A Novel Approach for Precipitation Modeling Using Artificial Intelligence-based Ensemble Models

Precipitation is the primary component of the hydrologic water cycle and its accurate prediction plays a significant role for the planning, management and design of hydraulic structures. The objective of the study intended to explore a new approach to increase prediction efficiency of precipitation in the arid, semiarid and humid zones. The approach was implemented using relevant data from stations in Iraq and Nigeria. Support vector regression (SVR), artificial neural network (ANN) and adaptive neuro-fuzzy inference system models (ANFIS) were applied for a single modeling of 30 years monthly average precipitation data. Thereafter, a nonlinear ensemble and 2 linear ensemble techniques were applied to improve prediction efficiency and reliability of the single models. Error measures as well as a goodness of fit measure were the criteria employed to assess the performance potential of the models. Based on the results of this work it was found that ensemble modeling could improve prediction accuracy of the single models as much as 38% in the validation phase. It was also confirmed that the performance of artificial intelligence-based models can efficiently be improved by the application of ensemble modeling in the arid, semiarid and humid study stations of Iraq and Nigeria.