Planning and Implementing Water Sustainability Projects
Across the Western United States, various agencies and water providers are planning and implementing water resource sustainability projects to respond to climate change and growing supply needs. In California, Groundwater Sustainability Agencies (GSAs) must reach and maintain groundwater sustainability by 2040 in accordance with the Sustainable Groundwater Management Act (SGMA). To meet this challenge, GSAs are developing projects such as water reuse, managed aquifer recharge, interagency water transfers, and aquifer storage and recovery (ASR) to effectively store and use groundwater.
Effective planning of groundwater sustainability projects that account for climate change typically requires using a groundwater flow model to simulate impacts of project configurations decades into the future. These groundwater model simulations can take hours or days to run, and adequately optimizing projects requires many iterative simulations to identify improved project configurations. To expedite optimization of sustainability projects using a groundwater model, Montgomery & Associates (M&A) developed a novel workflow using machine learning (ML) algorithms to autonomously plan, preprocess, post process, and evaluate physical model simulations. We call this workflow approach Machine Learning Guided Optimization (MLGO).
Machine Learning Guided Optimization
MLGO consists of a coupled, automated process where ML algorithms learn from inputs and outputs of a physical groundwater flow model to optimize project design and estimate project combinations that meet sustainability goals. The ML workflow consists of automatically identifying new project setups and combinations based on prior physical model results, taking into account optimization goals specified by the user. These user-defined optimization goals can reflect desired sustainability criteria, feasibility objectives, or supply metrics. This workflow combines the efficient computing power of ML and the real-world physics contained in physical groundwater flow models.
The illustration below describes how MLGO combines machine learning algorithms with a physical groundwater flow model to optimize systems.
Santa Cruz Mid-County Regional Water Optimization Study
M&A used MLGO to support the Santa Cruz Mid-County Regional Water Optimization Study (Study). The Study’s objective is to support the selection of water supply projects and management actions within the critically overdrafted Santa Cruz Mid-County Groundwater Basin (Basin), for long-term operations and shared regional benefits, including sustainable groundwater management and regional water supply needs. Water supply projects and management actions considered include ASR, Pure Water Soquel purified recycled water recharge (PWS), and interagency water transfers. Each project contains numerous implementation parameters, and the number of potential project configurations is practically endless.
For this work, MLGO was trained on the inputs and outputs of the existing calibrated physical groundwater flow model and tasked with identifying simulated project configurations that improve regional water supply while maintaining feasibility and sustainability. Under oversight of the groundwater modeler and constrained by user-defined realistic implementation parameters, MLGO learned from each iterative simulation and progressed toward project optimization over time. This process culminated with the identification of 4 robust water supply management alternatives representing a range of potential infrastructure investments and associated regional water supply improvements. Using MLGO helped deliver an improved final product, because the process was able to autonomously advance toward optimization and simulate many more project configurations than would have been feasible with manual optimization techniques.
The following schematic illustrates the general workflow of MLGO, which progresses autonomously until optimization is reached.
Other Applications
In addition to assisting GSAs with water supply optimization and project planning, M&A professionals can use MLGO for a range of groundwater management applications in water resources, mining, and environmental settings. We offer one of the largest, most experienced groundwater modeling team in the western United States, with over 20 professionals who are proficient in a variety of numerical groundwater flow and transport codes. Our predictive models can incorporate future uncertainties such as climate change or estimated future urban and agricultural water demand through scenarios or probabilistic simulations. We often use 3D geologic models to develop inputs and present results in a way that is accessible to project stakeholders. Our SGMA experts are adept at communicating complex modeling results to GSAs and interested parties to support long-term decision making for sustainable groundwater management.
Sobre el Autor
Patrick Wickham, PG, is a hydrogeologist in M&A’s Pasadena office who specializes in groundwater modeling and machine learning applications. Patrick will participate in the HydroML 2025 Symposium, hosted by University of California–Irvine May 27-29.
