**Active: 2024 - Present
Summary
What is the story of rooftop solar in New England? When policies encouraging the construction of ground-based solar photovoltaic arrays were rolled back in response to community pushback, did the pace of rooftop solar construction change? Where is rooftop solar concentrated, and what factors influence its construction?
These questions, and more, are part of the story of renewable energy, community reactions, and land use change in New England, building on existing work under Professor John Rogan and Extractives@Clark in collaboration with the Narragansett Bay Estuary Program (NBEP). To answer them accurately requires a multi-year rooftop solar panel dataset that does not yet exist.
This research project aims to create an open-access, high-quality, multi-imagery training dataset for rooftop solar panel detection, as well as creating open-source tools and reproducible workflows to create high-quality label data. The outcomes should go from source data processing, through a full-stack labeling interface, to model training and inference.
Goals
- Registry of rooftop solar by year in New England, usable by stakeholders. Rich metadata, findable, well-documented, online.
- Reusable model and methodology for creating rooftop solar databases from aerial imagery. Published on github, containerized, with a paper.
- Benchmark dataset for comparing machine learning models for rooftop solar detection, published through STAC (Spatio-Temporal Asset Catalog)
- Education and AI Readiness for Geospatial ML practitioners at Clark University
Outcomes
- Created reusable pipeline for converting online orthophotos into tiles for use in labeling.
- Created labeling interface that supports multiple simultaneous users.
- Ran community mapathon to annotate rooftop solar polygons using this tool.
Code Repositories
Repository for processing MA Orthophotos into tiles for serving with the labeling interface. This is necessary due to some projection and misregistration issues with the official tile service.
Label geospatial imagery collaboratively with a lightweight, customizable platform for multiple simultaneous users. This repo is currently set up only for the MA orthoimages created by the datagen repo, but the intent is to make it generalizable for other projects.
Online Stories
GIS Week mapathon expands rooftop solar dataset
Mapathon for Mapping Rooftop Solar Across New England at Clark University.
Unlocking Solar Potential: A New Dataset Takes Center Stage
Receiving the FARR-RCN grant for Mapping Rooftop Solar Across New England.