Prédiction de quelques normales agroclimatiques à l'aide de modèles empiriques

Predicting some Agroclimatic Data Using Empirical Models. The lack of climatic data in large areas of Africa is a major drawback in the efficient planning of land resources use. The objective of this study is to fill that gap whenever necessary, by estimating climatic data with empirical models using geographic parameters of the site such as its latitude, longitude, altitude and month of the year. Data from 66 agroclimatological stations were used, covering a region stretching between latitude 2 and 14N, longitude 3 and 23E, and altitude 13 to 1890 m. Average monthly maximum and minimum temperatures, rainfall and sunshine duration were computed for each of the stations, thus generating 792 monthly observations for each of the 4 climatic parameters. Polynomial simulations were used to derive predictive equations for these parameters, with coefficients of determination (R2) of 0.997, 0.990, 0.879 and 0.986, respectively. The values of the geometric mean error ratio were close to 1 for the 4 models (0.9998, 1.0035, 1.354 and 1.0107, respectively). Likewise, the geometric standard deviations of the error ratios for the respective models were 1.059, 1.116, 2.257 and 1.175. All these statistical evaluations are very satisfactory, which shows the very good fitness of the proposed models.