A Parsimonious Rainfall-Runoff Model for Flood Forecasting: Incorporating Spatially Varied Rainfall and Soil Moisture
Abstract
Compared to distributed models, lumped hydrological models for converting rainfall into runoff - such as units hydrographs - require less information, have a relatively simple structure, and do not suffer from over-parameterization. As an input to such models, though, we need to identify a single value of precipitation for the whole watershed, at each time step. Point depth measurements from rain gages are scarce, giving only very limited information regarding the areal coverage of rainfall. Furthermore, the response of a watershed to a rainfall event varies according to antecedent conditions, which can be indexed by considering either its initial soil moisture or else the magnitude of base flow at the beginning of the event. We explore solutions to such issues with a study on an 1,860 km2 sub-basin of the Obion-Forked Deer River system in Western Tennessee, which has a USGS stream-gaging station at Owl City. We use National Center for Environmental Prediction (NCEP) stage IV quantitative precipitation estimates (QPE) radar hourly precipitation data to analyze numerous events spanning 2010-2014. These data help us understand the spatial distribution of every rainfall event and identify its effective coverage and duration. To derive direct runoff for the basin, we utilize gaged streamflow data in conjunction with average radar precipitation, but only for those events with a spatial coverage in excess of 80% of the watershed's surface area. We use the European Space Agency (ESA) climate change initiative (CCI) soil moisture (SM) combined active-passive dataset V4.4 to find daily average soil moisture over our basin; these data give us the pre-existing watershed conditions. Finally, across all of the analyzed events, we check the correlations between the runoff coefficient, rainfall abstraction (-index), average precipitation depth, base flow at the beginning of the event, and soil moisture, and specify equations for calculating some of these variables, based on linear regression analyses. The validation of these equations with 11 storms of varying magnitude shows satisfactory results. Instead of representing the complex physical processes of runoff generation, we focus on the variability introduced in the hydrological response by the input rainfall and antecedent conditions. The unit hydrograph, derived from concurrent rainfall and streamflow records, is presented as a parsimonious rainfall-runoff model with very few parameters, that can use spatially varied rainfall data, and can be improved by using either soil moisture or streamflow data, as a measure of initial watershed conditions.
Publication Title
Watershed Management 2020: A Clear Vision of Watershed Management - Selected Papers from the Watershed Management Conference 2020
Recommended Citation
Sapkota, A., & Meier, C. (2020). A Parsimonious Rainfall-Runoff Model for Flood Forecasting: Incorporating Spatially Varied Rainfall and Soil Moisture. Watershed Management 2020: A Clear Vision of Watershed Management - Selected Papers from the Watershed Management Conference 2020, 183-196. Retrieved from https://digitalcommons.memphis.edu/facpubs/12965