Dataset Description
Due to the wide variety of landscape dynamics, crop types, growing seasons and agricultural management practices, mapping of cropland extent, crop types and irrigation practices at the global scale still remain very challenging tasks. Several attempts have already been made to come up with accurate global maps, but until now not one has succeeded in providing seasonal information at field level on a global scale. Current crop map layers either lack spatial detail or fail to provide regular updates. As such, the AFOLU community is still in need of a system that can provide seasonal global agricultural monitoring information at field level. For this reason, the European Space Agency, in collaboration with stakeholders in global agriculture like GEOGLAM, FAO, AMIS, has initiated the WorldCereal project to demonstrate the feasibility of producing seasonal update cropland and crop type maps based on open and free data.
The Open source European Space Agency (ESA) WorldCereal systems provides global seasonal updated maps for 2021 at 10 m resolution based on Copernicus Sentinel-1 and Sentinel-2 and Landsat 8 data. The WorldCereal system produces 4 hierarchical products. In Figure 1 the different WorldCereal products per season are shown, and below product definitions can be found.
Figure 1: Overview of WorldCereal products
Annual temporary crop map
This is the base product that is generated by the WorldCereal system. This is a binary map identifying land used for crops with a less-than-one-year growing cycle which must be newly sown or planted for further production after the harvest (FAO, 2023). Sugar cane, asparagus and cassava are also considered as temporary crops, despite the fact that they remain in the field for more than one year. The WorldCereal temporary crop maps exclude perennial crops as well as (temporary) pastures. These maps are generated once a year, the period being defined in a region by the end of the last growing season that is considered by the system.
Seasonal crop type maps (maize and cereals)
The WorldCereal crop type products provide binary maps for the maize and wheat growing seasons as defined by the global crop calendars, showing where maize and cereals are grown. Cereals include wheat, barley and rye, which belong to the Triticeae tribe. These crops were grouped together because their spectral signatures and growing seasons were too similar to reliably distinguish them at a global scale. The WorldCereal crop type maps are generated withing the respective annual temporary crop mask.
Seasonal irrigation maps
Seasonally actively irrigated cropland is therefore defined by the WorldCereal system as a piece of land that is extensively irrigated during a specific growing season where, without irrigation applied at regular intervals, crop growth would be significantly reduced or impossible. Incidental irrigation, such as irrigation that has been applied only during the sowing period of a crop, is not translated to actively irrigated cropland. A pixel can only be classified as being irrigated inside the annual temporary crop mask.
Seasonal active cropland maps
WorldCereal active cropland products indicate whether or not a pixel identified as temporary crops has been actively cultivated during each of the mapped growing seasons. In order for a pixel to be labeled as “active” during a particular growing season, a full crop growth cycle (sowing, growing, senescence and harvesting) needs to take place within the designated time period. Note that this active marker is not crop-type specific and will capture other crop types aside from cereals and maize as long as they show a similar growing season. This also means in practice that any crop grown (slightly) outside the predefined growing seasons will not be flagged as active cropland in any of the seasons covered by the system.
For COP26 these products will be demonstrated for 2021 at global scale. Products can be found at:
https://vdm.esa-worldcereal.org/
Usage
Users should keep in mind that the WorldCereal maps are datasets at global scale, generated with a single methodology applied over all regions. As such, the accuracy of the map may vary between locations and with scale. That said, the dataset can therefore be expected to be most useful for countries that do not have their own agricultural monitoring systems. In situ data on crop type and irrigation is of course very crucial for training the classification algorithms. This data is very fragmented, where in some areas a lot of information is available for several years and in other parts of the world no data is available. Although the methodology has gone through benchmarking and testing it is clear that areas with a lack of in situ data in the training database will have less accurate results than areas with sufficient in situ data. For all produced layers there is a confidence layer, to indicate areas where the model is less confident. Confidence layers of products can be found at
https://vdm.esa-worldcereal.org/
Methodology
The WorldCereal products are hierarchical in nature and are generated based on global crop calendars. In contrast to some other global products, the WorldCereal products are not produced for one single moment in time. Products are produced at the end of a certain growing season. A First step in the methodology was to define the growing seasons. This was done based on information coming from FAO, JRC-ASAP, Crop Monitor and NASA-Harvest, and where no information was available crop calendars were simulated (see Franch et al., 2022). These crop calendars were then stratified, based on start and end of season in 2023 WorldCereal production zones (AEZ) (see figure 2).
Figure 2: Global stratification based on crop calendar similarity. Each resulting zone serves as a WorldCereal map trigger to generate products based on local seasonality.
For each of these zones the WorldCereal products are created at the end of the growing season. As such, the global product is always a patchwork of products. First, the temporary crop map is calculated and all other products are linked to this map. Classification algorithms (CatBoost) for these products were trained based on the collected and harmonized in situ data sets. (in situ data can be found at
https://worldcereal-rdm.geo-wiki.org/)
From this data set its clear that distribution of in situ data is not homogenous over the world and that for crop type, irrigation and active cropland much less data is available. As such, the model trained for these products are trained on global data - whereas the temporary cropland maps are trained on stratified data.
Uncertainty and Accuracy
At the start of the WorldCereal project, minimum accuracies for the products were set for the different products. The overall product accuracy for cropland extent should be 80%, and for the croptype maps this should be 70%.
As mentioned in the data set section and methodology section, uncertainty and accuracy of the results is dependent upon the area of the World. Areas with large fields and where training data was available, will have higher accuracies than areas with small fields, complex seasonality and little to no training data. Accuracies are also product dependent, where cropland extent will normally have a higher accuracy than the irrigation product.
The validation approach differed from product to product, depending on quality and availability of reference data sets that were not used to train the models and hence available for independent validation. For the temporary crops, global validation results have an overall accuracy of 97.8% with a User accuracy for cropland of 88.5% and a producers accuracy for cropland of 92.1%. For the Crop type maps of maize at global scale, a user’s accuracy of 85.8% was reached and producers accuracy was 75.5%. For the Cereals, the figures are slightly higher with UA 93.6% and PA 77.7%.
Dataset Sustainment
The WorldCereal system is built upon the open and free data sets of the Copernicus Programme (Sentinel 1 and Sentinel 2) as well as Landsat 8. These data sets will be sustained by the different space agencies in the future.
Next to this, all algorithms and methods used in WorldCereal are open-source and the reference database will be further supported by GEOGLAM, through the GEOGLAM in situ data working group.
Technical Characteristics
Spatial resolution: 10m
Geographical coverage: Global
Temporal coverage: 2021-onwards (TBD)
Update frequency: Seasonal
Associated Guidance or User Manual
https://esa-worldcereal.org/en
Dataset Download
https://vdm.esa-worldcereal.org/
https://worldcereal-rdm.geo-wiki.org/
Points of contact for queries
Sven Gilliams
WorldCereal Project Manager
Flemish Institute for Technological Research (VITO)
Mol, Belgium
Email: sven.gilliams@vito.be
Zoltan Szantoi
WorldCereal Technical Officer
European Space Agency
ESA/ESRIN
Frascati, Italy
Email: Zoltan.szantoi@esa.int
