Publications by Simon Proud


Evaluating EO-based canopy water stress from seasonally detrended NDVI and SIWSI with modeled evapotranspiration in the Senegal River Basin

Remote sensing of environment. 159 (2015) 57-69

JRL Olsen, Rasmus Fensholt, Simon R. Proud, Simon Stisen

Satellite remote sensing of vegetation parameters and stress is a key issue for semi-arid areas such as the Sahel, where vegetation is an important part of the natural resource base. In this study we examine if additional information can be obtained on intra-seasonal short term scale by using the Shortwave Infrared Water Stress Index (SIWSI) as compared to Normalized Difference Vegetation Index (NDVI). We perform a spatio-temporal evaluation of NDVI and SIWSI using geostationary remote sensing imagery from the Spinning Enhanced Visible and Infrared Imager (SEVIRI). The indices and their seasonally detrended anomalies are evaluated using a gridded rainfall product (RFE2) and modeled actual evapotranspiration (ETa) for the Senegal River basin in 2008. Daily NDVI and SIWSI were found spatially highly correlated to ETa with r=0.73 for both indices, showing the importance of the north/south vegetation gradient in the river catchment. The hypothesis that short term evolution of index anomalies are related to canopy water status was tested by comparing 10-day averages of ETa with short term changes in daily NDVI and SIWSI anomalies, and moderate to strong coefficients of determination where found when anomaly variations where aggregated by Land Cover Classes (LCCs) with R2 values of 0.65 for savanna, 0.60 for grassland, 0.72 for shrubland, and 0.58 for barren or sparsely vegetated areas. This is higher than for the same method applied to NDVI anomalies, with R2 values of 0.57 for savanna, 0.50 for grassland, 0.32 for shrubland, and 0.57 for barren or sparsely vegetated areas. The approach of detrending NIR/SWIR based indices and spatially aggregating the anomalies do offer improved detection of intra-seasonal stress. However, quite coarse spatial aggregation is found necessary for a significant analysis outcome.


Show full publication list