Author's personal copy Can a 25-year trend in Soudano-Sahelian vegetation dynamics be interpreted in terms of land use change? A remote sensing approach Agne `s Be ´ gue ´ a, *, Elodie Vintrou a , Denis Ruelland b , Maxime Claden c , Nadine Dessay d a CIRAD, UMR TETIS, 500 rue J-F. Breton, Montpellier F-34093, France b CNRS, UMR 5569 HydroSciences, Universite ´ Montpellier II, Place E. Bataillon, Montpellier F-34093, France c Cemagref, UMR TETIS, 500 rue J-F. Breton, Montpellier F-34093, France d IRD, UMR ESPACE-DEV, 500 rue J-F. Breton, Montpellier F-34093, France 1. Introduction The Sahel/Sudanian belt in Africa, a rainfed agricultural zone, appears to be a particularly vulnerable environment subject to major changes due to climate and human activities. The region has undergone marked variations in rainfall in recent decades (Nicholson, 2005) that impacted both the population and the ecosystems, as well as a twofold increase in population since the 1960s (UNEP, 2004). The effects of climate change and human pressure on vegetation dynamics are not well understood. Remote sensing has proven to be a very efficient tool to characterize vegetation dynamics in large regions. In particular, images from the Advanced Very High Resolution Radiometer (AVHRR) on board the National Oceanic and Atmospheric Administration (NOAA) satellite have been widely used for global studies of Sahel/Sudanian environments due to the extensive archive (dating back to 1982), the free data, the suitable wavebands and the high temporal resolution. Using NOAA–AVHRR databases (GIMMS, Pathfinder or ARTEMIS), several authors interpreted time series of the Normalized Difference Vegetation Index (NDVI) in terms of ecosystem degradation or recovery, particularly in arid and semi-arid areas (e.g., Hountondji et al., 2004; Pouchin et al., 2002; Prince et al., 1998; Wessels et al., 2007). Because of subtle temporal variations and a broad scale signal, the results of these studies are sometimes difficult to interpret and may even appear to be contradictory. In several parts of the world, a number of authors have tried to relate the NDVI trends to land use. The spatial variability of the NDVI has been explained by ground-based data (e.g., Wessels et al., 2007), by using national or international databases such as FAOstats or PNUD (e.g., Olsson et al., 2005; Seaquist et al., 2009), Global Land Cover 2000 (Bai et al., 2008), or by interpreting high resolution images (e.g., Evans and Geerken, 2004; Li et al., 2004; Ruelland et al., 2010). However, the trend itself has rarely been linked to spatial land cover/land use dynamics. Evans and Geerken (2004) illustrated NDVI trends with a time series of high resolution satellite images of representative areas, but in a descriptive way. The difficulty in obtaining dynamic ground data sets or land cover maps at a regional scale and with a sufficient time depth is one of the reasons for this gap in research (Ruelland et al., 2011). Global Environmental Change 21 (2011) 413–420 ARTICLE INFO Article history: Received 13 July 2010 Received in revised form 21 January 2011 Accepted 11 February 2011 Available online 10 March 2011 Keywords: Land use changes Human impact Sahel NDVI time series Rainfall Trend ABSTRACT This study is based on the premise that, in the Sahel/Sudanian belt of Africa, the main determinants of interannual variation in vegetation dynamics are rainfall and land cover type. We analyzed the spatio- temporal sensitivity of the NOAA–AVHRR 8 km-resolution vegetation index (NDVI) to (i) annual rainfall variability (0.58  0.58 resolution) acquired over a 25-year period (1982–2006); and (ii) land use changes in the different eco-climatic regions of the Bani catchment in Mali (130 000 km 2 ). During the period 1982–2006, there was no clear trend in rainfall over the catchment, whereas there was a strong positive trend in the NDVI, both when the NDVI values were corrected using annual rainfall variability and when they were not. We divided the catchment into three eco-climatic regions based on the relationship between the annual NDVI and rainfall. In each region, we analyzed the observed greening in relation to changes in land use after correcting for the effect of annual rainfall on the NDVI. Results show that there is a mixed level of agreement between the land cover changes at the grid cell scale and the spatial pattern of the NDVI trend. Increased cropping does not explain the increase in the annual NDVI, except in the Sahelian part of the catchment. We hypothesize that the natural vegetation dynamics related to the non- linear rainfall patterns during the 25-year study period were responsible for these results. ß 2011 Elsevier Ltd. All rights reserved. * Corresponding author. Tel.: +33 4 67 54 87 54; fax: +33 4 67 54 87 00. E-mail addresses: agnes.begue@cirad.fr (A. Be ´ gue ´), ruelland@msem.univ-montp2.fr (D. Ruelland), maxime.claden@teledetection.fr (M. Claden), nadine.dessay@ird.fr (N. Dessay). Contents lists available at ScienceDirect Global Environmental Change journal homepage: www.elsevier.com/locate/gloenvcha 0959-3780/$ – see front matter ß 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.gloenvcha.2011.02.002