A New Framework for Process Based Computationally Efficient Semi-distributed Hydrological Modelling

Abstract

While physically based distributed hydrological models are widely used to simulate spatially distributed hydrological fluxes, the computational time/effort associated with their use restrict their application in large catchments. In the work presented here, to reduce the computational time/effort in distributed hydrological modelling, catchments are first disaggregated into contiguous, topologically connected Hydrologic Response Units (HRUs), and later equivalent cross-sections (ECSs) formulated. To delineate the HRUs, a catchment is divided into four contiguous landforms and series of Strahler’s first order sub-basins on the basis of similarities in topographic and geomorphologic attributes. The adequacy of HRUs delineation is ascertained by unsaturated soil moisture movement modelling on hillslope cross-sections. The delineated HRUs are also verified by high resolution aerial photograph analysis and field work to confirm the existence of topographic and geomorphologic thresholds used in the delineation process. As a means to further enhance computational efficiency, the delineated landforms in the HRUs framework are next used to formulate a single or multiple ECSs in each first order sub-basins. The ECS refers to a cross-section which is representative of a part or an entire area of a first order sub-basin. To formulate the ECS, the topographic and physiographic variables of the entire sub-basin or part of the sub-basin are weighted in a systematic manner. The formulated ECSs for every sub-basin are modelled using a 2-dimensional Richards’ equation based distributed hydrologic model. Model parameters are obtained on the basis of actual topographical and physiographical features of a catchment. The simulated fluxes from the ECS approach are evaluated at the first order sub-basin scale with the distributed modelling fluxes and soil moisture observations, and found consistent. The ECS simulated fluxes at large catchment scale are compared with the observed runoff data and also found consistent. The proposed spatially distributed modelling framework using the ECS approach shows significant reduction in computational time/effort, i.e. ~4 to ~23 times at the scale of first order sub-basin and ~40 times at the scale of large catchment, while the loss of accuracy is negligible. Development of this approach will improve the usability of distributed modelling for large catchment scale simulations.