Abstract
Monthly water balance models are valuable tools in water resources management. Unfortunately, most of these models suffer from structure complexity, large data requirements, occasionally low performance in simulating runoff, high seasonality of model residuals, and/or limited application in snow climates. By adding a seasonal component (SC), this study aims (1) to enhance the performance of the parsimonious GR2M monthly model (with only two free parameters) and (2) to increase its applicability in snow climates, without any further data requirements (as only precipitation and potential evapotranspiration monthly series are required to predict monthly runoff). The proposed model is examined at the catchment scale using 710 stations from the Hydro-Climatic Data Network, selected to cover various climatic conditions across the continental United States. Various performance criteria are calculated to compare the proposed GR2M-SC model with the original GR2M such as the Nash-Sutcliffe and Kling-Gupta efficiencies, the coefficient of determination, root mean square error and bias. Two split-sample sets are used to calibrate and validate the models. The calibration process is applied once using the first two-thirds of the dataset, and again using the last two-thirds, and each time the remaining sub-set is used for validation purposes. Furthermore, a seasonality measure is calculated to estimate the strength of seasonality in model residuals. The proposed GR2M-SC model clearly outperforms the original GR2M model as well as GR2M+, which is an improved 4-parameter version of the GR2M model with capabilities for handling snow accumulation and snowmelt. The seasonality of the proposed model residuals was also reduced.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have