T- I

Thematic Mapper Research in the Earth Sciences

Tit le of the Project:

Small Scale Patches of Suspended Mat t e r and Phytoplankton in the Elbe River Estuary, German Bight

and Tidal Flats

- Progress Report -

(bASA-CR- 182378) XtELAfXC LSAE€Eli EESEBBCR N88- 16 180 I b T E € liBti48 SCIEECZS: S E A L & ZCALE EATCIIES CE SUSPENDED L A 9 P E b A L E PHYZCEAAIKTCI IEI THE ELBE ElVEB E S ' I U A B E , G E b B A l EIG-EI T I D A L Unclas PLA3S Erogresr beFcrt (GKZS-fcrecbungszeatrun G3/43 0 1 18979

Principal investigator: H. Grass1 Responsible scientists: R. Doerffer, J. Fischer,

C. Brockmann and M. Stossel

G I< S S For s c hu n g s z e n t r u m G e e s t h ac h t FRG

October 9, 1987

https://ntrs.nasa.gov/search.jsp?R=19880006798 2020-05-24T20:55:59+00:00Z

1 1 1 3 I

' I 3 I I 1 1 ' I

Contents

1 Introduction 1.1 T h e Scientific Problem . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Experimental Objectives . . . . . . . . . .. . . . . . . . . . . . . . 1 .3 Project Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 The Field Experiment 2.1 Objectives and Strategy . . . . . . . . . . . . . . . . . . . . . . . 2 .2 bXethodology of the Experiment . . . . . , . . . . . . . . . . . . . 2.3 R e s u l t s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 Evnluation of a Thematic Mapper Scene 3.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Evaluation Procedures . . . . . . . . . . . . . . . . . . . . . . . . 3.3 R e s u l t s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Radiative Transfer Model . . . . . . . . . . . . . . . , . . . . . . 3 . 5 . . . . . . . . . . . . . 3.6 Detection of JVater Substances . . . . . . . . . . . . . . . . . . .

Determination of Atmospheric Properties

4 Conclusions

5 Further effort

Lit ernt ure

1 3 13 13 14 19 25 35

38

37

35

Chapter 1

Introduction

1.1 The Scientific Problem

T h e coastal region of the German Bight is characterized - as other areas of the Xorth Sea - by a complicated distribution of different water masses containing high Concentration of various substances, patticulary suspended mat te r and phytoplankton . These two concentration parameters are of high importance in ecological studies: Suspended matter has to be regarded as the main carrier of various inorganic and organic substances including pollutants a n d as the main substrate for biochemical processes. T h e phytoplankton distribution indicates areas of high biological activity (production, grazing, remineralisation etc.).

Present techniques to investigate or monitor coastal water masses of the North Sea are insufficient, because they can provide only a raw idea about the gross mass of substances involved in ecological processes, their distribution dy- namics a n d the transport mechanisms of the solid water constituents. Previous investigations indicate t h a t transport and ecological processes -as phytoplank- ton growth - are linked to small scale hydrographic features namely local fronts, eddies , convergence zones and areas of high turbulence. As main parameters, which control the formation of these features, one has to consider the shallow water bo t tom topography, the complex topography and current regime of the wadden sea, the fresh water run-off of the Elbe and Weser estuaries, the tidal currents a n d the general circulation of the North Sea.

Aicraft d a t a show tha t small scale temperature fronts a n d convergence zones may occur with sizes smaller than the geometric resolution of NOAA a n d CZCS da ta . These d a t a indicate also tha t smallscale structures are impor tan t for the t ransport of solid mat te r , which is characterized by continuous sedimentation a n d resuspension, as well as for the occurrence of high plankton concentrations.

I t is obvious tha t the small scale distribution pat terns require much more at tent ion. Its investigation is an important prerequisite for the understanding of the ecological process in this type of coastal zone.

2

J

3 l l ‘ 1

3 1 1 3 I 1 13 ‘3 !

T h e combination of high geometric resoiution along with the spectral capa- bility of the Thematic Mapper provides the unique opportuni ty to get a much better insight into this coniplex marine system and to improve the s t ra tegy for environmental monitoring programmes.

1.2 Experiment a1 0 b j ect ives

At the beginning of this T M research projects the following objectives were formulated:

T h e overall goal is to understand the mechanisms leading to the complex distribution pat terns in a shallow coastal region, which is influenced by a n es- tuary.

Special hypotheses are:

Bot tom topography and wind field control the tide depending formation of small scale water bodies, eddies and fronts.

0 T h e sediment t ransport mechanism is based on multiple resuspension a n d sedimentation events, which liappen on small scales as a function of areas with variable turbulence mainly driven by the tidal currents.

0 High phytoplankton concentrations are related to samll scale fronts.

For the application of TM d a t a i t will be necessary furthermore to test the following hypotheses:

T h e TJI allows us to determine a t least t he concentration of the total suspended matter concentration a n d to map their horizontal distribution.

A separation of the contribution of suspended matter a n d phytoplankton is possible.

Dynamic processes can be studied by combining T&l da ta , ship d a t a a n d multiple aerial surveys within short intervals.

1.3 Project Strategy T h e s t ra tegy of the investigation contains the following elements:

0 Model simulation of the TM a n d response s tudy with respect to quanti- tive measurements of suspended matter a n d phytoplankton chlorophyll. This s tudy should lead to a n atmospheric correction procedure and to algorithms for deriving the concentration parameters.

Field experiments with ship a n d aircraft for ground t ru th a n d for analyzing the tide depending dynaniics of the distribution phenomena.

3

Analysis of Thl d a t a with respect to the information content, distribution maps and the relation between the distribution of mainly suspended niat- ter a n d parameters as bot tom topography, the wind field a n d the current regime.

4

Chapter 2

The Field Experiment

2.1 Objectives and Strategy

T h e objective of the field experiment was to collect ground t ru th d a t a for the calibration of the T M d a t a in terms of radiances and concentration parameters a n d for the atmospheric correction. Furthermore the ship d a t a should give a link to parameters, which are not directly accessible by radiance measurements; this is particularly the relation between water topography, the actual water depth a n d the suspended matter distribution a t the surface. For this reason we selected sections in the vicinity of the Elbe mouth, where we expected a strong variation in the suspended matter distribution because of a distinct bot tom topography. These sections with a length of 10 - 20 km were traced with our research vessel 3 - 4 times during a tidal phase. A prerequisite to meet the same sections again a n d to identify the ship tracks within the satellite image was a n accurate navigation. T h e Syledis navigator equipment allowed us to position the ship with a n accuracy of about 1 m . In addition to the profiles t ime series of the same parameters were recorded during anchor stations. In addition a research aircraft of the DFVLR was operating the Daedalus Themat ic Mapper Simulator a t some days of the Experiment.

2.2 Methodology of the Experiment

T h e experiment was carried out during the period May 1st - May 17th 1986. T h e area a n d the profiles shows Fig. 2 .1 .TJI overflights were expected for the 1, 8 a n d 17 of May. During this t ime a total number of 29 horizontal profiles were recorded with the following instrumentation:

0 water w a s continuously pumped from about 1.5 111 depth through a ME- TRAJVATT turbidity meter, a n AMINCO fluorometer for chlorophyll con- centration a n d a BIE cell to measure temperature a n d salinity.

5

a SUAREZ in s i tu scatterometer was used to determine the t ime lag be- tween the in si tu turbidity and the response of the turbidi ty meter in the flow cuvette.

T h e water leaving radiance was measured with a DFVLR 12 channel filtherwheel spectrometer continuously froin the bow of the ship with a resloution of 2 spectra per second. This high rate was necessary to allow integration with a clearance procedure to achieve the 10 m resolution of the other instruments.

An irradiance meter mounted a t the top of the ship was used to record the downwelling global radiation.

T h e actual water depth was measured with a n ATLAS echosounder.

All d a t a including radiance spectra, turbidity, downwelling irradiance, water temperature , salinity a n d water depth were stored together with t ime a n d posi- t ion i n the central computer every ten meters during cruising a n d every minute during anchor position.

For calibrating the instruments water samples were taken whenever s t rong changes in concentrations were observed on the plotter. These water samples were analyzed for:

suspended mat te r dry weight

aborpt ion of the filtrate a t 350 a n d 420 n m wavelength t o determine the Gelbstoff concentration

chlorophyll a n d phaeophytin concentration

conductivity a t a temperature of 20’ C.

All ship d a t a a n d the values of the laboratory analysis were given to a B U L L SPS 9 UNIX computer for further analysis .

2.3 Results

The shipborne data have been analyzed up to now with respect to the following questions:

0 relation between turbidity a n d total suspended mat te r d r y weight,

typical scales in the variation of turbidity,

cross correlation between turbidity and water depth.

I 13 l J l 3 7

Figure 2.1: Location of the research aerea.

ORIGINAL' PAGE IS OF POOR QUALITY

I " I' 7

I ' -1

I d -1

i 3 I

11 I

‘I 1 I

T h e relation between the turbidity, measured by the METRAWATT continuous flow instrument , and the total suspended matter dry weight shows a high linear regression with a confidence level > 99% (Fig. 2 . 7 ) . This excellent relationship was surprising, because we normally find within the a rea a s t rong variation of particle sizes and composition.

A typical horizontal suspended matter profile of 15 km length together with the depth profile is presented in Fig. 2 . 3 . It demonstrates the high concentration and variability of suspended matter in the coastal zone which is characterized by shallow water, strong currents and a complex system of sand banks a n d tidal flats. T h e relatively smooth profile indicates tha t changes in the concentration happen on a scale of 100 to 1 km. At the present s ta te of the evaluation we are not sure if higher frequencies are suppressed by the measuring technique, because visual observations show often smaller patches at specific tidal phases. T h i s question has to be analyzed with the radiance spectra.

T h e correlation between water depth and suspended mat te r is shown in Fig. 2.4. T h i s figure confirms our hypothesis, tha t there exists a general relation between the water depth a n d the suspended matter concentration a t the sur- face. T w o reasons may be responsible for this result: T h e t idal current a n d the turbulence become stronger whenever the water has to flow over shallower areas; the vertical turbidity profile decreases from the bot tom to the surface.

As the next s tep we have tried to get information about typical scales by auto- a n d cross-correlation. T h e results of this analysis a re shown in Fig. 2.5. Indicated are the correlation coefficient as solid line and the 95% confidence levels as dot ted lines. T h e water depth is significantly correlated until scales up to 1 . 3 k m , the suspended matter up to 800 m.

T h e cross correlation between both parameters confirms the expected sig- nificant negative correlation up to a distance of 1 . 2 k i n .

T h e analysis is a confirmation for our hypothesis, that a main factor control- ling the horizontal distribution of suspended matter fields is the water depth.

i

:I .I 3 1 8

3

1 1

I7 -.

J

I

I I I 3

T M M A I ' 8 6 Correlation coefficient: 0.98

250.0 1

0.0 ! I 1 I I I 0 .0 50.0 100.0 150.0 200.0 250.0

Turbidity (fe)

Figure 2.2: Correlation between turbidity measured with the hlETRAWATT turbidi ty meter a n d total suspended matter dry weight as determined from water samples.

3

1 9

J

90.0 -

T h.1 h l A I ' 8 6

Profile 1 from May 17,1986

10.0-

0.0

3 3 1

I I I I I I I I 1

1

3

0.0

-5.0

-10.0

-15.0 , ' I ' I ' , ' I ' I ~ I 1

.o 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 Profile length (km)

T i m e (UTC + 2)

~ ~ I ~ I ~ I ~ I ~ I ~ I ~ I ~ I ~ I

9.40 10.00 10.20 10.40 11.00

Figure 2.3: Horizontal profile of suspended mat te r and water depth .

10

3

7

.

1

3 1 I 3 3 3 a I J I

1

T h i hi A I ' 8 G Correlation coefficient: -0.537

100.0

go.o]

80.0

80.0 W E" 70,0!

0.0 I I 1 I I I 0.0 2.5 5.0 7.5 10.0 12.5 15.0

Depth (m)

lo.ol Figure 2.4: Correlation between waterdepth and suspended mat te r concentra- tion in M 1.5 m depth.

11

3 1

-

'1 1 3 I :I 3 11 :I J 3 a 3 I J I 3 3 I

........... .............

. . . . . . ..... .......... . . _ _ . . _ .

.............

-0.5{ U

3

Q - l . o ~ , . ~ ~ ~ , , ~ . . , , , , , I , , , , , , , , , l , , , , ,- +i-;.-q 0.0 1000.0 2000.0 ~ 0 0 0 . 0 4000.0 5000.0 6000.0 7000.0 8000.0

Lor; (m) Cross Corre!ation between Depttl arid Turbidity

........................................... . _ 0 ........ 95% ........................... .....

........... L

I '7

........ . . . . ...

. . . ..... . . . . ............... ........ -r----F---1-r--1-, . . _ . , 1 ;. f7 (5 -1.0-

0.0 1000.0 2000.0 3000.0 4000.0 5000.0 6000.0 7000.0 ~ 0 0 0 . 0

Cross Correlation betweerl lurbidity arid Dept11 Lag (4

..................... 958 1.07

.................... 0 .......... 0 O S - . ......................... . . . .- ..' 4

- - .....

...... .... ........ ....... ......... R- , v----p-rrr'r++<-T

0.0 looo.o 2000.0 3000.0 4000.0 5000.0 6000.0 7000.0 8000,o

Lug (rrlj L

Figure 2.5: A u t o correlation of the suspended mat te r distribution arid the water depth a n d cross-correlation between both parameters of the profile of Fig. 2.3.

12

'I .I :1 . . .I :I 4 11 I 3 :I 1 :I 3 I] 3 1 3 1

Chapter 3

Evaluation of a Thematic Mapper Scene

3.1 Objectives T h e init ial steps in the analysis of TM d a t a were:

t o analyze the radiometric resolution and signal/noise ratio with respect to our task to derive concentrations of suspended mat te r and chlorophyll

0 to investigate possibilities for atmospheric correction

0 to analyze the information content of the T3I spectra over turbid water areas.

Further analysis will then deal with the scales and the relation between water dep th , the actual current and the suspended matter distribution derived from the TM scenes, the correlation with water mass distribution indicated by the thermal channel and with the information content of dry wadden sea areas.

3.2 Evaluation Pro cedur es

To evaluate the Thl scenes we had to develop or adapt the following procedures and computer programs:

0 procedure to read the tape and transforni the information into our image processing file structure

0 procedure to calculate pixel coordinates from geo geographical coordinates in order t o identify our ship course in the scene, t o plot the ship course in to the scence and to pick the pixel values out of the scene and add i t into our file with ship d a t a for comparing ship and TM da ta ;

13

7

3

I 3 I 1 ' 13

13 :I I1 1 -1 '3 -1 3 J

0 a classification program, based on a hierachical non supervised au tomat ic classification in order to separate different areas of water, sand banks, t idal flats, l and , clouds;

a factor analysis program which considers areas of only one class i n order to est imate the number of pararneters (factors) which determine the vari- ations in the T h l spectra a n d calculates the scores fcJr each factor in form of a n image.

Other programs have been developed for analyzing the effects of the atmosphere; they are discussed in chapter 4.

For our first analysis we used the TM scene of hIay lTth , which is nearly cloud-free, with small influence by haze a n d where we have good ground t ruth. Furthermore we have analyzed a scene of Aug. 1'231 in order to check the radiance dynamics over water and compare it with model calculations.

3 .3 Results Histogranims of all channels calculated only for the water a rea of the scene (Fis. 3.1) show the low Digital Numbers ( D S ) and the very narrow ranges. If one calculates the radiances a n d compares it with corresponding suspended mat te r concentrations (s.chap.4), i t becomes clear, tha t the signal/noise ratio is not always sufficient for deriving suspended mat te r concentrations with the required accuracy, a l though i t is possible to derive a distribution m a p for analyzing scales. For a detailed analysis we have considered the corresponding TM d a t a of our ship profile (discussed in chapter 2 ) . Fig. 3.2 shows the mean values, the s tandard deviations (hatched) a n d the overall range of the radiances along the profile. Obvious is the s t rong influence of the Rayleigh scattering of the atmosphere which is manifested in the steep decrease from blue t o red. Along the ship track the variations are rather small compared to the noise (Fig. 3 .3 ) . From this plot i t is obvious tha t channel 3 has the best signal/noise ratio with respect to suspended mat te r variations. Even the near infrared channel 4 is still influenced by the suspended mat te r scattering, while the blue channel 1 has the lowest variability, because i t is dominated by atmospheric scattering. An integration of 5 x 5 pixels improves the S/N and improves the s turcture i n the profile (Fig.

Factor analysis (DoerfTer,l981) opens the possibility t o es t imate how Inany independent factors determine the spectral variability i n the d a t a . To apply this method we have first separated the pure water area by a hierachical unsupervised classification procedure (Fig. 3 .5 ) . I t has been found t h a t channel 4 is sufTicient for a clear separation of water surfaces.

T h e factor analysis was then applied only to those pixels classified as water area. T h i s restriction is necessary, because otherwise cloud a n d land pixels

3.4) .

3 14

3 3 7

1

13 1 :I *I I1 1 1 1 'I .I 1 3 3 -1 3 3

~3 I

?

~

a ? ! t4 O i :- 1

0 1

5 d -

3 1

E l I

-1

2 0 '

L

% I

mean: 4.22 slond.dev.: 2.20

0.0 2 5 . 0 y1.0 75.0 !&.3 i z . 0 150.0 1 i s . o 2ob.o 22'5.0 2;O.o

0

0 1 ' -1 m

0-0- 5 9

s

9

O L ~ L L O L number

0

0 1 I 7 2 2 , 2 0

V

Y O Y ?

-1

/ I

TM c h o n n e i j

mean: 16.18 sLond.de.,.: ? . 6 3

9

'1

3.0 25.0 50.0 7 5 . 0 0 I k . 0 ISO.0 IiS.0 2Cb.O 22'5.0 2 9 . 0 dLg\ LoL number

0

Tfl chonneL4

mecn: 7 . 7 8 sLond.dev.: 2 . j O

dcgLLoL number dLgLLoL number

P ? J N O

nn

meon: 3.0s T sLond.dev.: 1.42 m ? N O nn

0

0

0

-1

-10

L 0

m -1 0

QF! s

9 0

0.0 2 5 . 0 so.0 7 5 . 0 1m.o 125.0 150 .0 175.0 iw.0 2 2 5 . 0 250.0 dcgLLoL number

TM chonneL6

meon: 99.41 stond.dev.: 6.36

meon: 3.0s sLond.dev.: 1.42

F i g . 3.3.1: Statistical para- neters and histograms of all 7 TM channels of only "water pixels".

Profi le 1 from May- 17, 1986 Bands 1-5, 7

45.0 1

0 . 0 - 1 , , 1 I ' d r , 1 I 2 1 - 7 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4

We 1 1 e II 1 a en g e (/L ni)

Figure 3.2: T h e variability of the T M radiances of the profile in Fig. 3 .3 .

16

j I -1 ' I , I I 40.0 7

35.0 -

25.0

20.0 -

15.0-

10.0

'r A I h l A I ' 0 6 Prof i le 1 f rom S l a y 17,190G

30.0;:

I I I I I I I I

0 .0 - I I I I I I I 1 14.0 16.0 2.0 4.0 6.0 8.0 10.0 12.0 0.0

m

E 3.

2 Y

25.0

20.0

40.0

35.0 l 'u 30.0 -I

z 20.0- :--- 0.0 2.0 4.0 6.0 8.0 10.0 12.0 11.0 16.0

Figure 3.4: Radianccs as i u Fig. 3 .3 al ter inlegrnl ion of 5 x 5 pixels.

18

1

TAI 2 T M 3 T b l 4 T > I 5 TM 6 TAI 7

-1

0.319 0.905 0.821 0.562 0.258 0.426

1 1 1 .I 1 -1 3 ' 3 11

1 TM 1 I Thl 2 I Th1 3 I TAL 4 1 TAL 5 1 T h I 6 1 T h l 7 I I i 2 ,

TbI 1 1 1.000 I 0.919 1 0.905 1 0.321 I 0.562 1 0.258 1 0.426 1 0.919 1 1.000 0.962 0.808 0.475 0.338 0.367

0.962 1.000 0.556 0.484 0.340 0.369

0.508 0.856 1.000 0.617 0.295 0.468

0.475 0.984 0.617 1.000 0.174 0.579

0.338 0.340 0.295 0.174 1.000 0.125

0.367 0.369 0.468 0.579 0.125 1.000

0.962 0.962 0.856 0.617 0.340 0.579

Table 3.1 : Ko r r ela t io n rn a t r i x and K o rn In u n ali t 5 ten

would change the covariance mat r ix and thus hide or suppress the variances of the radiances over water .

T h e resulting correlation mat r ix is given in table'(3.1). Th i s mat r ix indicates three groups of channels: channels 1-4 are highly correlated, also 1,4,5 and 7 while channel 6 is more or less isolated. Three dominating eigenvectors can be calculated from the covariances (Fig. 3.6). After orthogonal rotation using the Varimax criterion we can reveal 3 factors which determine 73% of the variances of all 7 T!vI channels over coastal waters. T h e strongest contribution to the conimon variances comes from channel 3, while channel 6 is nearly isolated (s. column H' in table 3.2). T h e factor loadings - which represent the correlation between the T M channels normalized radiances and the factors a re also shown in Fig. 3.6. From these loadings and the correlation mat r ix we calculated the regression mat r ix in order to estimate the factor scores from the normalized image, which are the manifestations of tlie factors for each pixel.

T h e result is very surprising: factor 1 shows very clear the distribution of suspended mat te r (Fig. 3.7), while factor 2 independently maps tlie scattering of the atmosphere (Fig. 3.8). T h e image of factor 3 is very similar t o the sea surface temperature distribution in channel 6, although i t seems t h a t also for this IR channel influences of the atmosphere are at least partly removed.

This example has demonstrated tha t i t is possible to separate formal factors which independently determine the variation of the T h I spectra and which can be related to real parameters as suspended mat te r , atmospheric scattering and SST. However, the application of this promising technique has to be further investigated within this project.

'

3.4 Radiative Transfer Model T h e radiative transfer equation both for atmosphere and ocean is solved with the matrix-operator method. Although a n aziniutlially resolving version of this code is a-flailable, the small scanning angles of tlie Themat ic Mapper of &So

19

' 1

Figure 3.5: TM szene after classification to seperate the water area.

20

-

1 1 1 1 I 1 1 I I 1 3 1 1 1

1

' 7

- CJ) C

71 0 0

.-I

-J VI_ c o 0 ..J 0 0 L

9 0 -

E i g e n v e c t o r s

--€I--

-f3-

I I I I 1 - 1 - - l

-6-

-e&-

e

c 0

*J

z- c 0.0 1.0

F a c t o r 1

I ' T u r b i d i t y 'I

2.0

--e

i 3.0 4.0 5.0 6.0

vecLor no. TH channel

, -] Figure 3 . 6 : Eigenvalues a n d factor loadings for the "water area". I

7.0

F a c t o r 3

I t S I - i r f a c e T e m p e r a t u r e I'

9 -1

I

8.0

I

i 1 I] 1 1 4

21

1 1 1

Figure 3.7: Ixuage of scores for factor 1 indicationg the suspended matter dis- tribu tion.

22

.

Figure 3.8: Iniage of scores for factor 2 indicating the atmosheric pa th radiance.

23

1

TAI 5 TM 6 TM 7

F 1 0.362 0.904 0.909 0.711 0.302 0.133 0.187 3.051

43 .553

F 2 0.376 0.241 0.252 0. IS3 0 .745 0 . 0 7 0.692 1.536

21.347

F 3 0.131 0.2iG 0.290 0.239 0.110 0.545 0.070

0.54s 7.934

0.336 0.159

Table 3.2: Rot. Factors and Kommunalitaten

around the nadir oKers the choice o f a n azimuthally averaged code, which exactly determines nadir radiances. T h e rough ocean surface is handled as published in Fischer and Grass1 (1984). T h e radiative transfer equation in the solar spectral range for a plane-parallel niedium, is given by

(f& + 1)L(TI P ' , d, X ) = WO(Tr A).

. J J qr, PLL, 4'1 A ) . P(T,PL, d , P'! i ' , A ) dP'dd' 2 r I

0 -1 -. +wo(;,X) r F o ( X ) P ( r , ~ , i , p 0 , 6 0 , X ) e =

(3.1) T h e radiance L and the phase function P are both functions of height z,

here expressed by the optical depth T = c ( : ' ) d ; ' , whereby the extinction coeficient c = JW) + C ( I ) + c(') + c ( y S ) in coastal water is either the sum of pure water, phytoplankton, suspended mat te r , and yellow substance extinction in water or the sum of Rayleigh and &lie scattering in the atmosphere. L and P also depend 011 aziiriutli and zenith angle and B ( p = cosB) as well as 011 wavelength A. T h e phase function P is a combiriation of all individual phase functions weighted with their respective sca t t e r in s coeflicients. T h e single scattering albedo wo is the ratio of the total scattering and extinction coefficient. T h e solar spectral irradiance Fo is taken from Neckel and Labs (1931).

Since tlie mat r ix operator method has been shown to give highly accurate solutions tlie main error sources in tlie solutions of Eq.(3.1) are improper optical parameters for given size distributions and compositions of distinct substance classes like suspended inat ter and phytoplankton.

An adequate simulation of Themat ic Alapper measurements also requires an adapta t ion of the radiative transfer code to the characteristics of the TiiI chan- nels (Barker, l931). Their poor spectral resolution does not resolve the changes of atmospheric and oceanic optical properties with respect to wavelength. As an f

1 1 I

example, the transmission of water vapour, according to a Inidlati tude summer atmosphere, is shown in Fig. 3.9. Additionally, the transmission functions of the sensor filters have to be introduced (Markhain a n d Barker, 1953). Tliere- fore, we have to clarify, which spectral resolution for the radiative transfer code is necessary, i n order to describe each of the T M channels.

For realistc variations of aerosol optical thickness between r, = 0.14 a n d 0 . i7 a t X = 550nm, as well as of water substances in coastal areas we simulated the T M channels 1 - 5 with 1, 5 , 10 a n d 30 nni spectral intervals. T h e calculations with 1 nni spectral resolution we accept as exact simulations. In table 3 .3 the mean, maximum a n d minimum variations 5 of upward radiances jus t below ocean surface a n d a t tlie top olat inosphere due to the reduced spectral resolution of tlie radiation code are shown. T h e criterium chosen for a sufficient discription of the T M channels is tha t tlie errors a re smaller than radiometric resolution ( table 3 .4 ) . T h i s is true for 1 0 n m resolution. Channel 5 may be simulated sufficiently accurate with 20 n m intervals.

3.5 Det ermiiiat ion of Atmospheric Properties

A main problem in processing TAI measurements is the determinat ion of the multispectral water-leaving radiance by removing the atmospheric p a t h radi- ance, which is primarily due to scattering by air molecules a n d aerosols a n d ocean surface reflection (Fig. 3.10). T h e accuracy of retrieved water substance concentrations is always l imited by the accuracy of atmospheric correction.

T h e assumption of a single scattering process allows to seperate the measured radiance LT.vf(X) of the TAI into a n ocean contribution t (X)L , (X) , which is reduced by t h e diffuse transmission of the atmosphere t ( X ) , a n d a n atmospheric par t d u e to scattering by aerosols La(,\) a n d molecules Lr(X) :

&7’.\f(X) = t ( x ) L w ( X ) + L a ( X ) L,(X). (3 .2 )

Following Gordon (1975,1981), the ocean leaving radiance L , ( A ) a t any wave- length X may be calculated froin satellite radiance L T J ~ ( X O ) , Rayleigh contribu- t ion L,(Xo) a n d aerosol par t L,,(Xo) a t a near infrared wavelength XO for which L, (Xo) = 0 :

i(X)L,(X) = L T M ( X ) - L r ( x ) - [ S ( x i X O ) ‘ ( L T . \ . I ( X O ) - L r ( X O ) ) ] i ( 3 .3 )

where S(X, A,) = [.a(X)F(X)]/[.~(Xo)F(Xo)] relates the optical thickness of the aerosols 7, a t different wavelengths a n d the spectral solar constant F ( X ) . If we introduce r a ( X ) - X-” , with the Angstrom-exponent a, we simply get

25

'I71 I T.1 4

i I 0 ,

0.55 0.60 0 0.95

waveleng t h ( /mi)

1.5 1.6 1.7 1 9 2 2.1

wave !en g t 11 (/. Lm)

100 i 'I; I 1: 80 I .-

P l 4 'IF1 3 nt 2

!-. I E 20

i d 0 I I

0 0.55 0 65 0 io 0 75 0 a0 0 85 0 90 0 9 5 1

I wavelength (/mi)

100

1 1

.- c o 60 .d

m v1 .- E 40 in

TM 5

C 2!)

J

I l !

'.' i 1 I

1.6 1.7 1.5 1 8 1 9 2 I I

2.1 2.2 2.3 wavelength (/.mi)

Figure 3.9: Transmit tance of watervapour (a) a n d other atmospheric gases (02, 0 3 , Cli4 and other) (b) , according to a Inidlatitude S u I i l l n e r otiuosphcre.

-.

a) spectral resolution: nm

I TAIL TJI2 I I I

5,ax [ 0.357 1 0.70 11 0.430 - 1 0.045 1 1 1 S,in 1 0.000 ~ 0.002 ;i 0.000 I 0.001 I

1 I Z = - O . O I j 2 = 2 0 0 0 0 , i s 1 0.067 i 0.009 i

2=-0.01 z=20000 I 2=-0.01 1 2=20000 1 0.043 0.004 1 0.005 0.001

0.02G 0.027 0.008 j / / 5,in i 0.000 1 0.001 1 0.000 I 0.003

Z = - O . O I 1 Z = ? O O O O \ 0.074 I 0.006 1

I 1 Tiif 1 TM2 1

I 6 I 0.222 i 0.014 i 0 .245 I 0.009 ' 2=-0.001 1 2=20000 j 2=-0.001 ~ 2=20000 1

I( ~ , U X 1 0.464 1 0.066 I1 0.513 I 0.043 -1 / / 6,in 1 0.000 I 0.003 1 1 0.000 I 0.002 d U

S I 0.005 1 0.016 1 I L a x I 0.236 1 0.025 I

J

0.00 j / I 0.033 I 0 .008 I /

I i 6,in I 0.000 I 0.001 II 0.000 ~ 0.003 ' I

2=-0.001 1 2=20000 2=-0.001 1 2=20000 I 5 1.000 1 0.048 1.110 1 0.031 1

' 6,az I 1.116 1 0.156 1 1.250 1 0.100 1

~~ ~ ~~

c ) spectral resolution: 20 nm

n T m 3 z=-o.o01 I 2=20000

Th14 z=-o.o01 I z=20000

I I !

6 1 0.639 1 0.019 11 0.071 S,az I 0.733 I 0.058 / I 0.052

J 0.006 I 0 .01 I n 6,in 1 0.005 i 0.006 ii 0.000 i 0.002 1

Table 3.3: Mean 5 , maximun 5,,,and minimum S,;, radiance differences to the 1 n m resolution for different coarser reselutions. All readiances i n LV/mzsrpnr:

l -

A % 58 L La

I 48

38

26

C

2 01 U c 9

W II

i U ie L

a ii 8

e,= 35"

1

1 2 3 4 . S

channel

I R a y l e l g h

n T h = 1 . 5

mi, = 2 , s

T , = 3,s

F a T , = 4 , s

L cn CI

C 38 2 al U c 9

W 9 L

W L

28

1 8

=I : 8

E U R a y l e I g h

R T , = 1 , s

mT2 = 2,s

E Z I T , = 3,s D T , = 4 , s

1 2 ;I 4 5

channe l

Figure 3.10: Upward radiances for TAI channels 1-5 for a clear atniosphere and for various atmospheric turbidities - indicated by the Linke - turbidity factor.

28

l -

Table 3.4: Radiometric resolution of tlie Themat ic Slapper in IV/77t29~p771 (re- ferring to calibration values for 1 DN from 17th of May 1986).

As shown in Fig. 3.11 tlie power law for .,(A) - A-" is valid for tlie continen- ta l a n d mari t ime aerosols used in this study. The Angstrom-exponent can be determined from

In r,(A~l) - In r,(XO,)

In A02 - In ,\Ol

if the aerosol optical thickness r, a t two wavelengths XOl a n d XOz is known. To show the importance of knowing the Angstrom-exponent for atmospheric

correction, the wavelenstli dependence of the extinction coefficients for mari t ime a n d continental aerosols are arpin given in Fig. 3.12, however, together with the TM channel intervals.

T h e Themat ic hlapper channels 4 and 5 can in principle (see Fig. 3.13) be used for a n estimation of aerosol type and concentration, as long as the concen- t ra t ion of suspended mat te r in water is low. As shown in Fig. 3.14 suspended m a t t e r concentration of more than 1 0 mg/l causes a non negligable contribution i n channel 4, while channel 5 remains unafected. In case of h igh concentrations we have to include the contribution of the ocean leaving radiance in channel 4 for a determinat ion of the Lngstrert i exponent. T h e ratios of cliannel 4 a n d 5 show clearly differences for mari t ime and continental aerosols (Fig. 3.13) . For- tunately this result is nearly independent o n atmospheric turbidi ty a n d solar elevation.

However, the biggest problem is the poor radiornetric resolution or low signal-to-noise ra t io of tlie TSI channels. T h e only way to overconie this restric- tion is a n integrating over many pixels. Since we know from CZCS rrieasure- ments t h a t the horizontal distribution of aerosols is Inuch more homogeneous t h a n for water consti tuents, the lower spat ia l resolution is no drastic drawback. In order to analyze possible errors of the atmospheric correction we considered three different atmospheric conditions ( table 3 .5 ) . IC we asslime a noise of i l D N in channel 4 a n d 5, a n increase of the number of integrated pixels reduces drastically the variations of cliannels 1-3, after subtraction of a tmospheric p a t h radiance. An integration of channel 4 a n d 5 by 3 3 x 33 pixels - nearly the pixel size of CZCS - further improves tlie atmospheric correction (see 3 . 5 ) . T h e resulting max imum variation of 0.14~~/7712srp7n in charinel 2 approxiiiiateiy cor- responds - according to Fig. 3.14 - to a change of 0.25 m g / l suspended m a t t e r concentration.

a = (3 .5)

29

Figure 3.11: Aerosol optical depth due to continental or marit ime aerosols as a function of wavelength.

30

.

maritime

I

\

continental

! I

t i

0 6 0 8 1.0 1 2 1 4

Wavelength (urn ) 1 6 I 8 2.0

Figure 3.12: Normalized extinction coeficients for niaritirne and continental aerosols after Shettle and Fenn ( 1 9 i G ) ; the TAI channels are indicated.

31

w 0

36 00=35', con t . aerosol

+ 00=51', c o n t . aerosol I I

u -

0 G 0 = 6 6 * , c o n t . aerosol

00=35', marit. aerosol X O o = 5 1 ' , marit. aerosol

V O0=66', marit. aerosol

15 -

I

5 -

E i , e i , 5 Z J 2 s 3~ ~5 L O ~5

Figure 3.13: Radiance ratios of channels 4 and 5 depending o n atmospheric turbidity; the contribution due t o molecular scattering is substracted. T h e upper curves a re for continental aerosols a n d the lower curves a re for Inaritirne aerosols; the different solar elevations a r e of minor importance for b o t h types of aerosol.

32

20 c

1 1 8 0

!

I

conl inrn ta l aerocol

T l l n k r = 3 0 TLII

200 4 0 0 6 0 0 800 i o 0 0

suspended m a t t e r (rn: ' I )

I

cnnt inrn la l a e r o c n l

Tltrikr = 3 0 TM7

i 0 0 '

I 3 0 Z O O 400 600 n n o i o 0 0

I zuapended ma t t e r (rng 1)

cont inenta l aerosol

T l l n k r = 3 0 Thl3

ront incnla l arrn5o1

Tlinkip = 3 0 Tv4

20:

I e j

Figure 3.14: Upward radiances for T M cliannels 1-4 depending on sus- pended mat te r concentrations and for various chlorophyll concentrations :

0 pg/Z (a) 5 pg/Z (O), 10 pg/Z (A) , 15 pg/Z (*) a n d 2 0 pg/Z ( x ) .The solar zenith angle is 00 = 50.7' .

I

L I I I t

T A I ~ I 2.76 1 0.53 I 0.48 j 0.26 I 0.22 j 0.07 J

TJI3 1 1.75 I 0.55 1 0 .32 1 0.18 1 0.15 I 0.05

I TMl T M 2

I

I

5.29 1.64 0.97 0.53 0 .43 1 0.14 3.66 1.15 0.63 0.35 0.31 I 0.10

I

a)lnaritime aerosol , T, = 0.1 at X = 550 ntn

n 1 1 1 3 1 5 1 9 I 11 I 33 TAl1 1 3 . 6 3 1 1.05 1 0.61 1 0.33 I 0.27 1 0.09

b) maritime aerosol , r, = 0.3 a t X = 550 nin

1 Th12 1 2.47 I 0.80 I 0.47 I 0.26 1 0.21 1 0.07 I \ I I I I I 1 T M 3 1 1.55 I 0.54 1 0 . 3 2 1 0.18 1 0.15 I 0.05

c) continental aerosol , T, = 0.3 a t X = 550 n111

I n 1 1 1 3 1 5 1 9 I 1 1 1 3 3 l

I TM3 I 2.07 I 0.67 I 0.40 I 0.22 i 0.18 I 0.06 I

Table 3 . 5 : Calculates errors of atmospheric radiance in tV/ni2 srprrt for channels 1, 2 and 3 caused by -C 1 DN in channels 4 and 5 depending on the number of pixels ( n x n) over which is averaged:

34

3.6 Detection of Water Substances A detect ion of water substances f rom T h e m a t i c l l a p p e r nie3surenieiits is oniy possible wi th channel 1, 2 a n d 3 , because of tlie high absorpt ion of water molecules a t longer wavelengths. An evaluation of suspended m a t t e r i n coastal waters is addi t ional ly restricted because of the low spectral resolution a n d the s iniul taneous occurance of phytoplankton a n d yellow substance.

A separa t ion of water substances is difficult even for CZCS nieasurements , which a r e more sui table for water monitor ing t h a n T M measurements (Fischer e t al., 1996). Whether the spectral resolution of tlie TAI channels 1, 2 a n d 3 is sufficient for a separat ion is not clear u p to now. However, the dominant water subs tance variations in coastal waters a n d estuaries a re dominated by suspended mat te r . Phytoplankton a n d yellow substance reduce t h e upward radiances, mainly i n channel 1 a n d less i n channel 3, a n d 3 , because of their absorpt ion. IVe expect a n underest imat ion of suspended m a t t e r concentrat ion for areas with higher a m o u n t s of yellow substance a n d phytoplankton (Fischer e t al., 1986). As shown i n Fig. 3.14 a n increase of chlorophyll f rom 0 t o 20 p g / l reduces t h e radiance i n channel 2 by 1.9 a n d i n channel 3 by 0 . 6 ~ V / 7 n 2 s r p m a t a suspended m a t t e r concentrat ion of 20 mg/l .

T h e low signal-to-noise ra t io of the Thl channels only allows a discr iminat ion of roughly 2 mq/l per DN (see Table 2 a n d Fig. 3.14). If we assume a S/N rat io of 1 D N , only concenrat ion classes of 5 mg/ l can be detected. h n avering over several pixels, i.e. 10 x 10 pixels, would enhence the the signal-to-noise ra t io by a factor of 10. T h e resulting pixel size of 300 x 3007n’ is still sufficient for most coastal water monitor ing tasks , as shown i n chapter 2.3 .

For t h e development of a to ta l suspended m a t t e r a lgori thm addi t iona l radia- tive transfer calculations will b e performed, i n order to set u p a s ta t is t ical basis, which allows us to validate concentrations retrieved from TM measurements .

35

Chapter 4

Conclusions

T h e factor analysis has shown that the suspended matter concentration, tlie atinsopheric scattering and the sea surface temperature can be retrieved as three independent factors, which determine the variation in the TAI d a t a over water areas. This method is also good for mapping the horizontal distribution of these parameters.

For the development of algorithms based on radiative transfer calculations i t is necessary to elaborate an atmospheric correction procedure. The spectral channels in the near infrared (4 and 5 ) open the possiblity to determine the Angstrom exponent better than for the coastal zone colour scanner, if one in- tegrates over about 30 x 30 pixels to improve the S/N ratio. The suspended matter distribution may then be calculated by the absolute radiance of channel 2 or 3 or the ratio of both.

Up to now we have no indication, whether the separation of clilorophyll is possible.

T h e distribution of suspended matter and sea surface temperature can be observed with the expected fine structure. Particularly the factor score image shows details, which have never been observed so far with other satellite da ta . At the present state we are optimistic in stating that the rnain task of our investigation concerning the suspended matter distribution will be reached with TM data .

T h e good correlation between water depth and suspended niatter distribu- tion as found from tlie ship da t a can now be analyzed for a n entire area by the synoptic view of the TM scenes.

3F

Chapter 5

Fur t lie r e ff o r t

I t is nearly inipossible t o compare the horizontal profiles recorded from tlie sh ip wi th t h e corresponding tracks i n the image because even a t ime lag of only half a n hour between TAX d a t a a n d the ship datacauses very marked changes i n t h e suspended m a t t e r concentrat ion a t a locat ion within this d y n a m i c a rea . T h u s a cal ibrat ion of Thl d a t a is not possible by a direct coniparison.

In order to link concentrat ion d a t a xneasured from the sh ip t o the radiances o f T h l , we have to analyze t h e radiance spec t ra measured from board t h e sh ip a n d then a p p l y t h e relation found to tlie ThI scene al ter a tmospheric correction. Another s t e p wiil be a comparison with the Coastal Zone Colour Scanner scene for t h e s a m e date .

T h e final investigation will concentrate on the analysis of scales using TI1 d a t a , t h e sh ip profiles a n d ba thymetr ic a n d current maps.

37

Bibliography

[l] J.L. BARKER, 1981: Relative Radiometric Calibration of Landsat T h l Reflective Bands , Significant Figures for T?II Dynamic Range Constants ; in: Landsat-4 Science Characterisation , Early Results I11 - 145ff .

[2] R. DOERFFER, 1381: Factor Analysis in Ocean Color Inter- pretation; in: Oceanography from Space, J.F.R.Gower, Ed. (Plenum Press, New York, 1981),pp.339-345.

[3] D. EISMA, 1981: Supply and deposition of suspended matter in the North Sea; Spec. Publs. int. Ass. Sediment. 5, 415.

[4] J . FISCHER, 1984: Reniote Sensing of Suspended Matter, Phytoplankton and Yellow Substances over Coastal Waters, Part 1: Aircraft hfeasurements; Mitt. Geo1.-Pal. Inst. Univ. Hamburg, SCOPE/UNEP Son- derbd. 55, 85.

[5] J . FISCIIER B IC. GRASSL, 1984: Radiative Transfer in an Atmosphere-Ocean System: An Azimuthally Dependent hlatrix- Operator Approach; Appl. Opt. 23 , 1032.

[6] J . FISCIIER, R. DOERFFER 8 11. GRASSL, 1986: Fac- tor analysis of multispectral radiances over coastal and open ocean water based on radiative transfer calculations; A ~ ~ I . opt. 25, 448-456.

[7] H.R. GORDON, 1978: Removal of atmospheric effects from satellite imagery of the oceans; Appl. Optics 17, 1631.

38

..

[S] 1I.R. GORDON, 1981: A preliminary assessment of the Nimbus-7 CZCS atmospheric correction algorithm i n a hori- z o n t a1 in 110 mog e neo us a t 1110s p her e ; in: Oceanography from Space, J.R.Gower, Ed., Plenum Press, New York, 257-266.

[9] B.L. MARKIIAM & J.L. BARKER, 1993: Spectral charac- terization of the Landsat Thematic >lapper Sensors; in: Landsat-4 Science Characterization, Early Results, Vol.11 235-2i6.

[ lo] IT. NECKEL 2 D. LABS, 1951: Improved Data of Solar Spec- tral Irradiance from 330 to 1250 nm; Sol. Phys. 74, 231.

[ll] E.P. Shettle & R.W. Fenn, 197F: hlodels of the Atmospheric Aerosols and their Optical Properties; ADARD Conf. Proc. 153.

39