ARTIFICIAL INTELLIGENCE TO AID TIMBER TRADE
Analyzed History Data Helps in Offer Comparison
It has been estimated that there are currently 632 000 forest owners in Finland. Not all of them have first-hand knowledge on forestry or timber trade. More often than not it is the seller who has the weaker card in a failed trade, unless the fault is caused by a clear and verifiable mistake during measuring or cutting.
Is the Forest Cover Information in Order?
A successful timber trade is usually based on an offer that matches the reality as accurately as possible. On the contrary, a failed trade is often due to a false estimation made in the offer phase about the wood amount or the sawtimber/pulpwood accrual. The state of the terrain and remaining woods after the harvesting is another essential meter for the successfulness of the trade.
Offer requests can be made utilizing a forest plan or forest stand data by the Finnish Forest Centre provided through the Metsään.fi service. The potential buyer usually makes an estimation visit, on which they base their offer. In any case, as up-to-date and accurate forest stand data as possible helps in making an offer that is on point.
Forest inventories that provide the data used to be made every 10 years. Presently, they are conducted every six years utilizing laser scanning, and the information is also updated in between the inventories based on forest use announcements and Kemera forest management grant execution announcements, among other sources. Nonetheless, there are still gaps left in the information.
The seller has an opportunity to fill these gaps for example by making an updated forest plan or by creating a timber trade plan with a consultant. Additionally, artificial intelligence and real-time satellite data offer solutions for updating the information. For example, the Finnish Forest Centre is already employing an automated felling monitoring. With the help of calculations based on satellite data, the forest resources can be estimated up to a tree species level and it can be predicted whether the timber to be traded will be fit for sawtimber or pulpwood.
Who Is Offering, How Much and for What?
In a traditional timber trade, the seller often pays attention especially to the sawtimber and pulpwood stumpage prices per solid cubic meter. However, it does not always pay to focus solely on the unit price. Arto Soikkeli writes in the 2/2020 issue of Metsänomistajat (in Finnish), a magazine focused for forest owners, that many other aspects affect which of the offers is ultimately the best one. For example, on sturdier forests to be cut than the ones having their first thinning, the cutting – and thus the timber type ratio – is the most significant factor that affects the end result.
Different companies have different requirements for measurement, cutting and quality that affect the result of the offer. It is worthwhile to pay attention especially to the sawtimber diameters and lengths. The comparison of measurement requirements can still be coped without specific skills or complex information. However, as the comparisons continue, things get more complicated. To make a more accurate comparison, it would be necessary to know what different companies collect with specific measurements, what kind of faultiness is tolerated and are any special timbers offered (plus, is it worth the trouble to harvest them).
Artificial Intelligence Rakes Through the History Data
The specialty of Finnish forest management associations is a cutting bank, in which the information of realized timber trades is collected and classified based on buyer, sturdiness and the felling type. In other words, the cutting bank tells how the buyers cut the wood they purchase. The forest management associations have taken the effects of cutting into account already for a long time when comparing offers for their customers.
So far, it has been very limited how much data can be processed in the comparison, which has made the comparison less accurate. However, thanks to machine learning, the validity of the comparison information can be significantly enhanced. The core idea of machine learning is that the more data and the more different variables the system gets to analyze, the more accurate it will become. That is, it develops itself better as it gets to utilize different variables in a more versatile way. The cutting bank of the forest management associations holds an enormous amount of information of past timber trades (over a million harvested cubic meters of wood), which the machine learning algorithms now get to rake through.
This means that the artificial intelligence compares the earlier cutting information of different buyers, the correspondence of the offer to similar timber trades and the outcome of the trades. In addition to the offer comparison, a forest owner gets to see how the offers made by the buyer candidates have held true in prior trades.
The offer comparison based on AI will get to action this fall as LeafPoint (site in Finnish), the forest system aimed at the forest management associations, has just begun its production use.
Forester, Product Manager (LeafPoint and WoodsApp)