Evaluation of the increased pre-harvest forecasting precision of sawlog supply by use of historical harvester data and wood properties models : a case study on Scots pine in northern Sweden

Abstract: Nordic wood procurement is customer-oriented and involves real-time steering of the procurement according to products and markets. The development of better products and increased process efficiency is important for industrial customers. Sawmills’ demand usually covers total volume, species, lengths, diameter, time of delivery and stock levels, but the development is moving towards a more specific demand targeting also wood characteristics. Thanks to StanForD2010 it is possible to store detailed data of harvested trees through harvester files from previously harvested stands in a standardized manner. Skogforsk has developed the tool hprImputation, which uses kMSN imputation to make yield forecasting of planned harvesting stands based on the known outcome from stored harvester data of similar stands. It is possible to combine the imputation tool with earlier developed models for forecasting wood characteristics, thereby en-abling forecasts on both stand- and log level. With the possibilities to measure qual-ity with 3D/X-ray scanners in sawmills, the forecasting precision on log level can be evaluated. The aim of this masters’ thesis was firstly to evaluate the perceived benefits of in-creased precision in yield forecasting from a value chain perspective and identify key forecasting variables for different perspectives of the value chain. Secondly, the aim was to evaluate the influence of applying the imputation method based on har-vester data and wood properties models on the forecasting precision for key varia-bles at the case company SCA. The study showed that there is a considerable need and value potential for more accurate and detailed forecasting, which would improve the management along the whole value chain from forest to sales of sawmill products. However, there is a need for development of analytical tools that enable a more standardised and transparent handling of the data. The imputation method developed by Skogforsk provided higher accuracy of fore-casting on stand level compared to traditional methods at SCA but is dependent on accurate input data which was best provided by airborne laser scanning data among currently available data sources. The wood properties model developed by Skog-forsk could provide accurate forecasts on mean heartwood diameter, but further studies should evaluate whether the models should be adjusted to varying stand age as is indicated in this study. Abstract This development could provide the missing link between stand characteristics and a sawmill’s outcome of specific products, which combined with high data transpar-ency and integrated analytical tools could boost the abilities of integrated forecast-ing along the value chain.