Details

Autor(en) / Beteiligte

• Lebel, T.
• Le Barbé, L.

Titel

Rainfall monitoring during HAPEX-Sahel. 2. Point and areal estimation at the event and seasonal scales

Ist Teil von

• Journal of hydrology (Amsterdam), 1997-02, Vol.188 (1-4), p.97-122

Ort / Verlag

Elsevier B.V

Erscheinungsjahr

1997

Quelle

Elsevier Journal Backfiles on ScienceDirect (DFG Nationallizenzen)

Beschreibungen/Notizen

• The water and energy balance models that will be used in HAPEX-Sahel to analyze the data collected during the experiment are strongly conditioned by the rainfall estimation accuracy over the areas of interest. Following the description of the rainfall conditions that prevailed during HAPEX-Sahel and the computation of statistics characterising the point rainfall processfor the Sahel, presented in a companion paper, it is examined here how accurate are the areal rainfall estimates provided by the EPSAT-Niger network at the event and seasonal scales. Using a geostatistical framework, it is shown that it is possible to infer a climatological variogram that represents the average variability of the event rainfields, within the limits imposed by the resolution and sampling window of the network. Average event rainfall estimation errors are derived. The density of the EPSAT-Niger mesoscale network (one station for 150 km 2) allows the estimation of the event areal rainfall over the 1° × 1° square with an average uncertainty around 5%. On the supersites (100–750 km 2) the average uncertainty can be as large as 20%, depending on the site and rainfall considered. This uncertainty is reduced by two-thirds, or more, when using the denser network designed specifically to cover the supersites. It is then demonstrated how the variability within the event rainfields and the space-distribution of the rainy events combine to determine the spatial structure of a rainfield resulting from the accumulation of several event rainfelds. Over the HAPEX-Sahel study area, the within-the-event-rainfield variability is dominant over the variability resulting from the distribution of the events in space. Consequently the correlation length does not vary much when shifting from the event to the seasonal scale. At an unsampled location the seasonal rainfall error magnitude is on the order of the average error at the event scale, divided by the square root of the number of events recorded during the season.