189_XXIX-part7

8/4/2019 189_XXIX-part7

1/8

A CONCEPTUAL FRAMEWORK FOR ESTIMATING CROP GROWTHUSING OPTICAL REMOTE SENSING DATAJ.G.P.W. Cleversl, RA.M. Bouman2, C. Bukerl & H.J.C. van Leeuwen!

lWageningen Agricultural University, Dept. of Landsurveying & Remote Sensing, The Netherlands2Centre for Agrobiological Research (CABO-DLO), Wageningen, The Netherlands

Commission number VIIABSTRACTFo r monitoring agricul tural crop production, growth of crops has to be studied, e.g. by usingcrop growth models. Estimates of crop growth often are inaccurate for non-optimal growing conditions.Remote sensing ca n provide information on the actual status of agricul tu ra l crops. This informationca n be used to in i t i a l i ze , (re)parameterize, calibrate or update crop growth models, an d i t ca nyield parameter estimates to be used as di rec t input into growth models: 1) lea f area index (LAI);2) lea f angle dist r ibut ion (LAD); and 3) leaf colour (opt ical propert ies in the PAR region).LAI and LAD determine th e amount of l igh t intercept ion. Leaf (o r crop) colour influences th ef ract ion of absorbed photosynthet ical ly act ive radiat ion (APAR) and th e maximum (potent ial) rateof photosynthesis of the leaves.In this paper th e above concepts fo r crop growth estimation will be fur ther elucidated and i l lust ratedwith examples fo r sugar beet using groundbased and airborne data obtained during the MAC Europe1991 campaign. A simple ref lectance model was used fo r estimating LAI. Quant i tat ive informationconcerning LAD was obtained by measurements at two viewing angles. The re d edge index was usedfor estimating th e leaf optical properties. Finally, a crop growth model (SUCROS) was re-parameterizedon t ime-series of optical ref lectance measurements to improve the simulation of beet yield.KEY WORDS: Growth models, Optical remote sensing, Growth estimation

1. INTRODUCTIONRemote sensing techniques have th e potent ial to provideinformation on agricul tural , crops quantitat ively ,instantaneously and, above a l l , non-destructively overlarge areas. In the past decades, knowledge aboutoptical remote sensing techniques and their applicationto fields such as agriculture ha s improved considerably(cf . Asrar, 1989; Steven & Clark, 1990). A lo t ofresearch has been devoted to land cover c lass i f i ca t ionand acreage estimation with considerable success.Another f ie ld of in te res t in agriculture is yieldest imat ion. Research in this topic , however, ha sindicated tha t remote sensing alone i s generally notcapable to produce accurate yield est imat ions. Thisha s prompted sc ient i s t s to look fo r other techniquesthat can be combined with remote sensing data to givebet te r resul ts . One of such techniques is crop growthmodelling.In th is paper, some views on l inking optical remotesensing data with crop growth models are presented.Some concepts wil l be i l lus t ra ted with preliminaryresul ts from th e MAC Europe 1991 campain over th e Dutcht e s t s i t e Flevoland.1 .1 Crop Growth ModelsFrom the 19th century, agricultural researchers haveused modelling as a tool to describe relat ions betweencrop growth (yield) an d environmental factors thatappear to govern crop growth. In our study, we makeuse of models and concepts that were developed in TheNetherlands by de Wit an d his co-workers (de Wit, 1965;Penning de Vries &van Laar, 1982; van Keulen &Wolf,1986; va n Keulen & Seligman, 1987; Spit ters et a l . ,1989). This type of models descr ibes the relat ionbetween physiological processes in plants an denvironmental factors such as so lar i r radiat ion,temperature an d water an d nutr ien t availabil i ty . Themodels compute th e daily growth an d development rateof a crop, simulating the dry matter production fromemergence t i l l maturity. Finally, a simulation of yielda t harvest time is obtained. The basis for thecalculat ions of dr y matter production is th e ra te of

189

gross CO2 assimilat ion of the canopy. Input datarequirements concern mainly crop physiologicalcharacter is t ics (e.g . maximum CO2 assimilat ion ra te ,resp i ra t ion and dry matter parti t ioning), s i techaracter is t ics ( lat i tude) , environmental characterist ic s (dai ly i r rad ia t ion , daily minimum an d maximumtemperatures) and the i n i t i a l conditions defined bythe date a t which th e crop emerges.The main driving force for crop growth in these modelsis absorbed solar radiat ion, and a lo t of deta i l isgiven to th e modelling of the solar radiat ion budgetin the canopy. Incoming photosynthet ical ly act iveradiat ion (PAR z 400-700 nm; more or less synonymouswith visible radiation) is f i r s t part ly ref lected bythe top layer of the canopy. The complementary fract ioni s potent ial ly avai lable for absorption by th e canopy.Subsequently, th e f ract ion of absorption by the canopyi s a function of solar elevation, leaf area index (LAI) ,leaf optical properties and crop extinction coeff icientsfor diffuse and di rec t f luxes (which in the i r turndepend on solar elevation, leaf angle distr ibut ion (LAD)and leaf optical properties). The product of th e amountof incoming photosynthet ical ly act ive radiat ion (PAR)an d th e absorptance yields the amount of absorbedphotosynthetically active radiation (APAR). The rateof CO2 assimilation (photosynthesis) is calculated fromth e APAR and th e photosynthesis- l ight response ofindividual leaves (Fig. 1). The maximum rate ofphotosynthesis a t l igh t saturat ion is highly correlatedto the l eaf nitrogen content (Fig. 2). The assimilatedCO 2 is then reduced to carbohydrates which ca n be usedby the plant for growth.Because of th is detailed modelling of th e solarradiat ion budget, this type of models is especiallysui table for the l inkage with optical remote sensingthrough th e us e of optical ref lectance models.1.2 Optical Remote Sensin&Crop growth models as described above were developedto formalize an d synthesize knowledge on the processesthat govern crop growth. When applied to operat ionaluses such as yield estimation, these models often appear

8/4/2019 189_XXIX-part7

2/8

10 0 200 30 0

Slay .1965 va n Lear a. Penning de Vries ,1972)( va n Loar 8. Pennmg de VrIes 1972b. Winzeler. 1980

40 0 50 0irrOdlonce (J rn- 2 S ~ l ) Figure 1. Relat ion between i r radiance an d ra te of grossCO 2 ass imi la t ion fo r ind iv idua l l eaves of wheat . From:van Keulen & Seligman, 1987, p. 43.ro te of ne tCO, assimilat ion1.5 (m g m- 's - ' )

10

0.5, -,

e" 0." .,'.0.01 0.02

0, .

:- Q a aa Q 00 a - 8 ::0 0

g' Q) .. :s o@

J' oo. 00 "0 . 4 . . :' : ~ o o : . . ..

0.03

:"\1 0 ', ," ....0.04 0.05 0.06 0.07

ni trogen content (kg kg')Figure 2. Relat ion between n i t rogen content in th e l ea f ,on a dry weight bas is , and i t s ra te of ne t CO 2ass imi la t ion . The di f fe rent symbols r e f e r to measure-

ments made by di f fe rent au thor s . From: van Keulen &Seligman, 1987, p. 47.to f a i l when growing condi t ions a re non-opt imal (e . g.pes t an d d i sease inc idence , sev ere drought , f r os tdamage). Therefore , fo r yie ld es t imat ion , it i snecessary to ' check ' modelling r e su l t s with some sor tof informat ion on th e ac tua l s ta tus of the cropthroughout the growing season (Bouman, 1991). Opt ica lremote sensing can provide such informat ion. There arethree ' key- fac tor s ' useful in crop growth models tha tmay be der ived from op t ica l remote sensing da ta : a)LAI; b) LAD; an d c) l e a f op t ica l p roper t ie s in th e PARregion.Ad a . The LAI during the growing season i s an importantvar iab le in crop growth modell ing. Also , th e LAI i sa major factor determining crop ref lectance an d i s of tenused in crop ref lectance model l ing (e .g . Su i t s , 1972;Bunnik, 1978; Verhoef, 1984). Th e es t imat ion of LAIfrom remote sensing measurements has received mucha t t e n t i on . Much research has been aimed a t determiningcombinations of re f lec tances , so -ca l led Vegeta t ionIndices , to correct for the effect of disturbing factorson the r e la t ionship between crop re f lec tance an d cropcha rac te r i s t i c s such as LAI (Tucker, 1979; Richardson&Wiegand, 1977; Clevers, 1988, 1989; Bouman, 1992a).Ad b. LAD ( leaf angle d is t r ibu t ion ) a f fec ts the processof crop growth because it has an e f f e c t on th ein terception of APAR by th e canopy. With op t i ca l remotesensing techniques it has been more di f f i cul to obta inquanti ta t ive information on LAD than on LAI. A solut ionmay be found by performing measurements a t di f fe rentviewing angles. Goel &Deering (1985) have shown tha t

190

Reflexion (685 nm) /%)7 0 r - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ~ 60 "5 0 ~ : : 40 r\30 , .2010 'I. , . - . : :00 10 20 30 40 50 60Chlorophyll a [ ~ g / c m 2 1

Figure 3. Rela t ion between red re f lec tance an dchlorophyll content of th e upper leaves of wheat crops.From: Schel lberg, 1990, p.75.measurements a t two viewing angles for f ixed sola rzenith and view azimuth angles are enough to al lowes t imat ion of LAI an d th e UD by th e in f ra redre f lec tance .Ad c. Leaf op t ica l propert ies ( l ea f colour) areimportant in th e process of crop growth because:1) they inf luence the f rac t ion of absorbed PAR, an d2) they can be ind ica t ive fo r th e nitrogen s ta tus (orchlorophyll content) of leaves which a f fec ts the maximumr a te of photosynthesis . Fig. 3 gives an example of are la t ion between l e a f ref lectance an d l eaf chlorophyl lcontent of a wheat canopy (Schel lberg, 1990) . Leafchlorophyl l content in i t s turn i s r e la ted to l e a fnitrogen con ten t . Leaf op t ica l p roper t ie s in the PARregion may be ascerta ined by spec t ra l measurements inth e v i s i b l e region (VIS) of the elect romagnet ic (EM)spectrum. However, a t lo w so i l cover th e measured signalwil l be confounded by s o i l inf luence. As a r e sul t , thes igna ls from a crop and from th e so i l background ared i f f i c u l t to separate , unless the spec t ra l so i ls igna tu re i s known so tha t a cor rec t ion of th e s igna lsi s possible . At complete coverage spect ral measurementsin the VIS of fe r informat ion only on l e a f colour .However, s ince the s igna l in VIS a t complete coveragei s r e l a t i ve ly low, it may be heavi ly confounded byatmospheric e f fec t s fo r which must be corrected. Onlyf ie ld (spectro )radiometers provide undistor ted s ignals .

Another measure of chlorophyll content may be offeredby th e so-cal led red edge index (blue sh i f t ) (Horle re t a l . , 1983) . The pos i t ion of the red edge i s def inedas the pos i t ion of the main in f lex ion po in t of the redin f ra red s lope . A decrease in l e a f chlorophyl l contentre su l t s into a s h i f t of the red edge towards th e blue.

2. LINKING OPTICAL REMOTE SENSING WITH GROWTH MODELS2.1 FrameworkTwo methods can be dis t inguished to l ink op t ica l remotesensing data with crop growth models. In the f i r s tmethod, c a l l ed I crop parameter es t ima t ion I, cropparameters are est imated from op t ica l remote sensingan d ' f ed ' into a growth model as input o r forcingfunc t ion . Mostly, crop parameters t ha t have been usedsuccesful ly so far are measures fo r th e fract ional l i gh tin te rcept ion by the canopy, namely LAI an d s o i l cover(Steven e t a l . , 1983; Kanemasu e t a l . , 1984; Maas, 1988;Bouman & Goudriaan, 1989). However, parameters l ikeLAD an d l e a f colour can also be used as i npu t in moreelaborate growth models such as those described in th i spaper. In sec t ion 2.3, some methods for der iv ing theseparameters from op t ica l remote sensing da ta wil l bepresented.In the second method, ca l led 'model re -pa ramete r iza t i on ' , crop growth models are re-parameter ized on t ime-

8/4/2019 189_XXIX-part7

3/8

se r ie s of remote sensing measurements. Maas (1988)presented a method in which crop growth model parameterswere ad jus t ed in such a way tha t s imula ted va lues ofLAI by the growth model matched LAI values t ha t werees t imated from ref lectance measurements. Bouman (1992b)developed a procedure in which remote sensing models(a .o . opt ica l re f lec tance) were l inked to crop growthmodels so tha t canopy ref lectance was simulated togetherwith crop growth. The growth model was then re-parameter ized to match s imula ted va lues of canopy re f lec tanceto measured values of re f lec tance . In a case study fo rsugar beet , the s imulat ion of (above-ground) biomasswas more accurate a f t e r re -parameter iza t ion than beforere-parameter izat ion on re f lec tance measurements. Th ere-parameter izat ion in th i s procedure i s governed bythe parameters which l ink th e crop growth model an dth e remote sensing model (s o fa r mainly LAI). Therefore,LAI, LAD an d l e a f colour , can in p r inc ip le be used inth e re-parameter izat ion when they are made exp l i c i tin both the crop an d th e op t ica l remote sensing model.Especial ly fo r LAD and l e a f colour , however, th i s wi l lt ake fu r the r r esearch .Cent ra l to our approach, in both methods, i s t ha t wemake use of so -ca l led / explanatory ' models as much aspossible . For crop growth th i s means th e use of modelst ha t are based on an unders tanding of physical an dphysiologica l processes of crop growth, and fo r opticalremote sensing the use of models tha t are based on anunders tanding of th e in te rac t ion of sola r r adia t ionwith vegeta t ion canopies . In the next sect ion the modelsused are br i e f ly descr ibed.2.2 ModelsSUCROS. In t h i s s tudy the used crop growth model wasSUCROS (Simple an d Universa l CROp growth Simula tor ;Spit ters e t a l . , 1989). I t is a mechanistic growth modeldescr ib ing the potent i a l growth of a crop as a func t ionof i r radia t ion , a i r temperature an d crop cha rac te r i s t i cs . The l i gh t pro f i le within a crop canopy is computedon th e bas is of the LAI an d th e ex t inc t ion coe f f i c i ent .At se lec ted t imes during the day and a t selected depthswithin the canopy, photosynthes i s i s calculated fromth e photosynthesis- l ight response of individual leaves.Integrat ion over th e canopy layers an d over time withinth e da y gives the da i ly ass imi la t ion ra te of the crop.Assimi la ted matter i s used fo r maintenance r e spi ra t ionand fo r growth. The newly formed dry matter i spa r t i t ioned to th e var ious p lan t organs . An importantvar iab le t ha t i s s imula ted i s th e LAI, since theincrease in l e a f area contr ibutes to next day ' s l i g h tin te rcep t ion an d thus r a te of ass imi la t ion .SAIL Model. Th e one- layer SAIL model (Verhoef, 1984)s imula tes canopy re f lec tance as a funct ion of canopyvar iab les (LAI, l e a f angle dis t r ibut ion an d l e a fref lectance an d transmi t tance) , so i l ref lectance, r a t iod i f f us e / d i r e c t i r radia t ion an d solar /view geometry( sola r zen i th angle , zenith view angle an d sun-viewazimuth angle) . Recent ly, th e SAIL model has beenextended with the hot spot e f fec t (Looyen e t a l . , 1991).Leaf i nc l i na ti on d i s t r i bu t i on funct ions used with theSAIL model are given by Verhoef & Bunnik (1981).PROSPECT Model. The PROSPECT model, as developed byJacquemoud & Baret (1990), is a radiat ive t r ans fe r modelfor individual leaves . I t i s based on th e genera l i zed"plate model" of Al len e t a1 . (1969, 1970), whichconsiders a compact theore t i ca l p lan t l e a f as at r ansparen t pla te with rough plane para l l e l surfaces .An ac tua l l e a f i s assumed to be composed of a pi le ofN homogeneous compact l aye rs separated by N-l a i rspaces . Th e compact l e a f (N = 1) has no i n t e r ce l l u l a ra i r spaces or the i n t e r ce l l u l a r a i r spaces of themesophyll have been inf i l t ra ted with water . The d isc re teapproach can be extended to a continuous on e where Nneed not be an in tege r . PROSPECT al lows to compute th e

191

400-2500 nm re f lec tance an d t ransmit tance spec t ra o fvery d i f fe ren t leaves using only three input var i ab les :l e a f mesophyll s t ruc tu re parameter N, chlorophyl lcontent an d water content . In the v is ib le region, mainlyl e a f chlorophyl l content determines l e a f opt ica lproper t ie s .

2.3 Crop Parameter Est imat ionLAIClevers (1988, 1989) has descr ibed a s impl i f ied , semiempir ical , re f lec tance model fo r est imat ing LAI of agreen canopy. Fi r s t , a WDVI (Weighted DifferenceVegeta t ion Index) i s ascerta ined as a weighteddif ference between th e measured NIR an d red ref lectancesin order to cor rec t fo r so i l background:

(1)with rir = t o t a l measured NIR re f lec tance ; rr = t o t a lmeasured red re f lec tance ; an d C = rs,ir/rs,r (rs,ir = NIRre f lec tance of the so i l ; rs,r = red re f lec tance of th eso i l ) .Secondly, th i s WDVI i s used fo r est imating LAI accordingto the inverse of an exponent ia l funct ion:LAI = - l / a . I n ( l - W D V I / W D V I ~ ) (2 )Parameters a and W D V I ~ have to be est imated empir ical lyfrom a t r a in ing se t , but they have a physicalin te rpre ta t ion . Another poss ib i l i t y i s to use a canopyre f lec tance model fo r theore t i ca l ly determining theparameters once the input fo r th e canopy re f lec tancemodel i s known (e .g . l e a f op t ica l p roper t ie s an d LAD).Fig. 4 shows simulat ion r e s u l t s using the SAIL modeli l lus t ra t ing th i s concept . A sensi tiv i y analysis usingthe SAIL model revealed tha t the main parameterinf luencing th e r e la t ionship between WDVI an d green( ! ) LAI was the LAD (Clevers & Verhoef, 1990; Clevers ,1992) .Bouman e t a l . (1992) a r r ived a t th e same formula t ionof the re la t ionsh ip between LAI an d WDVI through as imi la r l i ne of reasoning. They empi r i ca l ly foundcons is ten t parameters fo r var ious years , loca t ions ,cul t iva r s and growing condi t ions fo r some mainagr icu l t u r a l crops , sugar beet , pota to , wheat , barleyan d oats (Uenk e t a l . , 1992).

87 00000 spherical

00000 planophile6 11111111111 erectophile5

432

o ~ ~ ~ ~ - - , , - - ' - - . - - - r - - . - - ' - - - r - - ' - - - . - - . o 10 20 30 40 50 60WDVI (%)

Figure 4. Inf luence of LAD on a regression of LAI onWDVI as s imula ted wi th the SAIL model. From: Clevers& Verhoef, 1990, p.8 .Leaf angle distribut ion (LAD)Since th e LAD i s on e of the main parameters inf luencingth e r e la t ionship between WDVI an d LAI, informat ion ont h i s parameter i s very important . In Looyen e t a l .(1991) the poss ib i l i t i e s of acquir ing informat ion onboth LAI an d LAD by means of a so -ca l led dual look (two

8/4/2019 189_XXIX-part7

4/8

viewing angles) concept are i l lus t ra ted . Fig. 5 showsa nomograph of the simulated WDVI (SAIL model) a t anoblique viewing angle (52) plot ted against th esimulated WDVI a t nadir viewing fo r several LADs andLAI values. By plott ing measured WDVI values in to thisnomograph, an estimate of both LAI and LAD is obtained.

100,,-...

80W:=J 60o:Jmo 40I:>o:s 20

20 40 60 80 100WDVI-NADIR (%)Figure 5. Nomograph i l lus t ra t ing th e inf luence of LAIand LAD (in this graph each LAD consis ts of jus t onelea f angle) on th e WDVI measured from nadir and th eWDVI measured a t an oblique viewing angle of 52. Solarzenith angle of 36 and azimuth angle between planeof observation and su n of r as simulated with the SAILmodel (measurements are of MAC Europe 1991, CAESAR overf l igh t July 4th - Julian da y 185 -, 13.30 GMT).Leaf colourAs stated before, in pract ice i t will be very dif f icu l tto ascer ta in leaf colour unless leaves are analyzedin th e laboratory. A more pract ical measure may beoffered by the red edge index. However, Clevers & Buker(1990) have shown that this index is determined by bothLAI an d lea f colour (related to lea f chlorophyllcontent) . Two (independent) measurements are thereforeneeded, one more re la ted to LAI (l ike WDVI) and onemore related to chlorophyll content (l ike re d edgeindex). Fig. 6 i l lus tra tes th e simulated inf luence ofLAI and lea f chlorophyll content on th e posit ion ofthe red edge and th e WDVI (using a combined PROSPECT-SAIL model). In this study, the method of Guyot & Baret(1988) fo r determining th e posi t ion of th e re d edgewas appl ied, using only four wavelength bands. Firs t ,

740

E 7 3 0cCl)CJ"l720-0Q)-0 710

.2

80....

6 ~ 0 ) 0 C.330 82015

10 .S2fj

7 0 0 ~ - ' - ' - ' - ' - . ~ r - . - r - ~ ~ ~ ~ o 10 20 30 40WDVI (%) 50 60

Figure 6. Nomograph i l lus t ra t ing th e influence of LAIand lea f chlorophyll content on th e WDVI (from nadir)and th e posit ion of the red edge as simulated with acombined PROSPECT-SAIL model (measurements are of MACEurope 1991, Caesar overfl ight July 4th - Ju l ian da y185 - (WDVI), and AVIRIS overf l ight July 5th (rededge .

192

they estimated the ref lectance value a t the inflexionpoint halfway minimum (a t 670 nm) and maximum (a t 780nm) ref lectance. Secondly, they applied a l inearinterpolation procedure between th e measurements a t700 nm and 740 nm for estimating the wavelengthcorresponding to the estimated reflectance a t th einf lexion point . Measurements of the WDVI and th eposi t ion of the red edge may be combined for ascer ta i ning th e leaf chlorophyll content.3. MAC EUROPE CAMPAIGN - FLEVOLAND TEST SITESome of the above principles fo r l inking optical remotesensing with crop growth models will be i l lus t ra tedwi th preliminary resul ts from th e European mul t isensorairborne campaign MAC Europe in 1991. A descriptionwill be given of th e MAC Europe campaign in th e Dutchte s t s i te Flevoland and of the collected remote sensingand ground t ruth data.In th e MAC Europe campaign, in i t ia ted by the NationalAeronautics and Space Administration (NASA) an d theJet Propulsion Laboratory (JPL) , both radar and opticalairborne measurements were made over selected te s t si tesduring th e growing season of 1991. One of th e te s t si teswas Flevoland in the Netherlands.In th e optical remote sensing domain, NASA executedone overfl ight with th e AVIRIS scanner ( fo r systemdescript ion, se e Vane e t al . , 1984). In addit ion, th eDutch experimenters flew three fl ights with the DutchCAESAR scanner (for system descript ion, see Looyen e tal . , 1991). The radar observations made during MACEurope do not make part of this study and will no t beconsidered here ( the synergism between remote sensingdata from different sensors is th e topic of anotherstudy) .Test s i t e . The t e s t s i t e was located in SouthernFlevoland in th e Netherlands, an agricul tural area withvery homogeneous soils reclaimed from th e lake"IJsselmeer" in 1966. The tes t s i te comprised te ndifferent agricultural farms, 45 to 60 ha in extension.Main crops were sugar beet, potato and winter wheat.Due to hailstorms and night-frost damage of th e sugarbeet in April '9 1 some of the sugar beet fie lds weresown for a second time in la te April resu l t ing intosome yield differences a t th e end of th e season.Ground t ruth. Crop parameters concerning acreage,variety, planting date, emergence date, fer t i l izat ion,harvest date , yield and occurring anomalit ies werecollected for th e main crops (Buker e t a l . , 1992).During th e growing season, addit ional parameters weremeasured in th e fie ld. The selected parameters werethe estimated soi l cover by th e canopy (Fig. 7), th e

cover (%)1009080706050403020 -3+- sugar beet-a- potato10 wheat

0140 160 180 20 0 220 240 260 280 300day of year (20.5. - 25.10.)

Figure 7. Estimated average soil cover of sugar beet,potato and winter wheat for th e 1991 growing season,Flevoland tes t s i t e .

8/4/2019 189_XXIX-part7

5/8

mean crop he igh t , row dis tance, p lants per m2 , the so i lmoisture condi t ion an d comments about plant developments tage .Meteorological data. Daily meteorological data areneeded as input for crop growth simulat ion models. Forthe 1991 growing season these were obtained from theRoyal Dutch Meteorological Service (KNMI) fo r thes t a t i on Lelystad. Data consis ted of da i ly minimum an dmaximum temperature, daily global i r r adia t ion an d dailyprec ip i ta t ion .Leaf opt ical properties were invest igated with aLI-CORl abora to ry spectroradiometer a t the Centre fo rAgrobiological Research (CABO) in Wageningen. Theref lec tance or t ransmit tance s ignature of the upperan d lower surface of several leaves was recordedcont inuously from 40 0 to 1100 nm wavelength in 5 nms teps . The instrument was cal ibrated with a white bariumsulphate pla te .Field ref lectance measurements were obtained duringthe 1991 growing season with a portable CROPSCANradiometer (Buker e t a l . , 1992). Eight narrow-bandin te r fe rence f i l t e r s with photodiodes were orientedupwards to de tec t hemispherica l incident r adia t ion an da matched se t of in te r fe rence f i l t e r s wi th photodiodeswere or ien ted downwards to detec t r e f l ec ted radia t ion .Spec t ra l bands were located a t 490, 550, 670, 700, 740,780, 87 0 an d 1090 nm with a bandwidth of 10 nm. Theradiometer was cal ibra ted by point ing towards the su nwi th both types of photodiodes separa te ly . Percentagesref lectance were calculated by th e ra t io of th e signalsof both se ts of detectors . The sensor head of theradiometer was mounted on to p of a long metal pole an dposi t ioned three metres above the ground surface. Thed is tance to the crop was 2.5 to 1.5 m depending on thecrop he igh t . As th e diameter of the f i e ld of view (FOV28) was ha l f the dis tance between sensor and measuredsurface, the f i e ld of view varied from 1.23 m2 to 0.44m2 AVIRIS measurements. The ER-2 ai rcraf t of NASA, carryingthe airborne vis ib le - inf ra red imaging spectrometer(AVIRIS), performed a successful over f l ight over th eFlevoland t e s t s i t e on July 5th, 1991. AVIRIS acquires22 4 cont iguous spectra l bands from 0.41 to 2.45 ~ m . The ground re so lu t ion i s 20 m as it i s f lown a t 20 kma l t i t ude .CAESAR measurements. The CAESAR (CCD AirborneExperimental Scanner for Applicat ions in Remote Sensing)appl ies l inea r CCD arrays as detectors . I t has a modularse t -up an d it combines th e poss ib l i t i e s of a highspec t ra l r e solut ion with a high spa t i a l r e solut ion .For land appl icat ions three spectra l bands are avai lablein the green, red an d NIR par t of the EM spectrum. Oneof th e spec ia l opt ions of CAESAR i s th e capabi l i ty ofacquir ing data according to th e so-cal led dual lookconcept. This dual look concept consis ts of measurementsperformed when looking nadir an d under the oblique angleof 52. Combining these measurements provides informat i on on th e direct ional ref lectance proper t ies of objects(Looyen e t a l . , 1991). Successful over f l ight s over thet e s t s i t e were carr ied out on July 4th , July 23rd an dAugust 29th, 1991.

4. PRELIMINARY RESULTS MAC EUROPE 1991So fa r the concepts of chapter 2 have only (par t ly)been worked out fo r sugar beet , an d these r e su l t s wil lbe presented here .4 .1 Measurements of Leaf Optical Proper t ie sDuring July and August 1991 individual l eaves of sugarbee t were measured in the laboratory with aLI-COR LI-1800 portable spectroradiometer . During t h i s per iod

193

l e a f proper t ie s were r a the r cons tan t . The measurementsyielded for a NIR band (a t 87 0 nm) an averagere f lec tance of 46.0% an d an average t ransmit tance of48.4%. These values were re spec t ive ly 7.3% and 0.6%fo r a red (a t 67 0 nm) an d 15.8% an d 13.8% fo r a greenband (a t 550 nm). The average sca t t e r ing coe f f i c i entwas 0.144 for th e whole PAR region.4 .2 Estimating Leaf Angle Dis t r ibu t ion (LAD)Informat ion on LAD was obtained by means of the CAESARscanner in dual look mode. As explained in Fig. 5,measured WDVI values a t an obl ique an d nad i r viewingangle plot ted in to such a nomograph, based on th e actualrecording geometry and th e l ea f optical propert ies fromsect ion 4.1 , ca n yie ld es t imates of both LAI and LAD.Fig. 8 gives the resul t s of July 4 th fo r the CAESARscanner (note: CROPSCAN measurements over bare so i lyielded an est imate fo r C in Eq . (1) of 1. 15) togetherwith s imula ted curves fo r a spher ica l , uniform an dplanophi le LAD (LADs as def ined by Verhoef & Bunnik,1981). In t h i s f igure we have shown more r e a l i s t i c LADsas opposed to Fig. 5 with LADs consis t ing of j u s t on eangle . Resul ts fo r a l l three da tes showed tha t sugarbee t mostly matched th e curve for a spher ical LAD ra therwe l l , except fo r the beginning of the growing season(LAIo5: 20

/ , ' // / /

all ' //

, o " /,///;g/'/;;/

- - planophile LAD.------ uniform LAD-- - spherical LAD00000 measurements 04- jul-91

O ~ - . - - . - . - ~ - - . - - . - . - - . - - r ~ o 20 40 60 80 100WDVI-NADIR (%)Figure 8. Relat ionship between the WDVI measured fromnad i r an d th e WDVI measured a t an obl ique viewing angleof 52 for a spher ical , uniform an d planophile LAD (SAILmodel with a hot spot s ize-parameter of 0 .5 fo r sugarbeet) an d measurements obtained with CAESAR, July 4th -Jul i an day 18 5 - , 1991 (13.30 GMT). Solar zenith angleof 36 an d azimuth angle between plane of observat ionan d sun of 7 .4 .3 Estimating Leaf Area Index (LAI)Using the opt ica l l e a f proper t ie s found in sect ion 4 .1an d the spherical LAD in sec t ion 4.2 , the r e la t ionshipbetween WDVI an d LAI was simulated using the SAIL model(cf . Fig. 4). The regression of LAI on WDVI (Eq. 3)yielded fo r a an est imate of 0.418 an d fo r W D V I ~ anest imate of 57.5 (spherical LAD). Bouman e t a l . (1992)found fo r sugar bee t empir ical ly fo r a an est imate of0.485 an d fo r W D V I ~ an est imate of 48.4 , whereby theWDVI was based on green re f lec tance instead of redref lectance. CROPSCAN measurements an d SAIL simulat ionsyielded a ra t io between green an d red re f lec tances forsugar bee t of 1 .16. As a r e su l t , a value of 48.4 fo rWDVL corresponds with a value of 56.3 for a W D V I ~ basedon red ref lec tances . Fig. 9 i l l u s t r a t e s t ha t th esimulated r e la t ionship an d the empir ical r e la t ionship

8/4/2019 189_XXIX-part7

6/8

correspond rather well. Applying the empirical functionto measured WDVI values yielded temporal LAI signaturesfor the sugar beet f ields (Fig. 10).8

6

2

Bouman et 01.SAIL simulation

," /'/

///

/I

O ~ = ' - - r - ' - - . - . - - . - . - - r - ' - - r - ~ ~ ~ - - ~ ~ ~ o 10 20 30 40 50 60 70 80WDVI (%)Figure 9. Theoretical (SAIL model) and empirical (Boumane t a l . , 1992) relat ionship between WDVI (based on NIRan d re d ref lectances) and LAI for sugar beet .LAI9

876543 Be

-X - Bo2 ---8- Ee--- Fe-B - FrO ~ = = ~ ~ - - - L - - - - ~ - - - - L - - - ~ - - - - ~ - - - - - L - - ~

140 160 180 200 220 240 260 280 30 0day of year (20.5. - 25.10.)

Figure 10 . Temporal LAI signatures fo r sugar beet in1991.4.4 Estimating Leaf Optical PropertiesFig. 6 i l lus t ra ted th e influence of LAI an d l eafchlorophyll content on the posi t ion of the re d edgeand th e WDVI as simulated with a combined PROSPECT-SAILmodel. By plott ing both the measured WDVI (acquiredwith CAESAR) an d the red edge values (acquired withAVIRIS) in to such a nomograph for actual recordingconditions, an estimate of both LAI and leaf chlorophyllcontent i s obtained (see Fig. 6). Since AVIRIS datawere yet cal ibrated up to radiances and not toref lectances, WDVI values of CAESAR were used. Fo rcalculating re d edge values, radiances may be usedinstead of reflectances (Clevers & Buker, 1991). Resultsfo r sugar beet yielded an estimated chlorophyll contentof about 28 ~ g . c m - 2 , except fo r the beginning of thegrowing season (LAI

8/4/2019 189_XXIX-part7

7/8

5. SUMMARY AND DISCUSSIONA framework was presented to integrate crop canopyinformation derived from optical remote sensing withcrop growth models for th e purpose of growth monitoringand yield estimation. Within this framework, two methodswere described. In the f i r s t method, crop parametersthat play an important role in both the processes ofcrop growth and canopy ref lectance were estimated fromopt ical remote sensing data. These crop parameters wereLAI, LAD and l eaf colour. Fo r each parameter, anestimation methodology was developed or taken froml i te ra ture . The estimated parameter values ca n be usedas di rec t input into crop growth models. In th e secondmethod, a crop growth model cal led SUCROS was extendedwith a canopy ref lectance model to calculate remotesensing signals from th e growing crop. The extendedgrowth model ca n be re-parameterized, or calibrated,on t ime-series of remote sensing data. The main l inkingparameter between growth model and th e reflectance modelwas LAI.The framework was appl ied to data gathered during theMAC Europe 1991 campaign over th e Dutch t e s t s i teFlevoland. In i t i a l resul ts for sugar beet indicatedth e feas ib i l i ty of estimating LAI, LAD an d leaf colourfrom optical ref lectance measurements. As yet , theseparameter estimations have not been implemented in thegrowth model (SUCROS). A cr i t i ca l point to consideris th e precis ion and addit ional value of the parametervalues derived from remote sensing compared to th estandard values already used in th e growth model. Forinstance, th e LAD value tha t was derived for sugar beetfrom the CAESAR measurements confirmed the spher icalLAD current ly used in SUCROS (sect ion 4.2). Leafscat ter ing coeff ic ien t was found to be about 0.14(sections 4.1 and 4.4) instead of 0.20, but thesimulation of biomass is not very sensit ive to l eafscat ter ing coeff ic ien t in th e range of 0.1 to 0.3. Incontrast , much relat ive benefi t might be obtained fromthe est imat ion of l eaf colour expressed in l eafnitrogen/chlorophyll content. Especially the modellingof l eaf nitrogen status in canopies is extremelycomplicated (but equal ly important through i t s effecton maximum l eaf photosynthesis rate) an d actualinformation derived from optical ref lectance would bevaluable. However, it wil l take more research anddedicated experiments together with crop physiologiststo investigate th e potentials of opt ical remote sensingfo r the assessment of leaf (o r canopy) nitrogen status .The method of model re-parameterization was also testedon sugar beet. Fo r nine ou t of te n fields, th e simulatedyield was bet te r in agreement with actual ly obtainedyields af te r model re-parameterization than withoutmodel re-parameterization. Since th e re-parameterizationprocedure mainly concerned the cal ibrat ion of thesimulated LAI, these resul ts indicate the importanceof LAI for accurate growth simulation.Within th e framework for integrat ing remote sensingdata an d crop growth models, both methods ' cropparameter est imation ' and 'model re-parameterization'have to be used together in a combined approach. Cropparameter est imat ion ca n best be used to estimate cropmodel inputs (crop parameters) that are relat ivelystable throughout th e growing season, such as LAD andlea f colour of sugar beet . Model re-parameter izat ionis especial ly sui ted to cal ib ra te the simulation ofrelat ively dynamic variables , such as LAI.By using inversion techniques, remote sensing modelparameter est imat ion i s also possible . Depending onth e amount of physiological an d hydrological relationswhich could serve as a priori information, th e parameterestimation by inversion might be successful. Subsequently , crop growth model parameters may be adjusted sothat simulated values of LAI by th e growth model match

195

LAI values estimated af te r inversion of ref lectancemodels (cf . Maas, 1988).The framework ca n be extended to include otherparameters that are relevant to both crop growth an doptical ref lectance. Moreover, the framework ca nincorporate other remote sensing techniques as wel l ,such as radar , passive microwave or thermal remotesensing (Bouman, 1991).

ACKNOWLEDGEMENTSW. Verhoef ' i s acknowledged fo r providing th e SAIL modeland J . Goudriaan is acknowledged for providing th eSUCROS model. We are very grateful to S. Jacquemoudand F. Baret (INRA, Montfavet - France) fo r providingth e PROSPECT model. NASA is acknowledged fo r providingthe AVIRIS data in the framework of MAC Europe 1991.This paper descr ibes a study that was carried out inth e framework of th e NRSP-2 under responsibi l i ty ofthe Netherlands Remote Sensing Board (BCRS) and undercontract no. 4530-91-11 ED ISP NL of the Joint ResearchCentre.REFERENCESAllen, W.A., H.W. Gausman, A.J. Richardson & J.R.Thomas, 1969. Interaction of isotropic l i gh t with acompact plant l eaf . J . Opt. Soc. Am. 59: 1376-1379.Allen, W.A., H.W. Gausman & A.J. Richardson, 1970. Meaneffec t ive optical constants of cot ton leaves. J . Opt.Soc. Am. 60: 542-547.Asrar, G. (ed.) , 1989. Theory and applicat ions ofoptical remote sensing. New york: John Wiley & Sons,Inc . , 734 pp .Bouman, B.A.M., 1991. Linking X-band radar backscat tering and optical ref lectance with crop growth models.Thesis Agricultural University Wageningen, Wageningen,the Netherlands.Bouman, B.A.M., 1992a. The accuracy of estimating thel eaf area index from vegetat ion indices derived fromcrop ref lectance character is t ics , a simulation study.Int . J. Rem. Sensing ( in press.)Bouman, B.A.M., 1992b. Linking physical remote sensingmodels with crop growth simulation models, applied forsugar beet . Int . J . Rem. Sensing (in press .)Bouman, B.A.M., H.W.J. va n Kasteren &D. Uenk, 1992.Standard relat ions to estimate ground cover and LAIof agricul tural crops from reflectance measurements( in prep.)Bouman, B.A.M. &J . Goudriaan, 1989. Estimation of cropgrowth from optical an d microwave so i l cover. In t . J .Rem. Sensing 10: 1843-1855.Buker C., J.G.P.W. Clevers, H.J.C. vanLeeuwen, B.A.M.Bouman & D. Uenk, 1992. Optical component MAC Europeground t ru th report - Flevoland 1991. WageningenAgricul tu ra l Univers i ty , Dept. Landsurveying & RemoteSensing, Report LUW-LMK-199204.Bunnik, N.J.J . , 1978. The multispectral reflectanceof shortwave radiation by agricultural crops in relat ionwi th the i r morphological and optical properties. ThesisAgricul tural University Wageningen, Wageningen, th eNetherlands.Clevers, J.G.P.W., 1988. The derivation of a simpl i f iedreflectance model fo r th e estimation of Leaf Area Index.Rem. Sens. Envir. 25: 53-69.

8/4/2019 189_XXIX-part7

8/8

Clevers , J . G. P. W. , 1989. The appl ica t ion of a weightedin f ra red - red vegetat ion index fo r est imat ing Leaf AreaIndex by correct ing fo r soi l moisture. Rem. Sens. Envir.29 : 25-37.Clevers , J.G.P.W., 1992. Modelling an d synerget ic useof opt ica l an d microwave remote sensing. Report 4:Influence of l ea f propel,"ties on th e re la t ionship betweenWDVI an d LAI: a sens i t i v i t y ana lys i s with the SAIL an dth e PROSPECT model. Wageningen Agricul tura l Universi y ,Dept. Landsurveying & Remote Sensing, Report LUW-LMK-199202, 31 pp .Clevers , J.G.P.W. & C. Biiker, 1991. Feas ib i l i ty of thered edge index for th e detect ion of ni t rogen deficiency.Fif th In t . Coll . on Physical Measurements and Signaturesin Remote Sensing, Courchevel, France, ESA SP-3l9, pp .165-168.Clevers , J.G.P.W. & W. Verhoef, 1990. Modelling an dsynerge t ic use of optical an d microwave remote sensing.Report 2: LAI est imation from canopy ref lectance an dWDVI: a sens i t i v i t y ana lys i s with the SAIL model. BCRSrepor t 90-39, 70 pp.Goel , N.S. & D.W. Deering, 1985. Evaluat ion of a canopyre f lec tance model fo r LAI est imation through i t sinvers ion. IEEE GE-23: 674-684.Guyot, G. & F. Baret , 1988. Ut i l i s a t i on de l a hauteresolut ion spec t ra le pour su ivre I' e t a t des couverjtsvegetaux. Proc. 4 th In t . ColI . on Spec t r a l Signaturesof Objects in Remote Sensing, Aussois, France, 18-22January 1988 (ESA SP-287, Apri l 1988).Horler , D.N.H., M. Dockray & J . Barber, 1983. The re dedge of p lan t l e a f ref lec tance . In t . J . Rem. Sensing4: 273-288.Jacquemoud S. & F. Baret , 1990. PROSPECT: a model ofl e a f opt ica l proper t ie s spec t ra . Rem. Sens. Envir. 34 :75-91.Kanemasu, E.T. , G. Asrar &M. Fuchs, 1984. Applicat ionof remotely sensed data in wheat growth modell ing. In :Wheat growth modell ing: 357-369, Edited by Day, W. an dR.K. Atkin (Plenum Press , New York an d London, Publishedin cooperat ion wi th NATO Sc ient i f i c Affa i r s Divis ion) .Keulen, H. van & J . Wolf ( eds . ) , 1986. Modelling ofagr icul tura l production: weather , so i l s an d crops .Simulat ion monograph, PUDOC, Wageningen, The Netherlands.Keulen, H. van & N.G. Seligman, 1987. Simulat ion ofwater use, nitrogen nu t r i t i on and growth of a spr ingwheat crop. Simulat ion monograph, PUDOC, Wageningen,The Nether lands .Looyen, W.J., W. Verhoef, J.G.P.W. Clevers , J .T . Lamers,& J . Boerma, 1991. CAESAR: evaluat ion of the d u a l ~ l o o k concept . BCRS repor t 91-10, 144 pp .Maas, S.J . , 1988. Us e of remotely sensed informat ionin agr icul tura l crop growth models. Ecological modell ing41: 247-268.

196

Penning de Vries, F.W.T. & H.H va n Laar (Eds . ) , 1982.Simulation of p lan t growth an d crop product ion.Simula t ion Monographs. Wageningen: Pudoc, 30 8 pp .Richardson, A.J. &C.L. Wiegand, 1977. Dist inguishingvegetat ion from soi l background information. Photogram.Eng. Rem. Sensing 43: 1541-1552.Schell berg, J . , 1990. Die spektrale Reflexion vo n Weize n- ein Beitrag zu r Zustandsbeschreibung landwir tschaf t l i cher Kulturpf1anzenbestande durch Fernerkundung. PhDThesis , Rheinischen Friedr ich-Wilhelms-Universi ta t ,Bonn, 160 pp .Sp i t t e r s , C. J . T. , H. van Keulen & D. W.G. van Kraail ingen, 1989. A simple an d universa l crop growth simulator:SUCROS87. In: Simulation and systems management in cropprotect ion. Ed. : R. Rabbinge, S. A. Ward & H. H. van Laar.Simulation Monographs 32. Wageningen: Pudoc, 42 0 pp .Steven, M.D. & J.A. Clark (Eds.) , 1990. Applicat ionsof remote sens in g in agr icul ture . London: Bu t terworths ,42 7 pp .Steven M.D., P.V. Biscoe & K.W. Jaggard, 1983.Est imat ion of sugar bee t productiv i ty from ref lec t ionin the red an d infrared spec t ra l bands. In t . J . Rem.Sensing 2: 117-125.Sui t s , G. H., 1972. Th e ca lcu la t ion of the di rec t iona lref lec tance of a vegetat ion canopy. Rem. Sens. Envir.2: 117-125.Tucker, C. J . , 1979. Red an d photographic in f ra red l inearcombinations fo r monitoring vegetat ion. Rem. Sens.Envi r . 8: 127-150.Uenk, D., B.A.M. Bouman & H.W.J. van Kasteren, 1992.Reflect iemet ingen aa n landbouwgewassen. CABO Verslag156, 56 pp.Vane, G., M. Chrisp, H. Enmark, S. Macenka&J. Solomon,1984. Airborne Vis ib le / In f ra red Imaging Spectrometer:An advanced tool for Ear th remote sensing. IGARSS ' 84,SP2l5, 751.Verhoef ,W. , 1984. Light scat ter ing by l e a f layers withappl i ca t ion to canopy ref lectance model l ing: the SAILmodel. Rem. Sens. Envi r . 16 : 125-141.Verhoef, W. & N.J . J . Bunnik, 1981. Inf luence of cropgeometry on mult i spec t ra l re f lec tance determined bythe use of canopy re f lec tance models. Proc . In t . ColI.on Signatures of Remotely Sensed Objects , Avignon,France , pp. 273-290.Wit, C.T. de, 1965. Photosynthes i s of l e a f canopies .Agricul tura l Research Report 663, Centre fo r Agricul tur a l Publ icat ions and Documentation (PUDOC), Wageningen,The Nether lands.