Crop type maps identify the spatial distribution of crop types and underpin a large range of agricultural monitoring applications ranging from early warning of crop shortfalls, crop condition assessments, production forecasts, and damage assessment from extreme weather, to agricultural statistics, agricultural insurance, and climate mitigation and adaptation decisions. Despite their importance, harmonized, up-to-date global crop type maps of the main food commodities do not exist to date. To address this critical data gap of global-scale consistent, up-to-date crop type maps, we harmonized 24 national and regional datasets from 21 sources covering 66 countries to develop a set of Best Available Crop Specific masks (BACS) over the major production and export countries for wheat, maize, rice, and soybeans, in the context of the G20 Global Agriculture Monitoring Program, GEOGLAM.
Background & Summary
Ensuring food security is one of the major challenges we face in this century, especially in the face of a changing climate and a growing global population. With a rapidly growing demand for food, increasing conflicts, a highly interconnected global market, and increasingly extreme weather events, timely and accurate projections and estimates of global crop production are more important than ever1. Such estimates are a key component for well-functioning agricultural commodity markets and early warning and mitigation systems. One of the key international activities in support of transparent agricultural markets is the Group on Earth Observations Global Agriculture Monitoring (GEOGLAM) Crop Monitor for the G20 Agricultural Market Information System (AMIS) which provides a public good of open, timely, science-driven information on global crop conditions2. The AMIS and GEOGLAM initiatives were launched by the G20 Ministers of Agriculture following the food price crises in 2007/08 and 20103. While the GEOGLAM initiative is focused on enhancing crop monitoring capabilities, in support of policies, investments, and decisions in food security and agricultural markets using satellite and in situ Earth observations (EO), AMIS provides an inter-agency platform of economists and policymakers who work together to enhance food market transparency and policy response for food security. Bringing together the principal trading countries of agricultural commodities, AMIS assesses global food supplies (focusing on wheat, maize, rice, and soybeans) and provides a platform to coordinate policy action in times of market uncertainty.
In support of these activities, AMIS requested that GEOGLAM develop monthly crop condition assessments likely to impact production for these four main commodity crops. Foundational in providing such information is the identification of where each crop of interest is growing. Together with crop calendars, crop type maps enable the extraction of crop specific signals from satellite data during the agricultural growing season that can track crop development through the season and forecast yields ahead of harvest4,5. Despite their high value for trade and food security assessments, within-season maps at a sufficiently granular resolution to enable field to global-scale analyses of crop condition and crop yield do not exist across all of the world’s agricultural areas6,7. While for years this dearth was owing at least in part to insufficient satellite data and limits on computational processing8,9, today the principal challenges are the lack of high-quality ground reference data for calibration and validation of crop classifications10,11. Nevertheless, a range of crop type map products derived from satellite imagery does exist at national and regional scales (e.g.12,13). In addition, at the global scale, there are products such as the IFPRI SPAM-201014, M3-Crops15, and MIRCA200016 that provide information on crop type distribution based on sub-national statistics and a spatial allocation model at the 10Km resolution. While these represent the current state of the art for global crop type distribution, they are based on spatial models and subnational statistics rather than the spectral signal of a crop and they are at a very coarse resolution (10 km) and are out of date (i.e. represent croplands circa 2010). In short, they may represent the national or subnational total land area of each crop, though the spatial location of crops may not be correct, which presents a critical issue for their application in masking for within-season crop monitoring.
To meet the needs of the GEOGLAM Crop Monitor to accurately mask crop type with as up-to-date information as available, we developed a harmonized global set of crop specific maps for the four major grains (wheat, maize, rice, and soybeans) following an exhaustive identification and collection of the most recent, highest quality existing crop type maps at national and regional sources. Similar to the efforts by Fritz et al.17 and Waldner et al.18 that created a unified general cropland product based on existing cropland products, we designed a criteria system to assess the best data sets with regards to timeliness, accuracy, spatial resolution, and data source. The result is the first set of global crop type maps, at the 0.05 degree resolution, derived from satellite imagery, covering the major producer and export countries for the four main crops, referred to herein as the GEOGLAM Global Best Available Crop Specific Masks (GEOGLAM-BACS). These maps are used operationally within the GEOGLAM Crop Monitor in the creation of monthly global crop condition assessments and are updated on an annual basis as new crop type maps become available. The dataset is made publicly available with this publication on CropMonitor.org as well as on Zenodo and at the time of submission refers to version v.1.0.
Several steps were required for developing the GEOGLAM-BACS. These included the following, which are further detailed below:
Dataset collection and selection of a base layer
Criteria evaluation and scoring
Product selection based on scores
Generation of a unified map (updated as new data become available)
Data set selection
The Group on Earth Observation Global Agriculture Monitoring (GEOGLAM) required global crop-specific masks to provide accurate crop assessments for the primary crops (maize, wheat, soybean, and rice) that AMIS (Agricultural Market Information System) is focused on. In response to this requirement and acknowledging the need to ensure the spatial fidelity of crop locations while building upon years of community product development, we undertook a study of existing global, regional, and national crop-specific map products through exchanges with the various national and regional focal points that contribute to the Crop Monitor for AMIS2, the GEOGLAM expert networks, and literature reviews. Three gridded products based on different spatial allocation models and subnational statistics were identified at the global scale (SPAM, MIRCA, and M3), but the Spatial Allocation Production Model 2010 (SPAM-2010 physical area layer was selected as the foundation because it provides a gridded product containing the physical area of the GEOGLAM target crops at a spatial resolution of 10 × 10 km19,20. However, the SPAM-2010 physical area crop mask refers to data from 2010 or thereabout, rendering it out of date. Further, it is based on a statistical allocation model that evenly distributes the crop physical area density across croplands based upon the administrative level for which the statistics were obtained, and therefore can result in inaccurate locations of crops. Given the vital importance of accurate cropping locations, solely relying upon SPAM-2010 to mask current croplands is insufficient for remote sensing-based analyses.
Due to these features coupled with the ongoing mandate of the GEOGLAM Crop Monitor to report on current-season crop conditions, we needed to derive an updated product based on an updateable method for generating timely and accurate maps of major crops. The crop masks incorporated into GEOGLAM-BACS masks were either produced by GEOGLAM partners from moderate-resolution (10–250 meter) satellite data or generated by Agricultural Ministries or Councils of some countries (Table 8, Fig. 1). They therefore, represent the most up-to-date crop masks available at the time of submission. Additionally, the GEOGLAM-BACS masks used national crop production information and expert knowledge to refine the crop area of some countries, bringing high-value expert input from many corners of the globe into a unified product. The top five global producers of each commodity (Table 1) were prioritized as those countries account for between 82–97% of global crop production. The GEOGLAM-BACS integrates 16 out of the 20 crop-type masks for these top producers.
Several criteria were used to determine a product’s fitness for inclusion within GEOGLAM-BACS and are described below. In countries where no specific crop type maps were available, the SPAM 2010 data remains.
Selection criteria and scoring
Similar to the approach described by Waldner et al.21, five criteria were considered to guide the selection and evaluation of the crop specific maps. These consisted of seasonal specificity (SS), timeliness (Ti), spatial resolution (SR), accuracy (Ac), and data source (DS) (Tables 2–6). Each criterion has a range of scores with 4 or 3 as the highest score (depending on the criterion) and 1 as the lowest.
Seasonal specificity criteria
A key element for in-season crop assessments and yield forecasting is accounting for multiple growing seasons. As such, a requirement for the GEOGLAM-BACS was to account separately for winter wheat and spring wheat, for the first and second maize seasons, for the first and second soybean seasons, and for multiple rice growing seasons (up to three seasons). The Seasonal Specificity (SS) Criteria was designed to capture this information, where a score of 3 indicates all seasons captured, a score of 2 indicates that one of the multiple seasons is captured, and a score of 1 means all seasons are aggregated into 1 mapped layer (i.e. winter wheat and spring wheat are not captured separately, but rather are aggregated into one class of wheat) (Table 2).
Timeliness (Ti) criterion
In most growing regions of the world, croplands are dynamic with changes in planting occurring from one season to the next due to crop rotation, weather, agricultural markets, and policies. As such, a within-season crop-specific map at field scale would be ideal6,22,23,24,25. While there is a large push from the community to produce such products at scale (i.e.10,11), to date, such products are not widely available. Nevertheless, broadly speaking a recent season is generally more likely to represent crop specific extent relative to older products, especially at coarse resolution per pixel percent crop type maps such as the BACS product22. As such, the timeliness criterion is designed to give the highest score (4) to within-season products and the lowest score (1) to the out-of-date products defined as older than 5 years (Table 3). It is worth noting here that none of the crop products in BACS had a score of 4.
Spatial resolution criterion
A range of studies have characterized in detail the spatial resolution requirements for accurately mapping crop types, which account not only for field size but also account for field shapes, crop diversity, and overall landscape complexity26,27,28. However, for the GEOGLAM-BACS products, we took a simplified approach and considered that the higher spatial resolution products would generally be able to characterize crop types better than the coarser resolution products. While we acknowledge that higher spatial resolution does not always translate to higher accuracy, they do often perform better when produced with local training and validation datasets which is the case for the higher-resolution products integrated into the GEOGLAM-BACS dataset.
As such, the higher-resolution products (5–20 m) received the highest score (4) and the lowest-resolution products (>500 m) received the lowest score (1) (Table 4). The four resolution classes were defined based on spatial resolution requirements as articulated by the GEOGLAM community and defined in9.
The accuracy criterion scoring was based on the accuracy metrics available for each of the crop type products (Table 5). Where not available, scoring is based on an assessment against sub-national level published statistics. This criterion was based on four categories, with the above 90% reported overall accuracy score receiving a score of 4 and overall accuracy of less than 80% received a score of 2. Several of the products did not report accuracy at all, or did not report accuracy for a specific crop type or country, and therefore where possible we approximated the accuracy of these products based on a comparison with subnational-level official statistics. We recognize the issues with this approach, but given that we did not have any other data on their accuracy this was an acceptable approximation for this purpose. Where the accuracy of these products could not be discerned for these purposes, they received the lowest score (1).
Data source criterion
This criterion aims to capture the relative specificity of the products for the countries they cover. Generally speaking, products that are developed at a sub-national to national scale are based on models and ground data collected for that specific country and are specialized for that country. At the regional level, the mapped products and models are usually more generalized to map larger regions, and generally speaking have less national-specific data, and lower accuracies at the national scale. Likewise, products that are at the global scale, often rely on generalizations developed in locations with ground data and often extrapolate to map regions with little or no ground data. While great progress is being made in developing robust and scalable models utilizing sparse ground data sets10,11,29,30,31,32 still, for the most part, the sub-national and national-level products tend to have higher accuracies than do regional and global data. As such in this criterion, the national products received the highest score (3) and the global products received the lowest score (1) (Table 6).
Scoring & product selection
The scores for each criterion were then aggregated to compute the fitness indicator (FI) as follows:
Where FI is the Fitness Indicator, SS is the seasonal specificity score, Ti is the timeliness score, SR is the spatial resolution score and Ac is the accuracy score, and DS is the data source score (Table 7).
Overview of scores
The highest score possible accounting for the highest rank in each category is 18. While none of the products received this score (as none are in-season products at 10-meter resolution), several products received high scores, with eleven products scoring between 16 and 14, twelve products scoring between 13 and 11, one scoring 9, and one scoring 5 (the initial base map) (Fig. 2). This highlights one of the most pressing current gaps in crop specific maps, which is a lack of in-season crop type masks that are especially crucial for tracking specific crop development, input deficiencies, yield forecasting, planted area, etc. during the growing season itself. While many of the products were recent (within the past 1 to 3 years), a large portion of the major staple crop-producing regions of the world do not have publicly available recent crop type maps, notable countries are India and China, two of the world’s largest crop producers and consumers. An area of low scores is for SS in countries where there is more than one season per year for the same crop such as summer vs. spring-planted maize or winter vs. spring wheat. An important example of this is the differentiation between spring wheat and winter wheat in Russia or early and late planted soybeans in Argentina.
Generation of unified map
As the collection of national, regional, and global crop-specific masks are at a variety of spatial resolutions, all the masks were harmonized to a common resolution. The 0.05-degree resolution of the climate modeling grid (CMG) MODIS products was selected for the standardized product. This resolution was chosen to match the wide variety of climate datasets available at the CMG level for agricultural monitoring. Additionally, it has been shown that when detailed in-season crop masks are not available over areas with crop rotations, a coarse resolution mask, such as the 0.05-degree CMG, can be used to accurately monitor and yield forecast for specific crops using Normalized Difference Vegetation Index (NDVI) time-series data22.
The crop-specific SPAM 2010 products were converted to percent per crop, per pixel, and then resampled to CMG level from the original 10 × 10 km grid. Since winter and spring wheat have different growing calendars and require detailed monitoring during different periods, the SPAM-2010 v2 wheat layer was split out into separate seasonal masks. In locations where seasonal-specific crop masks could be obtained (Canada, United States, Russia winter wheat, and Kazakhstan spring wheat), they were used. For the remaining locations and seasons (Russia spring wheat, China winter and spring wheat, and Kazakhstan winter wheat), sub-national sown area statistics were used to identify the percent of the SPAM-2010 v2 wheat layer associated with each season. This was done by incrementally reducing the percent per pixel value across each sub-national region uniformly until the area of the crop mask within the sub-national region matched the reported sown area for that region.
The crop-specific masks used in assembling the GEOGLAM-BACS masks came in a variety of formats and spatial resolutions. The steps used to derive the final standardized format and spatial resolution vary depending on the source. Generally, each of the designated crops was separated and converted over to a raster grid at its native spatial resolution. Then each crop type gridded layer was converted to a binary grid of “1” for the crop and “0” for no-crop. Using the CMG grid, each CMG cell was then overlaid on top of the crop-specific raster where the number of crop cells was divided by the total number of cells covered by the CMG cell to get the percentage of the crop within per pixel. The final product was then assembled by replacing the data from within the resampled SPAM-2010 product where the national and regional products were found to be available (Figs. 3, 4).
The GEOGLAM-BACS masks are a collection of five global Geotiff files containing flat rasters at 0.05 degrees, one for each of the AMIS crops (Maize, Soybean, Rice, Winter Wheat, and Spring Wheat). A comprehensive metadata file accompanies the rasters. Values represent the percent of the specific crop per pixel, ranging from 0 to 100%. Within the rasters, values were converted to integers whereby the values are from 0 to 10,000, where 10,000 is equal to 100.00%. This was done to preserve small variations across the landscape of less than 1% while reducing the size of the dataset. The GEOGLAM-BACS will be updated regularly as new and more recent products become available. The updated rasters and metadata files will be pushed to the Zenodo repository as well as hosted on www.CropMonitor.org.
For those countries and regions that have had the original SPAM 2010 v2 data replaced, Table 8 identifies the specific products used. The data is accessible on the Zenodo platform33: https://zenodo.org/record/7230863#.Y1F3mnZKh9M.
The majority of the crop specific masks used in the creation of the GEOGLAM-BACS have associated validation and accuracy metrics that are publicly available (Table 9). Similar to work done by Pittman et al.34 and Pérez-Hoyos et al.35, the crop masks that do not have publicly available validation information were validated against reported sub-national sown area statistics for the specific reference year (Table 10). While we recognize there are many limitations to such an approach to validation, this was the only feasible way for providing some validation of these products. The comparisons against sub-national sown area statistics are provided in Table 10.
Known issues & required improvements
GEOGLAM-BACS, as indicated by its name, represents the best-available, most up-to-date globally harmonized dataset on the locations of the four major commodity crops wheat, maize, rice, and soybean. Still, there remain gaps as indicated by this technical validation, and enumerated briefly below:
While we have made our best effort to validate GEOGLAM-BACS, in many areas validation data for specific locations simply does not exist. All crop monitoring activities, including that of the GEOGLAM Crop Monitor, would be strengthened with quantitative validation that would empower more precise and risk-informed decision-making36. Critical to this effort is expanding the collection and availability of ground data for training and validation. Advances have been made in generating and interpreting labeled datasets to assist in these efforts (e.g.31,32,37), and should be further developed, particularly in data-sparse regions which often correlate with food insecurity38.
The GEOGLAM-BACS are not currently produced in-season. This is a major priority for monitoring crops as they develop throughout the growing season, yield forecasting, producing within-season cropped area estimates, damage estimates from extreme weather, conflict, etc. Relying on prior year information is the state of the science at global scale22, however with advances in public and commercial satellite missions, that together deliver sub-weekly or even daily observations at 3–30 m (e.g. Landsat 8, Landsat 9, Sentinel1 & 2, Planet Doves;9,39, and advances in cloud compute, ML, and digital data collection, within-season crop type mapping is within sight and has been articulated as a priority for GEOGLAM Essential Agriculture Variables (EAVs)25 and the European Space Agency’s WorldCereal project29,40.
In areas of the world where multiple seasons of the same crop exist – for example, maize and rice – in some cases seasonal specificity is lacking in the current GEOGLAM-BACS. Complementing crop type mapping with crop type calendar information is an important research area(e.g.62, again prioritized by the GEOGLAM EAVs.
There remain areas for which no recent satellite based crop type data exists, and the product relies on SPAM 2010’s statistical allocation model. It is unclear whether these locations are representative of true crop locations, and merit prioritization in future satellite-mapping activities.
Of particular importance to human livelihoods is the improvement of information in areas with acute food insecurity and/or characterized by smallholder agriculture. This requires fine spatial resolution data, which as mentioned is now available between public and private optical and synthetic aperture radar missions, as well as appropriate ground data, which is often the primary bottle neck.
As demonstrated by GEOGLAM-BACS, one of the most powerful tools in advancing global agricultural monitoring is international collaboration. The GEOGLAM-BACS is a public good that would not be possible without the inputs and substasive contributions of GEOGLAM partners around the world.
The primary use of this dataset is for underpinning EO-based, global, crop-specific condition monitoring for four of the major crop commodities in the major production/exporting countries, through the GEOGLAM Crop Monitor activity. As noted in the prior section (Technical Validation: Known Issues & Required Improvements), there is room for improvement concerning validation, extent, resolution, and seasonal specificity, and these limitations should be noted in the application of these data in different monitoring activities. Nevertheless, the data are already being operationally used within the GEOGLAM Crop Monitor Reports.
No custom code was used to generate or process the GEOGLAM-BACS masks. The software used in the assembly of the masks was ArcMap version 10.6.
Rezaei, E. E., Ghazaryan, G., Moradi, R., Dubovyk, O. & Siebert, S. Crop harvested area, not yield, drives variability in crop production in Iran. Environ. Res. Lett. 16, 064058 (2021).
Becker-Reshef, I. et al. The GEOGLAM crop monitor for AMIS: Assessing crop conditions in the context of global markets. Global Food Security 23, 173–181 (2019).
Cannes Summit Final Declaration. G20 Information Centre (2011).
Hao, P., Tang, H., Chen, Z., Meng, Q. & Kang, Y. Early-season crop type mapping using 30-m reference time series. Journal of Integrative Agriculture 19, 1897–1911 (2020).
Whitcraft, A. K., Becker-Reshef, I. & Justice, C. O. Agricultural growing season calendars derived from MODIS surface reflectance. International Journal of Digital Earth 8, 173–197 (2014).
Fritz, S. et al. A comparison of global agricultural monitoring systems and current gaps. Agricultural Systems 168, 258–272 (2019).
Kim, K.-H., Doi, Y., Ramankutty, N. & Iizumi, T. A review of global gridded cropping system data products. Environ. Res. Lett. 16, 093005 (2021).
Torbick, N. et al. Fusion of Moderate Resolution Earth Observations for Operational Crop Type Mapping. Remote Sensing 10, 1058 (2018).
Whitcraft, A., Becker-Reshef, I., Killough, B. & Justice, C. Meeting Earth Observation Requirements for Global Agricultural Monitoring: An Evaluation of the Revisit Capabilities of Current and Planned Moderate Resolution Optical Earth Observing Missions. Remote Sensing 7, 1482–1503 (2015).
Kerner, H. et al. Resilient In-Season Crop Type Classification in Multispectral Satellite Observations using Growth Stage Normalization, https://doi.org/10.48550/ARXIV.2009.10189 (2020).
Wang, S., Azzari, G. & Lobell, D. B. Crop type mapping without field-level labels: Random forest transfer and unsupervised clustering techniques. Remote Sensing of Environment 222, 303–317 (2019).
Johnson, D. M. & Mueller, R. Pre- and within-season crop type classification trained with archival land cover information. Remote Sensing of Environment 264, 112576 (2021).
Kussul, N., Lavreniuk, M., Shelestov, A. & Yailymov, B. Along the season crop classification in Ukraine based on time series of optical and SAR images using ensemble of neural network classifiers. in 2016 IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 7145–7148, https://doi.org/10.1109/IGARSS.2016.7730864 (IEEE, 2016).
You, L., Wood, S., Wood-Sichra, U. & Wu, W. Generating global crop distribution maps: From census to grid. Agricultural Systems 127, 53–60 (2014).
Monfreda, C., Ramankutty, N. & Foley, J. A. Farming the planet: 2. Geographic distribution of crop areas, yields, physiological types, and net primary production in the year 2000: GLOBAL CROP AREAS AND YIELDS IN 2000. Global Biogeochem. Cycles 22, n/a–n/a (2008).
Portmann, F. T., Siebert, S. & Döll, P. MIRCA2000-Global monthly irrigated and rainfed crop areas around the year 2000: A new high-resolution data set for agricultural and hydrological modeling: MONTHLY IRRIGATED AND RAINFED CROP AREAS. Global Biogeochem. Cycles 24, n/a–n/a (2010).
Fritz, S. et al. Mapping global cropland and field size. Glob Change Biol 21, 1980–1992 (2015).
Waldner, F. et al. A Unified Cropland Layer at 250 m for Global Agriculture Monitoring. Data 1, 3 (2016).
Wood-Sichra, U., Joglekar, A. B. & You, L. Spatial Production Allocation Model (SPAM) 2005: Technical Documentation. 87 (2016).
Yu, Q. et al. A cultivated planet in 2010 – Part 2: The global gridded agricultural-production maps. Earth Syst. Sci. Data 12, 3545–3572 (2020).
Waldner, F., Fritz, S., Di Gregorio, A. & Defourny, P. Mapping Priorities to Focus Cropland Mapping Activities: Fitness Assessment of Existing Global, Regional and National Cropland Maps. Remote Sensing 7, 7959–7986 (2015).
Becker-Reshef, I. et al. Prior Season Crop Type Masks for Winter Wheat Yield Forecasting: A US Case Study. Remote Sensing 10, 1659 (2018).
Defourny, P. et al. Near real-time agriculture monitoring at national scale at parcel resolution: Performance assessment of the Sen2-Agri automated system in various cropping systems around the world. Remote Sensing of Environment 221, 551–568 (2019).
Whitcraft, A., Becker-Reshef, I. & Justice, C. A Framework for Defining Spatially Explicit Earth Observation Requirements for a Global Agricultural Monitoring Initiative (GEOGLAM). Remote Sensing 7, 1461–1481 (2015).
Whitcraft, A. K. et al. No pixel left behind: Toward integrating Earth Observations for agriculture into the United Nations Sustainable Development Goals framework. Remote Sensing of Environment 235, 111470 (2019).
Duveiller, G. & Defourny, P. A conceptual framework to define the spatial resolution requirements for agricultural monitoring using remote sensing. Remote Sensing of Environment 114, 2637–2650 (2010).
Leroux, L., Jolivot, A., Bégué, A., Seen, D. & Zoungrana, B. How Reliable is the MODIS Land Cover Product for Crop Mapping Sub-Saharan Agricultural Landscapes? Remote Sensing 6, 8541–8564 (2014).
Löw, F. & Duveiller, G. Defining the Spatial Resolution Requirements for Crop Identification Using Optical Remote Sensing. Remote Sensing 6, 9034–9063 (2014).
Song, C., Lin, Y., Guo, S. & Wan, H. Spatial-Temporal Synchronous Graph Convolutional Networks: A New Framework for Spatial-Temporal Network Data Forecasting. AAAI 34, 914–921 (2020).
Tseng, G., Kerner, H., Nakalembe, C. & Becker-Reshef, I. Kenya Cropland Map and Non-Crop Labelled dataset https://doi.org/10.5281/ZENODO.4271143 (2020).
Tseng, G., Kerner, H., Nakalembe, C. & Becker-Reshef, I. Learning to predict crop type from heterogeneous sparse labels using meta-learning. in 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) 1111–1120, https://doi.org/10.1109/CVPRW53098.2021.00122 (IEEE, 2021).
Becker-Reshef, I. et al. GEOGLAM Best Available Crop Type Masks. Zenodo https://doi.org/10.5281/ZENODO.6511594 (2022).
Pittman, K., Hansen, M. C., Becker-Reshef, I., Potapov, P. V. & Justice, C. O. Estimating Global Cropland Extent with Multi-year MODIS Data. Remote Sensing 2, 1844–1863 (2010).
Pérez-Hoyos, A., Rembold, F., Kerdiles, H. & Gallego, J. Comparison of Global Land Cover Datasets for Cropland Monitoring. Remote Sensing 9, 1118 (2017).
Weiss, M., Jacob, F. & Duveiller, G. Remote sensing for agricultural applications: A meta-review. Remote Sensing of Environment 236, 111402 (2020).
ESA WorldCereal Data Collection. ESA WorldCereal.
Nakalembe, C. L., Kerner, H. R. & Becker-Reshef, I. Urgent and critical need for developing countries to invest in Earth observation-based agricultural early warning and monitoring systems to achieve Zero Hunger (SDG-2). in vol. 2020 SY025-02 (2020).
Claverie, M. et al. The Harmonized Landsat and Sentinel-2 surface reflectance data set. Remote Sensing of Environment 219, 145–161 (2018).
Cintas R., J. et al. Generating Winter Wheat Global Crop Calendars in the Framework of Worldcereal. in 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS 6583–6586, https://doi.org/10.1109/IGARSS47720.2021.9553083 (IEEE, 2021).
National Map of Crops 2018/2019 campaign. GeoINTA.
Skakun, S. et al. Early season large-area winter crop mapping using MODIS NDVI data, growing degree days information and a Gaussian mixture model. Remote Sensing of Environment 195, 244–258 (2017).
Xiao, X. et al. Mapping paddy rice agriculture in South and Southeast Asia using multi-temporal MODIS images. Remote Sensing of Environment 100, 95–113 (2006).
Gumma, M. K., Thenkabail, P. S., Maunahan, A., Islam, S. & Nelson, A. Mapping seasonal rice cropland extent and area in the high cropping intensity environment of Bangladesh using MODIS 500m data for the year 2010. ISPRS Journal of Photogrammetry and Remote Sensing 91, 98–113 (2014).
Gumma, M. K., Gauchan, D., Nelson, A., Pandey, S. & Rala, A. Temporal changes in rice-growing area and their impact on livelihood over a decade: A case study of Nepal. Agriculture, Ecosystems & Environment 142, 382–392 (2011).
Gumma, M. K. Mapping rice areas of South Asia using MODIS multitemporal data. J. Appl. Remote Sens 5, 053547 (2011).
INVEKOS Schläge Austria 2019. Offene Daten Österreich.
Agricultural use plots ALV 2016, Anonymous agricultural plot (2018) (PAA). Geopunt Vlaanderen.
Conab summer crop areas. Companhia Nacional de abastecimiento do Brazil (Conab).
King, L. et al. A multi-resolution approach to national-scale cultivated area estimation of soybean. Remote Sensing of Environment 195, 13–29 (2017).
Chang, J., Hansen, M. C., Pittman, K., Carroll, M. & DiMiceli, C. Corn and Soybean Mapping in the United States Using MODIS Time‐Series Data Sets. Agron. j. 99, 1654–1664 (2007).
AAFC Canada Annual Crop Inventory. Agriculture and Agri-Food Canada (AAFC).
Crop Map of England (CROME) 2019. data.gov.uk, (2021).
d’Andrimont, R. et al. From parcel to continental scale – A first European crop type map based on Sentinel-1 and LUCAS Copernicus in-situ observations. Remote Sensing of Environment 266, 112708 (2021).
France Registre Parcellaire Graphique (RPG). République Française géoservices.
International Food Policy Research Institute. Global Spatially-Disaggregated Crop Production Statistics Data for 2010 Version 2.0 Harvard Dataverse, https://doi.org/10.7910/DVN/PRFF8V (2019).
Basic Registration of Crop Parcels (BRP) 2019. Geo services - PDOK.
Dempewolf, J. et al. Wheat Yield Forecasting for Punjab Province from Vegetation Index Time Series and Historic Crop Statistics. Remote Sensing 6, 9653–9675 (2014).
Integrated map of coverage/land use of Uruguay. Ministry of Livestock, Agriculture and Fisheries (2019).
USDA NASS CDL. USDA National Agricultural Statistics Service.
CropScape - Cropland Data Layer. USDA NASS CDL Program.
Franch, B., Cintas, J., Becker-Reshef, I., Sanchez-Torres, M. J., Roger, J., Skakun, S. & Whitcraft, A. (2022). Global crop calendars of maize and wheat in the framework of the WorldCereal project. GIScience & Remote Sensing, 59(1), 885-913.
The authors would like to thank the GEOGLAM Crop Monitor partner organizations and especially the participants that donated crop masks to the effort. Additionally, the authors would like to thank the NASA Water Resources Program for the support through grants 80NSSC17K0625 (NASA Harvest), NNX12AJ91G, NNX17AL29G, and NNX16AP16G, NNX17AH48G.
There are no known conflicts of interest associated with this article and there has been no significant financial support for this work that could have influenced its outcome. The primary source of external funding for this work is from the NASA Applied Sciences Program in addition to the in-kind contributions from all the crop monitor partner agencies.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Becker-Reshef, I., Barker, B., Whitcraft, A. et al. Crop Type Maps for Operational Global Agricultural Monitoring. Sci Data 10, 172 (2023). https://doi.org/10.1038/s41597-023-02047-9