{
  "question": "How do resident patterns, workplaces, roads and daylight help interpret King County crash findings without substituting resident population for travel exposure?",
  "joins": [
    {
      "inputs": [
        "frozen WSDOT events",
        "full 2024 TIGER tracts"
      ],
      "key": "unique point-intersects-polygon",
      "cardinality": "many-to-one; ambiguous/unmatched retained",
      "output": "tract-crash-context-joined.csv",
      "check": "57785 events and 1654 severe events retained"
    },
    {
      "inputs": [
        "ACS 2024 estimates/MOEs/annotations",
        "2024 tract/place GEOIDs"
      ],
      "key": "text GEOID, leading zeros retained",
      "cardinality": "one context row per geography",
      "output": "acs-estimate-moe-annotations.csv",
      "check": "495 King tract keys; sentinel normalization; source-specific universes"
    },
    {
      "inputs": [
        "LODES 2023 WAC/RAC/OD JT01",
        "official LODES 8.4 crosswalk"
      ],
      "key": "2020 block code -> 2024 trct/stplc parent",
      "cardinality": "blocks aggregate before context joins",
      "output": "city-federal-context.csv; tract-context-features.csv; primary-job-flows-2023.csv",
      "check": "official uncompressed CSV checksums; OD destinations reconcile to WAC; current parents from crosswalk, not block prefixes"
    },
    {
      "inputs": [
        "2020-2024 highway archives",
        "2024 city reference"
      ],
      "key": "section intersects city; exact route/qualifier/mile key across years",
      "cardinality": "separate road-section evidence",
      "output": "state-highway-matched-section-pairs.csv",
      "check": "ReportingYear equals archive year; mainline only, unique keys and Hausdorff <=5m; AADT never summed into travelers"
    },
    {
      "inputs": [
        "event date/local clock/coordinate",
        "NOAA approximate solar equation"
      ],
      "key": "event, with explicit timezone and DST audit",
      "cardinality": "one derived condition per event",
      "output": "city-month-hour-solar-context.csv",
      "check": "unknown clocks retained; -0.833/-6 degree thresholds; sensitivity +/-0.5 degrees"
    },
    {
      "inputs": [
        "frozen Seattle outcome",
        "SDOT conditions"
      ],
      "key": "unique normalized report number + exact date",
      "cardinality": "one-to-one admitted match",
      "output": "seattle-physical-environment.csv; seattle-observed-light-versus-solar.csv",
      "check": "frozen severe outcome unchanged; unmatched, observed light and modeled phase kept distinct"
    }
  ],
  "analysis": "Descriptive area context, cohort decomposition, repeated-site traffic changes, time series and maps. No causal effect, individual inference or full-network travel-risk rate.",
  "open_measurements": [
    "Common all-road VMT or time-specific road-user volumes",
    "Comparable observed lighting/weather for Bellevue and Tukwila",
    "Historical actual on-duty officer-hours",
    "Intersection turning movements, pedestrian/cyclist counts and site-specific street design for follow-up"
  ]
}