You are here:
Statewide predictions of nutrient concentrations in rivers and streams (scripts)
Citation:
McManus, M. AND T. Linscome-Hatfield. Statewide predictions of nutrient concentrations in rivers and streams (scripts). U.S. Environmental Protection Agency, Washington, DC, 2023.
Impact/Purpose:
We propose to help the states and tribes gain more information from their monitoring data by combining data from multiple water quality monitoring programs so predictions of nutrient concentrations can be made at sites that are not sampled. We combined several different water quality monitoring datasets routinely collected by a state environmental agency for biannual Integrated Reports under the Clean Water Act. As part of our initial efforts with the Indiana Department of Environmental Management, we examined the spatial variation in total phosphorus concentration in the Patoka River watershed. We related that variation in total phosphorus to straight line distances among monitoring sites in the watershed as well as to stream network distances. Based on straight line distances, total phosphorus concentrations were similar among sites up to 10 kilometers apart. However, total phosphorus concentrations among pairs of sites based on upstream-to-downstream distances, indicated that the range of spatially similarity was up to 100 stream kilometers. For flow-unconnected sites, which are on different branches of the stream network, their similarity in total phosphorus concentrations often exceeded what was observed over straight line and flow-connected distances. Accounting for this spatial variation is needed to predict TP concentration on the river and stream network, and such predictions can help environmental agencies prioritize restoration and conservation efforts.
Description:
Travis Linscome-Hatfield, a statistician at Neptune, has created the 4 R scripts that can be run to produce 4 htmls for a spatial stream network (SSN) analysis. The scripts are meant to be "plug and play" in that an SSN object can be loaded for a specific Indiana basin and run through each of the 4 scripts. The 4 scripts cover: 1) Exploratory Data Analysis, 2) Multiple Linear Regression (MLR) Model Selection, 3) Spatial Stream Network Model Selection, and 4) Spatial Stream Network Model Prediction. Ecologists at Indiana Department of Environmental Management (IDEM) are the end users of the scripts as they will run them to make predictions of total phosphorus (TP) concentration in rivers and stream from 10 Indiana basins based on monitoring data they have collected. The covariates, or predictor variables, used in the modeling are from EPA’s StreamCat dataset, which contains land cover and landscape metrics calculated over the extents of catchments and watersheds. Model predictions are made to points on a basin’s stream network with the prediction points being at the midpoint of each 1-km stream segment. EPA and Neptune have been working with IDEM staff on the scripts since September 2021. The scripts enable SSN analysis to be done and are not meant to be a manual or textbook on spatial data analysis. EPA has given journal articles and reference materials to IDEM staff to provide more context to the analysis.