Remote sensing of biomass dynamics in drylands: Evaluating vegetation optical depth (VOD) using AVHRR NDVI and in situ data

Abstract:

Monitoring long-term biomass dynamics in global drylands is of great importance for global carbon cycle modeling and has been done extensively based on the normalized difference vegetation index (NDVI) derived from AVHRR (Advanced Very High Resolution Radiometer) observations. However, there are limitations from both the characteristics of NDVI (e.g. atmosphere and cloud contamination, saturation in densely vegetated areas, and affected by varying vegetation species compositions) and sensor related artifacts (e.g. orbital drifts, sensor changes). Being sensitive to the vegetation water content and not affected by clouds, the Vegetation Optical Depth (VOD) derived from satellite passive microwave observations can be an alternative to NDVI for monitoring biomass dynamics in drylands, yet further evaluations based on ground measurements are needed. In this study, we assess the capability of a long-term VOD dataset (1992–2011) to capture the temporal and spatial variability of in situ measured biomass data (herbaceous and woody foliage mass) in the semi-arid Senegalese Sahel. The GIMMS3g (Global Inventory Modeling and Mapping Studies, 3^rd generation) NDVI dataset is included for comparison purpose. Both VOD and NDVI reflect the temporal and spatial pattern of the ground data very well, however, the phenological metrics leading to the best correlations differ between VOD and NDVI. While the annual sum and maximum perform best for VOD, the growing integrals have the highest correlations for NDVI. Furthermore, VOD proves to be robust against typical NDVI drawbacks (species compositions, and saturation effects). Overall, in spite of the coarse resolution, the study shows that satellite passive microwave observation based VOD is an efficient proxy for estimating biomass production of the entire vegetation layer in the Sahel and potentially in other dryland areas.