The Man computer Interactive Data Access System: 25 Years of

271Bulletin of the American Meteorological Society

1. Introduction

The Man computer Interactive Data Access Sys-tem (McIDAS) has had a substantial influence on thecomputerization of the atmospheric sciences. McIDASmade its first major impact analyzing cloud drift windsderived from time sequences of geostationary satelliteimages. In its early years, McIDAS was used to pro-duce television broadcasts and was the unparalleledforerunner of the television weather graphics indus-try of today. In 1977, McIDAS provided a major im-petus to the computerization of the college classroomfollowing the National Science Foundation (NSF)-sponsored Interactive Video Displays for AtmosphericStudies Workshop. Eventually, the Unidata programwas born out of that workshop. McIDAS was the firstinteractive system in operation at the Forecast SystemsLaboratory, a research laboratory of the National Oce-anic and Atmospheric Administration (NOAA) in1980.

The McIDAS systems installed at the NationalMeteorological Center [NMC, now the National Cen-ters for Environmental Prediction (NCEP)], NationalSevere Storms Forecast Center [NSSFC, now theStorm Prediction Center (SPC)], and the National Hur-ricane Center [NHC, now the Tropical Prediction Cen-ter (TPC)] were used to help define the technicalrequirements for the Advanced Weather InformationProcessing System (AWIPS). AWIPS is one leg of theNWS Modernization Program, which, in addition toAWIPS, includes the new Doppler Weather Surveil-lance Radar, the Automated Surface Observing Sys-tem, a new generation of Geostationary OperationalEnvironmental Satellites, and the National Center forAdvanced Computer Systems. AWIPS is an advancedweather data processing system that will aid meteo-rologists and hydrologists to rapidly manipulate, dis-play, and analyze critical weather data.

Research and forecasting worldwide have im-proved using the geostationary atmospheric sounderanalyses and derived products first developed withMcIDAS. Since its initial use in a meteorological re-search project in 1973 (D. Martin 1973, personal com-munication), McIDAS has been through four majorarchitectural changes, has been distributed and usedglobally for both meteorological operations and re-search, and has set a standard for all other interactive

The Man computer InteractiveData Access System: 25 Years

of Interactive Processing

Matthew A. Lazzara, John M. Benson, Robert J. Fox, Denise J. Laitsch,Joseph P. Rueden, David A. Santek, Delores M. Wade,

Thomas M. Whittaker, and J. T. YoungSpace Science and Engineering Center, University of Wisconsin—Madison, Madison, Wisconsin

Corresponding author address: Matthew A. Lazzara, Space Sci-ence and Engineering Center, University of Wisconsin—Madi-son, Madison, WI 53706.E-mail: [email protected] final form 28 August 1998.©1999 American Meteorological Society

ABSTRACT

On 12 October 1998, it was the 25th anniversary of the Man computer Interactive Data Access System (McIDAS).On that date in 1973, McIDAS was first used operationally by scientists as a tool for data analysis. Over the last 25years, McIDAS has undergone numerous architectural changes in an effort to keep pace with changing technology. Inits early years, significant technological breakthroughs were required to achieve the functionality needed by atmosphericscientists. Today McIDAS is challenged by new Internet-based approaches to data access and data display. The historyand impact of McIDAS, along with some of the lessons learned, are presented here.

272 Vol. 80, No. 2, February 1999

meteorological data systems. Now it is headed towarda fifth architectural change.

2. WINDCO: Before McIDAS (1971–72)

a. MotivationIn the 1950s and 1960s, Professors V. Suomi and

R. Parent played a key role in establishing an earth-imaging program using instruments on several of theNational Aeronautics and Space Administration’s(NASA) first geostationary communications satellites,the Applications Technology Satellite (ATS) series(Suomi and Parent 1967, 1968; McQuain 1967). Thestriking images sparked many new research initiatives.In 1965, Suomi established the Space Science andEngineering Center (SSEC) at the University of Wis-consin—Madison to better focus his team’s researchefforts on extracting quantitative information from theimages.

Suomi was enthralled with the meteorological phe-nomena apparent in time sequence film loops of theimages. He wanted to derive quantitative measure-ments of the winds based on cloud motion. WhileF. Hasler, a Suomi student at the time, was the first tosuccessfully measure cloud motion using a cinematictechnique (Hasler et al. 1968), this technique could notbe automated. The early efforts were quite laborious,requiring pasted together printouts covering 1000 ft2

or more just to show a portion of one image usingprinter characters such as . , ; / 0 * M to simu-late a grayscale. Once these printout composites werecompleted, staff would walk around in stocking feetto find a landmark or cloud.

Suomi established an internal SSEC competitionbetween engineers trying to measure cloud motionwith an optical correlator and a group trying to applysoftware techniques to the problem. E. Smith andD. Phillips were the principal developers of the com-puter techniques (Smith and Phillips 1972). They wereable to measure cloud motion using a cross-correlationalgorithm to determine the cloud displacement in suc-cessive images, and then producing displacement vec-tors using a satellite navigation algorithm developedby Phillips (Phillips and Smith 1974).

Following this success, Suomi obtained fundingfrom NASA and NOAA to implement a proof-of-concept system dedicated to measuring cloud driftwinds. The requirements for the new system were chal-lenging. The only computer-generated video systemavailable at that time required the largest IBM main-

frame to run it; that solution was not viable. The newcomputer system had to allow the user to specify theimage coordinates of a cloud in at least three succes-sive images. Since the digital images were stored oncomputer tape, some method was required to map anddisplay coordinates to and from tape.

b. Design and capabilitiesThe proof-of-concept system used to demonstrate

winds processing was called WINDCO (Smith andPhillips 1972). This system used a new product in thetelevision industry: an instant replay analog disk. Theanalog disk could support image animations of up to520 500 × 640 frames at 15 frames per second (load-ing a frame took about a minute). The data were writ-ten by a Raytheon 440 computer to precise locationson the analog disk using a fixed timing track, also onthe disk. A joystick controlled cursor location with thesame timing track. When the user pressed a button onthe console, the frame number and cursor location ofthe target clouds were punched into paper tape. Thepaper tape data were transferred to punch cards by theRaytheon 440. The target location cards, navigationparameters, image tapes, and the analysis programwere submitted to the university’s Univac 1108 main-frame. The output was cloud displacement vectors

FIG. 1. User at the WINDCO terminal.


plotted over a map of the earth (Smith and Phillips1972).

c. ImpactsThe prototype system was demonstrated to offi-

cials from NOAA, NASA, and NSF on 12 April 1972(Schwalenberg 1972). The demonstration showed thatabout 1000 cloud displacement vectors per hour couldbe measured accurately from geostationary satelliteimagery. The attendees encouraged Suomi to build anoperational version of WINDCO. A major limitationrecognized in the design of WINDCO was the inabil-ity to intercompare other data sources with cloud driftwinds. What is now called data fusion was addressedin the design of the next generation.

WINDCO (and later McIDAS) realized a data pro-cessing method in which high-volume raw data withlow information density could be processed by thecomputer (with user input) to produce low-volumedata with high information density. Figure 1 shows auser at the WINDCO terminal.

3. First-generation McIDAS: The firstHarris System (1973–78)

a. MotivationWith encouragement from NASA, NOAA, and

NSF, a design team was formed shortly after theWINDCO demonstration. J. Benson was the systemdesigner and principal architect. E. Smith was the ap-plications programmer and architect; D. Phillips wasthe mathematician and programmer; T. Schwalenbergwas the electronics engineer; S. Sitts was the hardwareassembly engineer; R. Krauss was the software devel-oper and systems integrator; and J. T. Young participatedin the design and implementation by representing the

science requirements. Professor Suomi was the prin-cipal investigator (J. Benson 1972, personal commu-nication; Krauss 1972).

b. Design and capabilitiesThe first task was to find a computer fast enough

to handle the data volume, control of the display, andcomputational load. One computer system met the re-quirements at a reasonable cost: the Datacraft /5 (laterto become the Harris/5). This system had 96 kbyte ofprogrammable memory, 1 microsecond per memoryfetch or store, a 5-MB fixed hard disk, and a 5-MBremovable hard disk (see Figs. 2 and 3). From theirWINDCO experience, the design team realized that theuser should input control information to the computer.The computer, in response, should adjust the displayand make it continuously available to all applications.This design put the computer in the role of support-ing the user and handling as much of the process aspossible. No assumptions were made about the user’scomputer science expertise. The user’s role was tocontrol the processes and perform the judgment tasksthat were difficult for the computer. The user selectedand staged the satellite data, evaluated and correctedthe navigation, and selected the clouds.

In addition, the user evaluated the quality of thecomputed wind vectors. Display of both the vectorsand upper-level wind measurements from conven-tional sources was required for the quality assessmentprocess (Fig. 4). Image enhancement was also addedto the design requirements to allow vector selectionin low-light regions near the terminators, the borderbetween night and day. The ability to efficiently pro-duce vector graphics on a raster display was essentialto meet these needs (Whittaker and Dedecker 1977).This display design would continue until hardware

FIG. 2. First-generation McIDAS terminal. FIG. 3. The components of the first-generation McIDAS.

274 Vol. 80, No. 2, February 1999

allowed graphics to be displayed independently of im-ages. This original interface design, however, contin-ues to be used today.

The software was divided into logical modules andimplemented as separate programs. Each was tested,modified, and retested independently of the rest of thesystem. Each module was a tool that the user couldinclude in the processing strategy as needed. From thisimplementation strategy arose the tool-kit concept thathas been central to the evolution of McIDAS. Themodularity of the software facilitated another conceptthat has served McIDAS well: to imple-ment a component of the system in a waythat will work and iterate until it worksto the user’s satisfaction. Applicationprograms were written in FORTRAN,while system-level programs were writ-ten in Assembler.

McIDAS was able to accept data in-put from a variety of sources [memo-randum by J. T. Young on inputs andoutputs of the system (1972); Krauss1972]:

• ATS archive image data (via tape)• Synchronous Meteorological Satellite

archive and real-time data (1.7 MB s−1)• Earth Resource Technology Satellite,

a predecessor to Landsat data (viatape)

• Mariner planetary data (via tape)• FAA (Federal Aviation Administra-

tion) service-A and C conventional

meteorological data (75 baud; 1 baud~ 1 bs−1)

• computer programs• National Weather Service Radar real-

time data (1200 Hz)• hand-digitized base maps

The ability to accept such a widevariety of data input was done by build-ing one-of-a-kind hardware interfacingthat would take in the data stream andmake it available on the McIDAS system.Various data fusion applications wereimplemented to facilitate ground truthand advanced product generation (Fig. 5).

McIDAS offered a variety of outputformats, many similar to the output of theWINDCO system (Young 1972, personal

communication). All were “TV compatible,” or com-pliant with the National Television Standards Com-mittee. The demand for hardcopy output was addressedin this and subsequent generations, but was always sec-ond or third priority to new interactive capabilities:

• cloud displacement vectors for selected clouds;• analysis on the wind data (vectors, streamlines,

vorticity, divergence);• cloud temperatures, brightness, type, height, and

size information;

FIG. 4. Example of wind vectors.

FIG. 5. An ATS-III visible satellite image centered at 25.8°N and 80.2°W dis-played on McIDAS with overlaid radar data, hand traced from a previous display.(Radar data courtesy of W. L. Woudley, NOAA’s Experimental MeteorologyLaboratory.)


• statistics including means, distributions, and gra-dients of temperatures and brightness;

• false-color enhancements;• time sequences of aligned data;• remapped data;• conventional data overlaid on satellite/image data;

and• data for use by other computer systems.

A command line language was developed to allowusers to control the data display, the analysis of thedata, and the flow of data. This language consisted ofan action, or command, with an argument or param-eter list. For example, “YK T 500 1200 USA” wouldgenerate a graphic overlay of the 500-mb temperatureanalysis from 1200 UTC data covering the contigu-ous 48 states. The command “XK T 1200 USA” woulddo the same type of analysis using 1200 UTC surfacedata. The McIDAS system has continued to use thislanguage through all of its generations. McIDAS pro-vided an important interface between the user, thecomputer, and the database. True to its name, the“man” interacts with the computer to access data on thesystem. Professor Suomi often noted that accessing thedata streams available even to the first generation ofMcIDAS was like “drinking from a fire hydrant.”McIDAS allowed users to manage this untamed flowof data and derive the needed information from it.

c. ImpactsThe system integration was complete by June

1972. Fine-tuning for performance and complete func-tionality for winds continued for the next four months.During the same period, T. Whittaker designed andimplemented the ingest, storage, and display ofconventional surface and upper-air observations(Whittaker and Dedecker 1977). Local Weather Bu-reau regional radar images were ingested as well. InOctober 1973, a live video feed to the WHA-TV stateeducational television network was put in place(Young 1997, personal communication). A dailyweather program covering current conditions and fore-casts using the McIDAS feed was broadcast from theWHA studios. This was the first real-time computer-generated animation used in broadcast meteorology.

Over the next few years, a number of projects sup-ported the development of McIDAS. Image filteringtools were developed to support the processing of earlyLandsat data. The real-time ingestor for the Geosta-tionary Operational Environmental Satellite (GOES)was implemented in 1976. Tools for cloud height

measurement, a macro language, and satellite imagerenavigation were developed to support the data sys-tems tests in preparation for the First Global Atmo-spheric Research Program (GARP) Global Experiment(FGGE). Statistical image analysis tools were devel-oped to facilitate research on rain-rate estimation fromsatellite and radar data.

User pressure became so high that scientists werescheduled on the workstation around the clock.Therefore, in 1975, the analog disk output was splitto support two simultaneous workstations (then calledterminals). Later a digital drum was added as a framestorage device and after its failure, a digital disk wasused in the third workstation. In 1976, the system wasmoved to the University of Wisconsin—Milwaukee tobecome the first remote workstation supported by a pro-cessor in Madison. In this same period, external inter-est in McIDAS grew as it became better known. The U.S.Air Force Cambridge Research Laboratory receivedthe first export of this early version of McIDAS in 1975.

The implementation of the system as a collectionof tools, without the usual analysis of system require-ments and top-down design, came to be known asbottom-up design. McIDAS would not have been pos-sible without this type of design approach, because re-quirements continuously evolved. The implementersand users learned about interactive data processingthrough iterative development: as the implementerpresented new functionality to the users, users pro-vided feedback and additional suggestions to theimplementers.

4. Second-generation McIDAS: Thedistributed Harris System (1978–83)

a. MotivationIn 1977, work on satellite sounder science began

in preparation for the new geostationary sounder to belaunched on GOES-4 in 1979. A group of NOAA em-ployees under the leadership of Dr. W. Smith movedto Madison to develop the processing capabilities forthe Visible Infrared Spin Scan Radiometer (VISSR)Atmospheric Sounder (VAS). The McIDAS team hadto adapt to a new class of users, scientists who neededto develop applications as well as use the tool kit. Thesystem was expanded to include new tools for multi-spectral analysis.

The workload from the number of users, the satel-lite sounder processing, and the support requirementsfor the FGGE, which was to start in November 1978,

276 Vol. 80, No. 2, February 1999

required significant changes to McIDAS (Suomi et al.1977). This prompted the development of the secondgeneration of the McIDAS system.

b. Design and capabilitiesThe transition from the first to the second genera-

tion of McIDAS was primarily a move from a central-ized to a distributed processing system. The system,implemented in 1978, consisted of eight Harris/6 com-puters connected with a high-performance, SSEC-builtnetwork designed by W. Hibbard. (The network con-nections were known as “burn lines” since they wereconsidered very fast at the time.) Two computers(named Mom and Dad) were the database build sys-tems and the remaining six (named Abe, Liz, Sue,Ken, Eva, and Rex) were applications processorssupporting as many as 18 workstations. The Harris6024/6 computers had 144- or 192-kbyte memory,20–200 nanoseconds per memory fetch or store, and300-MB digital disk drives (on the database build sys-tems) or 80-MB digital disk drives (on the applicationprocessors). All computers had synchronous and asyn-chronous interfaces and available 9-track, 1600-bpitape drives. As in the first generation, applicationswere written in FORTRAN, and system-level pro-grams were written in Assembler. Late in the secondgeneration, the system was also running on an Amdahlcomputer system at the West German Space Agency.By 1978, portions of the system had been used on 16-bitminicomputers with some difficulty (Haig 1978).

The system continued to support the same activi-ties as the first generation of the McIDAS, but ingest-

ing satellite data became more important and difficultin the new system. For example, polar-orbiting satel-lite imagery was added to the data sources. [Polar Or-biting Environmental Satellite (POES), represented bythe Television and Infrared Observing Satellite(TIROS) series of spacecraft, is designated as “NOAA-##”, where ## is the consecutive number of the space-craft in the series.] This presented new and uniquechallenges across most aspects of McIDAS fromgeostationary data because of the changing perspec-tive and the continuous image strip nature of polar-orbiting data versus the discrete image and constantperspective nature of geostationary data.

Input sources for McIDAS in this generation, alongwith their nominal data rate, included the following:

• VISSR GOES/SMS satellite data (1.7 MB s−1),• VAS data from the new GOES sounder system

(2.11 MB s−1),• TIROS-N data via two Very High Frequency an-

tennas (1200 pixels s−1),• National Center for Atmospheric Research (NCAR)

model data via dial-up access (4800 baud),• NMC model data via dial-up access (9600 baud),• conventional data via the FAA 604 circuit (1200

baud), and• radar data via direct dial-up (1200 baud).

In addition to the first-generation output formats,McIDAS could now also send commands to the newGOES VAS satellites via dial-up access to GoddardSpace Flight Center and the groundstation at WallopsIsland, Virginia.

Archiving GOES/SMS data continued formallyduring this era, using three videocassette recorders tostore the raw signal from the VISSR and VAS datastreams (Fig. 6). The system used four 9-track tapedrives for archiving and restoring other data types toand from the system. During this era, the McIDASsystem saw some unique applications working with avariety of nonmeteorological data. The X-ray scans,aircraft thermal mappings, Landsat imagery, imagesof crystal structures, biochemical protein scans(Fig. 7), and planetary applications (imagery from theMariner Missions to Venus and Voyager missions,Fig. 8) were all tasks the McIDAS system took onduring this period (Haig 1978).

A typical McIDAS terminal at the time included acommand cathode-ray tube (CRT), keyboard, joystick,printer, color CRT with solid state image display (us-ing Intel technology), bisync communications to the

FIG. 6. Full Earth image from VISSR data from GOES-East(SMS II).


Harris computers, and solid state memory or analogdisk refresh (see Figs. 9 and 10).

c. ImpactsThis generation of McIDAS saw the expanded use

of remote terminals. During the first generation ofMcIDAS a remote terminal was set up at Universityof Wisconsin—Milwaukee and was connected toSSEC’s McIDAS system via a 9600-baud line. Later,a remote terminal was set up at the National Environ-mental Satellite Service facility in Kansas City. By1978, as many as 14 terminals, at the University ofWisconsin—Madison or else-where, were connected.

The McIDAS system wasused during this period for thetesting and acceptance of the newGOES VAS system. This effortincluded the Cooperative Insti-tute for Satellite Studies, SSEC,and NOAA. In addition, McIDASwas increasingly used in broad-casting, education, private sector,and government applications.

In November 1978, the FGGEproject started processing clouddrift winds for use in global dataanalysis. Cloud drift wind infor-mation was generated four timesper day from the five geostation-ary satellites. Production of thesewind datasets continued throughDecember 1979 and demon-

strated that McIDAS could sustain operational com-mitments for long periods of time. Figure 11 showsAsia in a satellite image taken during the FGGE year.

A major turning point in the evolution of McIDASoccurred on 15 July 1979. Congress conducted an in-vestigation into the loss of life in a Wichita Falls,Texas, tornado that spring. One portion of the inves-tigation was held at the SSEC. As a result, Congressdirected NASA to install McIDAS at NSSFC. Thesystem was installed in January 1981 (Norton 1981).Improved forecasts by the National Weather Service’s(NWS) Severe Local Storms unit soon demonstrated

FIG. 7. Scan of biochemical protein spots.FIG. 8. Voyager-II image of Jupiter, displayed on McIDAS.

FIG. 9. The components of second-generation McIDAS.

278 Vol. 80, No. 2, February 1999

the value of a system like McIDAS for real-time dataaccess and decision making (Norton 1985).

5. Third-generation McIDAS: Themeteorologist’s dream machine(1983–96)

a. MotivationThe third generation of McIDAS had its roots in

the installation of the McIDAS software on the

Amdahl computer at the West German Space Agencyin 1976, and later at NASA/Goddard Laboratory forAtmospheric Science in 1979. In addition, while thevisibility of operational projects raised the interest ofother operational facilities, all were concerned with thedependence on Harris computers, which were toosmall for their operations. A number of these poten-tial sites, including the People’s Republic of China,requested a port of McIDAS to an IBM 370–basedmainframe architecture. China funded the porting ef-fort and in 1983 the port to an IBM 4331 was com-pleted and delivered. The SSEC McIDAS facility waschanged to an IBM 4341 (five times faster than theHarris/6) by June 1983 and the Harris-based systemswere retired. Shortly after the implementation of theIBM-based system, it became obvious that this archi-tecture had lost the interactivity that was so valued inthe Harris-based architecture. This generation ofMcIDAS saw two eras of the IBM mainframe system;the “dumb” terminal era eventually gave way to the“smart” terminal era (personal computer or PC-basedterminals or workstations).

In June 1977, NSF sponsored a workshop on In-teractive Video Systems for Atmospheric Sciences atthe University of Wisconsin—Madison (Haig 1977).Representatives from about 80 colleges and universi-ties attended. A variety of applications of McIDAS foreducation were presented, as were applications fromother schools developing interactive systems. Thisworkshop ultimately led to the formation of the Unidata

program in the early 1980s under theUniversity Corporation for AtmosphericResearch (UCAR), which provides ap-plications software and near–real timedata distribution to UCAR members. Theneed for a low-cost, stand-alone, interac-tive terminal was identified following theworkshop. This was further motivationfor the development of a smart terminal.

b. Design and capabilities1) THE DUMB TERMINAL ERA

Third-generation McIDAS saw a re-turn to a centralized processing architec-ture (Suomi et al. 1983). While theterminal hardware included a computerchip, terminals were limited to control-ling the display and sending commandsto the mainframe. By this time, theSSEC-built terminals used solid statememory to refresh the display. The re-

FIG. 10. Users at second-generation McIDAS terminals.

FIG. 11. Asia, centering on China, during FGGE year 1978–79 from Japan’sGeostationary Meteorological Satellite. During FGGE, SSEC was asked to archivegeostationary data globally.


fresh memory was organized tooutput three frames simulta-neously. Two frames containedimages and the third, graphics.This configuration facilitatedtwo-frame, interactive imagecombination with an eight-levelgraphic overlay. With sufficientmemory, the workstations couldanimate up to 128 image frameswith 64 frames of graphics.Image and graphic frames couldbe written, erased, color en-hanced, and animated indepen-dently. The first implementationof McIDAS on the IBM main-frames was under OperatingSystem/Virtual System 1 (OS/VS1). Later, this was upgradedto IBM’s Multiple Virtual Stor-age (MVS) operating system.Eventually, this platform of McIDAS would be knownas McIDAS-MVS (Young and Fox 1994). McIDASterminals changed little from the second generation tothe early portion of the third generation of McIDAS(see Figs. 12 and 13). This new system had the fol-lowing hardware and software components:

Hardware• data-acquisition hardware• general purpose computer• local and remote terminals

Software• computer operating system• McIDAS control program• application and analysis programs

One of the advances of the software portion of thesystem was the development by M. Barrett of meteo-rological data files for storing conventional and otherancillary data, such as satellite sounding retrievals. Healso developed, along with J. Benson, the standardargument-fetching system that allowed an interfacebetween user applications and the subsystem ofMcIDAS. During this generation of McIDAS, assem-bler language was used for system programming andFORTRAN was used for the applications.

2) THE SMART TERMINAL (WORKSTATION) ERA

The McIDAS team quickly learned that the dumbterminal architecture was incapable of the tight user–

computer interactivity that was so valued in theHarris-based architecture. At the same time, PCs werebecoming more widely available. In 1984, develop-ment started on a stand-alone, IBM PC-based versionof McIDAS. Only some of the user applications avail-able in mainframe McIDAS were available in thisversion, called PC-McIDAS. Assembler with someFORTRAN was used to develop this version. The PC-McIDAS ran under the DOS operating system andused an Enhanced Graphics Array (EGA) or VideoGraphics Array (VGA) display. These workstationswere made available to users in 1985 (Ide 1987) andushered in the “smart workstation” era of McIDAS.

FIG. 12. The components of third generation of McIDAS.

FIG. 13. A user at a third-generation McIDAS terminal.

280 Vol. 80, No. 2, February 1999

In 1987, work started on the IBM Operating Sys-tem/2 (OS/2) version of McIDAS. This versionworked with the SSEC-built display hardware (Dengelet al. 1989) and the mainframe. This type of terminal,or workstation, allowed users to process data on eitherthe mainframe or the PC. This version used FOR-TRAN with minor use of Assembler. The first IBMmainframe smart workstation systems were deliveredin 1988 to the National Environmental, Satellite, Dataand Information Service; NMC (now NCEP); NHC(now TPC); and NSSFC (now SPC). While the PCworkstations were used to process some data, thesesystems still used the mainframe to build the databasesand run the major applications.

During the smart workstation era, McIDAS work-stations changed dramatically. In addition to the newEGA/VGA display, the tower of hardware that drovethe existing McIDAS color CRT displays was alsoavailable. This new workstation configuration wasknown as the tower workstation. The tower hardwarewas driven by the PC with 3-bit encoding to get 6-bitdisplay for imagery. These workstations saw over 15years of service worldwide. Some tower displays maystill be in use today.

With the demand for a greater frame space to dis-play more imagery at larger sizes, the WIDEWORDWorkstation was developed using a more compact setof hardware than in the tower. This display worked asthe tower workstations did, along with the McIDAS-OS2. This workstation was the first true 8-bit displayfor imagery.

Two additional display systems were developed aswell: the SSEC Display Adapter (SDA) and, later, thePresentation Manager (PM).

The SDA was the functional combination of theWIDEWORD and VGA displays onto one card in anexpansion slot on the Personal System/2 (PS/2) com-puters used to run McIDAS-OS2. With a second Su-per Video Graphics Array monitor, the display hadcapabilities similar to the WIDEWORD, for lessmoney.

The PM display mode used the OS/2 window man-ager (named “Presentation Manager” by IBM) to showthe text and image/graphics information in windowson the display’s normal desktop, right along with anyother applications. It was designed to be run withoutspecial hardware and thus overcame the restriction thatMcIDAS-OS2 could only be run on IBM PCs.

The growing use of high-performance, UNIX-based workstations by the research community and thedecision by segments of the federal government to use

UNIX-based data systems, drove the McIDAS teamto start yet another port of McIDAS. In 1990 work be-gan on porting the system to UNIX, with the first re-lease one year later (Santek 1991). FORTRAN usecontinued; however, system level programming wasdone using C. This McIDAS-X system initially ran onIBM RS/6000 on Silicon Graphics workstations, andlater expanded to Sun in 1992 and Hewlett Packard in1993.

Inability to repair the older hardware and availabil-ity of new McIDAS-X workstations in the mid-1990sbrought the sunset of the older display systems. In June1999, the PM display will be retired.

c. ImpactsIn January 1985, the U.S. Air Force Launch Sup-

port Center for the Eastern Test Range and theKennedy Space Center were the first recipients of thisnew generation of McIDAS for use with the spaceshuttle program. The Unidata program used the newPC-McIDAS, and, later, McIDAS-OS2 and McIDAS-X workstations. Up to 120 colleges and universitiesused McIDAS for daily educational and research ac-tivities. Other sites worldwide used this generation ofMcIDAS, including the Australian Bureau of Meteo-rology, the National Meteorological Institute in Spain,and NASA/Marshall Space Flight Center.

Pressure from the operational user community wasgrowing to stabilize the software, test and documentit more thoroughly, and establish a formal release pro-cess. In response, the McIDAS Users Group (MUG)was formed. The MUG provided software configura-tion management, organized the development of thedocumentation, and provided a help desk for regis-tered users. The user community appreciated the im-provements in McIDAS and its support. However, thedesire for professional software testing and documen-tation slowed the delivery of new capabilities and in-creased the cost of development. Gradually, membersof the research community felt disenfranchised by thisprocess because they saw McIDAS developers as un-responsive to their needs and under the primary influ-ence of the operational user community.

The impact of the port to UNIX was significant inthree ways. First, McIDAS became available to anexpanding segment of the research community, and theuser community grew. Second, the cost of support ofMcIDAS grew disproportionately because the lack ofstandards in the UNIX systems required far more test-ing and tailoring to the supported UNIX environments.Third, UNIX-based systems had enough processing


power to support all McIDAS applica-tions, reducing the need for an expensivemainframe computer. The increasedpower made possible more complex ap-plications (Figs. 14 and 15).

6. Fourth-generation McIDAS:A return to a distributedsystem (1996–present)

a. MotivationThe port of McIDAS to UNIX en-

abled the capability to move back to adistributed system. The move was moti-vated by the need to remove mainframe-processing elements from McIDAS. In1993, a rewrite of McIDAS was started.The functionality of old applications wasanalyzed and redistributed to a new setof applications. These new applicationsused a client-server data access conceptdeveloped by the McIDAS team calledAbstract Distributed Data Environment(ADDE) (Taylor et al. 1995). In 1996,the system-level interface portion ofMcIDAS-X was recoded in an effort toeliminate UNIX system dependencies.

b. Design and capabilitiesThe fourth generation of McIDAS is

a return to the distributed environmentpreviously seen in the second generation.The workstation versions of McIDAS(McIDAS-X and -OS2) from the thirdgeneration continued in the fourth. Thebiggest change was the sunset of theMcIDAS-MVS mainframe software.Without this central system collecting, ingesting, de-coding, and storing data, alternatives had to be found.The first step was developing workstations dedicatedto the ingest of satellite and conventional data. Laterefforts focused on developing a client–server proto-col for distributing the data.

In the meantime, the McIDAS-X system’s inter-face with the UNIX operating system was re-evalu-ated and recoded to allow for a more seamlessimplementation of McIDAS-X across the variety ofUNIX platforms. Although many of the applicationsare written in FORTRAN, C programming languageis used for both system-level and application-level

programs, to take advantage of UNIX capabilities.This work resulted in a system that allowed for betterintegration of the McIDAS environment with UNIXapplications, scripts, and other elements. It was alsothe foundation for software versions of hardware func-tions found in the previous generation of McIDASsuch as independent and bit-plane graphics and brief-ing windows.

The McIDAS-XCD (conventional data) platformwas the first success of the fourth generation. Builtahead of the full development of the fourth generation,this system was able to keep up with the rapid changesof conventional data circuit ingestion, decoding, and

FIG. 14. Storms over Texas showing county outlines.

FIG. 15. Streamline analysis over a classic spring storm.

282 Vol. 80, No. 2, February 1999

filing. Later in the fourth generation, this same sys-tem was able to handle both the NOAA Family ofServices circuits as well as the broadcast calledNOAAPORT, the NWS transmission system that sup-plies data to NOAA’s AWIPS.

The satellite ingesting effort was a two-step pro-cess. The first development was the McIDAS-XSD(Satellite Data) system. This system combined Hugheshardware and SSEC-developed software and protocol.It was developed initially for data from the JapaneseGeostationary Meteorological Satellite, and then, chro-nologically, for GOES, METEOSAT (the EuropeanMeteorological Satellite Organisation’s Meteorologi-cal Satellite), POES, and Defense MeteorologicalSatellite Program (DMSP) signals. Competing ingestsystems were CTA, Global Imaging, or site-developedsystems. McIDAS-XSD did not live up to expectations

and proved to be too expensiveto maintain and evolve. Thesecond effort was the SSECDesktop Ingestor (SDI). Thissystem, with only an expansioncard in a desktop PC running theSolaris x86 operating system,is able to inexpensively andcleanly ingest data from GOES,Meteosat, POES, and DMSPsatellites. In addition, SDI is ableto ingest conventional data(NOAAPORT) for use withMcIDAS-XCD.

With sources of data avail-able on mainframes and work-stations at the end of the thirdgeneration, the return to a dis-tributed environment began with

the development of the Data Distribution Environment(DDE). This demonstration package was a proof ofconcept for a client–server method to access data overlocal- and wide-area networks. In the previous genera-tion, data access required logging on to mainframeMcIDAS and downloading data using a private pro-tocol between McIDAS-MVS and McIDAS-X or-OS2. Now, data were available from a variety of serv-ers and to a variety of clients. As TransmissionControl Protocol/Internet Protocol (TCP/IP) andInternet communications became commonplace, theuse of dial-up communications or direct-wire commu-nications connecting workstations and mainframesdeclined.

After the success of DDE, the implemented ver-sion of this system, termed ADDE, was developed forMcIDAS-X and McIDAS-OS2. McIDAS-X can beboth server and client, while McIDAS-OS2 remainsonly a client in this system. With the sunset ofMcIDAS-OS2 platform and the important need to uti-lize a low-cost, low-end system, future work willstrive toward running the McIDAS-X software onIntel-based platforms (e.g., Solaris x86, Windows-NTwith Interix software, etc.). Workstations such asMcIDAS-XCD run McIDAS-X with the -XCD packbuilt on top. SDI units can also function as ADDEservers (see Figs. 16 and 17).

c. ImpactsFollowing the port to UNIX, the conversion to a

distributed system offered the hope of removing main-frames from the McIDAS architecture altogether, tak-FIG. 17. A user at a fourth-generation McIDAS workstation.

FIG. 16. The components of fourth-generation McIDAS.


ing advantage of advances in networking to reduce theoverall cost of a system. By 1998, most McIDAS-MVS systems were decommissioned.

With the tight budget climate of the late 1990s,development funds for completing ADDE applica-tions came slowly compared to user demand. Also,with focus on keeping existing functionality in thenew environment, few new scientific applicationswere added in the initial implementation of the fourth-generation McIDAS. As the fourth generation ma-tures, renewed emphasis will be placed on scientificapplications.

7. Conclusions

For 25 years, McIDAS has provided the meteoro-logical community with a standard for data access andtools for data analysis and data fusion, using a systemconcept that embraces a broad spectrum of users.Despite weaknesses, it continues to lead the commu-nity as a highly integrated system of applications tai-lored to scientific analysis of diverse databases (seeFigs. 18 and 19).

We have learned five principal lessons over the 25years of McIDAS evolution.

1) In an environment of change and evolution, thebottom-up design approach produces a systemmore satisfactory to users than a top-down ap-proach. Build basic tools in an environment thatallows users freedom to combine tools to accom-plish their purpose. Build tools to explore the sci-ence in the data, with little or no assumptions aboutthe user or the user’s environment.

2) Requirements are not our final goal, user satisfac-tion is. Therefore, include users on the develop-ment team. Use specific user requests to help un-derstand the general problems or issues. Design thegeneral solution to encompass the specific request.The designers must understand user problems bet-ter than the users do. Implement the solution in anevolutionary fashion. Evolve from something thatworks to something that works better based onfeedback from users at each revision.

3) Many research users will not be satisfied with thetool kit that is provided. They will want to modifythe source code to add algorithms specific to theirinvestigation. Therefore, the tool-kit software shouldbe organized into modules that are logical, easy-to-understand, extensible, and well documented.

4) A tool-kit user interface and documentation shouldmake it easy for the user to find out how to accom-plish a task.

5) The tool kit should consist of independent pro-grams, based on the science the users need, thathave been made a component of the system, anditerated until they work to the user’s satisfaction.

McIDAS will continue to evolve, adding capabili-ties and new data sources while staying current withnew technology. The McIDAS team looks forward tothe new opportunities and challenges presented by theEarth Observing System satellite series, particularlyby Moderate Resolution Imaging Spectroradiometerand Atmospheric Infrared Sounder instruments. Theexplosive development of the Internet will affect theevolution of the fifth generation of McIDAS.

FIG. 18. An infrared (11.0 µm) composite image in Mollweideprojection of GOES-8, GOES-9, GMS-5, METEOSAT-6, NOAA-12,and NOAA-14 satellite data with overlaid contours of forecast500-hPa NCEP Medium Range Forecast Model (MRF) geopoten-tial heights (m) made on McIDAS for 11 May 1998 at 1800 UTC.

FIG. 19. A global Mollweide montage image showing cloudcover from GOES, GMS, METEOSAT, and NOAA satellites aswell as sea surface and land surface temperatures from 14 April1995 at 1200 UTC.

284 Vol. 80, No. 2, February 1999

Acknowledgments. The legacy of Professor Suomi continuesin the evolution and contributions to McIDAS. The spirit of hisdynamic leadership continues to inspire the McIDAS team. Overthe past 28 years, the development of WINDCO and McIDAS hasbeen supported by many organizations. Of particular note areNASA, NOAA, NSF, and the Department of Defense. Thanks tothe McIDAS team over the years, especially Frederick Mosherand William Hibbard for their valuable insights and innovativesolutions to architectural problems. And finally, thanks to themany McIDAS users who have helped improve McIDAS throughtheir ideas and feedback.

References

Dengel, R., J. Benson, R. Lord, J. Plutchak, D. Santek, S. Shirshac,T. Whittaker, and J. T. Young, 1989: An OS/2 implementa-tion of PC-based McIDAS workstation. Preprints, Fifth Int.Conf. on Interactive Information and Processing Systems forMeteorology Oceanography and Hydrology, Anaheim, CA,Amer. Meteor. Soc., 41–43.

Haig, T. O., 1977: The McIDAS system. Proc. Workshop on In-teractive Video Displays for Atmospheric Studies, Madison,WI, National Science Foundation, Space Science and Engi-neering Center, University of Wisconsin—Madison, 171–178.

Hasler, A. F., J. Kornfield, and L. Jensen, 1968: A “blink” mea-suring technique for quantitative measurement of cloudmotions from ATS-1 spin scan camera pictures. Weather inMotion, Space Science.

Ide, J., 1987: PC-based McIDAS. Preprints, Fourth Int. Conf. onInteractive Information and Processing Systems for Meteorol-ogy Oceanography and Hydrology, New Orleans, LA, Amer.Meteor. Soc., 118–122.

Krauss, R., D. Phillip, T. Schwalenberg, and E. Smith, 1972:McIDAS—An interim report on the development of the Man-computer Interactive Data Access System. Space Science andEngineering Center, University of Wisconsin—Madison,36 pp. [Available from Space Science Engineering Center,University of Wisconsin—Madison, 1225 W. Dayton St.,Madison, WI 53706.]

McQuain, R. H., 1967: ATS-1 camera experiment successful. Bull.Amer. Meteor. Soc., 49, 74–75.

Norton, C., 1985: Centralized storm information system (CSIS).NASA Final Report NASA CR-175898, Space Science andEngineering Center, University of Wisconsin—Madison.

Phillips, D., and E. Smith, 1974: Geostationary satellite naviga-tion model. Studies of the Atmosphere Using AerospaceProbes. Annual Rep. NG-70-72, Space Science and Engineer-ing Center, University of Wisconsin—Madison, 253–271.[Available from Space Science Engineering Center, Univer-

sity of Wisconsin—Madison, 1225 W. Dayton St., Madison,WI 53706.]

Santek, D., T. Whittaker, J. T. Young, and W. Hibbard, 1991: Theimplementation plan for McIDAS-AIX. Preprints, Seventh Int.Conf. on Interactive Information and Processing Systems forMeteorology Oceanography and Hydrology, Anaheim, CA,Amer. Meteor. Soc., 177–179.

Schwalenberg, T. W., 1972: The WINDCO display system. Stud-ies of the atmosphere using aerospace probes. NOAA AnnualRep., Space Science and Engineering Center, University ofWisconsin—Madison, 1–15. [Available from Space ScienceEngineering Center, University of Wisconsin—Madison, 1225W. Dayton St., Madison, WI 53706.]

Smith, E., and D. Phillips, 1972: WINDCO: An interactive sys-tem for obtaining accurate cloud motions from geostationarysatellite spin scan camera pictures: Measurements from satel-lite platforms. Scientific Rep. AAS5-115042, Space Scienceand Engineering Center, University of Wisconsin—Madison,1–51. [Available from Space Science Engineering Center, Uni-versity of Wisconsin—Madison, 1225 W. Dayton St., Madi-son, WI 53706.]

Suomi, V. E., 1968: A color view of planet Earth. Bull. Amer.Meteor. Soc., 49, 74.

——, and R. J. Parent, 1967: ATS spin scan cloud camera andpre-launch calibration procedures: Studies in atmospheric en-ergetics based on aerospace probings. Annual Rep. WBG-27,Department of Meteorology, University of Wisconsin—Madi-son, 1–40. [Available from Space Science Engineering Cen-ter, University of Wisconsin—Madison, 1225 W. Dayton St.,Madison, WI 53706.]

——, F. R. Mosher, and D. P. Wylie, 1977: Final report on phaseI of development of satellite image processing techniques forthe FGGE. Space Science and Engineering Center, Universityof Wisconsin—Madison, 36 pp. [Available from Space Sci-ence Engineering Center, University of Wisconsin—Madison,1225 W. Dayton St., Madison, WI 53706.]

——, R. J. Fox, S. S. Limaye, and W. L. Smith, 1983: McIDASIII: A modern interactive data access and analysis system. J.Climate Appl. Meteor., 22, 766–778.

Taylor, W., J. Benson, T. Whittaker, and J. Rueden, 1995: Seam-less access to local and distributed data. Preprints, 11th Int.Conf. on Interactive Information and Processing Systems forMeteorology Oceanography and Hydrology, Dallas, TX,Amer. Meteor. Soc., 349–352.

Whittaker, T. M., and R. Dedecker, 1977: Processing and displayof satellite and conventional meteorological data for the class-room. Paper presented at 57th Annual Meeting of the Amer.Meteor. Soc., 17–20 January, Tuscon, Arizona.

Young, J. T., and R. J. Fox, 1994: Future Plans for McIDASWorkstations. Fourth Workshop on Meteorology OperationalSystems, Reading, UK, ECMWF, 240–247 pp.

Documents

The Man computer Interactive Data Access System: 25 Years of