Inferring spatiotemporal precipitation-discharge patterns of a snow dominated mountainous karst watershed using a hybrid physically based and deep learning modeling approach
Abstract
Snow dominated mountainous karst watersheds are the primary source of water supply in many areas in the western U.S. and worldwide. In these watersheds, precipitation, snowmelt, recharge, and discharge vary spatially and are controlled by complicated meteorological, topographic, and geologic heterogeneity. To represent the precipitation-discharge processes, we use a hybrid modeling approach that integrates a physically based snow model with a deep learning model based on the Convolutional Long Short-Term Memory (ConvLSTM) architecture. The hybrid modeling approach is tested on the Logan River Watershed in northern Utah with seasonal snow cover and variably karstified carbonate bedrock. We use the high-resolution snow model to simulate spatiotemporally varying snowmelt throughout the study area, and the ConvLSTM model simulates streamflow response to snowmelt and rainfall. The ConvLSTM model is trained and tested in a series of nested subwatersheds using observed streamflow along the Logan River. The spatiotemporal recharge-discharge patterns learned by the ConvLSTM model were then examined and compared with known hydrogeologic information from tracer studies. In addition to accurate streamflow simulation, the ConvLSTM provides insight into hydrologic connectivities at varying scales. Results suggest the hybrid modeling approach is a viable solution to spatially distributed hydrologic modeling and may be potentially transferable to other watersheds without site specific hydrogeologic knowledge.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.H31B..04X