Evaluation and Comparison of Remote Sensing Based Precipitation Products in Casamance basin, SENEGAL

Abstract

The role that rainfall can play in monitoring climate, floods and drought, but also in several socio-economic activities, namely agriculture and water resources management in Africa and especially in Senegal is very important. However, the availability and accessibility of these rainfall data are difficult due to the low number of functional rainfall station networks providing complete and reliable data, but also due to the high cost of these rainfall data. The use of remote sensing, through rainfall products, could be considered as an alternative to this problem. To be sure of obtaining complete and reliable data, evaluation and comparison studies of these satellite products are necessary to validate their effectiveness for efficient use in a given area. Our study is based on the evaluation and comparison of precipitation products derived from remote sensing in the Casamance basin, located in Southern Senegal. Under the influence of the Southern Sudanese climate, the basin covers an area of 20150 km2 with relatively flat landforms and low altitudes. There are two types of seasons: the rainy season and the dry season. Four remote sensing derived rainfall products have been considered; ARC 2, RFE 2, CHIRPS-0.05 and TRMM-3B42RT. The evaluation of these rainfall products, over the Casamance basin, was carried out at four-time steps: daily, monthly, annual and seasonal using statistical equations, statistical tests and an assessment of the accuracy of the rainfall products on the spatial distribution of rainfall. The results obtained from the daily assessment showed a weak correlation between the estimated rainfall data from remote sensing derived rainfall products and the observed data from the rainfall stations. However, ARC 2 and CHIRPS presented the best correlation and comparatively lower, RMSE, MAE and BIAS results. Evaluation of estimated monthly rainfall from remote sensing derived rainfall products showed good performance of CHIRPS-0.05 with the best statistical results (r=0.90, RMSE=66.80, MAE=32.01 and BIAS=0.99). ARC 2 and RFE 2 also performed well with correlations 0.85 and 0.70, respectively. However, RFE 2 obtained a high RMSE, MAE and BIAS. TRMM-3B42RT performed worst with a negative correlation (-0.55). Despite the good results recorded by the rainfall products at the monthly time scale, low correlations ranging from 0.46 to -0.31 were obtained as results from the analysis of annual rainfall. The observed maximum rainfall amounts were underestimated by ARC 2, CHIRPS-0.05 and RFE 2. However, CHIRPS overestimated the average amount of observed rainfall. The tendency of CHIRPS to overestimate precipitation and of ARC 2 and RFE 2 to underestimate it was further demonstrated during the spatial assessment of precipitation. This analysis also showed that the difference in altitudes could affect the accuracy of remote sensing products in estimating precipitation. Seasonal changes also have an impact on remote sensing derived rainfall products estimates. Overall, the results obtained highlighted the good performance of CHIRPS-0.05 and ARC 2 rainfall products and the poor performance of TRMM-3B42RT product on all time scales at Casamance basin in Senegal. On the other hand, RFE 2 performed well only with monthly time scale and spatial rainfall assessments and poorly with daily and annual time scale assessments.