For decades, the National Land Cover Database (NLCD) has been the go-to source on land cover for scientists, resource managers and decision makers across the United States. In 2024 it was reinvented as Annual NLCD, adding yearly maps where the legacy database came out every 2–3 years.
- First release: Collection 1.0, covering land cover and change from 1985 through 2023 for the conterminous United States.
- Source: mainly the long Landsat satellite record, plus other data.
- Classes: 16 land cover classes, based on a modified Anderson Level II classification system.
Land cover of the conterminous United States in 2023, from Annual NLCD Collection 1.0. Image from the USGS fact sheet on Annual NLCD
Why yearly maps?
Understanding both sudden and long-term changes — from cities to farmland, forests to coasts — matters for people, the economy and ecosystems. We need to know what drives change, what it leads to, and how management decisions, natural hazards and climate variability play out. That calls for land cover data over larger areas, longer periods and finer detail, updated more often.
Urban growth (red) in the northern Dallas–Fort Worth area, 1985–2023; Ray Roberts Lake (blue, top center) appears after the dam was completed in 1987. Image from the USGS fact sheet on Annual NLCD
How it is made
Annual NLCD blends the strengths of the legacy NLCD — with more land cover classes — and USGS's Land Change Monitoring, Assessment, and Projection (LCMAP) project — with year-by-year time series. Both came from the USGS Earth Resources Observation and Science (EROS) Center, which leads Annual NLCD with the Multi-Resolution Land Characteristics (MRLC) consortium of federal agencies.
It uses a purpose-built geospatial artificial intelligence technique on:
- Landsat Collection 2 U.S. Analysis Ready Data at 30-meter (about 98-foot) resolution;
- land surface change and land cover data;
- independent reference data to validate classes and estimate area;
- scenario-driven projections of future land use and cover;
- assessments of land change processes and consequences.
A satellite based on Landsat 8 and 9; Landsat satellites have imaged Earth's surface since 1972. Image from the USGS fact sheet on Annual NLCD
The six products
Examples below show Marysville, Washington, near Seattle, where the pink area on Smith Island was restored from grassland to tidal marsh as habitat for salmon, including Chinook salmon.
| Product | What it shows |
|---|---|
| Land Cover | the main surface cover each year, in 16 classes |
| Land Cover Change | how cover changed from one year to the next |
| Land Cover Confidence | how confident the model is that a label matches its training data |
| Fractional Impervious Surface | how much of the ground is impervious |
| Impervious Descriptor | roads versus other built surfaces — useful where rural "development" is mostly roads |
| Spectral Change Day of Year | the day in the year when Landsat detected an abrupt change in surface reflectance |
Land Cover. Image from the USGS fact sheet on Annual NLCD
Land Cover Change. Image from the USGS fact sheet on Annual NLCD
Land Cover Confidence. Image from the USGS fact sheet on Annual NLCD
Impervious Descriptor. Image from the USGS fact sheet on Annual NLCD
Spectral Change Day of Year. Image from the USGS fact sheet on Annual NLCD
More regions may follow, and yearly updates could extend the series. Legacy NLCD data remain available on ScienceBase.
What NLCD is used for
Since work began in 1992, NLCD has supported thousands of applications in the private, public and academic sectors: assessing ecosystem health, mapping biodiversity, understanding climate change, shaping land policy, and serving as a key layer for wildfire modeling, national water quality assessments, urban heat risk, carbon sequestration and conservation.
The reach of the legacy NLCD since 1992. Image from the USGS fact sheet on Annual NLCD
Impervious surfaces affect runoff and urban heat: farmland in Frederick County, Maryland (photo by Roger Auch, USGS), has far less than downtown Chicago (photo by Eileen Hornbaker, U.S. Fish and Wildlife Service). Image from the USGS fact sheet on Annual NLCD
Checking accuracy
Image interpreters classify land cover each year, 1985–2023, on thousands of 30×30-meter plots, using Landsat imagery, high-resolution aerial photos and other data. Phase 1 uses 5,000 randomly placed plots; phase 2 adds up to 5,000 more, sampled to capture change and rare cover types.
The 5,000 phase-1 accuracy plots. Image from the USGS fact sheet on Annual NLCD
Annual vs. legacy NLCD
| Annual NLCD (data through 2023) | Legacy NLCD (data through 2021) | |
|---|---|---|
| Release | yearly | every 2–3 years |
| Classes | 16 | 16, plus 4 more for Alaska |
| Years | every year from 1985 | 2001, 2004, 2006, 2008, 2011, 2013, 2016, 2019, 2021 |
| Method | three kinds of deep learning models in a geospatial AI system, trained on curated land cover data and Landsat time series | partner data, shape and context analysis, succession and trajectory analysis, and change detection combined with AI and machine learning |
| Source | Landsat | Landsat |
NLCD products are free to download from several sources, including cloud services (which may charge data egress fees), and EROS User Services supports users.
Sources
Based on Annual NLCD (National Land Cover Database)—The Next Generation of Land Cover Mapping, U.S. Geological Survey Fact Sheet 2025–3001; a work of the United States government in the public domain.
Licence: CC0 1.0 (public domain) · Adapted from pubs.usgs.gov
1
0
0
0
Comments
