DeepBase: A Deep Learning-based Daily Baseflow Dataset across the United States

High quality baseflow data is important for advancing water resources modeling and management, as it captures the critical role of groundwater and delayed sources in contributing to streamflow. Baseflow is the main recharge source of runoff during the dry period, particularly in understanding the interaction between surface water and groundwater systems. This study focuses on estimating baseflow using deep learning algorithms that enhance the estimation capabilities in both gauged and ungauged basins. Recognizing the shortage in accessible high quality daily baseflow data, our objective is to generate a daily baseflow dataset across the contiguous United States (CONUS) for 1661 basins from 1981 to 2022. This dataset provides valuable information for earth and environmental scientists, and water resource managers, enhancing our understanding of the water cycle. It also provides an important foundation for enhancing the study of baseflow contributions to extreme events such as droughts and floods. The dataset can be used as a new benchmark for future studies aimed at improving hydrological predictions and managing water resources more effectively.