WorldCereal: Empowering Global Crop Mapping with Custom Data and Cloud Tools

Custom Crop Mapping With Your Own Data

One of WorldCereal’s most exciting new features is the ability to create tailored crop type maps using your own field data. Every region and farming system is unique – now users can integrate their private in-situ reference data (alongside public datasets) to train crop classification models that are customised to local conditions. In practice, this means that an agronomist or remote sensing analyst can upload their own ground truth (for example, GPS-tagged crop observations or field survey data) into WorldCereal’s cloud platform and use it to generate crop maps specific to their area of interest. The system leverages state-of-the-art machine learning algorithms under the hood. However, you don’t need to be a data scientist to benefit, as WorldCereal handles the heavy lifting of model training and calibration automatically.

This custom mapping capability (see Figure 1) was introduced in a major platform update in March 2025. It represents a shift from one-size-fits-all global models to a more flexible approach, where users can define what crop types matter to them and obtain results tuned to their needs. You can still rely on WorldCereal’s default global models out-of-the-box, but now you have the option to improve on them by infusing local knowledge. For instance, a user in South America focusing on soybean and rice can train a model specifically for those crops and that region, using any available local reference data, and generate accurate maps that reflect the reality on the ground. This 'bring-your-own-data' approach opens the door for better crop monitoring in data-rich areas. With consent from the user, it can also help enhance the system’s overall knowledge base when contributed data are shared for broader model improvement.

Figure 1: Example of a custom crop type product for Ukraine produced using the WorldCereal system.

While the 2021 global products established the system’s baseline, users can now zoom in on their own regions – and even improve those maps with local data. These baseline maps demonstrate WorldCereal’s ability to consistently map cropland worldwide using satellite observations, providing a reference point against which customised results can be compared.

Not only does WorldCereal allow tailored model training, but it also provides an intuitive way to manage reference data. The Reference Data Module (RDM) is a companion tool where you can explore and download the publicly available in-situ datasets collected by the project (see tutorial), or upload and harmonise your own datasets via an easy web interface (see tutorial). All contributed data are kept secure – private data stays private to the user – but can be harmonised and used in the cloud processing workflow. By expanding this library of reference data (over 140 datasets and counting!), the community helps improve the default crop models as well. In fact, the WorldCereal team continuously updates the pre-trained global models as new reference data comes in, so every contribution makes the system more robust. This collaborative ethos of open science and user engagement is at the heart of WorldCereal’s mission. Would you like to learn more about reference data requirements for crop type mapping? Sign up for free to our Massive Open Online Course (MOOC).

A User-Friendly Processing Hub for Everyone

The WorldCereal Processing Hub marks a major step toward making crop mapping accessible to all. This cloud-based web portal lets users generate cropland extent or (soon) crop type maps by simply selecting a region, growing season, and map type—no coding or Python setup required. The system automatically pulls satellite data (e.g. Sentinel-1 and -2) and runs classification algorithms on the Copernicus Data Space Ecosystem (CDSE) cloud infrastructure (see Figure 2).

Originally limited to expert users via Python notebooks, WorldCereal’s workflows are now partially available through this intuitive web interface. While advanced users can still turn to the notebooks for full customisation, the Processing Hub offers a streamlined, user-friendly option with preconfigured settings. You can track job progress directly in your browser. As shown during our March 2025 webinar, even first-time users were able to create crop maps in real time. By lowering the technical barrier, the Hub brings powerful Earth observation tools within reach of a much broader audience. See the WorldCereal Processing Hub in action in this demo video.

To ensure transparency and build trust, WorldCereal also provides detailed documentation and examples. The open-source nature of the system means that advanced users and developers can dive into the code or tweak the Jupyter notebooks if they wish, but this is purely optional. Most users will find the out-of-the-box tools sufficient for their needs. By coupling a user-friendly interface with cloud computing muscle, WorldCereal bridges the gap between cutting-edge satellite technology and practical, everyday use by agronomists, policymakers, and even citizen scientists. It’s an invitation to anyone interested in crop monitoring to jump right in without needing a remote sensing background.

Figure 2: Overview of the WorldCereal Processing Module and its components. The system combines satellite observations, reference data, and cloud computing (through platforms like OpenEO and the Copernicus Data Space Ecosystem) to deliver crop maps via both code-based notebooks and the new web-based Processing Hub.

Powerful Algorithmic Core

At the heart of these powerful and customisable mapping capabilities lies the continued refinement of WorldCereal's algorithmic core. A key innovation is the integration of NASA Harvest’s Presto model, a lightweight deep learning feature extractor optimised for Earth Observation time series. Designed for efficiency and adaptability, Presto harnesses the power of self-supervised pretraining to learn generalisable spatiotemporal representations from unlabelled data. This greatly enhances WorldCereal’s ability to generate accurate predictions in downstream crop mapping tasks, even in regions where only limited labelled data are available. As a result, Presto enhances the system’s robustness and scalability across varying agricultural contexts.

Recent pilot projects have demonstrated its strong generalisation capabilities across diverse agro-ecological zones. Now a standard part of the crop mapping pipeline, Presto helps ensure consistent performance globally. This development was recently presented at the first ESA-NASA International Workshop on AI Foundation Models for Earth Observation, where the WorldCereal team shared their experience in deploying a geospatial foundation model within a real-world, global-scale operational system.

“By opening up our crop mapping algorithms to user input and making them easy to run, we’re turning WorldCereal into a community-driven tool. We can’t wait to see how users will apply it and help us improve it further.”

– Kristof Van Tricht, R&D Expert at VITO Remote Sensing and WorldCereal project lead

Explore WorldCereal and Get Involved

WorldCereal is publicly available and we encourage organisations and individuals to test the system, apply it to your use cases, and share your feedback. You can explore the public data products, experiment with custom mapping on the Processing Hub, and find documentation and tutorials on the WorldCereal website. A great way to connect with the WorldCereal community is through the online forum, where users can ask questions, report experiences, and discuss ideas. Early adopters are already exploring the possibilities: for example, an independent FAO analysis earlier this year highlighted the value of WorldCereal for global crop monitoring, and more use case studies are expected to emerge as access broadens.

Looking ahead, the WorldCereal roadmap is full of improvements driven by user needs. In upcoming releases, users can expect:

• Updates to the default models: higher accuracy and support for additional crop types, thanks to continuous learning from new data.
• Easier scaling of map generation: the ability to map larger areas in one go, leveraging integration with Europe’s new CDSE cloud infrastructure for on-demand processing.
• More harmonised reference data: an expanding pool of open in-situ datasets, further improving model performance and regional relevance.

Crucially, WorldCereal remains a community-oriented initiative. The project’s consortium is not only improving the technology but also investing in user engagement activities – from webinars and training to collaborative pilot projects – to ensure the platform truly serves its users. If you have in-situ datasets or expertise to share, you can contribute to WorldCereal’s open Reference Data Module and help enhance the system for everyone. Do you have ideas or requirements for new features? The team would love to hear from you. And subscribe to the newsletter if you don't want to miss any WorldCereal updates.

Join Us at Living Planet Symposium

Come meet us at the ESA Living Planet Symposium 2025 in Vienna – our experts will be ready to demonstrate the system, answer your questions, and gather your feedback. From conceptual insights to interactive demos, the WorldCereal team will be showcasing how Earth Observation is being harnessed for operational, global-scale agricultural monitoring. Whether you're a policy expert in global food systems or a technical EO professional looking for practical tools, our sessions are designed to inform, inspire, and empower:

Can’t make it to the symposium? No problem: head over to the online WorldCereal platform and start a run, then join the community forum to share your experience or ask any questions.

Together, let’s cultivate a new era of global crop monitoring. Whether you’re a crop analyst at an international agency, a researcher, a farmer, or simply curious about Earth observation, WorldCereal offers you the tools to make sense of what’s growing on our planet. Dive in, create your first custom crop map, and help us shape the future of agricultural intelligence. Feel free to reach out directly via our contact page or forum – the WorldCereal team is ready to support and learn from you. Let’s grow this community and capability, one season at a time.