En jämförelse mellan två datorprogram för utbytesberäkningar

Abstract: This thesis is made on behalf of SCA's Forestry Division and is a comparison of two bucking simulation tools: Aptan (manufactured by Skogforsk) and Best/upr (from Berget Systemdesign). The data is based on 32 stands from SCA's forests in Norrbotten and Västerbotten. The objective is to compare the simulations created by the tools with the harvested volumes from those stands. Three variables are needed for the bucking simulation; these are the diameter distribution (at breast height), the relationship function between diameter and height, and the taper coefficients. The volumes of five different assortments from the stands are compared. The assortments are as follows: sawlogs from Scots pine (named TT), sawlogs from Norway spruce (GT), coniferous pulpwood (BM), hardwood pulpwood (LM) and the total volumes from each stand (TOT). The thesis consists of four separate studies. Study 1 is the main evaluation of the systems and is the reference for study 2, 3 and 4. The data needed for the simulations in Study 1 has no damaged trees, which degrades sawlogs to pulpwood. The result of study 1 shows that Aptan overestimates the volume of pine sawlogs (TT) by 19 % and Best/upr underestimates it by 10 %. The prognosis for the rest of the assortments are as follows. GT: Aptan +47 %, Best/upr +40%. BM: Aptan -35 %, Best/upr -36 %. LM: Aptan -18 %, Best/upr -29 %. TOT: Aptan +6 %, Best/upr -12 %. The reliability of the differences above are relatively good for TT, BM and TOT when the standard error (SE) was between 2 and 4 %. The SE for GT and LM was between 10 and 18 %. In study 2, regional and local characteristics of height and taper is examined. The thesis area was divided into five separate regions and then the diameter-height relation function and taper coefficients were compared with those of study 1. The differences were marginal and that is probably more because of small differences in the forest characteristics between the regions rather than poor algorithms in the programs. Damaged sawlogs which get degraded to pulpwood are the objectives of studies 3 and 4. How many damaged trees are there in SCA's forests and how do the programs respond to changes to input of damaged trees? Based on experience at SCA, 30 % of the logs are damaged in some way and this is the value used in study 3. The prognosis results improved for all assortments, 2 to 13 percentage points, except for TT in Best/upr which declined by 3 %. Aptan has a special way to get information about damaged trees from previous harvests. That method was tested in study 4 and the results are less obvious than in Study 3 but still indicate a significant difference compared to study 1. Neither study 3 nor 4 have found the right input level of damages and it is impossible to distinguish this optimal level and how well the programs respond to damages based on this thesis. One thing is for sure: damaged trees are highly correlated to the quality of bucking simulations and this subject needs further research. The results are affected by both random and systematic errors. Some of them are related to the method this study is based on. If the errors can be decreased during practical work, e.g. by efficient sampling of trees, then the bucking simulations can be more precise.