Machine Learning and other Enabling Technologies to Advance Watershed Science and Co-Design Strategies

Abstract:

While watersheds are recognized as the Earth’s key functional unit for assessing and managing water resources, developing a predictive understanding of how watersheds respond to perturbations is challenging due to the complex nature of watersheds. However, several emerging technologies are starting to reveal their promise for greatly enhancing the predictive understanding of watershed hydrobiogeochemical behavior, including machine learning (ML) and artificial intelligence, exascale computing, 5G, and cloud based tools. We envision a future where these emerging technologies are integrated into ‘co-design’ strategies, where watershed sensing, data and modeling systems are ‘born’ to communicate with each other across multiple scales. Such co-design strategies hold significant potential to accelerate optimized, near real-time natural resource management and to address complex watershed questions.

Toward developing a co-design strategy to improve prediction of watershed hydrobiogeochemical response to perturbations, we illustrate the value of ML approaches at the mountainous East River, CO headwaters catchment of the Upper Colorado River Basin, which is the site of the Watershed Function SFA project. Examples include the use of ML to estimate the distribution of plant communities and microtopographic controls over plant ecosystem niches; ML approaches for delineating watershed zones that have unique distributions of above-and-below ground properties relative to their neighbors; wireless sensor networks that can autonomously ‘watch’ fluids move between watershed compartments; ML algorithms to extract, transform, and integrate data with physics-based models; ML-assisted adaptive meshes for watershed reactive transport models to simulate the influence of fine scale processes on larger scale behavior; hybrid models that can use time-lapse data to predict watershed properties. More information about the project is provided at watershed.lbl.gov.