New Copernicus High-Resolution Layer on Croplands Launched for Europe

Mapping Cropland Types and Practices Across Europe: What’s Inside the HRL Cropland Product?

The HRL Cropland product offers a comprehensive view of agricultural land use across Europe. It consists of several status layers that provide information not only on what is grown, but also how and when it is cultivated. Together, these layers help track farming practices, support policy implementation (EU Biodiversity Strategy for 2030, the Common Agricultural Policy (CAP) and the European Green Deal), and improve our understanding of cropland dynamics over time.

“With the release of the new HRL Cropland layer, we’re closing a long-standing gap in the Copernicus Land Monitoring portfolio. This dataset delivers unprecedented detail on crop types and cropping patterns across Europe, supporting more informed assessments of cropland use and change over time.”

– Kasper Bonte, R&D Professional, VITO

Crop Type Layer (CTY)

The HRL crop type product is generated through an advanced, scalable workflow designed to deliver detailed and reliable agricultural crop type information across Europe. At its core lies a state-of-the-art transformer-based machine learning model, capable of generalising across diverse landscapes and multiple years. Its integration into the openEO cloud processing framework ensures the approach is scalable and operational across large extents.

Multi-Source Input Data

The model leverages a rich combination of satellite and geospatial data sources:

• Sentinel-1 radar backscatter
• Sentinel-2 optical reflectance
• Meteorological data
• Topographic information (e.g., elevation and slope from Copernicus DEM).

Before being used in the classification model, this data is carefully pre-processed:

• Optical images are cloud-masked, aggregated into 10-day composites, and interpolated where data gaps occur.
• Radar data is similarly composited over 10-day intervals.
• All datasets are resampled to a common 10-meter spatial resolution and temporally aligned to ensure consistency.
• Topography and temperature grids are harmonised to match the spatial grid of the satellite observations.
  
  

Crop Classification Model

This harmonised, multi-source dataset is then input into a transformer-based machine learning model, which analyses the satellite time series to classify crop types. The model is trained on a large, harmonized reference dataset compiled from across Europe, primarily based on GSAA (Geospatial Aid Application) data derived from farmer declarations. This extensive training data enables the model to perform robustly across a wide range of agro-climatic zones, ensuring high generalizability and reliability of the resulting crop type maps.

Figure 1: Spread of used training data for crop type model

Crop Type Classes

The product distinguishes 17 crop type classes (Figure 1), carefully selected to reflect the most common and agriculturally significant cultivated crops in Europe:

• 13 annual crop types, including key crops like cereals, maize, rapeseed, and vegetables.
• 4 permanent crop types, such as vineyards and orchards.

To handle classification uncertainty, the model also includes:

• An “unclassified annual crop” class for pixels where no annual crop reaches a reliable confidence threshold.
• An “unclassified permanent crop” class for permanent crops with high uncertainty.

These classes provide transparency on classification confidence and help users interpret the data more effectively.

Following classification, a series of post-processing steps improve spatial coherence, correct inconsistencies, and enforce a minimum mapping unit of 0,25 hectares. This ensures that the final crop type maps reflect realistic agricultural field structures. Additionally, an inter-annual consistency check is applied to improve classification stability for permanent crops across multiple years. The major crop types are mapped with high performance, for the permanent crops and fresh vegetables some more confusion does exist.

The result is a high-resolution, annually updated crop type map (Figure 2) that supports environmental monitoring, agricultural policy implementation, and land use planning across the European landscape. As this product will receive regular updates it is a true European counterpart to the well-known USDA cropland data layer.

Figure 2: Zoom into crop type product for a region in France for period 2017-2021

Cropping Pattern Layers (CP)

While CTY tells us what is grown, the Cropping Pattern (CP) layers reveal how agricultural land is used throughout the season(s). These 10-meter resolution layers provide insights into field-level management practices and are applied exclusively to annual arable land, excluding permanent crops, which lack well-defined seasonal cycles. As with the CTY layer, OpenEO has been largely used to process the CP layers, supporting scalable, cloud-based generation of products across the European continent.

At the heart of the CP layers is the detection of crop emergence and harvest events, which are estimated using a combination of Sentinel-1 and Sentinel-2 satellite data. These sensors work together to track key seasonal changes at the field level, offering robust temporal coverage even under cloudy conditions. Importantly, the emergence and harvest models have been trained using ground-truth observations, ensuring accurate temporal detection across diverse European agro-climatic regions.

The identified emergence and harvest timings form the foundation for all CP layers, which are divided into five thematic domains: main crops, bare soil, secondary crops, fallow land, and cropping seasons.

1. Main Crops (MC)

This group captures key characteristics of the main growing season, including:

• Emergence date
• Harvest date
• Season duration

Each of these characteristics is accompanied by a confidence layer, indicating the uncertainty (in days) of each date. These indicators support monitoring of seasonal variability and weather-related impacts on crop development.

2. Bare Soil (BS)

These layers quantify periods when land remains uncovered by vegetation:

• Before the emergence of the main growing season
• After harvest of main growing season

They help assess risks related to erosion and nutrient leaching, with uncertainty again expressed in days, based on the precision of detected emergence and harvest events.

3. Secondary Crops (SC)

Secondary crops (often cover crops) planted outside the main season are tracked via:

• Secondary crop type identification
• Emergence date
• Duration

The seasonal type is categorised (short/long season, summer/winter period), supporting evaluation of soil conservation practices and off-season land use. Confidence layers are also included.

4. Fallow Land (FL)

Fallow land is defined as temporarily unmanaged cropland. Two layers capture:

• Presence of fallow land
• Duration left fallow

Detection is based on CTY confidence values, and indicators like harvest signals and vegetation activity. This layer can reveal relevant information on the dynamics of inactive cropland over time.

5. Cropping Seasons

These layers describe how cropping activity evolves over time:

• Number of crop cycles per year
• Main crop rotation diversity over three years

Rotation metrics help identify monocultures or healthy crop diversity. The crop rotation diversity is derived purely from CTY data and is currently available for 2017–2021 in rolling 3-year windows.

Future HRL Cropland Releases

The release of the 2017–2021 HRL Cropland layers marks a major milestone in European agricultural monitoring, but the work doesn’t stop here. Currently, the production of the 2022–2023 HRL Cropland layers is well underway. These updated products will continue to build on the established methodology and are expected to be completed by the end of 2025.

Looking ahead, production for the 2024 reference year will also be initiated soon. A key enhancement in this cycle will be a retraining of the crop type classification model using an expanded and updated training dataset. This retraining will improve the model’s robustness and further refine its ability to distinguish between complex crop classes across diverse agro-climatic zones. The release of the 2024 HRL Cropland products is scheduled for April 2026, completing the full time series from 2017 to 2024 and ensuring uninterrupted continuity in agricultural land monitoring. Importantly, this next release will also mark the inclusion of the United Kingdom, which will be covered consistently across the entire time series.

Join Us at Living Planet Symposium

Curious to learn more about the new HRL Cropland products and how they can support your work in agriculture, land monitoring, or policy? Visit the VITO booth at the Living Planet Symposium 2025 in Vienna during June 23-27. Our Remote Sensing experts are looking forward to answering your questions, showcasing the data, and exploring how the Copernicus HRL Cropland layers can bring added value to your applications.

Join us at one of our dedicated poster sessions to hear more about the development, applications, and future outlook of the HRL Cropland layer:

Let’s connect in Vienna and talk cropland mapping! Cannot make it to Vienna? Feel free to contact us online.