干旱区生态环境知识资源中心

The variability of rainfall is a key component determining how the continental surfaces react to the atmospheric forcing. When studying the impact of climatic fluctuations onto the water resources, it is thus of paramount importance to evaluate to which extent the atmospheric models used in this kind of studies are able to reproduce the variability of the rain process, both in space and time. First among these are the general circulation models (GCM) with coarse resolution, which has two consequences: (i) a simplified parametrisation of convection; (ii) a scale of representation of rainfields which is not adequate when it comes to use them as inputs to hydrologic models. Since linking GCM's and regional hydrologic models is the corner stone of impact studies, it is necessary to analyse the consequences of this gap in scales and to find ways of bridging it. As a preliminary step in that direction, a comparative analysis of the observed and of the GCM rainfall variabilities is carried out for a tropical semi-arid zone of West Africa displaying a high sensitivity to climatic fluctuations. Over tropical regions the GCM used here (LMD-6) has a space resolution of 1.6 degrees in latitude and of 3.75 degrees in longitude. The comparative study shows that the errors of the GCM rainfall outputs may be traced down to two fundamental shortcomings: (i) a wrong seasonal cycle, probably linked to problems in representing the large scale circulation; (ii) an unrealistic simulation of the mesoscale convective systems that are responsible for 90% of the rainfall over this area. This latter problem is especially damaging from an hydrological point of view, as shown from a detailed analysis of high resolution rainfall observations. Even though it is possible to design rainfall desaggregation models producing realistic small scale rainfields from large scale rainfields, such models are of limited utility as long as atmospheric models are nor able to produce a realistic climatology in term of number and magnitude of convective systems.