Remotely mapping stormwater facility footprints and storage volumes
The Issue: Stormwater managers often have incomplete information on the location and storage volume of stormwater facilities within their jurisdiction.
How USGS will help
USGS is working to develop an artificial intelligence/machine learning (AI/ML) deep learning model to remotely map the footprint of stormwater facilities and estimate a storage volume for each facility using high-resolution geospatial data layers.
Problem
There is currently no regional, comprehensive database of stormwater best management practices that includes stormwater storage estimates for the Chesapeake Bay watershed. Existing databases may only include stormwater facilities on public land, use inconsistent naming of facilities, and provide limited information on the size and storage capacity of each facility.
Objective
This project will use an AI/ML deep learning model to remotely map surface stormwater facilities within 30 small suburban and urban watersheds located in the Piedmont ecoregion with the Chesapeake Bay watershed. Each facility will have its footprint mapped and have an estimated stormwater storage volume based on high-resolution topographic data. Stormwater facilities to be mapped include swales, dry detention ponds, wet detention ponds, bioretention, and sand filters.
Project Approach
Existing high-resolution lidar-derived elevation data and imagery will be used to develop a set of predictor raster datasets as model inputs. A training dataset will be generated from existing municipal datasets and hand-digitized facility footprints. Predictor rasters and the training datasets will be input into an AI/ML deep learning model to map facility footprints. USGS will then develop an open-source workflow to estimate facility stormwater storage volume based on mapped footprints and elevation data. Results from the study will be published as a journal article and in a USGS data release.
Understanding the Effects of Stormwater Management Practices on Water Quality and Flow
Fairfax County Water Resources Monitoring Network
The Issue: Stormwater managers often have incomplete information on the location and storage volume of stormwater facilities within their jurisdiction.
How USGS will help
USGS is working to develop an artificial intelligence/machine learning (AI/ML) deep learning model to remotely map the footprint of stormwater facilities and estimate a storage volume for each facility using high-resolution geospatial data layers.
Problem
There is currently no regional, comprehensive database of stormwater best management practices that includes stormwater storage estimates for the Chesapeake Bay watershed. Existing databases may only include stormwater facilities on public land, use inconsistent naming of facilities, and provide limited information on the size and storage capacity of each facility.
Objective
This project will use an AI/ML deep learning model to remotely map surface stormwater facilities within 30 small suburban and urban watersheds located in the Piedmont ecoregion with the Chesapeake Bay watershed. Each facility will have its footprint mapped and have an estimated stormwater storage volume based on high-resolution topographic data. Stormwater facilities to be mapped include swales, dry detention ponds, wet detention ponds, bioretention, and sand filters.
Project Approach
Existing high-resolution lidar-derived elevation data and imagery will be used to develop a set of predictor raster datasets as model inputs. A training dataset will be generated from existing municipal datasets and hand-digitized facility footprints. Predictor rasters and the training datasets will be input into an AI/ML deep learning model to map facility footprints. USGS will then develop an open-source workflow to estimate facility stormwater storage volume based on mapped footprints and elevation data. Results from the study will be published as a journal article and in a USGS data release.
Understanding the Effects of Stormwater Management Practices on Water Quality and Flow