Mapping vegetation structure in West Africa

Kurzfassung

West Africa is characterized by a strong climatic North-South gradient, which translates in a clear vegetation zoning and a gradient from hardly vegetated desert to dense evergreen rain forests. Despite the many studies conducted in the region, little is known about the vegetation structure and its phenological characteristics at adequate spatial detail and coverage. The availability of Sentinel-2 together with Landsat generally allows for frequent observations and hence the retrieval of phenological parameters such as start of season, end of season or length of growing season. However, even the combined Sentinel-2/Landsat dataset does not consistently provide sufficient observations for such kind of analysis in cloudy regions, in particular in the Guinean savanna. We therefore explored the use of the Sentinel-1 Synthetic Aperture Radar (SAR) record to characterize the temporal patterns of forest ecosystems throughout West Africa. The higher observation frequency compared to optical data due to the system’s capability to penetrate clouds is advantageous in deriving temporal patterns while pre-processing is key in such analyses. Our findings show that single pixel time series without any speckle treatment fully blurred the “true” temporal backscatter variation with speckle-related noise. Speckle reduction leads to more obvious temporal patterns which reflect the North-South gradient and allow for more detailed assessments. We found good agreement of SAR-derived time series with NDVI time series in relatively dry savanna regions demonstrating the comparability of SAR phenology with phenology based on optical remote sensing. Yet, there are differences in some cases that need further exploration.