6533b85dfe1ef96bd12bef4d

RESEARCH PRODUCT

Spatial coherence and potential predictability of intraseasonal descriptors of the rainy season in Soudano-Sahelian Africa : application to the pearl millet crop in Niamey area

subject

description

The aim of this thesis is twofold : (i) fill a lack of knowledge about the spatial coherence and seasonal predictability of the intraseasonal characteristics (ISC) of the soudano-sahelian rainy season. These ISC are the seasonal rainfall amount (S), the daily rainfall frequency > 1 mm (O), the daily rainfall mean intensity (I), the dates of the onset and withdrawal of the rainy season, the mean length of the dry (LDS) and/or wet spells (LWS) ; (ii) document the climate-agriculture relationship over the Niamey area based on the millet crop example.From FRIEND-AOC daily rainfall records (1950-2000) for a 136 stations-network located in Senegal, Mali, Burkina-Faso and Niger, the spatial coherence, estimated through DOF, var[SAI], EOF, correlation, is not equivalent between the intra-seasonal characteristics. Seasonal rainfall and daily rainfall frequency anomalies have a substantial interannual spatial coherence. Conversely, the spatial coherence of daily mean intensity, onset and withdrawal dates of the rainy season and mean length of dry/wet spells interannual anomalies is weak. Consequently, the regional signal of the seasonal amount interannual variability seems mainly related to the in-phase modulation of the daily rainfall frequency. Potential predictability of seasonal amount, daily rainfall frequency, onset and withdrawal dates of the rainy season estimated from a 24-member ensemble of simulations made with the ECHAM 4.5 GCM forced by observed SSTs, is poor. On the other hand, the use of a model output statistics approached based on simulated rainfall or atmospheric dynamics enhances the skill of the seasonal amount and daily rainfall frequency hindcasts.Lastly, the relationship between the sowing date – rainy season onset date – and yield has been analysed using data collected from on-farm surveys of pearl millet crops, and rain-gauges records between 2004 to 2007 over the AMMA-CATCH Niger supersite. Results show that: (i) most of the farmers wait for the first rainy event greater than 10 mm to sow; (ii) the sowing waves are usually synchronized with the mesoscale onset date rather than the agronomic onset date; (iii) the sensitivity of grain yields, evaluated from SARRA-H crop model simulations initialized with different sowing dates (i.e. observed sowing date, agronomic and hydrologic onset date), is weak.