You are here:
WATER QUALITY AND STREAMFLOW ESTIMATION AT UNGAUGED WATERSHEDS USING MACHINE LEARNING.
Citation:
Mallya, G., A. Gupta, M. Hantush, AND R. Govindaraju. WATER QUALITY AND STREAMFLOW ESTIMATION AT UNGAUGED WATERSHEDS USING MACHINE LEARNING. World Environmental and Water Resources Congress 2019, Pittsburgh, PA, May 19 - 23, 2019.
Impact/Purpose:
In this study, we propose to use machine learning regression models, such as gradient boosting machines and random forest regression to predict daily streamflow and water quality loading at ungauged HUC-10 (approximately '105,000 acres') basins using watershed attributes, long-term climate data, soil data, and land use and land cover data as predictor variables. We will test the models on the Upper Mississippi River Basin and Ohio River Basin for water quality constituents, such as suspended sediment concentration, nitrogen, and phosphorus. The proposed method can be used by decision makers and water quality monitoring agencies to develop quick screening tools for identifying critical source areas using methods, such as risk-based watershed health assessments or TMDL analyses.
Description:
As part of the Clean Water Act, state, local and federal agencies collect water quality samples from water bodies, such as lakes, streams, and rivers to monitor their conditions. Though these datasets are available at several locations across the United States, they are usually sparse in both time and space. Water resources planners in several states use these water quality samples for Total Maximum Daily Load (TMDL) analyses by reconstructing water quality loading time series using surrogate variables, such as streamflow values available at nearby locations. However, to estimate water quality loading at ungauged locations, they have to rely mostly on physically-based hydrologic models that are often time consuming to set up. In this study, we propose to use machine learning regression models, such as gradient boosting machines and random forest regression to predict daily streamflow and water quality loading at ungauged HUC-10 (approximately '105,000 acres') basins using watershed attributes, long-term climate data, soil data and land use and land cover data as predictor variables. We will test the models on the Upper Mississippi River Basin and Ohio River Basin for water quality constituents, such as suspended sediment concentration, nitrogen, and phosphorus. The proposed method can be used by decision makers and water quality monitoring agencies to develop quick screening tools for identifying critical source areas methods, such as risk-based watershed health assessment or TMDL analyses. EPA Disclaimer: The findings and conclusions in this abstract have not been formally disseminated by the U.S. EPA and should not be construed to represent any agency determination or policy.