Modeling of periodic and stochastic component of monthly rainfall in Senegal

Résumé

Analysis of hydroclimatic variables is crucial for the elaboration of local sustainable development policies, especially in countries with a rain fed agriculture. Hence, modeling the rainfall behavior might be very helpful, particularly in designing strategies to struggle against the climate change effects. This study aims at seeking representative models that fit the monthly rainfall in Senegal. Thus, the simple Mann-Kendal trend test is applied to the raw available records in order to assess the trend significance. Autocorrelogram curves are analyzed to check periodic behavior of the series. Harmonic analysis is performed for the periodic component designing. The periodogram curve is involved to determine the number of significant harmonics in the fundamental period. The stochastic components of the monthly rainfall are modeled using appropriate autoregressive process, performed upon the remaining series after extracting deterministic components. The raingauges are chosen according to the three climate bands covering the territory of Senegal and all the time series ranges from 1970 to 2010. The Mann-Kendal test show that trends of the monthly rainfall series can be neglected in the modeling scheme. The autocorelogram analysis shows a periodic behavior of the whole series. The retained significant harmonics according to the periodogram analysis are of four at Saint Louis, Dakar, Kaolack and of three at Ziguinchor and Tambacounda. A first order autoregressive process has been identified to model the stochastic components. The exploratory analysis of the evolution in time of simulated and observed time series shows good correlation. Further, the comparison of statistic parameters of the series such the mean, the standard deviation, the coefficient of variation and of determination confirms the good skill of the proposed models. Therefore, models can be retained for any monthly rainfall data requirement in the corresponding zones.