Résumé
Climatologists, breeders and agronomists need tools to translate variable climatic conditions in semi-arid environments into probable performance, and to translate genotype characteristics into agro-ecological fit of cultivars. Although geographic adaptation of West African sorghum cultivars is largely related to local characteristics of the wet season, they require, in addition to drought tolerance, effective phenological mechanisms for temporal escape from drought and excessive humidity that would favour pests and diseases during sensitive development stages. A simulation experiment was conducted to predict the potential and attainable (water limited) growth and yield of three sorghum genotypes differing in plant type and response to photoperiod for combinations of five sowing dates and three sites on the N-S climatic gradient in Mali, for the period from 1971 to 2004. The models used were SARRA-H equipped with the phenological model impatience. Onset and end dates of the rainy season were estimated with the simple soil water balance model BIP. The resulting scenarios were evaluated on the basis of (i) escape from drought, (ii) escape of grain development phases from periods of high pest and disease pressure and (iii) the resulting "safe" periods for sowing. The latter took into account the agronomic advantage associated with early sowing, observed by farmers to minimize weed competition and decreasing soil fertility during the wet season. The results indicated that potentially high yielding and photoperiod insensitive "modern" cultivars display an advantage only in the north where the rainy season is short. Sensitive response of flowering to photoperiod was essential for more humid environments having a long wet season, resulting in appropriate seasonal timing of flowering and greater flexibility of crop calendars. The methodology is discussed with respect to its suitability for applications in crop breeding, geographic targeting of existing cultivars and studies on the agricultural impact of climate variability and change. (C) 2007 Elsevier B.V. All rights reserved.