You are here:
ROAR Project: Statewide Predictions of Total Phosphorus Concentrations in Indiana Rivers and Streams
Citation:
McManus, M., S. Sobat, K. Gaston, M. Cubbage, AND T. Linscome-Hatfield. ROAR Project: Statewide Predictions of Total Phosphorus Concentrations in Indiana Rivers and Streams. Presentation/Indiana Dept. of Environmental Management Office of Water Quality Watershed Assessment and Planning Branch, Cincinnati (virtual), OH, November 17, 2023.
Impact/Purpose:
We propose to help the states and tribes gain more information from their monitoring data by giving them a tool that will combine data from multiple water quality monitoring programs allowing for predictions at sites that are not sampled. We plan to do this by combining several different water quality monitoring datasets routinely collected by a state environmental agency for biannual Integrated Reports under the Clean Water Act. These datasets include: 1) network of fixed sites typically focused on larger rivers and streams to report trends at the sites, 2) randomly selected sites for reporting of water quality status for the state or sub-regions of the state, and 3) targeted sites used to monitor smaller watersheds, specific sources of pollutants, or are parts of total maximum daily load or restoration efforts. This combination can achieve the sample size, spatial configuration of headwaters to watershed outlets, and spatial density of sites to predict total phosphorus concentrations at unsampled reaches using spatial stream network modeling, which explicitly considers the branched and directional nature of stream networks. By producing a map of predicted concentrations, an agency could identify hotspots of high concentrations, verify those predictions, and prioritize what remediation is needed. Our research approach has two phases: 1) build the geospatial dataset of the monitoring sites, river and stream flowlines, and catchments with landscape metrics; and 2) use spatial stream network analysis to predict total phosphorus concentrations. A unique aspect of this project is that future monitoring readily provides an independent dataset to test predictions. In addition to the predictions maps, other products include the data, annotated code, and results in an accessible, transparent and reproducible format based on open science practices. By following those practices, we can train EPA region, state agency, and tribal scientists on making spatial predictions of water quality. This project is important as it enables states to more efficiently target resources to probable problem areas and to potentially threatened resources.
Description:
Our presentation to EPA's Region 5 Water Quality Monitoring Managers Meeting describes how we made statewide predictions of total phosphorus concentrations in Indiana rivers and streams. We outline the three datasets needed to make such predictions. The first being the Indiana Department of Environmental Managements monitoring data based on their fixed, probabilistic, and target stations. Second is the National Stream Internet (NSI) produced by the U.S. Forest Service. Third is the EPA's Stream Catchment, or StreamCat, dataset. Bringing those three datasets together provides the water quality monitoring stations snapped to the NSI flowlines of Indiana rivers and streams, and predictor variables from StreamCat so that a spatial stream network analysis (SSN) can be done to make predictions. That analysis was done using the open-source R software and 4 scripts that covered: exploratory data analysis, non-spatial multiple linear regression modeling, spatial stream network modeling, and spatial stream network predictions. We compare models from the regression and spatial stream network analyses and note that the SSN model is a better fit to the data and outperforms the regression model in prediction. Synched maps provide a way to compare the observed total phosphorus concentrations to the predictions, and those predictions are summarized by HUC12s using linked micromaps. We note some of the challenges to doing a SSN analysis, but point out advantages of having such predictions.
