{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreicv5ry6ljf3opjgjnahk64nmrpq2kcncymawboa3hoz6lrgyb7jny",
    "uri": "at://did:plc:3smsdppscbbmgh33ttbch2od/app.bsky.feed.post/3mm7vjdgxqf22"
  },
  "coverImage": {
    "$type": "blob",
    "ref": {
      "$link": "bafkreihwoqw5xd3teaquqn6wbbj7p77cepaizom6dge5l56xowlkghaure"
    },
    "mimeType": "image/jpeg",
    "size": 137526
  },
  "path": "/research-spotlights/keeping-humans-in-the-loop-improves-flood-forecasting",
  "publishedAt": "2026-05-19T12:57:47.000Z",
  "site": "https://eos.org",
  "tags": [
    "Research Spotlights",
    "California",
    "Earth science",
    "floods",
    "forecasting",
    "Geophysical Research Letters",
    "Health & Ecosystems",
    "machine learning & AI",
    "Modeling",
    "Nevada",
    "rivers",
    "California Department of Water Resources"
  ],
  "textContent": "Machine learning models cannot yet outperform modeling overseen by human forecasters when it comes to flood prediction, especially when making predictions of extreme events with long lead times, new research suggests. Credit: California Department of Water Resources, Public Domain",
  "title": "Keeping Humans in the Loop Improves Flood Forecasting"
}