How we did the match
For each river we fetched the upstream drainage polygon from the USGS Network Linked Data Index (NLDI). Projects with geographic coordinates are tested against the polygon using a ray-casting point-in-polygon algorithm. Those matches are tagged ✓ basin and are verified to be inside the river's actual drainage.
For projects without coordinates — or for rivers whose basin geometry is unavailable from NLDI (notably the Chilkat in Alaska, which sits outside the CONUS NHD dataset) — we fall back to county-inclusion matching using a hand-curated list of watershed counties. Those are tagged ~ county and should be treated as approximate. Two basins (San Joaquin and Dan) came back smaller than expected from NLDI — the query point was downstream of the relevant confluence — so projects in those watersheds may be matching via the county fallback. We will refine those specifically on the next iteration.
If you see a project missing from a watershed — or listed under the wrong river — email george@poweredbywho.com with the project and the correction.
Full methodology on each river, plus the specific threats and action items, is at mostendangeredrivers.org. American Rivers is the primary source. Support them directly if this report is useful — they do the watershed-level science we couldn't replicate.