f1 an j) - usda

39
li\ bfOf{Gt f1 A N J).S CJ-l fl K 15 JSC.18500 Co (JY --z- AgRIST ARS MINISYMPOSIUM DECEMBER 1-2, 1982 t. ~. Inventory Surveys Through ~ AgRISTARS ~ ~ 0 ~ .. ~ ... ~ I~ ·V····:~ ~. ~... ~.~J ' .. i: CD '. I~ :~>/~~~ Ih -(:{USDA-(:{NASA f:. USDC f:. USDI f:. AID -(:{ j NI\S/\ National Aeronautics and Space Administration Lyndon B. Johnson Space Center Houston. Texas77058

Upload: others

Post on 15-Mar-2022

1 views

Category:

Documents


0 download

TRANSCRIPT

li\

bfOf{Gt f1 A N J).S CJ-l fl K 15JSC.18500 Co (JY--z-

AgRIST ARSMINISYMPOSIUMDECEMBER 1-2, 1982

t .

~.

Inventory Surveys Through

~ AgRISTARS ~~ 0~ .. ~ ... ~I~ ·V····:~ ~.~...~.~J

' .. i : CD

'. I ~

:~>/~~~Ih-(:{USDA -(:{NASA f:. USDC f:. USDI f:.AID -(:{ j

NI\S/\National Aeronautics andSpace Administration

Lyndon B. Johnson Space CenterHouston. Texas 77058

FORWARD

Welcome to the AgRISTARS Mini Symposium.

AgRISTARS is a cooperative research effort led by the U.S. Department ofAgriculture and supported by the National Aeronautics and Space Administration,the National Oceanic and Atmospheric Administration of the U.S. Department ofCOll1nerce,the U.S. Department of the Interior, and the Agency for InternationalDevelopment of the U.S. Department of State.The goal of AgRISTARS is to determine the usefulness, cost, and extent to whichaerospace remote sensing data can be integrated into existing or future USDAsystems to improve the objectivity, reliability, timeliness, and adequacy ofinformation required to carry out USDA missions.The AgRISTARS program has been underway for three years now and a great deal offine technical work has been accomplished. While this work has been documentedin AgRISTARS reports and, to some extent, reported in literature and othersymposia, a need was perceived to provide an opportunity for a general technicalinterchange. This Mini Symposium addresses that need. The papers you will behearing cover a good cross-section of the work going on in the AgRISTARSprojects. There are also some very informative posters available in the lobbyof building 30 and in building 17, room 2026. We hope that you will take timeto visit these displays.

ii

CONTENTS

AGENDA

AgRISTARS MINI-SYMPOSIUM

WEDNE~L.AY, DECEMBER 1, 1982 ••••••••••••••••••• viiTHURSDrY, DECEMBER 2, 1982 ••••••••••••••••••• x

ABSTRACTS1-1 CROP Sl-~ESS INDICATOR ftl0DELSFOR LARGE AREA

ASSESSr .NT ••••••••••••••••••••••••••••••••••••••• 11-2 WATER L,~MAGE TO RIPE HARD RED SPRING WHEAT •••••• 11-3 NORMALIZATION OF NOAA AVHRR DATA FOR ANGULARANISTRlPHY •.•..••••••••••••.••••••••••••••••••••• 21-4 USE OF NOAA-6 SATELLITES FOR LAND/WATER

DISCRIt INATION AND FLOOD MONITORING ••••••••••••• 21-5 AVHRR LATA EVALUATION AND INFORMATION CONTENT 21-6 EVALUATION OF A NATIVE VEGETATION MASKING

TECHNI~UE ••••••••••••••••••••••••••••••••••••••• 31-7 FUNCTIUNAL EQUIVALENCE OF SPECTRAL VEGETATIVE

I NO ICES ••••••••••••••••••••••••••••••••••••••••• 32-1 EARLY ~EASON SPRING SMALL GRAINS PROPORTION

EST I MA T I ON ••••••••••••••••••••••••••••••••••••••• 32-2 A CROP AREA ESTIMATOR BASED ON THE CHANGES

IN THE TEMPORAL PROfILE OF A VEGETATIVE INDEX •••• 42-3 UPDATE ON A SYSTEM FOR LARGE AREA CROP

INVENTORY FROM REMOTELY SHJSED 'DATA •••••••••••••• 42-4 THE AGRICULTURAL INFORMATION SYSTEM SIMULATOR:

AN OVERVIEW AND AN APPLICATION ••••••••••••••••••• 52-5 INVESTIGATIONS OF THEMATIC MAPPER DATA

DIMENSIONALITY AND FEATURES USING FIELDSPECTROMETER DATA •••••••••••••••••••••••••••••••• 5

2-6 DEVELOPNENT OF A QUANTITATI VE BASIS FOR FEATUREEXTRACTION IN A VEGETATION MONITORING SYSTEM 6

2-7 AN AUTOMATED SPRING SMALL GRAINS PROPORTIONEST IMA TOR •••••••••••••••••••••••••••••••••••••••• 6

iii

AgRISTARS MINI-SYMPOSIUM

2-8 SM~PLING UNIT SIZE CONSIDERATIONS FOR REDUCINGDATA LOADS IN LARGE AREA CROP INVENTORYINGUSING SATELLITE-BASED DATA ••••••••••••••••••••••• 6

2-9 AN AUTO~IATED APPROACH TO LARGE AREA CROP INVENTORY,BASED ON COLOR AND TOPOLOGY ••••••• ~••~••••••• ~••~ 7

2-10 AUGMENTATION OF LANDSAT MSS DATA BY SEASAT SARIMAGERY FOR AGRICULTURAL INVENTORy ••••••••••••••• 7

2-11 DEVELOPtvlEIH, TEST AND EVALUAJION OF A COMPUTERIZED, .PROCEDURE FOR USING LANDSAT DATA TO ESTIMATESPRING SMALL GRAINS ACREAGE •••••••••••••••••••••• 8

2-12 EXPERIMENTS WITH AN EXPERT-BASED CROP AREAESTIMATION TECHNIQUE FOR CORN AND SOYBEANS ••••••• 8

3-1 LARGE AREA YIELD ESTIMATES DERIVED FROM PLANTSIMULATION MODELS •••••.••••••••••••••••••••••••• 10

3-l APPLICATION OF SATELLITE SPECTRAL DATA INESTIMATING WHEAT YIELDS ••••••••••••••••••••••••• 10

3-3 INTENSITY, HUE, SATURATION (IHS) DISPLAY OFTHREE CHANNEL INPUT ••••••••••••••••••••••••••••• 11

4-1 DEVELOPMENT OF CROP SPECTRAL TEMPORALAPPLICATIONS TO CORN-SOYBEAN SEPARATION ••••••••• 11

4-2 SUPPORTING RESEARCH IN PATTERN RECOGNITION -INTRODUCTION •••••••••••••••••••••••••••••••••••• 11

4-3 ESTIMATION AND IDENTIFICATION OF VEGETATIVECOVER TYPE USING MIXTURE DECOMPOSITION ANDDISTRIBUTION LABELING .•••••••••••••••••••••••••• 12

4-4 THE EFFECTS AND TREATMENT OF IIr-UXTUREIIPIXELS INPROPORTION ESTIMATION OF VEGETATIVE- COVER TYPE •• 13

4-5 A QUICK-LOOK EVALUATION OF THEMATIC MAPPER DATA •• 134-6 MULTITEMPORAL REGISTRATION OF LANDSAT AND AIC

~ULTIDATE DATA •••..•.•....••..•••.••.•..••••••• 144-7 NEASUREMENTS AND SCENE ANALYSIS OF CORN-SOYBEANS

AND S~~LL GRAINS FOR IDENTIFICATION, DEVELOPMENTSTAGE ESTIMATION AND CONDITION ASSESSMENT •••••• 15

4-8 THE USE OF VEGETATIVE CANOPY REFLECTANCE MODELSIN SUPPORT OF AgRISTARS •••••••••••••••••••••••• 16

4-9 SPECTRAL ESTIMATION OF LEAF AREA INDEX AND LIGHTINTERCEPTION BY CROP CANOPIES •••••••••••••••••• 16

;v

AgRISTARS MINI-SYMPOSIUM

4-10 MICROWAVE PROPERTIES OF AGRICULTURAL CROPS ATLONG WAVELENGTHS ••••••••••••••••••••••••••••••• 17

4-11 DEVELOPMENT OF JSC QUICK-LOOK IMAGE PROCESSOR •• 174-12 EVOLUTION OF REMOTE SENSING RESEARCH DATA/SYSTEM

~lD DATA MANAGEMENT •••••••••••••••.•••••••••••• 18

4-13 COMPARISON OF MINIMUM DISTANCE AND MAXIMUMLIKELIHOOD TECHNIQUES FOR PROPORTION ESTIMATION. 19

5-1 IMPROVH1ENT OF ~OISTURE ESTIMATION ACCURACY OFVEGETATION-COVERED SOIL BY COMBINED ACTIVE/PASSIVE MICROWAVE SENSING ••••••••••••••••••••••• 19

5-2 PRELIMINARY RESULTS OF THE COLBY AGRICULTURALSOIL MOISTURE EXPERIMENT ••••••••••••••••••••••• 20

5-3 PASSIVE MICROWAVE SENSING OF SOIL MOISTURE UNDERVEGETATION CANOPIES •••••••••••••••••••••••••••• 20

5-4 A MICROWAVE SYSTEMS APPROACH TO MEASURING ROOTZONE SOIL MOISTURE ••••••••••••••••••••••••••••• 21

5-5 REMOTE SENSING OF SOIL MOISTURE: RECENTAD VANCES •••••••••••••••••••••••••••••• ,•••••••• , 21

6-1 1981AgRISTARS DCLC FOUR STATE PROJECT •••••• ~•• 226-2 KANSAS LAND COVER STUDy •••••••••••••••••••••••• 226-3 THE USE OF LANDSAT FOR COUNTY ESTIMATES OF

CROP AREAS ••••••••••••••••••••••••••••••••••••• 22

6-4 AUTOMATED SEGMENT MATCHING ALGORITHM ••••••••••• 236-5 CLASSIFIER DESIGN FOR REGRESSIONS OF GROUND-

GATHERED WITH COMPUTER CLASSIFIED DATA ••••••••• 236-6 STRATIFICATION OF SAMPLED LAND COVER BY SOILS

FOR LANDSAT-BASED AREA ESTIMATION AND MAPPING •• 246-7 A CORRELATION ANALYYSIS OF PERCENT CANOPY CLOSURE

VS. TMS SPECTRAL RESPONSE FOR SELECTED FORESTSITES IN THE SAN JUAN NATIONAL FOREST •••••••••• 24

6-8 THE SIMULATION OF USDA SEGMENTS, FIELDS, ANDPIXEL SPECTRAL VALUES •••••••••••••••••••••••••• 25

6-~ USDA SMALL AREA CROP ESTIMATIN USING LANDSAT-AND GROUND-DERIVED DATA •••••••••••••••••••••••• 25

v

AgRISTARS MINI-SYMPOSIUM

6-10 EVALUATIUN OF USDA LARGE AREA CROP ESTIMATIONTECHNIQUES ••...•.••.•.......••...............•. 267-1 MAPPING FOREST RESOURCES USING TMS -'ATA •••••••. 267-2 DETECTING FOREST CANOPY CHANGE USING LANDSAT •• 277-3 HIGH ALTITUUE RADAR ASSESSMENT OF DAMAGE CAUSED

BY THE VOLCANIC ERUPTION OF MT. ST. HELENS ••••• 277-4 OKLAHOMA MID-CYCLE TIMBER INVENTORY PILOT TEST. 288-1 INVENTORY OF SOIL CONSERVATION PRACTICES USING

REMOTE SENSING ••••••••••••••••••••••••••••••••• 288-2 BUILDING A BRIDGE BETWEEN REMOTE SENSING AND

HYDROLOGIC MODELS •••••••••••••••••••••••••••••• 288-3 THE STRUCTURING OF A REMOTE SENSING BASED

CONTINUOUS STREAMFLOW MOUEL •••••••••••••••••••• 29

POSTER DISPLAY ••••••••••••••••.••••••••••••••••••••••••••••••••••••••••• 30-33

Poster displays are available in the lobby of building 30 and in building 17,room 2026. We hope that you will take time to visit these displays.

vi

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

1-1CROP STRESS INDICATOR MODELS FOR

LARGE AREA ASSESSMENTTerry w. Taylor, USDA/ARS, and

Frank Ravet, USDA/FASA prime objective of the EW/CCAproject is to provide a capabilityto moni tor and assess cropcondition over large area andrespona in a timely manner. Aseri es of crop stress indi catormoae1s were developed to alert acommodity analyst of potenti a1problem areas. The models functionas fi1ters to e1 imi nate thenecessi ty of devoti ng time andresources to examine in-depth 1argedata streams. Concentrati on isallowed on areas which are alertedas having a high probability ofstress occuri ng. $ubsequentanalyses can then be made usinganc; 11 ary, meteoro1 ogi ca1 andspectra 1 inf.ormation. Fi rs titeration models for wheat (springand winter), maize, sorghum, anasugarbeets have been developed.These are in vari ous stages oftesting and modification and have asimil ar structure consi sti ng ofthree major components: 1) aphenology model, 2) a soil waterbudget model and 3) a hazardroutine for stress definition. Themai ze stress model has been testedwith foreign (USSR) and domestic(Mi ssouri) meteoro1 ogi ca1 andground truth data. These resu1 tswill be presented.

1-2WATER DAMAGE TO RIPE HARD

REO SPRING WHEATArmand Bauer and A. L. Black

USDA/ARS, Mandan, NOGrain damage and/or yie1a loss canoccur when threshing of ripe hardred spri ng wheat is del ayed byrainfall or a combination ofrainfa 11 ana hig h hum id ity • Aknow1 edge of the effect of

1

intensity and persistency of waterrelated environment factors onsprouting is required to develop acapability to predict its impendingonset in ei ther standi ng orwi ndrowed crops.Misting at 1.97 cm/hr increasedspike water concentration 35percentage units the first 10minutes and 0.75 percentage unitsper mi nute thereafter. Wa terimbibition by grain occuredlinearly at a rate of 1.9percentage units per hour (Gordonet. a1 1977). Spi kes saturate atabout 150% water concentraction andgrain at about 100%. Theseequilibrated threshold spike waterconcentrati ons affecti ng sprouti ngis 42 to 45% of oven dried basi s.Water held in the c1umes andinterstitial area of the spikeevaporate more rapid1y than waterimbibed in the grain. Conditionswhi ch depress evaporati on afterrainfall enhance water imbibitionby the grain.Germi nation mechani sms in ripeni nggrain is not triggered until spikewater concentrations is reduced to14% or 1ess. Sprout; ng cou1d notoccur in spikes removed fromsealed-growing wheat which hadhigher than 14% water concentrationbefore wetting to 100% waterconcentration.Post cutti ng suscepti bi 1ity tosprouting differs with cu1tivars asdoes the time interval aftercutting when cu1tivars becomesusceptible. In the most spoutsusceptible cu1tivars 10% sproutingoccured within 7 days after cuttingand in the least susceptible morethan 35 days after cutting.Seeds which have taken up water andthen dried imbibe water faster withSUbsequent wetting and alsoincrease in susceptibility tosprouting. Sprouting is more rapid

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

at 13°C than at 24°C. Field grainyield losses on 10 cm stuob1e were15' and 20% less than 23 and 26 cmstubble respectively as rainfallforcea 1arger porti ons of thewindrow through the taller stubbleto effect soil contact.

1-3NORMALIZATION OF NOAA AYHRR DATA

FOR ANGULAR ANISTROPHY~. H. Duggin, D. Piwinski,

State University of New YorkG. Ryland, Lockheed Engineering and

Management Services Company, Inc.and V. Whi tehead, NASA/JSC

Empirical studies have demonstratedbeyond a doubt that target radiancedependS on view zenith (scan) anglein a systematic manner. Thedependence is mOdulated oyvariations in haze and cloud acrossan image. Similar studies showthat, while angular anistrophy inrecorded radiance can probably beca1 ibrateci out of the data (~hypothesis supported by empiricalstudies) variation in haze andcloud target radiance can causerandom fluctuations in targetradiance. This cannot only causeproblems in the identification of atarget, but a1 so can causeuncertainties in targetdiscrimination.Kesu1ts demonstrating the angularvari ation of target radiance, theform of the dependence on vi ewangle and the effect of unreso1 vedcloud are presented and the courseof present and future researchdiscussed.

2

1-4USE OF NOAA-6 SATELLITES FOR LANDI

WATER DISCRIMINATION ANDFLOOD MONITORING

T. Engman, USDA; D. W. Goss, USDA;and N. C. Horvath,

Lockheed Engineering and ManagementServices Company, Inc.

A tool for monitoring the extent (,fmajor floods has been developedusing data collected by theNOAA-6 Advanced Very HighResolution Radiometer (AVHRR). Abasic understanding of the spectralreturns in AVHRR Channel s 1 and 2for water, soil, and vegetation hasbeen reached usi ng a 1arge numberof NOAA-6 scenes from differentseasons and geographic locations.A 100k-Ijp table classifier wasdeveloped based on analysis of theref1ecti ve channel re1ationshi psfor each surface feature. Theclassifi er automatically separatedland from water and producedclassification maps which wereregistered to a global coordi natesystem. Testing of the classifierwas comp1 eted for a number ofacquisitions, including coverage ofa major flood on the Parana Riverof Argenti na.

1-5AYHRR DATA EVALUATION

AND INFORMATION CONTENTNicholas C. Horvath,

Mike L. MathewsLockheed Engineering and Management

Services Company, Inc.Data from the National Oceanic andAtmospheric Administrationsatellite system (NOAA-6 Satellite)have been analyzed to stUdy thei rnonmeteorologica1 appl ications andcietennine their useful 1imits. Afile of charts, graphs, and tableswas created from the productsgenerated in this study. Analysisof these products indicates thatthe Gray-f.'IcCraryIndex can discern

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

vegetation and associated daily andseasonal changes. It was foundthat the most useful data 1iebetween pixel numbers 400 and 2000on a given scan line. However, thedata were still considered quitevariable, so a new method ofanalysis \'/as performed whichidentifi ed procedures to stabi1izemost (-80%) of the variation in theindex. This has been accompl ishedby exami nati on of the sol arcorrecti on coeffi cients originallyused and the identification ofclouds. The se procedures can beeasily implemented. The metsatsystem seems best suited forproviding large-area analyses ofsurface features on a daily basis.

1-0EVALUATION OF A NATIVE VEGETATION

MASKING TECHNIQUEMaryaret C. Kinsler

Lockheed Engineering and ~anagementServices Company, Inc.

Foreign Crop Condition AssessmentDivision (FCCAD) has utilized acrop masking technique based onAshburn IS Vegetati ve Index (AVI).Early Warni ng Crop Condi ti onAssessment (EW/CCA) chose to usethis technique in the evaluation ofnative vegetation as an indicatorof crop moisture condition. A maskof the range areas (nativevegetati on) was generated for eachof 13 Great Plains LACIE segments.These masks were compared to thedigitized ground truth andaccuracies were computed. Ananalysis of the types of errorsindicates consistency in errorsamong the segffients.

3

1-7 FUNCTIONAL EQUIVALENCE OFSPECTRAL VEGETATIVE INDICESCharles R. Perry, Jr., USDALyle F. Lautenschlager, USDA

Numerous formul as, vegetati veindi ces, have been employed toreduce MSS data to a single numberfor use in assessing ground covercharacteristics such as plant type,plant leaf area, plant stress,total biomass, etc. There has beenmuch discussion in the literatureabout which index i~ superior. Theidea of two vegetative indicesbeing equivalent is formul ated ;nterms functional equivalence: Twovegetati ve indices are taken to beequivalent for making a certain setof decisions, if the decisions madeon the basis of the output of oneindex could have equally well beenmade on the basis of the output ofthe other index. The uti1ity ofthese ideas are demonstrated byshowing that several widely usedindices are equivalent.

2-1EARLY SEASON SPRING SMALL GRAINS

PROPORTION ESTIMATIOND. E. Phinney

Lockheed Engineering and ManagementServices Company, Inc.

M. C. Trichel, NASA/JSC

The value of information from acrop inventory system is determinedby its cost, accuracy, and when inthe growing season the informationbecomes avail able. The InventoryTechnology Development (ITD)proj ect of the Agri cul tural andResources Inventory Surveys ThroughAerospace Remote Sensing(AgRISTARS) program has developedan accurate, automated technologyfor early season estimation ofspring small grains from Landsat

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

MSS data. The technique is basedon a constrained linear model inwhich the observed spectralresponse of a scene is estimated asa linear combination of the majorelements in the scene. Theprocedure was evaluated over 100samp1e se~ents co11ected for cropyears 1975 through 1Y79 in the U.S.Northern Great Plains. Analysis ofthe test results inaicatedperfonnance that was substanti allybetter (n=luO, mean error 1.04%,standard deviation = 7.47%) thanthe automated at-harvest technol-ogies tested during the FY81-82AgR1STAkS Spring Small Grains Pilotexperiments or previous analyst-intensive at-harvest technologies.Further advantages include majorrel axati ons in requi rements formultitemporal registration (none),data storage and transmi ssion, andcomputation which are important inthe design ot smart sensor systems.

2-2A CROP AREA ESTIMATOR BASED ON THECHANGES IN THE TEMPORAL PROFILE OF

A VEGETATIVE INDEXJ. H. Smith and D. B. Ramey

Lockheed Engineering and ManagementServices Company, Inc.

~ost current crop area estimators,based on remotely sensed data,require the classification ofeither fields or picture elements(pixels) into crop types. Thispaper details current research intomethods Which estimate the changein crop proportion in a scene fromone year to another, withoutrequiring that individual fields orpixels be ldbeled as crop types.Instead, pixels are classified asvegetated or not vegetated, and theproportion of vegetated pixels inthe scene is plotted as a functionof ti~e for each of two years. Theplots are smoothed via polynomialregression, and the vertical

4

di stance between the curv€'s(profiles) forms the basis ofprofile change methodology.Results demonstrating thefeasibil ity of using the techniquewill be presented.

2-3UPDATE ON A SYSTEM FOR LARGE AREA

CROP INVENTORY FROM REMOTELYSENSED DATA

T. L. Baker, J. H. Smith,J. T. Ma1in

Lockheed Engineering and ManagementServices Company, Inc.

This paper presents an update onthe state of the art in large areacrop inventory from Landsatmultispectral image data. Inparticular, it describes progresswi th and improvement to theestimation system developed duri ngthe Large Area Crop InventoryExperiment (1975-77) and itsfollow-on Transition Year project(1978-79). Both were jointlysponsored projects of the NationalAeronautics and Space Adminis-tration, the U. S. Department ofAgriculture, and the NationalOceanic and Atmospheric Adminis-tration of the U. S. Department ofCommerce. The improved large areaestimati on technology is a productof and research ~ool for thecurrent joi nt venture of thesethree agencies in conjunction withthe Agency for Internati onalDevelopment of the U.S. Departmentof State and the U.S. Department ofthe Interi or, known as theAgriculture and Resources InventorySurveys Through Aerospace RemoteSensing program. Several candidatetechnologies under development aspossible improvements to the systemare also presented.

AgRISTARSMINI-SYMPOSIUMABSTRACTS

2-4THE AGRICULTURALINFORMATIONSYSTEM

SIMULATOR: AN OVERVIEWANDAN APPLICATION

D. b. Ramey and J. H. SmithLockheed Engi neeri ng and ~lanagement

Services Company, Inc.

A major application of remotelysensed data is the estimation ofagricultural proauction in foreignareas. The evaluation of suchproauction es'timates is difficultoue to the 1ack of ground-veri fi edcrop inventories for foreign areas.This paper ciescribes simulationsoftware designea to test theeffect of sample design, cloudcover, local area estimation biasand variance, and other factors onthe performance of Landsat-based1arge area agricul tural estimatorsin foreign (and domestic) areas.Results of a simulation comparisonof the effects of Landsat 4 orbitalcharacteristics with those ofearl i er Landsats on a speci fi clarge area aggregation systemconfiguration are presentea.

2-5INVESTIGATIONSOF THEMATICMAPPERDATADIMENSIONALITYANDFEATURESUSING FIELD SPECTROMETERDATAEric P. Crist, Richard C. Cicone

Environmental Research Institute oft~ichi gan

Features derived from the four MSSchannel s on Landsat 1, 2, and 3have proven to be of great value indetection of the cover type andcondition of agricultural fields.The Tasseled-Cap Transformation hasbeen widely used to capture thevast majority of data variabilityovera g r i c u1 t u r a 1 s c e ne sin twofeatures with direct physicali nterpretati on: Greenness, whi chcorresponos to the amount of greenvegetation in the field of view,ana orightness, which is related to

5

albedo or soil brightness. Otherfeatures such as the 7/5 ratio andthe Normalized Difference have beenfound useful in particular appli-cations.

The Thematic Mapper on the recently1aunched Landsat-4 i ncl udesdetectors sensitive to differentwave 1ength i nterv al sandI ordifferent bandwi dths than those inthe r~ss. In particul ar, TM Bands1, 5, and 7 are located in regionsof the spectrum unsampl ed by the~ISS (Tt1 Band 6, the thermal band,is not consi de red here si nce itscharacteristics are substantiallydifferent from those of the otherbands and previous sensors). Thesenew banas, and narrower bandwidthsin previously Sampled spectralregi ons, suggest that new featuresmay be available in TM data.Furthermore, the likelihood ofstrong carrel ati ons between atleast some adjacent band pairssuggests that a dimensionality-reducing transformation like theTasseled-Cap Transformation wouldbe of value for the TM as well.

This paper presents the results ofanalyses aimed at determining, bymeans of simulation, thedimensionality of H1 data overagricultural scenes, and theresponse of HI oata features (bandsor combi nati ons of bands) to thephysical characteristics of cropcanopies and soils. Field-measuredcrop spectra, and fi el d and1aboratory-measured soi 1 spectraa re used, along wi th Daveatmospheric model data andprelaunch sensor calibrationi nformati on, to simul ate Landsat-4MSS and TM inband reflectances,top-of-atmosphere radi ances, anddigital image signal counts forwheat, corn, and soybean plots.

AgRISTARS MINI-SYMPOSIUMABSTRACTS

The equi val ency of WI Bands 2, 3,and 4 with MSS Bands 1, 2, and 4,and the resultingIITasse1ed-Cap-equiva1entlltransformations of each, isdemonstratea. Al so shown is theincrease in complexity, in terms ofphysical interpretation, associatedwith the principle componentsderived from all 6 n; bands (again,excl uding the thermal band). Theresponses of TM spectral featuresto canopy characteri sti cs such aspercent cover, percent greenleaves, etc. ana to soilcharacteristics such as particlesize distribution and organicmatter are described. The resu1 tsill ustrate both the conti nui tybetween the MSS and TM sensors, andthe addea benefits potentiallyava i 1ab 1e through use of the ex traTM bands.

2-6DEVELOPMENTOF A QUANTITATIVEBASIS

FOR FEATUREEXTRACTIONIN AVEGETATIVEMONITORINGSYSTEMD. E. Phinney, J. H. Smith,

Lockheed Engineering and ~ianagementServices Company, Inc.

ana M. C. Trichel, NASA/JSC

The aevelopment of an obj ectivemethodology for evaluation ofalternative Landsat data pre-processing options, spectraltransforms, and feature extractionalgorithms is presented. Based onestimates of spectral separabil i tybetween a target cl ass and itsconfusion, analysis of variancetechni ques are used to eval uatepotential design options for largescale vegetation monitoringsystems. Case studi es arepresented for early season spri ngsmall grains separation and forbarley/other spring small grainsseparati on. It is concl uded thatthe basi s for effi ci ent, obj ecti veselection among alternative feature

6

extracti on approaches has beenestablisr:ed.

2-7AN AUTOMATEDSPRING SMALLGRAINS

PROPORTIONESTIMATORT. B. Dennis, R. B. Cate,M. r·. Smyrski, T. C. Baker

Lockheed Engineering and ManagementS( rv ices Company, I nc •

C. V. Naza reIntergraph Corporation

This paper describes a totallyautomated system for estimati ngspring small grains acreages within5- by 6-nautical-mile samplesegments as recorded in Landsatdata. This procedure was developedfor and tested in the fi seal year1981 U.S./Canada Spring SmallGrai ns Pi lot Experiment conduc tedat the Lyndon B. Johnson SpaceCenter as part of the Forei gnCommodity Production Forecastingproject of the Agricul ture andResources Inventory Surveys ThroughAerospace Remote Sensi ng program.The system was derived fromattempts to model some of the humanfunctions performed in the imageanalysis of Landsat data which wasroutinely carried out during theLarge Area Crop InventoryExperimer.t.

2-8SAMPLINGUNIT SIZE CONSIDERATIONSFOR REDUCINGDATALOADS IN LARGEAREACROP INVENTORYINGUSING

SATELLITE-BASED DATACeci 1 Ha11 urn

University of Houstonat Clear Lake CityCharles Perry, USDA

Crop inventorying personnel who usesynoptic information fromsatellite-acquired data mustcontend with large data sets. Thispaper reports on an approach forminimizing these data loads while

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

improving the efficiency of globalcrop area estimates usi n9remotely-sensed, satellite-baseddata. Results of a sampling unitsize investigation are given thatincludes closed-form, modeledallowances for, both, non-samp1 ingand sampling error variances.These model s provide estimates ofthe sampling unit sizes that effectminimal costs. A conservativenon-sampling error variance modelis proposed that is realistic inthe remote sensing environment.Thi s approach, in conj unction wi ththe sampling error variance model,permits a closed-form and viabledetermination of the sampling unitsizes.

2-9AN AUTOMATED APPROACH TO LARGE AREA

CROP INVENTORY BASED ON COLOR ANDTOPOLOGY

H. G. Smith, R. B. Cate, andT. B. Dennis

Lockheed Engineering and ManagementServices Company, Inc.

The concept of an automatedap~roach to crop inventory using acolor space representation (hue,value, and chroma) of mu1tidate MSSdata, combined with a spatialclustering technique for mid-summerestimates of winter and summer cropgroup wall-to-wall inventory hasbeen outlined, implemented, andtested. The results of the conceptfeasibility test compared favorablywith the results achieved in theLarge Area Crop InventoryExperiment (LACIE) Phase III 1977winter wheat inventory for 11 sitesin the same area. Extension of theconcept to di fferent crops anddifferent 1and surface cover typesis underway.

7

2-10AUGMENTATION OF LANDSAT MSS DATA BYSEASAT SAR IMAGERY FOR AGRICULTURAL

INVENTORYDaniel R. Nuesch, Dept. of

Geography, University of Zurich,Switzerland; Quentin A. Holmes,

Applied Intelligent Systems, Inc.,Ann Arbor, Michigan; and M. D.

Metzler, Environmental ResearchInstitute of Michigan, Ann Arbor,

1·1ichi gan

This paper explores the potentialof joint use of Landsat MSS andSEASA TSAR for agri c u1 tura 1inventory. We combi ne infonnati onfrom sensors which respond todifferent crop canopycharacteristics. Landsat MSS is apassi ve sensor which is responsi veto the presence of vegetativebiomass and chlorophyll absorption.SEASAT SAR is an active sensor inthe microwave region which isresponsive to canopy structure andits dielectric constant asdetenni ned by moi sture condi tions.The joint spectral attributes ofthese sensors affords an intriguingview of the agricultural scene.The high resol ution of the SEASATSAR brings with it the possibilityof refined definition of theboundaries of agricultural fields.SEASAT SAR data co11 ected overJasper County, Indiana, wasoptically processed, digitized andregi stered to landsat Segment 844consisting of seven MSSacquisitions. Digital SEASAT radardata was preprocessed using anon-linear isotropic filter whichremoved speck1e noise wi thout lossof spatia1 reso1 ution or spectralinformation as occurs withconventi ona 1 smoothi ng a1gori thms.The process developed resu1 ted inthe creation of two image featuresdubbed IItoneII and IItexturell• Thetexture image was in fact theextracted speckle noise and was

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

found to contain informationpertinent to crop canopyidentification.Results of this investigationrevealed that the finer spatialresol ution of SEASAT provides abetter definition of fielaboundaries than Landsat and thatthe features called tone anatexture can be used to improve cornfrom soybean labeling accuracies.

2-11DEVELOPMENT, TEST AND EVALUATION OFA COMPUTERIZED PROCEDURE FOR USING

LANDSAT DATA TO ESTIMATE SPRINGSMALL GRAINS ACREAGE

k. R. J. '·lohler, w. F. Palr.ler,M. M. Smyrski, T. C. Baker

Lockheed Engineering and ManagementServices Company, Inc •

C. V. NazareIntergraph Corporation

This paper describes thedevelopment, test and evaluation ofCAESAR, a computerized areaestimation technique recentlytestea at the NASA Johnson SpaceCenter.The techni que util izes deci sionlogic to test for characteri sticsdetermined by analysts to beimportant for crop identification.Regi stered Landsat mul tispectralscanner data which have beentransformed into Kauth-Thomasgreenness and brightness arerequired.The accuracy of proportionestimates obtained using CAESAR wascomparable to earlier results usingmanual techni ques which were verylabor-intensive. The primarysources of error were the selectionof acquisitions and the designationof biowindows. With correctacquisition selection/aesignation,the mean error was reduced from3.01 percent to 1.36 percent, and

8

the standard deviation was loweredfrom 11.31 percent to 6.19 percent.These results are simi1ar to thoseobserved duri ng developrnent andserve to illustrate the potentialof the teChnique.

2-12EXPERIMENTS WITH AN EXPERT-BASED

CROP AREA ESTIMATION TECHNIQUE FORCORN AND SOYBEANS

Michael D. Metzler,Richard C. Cicone,

Karen I. JohnsonEnvironmental Research Institute of

Michigan, Ann Arbor, MichiganJulie Odenweller

Space Sciences Laboratory,University of California,

Berkeley, CALarge area crop inventory usi ngspace remote sensing has been amajor focus of the AgRISTARSprogram with special emphasis givento small grain, corn, and soybeanproduction forecasting. To satisfythe need for timely estimates ofcrop production in foreign areas,ground based observations must notbe required by the crop inventoryprocedure. Several approaches toaddress ing this problem have beenpursued by various investigators.This paper describes anexpert-based appr.oach to theinventory of corn and soybeans.In the analysi s of data for cropproduction forecasting, an expertanalyst has available satellitedata in either numerical or imageformat, weather data, historicaldata, and years of experi ence.This expert can often produce verycredible results, though they arenot easily repeated and can be timeconsuming. In addition,transferring this expertise toothers is not a trivial task. Totake advantage of this expertisewhile introducing greaterefficiency, Objectivity, and

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

repeatability, an attempt was madeto proceduralize the expertmethodology in such a manner that anon-expert, and eventually amachine, coula emulate the expert'sanalysis.Achieving this ~roceduralizationrequired a thorough understandingof both the expert's methodologyand the physical basis for thephenomena seen in the sa te 11 itedata usea in the analysis. Thisled to the development of astructurea, hierarchical decisionlogic for crop identification whichwould guide a non-expert analystalong a path followed by ~heexpert. Supporti ng technology toenable the expert approach wasdeveloped as well, including thefeatures to be uti 1ized in thelogic, data normalizationtechniques to maintain temporal andspati a1 consi stency in thosefeatures, and methods of presentingthe data that would convey maximuminformati on.The crop inventory procedure thusdeveloped demonstrated thecapabi 1 ity· for maki ng accurateestimates of crop acreage usi ngnon-ex pert anal ys ts, but in aninefficient, time consuming manner.Overcoming this lack of efficiencyrequired mechanization of many ofthe tasks which remained in thedomain of the analyst due to theirjudgemental nature.The solution to this difficultproblem lay in the choice of astaged approach to cropidentifi cation. In this approach,the machine would makeprogressively more difficultdecisions, with each stage of theprocess building on the accumulatedlearning of the previous stages. Atotal of four stages were developedwith the first stage being simplyan automation of the totallyobjective portions of the analyst's

9

crop identification logic, and thesucceeding stages designed tohandle the more judgemental of theanalyst's decisions.The results of tests of theprocedures val idate the conceptof an expert-based system, anddemonstrate the success with whichan objective, proceduralizedexpert-based system may beautomated to achieve both accuracyand effi ci ency. The chall engeahead is in adapting this approachto agronomic candi tions· which areconsiderably different from thosein the U.S. Corn Belt.

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

J-lLARGE AREA YIELD ESTIt~TES DERIVED

FROM PLANT SIMULATION MODELSSharon LeLJuc

Center for EnvironmentalAssessment Service

Tom HodgesUniversity of r·iissouri-Columbia

Plant simulation models areconceptual hypotheses on the growthand development of a crop. Theseprocesses are expressed as a systemand are the basis of a computerprogram to output these ideasquantitatively. The goal is to usethese moael s to provide estimatesof the yield over a largegeoyraptlic area. The effort wasstarted duri ng 1982 and is beingex tended to other geograph ic areas,other models and other crops. Theresults presented will be forestimati ng the spring wheat yi e1dfor IJorth Oakota using the TexasA&N wheat (TAMW) model. Variationsin the application of the modelwere tried. Summer fallow andconti nuous croppeo conditions wereconsiaered. Planti ng density waschanged. Two different methods ofsunmari zing tne c1imatic data forinput were trieo.Results show the simulated yieldestimates follow the annual changesin ODservea yield. Statisticaladjustment done objectively may beusea to improve the accuracy of theyi e1d estimate for the 1arge area.Resources required to provide theseestimates for a number of locationsare significant. Quality controland surrrnarization of c1imate datafor input to the moae1 requirescomputer and human resources.Oai1y temperatures, precipitationand solar radiation are required.Uperation of the models andobjecti ve stati stica1 adjustmentmay be accompli shed with the ma inresource being computer time.

10

Quality of the yield estimates willbe examined and compared to theyield estimates from the simplecrop yield regression modelsrequiring monthly temperature andprecipitation as input.Val ue of these model s is ti~ecapabi1 ity to operate them and geta yie1d estimate for areas whereunusual crop conditions aresuspected. Also, they can be usedfor a 1arge number of areas and atmany different time periods.

3-2APPLICATION OF SATELLITE SPECTRAL

DATAIN ESTIMATING WHEAT YIELDS

Tom Barnett, NASA-JSC/Co1umbia, MO.Using Landsat segment-averagedval ues of greenness (Kauth andThomas, 1976) for wheat in the U.S.Great Plains, 1978 and 1979, aconsistent linear relation wasestablished between greenness atheading and end-of-season yield.Both wi nter wheat and spring wheatacross the Great Plains appear toexhibit identical values of slopewith slightly different intercepts.Tests at a smaller scale onfield-average spectral and yielddata for 1975-1978 confirm thevalue of the slope. Application ofthe winter wheat yield: greennessrelation to 1981 spectral data fromthe NOAA-6 AVHRR sensor for 25 migrid cells over Texas, Oklahoma,and Kansas gave very satisfactoryestimates of final yield at CRDlevel.

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

3-3INTENSITY, HUE, SATURATION (IHS)DISPLAY Of THREE CHANNEL INPUT

Russ Afilbroziaak, IWAAPast work in color display has usedthe direct assignment of red, blue,and green to three channel impact.The two channel displays, green =blue and red, now usea by severalyroups, can De shown to be theleast effective way of displayingthe information. ~ore effectiveways need to be developed andtested in an operationalen.vironment. Several possiblechoices of candidates are:

o Index presentation - B & W,false color

o Spectral - IHS, RBGo Revi sed coorai nate system -

IHS, RBGInitial tests indicate that themost effective method is a modifiedpolar coorainate display in hue (H)and intensity (1) of the IHSsystt:m.

4-1DEVELOPMENT OF CROP SPECTRAL

TEMPORAL PROFILES APPLICATIONSTO CORN-SOYBEAN SEPARATION

G. D. Badhwar, NASA/JSCThe temporal development of abiological system is a key to itsidentification. In case of crops,this development manifests intemporal change in the spectralproperti es. A model has beenformulated that describes thechange of spectral properties as afunction of time in terms of knownbiophysical properties of crops.Features extracted using this modelhave been applied to separatecorn-soybean in both full seasonand in early season work.

11

4-2 SUPPORTING RESEARCH IN PATTERNRECOGNITION-INTRODUCTION

R. P. Heydorn, NASA/JSCThis paper discusses some of theresearch issues related in the useof remotely sensed data for 1andcover identification and inventory.These issues are discussed in termsof questions related to datarepresentation and inference. Indata representati on one is tryingto transform a given set ofmeasurements to values which bringout properties that discriminatebetween land cover categories whilesurpressi ng unwanted backgroundeffects. Inference coversquestions related to theclassification of cover types givena specific representative of themeasurements or questions relatecito the quanti ty of cover typematerial in an area.Si nee much of the research hasconcentrated on estimati ng thequantity or proportion of amaterial, most of this paper willdeal with inference issues relatedto the use of classifier basedproporti on estimators or di rectproportion estimators. Theseissues are discussed in tenns ofthe bias and variance of theestimators. When the II spectralseparation" between crops isreasonably large, as is generallythe case between corn and soybeanswhen several measurements areavailable throughout the growingseason, classification methods haveshown to perform well. However,when this separation is low, as isthe case between individual smallgrains (e.g., between spring wheatand spring bar1ey) other somewha tmore complex methods are suggested.For these cases a method based onthe "decomposition of mixtures" isbeing studied. In this approacheach crop is represented by itsspectral probability distribution

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

and the total probabilitydistribution of the scene as amixture of these cropdistributions. The mixingproportions are taken to be theestimates of the amounts of theindividual crops in the scene. Ifthe crop distributions are membersof a known (ana identifiable)family of probability distri-butions, then these estimates aretheoretically unbiased.

4-JESTIMATION AND IDENTIFICATION OFA VEGETATIVE COVER TYPE USING

MIXTURE DISTRIBUTION DECOMPOSITIONAND LABELING

R. K. Lennington, C. T. Sorenson,T. G. Lee, ana S. S. Shen

Lockheed Engineering and ManagementServices Company, Inc•

A funaamentally important prOblemin the analysis of remotely senseddata has been the characteri zationof the di stribution of spectralmeasurements for cover types ofinterest. Such a characterizationis an implicit or explicit part ofthe training of any classifierbased on mUltispectralmeasurements. It also forms thebasis for most unsupervisedclassification or clustering suchdata. Proportions of cover typesof interest may be directlyestimated as the prior probabil ityof distributions identified ascharacterizing each cover types.For these and other reasons, itappears natural to formulate theproportion estimation problem interms of a mixture of unaerlyingdistributions. This mixturedescribes the whole image and iscomposed of a sum of simplerdistributions, each with somespecified proportion. The usualassumption has been that theseunderlyi ng simpler distributionsare mul tivaria te normal. Fo r

12

agricultural applications, the realquestions in such a formulation arewhether crop categories of interestmay be well represented by a smallnumber of such underlyingdistribution, whether theunderlying distributions themselvesmay be resolved from the overallmixture distribution, and whetherthe resolved distribution may beidentified or labeled.This paper describes recent workdevotea to examining thesequestions for Landsat data.Mixture distribution resolution wasaccomplished using the CLASSYclustering algorithm developed byLockheed at the Johnson SpaceCenter. The features used werederived from parametric curvesfitted to multitemporal greennessdata and the pixels examined wererestricted to those that arereasonably pure.We show that small grainsdistributions may be wellrepresented by a set of mixturecomponent distributions. Directproportion estimates for smallgrai ns as computed from groundtruth labeled componentdistributions are presented for 18Landsat segments. These proportionestimates are compared to theground observed proporti ons ofsmall grains in these images. Inaddition, evidenced that mixturedistributions components may beidenti fied by predi cti ng thefeature distributions associatedwith small grains is presented.

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

4-4THE EFFECTS AND TREATMENT OF

"MIXTURE" PIXELS IN PROPORTIONESTIMATION OF VEGETATIVE COVER TYPE

k. S. Chhikara, M. B. MerickelLOCKheed Engineering and Management

Services Company, Inc •A. H. Feiveson

NASA/J SCFrom the viewpoi nt of spectralsignatures of a vegetative covertype and its proportion estimationin a segment, "mixed" pixelspresent an unidentifi able problem.Due to lack of understanding oftheir spectral characteristics,these have been treated in thepast like pure pixels whenclustering and classifying thesegment data. It has been seenempirically that this approachcauses higher mean square error inthe proporti on estimati on ofvegetative cover type.~ackground radiation and atmospheremay cause substantial adjacencyeffect on the spectral measurementof a pixel. It has been arguedthat the spectral measurement of amixed pixel can be considered as a1inear combi nati on of spectralmeasurements representi ng thevegetative cover type components ofthe mixed pixel. Based on thismodel, we have investigated theeffect of mixed pixel s on theproportion estimation of avegeta tive cover type. Undercertain assumptions for thespectral class distributions, it isshown analytically that thetreatment of mixed pixels as ifthese are pure can increase thebias and the mean square error of aproportion estimate considerably.Several edge detection techniqueshave been investi gated for thedetermination of boundary pixel s.A method based on the spatialapproach has been developed todetect boundary pixels and then, to

13

allocate each such pixel to thehomogeneous class of pixels thatare spectrally closest to it. Themethod has been applied to a numberof segments and is seen to effectimprovement on the proportionestimati on of vegetati ve covertype.In addi ti on, another proporti onestimation method has been proposedand is be ingin vest igate d • Ifcertain underlying assumptions holdtrue, this method woul d provi de adirect proportion estimationprocedure that does not requiredetection and hence, any specialtreatment of mi xed pi xel sin asegment.

4-5PRELIMINARY EVALUATION OF THEMATIC

MAPPER DATA FOR AGRICULTURALINVENTORYING APPLICATIONS

D. Pitts, G. BadhwarNASA/JSC

S. Yao, M. Amis, S. Shen,J. Artley, C. Sorensen, J. Carnes

Lockheed Engineering and ManagementServi ces Company, Inc.

In order to gain famili arity withthe Landsat-D Thematic Mapper (TM)data, and to get some preliminaryindications of how this data can beused in agricultural inventoryingapplications, evaluations wereperformed on initial TM data setsproduced by G5FC. The evaluationsused TM data acqui red over WebsterCounty, Iowa, on August 2, 1982,and over Mississippi County,Arkansas, on August 22, 1982.As part of the evaluation, the JSCregistration processor was used toevaluate the band-to-bandregistration accuracy. Theaccuracy was better than one pixelfor all bands. As expected, theregistration accuracy among thefirst four bands and bands 5 and 7.

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

by comparing a TM image with a USGS7-1/,l. quadrangl e sheet map, thegeometric fidelity of the H1 imagewas founa to be excellent.Principal components analysis wasperformed on the TM Oata to get anindication of the informationcontent in the ciata. The resultsindicated that the six bands in thevi sible through mi d-infrared can beconverted into three principalcomponents which account foressentially all of the variabilitj'in the aata. This confirms anearlier result obtained using fieldmeasurement aata.Eval uati on of band countaistributions for corn anci soybeanssllowed that data in n\ band 4 mayproduce better corn/soybeansseparation than tviSSdata. The nldistributions were less skewed thandistributions using MSS data. Thisshoula produce improvedclassification results.In oraer to eval uate the effect ofincreased resolution onc1ass ifie ation, an un sup e rv isedclustering algorithm (CLASSY) wasappl ied to the Tttlaata and tosimul taneously acqui red ~1SS data.By observing features which werenot resolved into separate clustersusing TM data, an indication of howthe incrtased Tt-iresolution wouldaffect classification was obtained.The results of this evaluationindicated that using TM data wouldproduce separate classes for smallfeatures such as roads andhomesteads, while using MSS datawou1 d not separate these smallfeatures.

14

4-6MULTI-SENSOR REGISTRATION OF

EARTH RESOURCE DATAR. D. Juday, NASA/JSC

S. S. Yao, Lockheed Engineering andManagement Services Company, Inc.

P. Anuta, ERIMand S. Ungar, GISS

The effective utility of theremotely sensed earth resourcesdata gathered by satell ites andaircraft can increase many fold ifacqui sitions over the same groundareas at different times can beprecisely spatially aligned. Twoprocessors are used to accompl ishthis task. The JSC regi strationprocessor is used to registerLandsat multispectral and ThematicMapper data, while the GoddardInstitute for Space Study (GISS)Processor is used to accomplishaircraft Thematic Mapper Simulator(TMS) data registration.The JSC regi strati on processorbuilds upon the technologydeveloped for both the GoadardSpace Flight Center's Master DataProcessor (MOP) and the LACIEprocessor. It accepts both the MOPlip II formatted as well as "AIIformatted tapes as input. It makesuse of all the available ancillaryinform ation for a pair 0 facquisitions in accomplishing theimage-to-image registration. Theoutput images can be resampled intodifferent pixel sizes as well asplaced into a variety of mapprojections. Subpixel registrationaccuracies are achieved by crosscorrelating edge image patches fromboth the reference and theregistrant images and iterativelylocating the correlation peakoffsets to a fraction of a pixel.

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

The GISS processor makes use of theaircraft navigation and attitudeinformation to correct thedistortions caused by aircraft yaw,pitch and roll variations. Controlpoints are used to tie the imageryobtained to the desired mapcoordinates.In this paper, the details of theregistration methods for bothprocessors are described. Also,performance eval uati ons andrepresentative operationalcharacteristics are given.

4-7MEASUREMENTS AND SCENE ANALYSIS OFCORN-SOYBEANS AND SMALL GRAINS FOR

IDENTIFICATION, DEVELOPMENT STAGEESTIMATION AND CONDITION ASSESSMENT

Marvin E. BauerPurdue UniversityEdward T. Kanemasu

Kansas State UniversitylSlane L. Blad

University of NebraskaJ. Clifford Harlan

South Uakota State UniversityThe goal of the measurements andscene analysis research is toprovide quanti tati ve informati onand models on the distinguishingbiophysical and radiometriccharacteristics between cropclasses (e.g., corn-soybeans,spring wheat-barley) and cropattributes within classes (e.g.,development stage, leaf area indexor moi sture stress of corn). Theapproach is to first identi fy thekey cultural and biophysicalfeatures which enable crop type,stage and condition to beidentified using remotely sensedspectral measurements and then todetermine, by empiricalcharacterization and modeling, howthe biophysical-agronomic variablesare manifested in the spectralresponse. In this step, thefunctional relationships between

15

biophysi cal and spectral vari ablesare being determined. This isfollowed by development of model sutilizing spectral inputs,particu1 arly Landsat MSS and TMdata, and determi nati on of thesensitivity of the models to soil,crop and environmental factors.The approach includes measurementsof cultural practice experiments atten agricul tural experimentstations in the Corn Belt and GreatPlains states, along withmeasurements of commercial fielasat test sites in Iowa and NorthDakota. The experiments includetreatments of soil type, plantingdate, plant popul ati on and rowspacing, species and cultivar,fertilization and moisture level.Agronomic data include developmentstage, 1eaf area index, and othercanopy descriptors. the primaryspectral measurements are made withan eight-band radiometer with theTM spectral bands; other sensorsinclude radar and airbornemultispectral scanner.The primary objectives, experimentdesigns, measurements, and keyanalysi s resul ts of experimentsbeing conducted by severaluniversities including Purdue,Kansas State, Nebraska, and SouthDakota State will be described.The results include use of the datain spectral-temporal profile modelsfor crop identification anddevelopment stage estimation,relationships of spectral variablesto canopy leaf area index and lightintercepti on, effects of nutri entand moi sture stress on spectralresponse, and spectral charact-eristics of different species asfunction of develoment stage,cul tura1 practi ces andenvironmental factors.

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

4-8USES OF VEGETATIVE CANOPY

REFLECTANCE MODELNarendra Goe1, ASTEK Consulting

Several potential uses of theVegetati ve Canopy Ref1 ectanceModel, in conjunction withatmospheric scattering models forvegetative mapping will bediscussed. They include thedetermination of agronomicvariables like leaf area index andleaf angle distribution from canopyreflectance aata, and thequantification of "distortion"caused by the atmosphere in thereflectance as measured bysatellite-borne sensor.Recent progress made in these areaswill be highlighted.

4-!ISPECTRAL ESTIMATION OF LEAF AREA

INDEX AND LIGHT INTERCEPTION BYCROP CANOPIES

Marvin E. Bauer,Craig S. T. Daughtry

Purdue UniversityEdward T. Kanemasu

Kansas State UniverSityJohn ~1.Norman

University of NebraskaIf agronomic variables related tocondition and yield could beestimatea from multispectralsatellite data, then crop growthand yield mOdels could beimp1ementea for large areas. Theobjective of research at Purdue andKansas State Universi ties has beento develop approaches for combiningspectral and meteorological data incrop models. Leaf area inaex (LAI)is a key variable related to growthand infrared reflectance of cropcanopies; it in turn is directlyrelated to the light interceptionof canopies, a fundamental para-meter photosynthesis, evapotrans-piration and yield of canopies.

16

Reflectance data have been acquiredover several growi ng seasons ofcorn, soybean, and wheat canopies.Treatments included planting date,row width, plant population,cu1tivar, and soil type. Agronomicdata included leaf area index(LA!), percent soil cover, biomassdevelopment stage, and grain yield.The spectral variable greennessexp1ained 78% of variation in LAIover all treatments of corn.Single date observations of LA! orgreenness had little value inpredicti ng yie1ds. The proporti onof solar radiation intercepted(estimated from reflectance data)when accumu1 ated over the growi n9season accounted for approximate1yy65% of the variation in cornyi e1ds. Similar resu1 ts have beenobtained for soybeans and wheat.The concept of estimatingintercepted solar radiation usingspectral data represents a viab1eapproach for mergi ng spectral andmeteorological data in crop yieldmodels. We are currentlyassembling the necessary data toevaluate the concept using Landsat(·1SSdata, as well as obtainingdirect measurements of canopy lightinterception.In assessi ng the capabil ity ofremote sensing to e.stimate LAI, itis critical to address the abilityto make accurate in situmeasurements since remote sensingexperiments can be no moreconclusive than the groundobservational data on which theyare based. The natural variabilityin LAI and the accuracy andprecision of direct measurementsare being assessed. Addi tionally,we are developing and evaluatingindirect, non-destructive methodsof estimati ng LA!. These methodsoffer a more rapid means ofgatheri ng thi s informati on, mayprovide additional information(e.g., leaf angle distribution),

AgRISTARSMINI-SYMPOSIUMABSTRACTS

and may improve the accuracy ofestimates. The leading example ofthis approach is the work by Normanwho has inverted a model ofraaiation in plant canopies toestimate canopy structure variablesfrom ground data. The mOdeli nvers i on can be dri ven by ei therpercentage of sunf1 eck area as afunction of sol ar e1 evation or theproporti on of canopy gap as afunction of zenith angle which canbe acquired from hemisphericalphotographs.

4-l()MICROWAVEPROPERTIESOFAGRICULTURALCROPSAT LONG

WAVELENGTHSJack F. Paris, NASAjJSC

The mi c rowave properti es 0 fagricu1 tura1 crops at shortwave1 engths ( 1ess than 4centimeters) are well understood asthe resu1 ts of extensi ve researchin the 1970's by investigators atthe Uni ve r s i ty 0 f Ka n sa s • I ngeneral, it is known at thesewavelengths that the wet biomass ofthe crop' canopy is a domi nantparameter. Also, the use of crosspolarized radar was found to beminimal for the shorterwavelengths. Row structure and rowdirection effects and soil moistureeffects have been found to be smallfor the short wave1 engths. Atlonger wavelengths, little wasknown about the active microwaveproperties of crops before 1980.In the present paper, the resu1 tsof experiments conducted in 1YSOand 1Y81 over bare, plowed fi e1ds(row structure and row directionexperi ments) and over corn,soybeans, wheat, bar1 ey, andsunflower fields (vegetation canopyexperiments) by the presenter aregiven.

17

In general, many of the resu1 tsnoted by previous investigators inthe short wavelength portion of themicrowave region did not hold truein the long wavelength portion.Row structure and row d i recti oneffects can be quite large (up to20 db), especially at L-band (20em) for fields that have beenprepared for flood irrigation.A1so, the useful ness of crosspol ari zed radar measurements ismuch greater for long wave1 engthsthan for short wave1 engths.Indeed, the separation of maturecorn from mature soybeans waspossi b1e at C-band only wi th theuse of the HV channel (hori zontaltransmi t - verti ca 1 recei ve) •Exami nati on of the frequencydependence of the radarbackscattering coefficient for cornand soybeans reveal ed an apparentresonance effect probably affectedby the closeness of the si ze ofscatteri ng e1ements in the canopy(leaves, stems, and fruit) to thatof the microwave radi ati on. Suchunique scattering properties couldlead to the development of arobust, single-date cropclassification procedure usingmu1tifrequency and mu1ti-polarization radar data.

4-11DEVELOPMENTOF THE JSC THEMATICMAPPER(TM) QUICK-LOOKIMAGE

PROCESSORJ. Gilbert, NASAjJSC

With the launch of Landsat-D inJuly 1982, the remote sensingcommunity has been provided a newearth orbiting multi-spectralsensor whi ch has increased spati a 1resolution and superior spectraldi scrimi nati on than any previouslylaunched Landsat instrument. Thisnew instrument, the ThematicMapper, is a forward step in theprogression of remote sensing

AgRISTARSMINI-SYMPOSIUMABSTRACTS

technology, incorporating theknowledge and experience gleaneQfrom earl y scanni ng 'i nstrumentendeavors as well as utilizing thelatest available satellitestability and pointing capability.

The immeaiate concern of landsatground processi ng segments becamethe accommoaation of a data volumeoutput from the Thematic Mapperwhich is seven times as great asmulti-spectral scanner output forthe same scene coverage. Often,ex i sti ng mu1 ti -spectral scannerprocessors were in p1ace si nce th{early 1970·s and were not capablE:of hand1 i n9 such vol urnes of data.The prob1 em consequently became amanifestati on of the adaptati on ofexisting systems, whereverfeasible, and in many instances,the implementation of newprocessors designed for increaseddata volume handling.

This paper presents the developmentof the Thematic Mapper groundpre-processing capability of theEarth Observati ons Data Laboratory(EDOl) at the Johnson Space Center.The implementation of newprocessing elements and theutilization of existing syystems inthe EOOLto hand1 e Themati c Mapperdata are reviewed in detail. TheEVOl Themati c Mapper Pre-processorhas been designed to provideAgRISTARS researchers withsatellite imagery data overselected areas of interest in datasets whi ch are si zed to becomputa ti ona 11y manageable. Inaddition, the EOOLThematic MapperPre-processor includes thecapability to provide imageproducts in support of researchactivity.

18

4-12EVOLUTIONOF A REMOTESENSINGRESEARCHDATASYSTEMANDDATA

MANAGEMENTM. Alexander, NASA/JSC

The United States Space Program haSgreatly enhanced man IS abi1 ity tomonitor and examine the earth, itsweather, and its oceans through thecollection of synoptic data fromorbit. The technology to manageand analyze this remotely senseddata has not kept pace with theamounts of data available nor withthe ever increasing demands of theresearcher for ready access to tt·edata and data processingfaci 1iti es. As the quanti ty andvariety of data have increaseddramatically, so have the problemsof data arc hi val, a na1y s is, andcommunications.

This paper presents theevo1 uti onary development of theEarth Observati ons Data Laboratory(EOOl) data systems 1980-1982 as ameans of sol vi ng these researchre1 ated prob1 ems. Thi s transi ti onoccurred duri ng a peri od of rapi dtechnological change accompanied bybudget, program, and personnelreductions.

The choice of commerciallyavailable off-the-shelf IBM p1ug-compatible hardware; VM370/CMS/-OS/RSCS-Networking/ADABAS systemsoftware; SAS/HISl applicationssoftware; as a the keystone datasystem in the EOOl, has beenessential to the success of theEODl transition.

In addition, the Earth ResourcesResearch Division (ERRO) maintainsa large collection of data sets anddata bases. A1though these databases are related in that they areall earth observations, the dataare almost exc1 usively independentwith no shared software capabil-ities or electronic data 1inkage.

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

proportions as an alternative tothe maximum likelihood procedurescurrently employed, and theperformance of the MDE and MLE onboth normal and non-normal data hasbeen investigated.

In order to provi de basi c datamanagement requirements in additionto providing the capability ofmanaging the above mentioneddisparate data sets, a Data BaseManagement System (DBMS) wasinstalled in the EODL. The logicand selection criteria of the DBMSareas are discussed in this paper.

4-UCOMPARISON OF MINIMUM DISTANCE AND

MAXIMUM LIKELIHOOD TECHNIQUES FORPROPORTION ESTIMATION

W. A. Woodward, W. R. Schucany,H. Lindsey, and H. L. Gray

Southern Methodist UniversityA cODlllonobjective in agriculturalremote sensing is the estimation ofthe crop proporti ons al' ••. am inthe mixture density

mf(~) = k~l ak fk(~)

where m is the number of crops andfor each crop, fk(~) has been takento be the reflected energy in fourbands of the light spectrum,certain linear combinations ofthese readings, or other derived"feature" vari abl es. We haveexaminedestimation

minimumof the

distancemixture

19

5-1IMPROVEMENT OF MOISTURE

ESTIMATION ACCURACY OF VEGETATION-COVERED SOIL BY COMBINED ACTIVE/

PASSIVE MICROWAVE SENSINGFawwaz T. Ulaby, MYron C. Dobson,

and David R. BrunfeldtRemote Sensing Laboratory

University of KansasThe measured effects of vegetationcanopies on radar and radiometricsensitivity to soil moisture arecompared to first-order emissionand scattering models. The modelsare found to predict the measuredemission and backscattering withreasonable accuracy for variouscrop canopies at frequenciesbetween 1.4 and 5.0 GHz, especiallyat 0 ~ 30°. The vegetation lossfactor, L(O), increases withfrequency and is found to bedependent upon canopy type andwater content. In additi on, theeffecti ve radi ometri c powerabsorption coefficient of a maturecorn canopy is roughly 1.75 timesthat calculated for the radar atthe same frequency. Compari son ofan L-band radiometer with a C-bandradar shows the two systems to becomplementary in terms of accuratesoil moisture sensing over theextreme range of naturallyoccurring soil moisture conditions.The combi nati01') of both an L-bandradiometer and a C-band radar isexpected to yi el d soil-moi stureestimates within +/-25% of truesoil moisture even for a soil undera "lossy" crop canopy such as corn.

AgRISTARS MINI-SYMPOSIUM ABSTRACTS5-2

PRELIMINARY RESULTS OF THE COLBYAGRICULTURAL SOIL MOISTURE

EXPERIMENTJohn C. Richter

Lockheed Engineering and ManagementServices Company, Inc.

Jack F. ParisNASA/JSC

NASA conducted a major soilmoistureexperiment in the summer of 1978 ina test site near Colby, Kansas.Aircraft radar scatterometersmeasured the backscatteri ngcoeffi cient of about 40 fie1ds atseveral wavelengths, look angles,and polarizations on six dates. Oneach f1 ight date, ground truthteams collected data on thedistribution of soil moi sture heach fie1d with depth and over thehorizontal extent of the fields.They noted or measured other fieldproperties as well as vegetationcover type and amount, rowdirection, row structure, andsurface roughness.To increase the quality of the datato be. analyzed, we designed andimplemented software packages onthe AS300U computer system at theNASA ~ndon B. Johnson Space Centerto locate accurately the horizonta1position of each area viewed by thesensors during the flights for eachradar configuration. Also, weincluded checks on signa1-to-noiseratios and other engineeeringcoefficients. Having access to lowaltitude photography taken duringthe f1ight 1ine runs, we were ab1eto locate the sensor footpri ntsmore accurately than otherinvesti gators who have beenanalyzing the same data.The best single radar configurationfor sensing soil moisture (given asthe average vol umetri c moi sturecontent in the upper 5 cm of soi1)was in the C-band wi th apolarization combination of

20

horizontal transmit and horizontalreception (HH) at a sensor lookangle of 15 degrees with respect tothe nadir. This resu1t confi rmsthe findings of several otherresearch groups, however, t~ecoefficient of determination R ,was quite high (0.85) as comparedto a value of 0.65 obtained on thesame data set by previousinvestigators. This significantdifference in results was due tothe data hand1 ing procedures usedin the present analysis.

5-3PASSIVE MICROWAVE SENSING OF

SOIL MOISTURE UNDERVEGETATION CANOPIES

Thomas J. Jackson,u. S. Department of AgricultureThomas J. Schmugge, and

James R. WangNASA Goddard Space Flight Center

Vegetation cover has a significanteffect on the abi1 ity of passivemi crowave radi ometers to detectchanges in near surface soilmoisture. A quantitative techniquefor isol ati ng the effect ofvegetation was developed using atheoretical model as the basis of aparametric approach. This approachwas evaluated using data collectedby truck mounted sensors overexperimental plots. Resu1 ts showthat a microwave radiometeroperating at a 21 cm wavelength canprovi de vol umetri c surface soi 1moisture estimates to approximately5% of accuracy for fie1ds coveredwith moderate vegetation. Inaddition, all of the data requiredfor applying the parametric modelcan be measured using remotesensing.

AgRISTARSMINI-SYMPOSIUMABSTRACTS

b-4A MICROWAVESYSTEMSAPPROACHTO MEASURINGROOTZONESOIL

MOISTUREk. W. Newton

Texas A&~'l Universi ty

The current availability of waterfor agricu1 tura1 purposes hasbecome a severe problem in the lastfew years especi ally in Texas andOklahoma. Development of efficientand effecti ve methods ofregulating, monitoring andutilizing the remaining waterresources are cruci a1 to taki ngaction to minimize the problem.This document describes an approachto developing a technique ofmonitoring soil water conditionsover large areas for potential useto agricu1 tura1 managers. Thisapproach utilizes an orbitingpassive microwave remote sensingsystem to esti mate near surfacemoisture and a deterministic soilwater model to predict the moistureat root zone depth based on thenear surface soi 1 moi stureestimate.

Thi s microwave systems approach toremotely measuring large area soilwater information is currently inthe development and evaluationstage. This issue has beenaddressed thus far by dividi ng theprob 1em into two issues. One isthe eva1 uation of the abi1 ity tomake a 1arge area surface soi 1moisture estimate using an orbitingpassive microwave system (Newton eta1., 1979). The second is thedevelopment and evaluation of themodel that is capable of utilizingthe near surface soi 1 moi stureestimate as an input along withsoil characteristics to predict themoisture within the root zonedepth. The prob1 em has beenapproached by developing computer

21

simu1 ati on model s to eva1 uate theapproach and to val i date theseresults where possible with actualfield experimentation.

Experimental microwave and groundfield data have been acquired fordeve1 opi ng and testi ng a root zonesoil moisture prediction algorithm.The experimental measurements havedemonstrated that the depth ofpenetration at a 21 cm microwavewavelength is not greater than 5Cln. Previ ous work by Jackson(1980) i ndi cates that the moi sturein the lower prOfile (below fivecm) can be estimated with a 0.06standard error by using surfacesoil moi sture for the top 5 cm.Thi s document presents simu1 ati onsof brightness temperature over baresoil condi ti ons that can beutilized to test the lower profileprediction scheme of Jackson(1980). In addition, fieldexperimental measurements exi st tovalidate this work.

5-5REMOTESENSINGOF SOIL MOISTURE:

RECENTADVANCEST. Schmugge

NASAGoddard Space Flight Center

In the past few years there havebeen many advances in ourunderstanding of the microwaveapproaches for' the remote sensi ngof soil moisture. These include amethod for estimating thedependence of the soi1's dielectricconstant on its texture; the use ofpercent of field capacity toexpress soil moisture magnitudesindependent of soil texture;experimental and theoreticalestimates of the soil moisturesampling depth; models fordescri bi ng the effect of surfaceroughness on the microwave responsein terms of surface height varianceand the horizontal correlationlength; verification of the ability

AgRISTARS MINI-SYMPOSIUM ABSTRACTSof radiative transfer models topredict the microwave emission fromsoils; experimental and theoreticalestimates of the effects ofvegetation on the microwaveresponse to soil moisture; andsimulation studies indicating howremotely sensed surface soi 1moi sture may be used to estimateevapotranspiration and root zonesoil moi stur~.

6-11981 AGRISTARS DCLC

FOUR STATE PROJECTJames W. MergersonCharles E. Miller

Ma rtin OzgaNarti n Holko

Sherman WiningsPaul Cook

George A. HanuschakThi s paper summari zes the workperformeo under the major crop areaestimation element of the 1981AgRISTARS (Agriculture and ResourceInventory Surveys Through AerospaceRemote Sensing), DCLC (DomesticCrops and Land Cover) Project.The DCLC objective of providingtimely, more precise year-end stateand sub-state crop area estimatesfor SRS was accomplished. Corn andsoybeans pl anted area estimateswere provided for Missouri andIowa. liarvested wi nter wheatestimates were provided for Kansasand Oklahoma.

22

6-2KANSAS LAND COVER SURVEY

George May, USDAGregory S. Burns,

Marty Holko,Jim Anderson, ERL

During FY81 a state level landcover study was conducted inKansas. This survey utilized thearea samp1e frame and methodo109Ycurrently used by USDA, Stati sticalReporting Service to provideprobability estimates of cropacreages. All land within the 435June Enumerative Survey segmentswas enumerated into 17 cover types.These ground data were used inclassifying Landsat data to producea land cover classification for theentire state. Statistically based,acreage estimates were obtai ned bydeveloping regression relationshipsbetween the ground and classifieddata. Regional land cover maps andassociated regression acreageestimates were also produced. Apresentation on the results of thisstudy was given to various stateand federal agenci es in an effortto provide 1and cover informati onthat coul d benefi t resourcemanagers.

6-3THE USE OF LANDSAT FOR COUNTY

ESTIMATES OF CROP AREASGail Walker

Richa-rd SigmanStatistical Reporting Service

USDAThe purpose of this report is todevelop and compare estimatorswhich use Landsat data to estimatecrop areas at the county level.This report extends theBattese-Fuller estimator to astratified sample design andevaluates the Huddleston-Rayestimator and variations of theBattese-Fuller estimator on asix-county area in South Dakota.

AgRISTARS MINI-SYMPOSIUM ABSTRACTSFor SRS Landsat studies, theauthors recommend repl aci ng theHuddleston-Ray estimator with oneof the favorably evaluatedestimators in the Battese-Fullerfamily.

6-4AUTOMATED SEGMENT MATCHING

ALGOR ITHMMaria Kalcic, NASA/NSTL/ERL

The USUA/Statistical ReportingServ ice uses two stages in theregistration of digitized samplesegments to Landsat M5S data. Thefirst stage uses a control point-based transformation applied to theenti re area of study. 5ince thefirst stage, or global registra-tion, does not produce acc uracywithin a required one-half pixel,on a per segment basis, a secondstage registration is required.This second stage, or local segmentshi fti ng, has been performedmanually by overlaying plots of thedigitized segment and fieldboundaries to grey-scale plots ofthe Landsat MSS data, and thenshifting the plots to produce theproper 1ine up. The second stageregi stration has been automated asa result of the Domestic Crops andLand Cover/Scene-to-Map Registra-tion Task. The Automated SegmentMatching Algorithm (ASMA) usesedge-enhancement and within fieldvariability as inputs to computethe necessary adjustment to eachsegment's computed location.The algorithm results were comparedto the manual shifting results andproduced root mean square errors of18.86 and 25.21 meters in the lineand element directions, respec-tively. These errors meet theU50A/5RS requirement for a one-halfpixel registration accuracy.

23

6-5CLASSIFIER DESIGN FOR REGRESSIONS

OF GROUND GATHERED WITH COMPUTERCLASSIFIED DATA

R. P. Heydorn, NASA/JSCAn estimator which is based on aregression of June EnumerativeSurvey (JES) data with computerclassified Landsat data is beingappl ied by SRS to estimate statecrop acreages. The purpose forusing the Landsat data is toincrease the precision of theestimates based on the JES dataalone. Recently, regressionmethods have also been consi deredfor again using the Landsat data toimprove the JES estimates at thecounty 1evel • In each of thesemethods regressi on parameters areestimated using not only data fromthe county of interest but alsofrom surround ing counti es. Due tocounty-to-county differences, datafrom one county need not have thesame statistical properties as doesthe data from the collection ofcounties. To minimize bias in thecounty estimates it is importanttha t the data fi ts any assumedregression model.This paper discusses the results ofa NASA/JSC study to investigate theproperties of various classifierdesi gns that can infl uence theshape of the regression function.Some classifier designs which canlead to linear regressions incertain cases are di scussed. Thepaper also pr~sents a relationshi pbetween the R of a regression andthe omission/commission error ratesof the classifier.

AgRISTARS MINI-SYMPOSIUM ABSTRACTS6-b

STRATIFICATION OF SAMPLED LANDCOYER BY SOILS FOR LANDSAT-BASED

AREA ESTIMATION AND MAPPINGE. R. Stoner, NASA NSTL

The distribution of agriculturalcrops and other 1and cover typesfrequently is closely associatedwith certai n soi1sand 1and forms.One implication for Landsat-basedarea estimation and mapping is thata given pixel occurring within acertain soil unit does not in facthave an equal probability offalling into anyone of the landcover classes. For example, 99% ofall tobacco fields in RobesonCounty, North Carolina occur onwell or moderately well drainedsites , 1eaving 46% 0 f the countyland area on which tobacco isunlikely to occur.Another implication of soil/landcover relationships for spectraldiscrimination of 1and cover typesis tne possibility of soil-inducedspectral variations related tosoil-specific management practicesand background reflectancecharacteristics. In the RobesonCounty site, crop developmentcycles can be expected to differbetween soybeans grown on welldrained uplands and poorly draineddepressi ona1 bays. Other cropssuch as tobacco, are grown in widerows and achieve only partialground cover, wi th the resu1t thatsoil ref1ectance predorninates overthe response of green vegetativecover.Samp1 ed 1and cover can bestratified using mapped soilinformation available in typicalcounty soi 1 su rveys. For thepurpose of the Robeson County 1andcover study, soil information wassimplified into 5 general soilclasses whose characteristics ofnatural drainage, physiographic

24

position, organic matter content,and surface color are homogeneousin regards to properties that wouldbe expected to influence LandsatMSS response. Individual fieldswi thi n June Enumenati ve Survey(JES) segments were coded by soilclass grouping to effect thestratification of sampled landcover by soils. Cover typeinformation was available for corn,soybeans, cotton, tobacco, hay,pasture, and forest. A mul tidateset of Landsat MSS data from the1980 growing season was used in theanalysis.

6-7A CORRELATION ANALYSIS OF PERCENT

CANOPY CLOSURE YS. TMS SPECTRALRESPONSE FOR SELECTED FOREST SITES

IN THE SAN JUAN NATIONAL FOREST,COLORADO

M. K. ButeraNASA/NSTL/ERL

This investigation tested thecorre1 ation between canopy closurewhich is an indicator of forestbiomass, and individual ThematicMapper Simulator (TMS) bands forselected forest sites in the SanJuan National Forest, Colorado.Percent canopy closure wasdetermi ned for 30 sites, each 25acres in area,. from aerialphoto-i nterpretati on and groundsurvey. The sites were selected torepresent a range of canopy closurefrom 0 to 100%. They were a1soselected from plateaus with slopeE;:; 10% at an elevation of

approximately 9,000 ft. Thiscondition minimized the effect ofslope as a variable in theanalysis. The forest cOlilTlunitieswere dominated primarily byponderosa pine and aspen. For eachsite, the mean response per bandwas calculated from TMS dataacquired over the study area onSeptember 1B, 1961. A linearcorrelation and regression analysis

AgRISTARS MINI-SYMPOSIUM ABSTRACTSwas performed on mean TMS responseper band per site versus % canopyclosure. Correl ation coeffi cientsfor TMS bands 1-7 were -.757,-.663, -.666, -.088, -.797, -.579,and -.763, respecti vely. Usi ngband 5 data, a model regression inthe fOnT! ofarcsi n V%Canopy closure = y =114.69 - 2.363 ch 5

was applied to the data, creating amap of preaicted % canopy closurefor the study area. The fOllowingconclusions were made:1. TiltSbands 1, 5, 7, essentially

wavelength intervals notcovered by MSS data, provedmost significant in relating %canopy closure to spectralresponse.

2. The negative correlations wereprobably caused by a spectralcontribution from thebackground (dry soil, senescinggrasses) with higherref1ectivity response than theforest canopy.

3. For a given ecosystem the bestpredictive model is achievedwhen condi tions of greatestspectral contrast betweenbackground and forestvegetation exist.

6-8THE SIMULATION OF USDA SEGMENTS,

FIELDS, AND PIXEL SPECTRAL VALUESJ. C. Lundgren and Y. Tsong

Lockheed Engineering and ManagementServices Company, Inc.

The evaluations of procedures whichuse classification of Landsatspectral data to estimate cropproporti ons have been inconcl usivein some cases due to an inadequatedata base. Furthermore, ourunderstanding of the classification

25

process cou1 d be greatly improvedwith a larger data base.A simulation procedure has beendeveloped which could create such alarge data base by simulatingrandom segments, fie1ds, and pixelspectral values. This procedureattempts to create simulated valueswhich are similar or equal to theactual values found in our 1979 setof 33 Mi ssouri segments. Itsimulates segment sizes, fieldsizes, crop proportions, and apercentage of edge pixels which aresimilar to the actual data.The procedure simulates four-channel spectral val ues at thepixel level and the twoacqui sitions found in our Mi ssouridata set. It simulates a withinfield vari ation of pixel spectralvalues, a between field (withinsegment) variation, and a betweensegment variation. Correlationsamong the channel s and betweenacquisitions are simulated throughthe use of principal components.Spectral values for edge pixels aresimul ated through a linear mixingof adjacent fie1d effects.This paper also presents a pre-liminary comparison of simulatedand actual data.

6-9USDA SMALL AREA CROP ESTIMATION

USING LANDSAT- AND GROUND-DERIVEDDATA

M. L. AmisLockheed Engineering and Management

Services Company, Inc.The USDA approach to cropestimation for 1arge areas such asa state or a crop reportingdi strict is to regress survey data(ground truth) onto Landsatclassification results. Thisestimator can produce unbi asedestimates with measurable precision

AgRISTARSMINI-SYMPOSIUMABSTRACTS

for such areas. However, thepossible lack of adequate surveydata in a single county or smallgroup of counti es requi res the useof sampling units from the entirestate or analysis district in orderthat the regressi on model may beapplied to these slllaller areas.within this context, the regressionestimates can be biased.

Thi.) papEr sun;mari zes thephilosophy, evaluation, and resultsof three approaches to small areaestimation; the problems intrinsicin these approaches; and theresearch directions in which thesedifficulties led.

6-10EVALUATIONOF USDALARGEAREACROPESTIMATIONTECHNIQUES

Sylvia ShenLockheed Engineering and Management

Services Company, Inc.

This paper describes the results ofthe Domesti c Crops ana Land CoverClassificaion and Clustering studyon large area crop estimation usingLandsat and ground truth data. TheUSDAI s EUITOR system regi sters anddigitizes the ground truth and rawLandsat data. It clusters,classifies, and develops areaestimates by regressi ng the groundtruth hectarage for a gi ven croponto the n urnbe r 0 f pix e 1 scl assifi ed into that crop for eachsegment. A research program wasconducted to eval uate theperformance of EDITOR and makeselectea improvements to componentsof EUITUR. In was found that theuse of mul ti temporal data, overunitemporal, significantly improvedcrop hectarage estimates.Performance measures on anindependent test set and ajackKni fed test set decreased,indicating that the EDITORprocedure of usi n9 a si n9l e dataset for training the classifier,

26

developing the regressions, andevaluating the results leads tooveroptimistic performance esti-mates. An alternative clusteringalgorithm, CLASSY, when substitutedfor the EDITOR cl usteri ng method,produced improved estimates. Useof a simpler classifier, namely~1ean Square Error Classifier, didnot produce significantly betterhectarage estimates but showed moreextendibility of the regressionlines to an independent test set.

7-1MAPPINGFORESTRESOURCESUSINGTHEMATICMAPPERSIMULATOR(TMS)

DATAStephen D. Degloria

University of California-Berkeley

A quantitative evaluation of TMSmul ti spectral data is requi red todetermi ne the extent to which TMdata will serve as a framework forthe spati al estimati on and mappi ngof resources critical to forestmanagement and planning. 11-1Sdatawere acquired by the U-2 aircraftover the Plumas National Forest(PNF), California, during October1981 through August 1982 coincidentwith systematic ground datacollection efforts. The October1981 mission focused' on developmentand eva1 uati on of day-ni ghttemperature difference images forthe spatial estimation of soiltemperature regimes. The August1982 mission focused on determining(1) the abi 1i ty of these data todetect and identify critical forestresources, and (2) the spectralvariability of TMS data overdiverse terrain. Relationshipsbetween the spectral data from theseven bands of the TMS and severalforest resource variables includetree and brush species composition;basal area of commerci a1 coni fers;height, age and DBH of dominant

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

tree speci es; average stand crowndiameter; average stand density;timDer site; soil family; 0 horizonthickness; soil temperature(surface ana 5U cm depth); and thetopographic variables of elevation,slope, ana aspect. Based on thisand previous work, a mul tivari atesoil temperature mapping functionis being developed and applied insupport of the PNF Soi 1 ResourceInventory (SRI). Future activitiesinclude the integration ofLandsat-4 Tr-' and fviSSdata into themul ti -di mens i ona 1 data base forevaluatins the spectral, spatial,radiometric, and geometriccharacteristics of these data inforest and rangeland environments.

7-t..DETECTING FOREST CANOPY CHANGE

USING LANDSATRoss F. Nel son

Goddard Space Flight Center

Multitemporal Landsat multispectralscanner data were analyzed to testvariolls computer-aidea analysistecnniques for detecting signifi-cant forest canopy al terations.Three da ta trans forma ti ons ,differencing, ratioing, and aaifference of ratios, were testedto determine which bE:st aelineatedgypsy moth defol i ati on. Responsesurface analyses were conducted todetermine optimal threshold level sfor the individual transformedbands ana band combinations.Results inoicate that, of the threetrans forma ti ons i nves t i ga ted, adi fference of rati os (band 7/band0) transformati on most accuratelydel i neateu forest change due togypsy moth activity. Band 5(U.6-U.7 micrometers) ratioed datadid nearly as well, however, othersingle banos and band combinationsdid not improve upon the band 5ratio results.

21

7-3HIGH-ALTITUDE RADAR ASSESSMENT OFTHE DAMAGE CAUSED BY THE VOLCANIC

ERUPTION OF MOUNT ST. HELENSR. D. Dillman

Lockheed Engi neeri ng and ~lanagementServices Company, Inc.

R. E. Hinkl eu.S. Air Force, Washington, D.C.

Following the volcanic eruption oftvlOunt St. Helens on May 18, 1980,the surrounding area was obscuredby varyin~ amo~nt of clouds and ashfor 30 days. A total view of theoamaged area was neededimmediately. This need was metwithin 3 days by acquiringhigh-altitude side-looking radarimagery. This imagery was analyzedusi n9 only the characteri sti cs ofthe radar returns in conjunctionwith preeruption high-altitudephotography. The analyst was ableto establ ish the areal extent ofthe changes in 1akes, topography,and damage to timber caused by theeruption. The three radarcondition maps were compared toposterupti on photography coll ectedon June 19, 1980, and other damagecondition maps. These comparisonsshow good agreement for both theboundaries between classes ofdamage and the types of damagedefi ned by the radar imagery. A·major factor. in the totalexp 1oi tati on of the radar imagerywas the availability of imageanalysts trained and experienced ininterpreting the characteristics ofradar returns from naturalvegetation. This study shows thathigh-resolution radar data canprovide important information onthe damage to 1arge areas whenobscuration prevents the use ofother types of imagery.

AgRISTARS MINI-SYMPOSIUM ABSTRACTS

7-4OKLAHOMA MID-CYCLE TIMBER

INVENTORY PILOT TESTB • B • Eav

Lockheed Engineering and ManagementServices Company, Inc.

R. E. HinkleU. S. Air Force, Washington, D.C.

C. E. ThomasU. S. Forest Service

A large scale demonstration of aforest survey technology which useshigh altitude panoramic photographyin conjunction with two-phasesampling was conducted in Oklahomaduring the 1981 mid-cycle timberinventory update. The resul tsindicated that 4.3 of the 10.5million acres of eastern Oklahomaare in corrmercial forest land. Thetotal growi ng stock vol ume wasestimatea to be 2.0 billion cubicfeet.The fact that the growing stockvol ume estimate has a standarderror (4.6 percent) only 2.2percent 1arger than that of thefull-cycl e survey wi th only tenpercent of field work is anindication of the efficiency of thesystem.This project demonstrated that highaltitude panoramic photography incombination with two-phase samplingcan be used to efficiently satisfythe requi rements for mi d-cyc 1etimber inventory updates.

8-1INVENTORY OF SOIL CONSERVATIONPRACTICES USING REMOTE SENSING

R. H. Griffin, IINASA/NSTl/ERL

An examination of the size andshape of the 119 soil conservationpractices listed by the USDA/SCS inthe National Handbook ofConservation practices was made.

28

This investigation revealed thatabout 20 could be detected by ascanner with the resol ution of thethematic mapper. Although thereare 22 practices that cannot bedetected using remotely senseddata, it is likely that themajority of the remaining practices(77) coul d be detected w; th ascanner \'lith an Instantaneous-Field-of-View (IFOV) of about 15meters.Another area of investi gation hasfocused on the use of remotelysensed data and other mappedinformation to identify potenti alerosion hazards. For the studyareas that have been invest;gatedto date, elevation informationobtained from the NCIC tapes(1:250,000 topography maps) havenot been adequate for detenHi ningslopes zones for erosion hazardassessment. However, the use ofdigitized soil survey data alongwith Landsat MSS data have beenused to identify areas wherepotential erosion hazards exist.

8-2BUILDING A BRIDGE BETWEEN

REMOTE SENSING AND HYDROLOGICMODELS

Eugene L. PeckHydex CorporationEdward R. Johnson

Georgia TechThomas N. Keefer

Sutron CorporationRemote sensing technologies provideindirect measurement of landcharacteri stics, vegetative cover,and the states of water in thehydrologic cycle. However, the useof this valuable infonnation formodeling the hydrologic cycle hasbeen very 1imited. Thi s 1imiteduse is a result of two factors.First, there is not a one-to-onecorrespondence between parametersand states of hydrologic models and

AgRISTARSMINI-SYMPOSIUMABSTRACTS

cOrllfilon remotely sensed vari abl es.The second, is our i nabil ity toeffectively combine remotely sensedi nformati on wi th standard~easurements to improve esti@atesof mean areal values of hydrologicvariables.

The above 1imi ti ng factors are thesubject of completed and on-goingNASA contracted research and arereported here. The studies havebeen accompl i shed in four parts: '

o Review of Hydrologic ~Iodelsfor Evaluating Use of RemoteSensing Capabilities

o Strategies for Using RemotelySensed Data in Hydrologicf.1odels

o Combining Remotely Sensed andOther Measurements forHydrologic Areal Averages, and

o Updating of Hydrologic ModelsUsing Remotely SensedMeasurements

Objecti ves of the research arediscussed and include developmentof techni ques for (1) improvementin operational hydrologicforecasting, (2) enhanced knowledgeof the states of water in the1i thosphere for drought and cropyield studies, and (3) improvedestimates of mean areal averagevalues of hydrologic variables forwide agricultural application.

8-3THE STRUCTURINGOF A REMOTESENSINGBASEDCONTINUOUSSTREAMFLOWMODEL

J. R. Groves and R. M. RaganUniversity of Maryland

landSat remote sensing has beensuccessfully usea to provi de 1andcover parameters for single eventhyaro10gic models. The potentialof space platform remote sensing toprov i de other data for thesynthesi s of conti nuous hYdrol ogi cprocesses is rapidly advancing, but

29

applications are limited by theabsence of model s structured toaccept the newly available oranticipated remotely sensed data.

The paper describes the developmentof the structure and testingprogram for a physically basedcontinuous streamflow modelspecifically designed toincorporate i nfonnati on obtai nedfrom space platform sensor systems.The 1i nkage and operati ng conceptsare similar to those of theestablished Stanford WatershedModel family in that the objectiveis to route the incoming rainfallthrough a series of submode1s thatsimulate individual hydrologicprocesses to produce estimates ofthe daily or hourly streamflow andthe redi stri buti on of moi sturestorage within the drainage basin.Too many of the components in theStanford Model family are based onregressi on ana1ysi s and cannot beestimated in tenns of measurablequanti ti es. Each component in theproposed model is physically basedand optimized to interface withremote sensing capabilities. Allinput data, both satell i te andground based, are incorporated intothe model through a gri d cellgeographi ca1 i nfonnati on system.The data base, designed to bedeveloped from digital imagery fromlandsat, TIROS-N, and GOESsatellite systems is to includeland cover, solar radiation, snowcover, vegetati ve stress, cloudcover, temperature estimates, andother physical quantities relativeto the synthesis of streamflow fromprecipitation. The strategy in thedevelopment of the components usedto simulate individual hydrologicprocesses is to use extensivenumeri cal experiments with comp1exhighly theoretical models to evolvecomputati onally effi ci ent func-tional relationships that willstimulate the individual.

AgRISTARS MINI-SYMPOSIUMPOSTER DISPLAY

Pl-lFLUOD STRESS MAPPED BY SATELLITE

HWEXDee G. ~~Crary, NOAA/EUIS/CIAD

P2-1UTILITY OF TM DATA FOR FEATUREIDENTIFICATION/DISCRIMINATIONR. R. J. Mohler, D. B. Ramey,

and C. L. DaileyLockheed Engineering and Management

Services Company, Inc.

P2-2SPECTRAL WAVELENGTH CHARACTERISTICS

OF TM, MSS, AVHRR(Inventory Technology Development

Project Quarterly TechnicalIntercnange Meeting)

P2-3JSC TM QUICK-LOOK STUDIES SCHEDULE

(Inventory Technology DevelopmentProject Quarterly Technical

Interchange Meeting)

P2-4TN IMAGER Y--A ~lAP BASE FOR GROUND

DATA COLLECTIONc. R. QuinonesLockheed Engineering and Management

Services Company, Inc.

P2-5T"'IDATA QUALITY

M. M. Smyrski, H. G. Smith,E. R. f'ilagness

Lockheed Engineering and ManagementServices Company, Inc.

30

P2-6, 7THEMATIC MAPPER RESOLUTIONw. F. Palmer

Lockheed Engineering and ManagementServices Company, Inc.

P2-8AUTOMATED VEGETATION CLASSIFICATION

USING SIMULATED AND ACTUALTM DATA

K. S. Nedelman, R. B. Cate,Lockheed Engineering and Management

Services Company, Inc.and R. M. Bizzell, NASA/JSC

P2-9LABELING TARGET DEFINITION

IMPROVED BY INCREASED TM SPATIALRESOLUTION

K. S. Nedelman and R. B. CateLockheed Engineering and Management

Services Company, Inc.

P2-10LABELING AND CLASSIFICATION

IMPROVED BY INCREASED SPECTRALCOVERAGE

K. S. Nedelman and R. B. CateLockheed Engineering and Management

Services Company, Inc.

P2-11INVESTIGATIONS OF TM DATA

(7-BANDS)K.S. Nedelman and R. B. Cate

Lockheed Engineering and ManagementServices Company, Inc.

AgRISTARS MINI-SYMPOSIUMPOSTER DISPLAY

P2-lLAUTUMATED At~LYSIS UF SINGLE

DATA TM SCENEK. S. Neaelman and R. B. Cate

Lockheed Engineering and Manage~entServices Company, Inc.

P2-13TM FEATURE SPACE ANALYSISJ. M. Jones, J. H. ~nith,

W. S. KossackLockneed Engi,eering and Management

Servic~s Company, Inc.

P2-14HI DATA--A MAPPING TOUL

C. L. DaileyLockneed Engi neeri ng ana l'4anagement

Servi ces Company, Inc.

P2-1bAUSTRALIAN DATA CATALOG

(1981-82 GROUND DATA COLLECTION)C. R. Quinones

Lockheed Engineering and ManagementServices Company, Inc.

P2-16CHARACTERIZATIUN OF AUSTRALIAN

LANDSAT MSS DATAE. R. Magness

Lockheed Engineering and ManagementServices Company, Inc.

P2-17SG-l AUTOMATED ESTIMATION OF SMALL

GRAINS PROPORTIUNSw. F. Palmer

Lockheed Engineering and ManagementServices Company, Inc.

31

P2-18A NON-PARM~ETRIC CLUSTERING

TECHNIQUED. B. Rame.y

Lockheed Engineering and ManagementServices Company, Inc.

P2-19CULTURAL AND ENVIRONMENTAL EFFECTS

ON CROP SPECTRAL DEVELOPMENTPATTERNS AS VIEWED BY LANDSAT

E. P. CristEnvironmental Research Institute

of Michigan

P2-20A COMPARATIVE ANALYSIS OF THE

MULTISPECTRAL SCANNER, ADVANCEDVERY HIGH RESOLUTION RADIOMETER,

AND COASTAL ZONE COLOR SCANNERFOR VEGETATION MONITORING

R. C. Cicone, and H. D. MetzlerEnvironmental Research Institute

of Michigan

P2-21APPLICATION OF U.S. BASED ANALYSIS

APPROACHES TO ARGENTINA CROPIDENTIF ICATION

C. M. Hay, J. B. Odenwel1er,University of California

at Berkeleyand B. L. Wood,

Technico1or Government Services

P3-1MODEL PREDICTING SOYBEAN GROWTH

STAGES FROM DAYLENGTH, TEMPERATURE,WATER STRESS, AND MATURITY GROUP

T. Hodges, University of MissouriColumbia, Missouri

AgRISTARS MINI-SYMPOSIUMPOSTER DISPLAY

P5-1BACKSCATTEK FRO/·1A VEGETATION

LAYER: TWO APPROACHEST. J. Schmugge, NASA/GSFC

J. A. Kong, MassachusettsInstitute of Technology

and R. A. Lang, WashingtonUniversity

P5-2RE~UTE SENSING OF CANOPY PHYTOMASS

ANU WATER CONTENTS. E. Hollinger, V. C. Vanderbilt,

C. S. T. DaughtryPurdue University

P5-3RAUAR RESULUTION REQUIREMENTS FOR

SOIL MOISTURE ESTIMATION FROMSPACECRAFT SYNTHETIC APERTURE

RADARM. C. Dobson, F. T. Ulaby,

S. MaezziRemote Sensing Laboratory,

University of Kansas

PS-4SUIL MOISTURE UNCERTAINTIES AND

THEIR AFFECT ON REMOTE SENSINGINTERPRETATIONS

T. J. Schmugge, NASA/GSFC,T. Mo, Computer Science Corp at

GSFC, and M. Owe, NASA/GSFC

PS-SAIRCRAFT MICROWAVE RESULTS OVER

USDA WATERSHEDST. Jackson, USUA/ARS, P. O'Neill,

and T. J. Schmugge, NASA/GSFC

32

P5-6ESTIMATE OF SOIL MOISTURE STORED IN

ROOT ZONE BY SURFACE MEASUREMENTSP. Camillo, Computer Science Corp.

at GSFC, and T. J. Schmugge,NASA/GSFC

P5-7SOIL AND ATMOSPHERE BOUNDARY LAYER

MODEL FOR EVAPORATION AND SOILMOISTURE STUDIES

P. Camillo, D. Gurney, NationalAcade~ of Science at GSFC, and

T. J. Schmugge, NASA/GSFC

P5-8PASSIVE MICROWAVE IMAGERY OF THE

U.S. FROM SATELLITE AT FREQUENCIESBETWEEN 6.6 uh 37 GHz.

E. G. Njoku and B. Gochman,Jet Propulsion laboratory-CA

P5-9RE~OTE SENSING AND THE SCS RUNOFF

CURVE NUMBERTom J. Jackson, USDABeltsville, Maryland

PS-IOMICROWAVE MEASUREMENT OF SOIL

MOISTURET. J. Jackson and E. Engman

USDA/Beltsville, Maryland

PS-llMONITORING MOISTURE CONDITIONS WITH

INFRARED REMOTE SENSINGJohn C. Price, USDA/ARS,

Hydrology laboratoryBeltsville, Maryland

AgRISTARS MINI-SYMPOSIUMPOSTER DISPLAY

P6-1THE USE OF A COUNTY-WIDE DIGITAL

DATA ~ASE FOR SOIL EROSIONPREDICTION

Michael A. Spanner, TechnicolorGovernment Services, Inc., Ames

Research Center; James A. Brass,ana uavid L. Peterson, NASA/Ames

Research Center

P6-2ANALYSIS OF DATA ACQUIRED BYSYNTHETIC APERTURE RADAR AND

LANUSAT MULTISPECTRAL SCANNER OVERKlRSHAW CLUNTY, SOUTH CAROLINA

DURING SUMMER SEASONS. T. Wu, NASA/NSTL

P7-1DEMONSTRATING USE OF GIS FOR

PLANNING ON NATIONAL FOREST LANDSM. L. Mathews,

Lockheed Engineering and ManagementServices Company, Inc.

P7-2INFORMATION CONTENT OF THEMATIC

MAPPER SIMULATOR DATA IN AFORESTED REGION

James A. Brass,NASA/Ames Research Center;

Michael A. Spanner,Technicolor Government Services/

Ames Research Center;Oavid L. Peterson,

NASA/Ames Research Center; andJoseph J. Ulliman,

College of Forestry, University ofIdaho

33