Automated Cropland Classification Algorithm (ACCA) for California using multi-sensor remote sensing

Zhuoting Wu; Prasad S. Thenkabail; James Verdin

doi:10.14358/pers.80.1.81

Home/Archive/USGS

USGS · 70098947

Automated Cropland Classification Algorithm (ACCA) for California using multi-sensor remote sensing

Zhuoting Wu·Prasad S. Thenkabail·James Verdin

View original source ↗Export metadata ↓

Abstract

Increasing pressure to feed the growing population with scarce water resources requires accurate and routine cropland mapping. This paper develops and implements a rule-based automated cropland classification algorithm (ACCA) using multi-sensor remote sensing data. Pixel-by-pixel accuracy assessments showed that ACCA produced an overall accuracy of 96 percent (K hat = 0.8) when tested using independent data layers. Furthermore, ACCA-generated county cropland areas showed high agreement (R-square values 0.94) when compared with three independent data sources: (a) US Department of Agriculture (USDA) cropland data layer derived cropland areas, (b) county specific crop acreage data from the Farm Service Agency, and (c) the Census of Agriculture data for the 58 counties in California. Our results demonstrate the ability of ACCA to generate cropland extent and areas over space and time, in an automated fashion with high degree of accuracies year after year, greatly contributing to food and water security analysis and decision making.

Explore related subjects

Photogrammetric Engineering & Remote Sensing Article

Source-reported bounding extent: 32.534156° to 42.009247° latitude; -124.409591° to -114.134768° longitude. This indicates report coverage, not an exact sampling location. View area on OpenStreetMap.

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Zhuoting Wu, Prasad S. Thenkabail, James Verdin. 2014. Automated Cropland Classification Algorithm (ACCA) for California using multi-sensor remote sensing. https://doi.org/10.14358/pers.80.1.81

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

KEEP EXPLORING

Related USGS reports

USGS2025

Landsat-derived rainfed and irrigated-area product for conterminous United States for the year 2020 (LRIP30 CONUS 2020) using supervised and unsupervised machine learning on the cloud

Accurate maps of irrigated and rainfed croplands are crucial for assessing global food and water security. Irrigated croplands yield two to four times more grain and biomass than rainfed croplands. To meet rising food demand, the proportion of cropland that is irrigated must be increased globally. Because agriculture uses 80% to 90% of global fresh water, understanding changes in cropland extent, crop type, and irrigation is critical for meeting nutritional needs sustainably. The United States has one of the most productive rainfed and irrigated croplands in the world and is a leading producer and exporter of agricultural crops. Precise maps of irrigated and rainfed croplands in the United States are crucial for assessing the current and the future agricultural production capacity in supporting food security. We developed a 30-m resolution rainfed and irrigated area map for the conterminous United States derived from 2019 to 2021 multi-date Landsat-8 data (LRIP30 CONUS 2020). A total of 96 harmonized spectral bands comprising monthly median value composites of eight bands (blue, green, red, NIR, SWIR1, SWIR2, TIR, and enhanced vegetation index [EVI]) were used. A cropland mask was then applied, and reference data were sourced from various sources. A pixel based supervised random forest classifier, and pixel based unsupervised ISODATA clustering classifier were implemented on Google Earth Engine and the ERDAS Imagine workstation to classify, identify, map, and assess accuracies of irrigated and rainfed cropland areas. The LRIP30 CONUS 2020 product achieved an overall accuracy of 93.9%. The irrigated and rainfed classes had producer's accuracies of 90.2% and 95.7%, respectively, and user's accuracies of 90.8% and 95.4%, respectively. The total net cropland area was estimated at 139.4 million hectares (Mha), of which 94.9 Mha (68%) was classified as rainfed and 44.5 Mha (32%) was classified as irrigated. State level summaries highlight regional differences and their implications for national and global food and water security.

conterminous United States↗

USGS2025

Thirty years of the U.S. National Land Cover Database: Impacts and future direction

The National Land Cover Database (NLCD), developed through the Multi-Resolution Land Characteristics Consortium, was initiated 30 years ago and has continually provided critical, Landsat-based landcover and land-change information for the United States. Originally launched to address the lack of national-scale, moderate-resolution land-cover data, NLCD has evolved from the pioneering 1992 dataset into a comprehensive, annually updated product suite. Key innovations include the introduction of impervious surface mapping, forest canopy mapping, standardized Landsat mosaics, national-scale accuracy assessments, continual evolution of deep learning and artificial intelligence methodologies, and a transition toward operational, change-focused monitoring. The NLCD has become an essential resource for scientific research, land management, and policy development, with extensive adoption across federal, state, and local agencies; academia; and the private sector. The NLCD data underpin a wide array of applications, including biodiversity conservation, urban planning, hydrology, human health studies, and natural hazard assessment. As new global and high-resolution commercial land-cover products emerge, the NLCD continues to distinguish itself through its temporal depth, federal backing, and thematic consistency. Moving forward, the NLCD will maintain its niche as the leading, moderate-resolution, long-term land-cover and land-change dataset for the United States, ensuring continued support for broad national applications while complementing higher-resolution and global-mapping efforts.

Photogrammetric Engineering & Remote Sensing↗

USGS2022

New generation hyperspectral sensors DESIS and PRISMA provide improved agricultural crop classifications

Using new remote sensing technology to study agricultural crops will support advances in food and water security. The recently launched, new generation spaceborne hyperspectral sensors, German DLR Earth Sensing Imaging Spectrometer (DESIS) and Italian PRecursore IperSpettrale della Missione Applicativa (PRISMA), provide unprecedented data in hundreds of narrow spectral bands for the study of the Earth. Therefore, our overarching goal in this study was to use these data to explore advances that can be made in agricultural research. We selected PRISMA and DESIS images during the 2020 growing season in California's Central Valley to study seven major crops. PRISMA and DESIS images were highly correlated (R 2of 0.9–0.95). Out of the 235 DESIS bands (400–1000 nm) and 238 PRISMA bands (400–2500 nm), 26 (11%) and 45 (19%) bands, respectively, were optimal to study agricultural crops. These optimal bands provided crop type classification accuracies of 83–90%. Hyperspectral vegetation indices to estimate plant pigment content, stress, biomass, moisture, and cellulose/lignin content were also identified.

California↗