Modélisation et prédiction multi-site des précipitations annuelles et mensuelles dans le bassin versant du Cheliff par le filtre de Kalman (Agérie)

Abstract

The objective of this work is to test the forcefulness of the discrete Kalman filter in the hydrological domain and more precisely in the multi-site prediction of rainfall in the Middle Cheliff watershed area (Algeria), the improvement of the error of prediction for an annual and monthly step, and for a spatial and temporal dimension. To do this, rainfall data from five stations are used in our study area over a period of 11 years (2000-2010). The results obtained show that the differences between the observations and the predictions are minimal and this for the different steps and dimensions. Finally, the performance of the Kalman filter was checked using trace graphs, FK gain and standard deviations, as well as calculating the mean relative error of the predictions, which is less than 10%. It was also found that the results obtained for the monthly step are more precise than those obtained for the annual step, which allows to say that the Kalman filter is better for the monthly step compared to the annual step.