Science et changements planétaires / Sécheresse


Representation of rainfall in regional climate models and application to millet yield estimations in Senegal Volume 23, issue 1, Janvier-Février-Mars 2012


See all figures

Université Cheikh Anta Diop de Dakar (UCAD) Laboratoire de physique de l’atmosphère et l’océan Siméon Fongang (LPAO-SF) École doctorale Eau, qualité et usage de l’Eau (EDEQUE) BP 5085 Dakar-Fann Dakar Sénégal, Centre d’étude régional pour l’amélioration de l’adaptation à la sécheresse (CERAAS) BP 3320 Thiès Escale Thiès Sénégal, IRD Laboratoire d’océanographie et climat : expérimentations et analyses numériques (LOCEAN) Institut Pierre Simon Laplace 4, place Jussieu Case 100 75252 Paris cedex 5 France, University of Tokyo Tokyo Japan, Cirad UMR AGAP F-34398 Montpellier France, Africa Rice Center AfricaRice Sahel Regional Station BP 96 Saint-Louis Sénégal, CNRS Laboratoire de météorologie dynamique (LMD) Institut Pierre Simon Laplace Ecole Polytechnique Aile 5 - LMD Route de Saclay 91128 Palaiseau cedex France

The strong influence of climatic factors on agriculture and food security in sub-Saharan Africa in addition to climate change perspectives have prompted the scientific community to document the impacts of climate in this region. However, if many studies quantifying the impacts of climate rely on downscaling, very few address the uncertainty associated with their use. However, the choice of a particular method and of a particular regional model can strongly influence the final result since crop models are very sensitive to the quality of the input climate forcing. The objective of this study is to address this issue by analysing the dispersion of rainfall provided by eight regional models and how this dispersion spreads in the estimation of millet yields in Senegal. The SARRAH crop model is used to simulate millet yields. The study shows that there is a wide dispersion in the representation of rainfall from one regional model to another (and even sometimes for the same regional model with two sets of parameters) at both the seasonal and intra-seasonal scales. These biases introduce significant errors in estimating the agronomic impacts, which might invalidate conclusions about the impacts of climate change based on the use of a single regional model. The use of a bias correction method is indispensable.