Remote Sensing and Hydrology (Proceedings of a symposium held at Jackson Hole, Wyoming, USA, September 2010) (IAHS Publ. 352, 2012). Copyright  2012 IAHS Press 221 Parameterization based on NOAA-AVHRR NDVI to improve conceptual rainfall–runoff modelling in a large West African catchment ALAIN DEZETTER 1 & DENIS RUELLAND 2 1 IRD, 2 CNRS, UMR HydroSciences Montpellier, Maison des Sciences de l’Eau, Université Montpellier II, Case Courrier MSE, Place Eugène Bataillon, F-34095 Montpellier Cedex 5, France alain.dezetter@ird.fr Abstract A conceptual hydrological model (GR2M) is applied to a large, poorly gauged catchment in West Africa. The purpose is to simulate the rainfall–runoff relationship at a monthly time step over the period 1982–2000 during which marked hydro-climatic changes took place. The model is based on two parameters: X1 for the production function and X2 for the routing function. The size of the production reservoir is usually fixed over time by using data from the FAO soil map. Hydroclimatic data consist of observed series of rainfall, PET and discharge data. The advantage of calibrating the size of the production reservoir by using spatiotemporal satellite NDVI data from NOAA-AVHRR images is investigated. Indeed, in a context of substantial climatic variability, or even of non-stationarity of the observed series, it may be difficult for conceptual models to reproduce runoff precisely over long periods of time. Introducing a spatiotemporal vegetation signal using NDVI data enables partial capture of the effect of the climatic and environmental variability on the functioning of the catchment. Calibrating the model using these additional forcing data significantly enhances the simulation results at the basin outlet, whatever the spatial complexity considered within the watershed through lumped or semi-distributed applications of the model. This study is a first step towards the design of a production function accounting for the spatiotemporal variability of a vegetation index. Key words hydrological modelling; GR2M; NOAA-AVHRR; NDVI; parameterization; Bani River INTRODUCTION Climate change and its effects on the environment and society are at the heart of current political and scientific preoccupations. The issue is vital for regions that are already weakened, such as West Africa. The effects of the decline observed in rainfall series in this region since the 1970s (Lebel et al., 2003; Ruelland et al., 2012) have sometimes been contradictory. Large West African rivers have suffered severely from drought, with runoff deficits sometimes exceeding the rainfall deficit by a factor of three, as in the Bani River catchment in Mali (Ruelland et al., 2008a). Paradoxically, small Sahelian hydrosystems have attenuated or even compensated the rainfall deficit (see e.g. Mahé et al., 2003; Séguis et al., 2004). The mechanisms underlying these contrasting behaviours are far from fully understood. However, in particular in Sahelian regions, it is possible that strong climatic variations combined with marked changes to landscapes under the effect of anthropogenic pressure may have resulted in substantial changes to land covers and hence runoff capacity or evapotranspiration. In this context, hydrological modelling is aimed at providing robust simulation tools and also facilities for understanding processes. The West African tropical zone, marked by strong inter- and intra-annual variability in precipitation and vegetation, also has an obvious shortage of reliable measurements for showing this variability with accuracy. The use of physically-based distributed models often comes up against this lack of data and remains limited to comparatively small experimental basins and to short periods of time (Andersen et al., 2001). In contrast, conceptual lumped and semi-distributed models are reputed for their robustness and ease of use. They provide space and time scales compatible with the evaluation and forecasting of water resources as they can be applied to large, poorly-gauged catchments over long periods of time (see e.g. Dezetter et al., 2008; Ruelland et al., 2009, 2010b, 2012). The GR2M model thus provided a satisfactory simulation of water resources in West Africa (Dezetter et al., 2008; Ardoin-Bardin et al., 2009). It also showed its limits in a context of marked hydro-climatic variability under non-stationary conditions (Diello, 2007; Paturel et al., 2009).