Fuzzy logic-based rainfall-runoff modelling using soil moisture measurements to represent system state

Initial catchment state, such as soil moisture, strongly controls rainfall-runoff transformation processes. However, due to the high spatial and temporal variability of soil moisture, point measurements may not always be suitable to represent the actual system state of a whole catchment as required in distributed catchment modelling. In this study a fuzzy rule-based system (FRBS) using the Takagi-Sugeno-Kang approach has been developed using soil moisture and rainfall as input variables to predict the actual discharge at the catchment outlet. Four soil moisture probes from the hydrological test site Durreych (Black Forest, southwest Germany) were selected, each of them representing a particular runoff generation process (saturation excess flow, infiltration excess flow, slow and fast interflow, return flow). After manual calibration, the simulated peak discharges were very similar to the measured values. Furthermore, the pattern of rule activation in the FRBS reflected the complex, highly nonlinear behaviour of the catchment. Thus, in the FRBS framework, the measurements of soil moisture at representative locations could be used as representation for the actual system state, allowing for an entirely data-driven prediction of the runoff response using rainfall.