Landsat Collection 2 analysis workflow: satellite imagery, surface reflectance, and change detection in GIS
Landsat Collection 2 is the USGS reprocessed archive of Landsat Earth-observation data, with Level-1 imagery available worldwide and Level-2 surface reflectance and surface temperature products available for Landsat 4–9 scene data. The collection also includes U.S. Analysis Ready Data tiles for the conterminous United States, Alaska, and Hawaii. This page focuses on a defensible workflow for finding scenes, selecting processing levels, applying quality masks, harmonizing dates and sensors, calculating spectral indices, and relating satellite observations to physical features and geographic regions.
Landsat Collection 2 is best understood not as one single raster layer, but as a family of consistently processed Earth-observation products. The archive includes scene-based Level-1 data, global Level-2 surface reflectance and surface temperature products for applicable sensors, and tiled U.S. Analysis Ready Data designed for time-series work. The collection supports decades of observation across forests, farms, cities, waterways, shorelines, mines, roads, and other visible or thermally observable features of the land surface. A strong GIS workflow begins by deciding what physical property is being measured, which sensor and processing level are appropriate, and how clouds, shadows, snow, water, aerosols, saturation, and missing observations will be handled.
How EDK fits
For: GIS analysts, remote-sensing practitioners, environmental researchers, land managers, urban and regional planners, geospatial engineers, and developers building reproducible workflows from public satellite imagery.
Landsat Collection 2 provides a time-indexed observation layer for understanding how physical places change. E2’s world-model perspective is relevant when these observations need to be connected to identifiable geographic features such as buildings, roads, agricultural parcels, waterways, vegetation, infrastructure, or administrative regions. Landsat does not by itself provide a complete object inventory or address layer; it records sensor observations that can be joined to geographic entities and analyzed across space and time. In that broader workflow, Landsat can contribute evidence for finding and comparing land-surface conditions, while a geographic world model supplies the places, features, boundaries, and computational context needed to interpret those observations.
- •Compare Landsat-derived vegetation, water, thermal, or built-surface indicators across known geographic features such as parks, watersheds, road corridors, neighborhoods, construction areas, or infrastructure sites.
- •Compute geographic summaries over a consistent set of places: for example, rank regions by vegetation loss, measure water occurrence near roads and buildings, detect surface change around development footprints, or compare seasonal conditions across cities.
- •Use Landsat observations as one evidence layer alongside roads, buildings, public assets, environmental features, and other location data when investigating how the physical world changes over time.
Dataset profile
Formats
Cloud Optimized GeoTIFF for Level-1, Level-2, and U.S. ARD raster products.
XML and JSON metadata included with U.S. ARD tile packages and product packages.
CSV and Parquet machine-readable metadata through the USGS Bulk Metadata Service.
Shapefile grids for U.S. ARD tile boundaries and Landsat reference systems where provided by USGS.
Browse imagery and map-oriented representations through USGS discovery tools; these should not be confused with the analysis products.
Access methods
EarthExplorer interactive search and download.
EarthExplorer API and documented bulk-download workflows.
USGS commercial-cloud access options for supported Landsat products.
USGS Bulk Metadata Service for catalog-scale metadata discovery.
GloVis and LandsatLook for browsing and scene discovery.
Local GIS and cloud-native raster tools capable of reading COGs and interpreting GeoTIFF metadata.
Quality and limits
Clouds, cloud shadows, snow, haze, aerosols, saturation, fill pixels, and view or illumination conditions can make observations unusable or incomparable.
The collection contains sensor-specific differences across Landsat generations; band names and spectral responses should be harmonized before multi-sensor analysis.
Landsat 7 imagery can contain scan-line-corrector-related gaps in affected acquisitions, requiring explicit handling in composites and time series.
USGS documents surface-reflectance overcorrection in some bright, snowy, icy, or water conditions and recommends using aerosol and pixel-quality information.
Surface-temperature data can be missing where supporting ASTER Global Emissivity Dataset information is unavailable and can show documented artifacts.
The Landsat 4–7 Collection 2 QA clear flag has a documented discrepancy under certain water and cloud conditions; analysts should use the recommended component bits instead of relying on that flag alone.
A 30-meter observation is not equivalent to a parcel, building, road centerline, address, or asset record. Mixed pixels can combine vegetation, soil, roofs, pavement, and water.
Historical results can change after reprocessing or corrections; workflows should preserve product identifiers and processing metadata.
Analysis workflows
Define the physical question first: vegetation condition, water presence, burned area, thermal pattern, built-surface change, or another measurable signal.
Select the appropriate product level and geography. Use Level-2 for many surface analyses, U.S. ARD for tiled U.S. time series, and Level-1 when specialized processing or original scene structure is required.
Search by area, date range, sensor, cloud conditions, and collection in EarthExplorer or through a documented API.
Read the product guide and metadata to confirm band meanings, scale factors, offsets, valid ranges, fill values, projection, acquisition time, and processing date.
Apply QA masks for cloud, cloud shadow, snow or ice, water, aerosol uncertainty, saturation, and fill according to the analysis objective.
Calculate indices or physical proxies only after masking and scaling. Preserve intermediate rasters and the exact formula used.
For time series, harmonize seasonal windows, sensor bands, projections, resolution, compositing rules, and valid-pixel thresholds.
Join raster outputs to geographic features such as watersheds, parks, roads, buildings, neighborhoods, infrastructure sites, or administrative areas using zonal statistics or raster-vector overlays.
Validate apparent changes against imagery, higher-resolution data, field observations, or authoritative local records before treating them as real-world events.
Record product IDs, acquisition and processing dates, collection version, masks, thresholds, software versions, and reprocessing notices for reproducibility.
Value inside EDK
What Landsat Collection 2 contains
The Landsat Collection 2 archive contains imagery acquired by Landsat satellites beginning in 1972. USGS describes the collection as a reprocessed and standardized inventory intended to improve geometric consistency, calibration, metadata, and interoperability across the long-running Landsat record. Level-1 products are scene-based and include worldwide coverage for Landsat sensors. Level-2 products add derived surface measurements, including surface reflectance and surface temperature, for applicable Landsat 4–9 observations.
Surface reflectance is intended to reduce the effect of atmospheric conditions and make observations more comparable across dates. Surface temperature is derived from thermal infrared observations and supporting inputs, but it is not available for every scene or pixel. The U.S. Analysis Ready Data product reorganizes selected Level-2 inputs into a fixed tile grid and common Albers Equal-Area Conic projection for the conterminous United States, Alaska, and Hawaii. ARD packages include top-of-atmosphere reflectance, brightness temperature, surface reflectance, surface temperature where available, quality-assessment bands, and XML or JSON metadata.
- •Level-1 scene products: worldwide Landsat imagery and associated metadata.
- •Level-2 scene products: global surface reflectance and surface temperature products for applicable Landsat 4–9 observations.
- •U.S. ARD tiles: analysis-ready products for the conterminous United States, Alaska, and Hawaii.
- •Quality layers: cloud, cloud shadow, snow or ice, water, aerosol, saturation, fill, and related quality information.
- •Metadata: product identifiers, acquisition and processing information, calibration details, geometry, and product-specific attributes.
Geography, temporal coverage, and update cadence
Geographic coverage depends on the product. Landsat Collection 2 Level-1 scene products are available worldwide across the archive. USGS identifies global Level-2 scene products for Landsat 4–9, beginning with the Thematic Mapper era, while U.S. ARD is limited to the conterminous United States, Alaska, and Hawaii. The archive is therefore global in its principal scene products but not every derived product is global.
The time span also varies by satellite and sensor. The USGS bulk metadata service lists Landsat 4–5 Thematic Mapper Level-1 and Level-2 records from 1982–2012, Landsat 7 records from 1999–2025 on the current service page, and Landsat 8–9 records from April 2013 to the present. Older Multispectral Scanner records extend the Level-1 archive back to 1972, but corresponding Level-2 products are not available for all early sensors. Analysts should query the current catalog for the exact date range, sensor, processing level, and geographic footprint rather than assuming that every product exists for every year.
- •Worldwide Level-1 scene coverage across the Landsat archive.
- •Global Level-2 scene coverage for applicable Landsat 4–9 products.
- •U.S. ARD coverage limited to the conterminous United States, Alaska, and Hawaii.
- •Historical coverage is sensor-dependent, with the broader Level-1 archive beginning in 1972.
- •No single universal update schedule; new acquisitions, metadata updates, and reprocessing events occur on different timelines.
Formats, access methods, and a defensible workflow
USGS distributes Landsat Collection 2 Level-1 and Level-2 image products as Cloud Optimized GeoTIFF files. COG structure allows software and cloud workflows to request image windows or bands without downloading an entire raster in every situation. U.S. ARD products are also stored as COGs, with associated XML and JSON metadata. The USGS bulk metadata service provides machine-readable metadata in XML, CSV, and, according to the current service documentation, Parquet formats.
The main interactive access point is EarthExplorer, where users can search by location, date, cloud cover, dataset, and other criteria. USGS also documents bulk download options, an EarthExplorer API, commercial-cloud access, and specialized services for metadata and processing. GloVis and LandsatLook are useful for browsing or discovering imagery, but the exact access path depends on whether the workflow needs Level-1 scenes, Level-2 scenes, U.S. ARD tiles, or another Landsat science product.
- •Choose Level-1, Level-2, or U.S. ARD according to the measurement and geography.
- •Use EarthExplorer, documented APIs, bulk services, or cloud access for programmatic acquisition.
- •Read scale factors, offsets, valid ranges, fill values, and band definitions from the product documentation.
- •Mask invalid pixels before computing NDVI, NDWI, NBR, land-surface temperature, change metrics, or regional statistics.
- •Record acquisition date, processing date, sensor, scene or tile identifier, collection, and quality rules.
Suitable GIS and spatial-analysis workflows
Landsat Collection 2 supports raster-based analysis at approximately 30-meter multispectral resolution, with thermal products and panchromatic or sensor-specific bands requiring careful interpretation. Common workflows include vegetation-index mapping, crop and forest monitoring, burned-area studies, surface-water screening, urban expansion analysis, mine and quarry monitoring, shoreline change, drought indicators, thermal pattern analysis, and long-term land-surface change detection.
For a single date, an analyst can select a cloud-minimized scene or tile, apply QA masks, calculate indices, clip the result to a boundary, and summarize pixels by region. For a time series, the more difficult work is temporal normalization: selecting comparable seasons, managing gaps caused by clouds or snow, accounting for Landsat 7 scan-line-corrector-era gaps where relevant, and avoiding false changes caused by sensor differences or processing artifacts. Median or percentile composites can reduce noise, but the compositing rule should be documented because it changes the meaning of the output.
- •Vegetation and crop monitoring using reflectance bands and indices such as NDVI or EVI.
- •Surface-water and wetness analysis using spectral indices, QA masks, and time-series composites.
- •Burned-area and disturbance detection using indices such as NBR and multi-date comparisons.
- •Urban and infrastructure analysis using built-surface indicators, thermal patterns, and change detection.
- •Zonal statistics, raster-vector overlays, regional ranking, and proximity-based comparisons with other geographic data.
License, strengths, and tradeoffs
USGS states that Landsat data are public domain, including data stored in cloud environments. That supports academic, government, commercial, and open-source workflows, subject to the need to preserve accurate provenance and follow any separate terms that may apply to software, derived services, or accompanying datasets. Public-domain status does not remove the need to understand product documentation or validate results before using them for consequential decisions.
The central strength of Collection 2 is continuity. A long, reprocessed archive makes it possible to study land-surface change across decades rather than relying only on recent imagery. Standardized Level-2 products, COG delivery, machine-readable metadata, and U.S. ARD tiling also reduce repeated preparation work. The main tradeoff is that the archive remains an Earth-observation record rather than a complete, current inventory of physical objects. Clouds, shadows, snow, atmospheric conditions, missing observations, sensor differences, spatial resolution, and algorithm assumptions can all affect interpretation.
- •Public-domain imagery and metadata support broad reuse.
- •Long historical continuity supports multi-year and multi-decade change analysis.
- •COG delivery and analysis-ready tiling support scalable raster workflows.
- •Results can be affected by clouds, shadows, snow, aerosols, atmospheric correction, sensor gaps, and spatial resolution.
- •Known product artifacts and reprocessing events require version tracking and validation.
- •Landsat observations are not a cadastral record, address layer, building inventory, or live monitoring feed.
Questions teams ask
Is Landsat Collection 2 one dataset or several products?
It is a family of related products. The main categories are scene-based Level-1 imagery, scene-based Level-2 surface reflectance and surface temperature products, and U.S. Analysis Ready Data tiles. Their geography, contents, formats, and suitable workflows differ.
Does Landsat Collection 2 cover the whole world?
The Level-1 scene archive is worldwide, and USGS provides global Level-2 scene products for applicable Landsat 4–9 observations. U.S. ARD is not global; it covers the conterminous United States, Alaska, and Hawaii.
What should be masked before calculating a Landsat index?
At minimum, inspect the product’s fill, cloud, cloud-shadow, snow or ice, water, aerosol, and saturation quality information. The exact mask depends on the scientific question. Analysts should also review current USGS known-issues documentation and avoid relying on a single QA flag when the documentation identifies sensor-specific discrepancies.
How often is Landsat Collection 2 updated?
There is no single collection-wide update interval. New satellite acquisitions continue to be processed, bulk metadata are updated daily, and selected products may be reprocessed. For reproducibility, record the product identifier, acquisition date, processing date, download date, and any relevant reprocessing event.
Can Landsat identify buildings, roads, or construction directly?
Landsat can provide spectral and thermal observations that may support regional or multi-date analysis of built-up surfaces, land disturbance, and development patterns. At approximately 30-meter multispectral resolution, it should not be treated as a complete building, road, address, or construction-permit inventory. Those interpretations require additional geographic data and appropriate validation.
Keep exploring
landsat collection 2 download and formats
A complementary page focused on obtaining Landsat Collection 2 products, understanding file formats, and choosing access methods.
global surface water analysis workflow
A related Earth-observation workflow showing how raster water observations can be analyzed across geographic regions and combined with other spatial layers.
Sources
Landsat Collection 2
Official USGS overview of Collection 2 products, collection improvements, tiers, cloud-optimized formats, and documentation.
Open sourceLandsat Data Access
Official USGS documentation for EarthExplorer, Level-1 and Level-2 scene products, U.S. ARD tiles, bulk access, and cloud access.
Open sourceLandsat Collection 2 U.S. Analysis Ready Data
Official USGS description of U.S. ARD geography, product contents, COG format, metadata, access, and documented artifacts.
Open sourceBulk Metadata Service
Official USGS source for current sensor date ranges, machine-readable metadata formats, and daily metadata updates.
Open sourceLandsat Collection 2 Known Issues
Official USGS documentation of known surface-reflectance, surface-temperature, NoData, aerosol, and quality-assessment limitations.
Open sourceLandsat Data Distribution Policy
Official USGS policy identifying Landsat data as public domain.
Open source