干旱区生态环境知识资源中心

Reference Evapotranspiration (ETo) is one of the major components of the hydrological cycle that is very essential in water resources planning, irrigation and drainage management and several other hydrology processes. In irrigation system and design, the prediction of ETo is vital and indispensable for the quantification of crop water needs. This study investigates the capabilities of hybridized fuzzy model with firefly algorithm (ANFIS-FA) for predicting daily reference evapotranspiration over Burkina Faso region. Metrological information at Bobo Dioulasso, Bur Dedougou, and Ouahigouya stations, in Sahelian, Sudano-Sahelian, and Sudanian zone, are used for modelling development. Six different climatic input variable combinations corresponding to 6 models are inspected. The daily Penman-Monteith reference evapotranspiration for the time-period (1998-2012) are used to train and test the models. Several numerical indicators in addition to Taylor diagram are considered to evaluate the performance of the models. Results indicated that the hybrid ANFIS-FA model outperformed the classical ANFIS-based model for all three stations and the model with full inputs climatic data gave the best results. Furthermore, ANFIS-FA is performed the best for Bur Dedougou (Sahalian-Soudanian region) and less at Ouahigouya (sahalian region). In quantitative terms and for instance Bur Dedougou station, ANFIS-FA model increased the prediction accuracy remarkably (SI = 0.043, R-2 = 0.97, MAPE = 0.035 and RMSE = 0.24) compared with ANFIS-based model (SI = 0.068, R-2 = 0.89, MAPE = 0.037 and RMSE = 0.378). Results revealed the influence of utilizing nature-inspired firefly algorithm to substantially improve performance of the classical ANFIS model optimization for the applied application. Also, the applied modelling strategy can bring a trustful expert intelligent system for predicting reference evapotranspiration at the west desert of Africa.