Improving the Estimation of Basal Area in Harvest Planning

In the forestry industry, there are several reasons why today’s forest management plans do not always match reality. Factors such as lack of objectivity from the planner or even minor negligence can have serious consequences for the final outcome. The level of uncertainty is so significant that one should account for a possible discrepancy of up to twenty percent, which, in numerical terms, can mean a difference of around SEK 500,000 in value for an average-sized forest property in central Sweden.

Harvesters, the machines used for logging, generate more data than a fighter jet. This volume of data is far beyond what humans can process, but it is easy to imagine the opportunities it presents for AI.

Purpose

Sveaskog, Sweden’s largest forest owner, conducts forest inventories using statistical methods and performs thinning and regeneration harvesting across more than 40,000 hectares each year. Before harvesting a specific area, extensive preparatory work is required.

A planner goes out to the designated forest area equipped with a relascope to measure and estimate approximate tree density, identify different tree species, and assess the average height and thickness of the trees.

Traditionally, a set of statistical formulas is applied to this information to estimate the potential yield from a particular area. Sveaskog partnered with Tenfifty to determine whether historical data and predictive modeling could be used to improve forecasts of actual outcomes from harvested forest areas.

In this project, Tenfifty’s AI experts helped Sveaskog model the structure and characteristics of the forest. Based on the planner’s measurement points, our algorithms incorporated additional information such as:

• The forest’s altitude above sea level

• Adjustments made in nearby areas

• The identity of the individual conducting the measurements (as estimation errors vary between people)

• The forest’s geographical location within Sweden

• Growth conditions

• Type of soil and terrain