Symposium on Remote Sensing for Resources Development and Environmental Management / Enschede / August 1986 
269 
Spectral and botanical classification of grasslands: Auxois example 
C.M.Girard 
Institut National Agronomique, Paris-Grignon, France 
ABSTRACT:The author has performed botanical observations in the field, and in situ reflectance measurements 
over a French test-site at four different periods during the growing season.Factorial analysis of botanical 
relevés leads to agroecological units definition which are then spectrally characterized. 
RESUME: Des observations botaniques et des mesures de réflectance sur le terrain ont été réalisées sur une 
zone-test française à quatre différentes périodes.L'analyse factorielle des relevés conduit à la définition 
d'unités agroécologiques qui font ensuite l'objet d'une caractérisation spectrale saisonnière. 
1 INTRODUCTION 
Grasslands have been less studied than crops because 
of their less direct economical value: they corres 
pond mainly to a domestic use, and of their greater 
complexity in terms of species, canopy structure, 
phenological stages. Till now the main works per 
formed on grasslands concern: 
1. Mapping of grasslands,or rangeland units through 
aerial photographies or satellite data( Everitt and 
all 1985, Brown and all 1983...) 
2. Evaluation of green biomass or forage produc 
tion from spectral measurements in the field(Grou- 
zis and all 1983, Thalen and all 1980, Richardson 
and all 1983...) or from multispectral photography 
or MSS data (Curran 1983, Tucker and all 1983 *..) 
The work presented here has for aim to determine 
a methodology enabling to classify grasslands on 
behalf of their botanical composition and their 
reflectance behaviour in order to better use remote 
sensina data for their evaluation and monitoring. 
2 MATERIAL AND METHODS 
This research was based on 98 botanical stations of 
grasslands situated in Auxois(Cote d'Or, France) 
test-site of 220 sc km centered on the citv of 
Vitteaux (47° 24'- 4° 33'). 
The area is characterized bv a temperate climate 
with a sub-montainous tendancv. Winters are wet and 
fresh while summers are moist and hot. 
Geomorphology is characterized by plateaux cut 
by valleys oriented NNW-SSE. 
Plateaux correspond to calcareous rocks, while val 
leys and their flanks correspond to marl covered by 
more or less deep colluvium deposits. 
Crops are located on plateaux or on alluvial cones 
grasslands occupy flanks and bottom of the valleys. 
Soils vary according to slope but are mainly brown 
soils. Their characteristics important for grasslands 
are: depth and water capacity. Soils on top of the 
flanks are shallow with low water capacity, while 
soils in the bottom of the valleys are deep with 
high water capacity and hydromorphic features. 
Data were collected from mid-April to mid-July du 
ring 1980, 1981, 1982, 1983 and completed in 1984, 
1985, in order to survey the period of maximum vege 
tation. The stations area vary from 25 to 60 sq m. 
Botanical data were collected according the phyto- 
sociological method (Guinochet 1973). Present species 
abundancy is noted according to Braun-Blanquet's 
coefficients (Braun-Blanquet 1925), as well as phe 
nological stages for the more abundant species. 
Spectral data were collected using an ISCO SR 140 
spectroradiometer, mounted on a rigid support at 1.5 
m high.This spectroradiometer works in the visible: 
380- 750 nm with a 25 nm spectral resolution, and 
in the near infra-red:750- 1450 nm with a 50 nm spec 
tral resolution.Its field of view is of approximately 
100°. 
It enables to measure grassland exitance:M. Solar ir- 
radiance is measured by the same spectroradiometer and 
we worked on reflectance: R=M/E. 
Spectral data were collected mainly on clear sunny 
days during midday hours. They were referenced to a 
barium sulfate calibration panel approximately each 
30 mn. 
Both botanical and spectral data were handled using 
principal component analysis.(PCA). 
3 RESULTS 
3.1 Botanical data 
Numerical handling of botanical data enabled to rela 
te grasslands stations to the following phyto sociolo- 
gical alliances: Xerobromion erecti Br.Bl et Moor 38, 
Cynosurion cristati Tx. 47, and Agropyro-rumicion 
crispi North 40, corresponding respectively to: 
xerophilous natural rangelands, mesophile or mesohyg- 
rophile grasslands, and wet grasslands. 
But we continued this analysis keeping only the 
more abundant species.This is justified by the fact 
that the agronomical value of grasslands is depen 
dent on the more abundant species and their area 
coverage. 
Abundancy coefficients were converted in per cent 
coverage (Van der Maarel 1979),and a PCA was perfor 
med with all relevés at each period (April, May, 
June, July) for the 26 more abundant species. 
Relevés cluster in 8 units called agro-ecological 
units which are characterized by species and their 
varying coverage through time, they correspond to 
various soil characteristics : depth, water regime, 
trophic level, pH... (Figure 1). 
Table 1 gives a model for analyzing phytosociologi 
cal relevés according to abundancy of different spe 
cies through spring. This model is used for grass 
lands in calcareous Burgundy and may be extended to 
other temperate areas with slight modifications. 
It has been validated with three stations situated 
in a neiahbouring area and which wi 11 serve for 
testing the spectral behaviour model.