Improved Accuracy of Watershed-Scale General Circulation Model Runoff Using Deep Neural Networks

Abstract

Projecting impacts of climate change on water resources is a vital research task, and general circulation models (GCMs) are important tools for this work. However, the spatial resolution of downscaled GCMs makes them difficult to apply to non-grid conforming scales relevant to water resources management: individual watersheds. Machine learning techniques like deep neural networks (DNNs) may address this issue. Here we use a DNN to predict monthly watershed-scale runoff (i.e., stream discharge divided by watershed area) from monthly gridded and downscaled Coupled Model Intercomparison Project Phase 5 (CMIP5) GCM hydroclimatic fluxes (i.e., precipitation, evapotranspiration, and temperature). We used hydroclimatic fluxes, biotic, and abiotic characteristics from 2,731 watersheds in the conterminous United States to train and test a DNN that can predict watershed-scale runoff. The DNN described 93% (Pearson’s correlation coefficient = 0.962) of the variability in observed runoff and was temporally and spatially robust. The median absolute error (MAE) of DNN predictions was approximately 25 percentage points lower than that of gridded, downscaled GCM runoff or monthly normal runoff (i.e., 30-year average of runoff observations at the watershed-outlet). DNN monthly runoff predictions had the lowest MAE of all the grid-to-watershed-scale conversion approaches we tested, including: linear ridge regression, support vector machines, extreme gradient boosting, and artificial neural networks. We demonstrated why using DNNs to convert gridded GCM hydroclimatic fluxes to watershed-scales is relevant to water resources research and management. We also provided a methods guide for hydrologists interested in implementing machine learning techniques.

Using an optimized soil and water assessment tool by deep belief networks to evaluate the impact of land use and climate change on water resources

“…Because of its ability to simulate extremely complex and unknown processes, it has been widely used in water sciences. 17 Various researches have been conducted on the effect of land use and climate changes on water sources and the hydrological status of the watershed. For example, Marin et al 18 evaluate climate resource vulnerability due to climate change using the SWAT model, and based on the results of the analyses, a downward trend was observed for discharge at the end of the 21st century (up to 54%) and also a downward trend was observed for the surface flow up to 41%.…”

“…Neural networks are powerful tools that are made with the simple mimicry of the human biological nervous system. Because of its ability to simulate extremely complex and unknown processes, it has been widely used in water sciences 17 . Various researches have been conducted on the effect of land use and climate changes on water sources and the hydrological status of the watershed.…”

A Systematic Review of Deep Learning Applications in Streamflow Data Augmentation and Forecasting

“…Their results suggested that different architectures could be a better fit for different watersheds. Similarly, Ghose (2019) compared ANNs and RNNs for the Dhankauda watershed of Sambalpur, Odisha, India and showed that RNNs outperformed ANNs for their data in monthly streamflow forecasts.In 2020,Rice et al (2020) trained a deep ANN with precipitation, evapotranspiration, and temperature for 2,731 watersheds in the conterminous United States. They compared their method to some machine learning algorithms such as SVMs, linear regression, and extreme gradient boosting.…”