Swedish researchers develop method to reduce bias in forest mapping

Scientists at the Swedish University of Agricultural Sciences (SLU) have developed a statistical technique that addresses a persistent problem in forest assessment: systematic bias in maps created from remote sensing data.

Forest researchers have long observed that predictive maps tend to overestimate areas with low tree volume while underestimating high-volume forests. This "design-bias" trend masks true forest variability, complicating efforts to identify biodiversity hotspots, plan timber harvests accurately, and monitor carbon stocks.

The SLU research team evaluated different regression techniques using Monte Carlo simulations based on forest data from northern Sweden. Their proposed method, named MAVGAR, showed strong performance in reducing bias.

The study, published in Forest Ecosystems, offers practical guidance for selecting estimation methods based on data quality and mapping objectives. The researchers emphasize that the choice of method involves trade-offs between bias reduction and overall accuracy, measured in terms of root mean square error (RMSE), and should be guided by the specific goals of each assessment.

The study, published in Forest Ecosystems, offers guidance for selecting estimation methods based on data quality and mapping objectives. Accurate forest maps support sustainable forestry, conservation planning, and climate reporting. The choice of statistical method should reflect both the characteristics of the available data and the specific priorities of the assessment, whether that means prioritizing low-bias estimates of high-volume forests or minimizing RMSE across the entire landscape.

The findings are directly relevant for national forest inventories, carbon accounting initiatives, and biodiversity monitoring programs that increasingly rely on remote sensing data for large-scale assessments across diverse forest landscapes.

DOI Link:

https://doi.org/10.1016/j.fecs.2026.100428