T212 Machine Learning in Hydrogeology

Thursday, 29 October 2020
1:30 PM - 5:30 PM
Meeting Rooms

USING MACHINE-LEARNING PREDICTIONS OF GROUNDWATER SALINITY TO ASSESS WATER AVAILABILITY IN THE MISSISSIPPI RIVER VALLEY ALLUVIAL AQUIFER

Paper number: T212-4
Program: Technical Sessions
Time: 2:05 PM – 2:20 PM

Authors

Abstract

Long-term water level declines in the Mississippi River Valley alluvial aquifer (MRVA) have raised concerns about the availability of groundwater resources within the Mississippi Embayment (ME) in the south-central United States. Changes in water quality as water levels decline may limit future availability of groundwater resources of the MRVA for irrigation, public supply, and domestic use. To quantify water resources of the MRVA the United States Geological Survey (USGS) Mississippi Alluvial Plain (MAP) regional water availability study is working to improve water-budget estimates of an existing regional groundwater-flow model. Boosted regression trees (BRT) machine-learning (ML) models were trained on available groundwater-quality data to predict groundwater salinity (specific conductance and chloride) across the MRVA. Explanatory variables for the BRT ML models included water level and well construction information, surficial variables (such as geomorphology and soils), among others. These models build on previous ML models of salinity in the ME and extend the modeling predictions to the coast where salinity increases closer to the Gulf of Mexico and datasets are sparse. ML models were able to predict spatial distributions of specific conductance and chloride where discrete samples were not available. Preliminary model predictions of specific conductance were used to correct subsurface resistivity data from a regional Airborne Electromagnetic (AEM) survey in areas where salinity concentrations are elevated. Corrected AEM data were fed back into the ML models as explanatory variables to explore how variations in aquifer sediment texture, as indicated by changes in resistivity, and hypothesized connections to underlying, more saline aquifers affect salinity. Using ML to predict spatial distributions of salinity across the MRVA will help identify sources of water to the MRVA and drivers of groundwater quality. Information from the ML models will be used to evaluate a regional groundwater-flow model used to help answer questions of groundwater availability and allows water-resource managers to make informed decisions regarding water use from the MRVA.

Geological Society of America Abstracts with Programs. Vol. 52, No. 6

doi: 10.1130/abs/2020AM-355904

© Copyright 2020 The Geological Society of America (GSA), all rights reserved. Permission is hereby granted to the author(s) of this abstract to reproduce and distribute it freely, for noncommercial purposes. Permission is hereby granted to any individual scientist to download a single copy of this electronic file and reproduce up to 20 paper copies for noncommercial purposes advancing science and education, including classroom use, providing all reproductions include the complete content shown here, including the author information. All other forms of reproduction and/or transmittal are prohibited without written permission from GSA Copyright Permissions.