Time‐series validation of MODIS land biophysical products in a Kalahari woodland, Africa

Abstract

Monthly measurements of leaf area index (LAI) and the fraction of absorbed photosynthetically active radiation (f APAR) taken at approximately monthly intervals were collected along three 750 m transects in a Kalahari woodland near Mongu in western Zambia. These data were compared with MODIS NDVI (MOD13, Collection 3) and MODIS LAI and f APAR products (MOD15, Collection 3) over a 2 year period (2000–2002). MODIS and ground‐measured LAI values corresponded well, while there was a significant bias between MODIS and ground‐measured f APAR even though both MODIS variables are produced from the same algorithm. Solar zenith angle effects, differences between intercepted and absorbed photosynthetically active radiation, and differences in measurement of f APAR (photon counts versus energy) were examined and rejected as explanations for the discrepancies between MODIS and ground‐measured f APAR. Canopy reflectance model simulations produced different values of f APAR with the same LAI when canopy cover was varied, indicating that errors in the estimation of canopy cover in the MODIS algorithm due to the land cover classification used are a possible cause of the f APAR discrepancy. This is one of the first studies of MODIS land product performance in a time‐series context. Despite a bias in f APAR, our results demonstrate that the woodland canopy phonology is captured in the MODIS product.