The applied satellite data includes medium spatial resolution normalized difference vegetation index (NDVI) and Fraction of absorbed Photosynthetic Active Radiation (FAPAR) from MODIS (500 m; 2000-2015), PROBA-V (100 m, 2014-2016), SPOT-VGT (1 km; 1998-2010), GIMMS-3g (8 km; 1982-2015), as well as passive microwave vegetation optical depth (VOD) (25 km; 1992-2011), and satellite based rainfall estimates ARC2, CHIRPS (both 1981-2015). Very high spatial resolution (~0.5 m) data included imagery from WorldView-2, QuickBird-2 and GeoEye-1 commercial satellites. Field observations on woody cover, woody species, herbaceous and foliage biomass were available over three decades from field sites in Senegal, Mali and Niger.
First, long-term VOD and GIMMS-3g data were compared with field observations from Senegal (1987 to 2011) with highly satisfying results, especially for the VOD data that had not been validated with time series of field observations before. The results are published in
Tian et al., (2016) doi:10.1016/j.rse.2016.02.056
The applied methods for estimating biophysical parameters followed the scheme of linking variables measured in the field with satellite data to extrapolate field observations in space and time. Here this project made use of the high temporal frequency of NDVI and FAPAR time series which describe the seasonal characteristics of the phenology of the plants, which is very different between woody and herbaceous species. Whereas most woody vegetation keeps green leaves over the entire year, herbaceous grow only during the wet season (July-September) and wilt towards the start of the dry season. This fact was used to estimate the fraction of woody and herbaceous cover separately, and build models to estimate the biomass production per year. Results are woody cover and biomass maps with an unprecedented accuracy, published in:
Brandt et al., (2016) doi:10.1016/j.rse.2015.10.036
Diouf et al. (2015). doi:10.3390/rs70709122
Diouf et al. (2016). doi:10.3390/rs8080668
Then we applied above mentioned techniques to map changes over time, and found a very heterogeneous pattern of woody cover changes, but a generally positive trend for Sahel. Taken together, the results were an unprecedented synthesis of woody cover dynamics in Sahel and pointed to land use and human population density as important drivers, however only partially offsetting a general post-drought increase. Results are published in:
Brandt, et al. (2016) doi:10.1016/j.rse.2016.05.027
Tian et al., (2016). doi:10.1111/gcb.13464
We used then very high spatial resolution satellite data to illustrate changes woody vegetation, as well as to train PROBA-V data and derive a woody cover map at a high level of detail, which allows to draw conclusions on the relationship between human management and woody cover. We found large scale increases and die offs in woody vegetation. Moreover, we found that in semi-arid Sahel, the woody cover in farmland is greater than in neighboring savannas. Results are published in:
Brandt et al., (2017). doi:10.20944/preprints201703.0173.v1
Brandt et al., (2017). doi:10.3390/rs9010039
Finally, we applied data sets on land use and human population in relation to NDVI and woody cover changes. We found that in Sahel, a decreasing greenness (NDVI) trend is coupled with an intensification of cropland, i.e. less fallowed and more cropped land. Moreover, we found that for entire Africa, woody vegetation in sparsely populated drylands shows a considerable increase over the 1992-2011 period, but woody vegetation decreased in densely populated areas, being a sign for deforestation. Although the decreases appear in areas that provide numerous ecological services (e.g. biodiversity, carbon storage), the general increase in dryland areas challenges the widely held view of an ongoing reduction of all woody vegetation in Africa. Results are published in:
Tong et al., (2017) doi:10.1016/j.rse.2017.01.030
Brandt et al., (2017). doi:10.1038/s41559-017-0081
The methods developed in this project were are transfered to other study areas (China, Syria):
Tong et al., (2016). doi:10.3390/rs8050357
Tong et al., (2017). doi:10.1016/j.jag.2016.09.013
Eklund et al. (2017) doi:10.1088/1748-9326/aa673a
