BRDF CORRECTION ON AVHRR IMAGERY FOR SPAIN
H. Heisig
Institut für Photogrammetrie und Fernerkundung, Universität Karlsruhe, D-76137 Karlsruhe heisig@ipf.uni-karlsruhe.de
TS 3 YF Youth Forum - Remote Sensing
KEY WORDS: Remote Sensing, Radiometry, Correction, Model, Land Cover
ABSTRACT
The free reception of Advanced Very High Resolution Radiometer (AVHRR) data keeps a growing interest for the remote sensing
community concerned with vegetation studies on regional and global scales. However, the use of AVHRR data for long-term
quantitative monitoring requires consistent surface reflectance data. This implies correcting the effects from varying sun sensor target
geometries in multitemporal AVHRR data sets described by the Bidirectional Reflectance Distribution Function (BRDF). In a joint
research undertaken by the Geography Department of the University of Alcalá (Spain) and the Canadian Centre for Remote Sensing
(CCRS) a new semi-empirical model, the Non-linear Temporal Angular Model (NTAM), has been applied for the BRDF correction
of AVHRR data for Spain. Reflectance data derived from AVHRR channels I and 2 were corrected for BRDF effects by
normalization to a standard viewing geometry. The study period was between May and September 2002. The required correction
parameters were derived by model inversion. In order to obtain a good representation of the temporal and spatial dimensions of the
data to be corrected, a sampling scheme was implemented. The land cover based approach of the NTAM was accounted for by the
use of CORINE Land Cover data from Spain. A cloud mask algorithm was used to exclude cloud contaminated observations in the
sampling data. The evaluation of the statistical parameters obtained from model inversion showed good results for practically all land
cover classes in the two channels. Derived parameters allow for visualization of angular dependence of reflectance for different land
cover classes.
1. INTRODUCTION
Current research work at the Geography Department at the E
University of Alcalä is strongly focussed on the field of fire risk dE; dL,
assessment (Chuvieco 2002). Since 1998 the department T
‘receives AVHRR data in the High Resolution Picture " T
Transmission (HRPT) format through an own receiver. These
data are used for estimation of plant water content with the |
purpose of incorporating these estimations in fire risk indices | |
(Aguado et al., 2003; Chuvieco et al., 2003).
Sun-synchronous polar-orbiting satellites like the National
Oceanic and Atmospheric Administration (NOAA) AVHRR
sensors are operated on must have such an orbital inclination
that the rate of the resulting precession compensates for the
motion of the Earth around the sun (Cracknell 1997). The
satellites orbit tracks projected on the Earth's surface reveal à Figure 1: Concepts and parameters of the BRDF (Sandmeier &
parallel shift of about one degree east per day. This results in Itten, 1999; modified)
changing view angles for the same target within consecutive
day passes. where:
0, 2 zenith incidence angle
BRDF 0, = zenith reflection angle
¢; = azimuth incident angle
Vegetation, as most other surfaces, does not scatter irradiance 0, = azimuth reflection angle
in equal quantities in all directions. In fact, it shows a behaviour ¢ = phase angle
far from being Lambertian (Dymond et al, 2001). Its
reflectance depends on the angle of observation as well as on Parallel to the enhancement of BRDF measuring methods
the angle of the incidence of the solar radiation. This bi- (goniometers, measurements from helicopters), considerable
directional dependency lead to the idea of the BRDF effort has been spent on development and testing of analytical
(Bidirectional Reflectance Distribution Function). The BRDF and empirical models to describe the BRDF (Chen & Cihlar,
describes the ratio of radiance dLr [W m” sr! nm'!] reflected in 1997: Chopping, 2000; Hu et al., 1997; Ni et al., 1999; Ni & Li,
one direction (8, $) to the irradiance dEi [W m^ nm"] from — 2900; O'Brien et al., 1998; Roujean et al., 1992; Wanner ct al..
direction (8; $) (Sandmeier & Itten, 1999). Figure 1 displays 1995), BRDF models potentially allow for prediction of
the angles that the BRDF is dependent on. reflectance for any desired sun sensor target geometry.
40