Watershed Model Parameter Estimation in Low Data Environments

Abstract

Watershed models are essential for evaluating the impact of watershed management; however, they contain many parameters that are not directly measurable. These parameters are commonly estimated by calibration against observed data, often streamflow. Unfortunately, many areas lack long-term streamflow records, making parameter estimation in low data environments (LDE) challenging. A new calibration technique, simultaneous multi-basin calibration (MBC), was developed to estimate model parameters in LDE. Three Soil and Water Assessment Tool (SWAT) model initializations for USGS gages with ~ 2-year records in the Lake Champlain Basin of Vermont, USA, were evaluated by comparing MBC and the commonly used similarity-based regionalization (SBR) approach, where calibrated parameters from a watershed with an extended data record are transferred to the LDE receptor watersheds. In MBC, each watershed is initialized, and observed flows from each initialization are aggregated to generate a combined streamflow record of sufficient length to calibrate using a differential evolution algorithm. New hydrological insights for the region: Using this new MBC method, we demonstrate improved model performance and more realistic model parameter values. This study demonstrates that short periods of hydrological measurement from multiple locations in a basin can represent a system similarly to long term measurements and that even short records taken at multiple locations significantly improve our hydrologic knowledge of a system as compared to relying on the similarity of a basin with a long record of flow. In addition, this study revealed that the hydrologic response is mediated by the interplay of very low soil-saturated hydraulic conductivity (Ksat) and cracking soils. As a result, even if Ksat is very low, cracking clays have a large impact on runoff production Garna et al. (2022). IntroductionWatershed scale hydrological models are often used to quantify and predict diverse and complicated hydrologic system behavior and are often used to manage environmental and water resources. A process based hydrological model consists of equations and parameters that describe watershed processes such as snow accumulation, groundwater flow, runoff generation, and

Employing higher density lower reliability weather data from the Global Historical Climatology Network monitors to generate serially complete weather data for watershed modelling

“…Due to the potential influence of changing conditions on hydrological simulations, the conventional split‐sampling technique may fail to encompass the full range of variability in the watershed. Opting for full‐time series calibration, as recommended by Arsenault et al (2018), Garna et al (2023), and Singh and Bárdossy (2012) proves to be a more robust alternative. Thus, we performed the calibration of our model for daily streamflow simulation across the entire duration of the dataset (2010 to 2021).…”