ERRORS IN REMOTE-SENSING OF INTERCEPTED PHOTOSYNTHETICALLY ACTIVE RADIATION - AN EXAMPLE FROM HAPEX-SAHEL

Canopy PAR interception (IPAR) is an important variable in many models of canopy photosynthesis and net primary production. IPAR can be esti mated remotely using spectral vegetation indices (VI). However, IPAR e stimates by this method are subject to errors (E) associated with (i) the estimation of fractional interception (f(PAR)) from a VI measureme nt (E-F), (ii) the accuracy of estimates of incident PAR (E-T) and (ii i) the interaction of these errors (E-X). During the HAPEX-Sahel exper iment in Niger in 1991 the temporal evolution of a VI was measured ove r the study sites with a radiometer mounted in a light aircraft. Groun d measurements of f(PAR) and incident PAB on HAPEX West Central shrub fallow, grass fallow, degraded shrub fallow and millet sites were avai lable. The f(PAR) measurements were used to determine the relationship between the VI and fractional interception. The aircraft measurements were then used. together with remote estimates of incident PAR from t he TOMS satellite. to estimate IPAR through the season and to quantify the error sources. Both E-F and E-T can result in large absolute and relative errors in IPAR estimates, E-X was generally small. In ten-day summations on the shrub and grass fallow sites, the total errors were mostly less than 20% of the round measurements. On the millet and deg raded shrub fallow, the ten-day errors were more substantial. The erro rs at short time steps (1-10 days) cancel when summed for the whole gr owing season. In applications where the precise daily errors in IPAR e stimates cannot be quantified, the statistical uncertainty of the esti mates can be evaluated if the uncertainty in the input variables is kn own. The standard error of IPAR estimates is then dependent on the res idual variance in the regression between vegetation index and f(PAR) a nd on the mean square error of the remote estimates of incident PAR.