Evaluation of a Tank-Based Hydrological Model for Streamflow Simulation at the Rio Grande, Big Bend National Park, Texas

Hydrological modeling plays a crucial role in assessing water resources and developing strategies for sustainable watershed management. This study evaluates the performance of a systematically recalibrated and rigorously validated adaptation of the Sugawara Tank Model in reproducing streamflow at the Rio Grande site near Rio Grande Village, Big Bend National Park, Texas, USA (USGS site 08375300). While the Tank Model concept has existed for decades, its application to a semi-arid, data-scarce transboundary catchment at sub-daily temporal resolution, coupled with a systematic comparison of four derivative-free optimization (DFO) methods for automated parameter estimation, represents the novel contribution of this work. Specifically, the Nelder-Mead simplex, Powell's method, Pattern Search (GPS), and COBYLA algorithms were evaluated and compared for their calibration efficiency and convergence behavior, a comparative DFO framework that has not previously been applied to Tank Model calibration in arid transboundary river systems. The model conceptualizes the catchment as four vertically arranged tanks representing surface runoff, shallow soil infiltration, and deep groundwater flow. A further novelty lies in the hourly calibration framework applied over an exceptionally long 17-year record (2008 to 2025), enabling the model to capture both episodic flash-flood dynamics and prolonged low-flow drought periods characteristic of the Rio Grande. Model performance was evaluated under high-flow, typical, and low-flow conditions using Nash-Sutcliffe Efficiency (NSE), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and correlation coefficient (R²). Results demonstrate successful reproduction of hourly, daily, and monthly hydrographs, capturing seasonal variability and extreme events. These findings establish that a computationally parsimonious framework, rigorously calibrated on high-resolution long-term data, provides a scalable and transferable tool for flood control, drought management, and sustainable water allocation in the Rio Grande watershed and similar semi-arid catchments.