We believe you should be able to see the provenance of every score we produce. Not as a legal disclaimer buried in a footer, but as part of what the score means. A score built from modeled estimates tells you something different than one built from measured readings at a nearby sensor. Both are useful. They're not the same thing.
We label which is which. Always.
measured sensor · modeled estimate — coverage thins in rural areas, and we show the gap rather than inflate confidence. (Sensor points shown are illustrative placeholders pending live ingestion.)
Explore the data catalog → — every open dataset on its own page, with a map of just that layer (FEMA NRI, Census ACS, CDC PLACES, OpenStreetMap, USGS PAD-US, EPA SDWIS), derived from real Ashland data.
Most environmental data that exists at a national scale is modeled — computed from mathematical models that take inputs like traffic counts, road geometry, and atmospheric conditions and produce estimates of what noise or air quality would be at a given point. These models are built by serious scientists, validated against real observations, and used by regulatory agencies. They are also estimates. They can be wrong at the local level in ways that show up when you actually stand at the address.
Measured data — readings from physical sensors — is different. A PurpleAir sensor on the corner of a specific street, calibrated against a regulatory reference monitor, tells you something modeled data can't: what the air was actually like at that location, at that time.
We use both. Where we have measured data nearby, we use it and show it. Where we don't, we use the best available model and tell you that's what it is.
We also tell you the sensor density. In rural areas and small towns — exactly the places many people using still are looking for — sensor coverage can be thin. We don't inflate confidence where coverage is sparse. We show you the gap.
What we use: The US DOT National Transportation Noise Map (aviation, highway, rail; modeled dB(A) Lden at ~30-meter resolution). FAA Day-Night Average Sound Level (DNL) contours for airports. State DOT Annual Average Daily Traffic (AADT) counts.
What this means: Modeled estimates of average noise levels. They capture the shape of the noise landscape well — the quiet valley versus the noisy arterial — but flatten time. A street calm at noon and loud at 7am gets one number. We'll layer in time-of-day acoustic data as it becomes available.
The limit: The DOT noise map is designed for tracking trends, not evaluating individual addresses. We use it at the neighborhood and block level.
What we use: EPA AirNow (the regulatory reference network, real-time AQI). PurpleAir (low-cost PM2.5 sensors, hyperlocal, requires calibration). OpenAQ (the open aggregator that unifies both).
What this means: AirNow is authoritative and sparse. PurpleAir is dense and requires the EPA correction equation to be accurate. We apply that correction. Where both sources exist nearby, we show both.
The limit: PurpleAir coverage is concentrated in populated areas. Rural areas may have no nearby sensors. We show coverage density on the map above.
What we use: VIIRS satellite night-light composites, processed into Bortle class and sky brightness (mag/arcsec²). Published annually by NOAA.
What this means: Satellite-derived. It reflects sky brightness as seen from above the atmosphere — a good proxy for what you'd experience on the ground, with some limitations near large bright sources.
The limit: VIIRS resolution is ~500 meters. The number is for the general area, not your specific backyard.
What we use: FEMA National Risk Index (18 natural hazards, county and tract). FEMA National Flood Hazard Layer (parcel-relevant flood zones). NOAA climate data.
What this means: The FEMA NRI is a relative risk index — it shows where risk is higher or lower, not a precise probability. The flood maps reflect FEMA's current mapping, which is incomplete and in some areas outdated.
The limit: The FEMA NRI web application was retired in 2025; we use the downloadable data, which is current. Forward-looking climate projections are not yet in this version.
What we use: EPA Safe Drinking Water Information System (SDWIS) for public water-system violations. EPA UCMR 5 data for PFAS occurrence. USGS Water Quality Portal for surface and groundwater.
What this means: This covers public water systems. Private wells — common in rural areas — are not in the federal database. If you're considering a property on a well, we flag that and recommend local testing.
What we use: USGS Protected Areas Database (PAD-US) for public lands. NDVI for tree canopy and vegetation density. OpenStreetMap for parks, trails, and nature access points.
What this means: PAD-US is authoritative for protected public land. NDVI is a satellite-derived proxy for green cover. OSM data is contributed by a global community and is generally reliable in populated areas.
What we use: Census / ACS for demographics, residential tenure, and density. County Health Rankings for social associations and health factors. USDA Food Access Research Atlas for grocery access. USDA National Farmers Market Directory. OpenStreetMap for points of interest.
What this means: Census data is the most reliable source for these dimensions but is collected every 5–10 years. The picture is real but not real-time.
We don't scrape Zillow or Redfin. Their data is not ours to take. We use listing APIs from licensed data providers — or listings you share with us — and we tell you which is which.
We don't use your personalization data to build a profile of you. Your weights are yours. See the privacy page.
We don't present modeled data as measured, or estimated data as precise. If we don't know something, we say so.