Are Equivalent Cross Sections the answer to the computational woes of Distributed Hydrologic Modelling?
Abstract:Distributed modelling or conceptual hydrologic modelling – this is a dilemma that hydrologists have grappled with since long. While distributed hydro-ecological models are conceptually elegant and physically defensible, are they practical to apply given the significant computational burden they come at? One possible way of improving their computational efficiency is presented here.
A new approach of modelling over an equivalent cross-section (ECS) is investigated. A homogenization test indicates that the representation of soil type is most critical in forming the ECS. If the soil type remains same within the sub-basin, a single ECS is formulated. If the soil type follows a specific pattern, i.e., different soil types near the centre of the river, middle of hillslope and ridge line, three ECSs (left bank, right bank and head water) are required. ECSs are formulated for 8 first order sub-basins and simulated using a 2-dimensional, Richards’ equation based distributed hydrological model. Simulated fluxes are multiplied by the weighted area of each ECS to calculate the total fluxes from the sub-basins.
To assess the accuracy of the ECS approach, the sub-basins are also divided into equally spaced multiple hillslope cross-sections. These cross-sections are simulated in fully distributed settings using the above model. The simulated fluxes are multiplied by the contributing area of each cross-section to get total fluxes from each sub-basin referred as reference fluxes. At the first order sub-basin scale, results show that the simulated fluxes using an ECS approach are very close to the reference fluxes and computational time is reduced of the order of ~4 to ~22 times compared to the fully distributed settings. Overall, the accuracy achieved in dominant fluxes, transpiration and soil evaporation, is higher than the other fluxes. Over a larger catchment with 822 sub-basins reasonable accuracy in simulated runoff against observed discharge is achieved. As a result, this approach provides a great potential for implementation of distributed hydrological models at regional scales.