Have you ever wondered how we know how much deforestation has occurred in the Amazon or the Chaco? Or how cities have expanded over the last 20 years?
The answer lies in land cover and land use classifications

What Are Land Cover and Land Use Classifications?

These classifications are products derived from the advanced processing of remote sensing data collected by sensors onboard satellites, drones, or airplanes.
Their main goal is to characterize the observed surfaces in these images by grouping pixels into defined classes such as water, forest, grassland, or cropland.

Through this process, we obtain detailed information about the location, extent, size, and spatial patterns of land features. While the concept might sound technical, the applications are numerous and highly practical.
In simple terms, land cover and land use classifications allow us to:

1. Distinguish surface features (e.g., tell a forest from a crop field).
2. Quantify their areas (e.g., estimate how many hectares of forest exist).
3. Analyze their spatial patterns and temporal changes.

From Satellite to Map

The figure below shows a satellite image and the corresponding land use and land cover classification derived from it.
In this case, the classification represents the conditions in 2013, over an area of one million hectares, where ecosystems are strongly influenced by seasonal variations in rainfall and temperature.

To assign each pixel a specific class, our team of specialists analyzed ten years of Landsat 8 satellite imagery, along with several reference datasets.
Using this information, we identified representative samples for each class, trained and validated multiple machine learning models, and then applied the final model across the entire study area.

The result is a map where each pixel is automatically assigned the most probable class, within a statistically representative and reportable margin of uncertainty.

In other words, each land cover and land use class is associated with a probability value, indicating how likely it is that the classification accurately reflects the actual conditions on the ground.

What Are These Products Used For?

As mentioned earlier, these data allow us to quantify and compare land use and cover across different places and time periods.
By repeating the classification process at various points in time, we can observe how landscapes evolve.

In this example, we analyzed the dynamics of a protected area in Paraguay over a ten-year period, using land cover and land use classifications.

The results show that the dominant cover in the area was flooded savannas, followed by seasonally flooded savannas and irrigated croplands.

When comparing both the maps and the figures, we observed an increase in irrigated crop areas and a reduction of savannas.

By 2023, the presence of irrigated croplands had become even more evident. The large, regular shapes of these fields suggest commercial-scale agricultural activities, which could pose a serious threat to the ecological integrity of the protected area.

These spatial patterns may also indicate weak protection enforcement, land tenure conflicts, or even a recent designation as a protected area, but these are to be confirmed with additional information often from the field, published research and other sources.

Beyond the Map

These observations only scratch the surface of what land cover and land use data can reveal.
Their potential goes far beyond mapping: such datasets can be used to evaluate compliance with environmental regulations, monitor the condition of protected areas at local or national scales, and analyze long-term trends in the conversion of natural covers to anthropogenic uses.

Ultimately, land cover and land use classifications not only allow us to see change, but also to understand it and take action to protect our ecosystems.