Towards an Efficient Rainfall–Runoff Model through Partitioning Scheme

Abstract

Partitioning Scheme (PS) is one of the strategies that could play a constructive role in improving the performance of conceptual rainfall-runoff (CRR) models. The main objective of this paper is to develop a Rainfall Runoff-Partitioning Scheme (RR-PS) with the ability of distinguishing different flow regimes in a basin and simulating each regime separately. The model utilizes a combination of PS and "regular" procedures and is equipped with Fuzzy C-Means (FCM) and Seasonal Partitioning (SP) to recognize different flow regimes. In addition, to avoid excessive increase of the model parameters caused by PS, sensitivity analysis is used to recognize the sensitive parameters. The PS part of integrated model is only run for the "sensitive" parameters and the "regular" part of model is implemented for the "less-sensitive" parameters. Data from three different scale basins in USA and Iran are used to evaluate the models. A HBV-Light (Hydrologiska Byråns Vattenbalansavdelning-Light model) based CRR model (Improved HBV-IHBV) is developed in FORTRAN (Formula Translation) with several modifications for testing the procedures. The results show that in all cases IHBV-based models using PS method perform better than the regular IHBV model. Overall, IHBV-SP demonstrates better performance than the other PS based models. Further improvement is reached by adopting user-defined seasons in IHBV-SP through optimization.