SL-3 SJ90A color and color infrared photographic products, lhe SL-3 SI90B color photographic products, and Ihe June and August eclor and color infrared supporl air­craft photography .

Wilh the overlays in posilion Oil Ihe photographs, each interpreter independenlly examined Ihe train.ing, cells until he was satisfied that he could identify Ihe classes. Next. he examined the available test cells and classified Ihem inlo lhe Region Class and then the Se ries class. [n trus study deSign, interpreters were asked to first identify the Region and then the Series level classification (or each sample before moving to the next. (Because Aspen is the only prominent Deci­duous species in the area, the Region and Series classifications were the same in this instance.)

The same sample cell locations were used again to interpret the aircraft support photogr3J.lhs_ Cell loca­tions were transferred directly from· the vegetation and topographic maps to overlays on the 3eriaJ photo­graphs.

A factorial design for analysis of variance was used to test fo r differences between the appropriate fac­torS of fLIm type , photo sca le, flight date, interpreter, and plant community class. All factors we re con­sidered to be fixed effects. The highest order inter­action teml was used as the e rror tenn to obtain the F statistic.

Plant Community Mapping

U.S. Forest Service vegetation Iype-maps were used as base data to draw plant communi ty classific­ation maps at both the Region wd Series level for areal comparison with maps drawn from Skylab and support aircraft photographs. The base map covered 138 kin' (54 sq mi) within the Manitou area (fig. 11) and was initially drawn to the Series leve\. Region level maps were made by consolidating Series classes into lhe three Region classes of Coniferous, Deci­duous, and Grassland.

The base map underestimated the classes. This was due to the difference in criteria used to identify Wet Meadows in the development of the base maps. For instance, the original timber-type base maps had no Wet Meadow category, and the range type·maps had a very limited Wet Meadow category which did not cor­respond with ground information from the Manitou site . Therefore, artificial Wet Meadow class bounda· ries were determined from the or iginal type·maps and ground information.

The phot ographic producfs used were the SL-3 (August) S190A color transparencies and lhe MX-248 1:100,000 scale e IR trallsparencies. These we re not truly comparable ftIm types, but they offered the best photographic quality available from the two sources. The area to be typed was imaged wi thin one frame of the aircraft photography, thereby reducing iOlerframe difrerences. Photographs from S1.-2 SI90A and SL-3 SI90B could not be used because of cloud cover over the Manitou area.

Vegetation maps were made to the Series level from stereoscopic examination of the Skylab and support undernighl photographs separatel y, and were independent of each other. Once the maps were com­pleted, they were brought to the same scale for com· parison using a Bausch and Lomb Zoom Transfer Stope.

Areal comparisons between the base map and the Skylab and underfHght maps were accomplished using planimetric methods.

Cultural Fea1ure Mapping

To examine ways of using Skylab photographY to

39

interpret and map cultural fea lures, an area of 156 km2 (61 sq mi) was selected within the study sile. Cultural features were mapped from Skylab and wlderllight photographs and compared with cultural fea lUfes traced from a 1956 U.S. Geological Survey quadrangle.

Working under the Zoom Transfer Scope (ZTS), interpreters used S1.-3 S190B photographic produ cts enlarged to twice their original scale-l :500,000- and underflight CIR photos at I : 100,000. In att empling to map from the Skylab material , the interpreters found that the majority of the detail was lost in transferring information to a map usillg a ZTS. There­fore , they first placed an acetate overlay directly on the Skylab photo and wilh the aid of a stereo­microscope and a needJe, they etched the locatioll of road systems into the overlay . The overlay. enlarged on the ZTS, was used as a reference to map the other cuHural features from the original Skylab photos.

Next, a photo-revised cultural fea twe map was made, using additional informat ion obtained from the underfligtu photos. Because resolution on the under­fli ght photography was better, the road system net was transferred under the ZTS directly to the overlay base map without the etching procedure . Then we mapped the other cultural fea tures from the under­flight photos.

Foliar Cover Es1jmation

Large-scale (I :600) 70·mm CIR photographS taken rrom a Forest Service aircraft we re used to determine the proportionate amounts of live vegetative foliar cover and bare ground for selected grassland sites within the study area . TIle ultimat e intent is to use these kinds of data in quantitative sampling of ground surface characteristics as imaged in Ihe satellite photographs.

Transects measured on the ground were located on the large-scale aerial photographs, and foliar cover was measured using three means of lmage magni fi ca­tion : a zoom stereoscope. a hand-held measuring micrometer, and the viewing screen of a GAF micro­densitometer.

Unear regression correlation analysis was used to determine the relationships belween image-measured and ground-mea$Ured estimates.

Plant Com munity Classification by Microdensi1ometer

A GAF model 650 microdensitometer (fig. 13) was used to evaluate the Skylab photographic prod­ucts for classifying plant communities. (A micro-

densitometer measures the opt ical density of an image on a photographic transparency using a cali­brated light source.) Skylab S 190A nonnal-color and CIR photos from SL·2 (June) and SL-3 (August), and S I90B normal-color photos from SL-3 were used in Ule evaluation.

All normal<olor photographs were examined using a green filter in the op tical system; the CIR photos were examined with a red filter. These filters were inserled into the light -beam palh for enhancement of the color and CIR vegetation signatures. An effeclive circular aperture covering an image area of 6640 square microns was used with the S I90A material. For Ule SI908 materiaJ , an e ffective circular apertu re cove ring an image area of 41 ,500 square microns was used. These effec tive areas provided a circular image with the diameter approximating the side dimensions of the ~rnplc cells used for visual interpretation.

The same sample ceUs used in the visuai interp reta­tion test were measured with the microdensitometer. Optical image density measurements were made of each sample cell. The calibrated light beam of the microdensitometer was aligned with the cell location by means of the sample-ceU overlay used for visual interpretation. (Compen sating allowance was made for the apparent optical density of the overlay.) Values from all sample cells were obtained.

A two-tailed t test was used to detennine presence of significant difference between optica1 density sample me:lnS for both the Series and Region level classifica tions. Sample variances were assumed equal .

40

Figure IJ-A mkrodel1silo me ler (Gen, enlt AI1i1ine ~nd Fitm Corp. Modd 650) was u$cld to obtain optic.3t ima~ density. The plKllOJruph on the moveable image­hoklin& stait is SClnr>eu by the light beam lube. The openlof aligl'ls tM Skylab sample cell in the vicw SCJ'ccn before readi n& the appatt'nt opticsl density rf8.istertd on a digit1l display. The ~w;na SCl'ffn and. stq;e also aW:led in the estimatk> n of fo liv cover from larac-scalr photo.$..

Results and Discussion

Results indicate th:lt Skylab and support aircraft pholograph1c products can be successfully used for classincation and areal mllpping of native plant com­munities to the Region level, by visuallnlerpretation. Results of Series level classification from these same photo products are dependent on date, scale, and film type: . Areal mapping at the Series level required for­mation of class complexes_

Optical denSities measured on Skyl3b pholos with a microdensitome ter provided acceptable classifica· tions at the Region level. The Se ries level or classifica­tion, however, showed film·type and seasonal dependency.

Visual interpre tation provided successful cull ural feature mapping from both Sky lab and aircraft photograph.s; however, better resolution of the air­craft photos permitted mapping of finer details.

Very large scale aircraft photos were used to esti­mate quantitative parameterS of specific plant com­munities. The results were successful, but dependent upon species diversity

The result s are discussed in more detail in the following sect ions.

Plant Community Classification bV

Ph otoi nterpretation

Region l.evel- Skylab color infrared pho tographs taken with the e~lI'th terrain came,a (S I90B) should

be acceptable for Region level classifkation in a stratified sampfulg design. To ! Irati(y Stries level plant communities, however, high·a1titude CIR ph o tog.raph~ can be used only jf some classes are combined .

Both the uteUi1e and unde,night aerial photo­graphs provided acceptable results for interpreting both the Conifer and Grassland Region classes (table J 2). The lowest accuracy level obtained was 89.8 percent for the Grassland class from the JWlt S I90A CJR photos. The remainder o f the accuracy levels for these two Region classes were above 95 per~nl

regardless of ftJ.m. date. scale, or interpreter. There was no Significant difference between the color and CIR mm.

The Deciduous catego ry . which in this study is Aspen, was date, fIlm-type, and scale dependent; only CLR provided acceptable accuracies (table 12). One acceptable level was obtained for the June photos, and the August date provided two acceptable levels, all at different scales. Interpretation accuracy on 1 :50,000 scale August pho tos was not as acceptable as on the June photos of the same scale, probably because of interpreter fatigue or the testing sequence used .

Di fferences between photointerpreters were gener· ally small and nonsignificant ; however, the interpreter most knowledgeable about the physiography and

vtgela tion o f the test Mea was more accurate , though not s.ignilicantly so.

The great majority of commission erron for both color (4 1.2 and 52.9 percent aircraft , 59.9 and 72 .3 percent Skylab, June and August respectivel y) , and color Infrared (24.1 and 15.4 percent aircraft , 56.7 and 51.5 percent Sky lab, JWle and August respec­tively) occurred when the Deciduous class was iden ti­fied as Conifer. Because of the heterogeneous mixing of these two classes of vegetation in the Deciduous test ceUs (fig. 14) interpreters had difficulty in deciding whether Deciduous dominated within a lest ceU. Very few of the Conifer class were committed to the Deciduous class, however. because in most of the Conifer test cells Conifer dominance was obvious. Those test ceUs that were correc tly identified as Deciduous were homogeneous (jig. 14). For the June and August dates, on CIR film , thert were signifi­cantly (p • 0.99) fewer commission errorS of the Deciduous class to Conifer than on colOr film. AJso on CIR rUm, for lhe August dale, there were signifi· can lly (p • 0.90) fewer Decid uow class commission errors to Conifer than for the June date. There was greater separability belween Deciduous and Conifer using CIR, and an even greater separation occu n ed at the August date because leaf-color change of the Deciduous class had begun. Deciduous leaves at the JWle date were only half to two-thlrds developed and

Tab!;: t 2- A(CUfQC)' of ~;11M21 t:/ossijicQt iOiI of piOIl! CONlI1l1mitia III Rtglol1 It.'d.

Monltou. Co/orodo. tnl lite'

Date, source, Deciduous Coni fe r Grassland and 5Cale of

Color I Color I CIR Coiol I CIR pho toi,~ph$ cr.

JUNE 1973 Percent Aircraft , RB-57:

1:50.000 88.9 99.4 100.0 1:100,000 58.8 71.1 99.7 98.9 100.0 t OO.O 1 :400,000 63.9 99.7 98.4

Skyiab 5190A : 1:2.800,000 33.4 40.0 100.0 97.3 9'1.5 89.8

AUGUST t973 AimaCt RB-5 7:

1:50,000 17.8 98.2 J 00.0 1:100.000 44.2 88. 1 98.9 96.8 100.0 100.0 1 :400.000 SO.O 87.9 98.2 98.6 98.1 98,2

5kybb 5 t90B: I : 1.000.000 25.0 98.4 95.8

Sky tab 5190,,: 1:2,800,000 ZO.li 4 I.S 99.3 97.9 100.0 95.8

I Mean percent correct for IWO ifltellHctcu .. Not aU film types were available for all scales and dotes.

41

gave a low infrared reflectance relative to the sensor alt itude. Therefore , deciduous foliage was difHcult to separate when mlxed with conifers.

Commission errors of Ihe Deciduous class 10 the Grassland class type were often made on the Sky lab pho tographs (percentages were 6.7 and 7.1 color; 3.3 and 7.0 CIR, for June and August respectively). The extremely small scale resulted in difficulty for the interpreters in sep~rali on of these two classes in eco tonal situations. This was furtl).e r complicated by the subtle ty of color differences between the Deci· duous class and some members of the Grassland class in bOlh the June and AugltSl photographs of both HIm types. Another reason for commissio n errors was the lack o f topograph ic (stereo) rel ie f near the mo un tain/grassland in terface, Some misclassification was introduced by misregislralio n of lhe photo ove r­lay.

A few Conifer-te-Grassland commission errors were made on bo th film Iypes. Most of these errors we re committed on Skylab photographs, because of class mixing and misregistration of the sample loca­tion overlay.

Series Leve l, Forest - In general, of the five Iree categories included, only Aspen, the lone Deciduous Series class, W3S consistently cla~ified with an accept­able levd of accuracy and then only on the August

CIR aircraft photos (table 13) . At the I : 100,000 and 1 :400,000 scales accuracy exceeded the 80 percent level. On the satellit e photos, no acceptable accuracy levels for the Aspen Series class were achieved.

The Do uglas-fir, Lodgepole Pine, Ponderosa Pine, and Spruce/Fir classes were occaSionally classified with accuracies of 80 percent or grea ter, and then usually with the early-se~son and smaller scale photos (table 13).

The greater accuracy at smaller scales was achieved hecause species mixtures bl ended into a more appar­ent homogeneous unit, with the dominant species signature predomi "ating. The decreased resolution at these smaller scales allowed the interpreter to classiJY the dominant characteristics with greater ease. How­ever , some classificat io n d ifficulty was caused by species mix ing and somewhat broad color signa tures.

There we re commission erro rs fo r all categories. The overlap in geographic r3:nge caused Aspen to be generally confused with Spruce/ Fir regardless of date, scale , or film type . Tlte sma[[est such error (5.9 per­celli) was ob tained from the August CIR aircraft photos. At tllis date, CIR provided the necessary separation between th e fall foliage coloration of Aspen and the dark green foliage o f Spruce/Fir. Douglas-fir and Ponderosa Pine were most often con­fused with each other for all film types, dates, and

Table I) - Accurocy oillisuo! c/asrijlcQtion of p/Qll t commullities at Series level- Conilefs l aud Groft/and- Manitou. Colorodo. test sile' .

Conifers Grassland Date, SOlilce-,

D~u&bs· Lodgqw te Pon<lerosa Moun tain and scale ~f Spruce-Ifir Shortpass Wet Mead~w

photographs r" Pine Pine Bunch~n$$

Cotor I CtR Color I CIR Color I CIR Color I CIR Color I CtR Color I CIR Color I Ct R

JUNE 1973 ---- • -- Percellt --- -- - ._-Ailcrafr. RB·57:

1:50,000 46.1 215 71.8 ".1 18.6 92.9 100.0 1: 100,000 40.1 49. 1 58.9 58.4 77.5 79.1 82.5 15.1 90.0 12.1 100.0 95. 9 88.9 94.5 1:400,000 45.4 45.2 n.1 11.8 83.4 100.0 92.9

Sky lab S190A : 1: 2,800,000 18.2 81.3 80.0 10.0 80.) 68.2 75.0 60.1 62.5 54.2 96.9 100.0 77.) 40.9 AUGUST 1913

Aircraft RB-5 1: 1:50,000 48.0 ) 3.4 '/3.2 69 .4 85.7 100.0 100.0 1:100.000 56.9 59. J 52.3 48.4 69.9 10.7 18.3 77.3 88.8 74.5 95.9 95 .9 94.5 94.5 1 :400,000 43.6 39.2 42.9 54.8 73.1 '/ 5.1 83.8 66 .8 85.5 80. 9 100.0 100.0 92.9 100.0

Sky tab 5190B : 1. 1,000,000 76.3 66.1 15.7 75.0 38.4 100.0 100.0

5kyiab S190A: 1 :2,800,000 65.9 65.8 72.9 66. 1 73.1 82.6 10.0 48.6 71.7 20.0 100.0 100.0 75.0 87.5 ---- -

I Aspen wu the only Series c J~ $$ of decid uous t rees; per~nt Jccuracy is shown In tab/I' 12. 1 Mean percent oorl"t'ct for two inlerpreters. Not all film typ~s were Jnibbt e (~r all sc::.Jes and da les.

42

Fl$lue 14-A;tttaft support pllotogu.phy (ClR, J :50,000) wa~ u5ed with acetate overlays showina sample training cells (labdtd) and test orlls (numbered) for m .. l classification of native plant commurutiea. In the August photo (abon), note the Deciduous/Conifer (AspeIl/Pon4uosa Pine) muilng In cell 68, the YI,Jiability willUn and between the Ponderosa. Pine cells, and the similarity between the Sproce/FiI (SF) and Ponderosa Pine cell (69). In the June photo (b~IQw).lhe test celli repreRnt ltIe three Grassland communities elassilifd; Wet Me~ow (66), Shortpus (67), and Mountain Buncltgrass (68). Hen there is maximum distinction between the three Series cl.assel, bill cell 68 apprOI/;:hts the ecotone belWte.1\ Mountain Bllnchgrull and Sl'Iortgrass. The ~e of the photo shown is approximately 1:45,000.

43

scales. Cenerally, both Douglas-fir and Ponderosa Pine had the least com mission errors (15.4 percent) on early-season (June) color and CJR satellite photo­graphy. At the same date, and on both color and CIR, commission errors were grea test on the underflight aircraft photos (26.7 percent) and were essentially the same (nonsignificant differences) on the late­season (August) photography.

The large nwnber of commission errOfS on under­flight aircra ft photographs resulted because the Pond erosa Pine and Douglas·fir ove rlap in geographic range but occupy different physiographic positions in the landscape. Douglas·fir generally occurs on north and east slopes while Ponderosa Pine gene raUy occurs on south and west slopes. In some topographic posi­tions, however, depending on the degree ofs.lope and aspect, the two species form a heterogeneous m ix­ture. These fac ts cause seve re interpretation problems

in moun tainous terrain: ( l ) where the two species mix, separation is difficult and (2) in heavy shadows on north-facing slopes, it is difficult to determine where Ponderosa Pine forests Stop and Douglas-fir forests begin . Therefore, commission errors occur, from one species to the other, on the larger scale photos. Difficulty also arises when the interpreter makes a decision to identify the sample as one type or the other based on crown density. His estimate may in fact be accurate, but lhe criteria for decision may differ for the vegetation base maps, or the maps may not be entirel y accurate; base maps were based upon merchantable tim ber, not cover-type dom­inance , Also, the smaller scale (satellite) photos provide a more synopti c view with less resolution, (h ereby allowing inference to be made based on dominant colo r signa ture and topograpltic aspect, for easier separat ion of community Series.

Interpre tations of the Lodgepole Pine and Spruce! Fir community Series were also often confused. In general, the late-season (August) satellite photos, rega rdless of film type, provided fewer commission e rro rs (9.9 pe rcen t) between the two Series classes. One exception was Ihe lower commission e rror (2.7 percent) from Sp ruce/Fir to Lodgepole Pine on late­season CIR undertlight photos.

The confusion between the Spruce/Fir and Lodge­pole Pille classes most frequently occurred where the geographic range of the two classes overlapped. Within such areas, especially on shadowed north slopes, color signatures and image textures were difficult to sepa rate. Also. lhe lower resolution on the satellite photos, and the synop tic view they provided, resulted in fewer commission errors between classes. Classification of Spruce/Fir, however, was better on

44

the underflight photos regardless of date (season) or film type. The Lodgepole Pine class was correctly identified more often from the satellite photos. That is, fewer commission errors were made to Lodgepole Pine or Spruce/Fi r. More commission errors went to other tree Series that overlapped or were more diffi­cult to interpret without sound orientation and vege­tation knowledge. In general, the photointerpreter with the most knowledge o f the study area classified the tree Series most accurately, using inference and deductive reasoning as well as image signatures.

GrassJond- The three grassland Series cl asses (jig. 14) were cOllectly classified, with mean accuracies of SO percent or better from the underfiight photos regardless or HIm type or season (table /3).

In bo th undernight mId satellite photos, most of the commission e rro rs occuned be tween the Moull­tain Bunchgrass and Shortguss classes and between the Wet Meadow and Mount ain Bunchgrass classes. Mountain Bunchgrass was most often classified as Shortgrass because they rorm an ecotone o r contin­uwn. They were difficult to claSSify, especially on the satellite pho tos. The underlliglll photos, especially the 1:50,000 scale with improved resolution, allowed a better estimate or the Series class boundaries and therefore had fewer commission errorS. Even though the satellite photos provided a more synoptic view, the reduced resolution did not allow Series separa­tion. For Grassland , these characteristics appear to be a disadvantage, although for the tree Series classifica­tion , they allowed greater accuracy.

The Wet Meadow Series was generally cl3ssified at an accuracy of 80 percent or greater from both underllight and ~tel1it e photos (table / J). Results for (he underllight pho tos were most consistent, how­ever, and were no t dependent on date o r film type. Few commission errors (2.7 percent) occurred for trus Series class on under Oight photos. Commission errors thal averaged 2 1 percent did occur fo r the satellite photos. Errors fo r this class were made to the Mount ain Bunchgrass class, primaril y because of natural mixing near mountain slopes. High-density Mountain Bunchgrass provided the same color signa· ture as lower·density Wet Meadows on the very small scale satellite photos.

Plant Community Mapping

Late-season (SL·3) S I90A Skylab color photos and late·season (MX-248) under!light erR photos were used to detennine vegetation boundaries and estimate their area) ex ten t (fEg. 15).

Region Level-The Skylab and underflight photos were both successfully used (accuracy greater than 80 percent) to map the areal extent of the Conifer and Grassland Region classes (table 14). The extremely small scale of the Skylab material and its lower reso­lution caused a reduction from the base map mea of about 17 percent in the Grassland class. AI the Manitou site, Grassland occurred in long narrow stringers between the tree stands. Manitou Lake was distinguishable in the Skylab photos, but it occurred within one of U~se narrow stringers and waS too small to be measured planimetricaUy. For compar­ability, it was included with the Grassland class in al1 three Region maps.

The Deciduous (Aspen) class w.as not successfully mapped from either the WlderOig.llI or the Skylab photm (table 14), because of stand sizes and class moong betwun Deciduous and Conirer. No Deci­duous class areas were mapped from the lower­resolution Skylab photos.

Small areal differences (about 2 percent) between the truee maps were attributed to the mapping proce­dure, planimetric error, and Marly 20 years differ­ence belween the o riginal source base map compila­tio n and the current mapping a tternpts.

Series Level-The Mountain Bunchgrass and Short­grass Series classes could not be distinguished well enough in the Skylab photos to be mapped separ­alety. Therefore, they were combined into a single

Figure IS-Plall! communities wele mapped at the Se,ies level by visual interpreta!iOIl at 1:50,000 fOJ a unit within the Manitou site. The base map was compiled fronl U.s. Fo~st Service timber and range vegetatiotl rype-map1- The undernight map was compiled from IIIe--K&$On, J: I 00,000 8Cillt CIR airt:rafl pholOs. The Skylab map WlIS compiled flom late-scaSOtl (SL-3) S190A color photos. Successful mapping required formation of several complexes of the individual Serie$ c\uses. The map scale as shown Is approximately t : i 10,000.

legend

• Spruce/Fir/Lodgepole Pine Complex

• Ponderosa Pine

• Dougl as-fi r

• Ponderosa Pine/ Douglas-fir Complex

• Aspen

:J Upland Grassland

Ii Wet Meadow

Base map

Skylab map

4S

Table 14-Area comparisons be/ween base map, underflight map, and Skylab map for Region and Series classification levels, Manitou, Colorado, test site

Classification Base map I Underflight map' I Skylab map' level and type

Acres Acres Percell!

Acre~ Percent

of base of base Region:

Deciduous (Aspen) 786 285 363 0 0 Conifer 27,687 28,202 101.9 28,923 104,5 Grassland (w/lake) 2,690 2,676 99,6 2,240 83,3

All classes 31,163 31,163 31,163 (12,611 hal

Series: Aspen 786 285 36.3 0 0 Ponderosa Pine 15,729 7,744 49,2 4,887 3J.J Douglas-fir 10,269 8,630 84,0 6,546 63,7 Ponderosa Pine/Douglas-fir complex 25 ,998 26,707 102.7 25,788 99.2

Lodgepole Pine 1,434 Spruce/Fir 163 (') (') Lodgepole Pine/Spruce/Fir

complex 1,597 1,495 93.6 3,135 196,3 Grassland' 2,297 2,092 87.3 2,240 93,5 Wet Meadow 277 566 2043 0 0 Lake 16 IS 112.5

All classes 31,'163 31,163 31,163 (12,611 hal

'Late season, CIR, 1 :100,000. 1 Late season, color, S190A, 1:2,800,000. J Area not mapped separately . • Includes Mountain Bunchgrass and Shortgrass.

Series class. With this Series combination, both the underflight and Skylab photos were used successfully (> 80 percent accuracy) to map the grassland areas (table 14).

The Wet Meadow class was overestimated from the underflight photos when compared with the vegeta­tion base map (table 14)_ Furthennore, because the Wet Meadow class occurred as long narrow stringers, both within the grasslands and intermingled with the tree classes, the lower resolution of the Skylab photo­graphs would not permit mapping any of the Wet Meadow class (fig. 15).

The Douglas-fir Series was the only individual Conifer class that could be mapped successfully (more than 80 percent accuracy) from the underfligllt photos (table /4). Attempts were made to map the Lodgepole Pine and the Spruce/Fir Series but no area figures could be derived. Species mixing at ecotones, and slope/aspect relationships within these groups made class separation extremely difficult.

To map all the tree Series classes with acceptable accuracy, we formed two class-complexes: Ponderosa Pine/Douglas-fir and Lodgepole Pine/Spruce/Fir. The

46

Ponderosa Pine/Douglas-fir complex was mapped within 1 to 3 percent of the base map estimates from both underflight and Skylab photos. These small dif­ferences can be explained by mapping and plani­metric error. The Lodgepole Pine/Spruce/Fir complex was mapped within 8 percent of the base map esti­mates from underflighl photos, but was over­estimated by nearly 100 percent from Skylab photos. This difference was also due to mapping error: Aspen and Wet Meadow classes which were not discernible in the Skylab photos were included in the complex, and areal measurement eHor increased with smaller mapping units.

The Aspen Series class could not be mapped from either the underflight or Skylab photos at the Mani­tou Site, as explained earlier.

Manitou Lake was successfully mapped from the under flight photos and was distinguishable in the Skylab photos; it was too small, however, to be measured planimetrically (table 14). Because Manitou Lake was not included with the grassland complex or Wet Meadow Series classes, the areal estimate ob­tained for these two classes was more representative

\ , , /

1 1 \ j . /.- . ' , . ,

, I, \

~ ... ~r \. '

~ ,:~'

r " " -

Base map

,

- •

•.... -..... c. •.• __ 1-., .. _ ... "., --.-.~.

C-,* . _ ...... ,

.---

I ' .

,~ ,-' I , -' . -

.. d. ) "--=-. I - 1-/':"" ... ,_ ---<::J -'-'4,. ~-,.. .

/ •

Photo-revised map

Pcw" 16- Cultuul features wert nupped by visual interpreulion at 1:24,000 (JOOl all wea or 156 bD' (61 Iq mi) w;tllin the Manitou Ate. The photo-reviMd m~ was compiled using the 1956 VSGS Qulldrangle and UJe 1973 lire,,!t photos. The bu11d1nas grouped near the rH(tvOi:r (upper left) in the SkyJab map were Mhntitled u w.:1e buildings In the undttflight nap, but for simpllcity they we;e conPderee part of the bllildina oomplex at Lake George vUlqe. Key to features mown on photo-revised map appUtS to aJi mlp5

shown.

47

than that of the Region classification. Water level difference between the two map compilation dales was responsible for the difference between base map and underflighl water areas.

Cultural Feature Mapping

Skylab S1908 color photographs (August 4, 1973) e nlarged two times, and underflight CIR ] : 100,000 scale photos (August 9, 1973), were used to update a cultural·feature inventory map .(ftg. 16). Ground truUI information was obtained from the most recent (1 956) U.S. Geological Survey quadrangle and was revised from the 1973 underflight photography.

The features most casily identified from both Skylab a.nd underllight photos were paved highways and gravel roads. Lack of scene contrast in open areas, however, prevented differentiation of small seg­ments of gravel roads that passed under tree canopies or across treeJess areas. The problem was most severe on Sk.y13b photos. Most gravel road systems couJd be aligned on the underflight photos; larger scaie, better resolution , and occasional openings in tree canopies allowed at least partial viewing.

Identification of infrequently used dirt roads o r jeep trails was no t possible 0 11 Skylab photos alld was very difficuh 0 11 the unde rflight photos, because or canopy rover and related shadows, low scene con­trast, and lack of resolution of the narrow tracks.

UtililY co rridolS through forested areas, and those const ructed wjtltill the past 10 years, were identi· fiable from bo th Skylab and underflight photos. The kind of utility corridor could be detennined as either pipellne easement or power line from the underflight photos, but not from Skylab photos. Vegetation regrowth usually prevented detection of corridors older than 10 years rrom Skylab photos.

Earth excavation activities, including underground mining evidenced by refuse dumps, open stone quar­ries, and sand or gravel pits, could be identified on Skylab imagery onJy as a group, because of low resolution and lack of scene contrast. The specific activity could usually be identified from the larger scale underfHghl photos. These excavations were shown on the USGS base map use<! as ground trUlh.

Few individual residential or smaller commercial buildings were detectable from Skylab photos, whe reas most were detectable from the underOight pho tos. From Skylab pho tos, structures could not be diffe rentialed as to kind , except for large r industrial facilit ies. From undernighl photos, most individual buildings could be differentiated except those in close proximity to one another Or near gravel and paved areas. In general, identification was based on color

48

and refl ectivity of the roofing material , building size, and scene COntrast . Buildings with bright-colored or highly refl ective met·a! roofs, located on grassy areas, were the easiest to identify.

Cemeteries couJd no t be differentiated on either Skylab or ullderflighl photos, even though they were gene rally loca led in treeless areas a.rtd the grassland vegetation inside the sile d iffered from that outside.

WindmWs could not be identified on either Skylab or underflighl photos; however, areas wltich appeared to be windmill sites could be detected on the under­flight pho tos, These areas were either more lush in vegetation than surrounding areas, owing to water seepage and little animal use, or they were more denuded because animal use was concentrated there. The mills themselves were too small to be seen.

In cultural· feature detection, to aid identification and differentiation of particular features, convergent and divergent inferences can be made by well-trained, knowledgeable int erpreters.

Foliar Cover EstimatiDn

Estimates of plant foliar cover. ba re soil, and litter were made from two grassland sites (Antero and Eleven Mile) in the South Park portion of the study area. Estima les made from aerial photos by three viewing methods wele compared with ground esti· males using linear regression techniques. A coe fficient of 0.75 accounts for abou t 50 percent of the varia­tion in the dependent variable M, therefore only correlation coerficients of 0 .75 or greater were con­sidered valuable . Attempts were made to estimate foliar cover by plant species; however, strong vignet­ting at Ihe photo edges and extreme emulsion reticu­lation limited estimation to three cover classes: shrubs, herbaceous vegetation and litter, and bare soil. Results were 8S rollows:

Technique and class: Monocular, miC10meler

Bue soil Herbaceousjljl ter Shrub

SII'reo. micrometer BlIe w iJ Hcrbaccoosnit iel Shrub

Monocular. mierodensi tomelcr Bare soU Herbaceous/1it ter Shrub

..- Signlflcant al O.OS level.

Correia/ion coefficients

Elepen Mile Anlero

0.8423 0.0780 0.6836"- -0.3698 0.4246 -0.7906

0 .8180· 0 .11 99 0.8099* -0.33 15 0.4951· -0.8396*

0.9199* -0.5285* 0.8867* -0.1433

-0.0954 -0.7007

As shown , the first technique used-the stereo· scope and measuring micrometer with monocular optics- provided only two correlation coefficients of 0.75 or grea ter for the three cover type classes at the Iwo sites. The second lechnique-stereoscope and micrometer with stereo photos-provided three cor· relation coefficients greater than 0.75 for the cover classes. Generally the correlation coeffici ents we re higher for all o ther cover classes within both sites using this technique.

Nevertheless, the micrometer used with the stereo­scope was a source of error and inconvenience, for two reasons. First, the micrometer was difficult to align with the imaged transect location markers, and the alignment, once accomplished, was difficult to maintain , The !imallest movement of the instrument caused misalignment , and alignment and measure­ments were difficult o r impossible to ~peat precisely. Second, !.he micrometer increments became so large under 20X magnincation that it was difficult to "see beyond" the markings onto the image even USing stereo pai rs.

The microdensilometer, used as a viewing and scanning device, proved most convenient and precise. It allowed the transect location markers 10 be aligned precisely with the viewing screen crosshairs under lOX magnification and the alignment and measu re­ments were repealable. The viewing screen also aJlowed fo r smaller incremental meaSUrenlcnts (0.0001 mm). TIle major disadvantage was lack of stereo capability. As shown in the tabulation, three correlation coefficients were significantly improved by this technique-the bare soil and litter/herbaceous classes on the Eleven Mile site and the bare soil class on the Antero site.

The test results showed Ihat the coHelation co­efficients for the three classes were generally higher for the Eleven Mile site than for the Antero site, regardless of fechnique. The photographs themselves were similar in all respects. Apparently spatial ar­rangement , stature , and amount of ~egetation at the two sites accounted for the difference . Generally, the vegetation at the Eleven Mile site was less complex, less diverse, essentially lacking in &luubs, and of lower stature than that of the Antero si te (jig. 17). The vegetation at the Eleven Mil e site allowed a more precise estimate of the ground-cover classes. That is, the majority of the vegetative association was matte­formi ng blue grama and frinsed sagebrush (Artemisia fn'gida WilId .), interspersed with definite areas of bare soil and litter. The fringed sagebrush and bare areas were readily interpretable; the remainder was left in the herbaceous/litter class. At the Anlero site, fringed

49

Eleven Mile

Antero

P"igure J 7-These two sita; were used to compare on1P'Ound and aerial photo estimates or ptant foJiu cover at two Jfassland si tes within tJte South Pit);; portion of the Btudy :uta.. The vegetation at the Eleven Mile li te wu less diY'eJ1llll. tacke4 shrubs, and was of a tower stature than that of the Antero site. The plot fume pictured is J m' (3.28 sq It)_

sagebrush was scarce and rabbitbrush (ChrysolJulm­nus spp. NUll .) and actinea (Hymenoxys spp. Casso) were the taller components of the' system. These taller plants cast shadows which obscured the low­growing herbaceous plants in the understory.

Resulls also showed that the range in bare soil estimates between transects within sites was signifi­cantly different (p = 0.99), even though means for

the sites were not (p = 0.95). Individual transect estimates ranged from 10 to 54 percent at the Eleven Mile site and 15 to 45 percent at the Antero site.

Indications from these results are that bare soil, shrubs, and total vegetative cover can be estimated from large·scale photos in two-dimensional space avoiding plant shadows. Also. more accurate and pre· cise estimales could be made by using photography that is nOI vignetted nor severely reticulated.

Plan1 Community Classification by Microdensitometer

The MDT was used to poinl-sample the opt ical image density of vegetation sample cells of estab· lished cl~s imag~d on Skylab photographs. Stan­dard I tests fo r unpaired plots and unequal sample sizes were used to determine whether there was a significant (p ;.t- 0.95) difference between optical den­sity sample means fo r the Region and Series leveJ

vegetation classes. If 3. significant difference was found between classes, then the classes can be sepa­raEed ; this is brought out in the following discussion of the results.

Region LeveJ- The Region level vegetation classes couJd be separated from each other except for the Deciduous (Aspen) ve rsus Conifer combination. This separation depended on film type. primarily because classes, though dissimilar in visual colo r signatUre , had similar optical density on ClR film . Ecotonal mixing of individual type classes also QCcurred within these two broader classes. Therefore, the mean optical den­sity seemed to be more dependent upon Ihe mixture than the pho to date. Deciduous and Conifer classes were separable, however, on colo r n1m, even though separability depended somewhat on date. The proba­bility of a correct classiOcation was grea ter on June SI90A pho tos (p = 0.99) than on August S I90A photos (p = 0.95), probably because the Deciduous

Table 15-SignifiIXlIIce of differellcet be/ween ml'an optical denSity ' poluts f~ plant community c!a~l's at the Region (Forest) and Series (Grast/and) 11' 1'1'/ 011 Sky/ab photographs, Manilol<, Colorado, test silt

Skylab SL 90A Classification level and type June Au&ust

Colo! I Region:

Deciduous 3_32

". Conifer 3.45'"

Deciduous 3.32

". Grassland 3.01 ~.

Conifer 3.45

" G rassland 3.01 '"

SefJe S: Mountain Bunchgran 2.98

" Shortgrass 2.99

Mountain BUl1chg.ass 2 98 -. Shortgrass 2.99

Mounta in Bunchgrass 2.98

" Wet Mcadow 3. 12'"

Shollglass 2.99 ... Wet M~'~dow 3.12'"

t Diffelrnce significanl at p = 0.95 . •• Difference significant at p" 0.99. ' Density (0) = -log,o transmittance (T).

CJR Color I CJR

3.86 4.08 2.06

3.94 3.88·· 2.23

3.86 4.08 2.06

3.42** 3.36·· \.79·

3.94 3.88 2.23

3.42U 3.36" 1.19 ' •

3.47 3.38 I'" 3.42 J.21 · " I.H"

3.47 3.38 1.86

3.42 3.21n 1.15'"

3.41 3.38 1.86

3.38 3.71 " 1.8 t

3.42 3,2 1 1.15

3.38 J.77" 1. 81

50

Skylab 5190B, August. colol

3.90

3.63 u

3.90

3.21-·

3.63

3.21'"

3. 15

3. 13

3. 1S

3. )3

3. tS

3.63"

3. t3

3.63 · "

FiJure lB-Skylab coLO<' photos with Slmll*'s</'Ullt overlsy! wert w.ed for both visual and miCfode..sitoll1el· rk cLsSIIifiealion of plant communities at the Rtjion aoo Series In ti. In this example (519OB. "\11\1$1, 1:900,000), note the diLrM:ulty in d iscrimiJ\alina between squares 1 and 2 (center ). On cheground , )(Iuare I wu idenLifeed as Shortgrus and square 2 u Mountain BtlRchvl5S. The overlap betw~n the two $Quam forms an ecoto ne between the two classes along a rKlge. Square 39 (lower ant") is • typical Mountain BIII.chgnss site at a higher ekvation. Two Wet Meadow situ (WM) dn be ~n left or center. Tht area i.J\\Iged is South Park.

class was in the midJeaf to full·leaf stage at the early date and the actively growing leaves produced a very strong "light green" reflectance. The new deciduous leaves were lower in optical density than the darker needles of the Conifer class. The late-season Decidu. ous, with mature leaves, cessation of leaf activity, and some leaf drop at higher elevations, had an optical density nearly as high as that of the Conifer.

Series Level-The MDT interpretation results for classification to the Series level were varied. For in· stance, all three grassland Series tested could be sepa· rated from each other on late-season S I90A color photos. The differences were highly significant (table IS). Mountajn Bunchgrass and Wet Meadow, and Shortgrass and Wet Meadow, on both the early· and late·season normal-color photos could be separated with IUghly Significant differences. Mountain Bunch· grass and Shortgrass Series classes could not be sepa· rated, probably because of their very close phYSio­graphic proximity. These grassland classes are fre· quently difficult to separate on the ground. The two

SI

classes fonn a broad ecotone; Mount ain Bunchgrass communities gradually blend into the relatively flat · terrain Shortgrass communities within South Park (JIg. 18); both classes are short in statwe within the ecotone.

The results from CIR early-season SI9DA photos showed that the three grassland Series classes could not be separated (table 15), but results from late· season photos of this mm type did show that there was a hJghly Significant difference between the Moun­tain Bunchgrass and Shortgrass Series classes. Agai.n, the ecotone was partly responsible, and there was some positional inaccuracy of plot overlays on the very small·scale imagery. Time or year and poor rtlm qUality also influenced resulls. That is, vegetat ion was i.n an early stage of growth providing little infrued reflectance, and the fUm was of very poor quality, with a strong blue color saturation predominating.

Within the tree Series classes, only Ponderosa Pine could be separated from all other tree classes at any lime; this distinction occurred on late-season SI90A

and SI90B color photos (table 16). Ponderosa Pine could be consistenl1y separa ted from the other three Conifer classes regardless of date, rdm type, or scale. Douglas·fir was separated from the other three classes ill both the early·season ClR and Jate-season colo r SI90A photos (table 16). But as can be seen, Douglas.fir was not consistently separated from the other classes by anyone HIm type and season combi· nation.

Aspen and the Conifer classes could not be sepa· rated with any consistency, and the differences were date , HIm.type , and scale dependent (table 16). For example. Aspen could be separated from Lodgepole Pine in the early-season but not in the late·seasOn Sl90A color and CIR photos. Aspen was also sepa· rable from Spruce/Fir in both the early. and hIe­season CIR photos and in the early-season color photos. Ponderosa Pine and Aspen could be separated on the late·season $190A and SJ90B color photos; however, they could not be separ3ted on early.season SI90A color and CrR photos.

11 is nOI completely understood why these incon­sistencies occurred; however. we believe thai the stage of phenologicaJ development and Series class mixIng are both major factors. For example. Lodgepole Pine and Aspen frequently grow together on the same site. The two classes couid be separa ted on early·season color and CIR Sl90A photos, but they could lIot be separated in the Sl90A lale·season photos. Leaf development in the early-season Aspen was only 1/2 to 2/3 complete. :lnd this allowed better discrimina­tion (increased opt ical density) from Lodgepole Pine, Later season Aspen with mature leaves was not sepa·

rable from Lodgepole Pine. however, because their reneClance (optical density) was similar. In another example, Lodgepole Pine could not be separated from Spruce/Fir because of class mixing and ove rlapping of the physiographic range of Ihe two classes; both classes also had similar optical densities. Mosl incon­sistencies in Conifer class separatioo arc caused when the species grow in lhe same alea and have very similar optical mm densi ties. Where the Conifer classes were separable, this could be partly attributed to differences in crown cover and the homogeneity of the stands.

Applications Areal plant community mapping on an extensive

basis appears to be highly successful at the Region level. Intensive plant community mapping at the Series level appears to be applicable if the land mana­ger recognize s the constraint s imposed by mountain­ous terrain and the necessity to form community complexes in some areas to allow for interpretation difficulties caused by a natural mixing of dominant species and by terrain slope and aspect.

Plant community mapping can be done on range­land and forested areas as well as grasslands and shrublands where trees are not dominant or are ab­sent. This type of information can be useful to up­date existing vegetation maps. most of which are over 10 years old and are outdated by changes in manage· ment schemes or land use patterns and by catas· trophic events.

Extensive cultural feature mapping and monitoring can also be accomplished using Skylab-type photo-

Table t6-Sfgnlflcance of differel/ces in mean 0pfical density between plant community classes, Forest, Series level, Manitou, Co lorado, test site

Pores t Series Class

Aspen vs. Dougla9-flr Aspen vs. Lodgepote Pine Aspen vs. Ponderosa Pine Aspen vs. Spruce/Fir

Oougia9-fir vs. Lodgepole Pine Oouglu·fir vs. Ponderosa Pine Dougta9-flr vs. Spruce/Fir

Lodgepole Pine Vi. Ponde rosa Pine Lodgepole Pine vs. Spruce/Fir

Ponderosa Pine vs. Spruce/Fir

• Oifrereoce significa nt at p • 0.95 . .. Olffue nce signIficant at p. 0.99. RS· 11 0 $ignifi~nt differe nce.

Skylab S 190A

June August Skylab $190B,

Color I CI R Cotor I ClR August. color

•• '" "' • • .. .. "' "' •

"' "' •• '" .. .. .. n, • n,

n' • • no fl' .. • .. .. .. "' • ..

"' '" .. •• .. .. .. "' "' "' "' "' .. .. .. •• • •

52

graphic products. Depending on the mapping objec­tives, cultural mapping, however, can require supple­mentary aircraft photography at larger sCliles. For instance, an aircraft photographic scale of I ; 1 00,000 proved highly successful for most of the cultural features mapped in thls study. Mapping features ~ quires either lar~r scale photos or inference by a skilled interpreter trained in ecological principles. One ex.ample of the potential appUcation of cultural feature mapping is a dirt road system being developed within the Manitou area for access to a mountain home development. The roads, about 6.1 m (20 ft) wide, were discernible on the S190B color photos. Such mapping shows 3 land manager the size and extent of developing areas-information nol now readiJy available for large tracts of land, but required for tota1land management.

In areas where vegetatlon has not been mapped or has been mapped only superficially, Skylab-type photographs can be useful as a flIst level of st ratifica­lion for both mapping and multistage sampling of

natural vegetation resources. Stratification was espe­cially successful with vegetation units that display a high scene contrast-such as wet meadows within a grassland system-provided those units do not occur in narrow stringers as ecotones and are large enough to be resolved on Skylab-type photographic products.

Skylab-Iype photograplts, coupled with aircraft lUlderf1ig.hu, can provide a base for multistage sampl· ing schemes in order to stratify and quantify specific plant community systems through the estimation of plant community parameters. Ground data must be used as a final stage. In this study, large-sca1e (1 :600) aircraft photos and ground measurements were used to sample areas delineated on high·altitude aircrafl and Skylab photographs 10 estimate foliar cover of plants and ilirubs. Success appears to depend on the dlversity of plant communities in relation to the parameters measured.

Descriptive and dichotomous keys (table 17) may also be an aid to the land manager using aircraft and Skylab-type photographic products. These keys can

Tlble 17-Descripti.·e key 0/ p/lJrl l communiry charaCferilties imaged on NrlYoSeaso.1 (June) 1:400.000-scole CIR oirclTl/t phot(}$ (imoge characlautics detumined from tfllinirlJ: SlImp/e wll used for l'isuol interpretation)

Crown or Plant Color' Aspect foliar DiStribution' Textule Heighl and

community cover density resolution'

Forest: Aspen Veey red (11) or All; 60 to 90 percent ; Uniform to Even or Only rarely

deep red (l)} draina;ts mro. lohigh clumped;often rouPi flppa rent unless in Sliingers clumped

Doudas·ru Very d~p red (14) or GelleraUy 50 to 80 percent; Gener:aUy uni- Even or Not appalent-deep red brown (41) n(){th, east moo. toh4h form when pure mottled a uniform mass

Lodgel'Ole Deep red brown (41) or All; in 70 to 90 percent; Generally uni- SmO(lIh or Not apparcnt dark gray olive (Ill) saddles med. to high form when pure slightly

mottled

Ponderosa D~'k olive brown (96) OJ All; but 10 to 60 percent; Random, Slightly Rarely apparent Pine deep It'd brown (4 1) usually low 10 mcd. not uniform moUtoo

nOI north

Spruce/Fir Deep red brown (41 ) or All; SO to 90 peroont; Generally Slightly Not apparent dark red brown (44 ) commonly mcd. to high uniform mottled

Grassland: north

Mountain Very light All slopes 30 to 70 percent; Varies whh Smooth Not apparent Bunch- blui!h gree n (162) to nearly low to med. terrain; 'UM n,( unifo rm

Shortgrass Pale (XIr ple pink (252) 01 Fb, 10 to 50 percent ; Uniform Smooth to NOI apparent blue while (189) low to mcd . if pure moliled

Wet Meadow Deep red {I 3) or Flat or 30 to 70 percent; Uniform : Smooth or No! apparcnt d31k pink (6, drainages low 10 high neal water broken

'Co\or notation blSed on National Bureau of Standards System (NBS 1955). Overlap in colors is re$ult of natural mixing of species and terrain slope and aspect. , Dist ribution, height, and resolution of the dominant plant species within the sample celt are described as they appear to the photointerprcter.

53

be compiled at various levels of classification to char­acteriz.e the appearance of plant communities. The keys can be used in stratification of plant community units to measure area and to determine boundaries for sampling with very large scale aircraft photos to quantify foliar cover by individuaJ species. Keys can be effective if the user is aware of their cons traints within mountainous terrain-especially with smalJeJ

S4

scale photographs which provide synoptic landscape views. Descriptive characteristics, especially image color signature and texture, are affected by terrain slope and aspect (shadows), degree of plant sp~ies mixing, ecotones, and general "condition" of the plant community or species. It must also be recog· nized that image characteristics represent both plant and soil components.

FOREST STRESS DETECTION

Pond. ,asa Pine Mortolity from Mountain Pine Beatl.

Frede rid< P. Weber

The 110,,,,,,10" of m .. ' In OIIr S"lIlon', .... Ddl ..,d. 1M fQr.", .. I 1Mg. muhlagency tllk fm whkh $0 .. ,,11 million cion,,, Ire lpenl .. "h ye., In .... 10.1 and JrOIIod sufft)' .. SI . ... d.,rctk!n in tho f'llionai f ornl SyUm:l IU wdI .. On Sill' and pnn.I< lanob IdUaIly mo_ tho drtectlocl of illleci or disease ~ and Iht ~\I."on of imp,,' by • eoo,u" of Iho num!)e, of killed 01 dOlTllpd ,. ..... Only when dl ffilp eoustd by forest pe$11 o"n b. d.'oc\f!d .. ,Iy !n thc d ... lopm.nt o£lh. put buUdup •• on .... pid ond off..:!I ... remedial "",ion be Uk.n. "..ally and ,. lIlble detoction mothod. alO ,equJr.d ... Iu!r. darnW 01 \hr.-I of d ...... eJcttm<b 0\'0' Lor ....... and many own.rmipo.

lbit P' of II", SkyIob "w.uclo PrQIJ'" .... clniped 10 ...... the _ (101_ and compan.IM dTe<;Ii""""", of dll' pr<)II\lClI (rom throe "pili'. O)'ltem, of the Ea!!h Resoor ••• l::xperbn. nl Pock,I'" Datl .nalyN' .... , . m.delO det.rmln. th. be""n" of hlp Il"dal , .. <>Iution (S I90I:l) eomp •• ed .... ith 1m. prom;! Ip«lrai d""'iminatlon (SI90A and SI92). Ind IU1<"",1ie d...mcltion (SI92) . "",pam! with ImfIUI.I !nt0'P.mtimo (S I'KIA and 81'K18). We estab­!hhod lhoe exptrimftllal hypoll\e$b .Iu.. Slybh could DOl b& ...... ' 0;> <ie.ec, Iii ... III fOfests, md !ben l\lemp!~ .0;> disprovr. 11. Th8 Black Hill. Nation al FenQt ao~ .~""undJng lamll lIre,e ohosen .. the .he of 1111 Invtt. tgaU011.

The mQllJ\n.l..11 pine be.~e (pmd,.,.,ronUl PO'" d_ lIt>pk.) II a threat 01) pondw.P pine cPU­pandnou laWl.) Ihroughoul o.e central Rocky M<nmllill rqlOIl. and me Black Ifill. In p""kldor. Al:rc .. n ha", Me" ...... linoe Ihe mid·!9W·, to

d"ecl lind appro"" d..". nused by the ""'"III\aUI pin. \>H1l • . "I"U firs. remol e ....wn, , ... areb 10 imp.o, •• ertal de.ec'iOll and "I'P. aiul of ",,,,,,m.ln pl~ beetle dan,""" In lhe Bbck HUII ... n csll blishod In 1952 (Hell •• and ome,. 1959). J>\ninl I n epidemic ou.'b.nk of 1M ",,,,,,main plne Melk IrI tho .IIlH1i>e." Bbck Hi1b in 1963-64, the FOIes, ~ Renoot. SmJin& R.0IUrd> wort Unit lOOt aerial toio. photo­grapbo at I IOlk of 1:7920 ""'. the Mothe,n Black Hills. The .nulti"l hich·,"""u\J<)n 00101" "..up ...... cin .... re "oed 10 l run forest ...sou.c. m.nage,.. 10 locot. Inf."lIW n 'POI' .nd to count d •• d Ir .... In 1%5 the WIll be&on 8 yea .. or .1, ... deteclion

"

,o$Carch. Thil wo.k p,ovIded involUJble ""pcrlenoe In ' he .. qui,.." • • " fu' inJeCt damogo d.,ecUon Ind e • • h .... tion of impacl 01. ,.lOureo. Delalied Ilud ... alablisbed pidelhoes fOI aerial pho'opophy and allO i&JI'lfiod complex pbysiol"lJeal >nd en.Jronn>",laI .d .. iomhipl.f(cdUla 1n'"",e"'lion of muJlisporuai d.a, (Weber 1969; W<bel .00 Polcyn 1972 ; Webel and o!he,.. 1!n3).

Beslnnini; In 1972, a lANDSAT-I (ER'J'S.I) u· porl1flon l "'" COndl>(:l.d in Ihe Black HOI. to de .. ,. mine Ihe polentlal uuf\iJnUlf of low·reoolullon .. 1.1. II" I)"lIem. 10 dotect u d monilor fen"ll1,m I1\IJ map ""!It' .11on C,," , lyPH. Tht ,nulll 1)( 1M ex· pellment (Webt. and o thers 1975) Indlc,ued lhal lAtI.'DSAT·1 Ima$CtY c ... pn"";cIe information (0' b. oad ...... pImnIna but I;aIlr>Ol prori;le lp<Ciflc u~il

<lIimal'" of CO>"er 4yPt ac.ea,tea. The 1 ... 1 of dllllill. cltlon fo .... hkh .. tJlfloClOt)' aceuracie. we .. ob­IIlned hOI G"".tionable "tUlly fo, tho land pi,nne. I nri ,c",u,~ m.an.fIC'. W •• peculated thl! tlwo beat GYantilati,," appl!ellion to. UNOOAT·I dl" ... In pr<w\d.IrI« the 11m boo.! Ind of """Iif\caI1oo t>f land """ IIId eoftr-type dlSl .. _ Ile . .. of decidllOW oqota­lion. ocrCl of toni( .. ,"" "",,,(.>!lon, etc . Sum clel«-lioor ..... a flil",. In . pite Q{ OIl' bett dfoTU. J\rith .. rompu ... ·usilted P""""'''1 of digital I.p« flO!" m. l.rp .... U"" Df LANDSAT·! dl" In phot<l!P"lplLic: (o.m with specw Opli«! .Ie ..... " ..... 0 ."ccoutuJ lechnlque for del«llni m ... .

The SkyLab 11 1Od1el ...... deI.lJned 10 evaluale p,odoaell of • d irr .. rn' .. leOlle do,a-ooIl..,l ion 1)'.

lem, ";!h po .. '-ble """"n\4" ower Ihe lANDSAT.l

"',~

Study Area The Block llilli I." lite (N. /9) is an ar •• Of

10,200 I<m' (3938 Jq mI) in .... lIern $olllh OIkolll md eutetd Wyoml .... The foc ... of the lite 10 an eIlIplicai domoe atmdinc ...... 11.15 milion Ioecluel (1 M millinn ocr",). The _ Jmporl.lD' 10« Sp«0es, proridirl& mo,. 1Iwl 9S pe.ceP' of the Iotal tom ...... •

dol U"to'lJmbet" voIllme, II pOIldero&o pi .... GeoIoct· cally, lbe 1I1lcl< Hill. N •• iona.I !'ooClI porUon of 1M lOll .lle II on exposed ';'Y'.oUin. core .urrounded by oodimenlll}' fotmllions. The contnl formotion. II."

....... 19_n.._ .lOIoNo __ ........ ___ ........ _W.-. t:

,..,..,.. ..... _..;,~ .... SIMA ............. ___ ._9,.'7)._._ .... _ ... .... '_ ... ToN ,. k , I .... I .. ...u. """'". 4 . u.._I.<>6p __

el .... !lon ~I"""" J 2QI) lind 21 SO m (4000 lJl4 7000 fI). i, hi&Jtly diueI:'«1, ""lIh I .... .,0 .. of upooed IOU and ,urface lock. SUlfOUndlna I~. cennll cote .. e sedlmen'ary [,,", .. ,ion. of l'aleowlc 1""~lone, The 'opolf'lphy hOI. Is ICnt]y rollln,g, npecWly in ,he nortlt"'mc,n 8lack 111111, ... he,e ,be IImestQflO (onru I pl'tau ,.roeraJJy aboft 1800 m (5900 ft). The eastern put of tl>e BlIo:k Hlb roqlll", lho .... fDrmlUom ItI" senerllly .. Io'ftr tI ..... Ions. n.. radW4mdrilic drain. ",U .. n of the. ponnanenl uSIAlowh!a d' ...... 1n tIllt .... It st,Oftil)' ... Iden' Oft

"

u,cUile Imogery. Irmnedtlltly OIulide 1M ponderau pi... u>ne, ... hkh •• uruuncb the Blode HUll, it< I circul ll valley formed f,onl "ddlsh Trtusi< .nd ".",,1 ... IOfllhol. IIId ,andstone. Th. " ,ed VlUcy ,~ 0' II Is ctllod locally , II hl""y ..wbl. fI<I Skyllb St..l photappphy.

Out p .... lry In,,"lipllOll .. u. Moers 653 IatI" (252 "'l ntI) .... """ndint. !be ,.,ld·mlnlnc lown of Lad, Soutll Oako .... T..o .. b.t>loeb ...... chouo .... thIn Ih. aKa-Sa"")', with lI\ aKl of 3!M9 Ita (9159 acres). .u.I E~ ..... th 1II .... of 4142

hi (IO,2JS .,res). Duri ... the S{"l million in Sep­' embe, 1973 .• ,"I,d "'~ock. WI"O" Peaks . ..... oUlbluhod o~'iId. , h. ptimlfY ".0 In the Be .. lodS4 MO~nlaino of uJle,n WYorllin&. no,,~ wut o f !he 81.,0 Il ills. n.. WI"" """ks sut..bloc •. thou'" _, "'""yod on lhe Vound. wos modlod lh,ouc/> .. rial recunnllwnoe ... '''''YI . tId ne.iy acqul,ed .erUJ ,"",or p/1olo~y. The "0' of lhil ... t>-hlock ... u~I44 .... (10.140 ..... 0$).

In ' be LA.'1OSAT·! .tudy (W~b", ... d otho:n 1975) the mountain pine bull. InfoOlation opoll 0/1

the 1 ... 0 In, ........ study ~, ...... ,. id.ntlfltd 011

Sc-prtmber 1972 .orb! phQ,OJIap/ts (elK, 1:32.(00). The ERil.wood JUb.bloc.k com.ln. maay mOO"'"'" and .....n.l4ed nlOWl,ain pi ... b~ell.lnf<Sllllon'"nd is impOfllfU II • IfI Mil ionai ... 1 ... her. lillie buH. actl.-ity Is _ieoobl. du'In& ondomlc condilions. Thll .,el II fin. affecled. h~,. du,in& an ""plndl", bilk beelle population. and Ih .. ill good bl""".I .. " .. f<>r on bnp.nd l", opldtmk .... Ib,nk. The Sml)' tIIb-block it on Ih. wu' .ide of ,he ptima 'y u udy .. ea. Tfldhlonolly i. con ilin • • filii 1p«:lIum of _~n pine beetle "",mly. H". i. Is <;Ommon '0 bavc hilfl b«11. popu .... i ..... , . nd I...., ""'"",e. of .imbe, ar. killed a"nulily. III con ' ,UI to the EttaJe­wood ., ... the S.VCy sub-bloc:k hOI been ... ""'. of ptrpt, u.a] ,,' i.-I'y fo, .he las! J3 yeo,. Ind «,,,.oins I,," cdlicll mlllJ of ,he biJ"k b«!1c populllion. !l«t!e populilions b ... remained WIIlt.lntly hlW' In the ..... and II " 1pp"''''tIy I lOulte of .... 'Ieo ... hlc:h qJm'1I<s Ihe control p,oblem ill IWO lUrToundlnl Nulo.aI Fo,esl mlnqmrml ...... n.. SIlO)' ,u!>­block <:<lnl';", the hugtsl inf.'Io.I""''' .. 1oIcb .... hoped to delect .nd mop on Skyl.b Im ... ry. The W" 'on ""'" lUlr-bl<>ek _ added 10 1M Iludy beclUllt tho SW miaion did 00' lIu,fac'o, Uy pOO'''' pplllhe Black Hills 1.,1 ,1'_ duma the ben period f'" fo,.., II", .. Ik ... ,"'n. A1lhOlJllr the mo"nloin pi ... beclle d~. nor..,..... ,.,iou. dilnlF 1II,~e lie .. lodte MOIIn!Jlns. lhe W."en r •• k, If ... con .. in. numefOUt dead ponderOll pine. bolh oin"y and id sm. n croups. Durlna SOp! .... beI 1913, 'oc"""~ """'0)'1 Iftealtd ..... ' 11 modefllHlud inf.,. .. t\onI, showl", 2S '" "',,,., !,.., 1"" POUP . ... .,.1 of War,.., .....

Clsssllicstion Sys t6m Imerp, ... ,lon or all Black 1I.i14 IHt lile 1nI>t;<l)' -

pholOpor/'tio> '12Il'plfenci ... oo-Io! composiles. Of

eompute,-compaUble !ape d ... -iI t. ... d on rile 1tIn· .. "'''ell Clas$ifocalion of BlICk HiIh OCU)"llem """"' typtl (Iable lB). This hienrthy .... demed ' .... If].

"

T_ "_''''''t Hilt> NG_ fu"'~-.Jq f" __ ~"""'"

II U"-[[IT ......

l>oI" "'~u UI~I'Io,

W,,,, .... p......Jh~ F_",.,..r

II ~ In 1.0 .... 00<1 ponol •

R ........ .. Sun ... "'" a.elu

,~. _ I. c-rpo ........ " ... "'*"""- ... ww;i.,..""",.

' ,1IC:Ioo<Ied ""'k ""!<tOp ,.,. _""'"' ~

cally f"" .... ",itb 'i,«at, Oi' .. ,elli'e Imq:ol)' Odd Is IIIIt.bl. f", lito .nlh 8bclr Hi!b ta l Jill. The Rim: pmJomin'ft' "",,",.type d ............ 'SCOled in .be .m!< ..... 1I$<d os. b .... fo. mapp!Qe; lh" sub-bloc:ks in'M IHt lile. Fo, .... IYli. u,lna compu le"" Iilted '«hnlqut ••• .."., .... 11 rn<><ilfled cIIlSirlClliorr 1)'''

......... 'e<!ul red, .. foIIow1:

Tnul-..ii, !!o, • ...........,.. .. ,.fOiI ..... T~pO! <oorif<ItWlllil 'TypO I ,<><ril_ 'Typo 1 eonll ....... nIIl __ ,. .. oJl, -....

• .. 1'&11.,. Dry poo'." TJ .. L cooill.......ul 'Typo I ..,.;te...­TJpe J .,..;,., ... "~, T' .. ~

A1100uF lito .... jo, I.e .. , III e&leso,la III t«J/~ 18 ,omllne<! 1011.1<:1, " .... Iiot!. '" IIWIlith l cauoed p';"

m.rily by I<>P"Ilaphy ~.d to be accounled fO!. A .ysl.m for dnsifying mount. ln pme bee lle

inf.st.lioM by size "liS also d.-vrsed. Before ..... Iu.t· ing tho o,,!h ,.$OU,e .. "'elli,. d.n , lhe R.mo'e Senoing Rne.,cl, Work Unit hod cl .... lr .. d inf"'la· Uon. Into phOlO m.,. bloed on Ihe nu mbe r <>f Irtel idon,med in an Inf .... lion lpo' . For mi<ro>cale phologr.o~hy ..,d .. 1.Uile im;ogery. h""' .... ', spol size in met." "I'" mOre useful me •• ure," followo;

I.f .. ,,,km ,"" (mot"'): l.n< 'h"" ,0

10 '0 25 16 '0 SO 51 ' 0 100

101 'Q]OO 1oI~ ttl..,,)OO

I 10 J • '0 10

II '0 10 II '0 50 ~ 1'0100

'00_

10 lbe .ub-blotk •• hown'",", only l~O" $POU

s, •• ter Ihan 50 III (165 fl) in t~.long",' dtmen<lon w,,~ cI • .,jfied.

Skylsb Data September wi' !Mnur.ed in Ih. original r~rch

proVO .. 1 and dua roqu .. ts .. Ih. ooly "li,faclory dm. (S1.-3 mi"ionj 10 OOllecl dOla fO! OUr . ludl ... Ex"n, iv. doud cove, """CI)f.d Ih. Black Hill. on lhe Seplember 13, 1973 p ..... . nd EREP SImSOn w.re not aclivOled "" Ihe Seplen,btr 18 p ...... We elected, howe.,..,r, to ~o "" •• d with our ' .... "h pro" .... on I re,{rioled basis wllh • • • n .. b!. d.H.

An excellent ~nd c<>mplele dill .. t was ,..,ei •• d f,om the SI.-:: ove.p ... of the Bl ack Hill. I." ,i,. on June 9, 1973. Although Ih. Im.gory wu of exeev­lio"oJ quality, June io perhaps lh. wo,,1 possiblo linl' of ye., ' Q det..,! fore.t 01, = ,"suhing from moun­lain pi"" 1>ee1l. damage. Any dead !rut In the sil. hod 103' m(41 of thei' dioooimed foliage during the winle! .

A11~ough western Soulh Dakol> wlJ! <>b""ured by cloud "0"'" during Ih. Seplember 13 p .... the ,. w Jl

on. Bood flame of SI9011 ooior pb<>tOSf1phy with only. "';'P of douo , over the Bear Lodge Mounlairu. Icd une poorly expo",d frame of color IlI1d colo< inf"'ed (SI90A) phol"lllphy. The Sl90A fra",. cove,ed tht We"'rrun",l piece of the Black Hills Ie" sil. but IVOI badly und.re~posed and moo ofll,. tesl sl,,, was """,ured by douds. In addilk>n, the S I90A f,ame "" .. red 1I,e Be", Lodge Moumlins bUI, "8,io, w" pOOrly .xpo .. d over Ihe for«l. A 'pedal p''''' .... ing effon b~ lhe pholographic I. borolories 3t lohn· ron Space Cemer improved Ihe inte,pretability of the muluspe<:m.l imag ••.

r,blo 19-Tn-< """,.Iily <4."d ~y "",""to;" pift' iH<ll. [(I ,h. yun 1971 • ..,J J91J.' ",b·N""kt 1. :1. ~nd 4.' 8",<* Hi1lI. S. lJoJau. "" .1"

IM ... "h". T",", " "mb<.. of de"" r<=> oiz.e ,b .. ~wood (1) s..." (2) W",on (".,.,u.)

1972 ) t911 1972 11~1l h . h(.)

191)

L ..... '''''" 10 - ... - '" ". 1 0 to " 2.702 2,6ll 2.H2 un m 26 10 50 1.198 1.207 '" 3.(160 ,~

11 10100 Im~ '" - "" '" 101 10300 m "' - '" '" !>Ion 'h," .lOO - I.OSO - - -Totol "" 6.J3S UO< 11.),<2 1,910

R.oU".1~11nl 1.11/1 4.00111

'No ~'" , .. Jlahl. rOI Will." I'<>b. 1912 'Ar<.> of •• b l>I",,~ I; ].?49.1 h. (9.HUi ""1<'). 2, • • 1.2.0 n, (I 0.]]3.1 ''''''i. 4, . ,18 7.2 h .. (10.),<1 ""'''').

On Jlnuuy 18, 1974 . Ih. SlA mission pol$ed CI"~r Ihe nOllhem lip of the Black Hill . 1.01 <11 .. and

obUi".d co>"'go of lh~ SJr""y .1Id Englewood ",b­blot., with Ill. muhi.ptLtr:l1 c"'~.,.. A1100ul!h the Ie" ".' wal do .. of clo um, the ground WI> ""OW cov.red. A$ ~ reluh. the fo,e" veS""'lon '0'01 und.r· .xl"""d ()tl bolh the color ,nd eolOl infrar<d SI90A flim. The lOst site was 001 "overed by the Sl90B nOr SI92,y"""".

Ground Truth Th~ lrend .nd ",re.d of the nlMnl.in pin. beelle

in Ihe nmth.rn BJ.d Hills. which included Ih. Englewood ond S .. oy lub-blocks, w" monltOled with I :32,000 =I~ color inf, .. od (CIR) .erial phOIOSriphy. The .. phol0"'ph, wW' 1.ken in III. August Ol in early Sept.mixr 1972 and 1973 by th~ Re"",.e Sen.ing R .... rclt Work Unit. The infe.to. 110m; we,e deli""aled and r..,orded by spot 01:1.0 10 lut del~Hon by ,..tdlh~ and .olrcnoft .. n",,~ (lobi, 19). Olh.". Y',,1y ,"rver' we,. condu~led by Ihe Foro .. Semc. Rody Mountalo Region.' Th .... antluil .nrveys provided .. 1!rtu1., of barl: beetle domOllO ror Iho entir. mack Hill . Notional Pot .. , and Wer~ useful In this study.

Fot.,1 Service ' .""'roe photogropn. for the B .. r Lodge Mount .in. (II. seale <>f 1:15,840) W<l. u,ed 10 e<rablish !found trulh in Ill . WI".n Peaks 1Ub-

• "'Iom .,jon 00 fIl<. u.s. Fo,," So,.,.", ... !l"d:y .\l". nlOi. !lesion. Ili'-;';"n of Timbor M>_ <nt. Dc"""r. C<>!o,O<!.,.

bloo;k. Wl ..., fonllllalC WI tbtx plIoI ..... phs we.., In ...... &1l1li 1.Iko. Del. tho 41111 01 1M SI...J pholopaphy or 1.10. Bear L.od~ Noun ....

T ... IIIOfbllily c,,"umo .. Ithl ..... l!nP< .. ood and. Soooy IIi ..... lccu for 1972 .... 1973 •• --.a In ,rJ/>/, l~ Thuo d.u CO'O"O' tho period of 'h Skylib ,,,perlmom. bill MUIcI be .... only In pan 10 ...... JfOIIOd lnoth. T.- ldomtir .... II dHd 0fI 1M A,.." 1973 CIK phOIOPlphy bad lOOt ycl ....... lonhlhlt diacolotltlon I I tho l ime of lbe J ..... 1973 S1,2 mlulon, ..... liMed ... 11=, ~ •• In tho II, .. SW .... SIA ftIi ...... _ INdequat •. )lo,U"I, fOunll fDl' .bi: 111' ...... Pub ",b-block ('6bI~ 19) we •• Q!olalned r,,,,,, I .... A"I"" 1973 cdOf I : U,i40 '_01 ptlolOl'lp/I)' .

A. upandl", epidomi<' is md~nt from I oom. po ....... 01 1M """,.Illy lotaloln lbe 1.111 • . WlIerC!U

tIw 51..." IUb-Wock had hlBh moruUry """''' tOf both 1972 NOd 19l1.1 thnrofoid iIIe~ .. 1n ,"OftalU), • _ ........ l!.I&Ic_uod IIJII.Wock f ..... 1972 10 197). Tho _llllly ....... u Itt tho E.ttk-oud .... "loci< tOf 197) ......... ".tM-1IlIf9' fadad I.m ."" ..... 0-1 ~ ........ Iogi ... ~ JuI)' ..... A"""t 197J bef.,.e the phoI<lIJ1lplw we .. lo1en.

Infestatlon • ..,1c;l .,"ly bet"'" tho So..". end E." .... ood IUb-bloo;k •• In Ihe S'''''y ,ub-block 11, ... ...... amid.,obl. "1I"ptlOn of $lnan., prior I.test .. tIoftI 1n10 _~ "f)' Will infesllllM e.inl .... In both 19"/2 ond. 1973. By .""UUI. moll of tho n ....... '.,... ~o~ukrrl "",II In tho f.r~1lOd ",Jr. Woe\: well ... lhan SO m (16S f1) '" 11M In 1971. "'1m ...... ~ rpidenoic ill thla at ....... 197) ","",,1m pIoo ..... phy ...... oIecI COtIIIdonbl .... . UOOI of ...... lpotl and !be crelliroa of IIlII1)'

Infelul __ OWf SO m (,IIIW 19). Ff(Im thl CIR pIwr .... phy ror 197'2.439 ..........

pol,U' '!!p""n" ... each of nino C ..... ·,YI'f d_ lhuwn In ,1IbI~ 19 ...... seI«!ed (.om 'ho mil,. lott ,It. for the SIIyla pholoiruerpreutJon I". Slmplt polnl dook. _. bued QtI ,'''' Ivliilablbly of lho ..... pIo I~a ,.,.. pound chcd, dill.Ib~11oro 0( lhe ..... "Iee IhroooaJtout Ihl _lIdy are., and dortlntll_ .... 01 the Mmplo poinl ..nh • ...,..,1 10 Ihr .... noll/ld· ina 1anU;:lpe. A ""'I' .... "replied In W form or .. _ u ,. o .. :tlq Ihowl ... the -.ple pol" .... mkro. d ••

To _lUll, Ihl ""'" crut«l (,om SI91 dill lipel by oompu .. ' ..... I«I nuppi .... I,", type map <I""", (Of 11>0 lANDSAT·1 Itddy ('10' __ and olhen 197'). .... the .... ~H)'pe tIaYa (tCoU II). w. ....... Tho type ...... were drawn 011 oatal. ""'rb)'t, (""" inlap,t'lliina "f 1:31,OOD DIe Cnt ............ phocopapha.

"

Procedures

PMtllilll8t,l1t11iGlI

&fo" inleq>relllion bqan. oil "'''''o<:k bound· ,ria ..... m.o,k«l on the Skybb phOIOPIipha. Thr" rnel/todl ..... ........ '0 <101«;1. CO\Ollt, Iftd "'1' buk· beotl.k.illed um on thrt Skylab """, ... u. tllM. plilolOPlphl we ••• nmined "" • Rlehatdl )11M 3 '''''1 IIb1tt wsln& I 80-" ar.t lamb 240 z-, "ereomk,_ope. SltlCO tnd worn uptlar we,. dun.poI 10 proridt (ou. StpUIII riew" n","IIIe .. UOfoJ. All phCMo prod"" .. were .wlll'td Iro ... endlrC ordt, of ocaI. (1 :500.000. 1:100.000. 1:2SJ)OO) 10 tod_ Iho pctaibility of Inlnp<.I .. blu..

NUl. the ..... Skyl'" prod ... " ....... . iewed in the _ "'I)' .. bofo •• uoopt tlr ll ....... opllar .... . ,.~ .. Ith 1If~' opIia. 11M ..... fo\lt 'riew-!n&"Mowucued.

FlDaIy. I V.rbwt _'9'ojocllorl_ .... \dad 10 _ tho lin-. .... ""';Ilco .... flllO 01 11: 1. As ..... 0( In_ ~ 1dtntifW, doc .....,me.1ioa tado WI. ch ..... 10 293:1 • ....tlldo proriokd the 11l~1 Inlt,plet.lliroa ~. Raulwu ~ lctIIa 'ane«! f.om I :231.soo to I : 16.000.

Multllp'ttrll Stlnn .. Dill Analvsil

Tho Sl::yllb SI92 muililp.oc\,aI tean ... , dill ..... OfI2Iyud by the UniTtr.l)' or K_ SpICI Tee'" noi"D' Cen' .... O!>iy r ..... or tbro thI. leen ""'* at dcC1",,-",lie etIetD' .............. fuI ..... :

o.n 10 0.14 0"" 10 0.16 0 ...... 1.01 ). 10 ... US

10.10 IOl;UO

Do .........

III ' .. _

•• laf ... ", IBIr .. ", '""'""'" iaf ... ",

lbao ..... tho ba ..... r""nd 10 be .. 011 ~ ... M rot .... dot oct!on lJId .... lIlon "''Wi ...... /Ion toll IJ chon ..... of Ula .... 'e p~ on thl Johmon 59- Ccau, Int .. lrt1ioc fOmpUl .. Procaal ... lYlt_.

The Skylab 1Qrnf\O, dAtl . ..tIen plouo.!. (011""" a circula. 110. nol a luallhlliAo. Thilp"""trie dillo" l ion O«IIti bcca ... IhelnMt. II pGlftllld II In ""lIe of S3J· from Ih ... rtleol . NonnaIIy. tho ,aotllinl COftie>l II:~ dall could bo procaoed by compul .. witboul 100 "''''''' dil'llaaIty. b." """_ tho co.1aI ...... dilpbyN "". TV InOCIiIOf •• tI,IWI1 liar. tlrro dlscortloa "'"GeI",," bindtn 1M Iocallon .... ldentif ... u.. 01 \mdmuu ar.t FGODd ,",th r .....

NrlJI phOlotnP'" ond INP'. The,efor... 1<O"tlri< dlalorh"'" we,e remootd md llle ""an 1_ dal l " "li/1ltMd '" IhlJ . naly. by I camp.n., p~D'" de¥l'lOpOd I I KlIlla. Un, .... . ,y.

UnNlfllJIlt.n.d dll ' f()J Ih ~ SI""Y Ind ".JIll .. wood lub-llhx kl W<r. plOooucd by the KANDIDATS (K.n_ Di&iuJ IrIlASl' 1),11 System) Inl.' K!!'. dipuol multHm ... f'il~rn rtoCO$'IlIlon I)'Ile,n. Thit 1)'11"", '*' nonpl.""",ulc •• :rt llilca. Ihe IlIyn dtcilion ,ule. and , ..... l ib .. Ioot.~p ltPP.o.or;:h 10 t:lassiflcalion. C.I ....... \0 ~ ,~by lho 1)'1tem ... tboot lilted eatlior ... thb ' ...... 1 undCf "CIwlilalLooo Sylltm.~

I'un ..... ' infomllilon rtp"UI\II 1M KAt.'DIDA TS P'ottuilll I)'llem Oln be obilined f,om th~ Unlve,· Sl ly of KIMM SpICe Ttchn%o' ('onle • • La ... ,rno •. k ....... ()J lhe Rodr:y Ml)llnl"", I'orr:sc InoI ~ EJr.pt.lm<nl Slllion. ForI Coli ..... CoIouoclo.

R.sults ."d Discussio"

""ot ......... ti"" F,om O1InoJf)' ~Plllin.lloo or .u Styllb plio, ...

"ophk pru4\or;lS ,_ ....... _ .... Iflo:anl ly ""'DUd tho ..,1 .... " ... n .... of (, ...... 10 bo IUbjecl" 10 Quanlillli¥e ;"1.,pc.II'Ioo . ~I""""ular ond V ... III Ir:tleq><.' ........ _rc petroll"" only 00 .hree SI90fI ookn pftoIos £tom St..l 1l1li oac from SW. St .. _

or:vpIe IdInp<OIlrI_ wwt' prrr ......... "" tho lb," SI90tI ooIor p/MlIos from St...!. T ... mnI!ndtr ofH .. ~ororv'phk p ..... l •• urn","" did not ,nul ewt­tk""" of <karl u-. ro..l II ..... o. bilk bettie rnf ..... ioM.

P .... li¥e ...... IIS " '11 " S I90 ph~O&I.ph"" prodlOtli Wfie undentoOd >Imply Mid'nlllltolion of mounl,'" pine !>HIlI mrnmloo IfPOII IIId not • ,boli'y 10

counl lndi'rid"'" ca.~ It_. It.aoll1 art 11_ .. I counl at (" fHUtion '1',>1, wrlhln IUb-bIc:d< "udy ....... Funh<:rmocl. in lho prtlimiftar)' luminlli<>n. no ("feslllion >pOlS _1d _ Idenllfltd on color III eo!o. inf~ p.odW:lI r,orn Iho m""",,"I,a1 r;;r .... ' a I)"lIem. n ... ot .... all i4tnllf1r:.II"", .... ~ mm. un nomral-col", p. or;I""'lS from ",e ~"Ih 1 .. "",, ,,,nt,, (S19OB).

... romporiJan or ,",ulll for 1M &l1'O)' Iludy "., (IDbI .. 10) rewo1.d th o benefll of lItftOloCO,,1c .1 . .. ·· "'I. boU, iD "m" of hiJlr numbe, nf iden, iroed Inf .. I.lion< InoI f ..... coo,ml$lion Inor' . All of Ih. (0".01 klenl iflr:OIIIJIU "'er. 1IIId. (.om 'llollp of o4Q Infesmlo,," IarJ<1 Ihllt SO III (1601 fllo,..a1 di_ tlOD) .... """"'n on rhe ..... nd lruth my. In III 1n1e<pUtlliorl oU"",PIi • • U al ... Infa"'iocII on I ....

..

1I0000d l ruth map WI .... 10 Iar~ thad 100 m (328 (I) .. ere Ioellocl.

Iden,lrLCllion .... ilhin .he 1?naI .... .-. lIudy .... ' ''0''00 li'''~1f ....... It. (labI.l~. All <"",oct !d ,"tln· ell lon' within Ih. Engl.w ood il u<.ly ... , wert mode f,ntn I SlouP of 26 Inr.,Ullon. In tho WOIIII" tru lh: IhlY Were 1.'1'" ,h . .. 26 m (85 0 ) ~u' lCSS lhon 50 m ( 1601 (.).

'Tho op.bnum OICWilll ..:ale. U$ir'l ei.h .. tnonott,· Ilf 01" ste,f() optic. on thl i'.oom 2 4() mia...:<>p< . • 1S

obGor. 1:75.000 if,.. 10). This JCMo ..."bintd II •• rKlory rnacDiftcalloo with .euonlb!y !harp im .. deflnilion. II Is Impor •• n. 10 noll 1""1 Idcnlir>eallw. of the <karl ponde,,,",, pinr In lb. S190B toIor rum Wli prlnllrily by Ihe bri"l ,oddhb-o, ... color , •• he. llu.n by ...,."",ltl"" of lOY IU lu",1 differll'Oli f.om the ..... Ihy pine. '/lle tpmlly of the dud roJiIp 'tmlbtinl 00 ,ho It .... 01 the lime ofu.. J ..... 9. 19H ...... 10 ....... 1'" 1ft •• lho I ..... whoa tho dcad ..... d) .... ' ..... finl ..... _ 01*1<: . wrdoub.eoIIy p_led ... rdt ........ speaaculll reoulu. I'Urllla· """'. II .. low ......... diu ... the St..2 pasa.r:1y Irr 1M moo""", ",.d. l"lr~tllioa diff"lCIIh. In the .I_p IIt"in 01 tho Blac:k Il1II. moat infHtIl,.,... on _. and _I>-f."'1\& dopes Wh. ill lIrod ... IItd

... toe nOI ouIhle. ",. (".(m!".iIon of monoc .. a. and .,treoscoplc

oitwlnl ._lied., defiQIlt 1".et, .. ,Cf p,ef .. ...,. rot ...... Tho _ 1I"'&ibM be ... llt .. f It • ...., ..icwin& ..... I IHIo, tmclmcy UI _. rom .• , ....... In

tbo InICfp,,"UIIon 01 dud " ... 1fOUPI. Moot of the tooI't"' ...... , tIIo" ' ...... Ied f ...... OIHnt I"'c:ha of bo •• ooiI or npooed fooet.. flooc On MltII""t) infoo'.

lioa """S. II"",,, of ic>11 .... ,.lltlon wll" lbe V.riocon Yirw.,

T."'" :10 ..... ,""""" <01 Ior/ .... ,w. _I"" st,wb 51_ '*'. .,. __ ....t n"<O«"<>/M .......... lot ._ _ ...... h . 9Io<tffllll. I.Dlrl .........

.. _ of I0Il" .... ,_

,- ......... \0. -~

.""u.r<ak (''''' ..... , (,,,,- ('_.,ocIl eo",,,,, ,.~

, SOO.OOO 0 0 • • I tiXI.OOO " " " " , $0.000 " " " " I IS,/lOO , , , ,

I~ •• -, $(10.000 0 0 , 1 100.000 " " " " 1 JO.ooo " " " " I U.OOO • , • •

~ In/onta.ion or.,. c Mine toili~ • Ree'.ati"" housing are. b Sp .. rlioh ~nvon d Hardwood ''''.,'

fiprol6- lo'etp,.h ' ion 01 pho'", " 0 1<01< 01 I :7~,OOO ..... d ... ""i .... . 0 I>t "'''''' ofl.", ... fOl ....,"""

...... mo, incl""i .. ,II< d<fO<tio" 01 du<l """" ...... pi .... I'l101<> SU.aI·I51.,.. _oUr <ooukI ...... I<> ._, !f ...... , ..,..;bIo i",>1"'1 .. ;u, ,.,. ..nil ..... in com .... bot II. molu,ion 01 IS+ .... en .... 0<>' I'>Od """"Ih 10 ,""" .. i""iYidua! ,_ Of ""all_ 01_ II ... ,,,do .. would b. <0IIII"'" i . .... ndi~ d .... . ........ 11rYeiIklt)', _ >cal . .. opprw;"'.,oly , : 1(10,(100.

indicated lhi, le<hnkJue w .. inf .. ioo 10 micr<>Scoplc viewing of l'OIIspIT.",i .. on a aood'<ju'llity light tabl~. POT enmplo, with ,h. Vo,i.o!:on, 1:80,000 "' .. be!.t; at this ..... 1 •• eighl CO!loct id.ntlfIC"ions we,. /Md. In Ih. Savoy lUb·block and <me In Ih. En~.· wood .... b-block.

From "'" experie""" with phot,," from ,he SL-2 mission, we conclud.d thll il wu unfortunate 'ha' during the SW mission (I) "OJeo SI'KIB .010' .""e,ase w .. 1101 obtlined rOl the lieu Lodge MOllO' laiN. and (2)opp<>rtunily was nol I.k.n 10 Image the Blook Hills , .. , oile on Ihe S.plnnbe, 18, 1973, 1'''' when the . mi, ...... wu 01 .. , of cloudo, Whh mnnocullr viewi"3 of Ihe earth le"ain com ... pholo of tho WI .. on Puh ar.,. (September 13. 1973), the following ,.oult. we,. ootained,

"

Seale: I:SOQ.OOO , , 10100,000 • , I:SO,OQO , , !'l~,OOO 0 0

00 ... enmin.otion .of th. pboto revealed th.ol 01 · though large portioN of Ihe Be" Lodge Mounlains we .. fue of cloud. , Ih.,. wor. reve,'II oumulU$ cloud. in Ih. W .... n Pe.ks ... a. Aloo,. thln bye, of drill' cl<>udt OO.e.M mos' of the ."tdy ., .... Al· though tlu. laye, did not obJCIJro the o<ca, It ",,,,,,d 10 abSQ,b .. d/orln~ .. n«,led Ilghl, JOlrocli", from suoce"fnlld""Uf"'Alion of dud and dying "ee •. nu. phenomenon W," moS! app ... nl on KYero! fram .. of

lbe mullitpoclnd earn ... r""Ully. All 1M blf .. lliion spoto. idenlir .... We" Iltset Ihln 100 m (32& f'. Ion~'" dimenAon) Ind "'". <>rlrnled on "'1W~;><>o«I "'ope<. Th,.,., odditional spoil "".1., th.n 100 m (328 fl) "".,. not loollt<! in the SI9011 pnol"ll."hy b«a= of Ihei' I'I<I«n .. "t on •• bll\ltly stetI' norlt.. In.! tl<>uhwnt.fldT\& !lopu.

Muhi.pectrll Sl:anne. Data Analy,is

TIlt ,.,ulll "f lI,i. oxp."im.nl ",' ••• d/l11'I"'inling. Poor ..... 11' tll! be Il1fibulN to foul faclOfl: (1) 'lullity of the multispect.al dl" (J un. 9, 1973) "'u unsatiofaclOfY. (2) da, ........ nl>1 coIl«tN dIU;.,. tho po"" af ope.oh1l rapon .. 0( bHtlt-~mtd If'" (50 ... tembe.), (3)...,..ra1 chlflllli. or dllt lIomlilly II$<!d III r"'HI classlfk:ad"" _ •• not usoblt, lI!d (4) Ih. r ... !>aDds of ........ dati wuld no. be _rately • .po.. t<.td ""Ih "". ""olh". Pond .. _ p' "" " ... kiIccI by !be """,num pi ... beetle w .... nol d~""ltd ODd IM •• fo.e !hey ",et. nOI elwir .... by KAI'IDIDATS p.otnOi"" The miott&lst.ltiorl of MSS bandl ..... probably lhe prim.1)' RIOOI1 why the .datlvely....n beelle inf''''li,"" coo ld nol be d.locted,

lhing Ih. June 9. 1913 m,1\,l1peclfll JOlnn., dotl .nd th. KANDIDilTS p.t>Celilna .y.ttm. m.1\)' P'''' aJulll .ari.oIIQ~ of 111. 'pe(:lfol ba"d. 011 <1 , " nltlons .... r. al1fJ't1pled to Imp.o,. I ll d ~ooth Ih. dlla. V •• ioUI rombiruulonl of Ih. lpecml bonll' and vari .. lions In In. t.bI. l00k.up th,uhold <lI1 .. e. wo •• tned III .n effort to optlmlte clasaintllion ,uuln. All .\tempts, h"","",. " . .,. fuo !ksl. The ben tWliflc .. tion ..... In lbe En&Iewood "'~.:.:k. 1'01' ""lone<, wi,h t~nin;g datl.nd p.l.td IoISS bands 1 and 6 and 8 ll1d 12 (= ~Procedu."'') In toOIdlnlllOfl. Ihe bfU _ .... mi>claui/kaliofo rale fOf ;dI clataes ..... 12

~nt. The poorest I1tl1di1.sirieltion .. It .... 20 po.unl for _I me..,.,.... l.IJin, I'" dIll. lbe poor ... c ..... rlCatioo ...... for cunif .. t~po 2 wllh 35 pe.ct'~1 mi.d.Slir.ed. The .... r. mltduliOCll!on ,1'0 fOf oil loot data ..... 15 po","n\. Ott... combln ilion. of MSS dm bands .nd lobi. Iook",p 1I1.ethold v~ .... , ,,,, suIted In hlgller m!sc] .... fl<: •• !Qn ,,, ••.

Al1emp" to imp.ove cl .... rlCl1k>n by tl)'lnli dlf. f" •• nt d ... ""ootnl,,& pIOCtdu.u. bind pli, com· bin.lIon'!. !O bi. look.up docision rules. and modlned 11<" !fW lruth .. 11 w.,. not .ory ....... dl"'. ]n one .. t of txperin1On ... o .... .;all tI..,;flc.tlon accuracy ... 1' imp."""d by tcmovin& th. dead·" ... cl1e",l>' , cOl'l'>­binllli 111 eonif .. "I"I"'Iu. Ind Imp'0"'" lhe ,.Ii ... Hli .. I fo. tho lrlllSlUo.>n 01' • .,1)'. 1'h. ' .... 111 .... .., bette., but bcco ..... of miJIeg;,u l lion, 111 ••• , .... alM<! ... IIlL1voldabl. ".de-off bet"'et" Int ,"pul"" t...d· wood ... bare .... 1 lI!d .otk '" 1"'''Pfetult b ... 1OiI

"

and rod< .. h .. dwoodt. An Iltemp' .... mad. 10 .toOlve 'M conniet bo""een hlldwood.""" blO. soil .nd .oek, by modif}illl 'M ]>lob.bOnit. of dwiflCl' tion b.sed on lJound ob ....... "lon.-Ih. p.obabilit1 .. "" .. e .. oed u .... IgIlI11II flCIOfJ to lmp.o ... <:h.<sir,ca­Hon. R.'ul11 of Ih ... dltT.,.., • • "",binuion. of tet~· niques Ind p.IlU,w11ll p'l'ImtlOn Ir' quite . anlblo. bu' the best clwifll:lllol1 lhQwed only aboul 70 pere.nl of the 439 wnplc l'Qirll ' correc tly ld t nttfled .

Applications

or o.ll lhe Skyllb dill p.o,h .. ". col", phot.",. ..phy obta1nod in "fleo1OtOp~ co ...... ... ilh Ih. S! 908 conh t ...... tamtll wu bal for the ""'at"'" Ind lIJ>IMusaI at dead polld..,OIII pin. ~loed by moun­uIn pine bettlo. Fo, pUletl ..... r ........ ,OI bit .. <:(1ft,. "If. lib oilC .. n pholO&rlphy. """'Id be obllitltd ill Ille AUKU1'l Of .Idy s.,ptem .... dUriI!J !1M peat of diff_ln opeclnJ.enOClantt. bet ........ hnllhy and dyin& lIeeI. In .t<"P t .... ln Ute thl Black Hill .. wh.,. inr .. u!iom oco;u, IndOpendtn l of oIopo o. a.poel, .. !.!II'e phot .... l\Ihy obtained In la t .... mmc. rhould be pl .nned (01 middlY .. po.,,,. when the hishe. sun In~ •• occu •. CompldlOn. of SI9/)8 color pho ll>JnlPhy (ground ~.oIutlon ap,,'o~. IS 10 2() m or 49 to 66 fl) .nd highll ,uolution phol"i'aphic 'Y".m. ('\lCh ... hOl or lh~ U.2 hish·J1igbl 11"fln). indlC' " thaI ",tolllt. phOIGllJphy w!!~ ~ .uoIuUon 0( I to 2 m (3.27 10 6.54 ft) I'o;lJnd',esolvtd distinct would be Idul fo. det..,lion Ind Impacl .pp ....... or ., ..... in fo.atl. Photo,,"phk tffiul1lon, CO\.Ild be .,the. bi&h.,noIuUo.>m coIOf 01 ~0101 lnf.lrod, al· Iho"", the fa"" .. II pld .. ~d.

A11hou&h pllOu'I'ophy f.ORt 1M multi1pect • ..t came •• (SI90A) ..... of no .... fo.llmrdel<e'ion ill !he 81ad: IIlIb, bolh lhe muhibaM cepobllity ODd "'F u .. co"" •• ...,~ .... fu] to fOref! plan ... " 1M ."""',u mllllC"t. They Pfef." ed I 28,SJ( enllfle· mtnl (1:100.000 10110) of lbe colo. mfmed band from the muluopeclral Clm •• 1 I)'Jlem, M a worltina 1001, th. elR print "'II used 10 lid III updltinA existiT\& type.map •• In non'yIl.m .ol d In'onIOf)', and In pl.nnl"!! tunbe. sales. AU the resource m"" ..... Ind fo,." pl.nno" ClI!vustd 1<1 Ih. UI ",,~ HW, ,,&<"<1 that .uch I I"inl, rec.i"," OI1U ~."h yu •• wauld b. ofr""l ... In tllf)'inl 01.11 Ih.l, wOlk. Inl.'· pmllive aid. la mho .... lhe utUHy Orllle pholOClll'h wUllkl be IUY 10 dnd"!,, .nd inclijc\e '''''h II"", U I lemi>1l!e (lVeday showinf: adminlo""lIvo """ndary 10<111",,, .. . _.phi<: coonllnl •• pId ",",lay, IDd mmrnhip bOIIn<hries. Althoush no .rr.,.. "' .. mod • 10 Cleate typc.mapo with Sky~b photo ptodllClS,

.hey w.... used NCCaSfuUy 10 check .Ite .ype u pOini Ioclliono-.-U IefOUlt<: ".....,.,. fuq""nlly "'" ~Iopp/Iy.

TIle ,.....til <I the Un;"'IIiIy of ~ compu'.'· ..... ..s In..aIysis of S 1 ~2 mllhi$pecual ........ r dou of lhe Ill:oct Hiil Jhowed only Iimi.ed utilily for , ... _", .. lII2Iy$b. W. ''''''IfIi>e rhol tbe SI\I! oylI"'" wOlin upe. imen,"1 pacbJe. pe.haps!>col thOlllhl of

IS 0 prulOlype of. f"'''te """" .. S(2JUItt. E_ 00. rhc __ , penlHllllOnce and remlI." im>F producr did Il001 meacu.e .. p I" UpccUl!lon. A1thoU&h ,he p o,,", du<rfbU"" fClUllI~. 'UIOflIble (100..1 80 pe<t<:M for "'W' d l5Se<). IIWIY ~t land ..... d auct could not be iden,lr .. d and other. 10m ...,. iI<lI:1epubl. """, In poinl-by·pOin' cba:iroeJliml. The punuJl of O!'limlOm compu ••• ...mtM do_rlCouon was exhaultiw. bul Ih •• <Wlrds we,e minim •. A dUetl comporl""" of LANDSAT (ERTS) "'SS and SI92 MSS Imqery (Of' Ih. urn. >tudy bl<>ck,wU".,' pe,fOl'med. II Is ..........t f.om e.rlie, Olodics, bow· _ •• tJu,! 'he clas:iif'lCIlioo ..... 111 W<luki h .... ....., slrniIl' , ace,,,," M !he J>OOf quaUty of bInd 13 ditl (th.una! IR) the q ..... km of whether lbe bendlu of Jr>Odml ... ,eoolution tbennallmoserr from spICe ""do AlflltlCIntly to e!the, (I) tI .. defection " r 111" in roreal' ", (2) Ih. muhi$peet.al dwlflCltion ofl,nd IlJe. Anllysiro of only ,h. tW<l iQ'Od"lualily bind. of M.'iS dall Wll nol lIIequ". ror our purpcon. Detec· lion and ~IHdflc~t1on ' OC'l[OCY ",/f.rt>d bu ... oe of (I ) ml.mlimltioo of duo belWeeft band •• (1) Inld ... qUltt JPI'.laI ~lu!IQl'I of "'SS imll"ry, Ind (3).00 few bond. of """ble MSS dOlI for I fu! l aod oomp!e.e Inllyd!.

DU';1\iI I"" 211 YUrI rinc<: the Skybb bunch, .... bl v-e hAd the oppoflunhy to Ie .. Iik EI' dna for ....... u ... ful I I'pllcltl"", Ihll lie Dol connected

"

with ,he Blid Hills Itudy objecl''''': I. EnJuJ"I'I<ntJ "e S 19011 <:d ... and elR pIIOC,*

,.phy ha ... heed uad on ,be .... COM' of the Unll'" SU ... I" obi ... buelliIe doU (,om Ample 1ocMio .. "pp<O.ed r ... the ... ,iooaI I' ....... ry IlICellti"'" I'l'o-" ... (1'11'). ltgllon.nd exam; ... t ..... of FIP lit .. hi! In many instances provided 1n(""""I1 .... , othe,· "';,.. """nillb!<, on the """d,1Ion of pra:liot IIlC!O befo", ,be p."" .... _ br"",. AlIItOUJh SI90B reooIu'lon ...... adequale for ....,. QMninatlortl, II ..... inad<qlllie for O!loe ... Apin, resoIutlon of 1 to 2 m with elR film would ha .. been ideo1.

2. s. ... -..pi~ enlllgemenu "r e%, "hOI", f.om .... ..rtll I .... n carne.. of IUo!>dlni dead Entdmm ..,.,.... ~ ~ Parry) in tbe J"",.. )lm ... "i... oor,h_.. of Leo AiIlllOO, Ne ... )Ie""", . p.~ useful f ... for"" mllllFf1. Boawc "" "the, rtoOU."" plm'"",,""y ... K ....n.bI., the Sl<ytib pb"t""'pby .... used (or pllIIfIint ...... M Iart;e 'feU of !lIooi."Il dud limbe" """"SlIllie r.om el1illing road ...

3. Black Hills National I'orm phnne .. IIrd no II)bfCC mlJl .... 1M<! 1: 100,000 mllliornen .. of elR pl,ol"" f,nm the multispc<:u" came" I)'ltem. The phol ... we ... \lied rOt upd";1\iI .OY<. type·mops. for """'y<ltDl! ",.d ;n .. nlory. and for pl.Min.lim .... ,.d. Ilyout. All ...... ave<d thot rominuod lvall.bU· Ity IIrd use of the .. ttUile plIoI'" ..-otiId produce new If\d OC>S1 .. ffocUv-e applicltions. 1'0. ~ ... !tli u,lljty. ....... space: photos ohould be lvaillhl. ODCe • yel:!. Or 11 leut once . ach 3 y ...... Furthermore, oIJnoIt aU .equrn.menlt for remol. sen';", dIU oould be met with rpl ce·oequ,.ed C1R pho'ogrlpby hl";ng I ,.sol". .ion of I '0 2 m vound.,...,lved d!1tmce. U .. , plefe,e""" w .. for I Ite.eo pair covering. mrnimum of240.000 ho(S92,&OO ac ... ).

MEASUREMENT OF FOREST TERRAIN REFLECTANCE

Dotermination 01 Solar and Atmospheric Efled5 on Satellite Imagery

Robert W. oa""

A prt<licll mtthod of 00,,"<\111& .... Jljl~ ",dunce d.1> '0 .ccoonl fO! chan,", ;n '01", Inldi.,lce an~ atmospherIC dfec" io • tonl;nWllA lIud. SLId ••

","1bod w""W b< ~"rul f", • number of .noon" 1" pem,1r ,he """'1"';'011 of ipet",1 >il""'."'" of dir· 10,,/1' "'11"" .~ diffo .. o", ,ilt" La .11_ lcrome .n .. · lU"" of speon.1 vll i.liM. <I\IW<! by lm1pOlll chanl" In '<ltUliDn and "', •• , qlUlily, >lI d '" II .... pro .. I.thniqu .. f'" U1endi", >pertrol alAn"""" from oroo ., •• lu . ,,011>0, ,n comput."" .... d cb .... n. , oti"" of .... Iht. W'''a<'Y.

0 .... lppruKII to ' he probleol. of oetalnm,; """""I. ;uti"", Cotfficltllll t" corn. ' fur wIv .nd ."' .... ph<n.c .rr..,n is 10 l1udy the .emo'", ... sponoe 10 'o.-re1O of Known ",lit,!>" ... In II .. '''''pi .. , de· '"rip''''''' (h ... ,.llIte"""LI'ffd rnlior.a (N,) i • .. SUII,td p<opoltiun.1 to '."am relkct .. « \pl ... ilh I muh",lb'm. oodl'lcirn' ,.prt"~llnA Ih. pKlducl of 10,.1 irrtd~n.t ond a,m<>:ipl"'1c lL.n<mi">n~ .

Tho< ,od .. "". dIll . Iso ,,,,,Iudt III oddiol •• ,e,m replese,,'lng 'hI po,b ,odi.,,", \Np) "f ul"""o "' ... 10m! taol~II011 . Sp"",fic"Uy. 'ho oqu.1lio., Ilkll 'h" (Mm. n,um,n! 'he ,.n«1Of I. Lam b .... ""

., 111 .. ... . _ p . ,.p , '" In ... 1tid1 H • imdi;u"o<% :mol f - Ntm "a,,"ml".uct .

Rdlurln<t "10.,, uf , ... " '" nlore a, .... h., ,,;.,y 1'.llly in brigl,I"." when pInt .. d "" Inl' rhe" WI­,npondlng "".r.rtO ,od~""" .. 1 ... 1 yieLd III",.".<!<, ""':ow'" of .. mOlph.,., . fT«," ., ,he ,In,. of rhl

"'ltIL .. o .... ,flw.' n.. ,.I1.e',",. "",.wremont '",bnrq.e h., In

ad .. n ... ove ' .",,,e .>,1 • ., "t<i' h<.><l.'" Ihl! .ht dot. .,'~.II"''''' .... d n~' bt ,cmpoully ,oi,,,'iden' wllh Ille ", .. m'e .,".,Il .. hl. 'l'Itb .. ch"", ... dt><' "0' p,o\:>< ,100 ., ",o'p!"'" bu, ,,0<11 .. ,h. ,dl":'ion prov.,ries of ,I,. .. ,,,in . flunn, _ ...... onll po"O.I> ... m'n Ie·

fl<:..Uru:e p,ub.b!y do .. nol "'y .... ., ,bnt. The,.· f,,, •. tilt .. n""",,,.". dl" mill" bt .<:eep .. blo If 'cq"ltN wuhm , we.k 0' lWU of 1he ",.lltlC 0 ... · Ill!!" . An Onportinl oo""cl"a',on " 'Quld be Iho lI.hilily of Ir~PO'''''' ,",~el 0'9«" wclt .. pl ... 1

..

ph enoloCY. lOil moi"u, ••• r><! ..... lor tu,bld"y. Of <ou, ... I, " o4vluble '0 wu, k ur.dc, Ihe ... rne..J .. ;"adi",I~" condih01t" hlt .1 .... 1.Uir ... nlO, upc,i. .... « (un .. lime of '"Y . nd .imil .. cloud . nd hlZ~ col.'h'",~,}-p~t1kubrly if 1M ,..It" ore "",,·La,n· be,t;"'.

11,1. INPC' ,epom ~.oI' l of lbe . n,lymof 0 ..... 1 of Skybb (EREI') photos >nd one I,,""" "f ERTSI (LANDSAT.!) mulHSptcl ... l OCVIne' d" •. off ... ... ... co"ciuWon' .boul ,hi. txpcrlmtnl ..... d , u,ge$!1 IIOlnc po.llbl •• ppbc~'ia," of ltou 'ypc "f .adian ... me .... ,.",.nt, All clUI .. "'" I hlcJ' "" ..... loJil "" . ' we.n ... 'oll"e ,,,liln ,,,, and ,,,clln reflect."",. wuh ", •• ",,,ble ,~".. of path ,¥d,,,,,,,* '~"'I'I'_ f,urn III< ""kub'k-nl

Til< m..."tigat!on of ""or .nd "nlOSphtrk efTeo" IIt",.1 hum •• rI",. ",.n., .. 10 ... n .. ,·tro,,"d~ • ...,1 ,odiom.lIlc nlCUulOm.nll >I l id. in .n.ly .... 11,.1. lire irll.lg'Ny (Ilo~., .nd 0111<" 1975). r .... sirnU .. '"'Jell. 10,1" dIfferences in LANDSAT .1.", .. ,..",d ,arll",,<:< •• 1"" ...... ""'ed tl<t-..-n d,ff ... "t d .... of "".'11". lAN"DSAT· 1 "d .. oco •• lu«.1so diff. red by .. m""h .. JO pt,cen' rrom I'""nd·".usured ..Iu .. T bno .~.mple •. combined wllh put ""idonee of h ... dfee .. In ""ri,1 pIKH""'phy.led '" .<>lOl.h fQI I ,""Ihod of Kco~nl!", f", ' ''UI!"". 'n ,,;,1 .. I, .. dil nco and >!mll.ph ... " ,no.,f".~.

Th. p,osperl of ""'II( ex .. '",. comp~x ."",. I'h.n. ",,,dol,,,& It.ohn"l ........ uh ,Ire!, tktallod <;om· 1"""""101 ,,,,,,,odu'" did "III .p"..1 .0 lIS. Th") ... QUld ,equl,. ,,'" "lUch ..,n .... rt d".lup,,,en' a,,~ c."tly ... ·".loor'8th-by.", ••• I'''i,h ."mputtl;'n. '0 d.""e n""unlf", atnmspl,w" .",,{foc ieOI'. AI ,... II'­

<111.,. ",<)doJinS pruloal>ly 10q"l,e. 10m' kind ~f In

iiI. ,odlOmelr i< m ... u,em'n L Wi lh .... I111.·n'.,chod Iclle.:,., ... ,n"",reln.nll ..... huped to rl<YlS< I ';L"P!U ",nnl "f nudrn, rWi) < ..... rr", .. ulllhOl could hn .. 'ly """,fann .. ,,1101 . d ....

IM'e.d of 'O\ft'-ITl<lun,.d !Mu"menlal;'n 01"'"" .. 'M LANDSAT·' "udy. , Io,,··ny'lti .,,,,)~Ii wat

IIsr,d '" ,hi< study In (olio", ,onectarou. 11ms we OVOIdN 1'010 p,obleml ,ho diffi<ul'y of """,loin; ,,! '" o",,,.nded ft<i ld so,. in an "fI<ll h;orsh en ... ,.,.,·

.... "', and tlw d,mc"hy of IfI,UIII' row..,an 'ow'"

. ' 10' '0 'ho la,. "tom,e Itudy lie. 0.... ,~, III· It' ..... "' _lowed only I ,,"aU pOlli"" or 011<

V.NDSAT p~tu, •• itmont, w"""' ••• low-llyi .... IJ.

(f,ft ~.,u l<l h .... a<kq,,"o l)' $Impled dOlO .. of II,,. •• .. ,ily ,.,.,I •• d by LANDSAT.

Study Are s Tho poaibillty of acqairltc s.yIB _ ... on

...... 'hln OM 1111. for 1M IIIK. 1Iilb, South Oa!<ou, lIId the AtlanUl, Ceo,.., .. ta pvc Id hopo of de_",ln,lhe uli< or."".,......ne W"ecIlOn' 10 dill In l it •• IwO ""U. On. pouiblolppllut,ioII was ","11111,""",n' of ,empo,ll chaqa In .. "wive tt:pto.

fill ..... ,"' ... Anothe, Ipptt .. tl"" wll oipIllulf U· ," .. ilion f~ one .. 1Ht1od '0 l...,tM, In con,,,.u .. · aided dolSit"lmiofl .. ort.

Tho Ih.,,1"1 .,.c ... for '."HI IfnOCIaMO _ .. ,ei ........... 0-. ore, thoe SIoOlO)llnd E~ ... b-llloc:klln tho BlId< lIib .clMly " ...... July ~7. 1911. A mill"'" ..... lifo no...... ...... , '''0 .cully bb:kl of I"",,,,,un.tely ,",SO beo,.,<. (10.ooo lI:, n ) neo, C.nollt"". G.Ortil ( I" the or13'n,1 At lut. Itud), ".1) "" hnu.'Y 13, 1975. Beclu" Skyl.b ImIFf)' did not P'''''' '0 be 1.1,I.bl. fa, 0." und .,ni,tht cO"· "1&". nO , ttomp" """" m~de to ~omf>lli' 11m ... "h.ric offtc" OJ! Skylab dIll for u, •• IWO sllU. TIlt 8Iad Hilll nl&/" dll., howe_" ........ INlI),"" ud Ulmpa .. d "lIh UNDSAT·I dll.l.

[lu'inltbe fiul <by. 0( .'" Skyllb pnlI .. m _ ....... Ible to obUIA _'ote for • .condo!)" 01,. in notlbom C.IIfOfnio. 110';,11 only. oilY', _lCt, , .... oj" wu .ellaod IS kin (14 • ..un) ... .c of M.eddJrc, Calif"'nl' (/It 11). It WI. a "'''OW .hlp alon, Skyllb tr~k 63 U I.r.d/na r.om • poin' B km (5 mllet) ...... hw .. t of Whbb ylown RHO",,"' 10 tllo airport 011 ,100 IOUlhalll ed.,e 01 ~do.tilll- ' .... lInQl 01 35 Ian (12 "'tiel).

Tho .... _.a end 01 tho ~t If" Itt mocII.otel)' Keep tel ....... h~ woook4 tIo9olofflliud ""'"_ and d"", '""" (I"IIow 1Ibio .... D<:Iutt.I. C.wfo"u bbr;k ~ (Qwmu Minai')' ond "",,,,011 }joe oU. IQw...,., cluytt*p4 1Jciom.). PaUlltI or __ it. (Art"IOI'WIoI I".) ond o.h. r dll",''''' ....., ... If II.., p.-"t. l h ",nl.oI pollion and us""n end of lhe m,p lro p'edominlmly ml~.d oal<, chopanal,.nd "'It~,.land, with ........ ",,,I hom .. ileL

A f ..... faUow floW.lnd fIIk!! ofwin'" CIf',,' WftI

norood ., ,he II .... 0( tho S-kylob COWl ..... The wattn 0( WIIlskc)'town ItftooIYOir 0114 1M So~ ....... o'" RIvo-r ~ppc&<ed ~Iollorly ,u.btd dw,", 'M _ .... .

"

Ins trumentstion The inIUWDon.Hu", (or pllle.l .. \he noo-.y

"" ..... 'nillUod in a 1 .... "-<: .... Aero Commllldc:. 3008 ,"e .. !) ",,,,hro.d fm .erial photo,lIphy. TM fljulpmo .. t COJUitIed p."" •• lly of I ,odj"".., •••. It .. · dio~ me .. t, hlPsPted 'o<",do,. vidicon (rid.o) ...... ". f\dc:o .. ,... Ilftil, oncl.lde bllld,,_ rn, .... II Ik 221· r-•• for moo, ............ 'Ull _ JII"O"Idod by .llIbilled ({~ ... ncy and 001, ... , Ia~, ... , ,It..! ., SOO W"'" TIIJtIliq off the 2&.¥o1t d.e . ... cult ...,., .

~fllcll __ ..... aued by an Wp"' .. " ...... I ..

, .... 10._ ""'e"I IIId I dOW1l.a,d1'Qbttirol ""'l0-me .... Tho sllico:In diode deltc'on WON rut ..... ,0 mil'" tho blndwid,lu or UNDSAT.1 muIU'pcct .. 1 SCI/UI., and Skylab S I90 .. n., ... Tho .. f\e~I ... CI wu dul>'Cd from tho lire'oft ndla_ N ..... ",. tqUllioro 2. In whldl.he Ihhudt iI. ... mMd ' 0 be low . ........ '0 mlninllu 1'" tffocn of .unosp/lt.ic ,.,11.

• •

",.no ~

...... , .....

•

I

m • , --­H

-, • • ~ ..... ,. ...

~ ..... lI-n."""., .. __ .ot ..... , ....... _ .... _ ..... lS to. !U ""'"'" .. _II_ .. ~_.......,._. ... .......

">

I'IJ-< U - f ... _, ......... , IUJ"I m..,_' ol '.n .. D ._ .. _ ,_001 (M/f '" 1IP\U oW ... .... _ , ",,;p, <11>., _ ... '"_'' ' ... ,i< ........ . _io .... _to .idlloop" ...- _ ' ;Ioa

_ k<, TV .. <Hoi ........ _ .......... . .... ............ "",., ",olio". ..... , ~ ... OIl".

Wh.! " H OJ N • • r< kn" ... n I I III. '"'10 or .. "lin. ~Inlll'l. ''lU"I;O ... I and ~ ,'an ~ "ltd '0 find Ihr bum ' .. ",nHl " n",,;

• ", The Slytob SI90 A ph<Jtol '" b~ "IlIIlp"reJ ... i,h

In. reileetone<: dll. w,, ' ",."n.d hy I d\~I1.1 ",i<;ro­

denoi, .. "., ... I'rogronu we f, "'" ,''" t ~ e~n .. ,t II .. n'ie,od.""ity . ,1" .. t<l dlrf"lt d'n, jly ,nd , ub ",· <!U<lldy 10 .rr."II .. 111111 expoIu .. >I ,h. mill plo t ...

n.. .. diomel .. emplll)'<d (,or "",,"'''''"1 of Ie,· ,.in ,adutn« is In Opl""'''' 1 ... \><, .. 1\ ...... ~h""l 700F ... hh d,"," .,wI .... 1'" ""Ipt'" It IIr",l,r .. , ,hi <Op' .. 1 (rum I "Ii""" pIoOloo.1,O<!<' ill I phul"vullli<' ~""e,,'· • .,.v<llllg.< .'00' lno! p,,,,,Jd .. 'IX &0111 .. or linu, '"'pIiIUM. The 1"11 -,,,.1< VII". of II .. (1)0" •• ",<0, ..... "'"'" i, Ib<m' I • 10-1/\ ",.ix>", J . IO" 1V ,,( llsI" eut ' i*' on , ,U",oo de,. ",o. II Ii "'~" "' !lsili" [h." ".cuury ro, Ih;' nl"'d'''tnt !:w',""1< W , ll~,,~.i

upe F~lion w .. aboul ! m;"""""'11

Th, JJ<!i .. "eler ..... ell""'" bUi ll fO'luw n()l$t aDd hi&l' f''"'I~'l<:1 .upon .. . For . "I"., ...... "., WlPU' II II" I n,k'OOlnp ...... 1. lhe, 10111 01 1M iiI< .nd fall ,In><. 4"" II .. ,,,,,,,ionl efr.., of Iny und'f~'!OP'" «>mIn ....... in ,~. ou'plI' .. _ !J I .... han I mlUi· w..,,~. Thtf.r"re . ,hi. In,,,un .. nl is '""",""" .""w~, (UI n .. ,w«" .. n,," .,,«.f, Jp'od. up 10 ~S """rs J)eF..-wnd ! ]OO ,rulh\. "Ie '''rUt " I',k. 1 p.ot>< w .. oo",tructed r""n • nud""ed .IOHlI"",,, N",,~ linking. 13$·""" (illlIe"

lonl " Ill , pb"".dlffu"," jun<I'o" Sl ll~on dlo ..... wlll.1! I, 'lluked by ~ 6-<1tm di.rntltl 'perlll" . " .. fuu lt .n l n,ld of 'it ... (FOVII,·14 ,n~tirld ... nl or. r,,11 ,ngle "r ~.66 AI Ihr d .. i"lt<l expe[;",.n .. 1 0111' 110.1, "f 300 'n. 'M ""uDd ,=:>Iulioo .. 13 m lilhl mllSI"'" " I",",-",Iplnltrd ... ilh 50-",," lill ... ",,,,,,,,. ..a IWI Iht r,U"' ,,,' .1>< kno.

r",,1 ; .. l<Ii." .. ·• " , lte",rl .lIituoJe .. us " .. , ... ,ed ",II, ." .lc'''u''lCll''~k>g. oJeoi"",d .,td bU ill In 001

I ,~""'", ~ . I[ us.: ,nl ,lI y "'>t ch .. tho p"Ff .. ,t\> n\." d .. ,":I..,,,Il<, ." Ih. "" II .. s.e J fo, g;t,,,*llnB "db"c" dot. 1'1'1(1 "O'" • Jiffuot , nluun!O"I! in ,,,* luI' It,,· r.,-~ \lr Ito •• "".11 ..... , ,· ... io. hy. flb .. 0»';':11181'1

melll. and It. d,m..1J lly uf .. I>llft, I f. \Ii ..,011 I ...... ' .IU In Ihi IofF Aldu. 'I~dy Ir ... Ono 1_., boo II ...... n! .,.;'",0<1 oroly • .",11 port_ of 0 ... lANDSAT plcl~!."Itm.nl, wh ... " , Jow.Ily,n,"i:r· croft «JUld "" ... Idtq ... l.ly IImplod dNent of I .. ~U • 0nly,eooInd by LANDSAT

Study Are. The paIIa""ly of 8aI~u", Sicyltll _Of. on.

mo,. IhuI OM <lt1O for d .. Block 1111'" S6",~ MOl •• ,nd lho Alllnl • • Ceo",&. aw. p¥e '" 1000 ... 01 demonNII1,-, lbe ... of ,,~.Ic CO ... ,lioft. 10 d.lIl" lIIe .. two ..... One potd>k 'pplle.toOll"' .. _..., ..... nl of lmrporll dLo", ......... tlYi epee-1't,J ..... tura. AIIOlher opplkllion ...... Iip.lu,e .... I.n_ f,om ant .. b-bloc~ 10 .ncu"" ill compuler. .ldod cbAllbllot' work.

11M 0;« •• 11 .ylCem rOl lenUi .. fleel.""" ", ••• ..... m .. ' wfl /\own u~, 'h $000, I n(! Enpwood sub-bI!,\Cl<. 10 It. BlICk lun. Iludy , ... on July 17. 191J. A miuion w" .10l0 flown ooe. 1"fI'0 MUd)' \IIod.. of . pp,oxom.,ely -t()}O ~.el .. tI (10.000 >c'.') nu, Carro/iiC-"'. 000" .. (in It. 0111;011 ",ltn .. lIudy ' .. 0) "" hn ... ,y 13. 197$. 1H«.u. Skyl'" 1m ... !)' dltl MI provo ' 0 bt ..... II.blo for "", unll.1I1 ... c0¥-

lI ••• "" ,u.mr" .... It mado 10 «>mpu" IImOf. phork .ff." .. 00 Skylt~ dill f ... IhCM .wo . iIOi. , ,,. 8bck lIill. f1'1ln dOli. how ....... we .. 'natyud ond oompo •• d with LANDSAT·l da .. .

0.."", tho finll dl)'i of .ho Si<yllb I'fOIA'I'l "" _,e .Il10 to ob"in ~ .. (01 • 1I!<:Oft"'ry lite in """tot.n C.ur",,,La. Wilh OIIly I day" " ... Ot.. ' tel.

" " '"" .. IOCIed 1$ kill (24 "';I.,d ..... o( Redel .... Cabroonl. If., 111. 1 ... 11. "'''ow .. ,ip aJun. Skylab IrICk 63 OX"""',, r,om • pObll 8 kill (S nuIot) -'''-'. u( 1',"h ... ylQoO'1l RHo .... 10 dw abpo<l "'" llroo _.hoatt .d" of koddinl • clUt •• "" oilS kmm ....... ).

Th. _til .nd of It. .... 1 '~I II mo ..... tel)' ""p I.If.1n .. ill'l ..--Jed ""pour rnl.Md ~dr ... and dig .. pinot .".... -'HIt .. 000rtI.). C .. ,foo'" bLock .. k (Qottrt'll. tdouuJ, _filii Cllll)''' ih'e oak (~. ~1QIqHs loorbm). h ie"",, 01 nt_ni •• (ArrJOoUpllylo. tp.) ud 0',,", dllpo • ..J 1JIOcirs..., at.., pme .... Tho ... ".,at fIOI,ioro ,nil ell ........ of It. Itrip II. ",~lIJnll)' ", .. ed oak. chape"aJ, and postu,d.lId, ""th ....... 'UflIl hOllt<$i~

" rnt foJlow 1'iorId •• "" fIeldt of Wlnte",opl w.". noted II lhe ti .... oIllte Sk¥lah ~' .... the ........ or IIo-h ..... ylown Rttervoo- and the Soc_nto Rhet .pptned ,.I,.l .. ly lurbld dllll", the IM:fl*!.

"

Instrumentstion

The II'IMlUmeJltaU"" ro. plhe .... tlII ""''''1)' dati .... instaJJod in I Iwln-copn. M'O Conll!l1ll<lol SOOIl air<nll modirxd (0 ••• ,tal pIIo'OIl,phy, The eqwpmen. conwted p,imo,iIy of. ,00II ........ " 11 .. d.""" ...... ,. hJahtpHd record". Yidicooo (.idro) c ..... n.,""""'."'" ullit. """ ....... b,nd,.. lihn .11 (ffl, llJ. ,,-, roo mOIl "*",,,",.nl. WI. pOWIdool by IlItblbud ({reqI><llC)' and .oIt ... )lno .. ,tt ... ted OIl SOO win .. N"""'I off .ho 28...:>1t doC • • ,n:nI\ .".,.

R.fkd.ancc w ...... lIURd by In ..,.... .. d.,...t.oli .. irTIdll ........ 1 .. and a d<;Jwn .... 'd-poln1ilq; ,Jdio,

....... 1lle tilicon dio<le ""'<eIO .. w ... ffi'...a. ' 0 ", .. do tho ~>lldwodlh. oII..-\I'IOSAT. t mllllltp<eIl.1 Klane' and Si:Jlab S 190 II!IUOR. n.. .. fleelantll"011 doliftd ItOlll tho aiJ1: .. fl lid,,,,,,,, N, ";",fqur""" 1 . .. who<llthe .1111..". It ............ ' 0 boo low cftOUlh 10 mmInr._ lit< .IT<e" of atmospheric poth:

• ,

I" .. " ~,

,

•

1

~. ,---"

--­•

..........

~_Jt_".. .... >i .. ,.. ."'_ ..... __ ..... ...., U '"' In WIto! ..... __ • ~ __ tTlOono .-.

r ia .. ' lJ-~"'o""'.' ,.,. " ",,,II , .... , m .... '..., •• , or 1",,1l. I!'II." .. ....,. "",10004 (lei ' ", "'po,)..;.too OIp. ' 0<01'11<,. " ';p , ... , ,""',d ... IOdIo>m ..... UOI", I.' ...... itt04i>"". m"", "';'" ;"pu, prub<.ooJ jil,,,. '~on ,b ... lV moo l." .. ,ldl«;l. '..., .... . M ,><1""",,, l.r_1 ~,_ 0' ,.,1, 1 cam ... """'.,.

Wlltll It u, N. " •• '1"",0 II II .. tll<l' uf .. 1.lIlle (I'O".,liilhl •• quIIMm. I . <id 2 C'>n h. ~ .. d ,~lind tho bfam Irlnlmim,,,,.

• NS- Np

N • 'J>

Th. SkyiK< SI90 ~ rhoto. IU b< tOlll", .. d ,,"h III< ,.111:<110« dll' ..... , .. ,,, ... d by I ~i&lI'" mll:" ... donOllo",e,e,. I" <>arlm! we,. ",,,,,en hI CU"'·,,, ,''' mi<.odenl"Y y.lu", 10 d,fi"u\Ot dellSl,y "oj 1IIbM'· q .... nlly I<> .fleell ... r,~" UP"'"'" ., II .. 111", pi ....

Rld(omtltl

, .... ,.dio",el,' ''''ployw ' UI I~U'UI"".nl ur 'fl· .. In ,adl.n« it.n 01",""11: Lobo,.'n,,,,, M<>oitl 700F ",,111 dilolol....a .... ,'" "" '1"11_ II ~",plll"lll:c """I f,oon I IlIICOII pho.odludo in I po/>uIOV\Ol, ... ~"""". ''''001'. ,,1Od. ,rot PI"",I!lt$ IiA de<.des "r UP'" amplilud. Tho IIIU·..,.10 .,1 ... uf III< , ...... 1 "nAliI~ IOnBf illboul 1 ~ 10 .... 0' .boJul J • 10" W Ilf lish'

"""I'll' "" I whcu. d.I«lul. Ii II n,urr ~", •• Ih,,, __ y ('" Ihlt '~p"lImenl be.·." ....... ' ,,"unl.1 0JI<,"11on ..... bnu' I ",,,,,, .. n,p

..

Th' IOU({,,,, ... , w"' <"",,'" built (0' low noi .. and hl~h fo«jllC,,<Y ,espo"... F ~,. "l"o ,o-w.", Inf"u" ,I>< I mk""m,p ltvlL ,I>< '"IU of 'he m. . nd (.11 lun ... ~nJ 11 0<" 1"""io,,1 eff«1 of ."y u"derd.lnll'l,,~ <""dl llo". In 'he ""'po' w.'" " Ie .. ,han I mini­..,'md. Thordo,. . ,hi> ,,"'rumenl .. ",sponl;"'. 0"""11> r", In.,..J,emen".' lirerafl .pee'" up 10 4S ",.I«i p"r =~"d ~ JOO m,,Ih)

Th< InpUl ~1'1l,u pwh< "'IS """'"""'ltd (,om a ",...h"oed alw"in"m Wexl li"klDl' 1J5.."m c:u" ... .... 11 .rtil • rl,,,.,-diffuoed )",,<111>" s.ilko~ diode. ",hdl i. " ..... «1 by • 6·..,m di.unel« 'pe"ure. nl< ~lIIlI.nl r .. ld o f .i ... · ( FOV) i. 44 mini,><;IiI", Of •

lull ."p< ,,( ~ _6~_ Ar Ih~ dUlg,,"" upe.i" .. nl.1 .lti·

I"do of lOCI m. ,he '''''''''' ' .. .,1,,' .... II 11 m. l ighc m",'i"" .. ""'''''''''''''1'<1 ... ilh 50-mm rille .. m",,"' ­"" Mil,. (,ur" 01'11 .. Ion •.

T",.I ,n'3\liln«' •• ,]"",(, Illi.udc w., m .... """ IOli h .n ~1"""fOlIi<1 P"'~¥ dnlpd and b~1I1 in <KIf

L< ....... "'~_ II .... n ... Uy I"",d ... II,. pe,(OllllO!l«' r:t"~' .. IlI;'" o( trw un" ...... r ... plhe"'l& lid ... "" d .... ll", ' horn a diffu"', m""n!od In tho '''1' Mi"

, .... ~ nl lhe .,m.fr ..... u,,;ed by I riMI """," I"" .

zuidt 10 I fiI,~, Ind cltl~~'O' rhambtL An idenlical op'ical ""'mlrly \QI romlfUCled 10 u .. In labo.1I01')' callbuU" .... N~I')' cooino r:<>ffffIion 0( , ... d ... fQf 1_ .. n 1 .. 1< df« .. WH accompb<l,ed by I '''''"ory piomr.ric .odrn'quo l o..1ils of , ... .« .... "' ... for c ....... co"celion for "'" "",dis,..., p,obe ... onUabk on requnl; "'" In'roducIIM.) Tba eo.rt'tlio" II In Ihe ()r<l<, of + 10 perce". rOf I unil"­... loof6O".

Tht radilnt< .nd ,"><lil_ dill .... '. '«<Ir<i<d "" I 8 nub M.,1t 222 ponabl. ",,"II '«<I.de •. nnl "",­Ul')'. (If • .c:.1'0"'· ... d n!<:O,det ICCQffIlDOdl1U 'W<> m·

d'p<fId.", eN"""" of InllOl d.,. "" f .. , , •••. Full "'paRle fo, I modorau ampli,ude 11m Ie$! d»n 14 mllll_dl:. ",hlclr ",un. th .. , ... Vound Iflck M-In, .. "",d h. fI1(W.d ""Iy Ii>ou, 0.6 m o. I ... ,1wI $ pt"",n, of ,he d Uom<:Ie, of ,he """ bein, .... UUfed.

Vidl" E~ " ' p ... nl

To Ioett. p,oc''''y ,he flr&ht porn cone'l"'rrdlna to I ... "OW FOV ,&d'"nOO "",1"'""",,,,, SIIpp"r1

."'"Fry IIoowinl ,he wrrlHlJldln. '.It.", is tequi""'. A vidlco!t come", with malflclk IIpe ... eo.dtnl w" chooetl for II. ClIp.obiUty of , .. !-tin. m<uure""'"I. Th. camon"'"n cln ..... in • monltn' tht pi".'" be· I", ",couled Ind repllY tilt tlpt irnmtdI".Jy Ifw com1''''' 'ns I nIp" lint ,n be ee, tlln he 1t"l1own the ,ight lin and ,.cn,ded ,h. nc« ... ,y imatel')'.

In Iddltloo 'n ill , .. I·Inn. cap"bill,y, the .,dicoo come .. cl n opp' ox!male ,h. '''''Clra! lespo"" of the .adiometer ond ;rndi'nre ""'tt, by ~ .. offlltet .. tId • "lie,," dkxlt '''Iy pickup ,ube in t ~. CII1>C ... l'be l;r1lcon t"",1 I" I n RCA model 4S12 tube It,d 0 """,. 1111 .. ",iti.tty CUTV. slm lllI 10 ,he curv. or t he 1'hot<.>­dlodu uled I" t .... tadlometet and 1"'Wanee "", te' (fit. 2Jt Th o 'ilk"" vldle'ln camel, i, SO .. mit i.e in In. absolute '"mt ,h i T "",mil ope",iOll In the pllne requ ited. len. apenh'8 M fil l and I neu" . '.,J.", ;ty tiI,er of 10 po,c."t "'lI .. nllt ln" In I;ORIbl nllion WII" the b.ondp .. , nI'er.

The .. dleoo c.meti wu '.,tki lly mounted a n I mlpplnl Clmefl mount with Ihe ""'eU&ht<d "di<.>­ntolt. geCU TN tn 'he C",,"II hou,Il\iI. Two ,mIll clo$td~lTC\!lt TV "'0,,1' 0," "" "t "",d - Ollc In th. r,oo, of the II'c"r, 10 l id 1M pIlot. in .0.lptl"l 110". 'M fllshl tint, ond I"" olh« in tlr.e to .. f", ,h. mllIum.nlope,olo ...

Th. vlden ,",,,, "'" ,e_ded "" , (",,,me'(lo! . q",'ity h.lf.lmh 111'. ,.co,d" (VTIt). AlthouSh tho imI", resniutlon willi th;, d .. ice II IIml'ed to abou,

"

300 lV ~".. per pi<.ure height (the .aRleta ItIClf II apoblo of 700), !hI!; low-rou '<WIder <Int, ,I..,., playback of " .. rut pic'u ..... Analym con be """. dU<lN II "".mal opeed. ,"mb!< IInw tpeedl, n, It I lIop-oction 211in&-

n .. VTR employed '" tbb .. oily hod ,wo .. oiIabl< ,udio duru><l.. 0 ... durur.J .... uoed ro, """oe· kupirtg d.o.lI in.rodoccd by • microphone. On ,lit OIhe, dun ... 1. • 'ioniD, stgn.! "' .. ;.u.-oduced by ut nio pul .. lI<"omor, wbieh -Iy".;:hm"""dy udted .n e<enl mI'lu, "" 'M cltort m:norle •• This pt<Mded t;me--ba .. syncluoru .. tiort of 1110 VTR w;1Io ,be cIolf' record ...

In ml1dtinl; ,1>0 Skylob S I90A bud_. "'"' we .... IirnllN by tbe 1_ ItId fund. ,,"~abl • . We .,. ,em(H«i 10 mllctt lilt _petitti relJlOO'" or ,he fou, band ..... ' Irtilllrd bloct. .. nd·w!rlt. fItm. Three .. u of r~,.t1- we," requ;'.d ..... nt each (n, Tbe lrndianct. """er , .ad;OO1<tct, lad vldkoo in1q<t. W. oeJccIN off·,hMh.lf wl<>f]>llOn glass filt ... Irr OI""k Thick· ....... 10 ""' .. the tim. ood cor.! CONIrliots. Fo, .. d! 'pttt",1 bInd a cul-oo filte, w", combined "'il~ I tt- mI •• fOf .b""ption of lor!@; ...... kn~thJ. TIl< filtct ",,, Ind the;, thld:" ....... 'n millin .. tet1-, "",re:

A. lIoy, Y·SO (2.S) ond Scltou BG-18 (1.0) 8. HOYI R-60 (2.S). Ind HOYI HA·JO (3.o), . rrd

Scholl BG·20 (2 .0) C. Ilny. R-7ll ( 2.S) Ind lloYI ItA.30 (3.0) D. HOYI IR-80 (205) 000 Hoy. 8-370 (2.S)

,

.L"";;--"---i""--"---'''''---"--'ooo",'J We ... ""'l'" 1,,,,,,,,,,.,,,.1 .......,. ll_YIIo ipO<tral __ 0( tho Jiticorr .... j. «>0 ....... ;r..- 10 <-h .. or .... pl>n<od.iod .. Eoclt • ..., w .. _.~ ..,....U"01 '" It ........ t ..... ,

The spoc,nol '''pon ... of .he yldlcon Ind pI,oto. dlodu ""ere COJtIbined ",nh nit .. 'nlnrmi!tll>Ce du. .0 II'I;"" .. cu,..,. of .y.ttm 1j>e<'1rJI rupo"",,,. n-r. • pec tr.1 "''''''''''SO' of t il . 5190" COI ,",I ltalio", em· ployinl blIc k .. nd·",hilo fll nl w.~ ,lao <OOlpulcd (COfnrulaliot, IV3U.blt on ""I~nl). Th. film >v«lral .. n,ilOl1"uy ""'" ""<Ie like" inro IC<oun. , .... _U .. ,pe;:'nli 'nlnsmi":o.nce> of .11 op'k,1 WI"P"'''' '''' ,n rhe ryJlern. n.. bandw,dlh • • ", """' .... rized in 'e" ... of the Ioca.iofts of the half.~, poilU .... he'e .he lit$pOfta ,.""hes 50 pc ........ 0( ,he pOlk ... 1"" ( ta.blt:

21/. "l1h"u(h mlm\.atehe. between .he SkyLab come-,...

."d tho ,""lome";" In.trumenll "" 'pp',cnr, thoy ""y not .au .. lorg' o,ro" in ... ,d.b,nd n,e .. urtmcnli nf ,e ,nl" , . f1tc1anc •. 'file ,nosl impo't>n, diff'"",,,e occu', bet""un fil.er .. , II .nd ""len S. ",'h'rt ,Ilt ,.,Homet" ..... vidkoo pi,~ ur oXOtVl .r>trJY f,mn ,"" .ItOOllnf,ored roflection b. nd of ltv< oqetl!iOO. beainni",., .hour 700 n""'"""ef'llnm) Thi< conti,· tlon ...-wld rHUl, in • hither rellteranoo ... ~ (,om _ .uc .. r, me''''IU",,''1 .hlu tha. ,0 which .he

.. "llu. c"""" m;pon'is.

Skylab Data 1'0' the "oltl .. nt CalilO,nia .ite "ie "'4ueO(ed ,11<1

,tU,,,d phOl<>crap/lic ,u,. f'<lom milSK>l1 SIA. p ... 93 (JonullY n. 1974). IIWfld tr~<~ 6. Th. 11011 III· duckd dup!i .... fd"" fOf 5190". "",""in .. 73 throu.h 7$. Ind SI908. maplu .. 94. Tht wt"~m 4(1

""rcen' of.he airaaft flill" p",h <w<, IIIh"'h _ ott· lO,ned ,.n ... I1""" d ...... ., <Ioudf,H on tl .. SI90" pho,M. The S 1908 <MO.,. w .. ,u"t<d on .,.., b,e '0 include Iny of (he , ... , .i,,,.

"Ii .. mera ,""on. of the SI90" ?~k ..... ,. opt,.bl< .~"r' It"ion 5. ~ erl(!~lI f"nt< (n"",bo. 214) "'II bLtnk fm Ihi. rod Im,d. Tn. " •• n band (1l1Unn 6) Ind 1 .... no", .. kolor b.nd (.tati"" 4)

... re III. U Cl<P<>"'d. but the II" .. ;n'",rN ... ".Hi", bond. (Itllion. I, 2, arod 3) ",ore IIlgtoUy underox· -, .

Ihgh .' ...... lu.ion Imag .. f,,,,,, IIulo". 6 Jnd 4 .up· pOll.d n .. lnlngful ml<rodelUtl",,,,,lor IIClnnit1i inc,.· nt<nll ., IInaU aI 8 mi<,omlf¢ U}. Th, bb"k .. l1<1· wh ' .. Inf'aled imoge. of " •• ion. 1 and 1 ... e .. vel)' ".nul., and did no ..... ,it .... rm' ... (n[!feme"" smal· '" tll>n n m;",on .. Fextu,"" .... U-, .han 100 m (J27 ft) oc ..... _re I.rely detet.'" by ey~ on lb. In'",red inutl".

Procsdurss for Dsta Anslysss

The .xpe'~n t p,oouc.d ''''0 kInd. of d,u.o be reducod ' nd .n.lyud. Th. n", ..,,,,lo,.d oror,,"ft ,.dlo n". ""d Iff.d",,"" d .... recorded 011 I "rip oh"t; lh •• du. h.ad 10 be ,.duc<d '0 H ...... II"" ..n"'I1I"" -ah ... for ."""lk ." .. on lho "ound. Co .... n'ion.1 trutntu.\ and toPlpulO,.td.d .tthnique. ... ,. ult<l. "The ... ""d tir>d of ..... "'U ,Sqw. ima(llt.y. ",~ .. h ,.""lIed lllo «>n"H"nlott of piloto­.,."p/lic d",slty '0 >w.,enT ndianoo ... 1"" .. A new C'OmpUltt I~chni<tu~ ",11 rk .. lopcd .'illl;! .... n';. lomelric <.Lo,. I"ovickd by .n. Pholosnph!<: itch. 010101:/' OI.llIon of NASA IJSC, Otto p,he 'ed •• rlit, durinllhe . tudy of t!", LAto.'DSA T·I ,,'~lIi, .... ~. ala<; .,,,Iyud, 11 tlncribt<! be low und .. "R""l,, 000 Discu"",n."

Bei"", .".IYlI"l the ,adiance d.t>. __ rov_od 11 .. "ideo lip .. of ,he "Ith" .nd plOtted .he /lilht I"'th. on ... iI,bl< 1lIIp< .nd •• rial pflo'o ~rI(IIt. To Ilttd .he urse-.... I. ""'00 p.lll, on I :l.900.OClI).ICale "p"c pholos. It "'11 0("" IIC""""'y .0 ,,,,,.r., lhe p.,k, 10 I tlledlu", $CoOl •• "'eh .. 1'15 .000 ,u"",,,,, pho'Of;rlphy. Th'll. It "" .. possib le to d.r.,mioe

Tobit 11_ R,Io_ <>I.".. .... ,,_.. 'Yl.\t "" ~ .. _ • __ 1'"""

-.. ;",,,,,-.,,- -, fII'" "'" ,.;,_ tit1m- ___ .., "" 1IJ.""#fr-_ -C""'O'UI",o/

S"""" IlmlITC S"'I<»> lim. .. D S",,,,,, 'Ifi"'"' B 5 .. ,",. 6/fUru ~

JOO 10 80D 80010900 600 'o"lOO

"'"~

Hoif·.., ..... ,.,1"" .• """","If' lll'oi l~ 6_9 [OlH 6jO '0 180 IIOS '" 181 "O,0~56 11(15 10 %, 60. to"'" ~101U 60510 no Jl1"'94 091'06U .9S,o ~U

"

""" 14_ 11_ .. , ' _ , .......... , .... ""'" _ ......... ,~, ...... __ ..... cn;_ .. " ...... b7 SIIlO .. coma, ... 1IM 6.

which pall. of Ih. radiance d.,. "r.am ' .p" lent tf\itlVlly bo""".,...,u • .,.u Ih., II. ,uolvabk on Ilw AI.lli,.. im..,.,y.

~ "rip chan dot. """" .... pled .. ilh. dipliMI at inte,vllll eq .... lins app,aximatdy 20 m on thot IVO"nd. T I>< !lti<>/; of .. dian,", to "ra,hanoe H •• cal· CU1l1<d by ClIm?ut~r 10 111''''.'''. I Ilatl m .. m ar re/kclanoe nJ"",. C.H~""",, coefficient< ral ,he nod!oo'.'e>, h,od;,,,,,,, metcl. ond cho" ,,,_dc, ...... derlv<d in oor lol>omol)' no .. 'he lime of the . i ~'"ft

nif.hll by ulinS' 11th! """,ce ... "h <alib,. """ " . 00·

!,~'SO _:1 . '-.,' -

1 .... , .. "aSO --. j _." ........ , .. ,~",,, .,"" .... -... -.

I "". f ~" oc..

... ~ 1000'''opo .~.

• -" r -~ I ' .. • ..... •

e-.o. __ ,t, ., .,.1/, ..... ~"

",,", .. , ~ff ....... ""1 ..... " ........... F\t» .. lS · l_ .... .....,. ____ ~od , ..... ,"_I'*do<. .. _ ...

"

obk '0 • N~!lonil Bu,.au Qf SI",,,l.1ld! SOurC •. w. .. kcl.d port;"". of ,he d.o.u "'Ol", ,1\0, UI_ct~ ~ 10 IJ 1oornos<<IO""" .. no of int ..... on the rhOlOJ. . nd oornpu,td ,<10", •• 11.:.,. of Id k><:lltIa:.

SIIyl.b PhOlog .. phic!b1l

Tilt lI:u.llot. ,.d ... """ dOl. HI. de.~ f,om d,';. , ", m;",<><I<"<IIO","'.' (MDT) .can. m. d. On dupll· cat. phol'" of ,he , .. I .ite. u..u.g' P!lOt o~1<"lc O.\a SY"tfnl Mod.1 1010 uni t n,e NASA .. nsilQ!ne ' t\<: d.ta I. &Ivt" ... te,ml of n>3OIO""I. diffu!oe '!.nlity. C ..... <oUbtallQn Kole! belW.'" mln<><l<ndly and dif· f ..... d.ndty darll cllftJ;,tcrobly fot va,iom",.1n film

p'o,,""ios. MDT n" .... 'ical ape nu,., Ind 8* .... L in·

"'fAa! opllcal d ...... (W'iu 1913; Schmin ..... Allmon 1970). Hrnce il wu _ .. ry 10 calibrate microderul,y opint.1 \h~ difr .... de .... ,y Kale by wink ca!ibro\N llonJ>ltt of «1M' filmor proorldcd by JSC. Worki'" in dilfUfC lIen!lty uniu ..... ch_ nOi '0 ..... the Inletmodial . ' I .... dcoling wi Lh Ih . olialn~ fiIro .. nsitomell)', but ' 0 romp. ,.. m.u ur<d dupll . cate d. ".iJie. whh IIIQJe p,oduced by ,/0 •• xpo.ure ..I .... oWl lt<! 10 Ih. oripnaL folm. The 1)'11 .... lespor .... aorvc "'as oblIDItd by i>OIynoonia! """'eUna Io<hniq .... dotctlbed by Darli (1973).

1M IN/yJiI: 1M da •• flow ,. ... .....so.! ...-nDy In 110_ p..nd pam. (~ VJ. f l'Oll) .100 """ .... demn_te. ICM or !he ICU the, 0 his ... """ ... ....... I~ In 11 .. lin. pallo '0 ,,,.tilish II .. "' .... 01 dtMUy •• Iua .. hkh m.... be W ib""N. n Ul dIU Itpe .. e Cben , .. dy fo, the ron.1 ptoeeUing " .p. In lho 100«III<I pol" I scpo",. IIpo' OOCI.lining ,h_ JSC "","'oble, ",,"m _ lnaJyzed by • 1"""lm Clll.1ed R.-'STAB. A lea.,-tqUI' .. fi, 01 ,he """.o;w;!"";.y M ( .. dl&llal CUlIIU) '0 6iff ..... <kmI.y D p«><tuood I ...",nd-ool<. eq ... uon .-I. b I stand .. d error ","",1Iyl

lela !bon O.QI <liIT_ do ..... y 1IIll ... f", tIt.,h..., CI.lfI)'

iii,.. u .... (A boor f.,.." "". tried. boo, II p<oduad "llIldard "00" oIlbou. 0.04.) Con" .. ,.,,,, to diITuJr dtnslty """lei 'hen be mode .. i1h on oqu,i"" of ,ho 'om

")

fit .. hlch C,. C •• Inc! CO Ife mod_tina coefrldrotL In • third path. mocI.llIIt of Ibo ph<>,,,,,,,pItk

J)'Sten1 """"n" ", .... :oily n.all.oIl In In ~I!I"" for dupllclt. ,,"mlly thl! "'H IhI,d ",.It< 10 ktsJO of ,.1.11 ... updlur<. S<.>rne 0( ,he n'n~' fOneo' of den,ny «>111.1 be fit '0 • _"".l·ord. , .ur .... f .ocn 'ho ... quilion •• look-up ,abies "'. ,. ien ... ,.d '0 do ,ho .e ,,,.1 conv."ion.

Th . loot.up "bit. ,Ite .:oefficiem. of me d.nsi1.\' "", .. " ion equation. f,...., RN STA8 •• J>d .ho .i ••• "'0 111'" ~ •• put m.o • P .... 1rn rumcd XCAl. .. h,ch 110._ l1Id 1"11111 Oul • ",1.,10<. "';",,"," valut fOf .. tIt rnlclOdmli,y .-.du • . AnOlho. p...,.....,. DES. CAl. carnbinn th.t f ...... ion. of RNST AB ltId X("Al. " .lIizi", ,ho lool-up Ilble .nll l>o.h MDT Inpu' l~pU 1ft """ <:OmptllOf nln . DENCAL .. ri, •• 0" oxl"*'" • opt ••• w. 1I ... prln'OUI. buI 4 .... "'" ho., .he ~xlbjlllY of .. l.ct;nl cl,ff •• OfI' ~ombl nl!i"". of "'~ ... n;nl fUn' frum [he MDT \'pes" do RNSTA8 ond )(CAL.

0,,,:. ,ho ,ebu¥c U~I ptlOllau, was obl.ined. lho do .. .,.,.,..spondina 10 plnocul" inu&e block. _ .. idtnlif"'" and .... u WillItS """""IN (or -ach. Tlw ob1oIu,. uposu •• E a, lbe 1"1", ploD. _ coon. pOlK ~ _ add" ....... cfat. p1llo<tdtd by NASA A

f<lII~ ... lo" '" "'I"1 .. l."t ladianor S In ... ~ ~n;!I (h""'" 1968) was milk by Ih. cquaUOf1

)oj ... F' E .,T 'SI

in whIdt f .. camo'a lao. f .... ....m.. . ... Ut,."lIed n,.,.. ....... ond T" 'OII1I1I ..... ~IUN fI/IOMJO'1 IeIK. m~ •. _ wiDcIow.

Results find D iscussion fOf lbe WhiUtqo'(M'n 11: ••• _ •• ,. """,'N by

styLob p!K>l<11. .ho dm,ibutioru of "'~ta.",. (111" • ""oil "" tho _ dlY) _,. rocorded ."'" Alelll,. .-di. n"" ... 1 ......... ctlIl1pOlN (/16. 26). T ho v.en band .ewJII 1n<I.I<:;t" lhe pOU.iblJlly of '''0 diltl"" p<lpI>t..iOOJ of .dlKtlno;:&-<l/W orOOU>Cl 0.04 .nd "n/uhe, •• """"" 0.1 • . f'otsilly lit,., linu, COIld .. 1I0Il h r ... eed "" lhe da ••• 11\11 oxd ........ of 'M 'wo biP pOinll y;.Idl • ~,ffffftl(t .. pa.h tldiltlol 0(

only 1 peta"nl f ..... linn, ......... Tho ... po fill tlor no. ,.,noion I, .... ""Id Ito I I po'.con ...... ilow .Ior ",fleetlnee ""'0 ~k1 k u .... fo, ~l1c:o"oo ""prnd. "" .boo di".!bub"" 0( ""'lIt1C11 .nll ,dire· .""" nI .... in .he " " .'IlI u d ,HI ...... Multln",dli doll m .... , P''''.' p,obloonl.

Clmo,. " .. "'" S (rt4 band) malf...,tlltl ... d ,.,.,. 'M.'"ily _ .he Whi ..... y ..... " ., •• ",d lhe plto'o In .hll bond .... l1(li ,.k.n. BUI It ... n po_bl< '0 compu •• , he .. IoU ... ox,....." ... Iu", (rom ,cd. r.U,.d ... n .... """".' ......... of.ho fIO,mokolor I\Im (SO.JS6) f ....... ClJntea 1fallon" SI",. cl>< .ed· .. n. tl ... Io)'t"r In roIo. 111m • nOI ndl .. ively "M,,", ,0 .he rod ...... I.nlth$, .... did no, lIy 10 conv", '~ I"" IU,. 10 Iboolull .. dlln~'. N . ... ch.I .... r.~.c"n~. . "~ "pOw,,, .. ~ .. do .. ly ~" , .. I ... d In .h ... d .. ,Ion (~. 16). II lh.y " 0' . 111 Ihe olh., wa,.k",lh II. &ion •.

f urthe, ovlde"", "r the hnu, 'ololicoII' ,ip be, ,W",," ",fla:unor and " •• U". ,adI."c' c ..... hom • • ... tlidtl '" tbe Bloclr; Hilh on Au .... 27. 11171 (fit. 16) 11« ..... of 'lIOnlltlclln cohbntJon cIt'a. ~ II", •• """"" ho,~ In n ... " .. tlud f""" The ,,(.1111. .. dl .. "" >llu,", we'e ('Qm LANDSAT·I ifrAtc 1028·' 7121 of .h~ WlI • .,.n Ilh" on Au",11 20 . 1972. No l.ANOSAT dill from Au,ul! 10 Sopttlnlto, 1971 .. ere us.able bf •• ulI of cktud 1X1>" . A,·.,. Ildton~ •• 11 .... ~,. compu.d f""" lhe .y~.m­

"""<"Cioll compul .. liP" _ co,,""ed fntm dlJill1 CO<Inll '0 "...-11"1 unlll u~ .on_moon dill f,om I"",Opt;". NASA publ.,11iort1 (NASA 1912, n.o... 197J).

AD oorrol .. "", olllfflolonn eu_<.Ied 0.904. "". Film,; .""( lho ,2d"nCl eq ... '''''' I it ... Itd (t.IW 11/. Tho p .. h IId"n .... Iu", I,. IU", 'l hon 'lor ..... of .ho .. ''1I<>''od by i("",,"nd o,he," (19711 fur fou. dlffer.n, LA NUSAT In ...... W. do nO, und.,. .. ond ... hy ..... " ".110111 recoodtd """, p.olh ,..tI­...... chon _. ""Jon I . ...... Ioro 2 ,e",_"11 Ionte • • IW!o-ntlh •• "'" 11M • _'001"1' blDClwidlh. 0...,......18.....,.,. 01"1"0' lI.k loop flC'or " OU.,.. Iy I.S 10 U) IlItd ito tJw SASA .... WI_Irr 1"0-

SKYLAS LANDSAT - I ,-,~

/ • , -

/.-Sf.lTlOH (; , ~ • G,"n bond ,

/" , .. ~ • 1015$ 1 g N..,. ·,nl.o .. d ~ • ',-" •

~?' • J .. ~ /~~ST.lnOH ' ~

• !,~

'. ~ .. . .. .~

Rollocl~' (p 1 ,", 55 (; • , .• • , • -• • ~ .-' I • ••• . .-. • SlATIO N.

[101 $$ ~ , Red 'UP"""' • '. " .. " ... • " .. " , .•

Rel lt<' onCllp l Relo'", R'!I'c'~

~ip ... 16 _ R"._~ip. b,,,, ... 'ool~.ooI..n"' .. _., .. ,p""'~ ... fo< SI·yUb."" MilTS r", , otloooo b.""~ POT Slry\>b .... " . """0 J , _.or ... F..,' O ........ Uobl •• , .. "I .. ~'_' . . .. "OOCW1(O ,.,. no. ,," ,...,., ... of ,h. _..:0'" r~", 01 ....... '10,100 .. ,. ... <OIe.a.,,,,. P<>r LANOS.lT .. ., .... " " .. ..... _ dot • • r<fl,,1ON:<t 10 _. b .. d wu "","'1>1)< _ .. 06> ... It> Itt h ..... ., ...

1"_ 21-h'~ ...s .. 1t« .. Iutt •• <1 ","no'''''' _ffki",,, __ fro'" , ,,.;, .. qf 'h< ,<k,"" 1.,,_ .. ,';Ii .. "",1o_ • .wI ,_"' "p",,,..u

Sp.a", platfOtnl.

"" OOF . .,.,." It""", 0. bo"".

,od Ioandrilu. (non)

SIt~btb SI90A ~-'1710)9-0

" 1_ 1111011 4 2_101 10 aI!

LANDSAT· I . .lIIt1op<ctroi OCO_r

."'9< '" ,on ,-41)00 10 700 6_",J to m 1--10110 ~'1

""' .... , ..-'

" " .., ."

" "

Cor ...... """ --, '.K ." .K ~

"

cedu,. 10 cOn", ro, I wide, bind Kodak \\'1I1!~n 89 flit .. uRd in lho wnsi'OInete, .. ,hor Ih~n Iht ilClull ruahl rule". We liso fwnd oil"" ulICxpllind dil­c,ep.onclt. In compl"'" l.ound-rn<llU,ed ",dilnce dli' wllh ~aJUOI rUlm SJ90A mm d ..... i'y. lmOuntlns 10 U m""h U 40 p«~n'.· The malltl de .. rtfl f\"rlic,II""y.

The ..... a photo. indi"",.d oemiltln.p',ul dnu. 00 clffOS""lut doud. in lhe .ic,"il~ eM lhe teU IIle. Abo. l ""Y 11&11' ItId •• rilb1e cimHo,n, .. ,1 ..... nol~ &boo l J hou • • ft •• lhe Skyl'b o ... rll;p,t ... "11." lhe Ilte.,ft Oi$hl "'11. p"lfo,med. Therefo ... uodo· 1«led ice CI}'I,"h or ~he. ItlOW" mWl ' hi'. <on­IIlbuled 10 ,he 1"""",ly In nu,·inr .. ,ed p.lh ,idionce .

'So ...... , .. ""' ... ,"" ..., 5_ .......... _ '~1'. Sq,rot,.... .... __ "..""-." _. 001. I ($1'JOfo). ItUC.(I!H3 /U.SA L • • la ...... S- (hi..,. p_ J-lIp 10 ....

O ... "lIe the _'U .... rimlI ""il the IppU~ 01' _tOnI til_dow. in .ho fo, •• ' >! the ''''' or ,t. oj.e .. r. flillu ".u'ed ,ho, ,'" It(\eel>"", dOli woold he .. lid ro, 1 , 1 .. , .y. lIowfOtf. 4nly , ho tldLonOl and (r,odbnee dill IICqulred u!ly In the n,J,h1 ('" 'he irf'en ba lld co~Jd N cumJIltrf\! • • Iid (or b<,m I'"nlnlltl~nc~ Cllkul,tlon, ulins 'q""i<>o 3. 1 'he '''lllt WI' 0.&6 10.0S.

Applications This •• i1ec: • ...,. ..... u' ....... 1 ,ed\Q1que Ihows

mIlCh promh< .. I _111$ {lr oIIl.ln!na vobd nl .... of lho path udlm<:e «11111'011001' 0( I&ltll". imJc.'Y .1Id p<ob.bly 01' hlV~lltlllldl In-c .. h 1mlce'Y AI· lI,gqh 'hil method may ...,. ho tllo Itut oxpo",I .. ...... It n,'" bo I .~I .... bk Idju'W;1 '0 «In,;>",.,,,,".1 mtlho.>d. <II 10 melhod. u!ll ... I'OOUMI-billtd lIIS'no· me" .... Io .. '.qll" .... .,. .. u ...... "" «Ilnddenl 10 IIw ... ~Dlle .,...,~ I'Ll .. II. u ........ J ' OCClllpole 'lw

",cllf, RnOC'_ mt'~ .... "" Tho ...... ,.;IbNod .... hod dcwloped bJ RotoR .Itd Pucoc6; (Ilot;m IIId Ollwt. 1971). l1ut won .... '- oIoouIoI ., 10 .. btl!, det.nninc ,Ito: te!.o'ioIuhlJl of tho: slope 0( lhe ,,,,,enIun line \0 Iotal irTHilntl 1,.;1 b ..... 111INIfI;" ,_.

St •• ,1II .lIi1blel IIT«,,", lhe .lrcnfl n,UIU,.. me'" ledtnlqll( ntt'II to bo Invnl""..,. MOl' of .ho", ronce,,, ' ho ",011_ othaw do.~ly ,ho nl$ll' _ ... . , ...... mull maw' 11ooIe" llor I ... of 1&1<lbll ..... , . ¥t. TI_ .armbln UKt.tllo 0011' lflI;IIo ....... oww· 1tII...po ,. of d.oy. '""111/1 ,"~tode. and SfIOClraI .. ~.

Equalion 1 . ... lnch 1$ COftOtIftId ... "h 'P«nol tr. fee ... .. ,lclly .alid Oily .. indlrtduol .." ........ '111;

"

01 for 1>1''''''' blnd .. idlh>. The flc' .... r . II. P. and N" .,' .lllpoocmIly de"""",,nl. SUI"" \be broadband IMUU letTtt"ts II. ICI ... Uy "', ... als"..., "",od<n&th. tIM , ........ .. tould be cbeek,d "pimt ",.,. ... mnrnl. nlld. with IpoClroradtom<tm (!tull'rotn 197").

l'amblt ,ppitolllo", of lin .. , "'ntO'lpItc,ic 'ro llJ-fOl nll ,wn !n OompU I~' d.sslnG.1t1on I .... be"" dll· ruloleil In ,h. ht.r>tUfC (ROIl" and all, ... 1973: Pel· ter 1974. Kan 1971). The ob,.Oliv. iI 10 oxtmol .pe, trll .... "", .. (u, ..... I dalsoned ICeM'" do,. ct

'" I lIID'I .. IltN:IossU ......... ben lhe IIWII diITfJen"", .. , d ... III JQI ... _ .,,,,,,,,,,holi<: eff«'L Rowri"", tolIlllllOn I WI iIJ ..

'" ;n ... hkh ,he Rlbocript I,d ... I .. . ho WI"lonzlb bind. K_I", 1100 COH«1 _Ilkwnn .,Ind bi-onc

CIIOId CO""""' lho- ndDncc ct> Hi '0 .. flee .. """ ,", Pi beforl o~ ...... Abo. now .... of .. diadoI do .. Hi (.-it!! cooffocwnl$ a; .~d bJ .....Jd be ~ "'led 10 1M ..,. of p<n1<tU1 .... by .100 ~Ii"'"

In • lU4><r'\'t$<:d 01l"1'ico,;on ..:Iot_ ' .... wnn& trllll"'l "'". onr could , r .. ufomo Ihr dircrimin ... t (U"'11oot "",d In llor IIIIlyw of 1M old !hll _ , .. I_ ..,.10 of ./w _lb .... I. n. ""'11104 "'wid feqIIft

.100 .... amoun. of..-.poUl''''''. r ... u .... pIe ...... • G_ lin .. , cllnilkr ... plltl.ht ,rand_i"", III tho ""'III .. elor """ Iho (01' .... _ mlnilt '0_ lulII:tiun. Ihililt ~Ik"'" to ,/w,.,... do" ....

LITERATURE CITED

Aid",," , II. C. 19'11. s,.... .... "" "" ioN _ .04 ''''''''Y. "'''''0-

......... ~ .... '7(4) ,1.19.001, . ... AIII~ It, CH N, X , NO';';~. "'" '11'.1. G .... ,_.

ItH , POI" ' . ....... r : lMd _ <''-'''' .... b ... di" .. b .. ". Il10 .... 1,.. /. ~Y>l " ,,1on ol I!n s·! 1Io .. 'Of fo t<I!"'" ... ..,t.I "',_~,. llSI)" For .. , 5< ... It ... 1'>1'" P'SW·] U. ~7 V .• ill"" ,.",11< So ", """" rOle" , nd Ita .... v.x,. St • . , !I<,~<Ior . c.~ r .

C .. fkl~ . R. It. l~l. S ... ~I ... '''' . .. by '" 100'.""'«1" "'" .. .,hod..

R ... R'I' 4, 11 p. IlS1M Por ... 5«> .• S"."h ...... F .... t _Oil 11>",. ~>P. 51" .• 1'1>_ •• "ri.,

O" ... M,W. \ 91l . otpul " . ,.'.'10"1 Q/ <01<>< "" , .. ... I~rn .... t.ld

1<1 p."". lo<o;oilioD """ .... In k ' oTIQ'< S,n"'lIi 01 l:'",h It""""" 1, 4 Jl ~jJ. 51"0« In"" t'"!" 01 Te."" .... , T""'h<>IrI. , To" ".

o..i><Mlu •. It . f . 19n . ,,_ , ..,.","'" nf "", '"et. Id . n... ..... toll"''' '

.. ', ............ ~. _I .... "" «>lI<Op .. 01 ",,""1000 d.lOiIb,io • . £«>1 . .11_ ,. n . lO l ·1lll,

D .... ' . W. Coo L. ~ . ,.".10 .... . 0iI A. !. 11<._. 1911. 1 .... . .. of _U ..... _ ... '''''''''''."1 I ••

..... ."" 0,, ... 1 ... ...,. ,. p,,,,,, S. ... ' h In •. S,,"p. on It, mo •• $0 ..... of t." ..... V .... of J.IIcJt.. . An" "'<bo<. ! · 'lOS·j I I 1.

PrbodI. Il . 5. 1~9 . ........ ..,."..nII """" (of, .. .., p1"'op""lu'_

• Prfi<>._ .. d .. "" •• _ "" 11''''. _a 100< ... 1<>1;'" /. "ot. ........ 1 Col", rho'O&I . in 1'Iu. !d. .... ,UI Colo>! 1'110>1 .. ,. '11'0"'110"". 1.1010 . <>I ~l<ofI<I&, c. ..... IlIt. p, I ' Go I.'.

0n0n0I. fl , $ .. orod l . E. 1', ....... , 191! . )'"U ...... .. oI' iI»od •• not ooq_' UI ..... ..,. 10

14~,;r1 '" q .... if, _ IOHOO -"'''' ,_ F",", "POO' F .. , ... ~_. S ... " ocr .... N"S'" omu S- s.. .... "'ppI .. " p .

o.ioooII. Il . S .. _ fl. II. Fl ...... '''S. ~ I ..... ...,.: a-atkoIioII of ,.., __ ~

, ... ,. £ ...... , ... at Elt TS.1 .... roo r ..... on<! "..-,.1 _ ,... VSD... F_, $0 ... l .o. ...... r$W· lll. " p. ..... hdll< So.,k_, ~' .... , _ Il_ Eop. S<a.. IIo!~. Calif.

' idiot. Il . c.. T04, CoOl". IJ1J . E ........... at (11'$· ' do,. , ... , .... , ."' ._

..... ......" .. USDA P", ... $< .. Il ... '- r$W.(Il. i 7 p .. iDM. radii< s.,.,_ 1'0'''' "" 11.0 ... eop. $ .. ~ _""'. Cdr.

HdIor, l . C_ J . L e. .. , 00Id F. ' . lnlI/tl. 19l9 . ..... 101 _..., • .., _t It .. ..... 10 I., ... H ... ,

s .... l'llpo< "6. 1 p. lloct, )I"" .... F ..... ""' __ t . p. SIn .. Fon CollI ... Colo.

H_ b .. not •• G I' H. 0.. "" _III .. '01~''' r... _ .-.tlooul ...... _

"" "" .... 001 ........... 01 .No. lI.o_f~font/l . ... 17; (i ..,.,

H ......... Il. L 19 74 , Sr<trol ............. ,001 lAolyt<. of . ........

pion;" ,n .... "" ,--0 ....... ~ .... Pm<. ~I ..... So<. of /'bo",..,plieoI I .. ,,,,,,,. EAc. s..n;". _ !i<..-n

and. 1."",. S",. r.,. EM,h """" •. !5uo Di<oo. C>lif .. " 1.1;1"" !0·10. " 741. 11'9(1.10»

' __ •• Ii. ' %S. Opti<o! p-!oo,op.phi< _n .. l"...,. .,. . ..... ,. p.

1.1-76. 'olin W.." .. d Soa!: ).I ... V.,.k "'" 1.00<100. x.. n. 1:. P. P.

I'll. £ff""" of "_,<tic <Ott«'''' "" d . ";foeo';"", 1,0<1.10=1 Ei<>c!ro ... Co" It>< .. 11""""0 """"'",, Sy". 0;, .. UO."OR. !'ubi. LEe N"",beo TIoI 6oI 1-6 W.

" K, . .. 4 . L ... .wl S. POlO, 19n. Mol .... ' . lI«j. i~'io. of foo .. ! W"''''', ''''' bo m

>po<. ODd ';'"nl , ...... y .nc! "",,04 .... pl .,.. ,. Sp.o« R"","'" 1((. ho<. of 'lI< Fiflc.n,b 1'1 .... <)'

!oJ «1./; I IH ( 1. lI",ley . P. G., R. C. "'Id""" . • nd R.C. H.o. •.

( %9 .. '1 011ll_ .... p! ... of f", ... ,..., .. coo b, . site "",co p",-,,,,..opIIy. I . S,""ooo ~""., E.,,10 R<>o w . Air<>. I'r"l",n Sa, .. R .. . Vol . 2. A.!!k./fo, .• 00 So ... 8'11<1 <0.. ~"'SA !oJ"""", Spo<o<r . C .. ,., 110,",'0" p, 19·1 '0 1" 11 .

14 ...... Colo.! eomp.~ lt lGo L%O. III ..... 1000/; of _. M. _ll Colo! CG.

..10-" Nllloool A'_''''' .... Sp"", ~.bm";'""ioo. Godc!. o<! S_ FJ1Vo' C • • , ....

1911. Do .. _ ....... _ : NAIl ... E.,,~ a..o", ... T",~ s""'-~,., Doc:. No. 1ISJ)O;!.49. NAS ... ~~, I'1l1/0, C ..... , . G" ... 0eIt. IoIiJ. r . G-I4.

1ol.1looW ......... of S,o,od" ..... 1'$', no ISCC-HBS __ .., ~""I'" _ '"" ..

.1<-,. of .............. 1I.s. D<p.of C ....... NBS. CIo<tobo llJ ,

liIttIoIo. l . D~ M. Gooldlool."" S. jO>l;kolo. IJ'lJ. ~ • __ kKY ___ "'_1<>-

__ .... ly1ioulE .. TS-' .... _ .... _' do .. ..... _,ioO .. ,..-iIiIl _JOOOa. Vol. " T.,fo. rr ....... 'iDM, Soc. .... Thlnl f.ortlo ~. Ttdo. $oooll!oe-I 51'"1'. NASA GoddaJ d Spo.cc FIi&I>' C<n .... G_h<It. OW. po ! ~ s.Jn.

.._ of ,110 bMQI< _ ... It&O<,,,,b Wo,t Unit. 1m. _ ... "" ........ '""" .... oI.~ __ _

___ F" .... t "POO' I", J;ortII _ . s...-. 1'0_ . N~SA 0L"II<0 ofSt-e sa. .... Appl., "l po .....

.. '''~ .... Go Il. "10, G_ '0."",,"" _. 'Of I. 1 ....... ' _ s;o.

IofI<aI "..... .. 1 .. __ ....... ..-fico_ of _ ,,,,,,,,,,, -"" _ bjll', It. -. IBI' H.ftdto. • • _ 5<l.1'robl~ (h.f""' , I II ...

""" • • It. D .. oad I. C. Corltoo. Co-dui, ..... 1'73. t:COCUu_. __ roo ola"', ... _ 4 ....

USDA " _ ,, 5<0 _ t;e..q .. _ TOOk 1'. ",.. ~.p. 011 fIl< .. F", .. ,.,. $d . l.o~, 10",-'. F",..c .0<1 11.0_ r..... $,,,-""-'. 101 ... , . $1 p.

r~"." I. F. 1974. II ............... eIf",1> "" ,."'mlll< d"';' ''''·

,io. of ,,'''~'e "'"..m.IIIa,"' ..... ........ , ..... !'roc. 50<. """''''''I',1cal In"''''''. c.... 5<01110 . ... SUnnt" and ["_1 51". 10. IA:tlI 01>0. Is.. Diqo, CoIlf., A_, [HO.I~"I.j l '?J.lll.

Il_~ 11. . II., 1(. _ .. .... N. J. SNII. 1973. A "'Ml.!_ "" _ U., u;rs "' .. k>r """"

.... ,, __ ell", ... "'oc. of Tbltd EltTS Sy",p .•

"' ...... ''''' D.C. ..... I , -. J, I' I 7IJ·II(Io'. Sc ... in. II. C •• I •• , UOIIJ. It AI ........

Ino. 111_ of _ _ ... diff_ Il"g ...... 10}'. AwL Opti<>. 9(' ),171-11 • .

TIt" ..... , V. L. Itli. Ge_riooo .... ...,..... ........ , .. _ 0' ,100

EilTS !!ISS .,. __ "" ......... _ '1lIIo .. ,... hbI. Ii<>. "'·~HHOt. I<AlIAGo!Ward S ..... FUPt c.., ... ~. loW.,. 1>-1 10 Il-0l.

lI.lI. FMOSt-. 19' •. F_ s....,,_, ,oJ""", (t'Sl' Ui9.n~

U.s. FGfeOI Sen~ "'''' , " o, D.c~ 0.,.. 10 11" ' ·[·

U.s. F_ s...Ioo. l%ll. F",... 5...., N_ I.,. "'" __ Soot,I>­

...-0 Fom' Eq.S ..... """'. I<.c. p U 1'1.

'11' ..... . F. P 1"9. Remot. So""", Impll<odoA! of wo!O, • • (,cil 0""

_ mo,.., .... ;,. to. 1"'''''_ pia< .tI",kod hy h.t. ......... IId _uoI diJo_ mp ........ I'h. D. "' .... , U . ...... WIdl., A .... Aobo,. J'J I'

.".. ..... F.r .• R.C. _. F.G. s..iowllll, .<NI r . l . -1,1). ~ _ <w.iI" ... rio. ill "'" ..,.._ ... ,.,... . ...,. hy ... 1.1<,0« .... _ .. ,lid "",",P'I'_ ... ...... I. ...... £.1&10111 hi,. Syoap. .. It.,." .. s..,.;q of ~_, U"" . of Ni<:Io.. Eo";,.,.. II ... Iut.. ol !lido ........ AlbOl. l : ll l ·ll1

.... ...... F. P.,""" F. C.,.,.",. ... Int. R_ .......... 4 ... _ ill fo ... n._

"..,... Eq. J.I(l): I63-I1$.

"' ..... , 1'. P .. f;. II. 11 ..... 1>, ..... T. H. "' .. '< . .t". F ............. _rio .. I. E ..... '1oa of EIITS-l do .. 10< f_ ..... ~ __ L USDA F<>rerl

.sen.. 11. ... "'pet !'SW-tl}. ' 7 1' ....... I"a<IlI< So-<k­

..... FGfeOI ..... 11..0_ b p. Sta. • ......,. . CoIIf. "' ..... ,.,. 19". n...,;_..,. I. snE II ........ of _ .... ScI.

...:I Eot- p. Il'I.I" . 1010. -.. SoN: ,,_ YOIt. . ... 1-

Aldrich, Robert C., technical coordinator 1976. Evaluation of Skylab (EREP) data for forest and rangeland surveys.

USDA Forest Servo Res. Paper PSW-II3, 74 p. Pacific Southwest Forest and Range Exp. Stn., Berkeley, Calif.

Data products from the Skylab Earth Resources Experiment Package were examined monocularly or stereoscopically using a variety of magnifying interpretation devices. Land use, forest types, physiographic sites, and plant communities, as well as forest stress, were interpreted and mapped at sites in Georgia, South Dakota, and Colorado. Microdensitometric techniques and computer-assisted data analysis and sampling procedures were developed and tested against ground truth. Results indicate that only Skylab Sl9GB color photographs are good for classification of forest and nonforest land (9G to 95 percent correct). Both visual and microdensitometer techniques can separate range plant communities at the Region level (ECOCLASS system) with over 9G percent accuracy. Only mountain pine beetle infestations more than 26 m (85 ft) long could be detected. In a study near Redding, California, radiance from Skylab SI9GB and LANDSAT sensors was found linearly correlated with terrain reflectance.

Oxford: U629.l9[+587.7+44] Retrieval Terms: Sky lab; Earth Resources Experiment Package; pnotointer­pretation; microdensitometric analysis; remote sensors; forest classification; range inventory; plant communities; forest stress.

Aldrich, Robert C., technical coordinator 1976. Evaluation of Skylab (EREP) data for forest and rangeland surveys. USDA

Forest Servo Res. Paper PSW-1l3, 74 p. Pacific Southwest Forest and Range Exp. Stn., Berkeley, Calif.

Data products from the Skylab Eartll Resources Experiment Package were examined monocularly or stereoscopically using a variety of magnifying interpretation devices. Land use, forest types, physiographic sites, and plant communities, as well as forest stress, were interpreted and mapped at sites in Georgia, South Dakota, and Colorado. Microdensitometric techniques and computer-assisted data analys-is and sampling procedures were developed and tested against ground truth. Results indicate that only Skylab S190B color photographs are good for classification of forest and nonforest land (90 to 95 percent correct). Both visual and microdensitometer techniques can separate range plant communities at the Region level (ECOCLASS system) with over 90 percent accuracy. Only mountain pine beetle infestations more than 26 m (85 ft) long could be detected. In a study near Redding, California, radiance from Skylab S1908 and LANDSAT sensors was found linearly correlated with terrain reflectance.

Oxford: U629.19[+587.7+44[ Retrieval Terms: Skylab; Earth Resources Experiment Package; photointerpretation; microdcnsitometric analysis; remote sensors; forest classification; range inventory; plant communities; forest stress.

Aldrich, Robert C., technical coordinator 1976. Evaluation of Skylab (EREP) data for forest and rangeland surveys.

USDA Forest Servo Res. Paper PSW-113, 74 p. Pacific Southwest Forest and Range Exp. Stn., Berkeley, Calif.

Data products from the Sky lab Earth Resources Experiment Package were examined monocularly or stereoscopically using a variety of magnifying interpretation devices. Land use, forest types, physiographic sites, and plant communities, as well as forest stress, were interpreted and mapped at sites in Georgia, South Dakota, and Colorado. Microdensitometric techniques and computer-assisted data analysis and sampling procedures were developed and tested against ground truth. Results indicate that only Skylab Sl9GB color photographs are good for classification of forest and nonforest land (90 to 95 percent correct). Both visual and microdensitometer techniques can separate range plant communities at the Region level (ECOCLASS system) with over 9G percent accuracy. Only mountain pine beetle infestations more than 26 m (85 ft) long could be detected. In a study near Redding, California, radiance from Skylab S19GB and LANDSAT sensors was found linearly correlated with terrain reflectance.

Oxford: U629.19[+587.7+44] Retrieval Terms: Skylab·; Earth Resources Experiment Package; photointer· pretation; microdensitometric analysis; remote sensors; forest c1assification; range inventory; plant communities; forest stress.

Aldrich, Robert C., technical coordinator 1976. Evaluation of Skylab (EREP) data for forest and rangeland surveys. USDA

Forest Servo Res. Paper PSW·1l3, 74 p. Pacific Southwest Forest and Range Exp. Stn., Berkeley, Calif.

Data products from the Skylab Earth Resources Experiment Package were examined monocularly or stereoscopically using a variety of magnifying interpretation devices. Land use, forest types, physiographic sites, and plant communities, as well as forest stress, were interpreted and mapped at sites in Georgia, South Dakota. and Colorado. Microdensitometric techniques and computer-assisted data analysis and sampling procedures were developed and tested against ground truth. Results indicate that only Sky lab S1908 color photographs are good for classification of forest and nonforest land (90 to 95 percent correct). Both visual and microdensitometer techniques can separate range plant communities at the Region level (ECOCLASS system) with over 90 percent accuracy. Only mountain pine beetle infestations more than 26 m (85 ft) long could be detected. In a study ncar Redding, California, radiance from Sky lab S1908 and LANDSAT sensors was found linearly correlated with terrain reflectance.

Oxford: U629.19[+587.7+441 Retrieval Terms: Skylab; Earth Resources Experiment Package; photointerpretation; microdensitometric analysis; remote sensors; forest classification; range inventory; plant communities; forest stress.