Automated Scenario Assessment of Groundwater Table & Salinity Response to Sea-level Rise

Dipankar Dwivedi | Lawrence Berkeley National Laboratory



More than 1700 Department of Defense (DoD) sites in coastal areas worldwide are expected to be impacted by rising sea levels. Decisions regarding the management of these high-value military installations near and along the coast depend on the accuracy with which changes to groundwater levels and salinity can be predicted under realistic future conditions. A major challenge for decision-makers is that high-fidelity mechanistic models that relate the rising sea level to groundwater level rises and salinity intrusion are inevitably site specific, expensive to build, and difficult to use. To address these concerns, the objective of this project is to develop a flexible, transparent, efficient, and user-friendly toolset with multiple reduced-order models (ROMs) that can be employed by non-expert personnel to quickly estimate changes in groundwater levels and salinity intrusion using simplified representations of site-specific data. The novel scientific contribution of the project is the evaluation of a range of strategies to create the ROMs, including proper orthogonal decomposition, machine learning approaches, and high-fidelity mechanistic models, to identify which type of ROM best approximates the groundwater levels and water quality (salinity) at a variety of coastal sites. The resulting end product will be a risk management summary for site managers that carefully evaluates multiple ROM choices, thereby capturing the range of responses for groundwater levels and saltwater encroachment expected under a changing climate.

Back to Top

Technical Approach

Given that there are more than 1700 DoD coastal sites, the overarching hypothesis is that these sites can be categorized into a select few classes based on similar characteristics, which can then support ROM development across each class through a representative site analysis. To test the hypothesis and demonstrate the approach, the technical plan consists of six tasks. In Task 1, data will be compiled from a large number of coastal DoD sites to develop ROMs. Then the project team will classify this large number of sites into a few classes based on broad site characteristics (e.g., sedimentary, estuary, marsh, head- or flux-controlled). Task 2 focuses on developing and testing a variety of ROMs, including physics-based ROMs and surrogates. It is hypothesized that ROM choices can be developed for a subset of sites that are representative of individual classes and then broadly applied to the entirety of sites within that class. To enable the development of physics-based ROMs, Task 3 focuses on high-fidelity simulations at representative sites identified in Task 1. Task 4 evaluates the performance of ROM choices under a range of subsurface scenarios relevant to sites. Task 5 involves the development of a transferable and efficient software tool by involving relevant stakeholders. Finally, Task 6 provides technology transfer support to facilitate the use of the software by DoD personnel.

Back to Top


This project aims to provide DoD with the tools needed to arrive at better, more transparent, and defensible coastal site management strategies. Current simulation models are site specific, data intensive, and computationally expensive, rendering them unsuitable for broad applications. In contrast, ROMs present accessible and parsimonious solutions. Having an ensemble of ROM choices further ensures success in addressing coastal ecosystems’ complexity, especially when dealing with coupled flow and salinity transport simulations.

Back to Top

Points of Contact

Principal Investigator

Dr. Dipankar Dwivedi

Lawrence Berkeley National Laboratory

Phone: 510-833-0504

Program Manager

Resource Conservation and Resiliency