Three 
5-202. 
PROCESSING OF HIGH SPECTRAL RESOLUTION REFLECTANCE DATA 
FOR THE RETRIEVAL OF BIOPHYSICAL INFORMATION 
sntial 
7 
O > 
clectance 
E.M. Rollin and EJ. Milton 
NERC Equipment Pool for Field Spectroscopy 
Department of Geography 
University of Southampton 
Southampton S09 5NH, U.K. 
1990, 
¿m) of 
:ea, 
ABSTRACT: 
This paper examines how different processing, in particular different degrees of smoothing, affect the 
extraction of biophysical information from spectral data. Bi-conical spectral reflectance data collected during 
msing 
the EISAC’89 Campaign from an experimental grassland site on Tadham Moor, Somerset, UK were used 
for the analysis, together with canopy chlorophyll and water content variables. The analysis focuses mainly 
on the correlation between percentage water content per unit area of leaf and spectral parameters across 
:e 
the water absorption features at 1150nm. Correlation analysis revealed that both the wavelength position 
and magnitude of the peak correlation between first derivative and water content was sensitive to the degree 
of smoothing. This sensitivity is attributed to the fact that different levels of smoothing emphasise different 
scanners, 
spectral features across a spectral region where the two variables are generally highly correlated. The effect 
emphasises that the exact position of peak correlation between the two variables should be interpreted with 
caution especially across spectral regions where the correlation is generally high. 
)0-2500 
ETL-0573, 
KEY WORDS: EISAC’89, data smoothing, canopy water content. 
¡rbos , J., 
.ate the 
:e 
1. INTRODUCTION 
Several laboratory studies have examined the relationship between a variety of spectral indices and a range 
of measures of leaf water content (Gausman et al, 1970, Allen et al, 1971, Hunt and Barrett, 1989). Danson 
et al, (1992) described various measures of leaf water status which have been correlated with leaf spectral 
surements 
reflectance. They used the measure Specific Water Density (SWD) as the correlate in their analysis of 
laboratory leaf spectra. The problem of extending relationships identified in the laboratory to the field 
situation are well documented. Hunt and Barrett (1989) concluded that it would be difficult to determine 
the necessary parameters for the calculation of his LWCI (Leaf Water Content Index) from a practical 
application of remote sensing. He also showed that the sensitivity of spectral indices based on leaf reflectance 
i and 
cal 
in the SWIR alone, or on a combination of SWIR and NIR reflectance, was insufficient to be useful in the 
field situation. Many of the leaf water content measures used successfully in the laboratory experiments are 
difficult if not impossible to retrieve in the field. Similarly, many of the spectral indices using SWIR 
reflectance extracted from laboratory spectra are not accessible to the field investigator because useful 
surements 
irnational 
•ng, 
wavelengths are within regions of strong absorption by the atmosphere. In the field, the relationship between 
leaf water content and leaf reflectance is modified to the canopy scale and is further complicated by the 
influence of canopy geometry and soil background effects and the fact that real vegetation canopies comprise 
a mix of components other than just leaves. 
This analysis attempts to establish useful relationships between canopy spectral reflectance and 
F. 
Purdue 
canopy biophysical parameters from field data, focusing primarily on the correlations between the spectral 
parameters and the gravimetric water content of the canopy expresses as a percentage of fresh biomass 
weight per square metre area of canopy. The effect of different spectral processing, in particular different 
degrees of smoothing of the spectra, on the extraction of biophysical information from the field spectral data 
Imaging 
is examined. 
2. DATA COLLECTION 
The data examined were collected during the UJK. phase of the European Imaging Spectrometry Airborne 
357