Still huge errors in remote sensing assessments of global forests

The need for information about the state of the world’s forests is growing, in order to curb climate change and biodiversity loss. Under the Climate Convention, for example, countries are required to report changes in how much carbon is stored in biomass both above and below ground.

However, obtaining high‑quality data for large‑scale forest monitoring is often difficult. Some forests are inaccessible, and ground-based inventories are expensive and time‑consuming.

Getting closer to "good enough"

Using remote sensing data from space offers new possibilities. Satellite imagery, radar data, and spaceborne laser scanning are used in data models to estimate forest conditions on the ground.

But can the results from such remote-sensing-based monitoring be trusted? In a new study, PhD candidate Emanuele Papucci evaluated the quality of large-scale biomass estimation studies using remote sensing conducted between 1992 and 2022.

“The methods have improved, and we are approaching ‘good enough’, but several challenges remain before biomass estimates can become comprehensive and truly reliable,” says Emanuele Papucci.

Advances have been made in remote sensing technology: the use of lidar, radar, and digital aerial photogrammetry has expanded. Techniques capable of capturing forest structural properties have strengthened the reliability of biomass estimates.

Several challenges

At the same time, several problems persist, such as lack of field data for model calibration, signal saturation in high‑biomass areas, and misconceptions about statistical methods.

In their scientific article, Papucci and his colleagues highlight that statistically rigorous estimates based on remote sensing still face numerous challenges. Key future needs include:

• Obtain sensor data with stronger correlations to biomass to improve the accuracy of remote‑sensing‑based estimates.
• Gather larger, harmonized field datasets—at both tree and plot level—to better calibrate and validate biomass models.
• Apply statistical principles correctly and consistently to ensure robust inference and avoid systematic errors.
• Develop improved methods for domain estimation, enabling more reliable estimates across different forest types or categories.
• Strengthen quality assurance and quality control throughout all stages of data collection, processing, and analysis.

Large margin of error

Papucci’s supervisor, Professor Göran Ståhl, notes that the margin of error in remote‑sensing‑based estimates can be very large. Instead of just a few percentage points, errors of several tens of percent may occur when reporting forest conditions at the national level.

“Major global assessments based on remote sensing and models must be taken with a large pinch of salt,” says Göran Ståhl.

The calculations typically fail to account for method‑related systematic errors, which lead to underestimation in dense forests and overestimation in sparse forests. They also generally assume that a single model relationship can be applied across vast areas—something that often results in strong regional systematic errors.

According to Ståhl, the current lack of knowledge about statistical methods is a major obstacle to sound development of global assessments that rely on remote sensing.

“Those who produce these estimates must know how to correctly report margins of error— and I would argue that this knowledge is lacking today.”

Need of smart combinations

The technologies used in global inventories often rely on satellite imagery, radar, or spaceborne laser scanning, where the relationships between remote sensing data and ground conditions can be weak. But these methods can be improved. In Sweden, for example, forests are now laser scanned from aircraft at much higher resolution. Combined with high‑quality reference inventories collected on the ground, the errors become much smaller.

“You need smart combinations where remote sensing and ground measurements complement each other, allowing you to do better than what either data source can achieve on its own,” says Göran Ståhl.

Read A review of forest biomass assessments based on remote sensing reveals progress in methodological quality—but major challenges remain