What is Land Cover Classification? Classes and Data

Land cover classification is the process of labeling each part of the Earth's surface by the physical material that covers it. Examples are water, tree canopy, grass, crops, bare soil, or pavement. Remote sensing measures the surface from satellites and aircraft. In remote sensing, a classifier (a rule or model) assigns every pixel or segment of satellite or aerial imagery to one class. The classes come from a fixed list, the legend, and the result is a land cover map. Agencies, insurers, planners, and climate scientists use these maps to measure where forests, farms, and cities are and how they change. This article tests the claim that a map's accuracy score depends on its reference sample, and that two good maps of one place disagree.

Key takeaways

• Land cover is what physically covers the ground. Land use is what people do with it.
• Classification systems define the classes. Examples are the Anderson system and the National Land Cover Database (NLCD). Others are Europe's CORINE (Coordination of Information on the Environment) and the Food and Agriculture Organization (FAO) Land Cover Classification System (LCCS).
• Classifiers range from unsupervised clustering to supervised random forests and deep learning models.
• Every map needs an accuracy check against independent reference samples. On one reference set, three 10 m global maps scored 65% to 75% overall accuracy.
• Open maps range from 1 m regional products to 10 m and 30 m national and global ones, with Annual NLCD for every year from 1985 to 2025.
• In Wherobots, European Space Agency (ESA) WorldCover rasters and Overture land cover polygons are queryable in SQL. RasterFlow (Public Preview) classifies new imagery.

Land cover vs land use

Overture Maps defines land cover as the physical thing covering the land, and land use as the human use to which the land is being put.

	Land cover	Land use
Describes	Physical surface material	Human purpose
Examples	Tree canopy, grass, water, pavement	Housing, farming, recreation, industry
Primary source	Satellite and aerial imagery	Parcels, zoning, surveys, imagery
Golf course	Grass and tree canopy	Recreation

Imagery records land cover directly, because pixel values measure the surface. Land use needs context. A field of grass looks the same in imagery whether it is a park, a pasture, or a vacant lot. Many products combine the two and are labeled land use land cover (LULC).

Land cover classes and classification systems

A classification system fixes the classes and their levels so maps from different agencies and years stay comparable. Legends vary with purpose. NASA's Earth Observatory notes that land cover maps range from two categories for flood maps to seventeen for standard global maps.

Most systems share these types: water, forest or tree cover, shrubland, grassland, cropland, wetland, developed or built-up land, barren land, and snow and ice. The best-known five-class level is the top of CORINE Land Cover, Europe's inventory. Its five classes are artificial surfaces, agricultural areas, forests and semi-natural areas, wetlands, and water bodies. CORINE splits them into 15 classes at level two and 44 at level three.

The Anderson system, U.S. Geological Survey (USGS) Professional Paper 964 by Anderson, Hardy, Roach, and Witmer (1976), set a US federal standard for classifying land from remote sensor data. It fixed the first two levels of a national legend so maps from different agencies could be added together. It left levels three and four for agencies to extend.

The National Land Cover Database (NLCD) still uses a modified Anderson Level II, with two-digit codes:

Code	NLCD class	Code	NLCD class
11	Open Water	52	Shrub/Scrub
21	Developed, Open Space	71	Grassland/Herbaceous
22	Developed, Low Intensity	81	Pasture/Hay
23	Developed, Medium Intensity	82	Cultivated Crops
24	Developed High Intensity	90	Woody Wetlands
31	Barren Land	95	Emergent Herbaceous Wetlands
41, 42, 43	Deciduous, Evergreen, Mixed Forest	12	Perennial Ice/Snow

The developed classes are split by impervious surface: Developed, Open Space has less than 20% impervious cover, and Developed, Low Intensity has 20% to 49%.

Global products use the United Nations Food and Agriculture Organization's Land Cover Classification System (LCCS), which Di Gregorio and Jansen published in 2000. Unlike Anderson's fixed levels, LCCS sorts land through a series of two-way choices. It builds each class from separate diagnostic criteria, such as life form and cover density, so any legend can be translated into any other. ESA WorldCover defines its 11 classes with LCCS, from Tree cover and Cropland to Built-up and Mangroves. Global maps from MODIS, the Moderate Resolution Imaging Spectroradiometer, use the 17-class legend of the International Geosphere-Biosphere Programme (IGBP). The Dynamic World taxonomy stays close to the land categories of the Intergovernmental Panel on Climate Change (IPCC): forest land, grassland, cropland, wetland, settlement, and other.

How land cover classification works

1. Choose the imagery. Pick a sensor and season. Then build a cloud-free composite, one image made from many dates, often from each pixel's median.
2. Collect training labels. Label sample areas by hand or from an existing map.
3. Extract features, the inputs the classifier reads. Use spectral bands, indices such as NDVI (normalized difference vegetation index), texture, elevation, and the time series of each pixel.
4. Classify. Apply a classifier to every pixel or segment.
5. Clean up. Remove isolated pixels and enforce a minimum mapping unit, the smallest patch the map shows.
6. Assess accuracy. Compare the map against independent reference samples.

Ned Horning's guide to land cover classification methods for the American Museum of Natural History sorts classifiers into families:

• Unsupervised classification. A clustering algorithm such as k-means or ISODATA (Iterative Self-Organizing Data Analysis Technique) groups similar pixels, and an analyst names each cluster afterward.
• Supervised classification. An algorithm is trained on sample pixels an analyst labeled. Breiman's random forests (Machine Learning, 2001) take a vote of many decision trees. They became a standard land cover classifier because they handle many correlated features and rank the features that matter most. Belgiu and Drăguţ's review (2016) documents their spread through remote sensing.
• Object-based image analysis (OBIA). Pixels are first grouped into segments, which are classified by their spectra, shape, and context. Blaschke's Object based image analysis for remote sensing (2010) traces how the unit of analysis moved from the pixel to the segment. He links the shift to pixels becoming smaller than the objects being mapped.
• Deep learning segmentation. A convolutional network such as the U-Net (Ronneberger, Fischer, and Brox, 2015) labels each pixel from the pixels around it. With that context, the network separates a shadowed roof from water and a lawn from a crop field. Zhu and colleagues' Deep learning in remote sensing (IEEE Geoscience and Remote Sensing Magazine, 2017) and Ma and colleagues' meta-analysis (ISPRS Journal, 2019) record the field's move to these networks.

Resolution sets which classes a map can show. At 30 m a pixel mixes a house, its yard, and the street, so NLCD measures developed intensity by percent impervious cover. At 1 m, the Chesapeake Conservancy maps structures, roads, and tree canopy over roads as separate classes. In the header image of Rockville, Maryland, the 1 m map separates single houses and roads, and the 10 m map merges them into built-up blocks.

Accuracy assessment compares the map with reference labels at sample locations. It tabulates the result in an error matrix, a table of mapped class against true class, as Congalton set out in his 1991 review. The matrix gives overall accuracy, user's accuracy (how often a mapped class is right on the ground), and producer's accuracy (how much of a true class the map found). Olofsson and colleagues' good practices (2014) add a probability sample, where every location has a known chance of selection. They also add class areas estimated from that sample with confidence intervals, because pixel counts include every misclassified pixel.

Land cover classification examples

Well-known examples of satellite image land cover classification:

• NLCD and Annual NLCD. The USGS maps the conterminous US at 30 m from Landsat. Annual NLCD Collection 1.2, released in June 2026, covers every year from 1985 to 2025 with 16 classes and uses deep learning models.
• ESA WorldCover. A 10 m global map from Sentinel-2 and Sentinel-1 radar. ESA reports independently validated overall accuracy of 74.4% for the 2020 map and 76.7% for 2021.
• Dynamic World. Brown and colleagues (2022) trained a deep learning model on 10 m Sentinel-2 imagery. They publish class probabilities for each new Sentinel-2 image, so a map can cover any date range.
• Esri and Impact Observatory 10 m maps. Karra and colleagues trained a segmentation model on over 5 billion human-labeled Sentinel-2 pixels (IGARSS 2021) for annual global maps.
• Global forest change. Hansen and colleagues mapped 21st-century forest loss and gain at 30 m from Landsat (Science, 2013).
• Chesapeake Bay 1 m land cover. The Chesapeake Conservancy maps the watershed from aerial imagery and lidar, with 900 times more detail than 30 m NLCD.
• Pretrained models. Esri's Land Cover Classification (Landsat 8) model is a U-Net trained on NLCD 2016, so it outputs NLCD classes from new Landsat scenes.

Land cover maps measure impervious surfaces, such as roofs and pavement, where rain runs off, for flood risk models and stormwater fees. Forest extent and loss go into carbon accounting. Catastrophe models use vegetation near structures for wildfire exposure. Planners start tree canopy targets and heat mapping from a land cover map.

Land cover datasets

Resolution, update frequency, and class detail trade off against each other.

Dataset	Coverage	Resolution	Classes	Time span
Annual NLCD	Conterminous US	30 m	16	1985 to 2025, yearly
USGS LCMAP	Conterminous US	30 m	8	1985 to 2021, yearly
ESA WorldCover	Global	10 m	11	2020 and 2021
Dynamic World	Global	10 m	9, as probabilities	Each Sentinel-2 image
Impact Observatory 10 m LULC	Global	10 m	9	2017 to 2023, yearly
CORINE Land Cover	Europe	Vector inventory	44	Since 1990, updated periodically
Chesapeake LULC	Chesapeake Bay watershed	1 m	13 to over 50	2013/14 to 2021/22

Land cover change

Land cover change is the shift from one class to another between dates, such as forest to development or pasture to cropland. Annual NLCD publishes a land cover change product that joins the before and after codes into one number. 8123 means Pasture/Hay became Developed, Medium Intensity, and 7182 means Grassland/Herbaceous became Cultivated Crops.

Comparing two independent maps exaggerates change, because each map's errors look like change. ESA warns that differences between its 2020 and 2021 WorldCover maps mix real change with algorithm changes. USGS Land Change Monitoring, Assessment, and Projection (LCMAP) instead models each pixel's full record of observations with Zhu and Woodcock's Continuous Change Detection and Classification (CCDC) algorithm. CCDC fits a harmonic model, a smooth yearly curve, to every clear Landsat observation. It flags a break when new observations depart from the curve. The Chesapeake Conservancy reports its 2013 to 2022 change data captures 96% of land cover changes with a mapped accuracy of 77%. Change detection covers the methods in depth.

Land cover classification in Wherobots

Overture land cover in the Havasu catalog holds ESA WorldCover classes as polygons, with Overture land use alongside. Overture stores polygons for several map resolutions, marked by cartography.min_zoom and cartography.max_zoom, so an area summary must filter to one first. The WorldCover rasters are public Cloud Optimized GeoTIFFs. WherobotsDB reads them in place in S3 as out-of-database rasters, fetching only the tiles a query touches.

The queries below read these inputs:

Input	Format	Size	What the queries read
ESA WorldCover 2021 (v200), file N39W078	Cloud Optimized GeoTIFF in public S3, 10 m pixels	3 by 3 degrees, about 36,000 by 36,000 pixels	The 1024-pixel tiles that intersect the box, clipped to 1,555,200 pixels
ESA WorldCover 2020 (v100), file N39W078 (second query only)	Cloud Optimized GeoTIFF in public S3, same 10 m grid	3 by 3 degrees, same grid	The same tiles and pixels
Rockville box	Polygon written in the query as well-known text (WKT), in longitude and latitude (EPSG:4326)	One polygon, longitude -77.21 to -77.09, latitude 39.04 to 39.13	The clip boundary

This query counts WorldCover 2021 pixels by class in a box around Rockville, Maryland, the place in the header image:

-- WorldCover 2021 class counts for a box around Rockville, Maryland
WITH box AS (
  SELECT ST_GeomFromText('POLYGON((-77.21 39.04, -77.09 39.04, -77.09 39.13, -77.21 39.13, -77.21 39.04))', 4326) AS g
),
wc AS (  -- 1024-pixel tiles of the 3 by 3 degree file, read on demand
  SELECT RS_TileExplode(RS_FromPath('s3://esa-worldcover/v200/2021/map/ESA_WorldCover_10m_2021_v200_N39W078_Map.tif'), 1024, 1024) AS (x, y, rast)
),
c AS (   -- RS_Clip(raster, band, geometry, allTouched, noDataValue, crop)
  SELECT /*+ BROADCAST(box) */ RS_Clip(wc.rast, 1, box.g, false, 0, true) AS c
  FROM wc CROSS JOIN box
  WHERE RS_Intersects(wc.rast, box.g)
)
SELECT v AS worldcover_class, COUNT(*) AS pixels,
       ROUND(100 * COUNT(*) / SUM(COUNT(*)) OVER (), 1) AS pct
FROM (SELECT explode(RS_BandAsArray(c, 1)) AS v FROM c)
WHERE v > 0
GROUP BY v
ORDER BY pixels DESC;

One WherobotsDB query returns the class counts of the box's 1,555,200 pixels in about 20 seconds. Tree cover is 64.3%, built-up 23.2%, grassland 11.2%, permanent water 0.6%, bare or sparse vegetation 0.4%, and cropland 0.2%. RS_TileExplode splits the 3 by 3 degree file into 1024-pixel tiles, and only tiles intersecting the box are clipped.

Swapping the box for parcel, tract, or watershed polygons gives a zonal summary, the class shares in each polygon. WherobotsDB runs raster and vector in one engine; most stacks pick one.

Two WorldCover versions over Rockville

ESA produced WorldCover 2020 with algorithm v100 and WorldCover 2021 with v200. Both maps share one 10 m grid. This cross-tabulation counts the Rockville pixels for each pair of 2020 and 2021 classes:

-- Cross-tabulate WorldCover 2020 (v100) and 2021 (v200) pixels in one box
WITH box AS (
  SELECT ST_GeomFromText('POLYGON((-77.21 39.04, -77.09 39.04, -77.09 39.13, -77.21 39.13, -77.21 39.04))', 4326) AS g
),
t20 AS (
  SELECT RS_TileExplode(RS_FromPath('s3://esa-worldcover/v100/2020/map/ESA_WorldCover_10m_2020_v100_N39W078_Map.tif'), 1024, 1024) AS (x, y, rast)
),
t21 AS (
  SELECT RS_TileExplode(RS_FromPath('s3://esa-worldcover/v200/2021/map/ESA_WorldCover_10m_2021_v200_N39W078_Map.tif'), 1024, 1024) AS (x, y, rast)
),
pairs AS (  -- the two files share one grid, so tile (x, y) covers the same ground in both
  SELECT /*+ BROADCAST(box) */
         RS_BandAsArray(RS_Clip(a.rast, 1, box.g, false, 0, true), 1) AS v20,
         RS_BandAsArray(RS_Clip(b.rast, 1, box.g, false, 0, true), 1) AS v21
  FROM t20 a JOIN t21 b ON a.x = b.x AND a.y = b.y
  CROSS JOIN box
  WHERE RS_Intersects(a.rast, box.g)
),
cells AS (
  SELECT z.v20 AS class_2020, z.v21 AS class_2021
  FROM (SELECT explode(arrays_zip(v20, v21)) AS z FROM pairs)
  WHERE z.v20 > 0 AND z.v21 > 0
)
SELECT class_2020, class_2021, COUNT(*) AS pixels,
       ROUND(100 * COUNT(*) / SUM(COUNT(*)) OVER (), 2) AS pct
FROM cells
GROUP BY class_2020, class_2021
ORDER BY pixels DESC;

The query runs in about 20 seconds. Of the 1,555,200 pixels, 91.0% keep the same label and 9.0%, or 140,716 pixels, change. The largest moves are tree cover to built-up (2.6%), grassland to tree cover (1.6%), tree cover to grassland (1.0%), and bare or sparse vegetation to built-up (0.9%). Built-up land rises from 20.1% of the box in the 2020 map to 23.2% in 2021, and bare land falls from 1.8% to 0.4%. The 9.0% is label disagreement, which mixes real change with the algorithm change. A pixel can also change on the ground yet keep its label. Without reference labels, the 9.0% is neither an upper nor a lower bound on real change in Rockville.

Classifying new imagery with RasterFlow

The catalog also holds imagery for custom classification: Sentinel-2 seasonal mosaics at 10 m and USDA NAIP aerial imagery at 30 cm to 1 m. NAIP is the National Agriculture Imagery Program of the U.S. Department of Agriculture. RasterFlow (Public Preview) builds mosaics from imagery, runs computer vision inference on them, and writes the results as vectorized outputs on the same platform. For land cover, the input is cloud-free Sentinel-2 composites with all 12 L2A (surface reflectance) bands at 10 m. A segmentation model runs through the bring-your-own-model path, and the output is class polygons. This call builds a harvest-season mosaic for an area of interest and writes it to your bucket as Zarr, a cloud format for large arrays:

# Build a cloud-free Sentinel-2 composite as the input for a land cover model
from datetime import datetime
from rasterflow_remote import RasterflowClient, DatasetEnum

client = RasterflowClient()
result = client.build_mosaics(
    datasets=[DatasetEnum.S2_MED_HARVEST],
    aoi="s3://my-company-data/aois/project.parquet",  # one or multiple geometries
    start=datetime(2024, 1, 1),
    end=datetime(2025, 1, 1),
    bucket="s3://my-company-data/rasterflow/results",
)
print(result.first_row_mosaic)

The change detection workflow stacks two seasons or years of imagery for one model run. For change without training labels, RasterFlow scores the AlphaEarth embedding distance across years. An embedding is a list of numbers that summarizes each pixel. Classified polygons load into WherobotsDB, where a spatial join summarizes land cover by parcel, watershed, or county. WherobotsDB is built by the original creators of Apache Sedona and is 100% code compatible across all spatial functions.

Land cover classification research

Land cover maps from 1 km to 1 m pixels

Loveland and colleagues built IGBP DISCover (International Journal of Remote Sensing, 2000) with the IGBP legend. They used 1 km data from AVHRR, the Advanced Very High Resolution Radiometer. Friedl and colleagues' MODIS Collection 5 land cover (Remote Sensing of Environment, 2010) moved to 500 m. Gong and colleagues' FROM-GLC (2013) and Chen and colleagues' GlobeLand30 (2015) reached 30 m with Landsat. FROM-GLC stands for Finer Resolution Observation and Monitoring of Global Land Cover. Buchhorn and colleagues' Copernicus Global Land Cover Layers (2020) added cover fractions, the share of each class in each pixel, at 100 m. The 10 m generation came with Sentinel-2: ESA WorldCover (Zanaga and colleagues), Dynamic World (Brown and colleagues, 2022), and the Impact Observatory and Esri maps (Karra and colleagues, 2021).

The 1 m generation is regional. Robinson and colleagues' Large Scale High-Resolution Land Cover Mapping with Multi-Resolution Data (CVPR 2019) trained on the Chesapeake Conservancy's 1 m labels. The authors used 30 m NLCD as coarse labels to extend the model to the rest of the conterminous US.

Accuracy assessment and why two good maps disagree

Congalton's error matrix and Foody's Status of land cover classification accuracy assessment (2002) set the standard terms. Pontius and Millones' Death to Kappa (2011) argued that kappa, a statistic that adjusts accuracy for chance agreement, misleads. They split disagreement into quantity (wrong class proportions) and allocation (right proportions, wrong places). Stehman and Foody's Key issues in rigorous accuracy assessment of land cover products (2019), with the Olofsson good practices, made probability sampling and sample-based area estimates the standard.

The 10 m global maps show why. Venter and colleagues compared Dynamic World, WorldCover, and Esri Land Cover (Remote Sensing, 2022) against global reference data. They found overall accuracy of 75% for Esri, 72% for Dynamic World, and 65% for WorldCover. ESA's own validation sample gives WorldCover 74.4% and 76.7%, so the score depends on the reference sample as much as on the map.

Resolution and the mixed pixel

Woodcock and Strahler's The factor of scale in remote sensing (1987) showed that the right pixel size depends on three things. They are the information wanted, the method, and the spatial structure of the scene. Fisher's The pixel: a snare and a delusion (1997) argued that treating a pixel as one uniform patch of ground is a mistake. At 10 m, a pixel on a suburban street holds part of a roof, a lawn, a tree crown, and asphalt, and the map can record only one label.

Finer pixels bring occlusion: one object hides another from the sensor. Robinson, Corley, and colleagues' Seeing the roads through the trees (2024) built the ChesapeakeRSC benchmark from 1 m NAIP imagery of Maryland, with a class for tree canopy over roads. A U-Net trained on the benchmark reached 84% recall on visible roads and 63.5% on roads under canopy. Recall is the share of true road pixels found. It drops under canopy because the pixels above a covered road look like trees. RasterFlow ships the ChesapeakeRSC model. Its Maryland output is in the Havasu catalog.

Open problems

• Area estimation. Sample-based estimators correct pixel-count bias but need a probability sample of reference labels for every study area.
• Labels. Deep models need many labels, which are expensive. The Chesapeake Conservancy spent 10 months and $1.3 million on its 1 m map of the Chesapeake Bay watershed, as Robinson and colleagues report.
• Legends. A legend built for carbon accounting splits classes differently from one built for stormwater. Translating between legends loses detail.

What scale changes

• The pixel count. WorldCover 2021 ships as 2,651 Cloud Optimized GeoTIFF tiles of 3° by 3°, each at 10 m, and Robinson and colleagues' 1 m US map holds over 8 trillion pixels.
• The join. Decisions are made per parcel, field, tract, or watershed, so a raster-vector join sums map pixels inside each polygon. Jia Yu, Jinxuan Wu, and Mohamed Sarwat introduced GeoSpark (ACM SIGSPATIAL 2015) for distributed spatial joins. Yu, Zongsi Zhang, and Sarwat described its partitioning and indexing in Spatial data management in Apache Spark: the GeoSpark perspective and beyond (GeoInformatica, 2019). GeoSpark became Apache Sedona, and Yu and Sarwat founded Wherobots.
• Training. Kanchan Chowdhury and Mohamed Sarwat's GeoTorch (ACM SIGSPATIAL 2022), now GeoTorchAI, adds raster transforms and benchmark datasets for training land cover models at scale.

Conclusions

For land cover classification, the evidence shows that pixel size limits the classes, and the measuring method shapes accuracy and change figures.

1. Pixel size sets which classes a land cover map can separate. A 30 m pixel mixes a house, its yard, and the street, so NLCD scores developed land by impervious share. The 1 m Chesapeake map separates houses from roads. For single structures or roads, 10 m and 30 m products answer a different question.
2. A land cover map's accuracy score depends on the reference sample as well as the map. WorldCover scored 65% on Venter and colleagues' reference set, and 74.4% and 76.7% on ESA's own sample. Score candidate maps on one reference sample, and estimate class areas from a probability sample with confidence intervals.
3. Land cover change found by differencing two separately produced maps is label disagreement. Over Rockville, 9.0% of pixels changed label between WorldCover v100 and v200, mixing real change with algorithm change. Use one time-series method such as CCDC, and validate the change map against reference labels.
4. Summarizing a land cover map by area is a raster-vector query. Tile large rasters before clipping: RS_TileExplode turned a 36,000-pixel-wide WorldCover file into a 20-second count of 1,555,200 pixels.
5. Where no open land cover map holds the class needed, train a classifier and expect occlusion to cap recall. The ChesapeakeRSC U-Net reached 84% on visible roads and 63.5% under canopy in one 1 m benchmark in Maryland, so other regions need their own test.

Related terms

• Satellite imagery analysis: the computer vision tasks behind land cover maps
• Change detection: comparing imagery and maps over time
• NDVI: the vegetation index used as a classification feature
• Earth observation data: the imagery land cover maps are made from
• Raster vs vector data: land cover exists as both
• Sentinel-2: the 10 m imagery behind WorldCover and Dynamic World
• Landsat: the 30 m archive behind NLCD

Read more from Wherobots

• Raster processing at scale: the out-of-database architecture behind WherobotsDB
• AlphaEarth Foundations for agriculture change detection
• Zonal statistics on AlphaEarth embeddings
• Fields of the World: a GeoAI pipeline on RasterFlow
• Build a Sentinel-2 mosaic with RasterFlow

Query land cover rasters and polygons with spatial SQL with a Wherobots free trial at cloud.wherobots.com.