Utilisation du Modèle Agrométéorologique pour la Prévision des Rendements des Céréales au Mali : cas du mil, sorgho et maïs

Abstract

The conceptual models of yield prediction were made from water balance indices related to millet, sorghum and maize crops in the Sikasso, Ségou and Mopti departments. The objective is to use an agrometeorological model (AMS) to pre-calculate water balance indices from agrometeorological data. Crop yield data and water balance indices from AMS were combined to establish regression models. Conceptual models of yield prediction are configured from the combination of two to four water balance indices as explanatory variables. The water balance indices with a better predictability of yields are: TWR (total water requirements of the plant), INDXhar (end-of-cycle water satisfaction indices), Cr1a (Decades to which the Rangeland index crosses the line defined by 0.4*PET), WEXt (total value of water excess at the end of the cycle), SWi (Initial content in soil water) and ETA (Actual evapotranspiration at initialization, stages, flowering and harvesting). The best models developed were validated by comparing their estimates to actual values of returns. The cross-validation of the millet yield forecast model of the Ségou circle indicates the following statistics: R²cv = 0.51, RMSE = 144 Kg/ha and the statistics are R²cv = 0.45, RMSE = 137 Kg/ha for the sorghum from the circle of Sikasso. On the other hand, the validation statistics are not very significant for the Mopti circle with coefficients of determination R²cv = 0.27; R²cv = 0.20 and R²cv = 0.16 respectively for millet, sorghum and maize. In this study, conceptual predictive models developed from agronomic and climatic data provided acceptable estimates of millet, sorghum and maize crop yields in the Sikasso, Ségou and Mopti departments.