Predicting Rainfall Onset and Cessation Within the West African Sahel Region Using Echo State Network

Rainfall onset and cessation are essential factors for the agrarian society in the West African Sahel region. This information determines crops to plant, helps project crop yields, and reduces loss due to climatic factors. Rainfall onset has also been associated with disease outbreaks; adequate knowledge will help in epidemiology and health management schemes. Therefore, predicting the onset will be necessary for the economy of countries in the Sahel region. Historical daily rainfall data from 1901 to 2000 for eight locations within the Sahel region were obtained from the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) at a 0.5° × 0.5° grid resolution. Rainfall onset and cessation were determined using the Liebmann method based on long-term precipitation in the location of interest. Onset and cessation were predicted using a machine learning approach—Echo State Network. The root mean square error for rainfall onset was 15–24 days, while a range of 9–19 days was obtained for rainfall cessation. Results obtained in this study suggest that it is possible to effectively predict rainfall onset and cessation within the Sahel region. The predictions will enable farmers and government agencies to address agricultural, health, and environmental issues associated with the onset and cessation of rainfall in the region.