Improving Watershed Modeling via Big Data: Insights from Multiple Cases Studies

Abstract:

The big data evolution has begun to directly address the hydro-geoscience community’s limited-information challenges. Sensor-based data from airborne satellites and in-stream monitoring provide a trove of spatially and temporally dense information that can be mined for solving global flooding and water quality issues. However, these data developments have outpaced hydrological and biogeochemical process-based refinements in watershed models. We therefore ask several overarching questions: How can satellite- and other sensor-based measurements be linked with existing watershed-scale models to improve hydrological and water quality simulations? To what extent can these data address and minimize uncertainty in model outputs and improve internal process representation? We present multiple case studies and lessons learned therein regarding integrating sensor-based information into watershed models to improve hydrological and water quality simulations. First, we detail how integrating satellite measurements including Moderate Resolution Imaging Spectroradiometer (MODIS) Leaf Area Index, National Aeronautics and Space Administration (NASA) Soil Moisture Active Passive (SMAP) root zone wetness conditions, and LANDSAT 7 surface water extent data advances hydrological predictions and the physical realism in models across watersheds in multiple physiographic regions. Second, we highlight how U.S. Geological Survey nitrate sensor data offer high temporal resolution data for model calibration and how the use of data mining across large spatial and water quality data sets complements traditional linear statistical models. Our studies cumulatively suggest that integrating big data provides a needed trajectory toward advancing process-based hydrological and water quality simulations in watershed models.