Land degradation monitoring in Sahelian Africa from both NDVI and rainfall data

The African Sahel has been a very controversial region in the climatic change debate during the last decades. From the studies that announced a dramatic advance of the desert in the 1970s [Lamprey 1975] and lead to the first conference about desertification [Nairobi, 1977] to the hopeful articles assuming a global recovery from the great droughts of the 70s and 80s [Prince et al., 1998; Hermann et al., 2005], many works have been carried out about this region. The studies of recent years are often characterized by their emphasis in the analysis of climatic fluctuations, specifically inter-annual and inter-decadal variations, as it is necessary to avoid misinterpretations related to climatic variability. Even if some authors have noticed a recovery as a result of increased rainfall since the mid 90s, the rainfall level of the 50s and 60s has never been reached again. It is not clear if a true climatic change is happening nowadays in the region, leading to degradation and desertification as a consequence of carry over effects from precedent droughts and reduced rainfall or if the changes observed are principally human-induced. Thus, two main variables are the most responsible of the observed changes in the last decades: climatic variability, related primarily to inter-annual and inter-decadal rainfall variability, and the demographic boom. Obviously, discriminate between climatic and human induced land degradation/improvement in a regional or global scale is an important matter that should be treated carefully. Following Evans et al. (2004) and Wessels et al. (2007) we have developed a method using the residuals of the NDVI-Rainfall relationship to retrieve the vegetation response in function of rainfall, so that climatic variability is cleaned off and human degradation is localized. The only remaining question is that the carry-over effects of precedent years are not taken into consideration and as a consequence detected trends cannot be only attributed to human factors. The evolution of Rain Use Efficiency RUE, the ratio of net primary production and rainfall, is also used here, as some authors claimed from observed RUE trends during the period 1982-2002 that degradation occurred in large areas of the Sahel, and we wanted to test this assumption. The studied region is depicted at figure 1. The data used as a proxy of vegetation seasonal activity are taken from the GIMMS NDVI database, which is corrected for residual sensor degradation and sensor inter-calibration differences, effects of changing solar zenith and viewing angles, volcanic aerosols, atmospheric water vapor and cloud cover [Tucker et al., 2005]. The spatial resolution of this database is 8 km and it covers the period from 1981 to 2006 on a bi-weekly basis. The data used as rainfall is the CRU 2.1 rainfall product [Mitchell et al., 2005]. This product is based on the interpolation from in-situ measurements and it offers a good spatial resolution (0,5o); some problems can be found in highlands and very vegetated areas, that do not correspond to our study region. It covers the period from 1901-2002 (recently expanded) on a monthly basis. The GIMMS NDVI data have been re-sampled to match the CRU Rainfall at the same spatial and temporal resolutions. ha l-0 06 24 22 3, v er si on 1 16 S ep 2 01 1 Author manuscript, published in "LES SATELLITES GRAND CHAMP pour le suivi de l'environnement, des ressources naturelles et des risques, Clermont-Ferrand : France (2010)"