Annual National Land Cover Database
Description
This dataset characterises the land cover class, land cover change, impervious surface class, fractional impervious surface (%), and spectral change day of year at 30 m spatial resolution. Coverage is for the conterminous United States annually from 1985 onwards. Pixels are assigned one of 16 land cover classes.
2ce9cb24…Usage notes
New time points are delivered annually. Spatial coverage is limited to the conterminous United States. Land cover values are associated with a July 1st date, so changes after this date appear in the following year's data.
Variables
land_cover
Description
Index of the predicted land cover class. Predictions are made using deep learning models trained on National Land Cover Database 2019 labels, Landsat data, and a range of ancillary datasets.
Usage notes
Indexes represent the surface state on July 1st of each year, so changes after this date appear in the following year's data. Overall accuracy is 82.5%, consistent between years but lower in the eastern US. Highest accuracy is for water (96%, 93% user/producer accuracy) and tree cover (90%, 83%). Lowest accuracy is for wetland (69%, 74%) and barren land (43%, 57%). Out-of-place classifications (particularly developed and barren) may occur over water bodies. Linear artefacts may appear in the desert Southwest due to shrub/scrub and grassland/herbaceous confusion. Not all land cover changes are detectable.
| Index | Name | Description |
|---|---|---|
| 11 | Open Water | Open water with < 25% vegetation or soil cover. |
| 12 | Perennial Ice/Snow | Permanent ice or snow cover (> 25% total cover). |
| 21 | Developed, Open Space | Mixed constructed materials and vegetation (e.g. lawns, parks, golf courses). Impervious surfaces are < 20% total cover. |
| 22 | Developed, Low Intensity | Mixed constructed materials and vegetation (e.g. single-family housing). Impervious surfaces are 20%-49% total cover. |
| 23 | Developed, Medium Intensity | Mixed constructed materials and vegetation (e.g. single-family housing). Impervious surfaces are 50%-79% total cover. |
| 24 | Developed, High Intensity | Highly developed areas (e.g. apartments, commercial, industrial). Impervious surfaces are 80%-100% total cover. |
| 31 | Barren Land (Rock/Sand/Clay) | Areas of low vegetation (e.g. bedrock, desert pavement, scarps, talus, slides, volcanic material, gravel pits). Vegetation is generally < 15% total cover. |
| 41 | Deciduous Forest | Trees > 5 m and > 20% vegetation cover. At least 75% tree species shed leaves seasonally and simultaneously. |
| 42 | Evergreen Forest | Trees > 5 m and > 20% vegetation cover. At least 75% tree species maintain leaves year-round. Canopy is never without green foliage. |
| 43 | Mixed Forest | Trees > 5 m and > 20% vegetation cover. Neither deciduous nor evergreen species exceed 75% total tree cover. |
| 52 | Shrub/Scrub | Shrubs < 5 m with shrub canopy typically > 20% total vegetation. Includes shrubs, young or early successional trees, and stunted trees. |
| 71 | Grassland/ Herbaceous | Graminoid or herbaceous vegetation (> 80% vegetation cover). Not subject to intensive management (e.g. tilling) but can be grazed. |
| 81 | Pasture/Hay | Planted grasses and/or legumes for livestock grazing or seed/hay production (> 20% vegetation cover). |
| 82 | Cultivated Crops | Annual crops (e.g. corn, soybeans, vegetables, tobacco, cotton) or perennial woody crops (e.g. orchards, vineyards). Crop vegetation > 20% vegetation cover. Includes all actively tilled land. |
| 90 | Woody Wetlands | Forest or shrubland (> 20% vegetation cover) with periodically water saturated soil or substrate. |
| 95 | Emergent Herbaceous Wetlands | Perennial herbaceous vegetation (> 80% vegetation cover) with periodically water saturated soil or substrate. |
land_cover_change
Description
Index indicating land cover change between consecutive years, if present.
Usage notes
Indexes concatenate previous and current year land cover indexes (e.g. 9590), with no change retaining the current year index (e.g. 95). Changes are associated with July 1st, so changes after this date appear in the following year's data. Accuracy for matching the exact year of change is 13% and 16% (user/producer accuracy), increasing to 28% and 34% with a two-year tolerance. Accuracy is higher in the eastern US (49%, 54%).
land_cover_confidence
Description
Probability value for the land cover class. Values are calculated by transforming neural network outputs into probabilities.
Usage notes
Values indicate relative confidence that does not represent exact accuracy. Confidence may not always correspond to maximum probability across all classes. Processing errors can yield values > 100 or wrap to unexpected values. Occasional incorrect low confidence values can occur due to wrong classes being remapped during processing.
fractional_impervious_surface
Description
Percentage of a pixel covered with artificial impermeable surfaces. Data is generated using regression models trained on National Land Cover Database 2019 impervious labels.
Usage notes
Impermeable surfaces are defined as pavement, concrete, rooftops, and other constructed materials. Values represent the July 1st surface state, so changes after this date appear in the following year's data. Provides the basis for developed land cover class assignments. Processing errors can yield values > 100 or wrap to unexpected values.
impervious_descriptor
Description
Sub-index of developed land cover classes. Predictions are made using deep learning models trained on National Land Cover Database 2019 labels, Landsat data, and a range of ancillary datasets.
Usage notes
Indexes represent the surface state on July 1st of each year, so changes after this date appear in the following year's data. Out-of-place classifications may occur over water bodies. Not all impervious surface changes are detectable.
| Index | Name |
|---|---|
| 0 | Non-urban |
| 1 | Roads |
| 2 | Urban |
spectral_change_day_of_year
Description
Day of year when substantial surface reflectance changes were detected in Landsat data. Data is generated using a continuous change detection algorithm to detect deviations in spectral behaviour of Landsat data over time.
Usage notes
Values range from 1-366, with 0 indicating no change. Spectral changes represent abrupt non-seasonal surface changes that may or may not correspond to land cover change. High-intensity events (e.g. wildfire) produce both spectral and land cover changes, while other events (e.g. drought) produce only spectral changes. Low-intensity or ephemeral events (e.g. low-intensity fire, tree thinning, urban resurfacing) may produce spectral changes without land cover change. Linear artefacts (striping) can occur due to input data issues.
Pricing
This dataset has no data acquisition cost.
Resources
Provider
USGS
www.usgs.govThe United States Geological Survey (USGS) is a science agency of the U.S. government that provides scientific information about the Earth's natural resources, natural hazards, and ecosystems. USGS operates the Landsat satellite programme and the Earth Resources Observation and Science (EROS) Center, producing geospatial datasets on land cover, land change, and environmental monitoring used by researchers and decision-makers worldwide.