Enhancing Hydrological Simulation Efficiency by Improving Initial Soil Moisture Conditions through Reanalysis Data

Abstract

In the domain of hydrological modeling, accurately determining initial conditions such as soil moisture content is crucial for enhancing simulation efficiency and applying these models effectively in water resource management, flood prediction, and drought forecasting. Traditional methods often rely on a data-intensive warm-up phase to establish these conditions, which diverts valuable data from calibration and validation. Addressing this challenge, our study introduces an innovative methodology that utilizes an alternative global soil moisture dataset to optimize these initial conditions without the conventional warm-up phase, thereby aiming to improve both the accuracy and efficiency of hydrological simulations. We focused on the Block-wise use of the TOPMODEL (BTOP) and ERA5-Land reanalysis data, specifically analyzing three soil moisture variables within the Fuji and Shinano River Basin, Japan. Through a comprehensive correlation analysis, we examined the dynamics between these variables and employed a range of curve-fitting functions alongside advanced techniques, particularly Long Short-Term Memory (LSTM) networks, to establish a robust relationship between BTOP and ERA5-Land soil moisture variables. The LSTM, known for their effectiveness in handling complex time series data, were instrumental in capturing the intricate spatial and temporal correlations between the variables. To validate the efficacy of our proposed methodology, we conducted four hydrological simulation scenarios, meticulously designed to assess the benefits of incorporating ERA5-Land soil moisture data into the model's initial conditions. The results were compelling: simulations using the enhanced initial conditions significantly outperformed those without the warm-up phase and closely approximated the 'optimal' scenario typically reliant on extensive warm-up data. This study not only underscores the potential of using reanalysis soil moisture data to refine initial conditions, thereby revolutionizing water resource management and forecasting practices, but also presents a scalable solution that can be adapted to various hydrological models and scenarios. Consequently, our research contributes significantly to the ongoing discourse on improving environmental modeling and management practices, advocating for more precise, resource-efficient, and adaptable methodologies in hydrological modeling.