摘要：Satellites provide the only practical source of data for estimating biomass of large and remote areas such as the Alaskan Arctic. Researchers have found that the normalized difference vegetation index (NDVI) correlates well with biomass sampled on the ground. However, errors in NDVI and biomass estimates due to bidirectional reflectance distribution function (BRDF) effects are not well reported in the literature. Sun-sensor-object geometries and sensor band-width affect the BRDF, and formulas relating NDVI to ground-sampled biomass vary between projects. We examined the effects of these different variables on five studies that estimated above-ground tundra biomass of two common arctic vegetation types that dominate the Alaska tundra, moist acidic tussock tundra (MAT) and moist non-acidic tundra (MNT). We found that biomass estimates were up to 33% (excluding extremes) more sensitive than NDVI to BRDF effects. Variation between the sensors resulted in differences in NDVI of under 3% over all viewing geometries, and wider bands were more stable in their biomass estimates than narrow bands. MAT was more sensitive than MNT to BRDF effects due to irregularities in surface reflectance created by the tussocks. Finally, we found that studies that sampled only a narrow range of biomass and NDVI produced equations that were more difficult to correct for BRDF effects.
