Estimation of BRDF from AVHRR Short-Wave Channels: Tests over Semiarid Australian Sites

Abstract

Correction of time series of Advanced Very High Resolution Radiometer (AVHRR) short-wave reflectances for directional effects is investigated for four semiarid Australian sites. Two contrasting models for the bidirectional reflectance distribution function (BRDF), which satisfy the reciprocity condition, are fitted to all cloud-free observations within a sliding window centered on the current time. The window length is adaptive, but typically is between 10 and 20 days, corresponding to one or two repeat cycles of the polar orbiting meteorological satellites operated by the National Oceanic and Atmospheric Administration (NOAA). Observations within the window include satellite elevations ranging from one horizon to the other, so that the parameters of the BRDF models are well determined. In cloudy conditions the algorithm allows the time window to be extended in small steps up to an upper limit in order to secure sufficient data to fix the parameters of the model. For the test sites examined, the algorithm appears to be robust. Once the BRDF has been estimated, times series of reflectances at the top of the atmosphere are calculated for an observer at zenith and the Sun at fixed elevation. The corrected time series of reflectances are consistent and reveal structure not apparent in the raw data. In particular, times series of normalized difference vegetation index (NDVI) computed from the BRDF-corrected reflectances show the rapid response to rain expected for semiarid test sites. The accuracy of the fitted BRDF is validated for one of the test sites by comparing reflectances observed by the visible channel of the Visible and Infrared Spin-Scan Radiometer (VISSR) on the Japanese Geostationary Meteorological Satellite (GMS) against reflectances predicted by the BRDF model. The correlation between the two sets of reflectances is excellent. Because GMS samples regions of the angular domain not seen by AVHRR, the high correlation confirms that the BRDF is well represented by the model based on AVHRR data, even for geometries unsampled by AVHRR.