Assessing Sahelian vegetation and stress from seasonal time series of polar orbiting and geostationary satellite imagery: Limitations and new potentials
Research output: Book/Report › Ph.D. thesis › Research
Vegetation in the semi-arid Sahel has been widely studied using remote sensing data acquired from satellite based instruments. Due to the size of the region and limited ground based monitoring networks, remote sensing is the only feasible approach for quantifying spatial and temporal variation of vegetation across the region. The vegetation growth is mainly water restricted and since annual cumulative rainfall varies from year to year, so does the conditions for vegetation growth. Much of the current knowledge of Sahelian vegetation status and long term development is based on commonly applied vegetation remote sensing methodology. The objective of this Ph.D. study has been to examine limitations of these methods and to investigate the potential for improvement. Currently the most widely applied use of remote sensing data for vegetation analysis is based on the normalized difference vegetation index (NDVI), which combines red and near infrared (NIR) spectral regions. From NDVI data a greening of the Sahel have been identified since the 80s and attributed to increasing trends in annual rainfall for large parts of the region. One part of this thesis analyses time series of parameterized MODIS NDVI using the unique field data set from the Widou Thiengoly test site in northern Senegal. The field data have been collected under controlled grazing intensities. From this data a very clear effect of grazing on plant species composition and NPP/NDVI relationships is found. It is suggested that the varying NPP/NDVI relationships, combined with the large increase in livestock of the Sahel in recent decades, means that the greening of the Sahel cannot uncritically be interpreted as a positive trend in vegetation productivity due to increasing rainfall. It can also represent grazing induced changes in species composition which covers neutral or even decreasing trends in biomass production. For monitoring vegetation status on a shorter time scale in the Sahel, the NDVI may not be the most appropriate index. From previous research it has been suggested that the Shortwave infrared (SWIR) spectral region provide good sensitivity to canopy water content, which can make vegetation stress detection possible. Furthermore, the high frequency observations in the optical spectrum now available from geostationary instruments have the potential for detection of changes in vegetation related surface properties on short timescales, which are challenging from polar orbiting instruments. Geostationary NDVI and the NIR and SWIR based Shortwave Infrared Water Stress Index (SIWSI) indices are compared with extensive field data from the Dahra site, supplemented by data from the Agoufou and Demokeya sites. The indices are also compared with output from hydrological modelling of the Senegal River basin and a gridded rainfall product. It was shown that NIR/SWIR based indices are better suited for assessing water stress related changes in vegetation status, as compared to the more commonly applied NDVI. The hypothesis that short term variations in anomalies from seasonally detrended time series of indices could carry information on vegetation stress was examined and confirmed. However, it was not found sufficiently robust on pixel level to be implemented for monitoring vegetation water stress on a per-pixel basis. By detrending time series for the Senegal River basin, spatially coherent patterns of index trends were found, suggesting that real information is present in the daily anomalies of both NDVI and SIWSI. It was found necessary with spatial aggregation on quite coarse scale for significant analysis results when comparing with modeled actual evapotranspiration. The Dahra test site and the data acquired there are also presented in detail. Together with the AMMA Gourma super site in Mali and the Demokeya test site in Sudan, this is one of the few Sahelian sites providing observations of ecosystem properties and hydrological and meteorological variables over many years.
|Publisher||Department of Geosciences and Natural Resource Management, Faculty of Science, University of Copenhagen|
|Number of pages||190|
|Publication status||Published - 12 Jun 2014|