A multiple hydrograph separation technique for identifying hydrological model structures and an interpretation of dominant process controls on flow duration curves

Abstract

AbstractThere has been significant progress made towards understanding the process controls that govern the shapes of flow duration curves (FDC). However, the transferability of the knowledge gained to large spatial scales is still lacking, partly due to the fixed representation of the model structures representing the unobserved underground layer. In the context of improving large scale applicability/transferability of FDC studies: (1) we propose a novel approach to transform/restructure the representation of the subsurface component in hydrological models to be flexible and adaptable in any environment and (2) provide a thorough interpretation of the dominant processes that impact the shape of FDCs. The flexible structure is designed from a series of multiple connected linear reservoirs that are developed from a unique multiple hydrograph separation procedure, to estimate the number of distinct hydrological processes at play. The main results are: (1) wet catchments have more complex model structures (more reservoirs) than arid catchments; indicating that subsurface processes are potentially more present in wet catchments than in drier ones. Dry catchments have more near surface occurring processes, thus have less complex model structures. However, more complexity is potentially required within the internal reservoirs of dry catchments. (2) The strength of interference in the intermediate reservoirs in controlling the dominant processes towards the bottom determined the shape of FDCs, and this increased with aridity. The novel finding is that a model structure can be determined by the catchment's ability to generate flows as well as the strength of low flow persistence. Hence, we propose a conceptual framework for the development of flexible model structures based on the shapes of FDCs. The transformation of models to be flexible in this research, should contribute to studies that seek the transferability of models over large spatial scales.