Earth Observation Data

Land cover in sixteen classes on a 30 m grid, classified from satellite imagery, one class to a cell and one map for every year since 1985. For each parcel the release carries the share of its area in each class and the class covering the most of it; the sixteen shares add to one. The classes and their definitions are below. The publisher states that mapped features may not be in an exact geographic location and that the data may be corrected or replaced without notification.

An annual crop-specific land cover layer, produced from satellite imagery and extensive agricultural ground reference data collected through the growing season. Resolution is 10 m from the 2024 crop year and 30 m for 2008 to 2023. The release carries the share of the parcel in cultivated cropland and the single crop class covering the most of it. The publisher does not warrant any conclusions drawn from the data.

Two readings. Forest cover is the share of the parcel in the three forest classes of the land cover grid: areas dominated by trees generally greater than 5 metres tall covering more than 20 percent of the total vegetation. Canopy height is the average height of the top of the canopy for a stand, read from airborne lidar returns normalised to the ground surface where lidar has been flown, and from a 30 m modelled vegetation-height grid elsewhere; the release names which answered. Height is read only inside the mapped forest area.

Vegetation disturbance, mapped where an indicated decrease in vegetation cover is seen within a 30 m cell. Only confirmed disturbances are reported, with the date of initial disturbance: a loss is confirmed from subsequent cloud-free observations that show the detection persists, so how quickly a loss appears depends on how often the ground is seen clearly. The layers record loss of vegetation cover; no cause is attributed.

Three readings, each from a separate model and never blended. Tree carbon — live above ground, standing dead, and down dead wood more than 3 inches in diameter — is in tons per acre, from a 30 m model that imputes a measured forest inventory plot to each forested cell. Aboveground biomass, also tons per acre, is from a separate 30 m model built by matching satellite imagery from 2014 to 2018 to about 213,000 field plots measured in the same period. Soil organic carbon is the stock estimated for the top metre of the soil map unit. The publishers state that these are modelled estimates and a snapshot that changes over time.

Soil properties and interpretations for the soil map unit under the parcel, read from the national gridded soil database: detailed field-validated soil survey mapping, filled with a general soil map where detailed mapping does not exist. Ratings describe the map unit at survey scale, not a sample taken on the parcel, and the publisher states that onsite investigation may be needed to validate the interpretations and to confirm the identity of the soil on a given site.

Nine readings about ground that may move. Slide-prone terrain comes from a 90 m national model of slope and 100 m relief trained on 613,724 mapped landslides, which captures 98.9 percent of observed landslides within terrain covering 44.6 percent of the area. Mapped landslides, sinkholes, karst geology and underground mine workings are compiled from inventories kept jurisdiction by jurisdiction, which the publisher states vary greatly in quality, accessibility and extent. The set also carries measured vertical ground movement, the nearest Quaternary fault with its age and slip-rate class, the seismic site class from the time-averaged shear-wave velocity in the upper 30 m of the ground, and a modelled probability of permafrost within 1 m of the surface, whose overall map accuracy the publisher puts at about 85 percent.

The share of the parcel inside mapped wetland and deepwater habitat polygons, interpreted from aerial photography at 1:24,000; open water is counted separately. The mapping records the extent, approximate location and type of wetlands under a biological definition of wetland. The publisher states that regulatory agencies with jurisdiction over wetlands may define and describe wetlands differently, and that no attempt is made to define the limits of proprietary jurisdiction of any government or the geographical scope of any regulatory programme.

The distance from the parcel boundary to the nearest mapped lake, pond or wide river in the national hydrography layer. Features are traced from lidar elevation data where an area has been remapped; where it has not, the layer carries the older 1:24,000 geometry, tagged as such, so accuracy varies by area. A parcel on the shore reads zero.

The distance from the parcel boundary to the nearest mapped stream in the national hydrography layer. Features are traced from lidar elevation data where an area has been remapped; where it has not, the layer carries the older 1:24,000 geometry, tagged as such, so what counts as a mapped stream varies by area.

The stream nearest the parcel, how steeply it falls, and the length of stream over which that fall was measured. Gradient is the drop along the sampled stretch divided by its length, taken from a 10 m national elevation model. Under 300 m of sampled stream no gradient is reported and the length available is given instead.

Two readings. The distance to the nearest surveyed navigation channel, measured to the outer boundary of a completed channel survey. And, where a surveyed lake lies within 500 m, that lake's name, its depth at the point nearest the parcel, its deepest surveyed point and the year of the survey, read from mapped bathymetric contours.

Five readings. Flood: the zone covering the most of the parcel, the share of it inside the area of special flood hazard — land subject to a 1 percent or greater chance of flooding in any given year — and the base flood elevation, the flood having a one percent chance of being equalled or exceeded in any given year. Storm surge: the lowest hurricane category whose modelled surge reaches the parcel and the deepest modelled water on it, from an ensemble that keeps the highest surge in each grid cell across up to 100,000 hypothetical storms per category, run at high tide, which the publisher describes as the near worst case under ideal storm conditions rather than a forecast. Shaking: peak ground acceleration with a 2 percent and a 10 percent chance of being exceeded in 50 years, for a firm-rock site. Hurricanes: six-hourly best-track positions of all known tropical and subtropical cyclones since 1851, counted within 50 km. Tornadoes: recorded damage paths since 1950, counted within 5 km.

An index that quantifies the relative potential for high-intensity wildfire that may be difficult to manage, at 270 m, in five classes from very low to very high. The release carries the share of the parcel in the top two classes, the share that can burn at all, and the mean index; non-burnable ground and water are left out of the burnable share rather than scored low. The publisher states that on its own the index is not a map of wildfire threat or risk, is not a forecast for any particular season, reflects landscape conditions as of the end of 2020, and is not intended to replace a locally calibrated assessment.

Whether the parcel has burned in a mapped fire since 1984, how often, in what year, at what severity class, and over what share of its area. The mapping programme assesses the frequency, extent and magnitude of large wildland fires — wildfires and prescribed fires reported as larger than 1,000 acres in the western United States and 500 acres in the eastern United States — at 30 m. Fires are omitted where suitable satellite imagery was not available or the burn was not discernible in it.

Thirty-year averages for 1991 to 2020, taken over the parcel from a 4 km regional hydroclimate reanalysis of historical conditions: the output of a regional weather model forced with a global reanalysis, with enough temporal and spatial detail to resolve mesoscale atmospheric processes. The readings are annual precipitation, mean air temperature, the warmest and coldest month, peak snow water equivalent, days at or above 90 °F, days at or below 32 °F, and the yearly peak wind.

Long-term average global horizontal irradiation over the parcel: the total solar energy received on a horizontal surface, direct and diffuse together, in kilowatt-hours per square metre per day at 250 m. It is modelled from weather-satellite imagery and atmospheric data, not measured on the site.

Fine particulate matter as a yearly average on a 0.01° grid, about 1 km, estimated by combining satellite aerosol optical depth retrievals with a chemical transport model and then calibrating to ground-monitor measurements; the mean of the newest five years and the federal annual standard in force for the year sit beside it. Surface nitrogen dioxide is a yearly average on the same grid, from a machine-learning model built on process-based model output, road-network and land-cover predictors, evaluated against 9,913 ground monitors. The publisher states these estimates are primarily intended to aid large-scale studies and are unlikely to fully resolve gradients at the grid resolution.