Cloud detection algorithm comparison and validation for operational Landsat data products

Steven Curtis Foga; Pat Scaramuzza; Song Guo; Zhe Zhu; Ronald Dilley; Tim Beckmann; Gail L. Schmidt; John L. Dwyer; MJ Hughes; Brady Laue

doi:10.1016/j.rse.2017.03.026

USGS · 70186881

Cloud detection algorithm comparison and validation for operational Landsat data products

Steven Curtis Foga·Pat Scaramuzza·Song Guo·Zhe Zhu·Ronald Dilley·Tim Beckmann·Gail L. Schmidt·John L. Dwyer·MJ Hughes·Brady Laue

View original source ↗Export metadata ↓

Abstract

Clouds are a pervasive and unavoidable issue in satellite-borne optical imagery. Accurate, well-documented, and automated cloud detection algorithms are necessary to effectively leverage large collections of remotely sensed data. The Landsat project is uniquely suited for comparative validation of cloud assessment algorithms because the modular architecture of the Landsat ground system allows for quick evaluation of new code, and because Landsat has the most comprehensive manual truth masks of any current satellite data archive. Currently, the Landsat Level-1 Product Generation System (LPGS) uses separate algorithms for determining clouds, cirrus clouds, and snow and/or ice probability on a per-pixel basis. With more bands onboard the Landsat 8 Operational Land Imager (OLI)/Thermal Infrared Sensor (TIRS) satellite, and a greater number of cloud masking algorithms, the U.S. Geological Survey (USGS) is replacing the current cloud masking workflow with a more robust algorithm that is capable of working across multiple Landsat sensors with minimal modification. Because of the inherent error from stray light and intermittent data availability of TIRS, these algorithms need to operate both with and without thermal data. In this study, we created a workflow to evaluate cloud and cloud shadow masking algorithms using cloud validation masks manually derived from both Landsat 7 Enhanced Thematic Mapper Plus (ETM +) and Landsat 8 OLI/TIRS data. We created a new validation dataset consisting of 96 Landsat 8 scenes, representing different biomes and proportions of cloud cover. We evaluated algorithm performance by overall accuracy, omission error, and commission error for both cloud and cloud shadow. We found that CFMask, C code based on the Function of Mask (Fmask) algorithm, and its confidence bands have the best overall accuracy among the many algorithms tested using our validation data. The Artificial Thermal-Automated Cloud Cover Algorithm (AT-ACCA) is the most accurate nonthermal-based algorithm. We give preference to CFMask for operational cloud and cloud shadow detection, as it is derived from a priori knowledge of physical phenomena and is operable without geographic restriction, making it useful for current and future land imaging missions without having to be retrained in a machine-learning environment.

Explore related subjects

Remote Sensing of Environment Article

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Steven Curtis Foga, Pat Scaramuzza, Song Guo, Zhe Zhu, Ronald Dilley, Tim Beckmann, Gail L. Schmidt, John L. Dwyer, MJ Hughes, Brady Laue. 2017. Cloud detection algorithm comparison and validation for operational Landsat data products. https://doi.org/10.1016/j.rse.2017.03.026

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2026

On-demand global Landsat evapotranspiration product: Development, evaluation, and dissemination

Global actual evapotranspiration (ET) is one of the essential climate variables needed to understand and manage the relationships among food, energy, and water resources. The U.S. Geological Survey Earth Resources Observation and Science (EROS) Center launched a provisional ET product in 2020, offering on-demand, field-scale global coverage derived from Landsat data through the EROS Science Processing Architecture (ESPA) platform. The ESPA interface provides ET data for cloud-free Landsat overpasses starting in 1982 with Landsat 4 through the current Landsat 9. The ET data are delivered as a Provisional Level-3 Science product created using the Operational Simplified Surface Energy Balance (SSEBop) model. Landsat surface temperature and reference ET are the main model drivers along with vegetation index and net radiation for model parameterization. A large volume of Landsat-based ET orders (e.g., over 1,200,000 images from June 2020 through December 2025) around the world indicate increasing awareness and application of the ET data. The ESPA platform enables land and water resource managers and researchers to access a first-order ET product without requiring advanced knowledge of remote sensing technology or evapotranspiration modeling. We present the methodology and workflow of the on-demand Landsat ET product and its performance evaluations over diverse hydro-climatic settings. The product can help estimate field-scale consumptive water use and thus quickly and consistently assess historical water use, allocation, and budget to inform water management under changing environments. Future ET data aggregated to monthly and seasonal time scales are expected to enhance integration with decision-making tools and procedures.

Remote Sensing of Environment↗

USGS2026

A framework for integrating spatiotemporal deep learning methods with landsat for annual land cover and impervious surface mapping

Land cover information is essential for understanding Earth’s surface dynamics and how vegetation, water, soil, climate, and terrain interact. The National Land Cover Database (NLCD) has been the authoritative source for consistent U.S. land cover mapping. To extend NLCD’s temporal resolution and reduce production latency, we developed the Land Cover Artificial Mapping System (LCAMS)—a prototype spatiotemporal deep learning framework piloted as the foundation for the new Annual NLCD. LCAMS builds on concepts from legacy NLCD and the U.S. Geological Survey Land Change Monitoring, Assessment, and Projection (LCMAP) initiatives. It employs a loosely coupled two-stage architecture consisting of independent but functionally interdependent spatial and temporal models. Spatial models extract per-year information from Landsat data, while the temporal models refine the spatial outputs to enforce inter-annual consistency—critical for reliable land change monitoring. LCAMS produces annual 30 m resolution land cover and impervious surface outputs, with region-specific fine-tuning to generalize across diverse landscapes and temporal dynamics. Validation was conducted using an independent dataset of 1925 randomly sampled plots from five U.S. Landsat Analysis Ready Data (ARD) tiles spanning 1985-2021, selected for spatial and temporal variability. This dataset was used consistently to evaluate LCAMS, Legacy NLCD, and LCMAP. Using the NLCD legend, LCAMS achieved 72.1 ± 1.60% overall agreement, compared to 71.1 ± 1.7% agreement for Legacy NLCD. Using the LCMAP legend, LCAMS achieved 83.4 ± 1.22% agreement, compared to 84.6 ± 1.11% agreement for LCMAP. Overall, LCAMS delivers comparable accuracy while offering higher thematic resolution, longer temporal coverage, and automated production of annual 30 m CONUS land cover.

Remote Sensing of Environment↗

USGS2026

Towards global mapping of dynamic surface water extents using Sentinel-1 SAR data

We introduce a fully automated and scalable method for mapping surface water extents from single-acquisition Sentinel-1 synthetic aperture radar (SAR) imagery. This approach integrates adaptive thresholding of radiometric terrain-corrected SAR backscatter data, fuzzy-logic classification, region growing, dark land estimation, and a bimodality test to minimize false positives in low-backscattering areas and false negatives in high-backscattering areas. By combining these steps, the algorithm achieves classification accuracies exceeding 85% in detecting surface water extents across diverse environmental conditions. Accuracy was first assessed at meter scale using 52 PlanetScope scenes acquired worldwide in September–October 2019; the algorithm achieved 93% overall accuracy, 86% user's accuracy, and 94% producer's accuracy. Global robustness was then evaluated by processing every Sentinel-1 acquisition from 1 to 12 November 2023 and cross-comparing the resulting maps with 6561 temporally matched observational products for end-users from remote sensing analysis (OPERA) dynamic surface water extent from Harmonized Landsat and Sentinel-2 (DSWx-HLS) products. This large-scale test yielded 90% user's and 94% producer's accuracies, confirming reliable performance at continental extent. Additional case studies demonstrate the algorithm's ability to handle surface water extent in sand-dominated deserts, to track seasonal amplitude in Folsom Lake (California), drought-induced loss in Cerro Prieto Reservoir (Mexico), and rapid filling of the Grand Ethiopian Renaissance Dam. These results show that the method scales across local to global domains and maintains high accuracy, providing a practical tool for near-real-time monitoring of floods, droughts, and water-resource management. Because the approach is sensor-agnostic, it can be ported to forthcoming L- and S-band missions such as NASA-ISRO synthetic aperture radar (NISAR), broadening its applicability to future hydrologic observations.

Remote Sensing of Environment↗