Detta projekt kretsade kring två huvudspår för automatisk timmermätning. Den första delen handlade om att använda AI för automatisk objektdetektion vid stockmätning i sågverk. Bilder erhölls från två kameravinklar i sågverket Bygdsiljum, en vy som visar ändytor och en annan som visar mantelytor på stockarna. Dessa bilder hade sedan tidigare annoterats med märkningar som anger träslag (t.ex. gran eller tall), och även ett mycket litet antal andra typer av egenskaper (såsom vissa sällanfel). Målet var att förbättra en befintlig AI-baserad objektdeteksionsmodell (YOLOv7) genom att anpassa den bättre till sågverkets specifika miljö. Genom att experimentera med bland annat utökad träningstid och så kallad ” mosaic-augmentering” förbättrades modellens precision avsevärt relativt de experiment som tidigare gjorts inom ramen för detta ändamål, särskilt när det kommer till att skilja mellan gran och tall. Även tester med varierande mängder träningsdata visade att mer data kan förbättra resultaten, men att effekten avtar efter en viss punkt. Projektet testade också en unifierad modell som kan hantera båda kameravinklar samtidigt. Denna modell presterade stabilt och visade att det är möjligt att effektivt kombinera data från olika perspektiv. Modellen visade sig främst fungera i sågverksmiljön, men presterade sämre i en mer generell “skogsmiljö” (utomhus), vilket inte är förvånande. En av de huvudsakliga slutsatserna utifrån resultaten är att AI kan användas som ett hjälpmedel för att underlätta och snabba upp ytterligare annoteringsarbete i framtiden, vilket i sin tur kan ge upphov till modeller som presterar tillräckligt bra för sökta ändamål.

Den andra delen handlade om att undersöka användningen av hyperspektral avbildning med NIR för att upptäcka defekter (skogsröta och blånad) och klassificera träslag. Modeller utvecklades för att förutsäga olika typer av defekter. Gran- och tallprover, med olika grader av skogsröta samt blånad, samlades in från sågverk för att skannas och analyseras.

Proverna skannades med två kamerasystem, SWIR och VIS-NIR, men det är ännu oklart om de resulterande modellerna är tillräckligt bra för att implementeras i sågverksmiljö. SWIR-kameran har sannolikt mycket goda möjligheter att detektera träslag och barkförekomst. Även detektion av defekt respektive frisk ved kan sannolikt genomföras, men möjligheten att klassa defekterna som skogsröta och blånad är sämre.

De flesta testerna har genomförts på en renkapad yta, vilket är en brist. Hur bra systemet fungerar på smutsiga ytor behöver utredas ytterligare. De prediktioner som gjorts på smutsiga ytor är i nuläget inte tillräckligt bra, vilket till viss del är naturligt. Om stockänden är täckt med ett lager jord/lera så är det svårt att hitta defekterna. Man kan ev göra det om inträngningsdjupet är tillräckligt, men det är tveksamt om det är det.

Om resultaten är tillräckligt bra för att gå vidare och implementera en pilotuppställning beror också på kraven på systemet. Om det räcker med att defekter hittas utan att behöva klassificera dem i skogsröta och blånad, samt att kraven på hur stor del av ytan som är angripen är rimliga, så är det troligen fullt möjligt att ta ett steg till och skapa ytterligare ett projekt där systemet praktiskt testas i sågverksmiljö. I ett sådant projekt kan man dels få svar på praktiska svårigheter med mätmiljön, men det ger också en plattform för att samla mer data och bygga robustare modeller. I så fall måste det finnas enkla och effektiva stöd för annotering av defekter. På så sätt kan mer data samlas in, praktiska omständigheter testas och det kan eventuellt leda till ett fungerande system.

Background: The evaluation of genomic selection (GS) efficiency in forestry has primarily relied on cross-validation schemes that split the same population within a single generation for both training and validation. While useful, this approach may not be reliable for multigenerational breeding. To our knowledge, this is the first study to assess genomic prediction in Norway spruce using a large dataset spanning two generations in two environments. We trained pedigree-based (ABLUP) and marker-based (GBLUP) prediction models under three approaches: forward prediction, backward prediction, and across-environment prediction. The models were evaluated for ring-width, solid-wood and tracheid characteristics, using ~ 6,000 phenotyped and ~ 2,500 genotyped individual. Predictive ability (PA) and prediction accuracy (ACC) were estimated using an independent validation method, ensuring no individuals were shared between training and validation datasets. To assess the trade-off between comprehensive radial history and practical direct methods, we compared GBLUP models trained with cumulative area-weighted density (AWE-GBLUP) and single annual-ring density (SAD-GBLUP) from mother plus-trees. These models were validated using early and mature-stage progeny density measurements across two trials. Results: Despite the smaller number of individuals used in the GBLUP models, both PA and ACC were generally comparable to those of the ABLUP model, particularly for cross-environment predictions. Overall, forward and backward predictions were significantly higher for density-related and tracheid properties, suggesting that across-generation predictions are feasible for wood properties but may be challenging for growth and low-heritability traits. Notably, SAD-GBLUP provided comparable prediction accuracies to AWE-GBLUP, supporting the use of more practical and cost-effective phenotyping methods in operational breeding programs. Conclusions: Our findings highlight the need for context-specific models to improve the accuracy and reliability of genomic prediction in forest tree breeding. Future efforts might aim to expand training populations, incorporate non-additive genetic effects, and validate model performance across cambial ages while accounting for climatic variability during the corresponding growth years. Overall, this study offers a valuable foundation for implementing GS in Norway spruce breeding programs.

Spatial information on wood structure and chemistry is crucial for understanding wood functionality. We present a high-throughput and high-resolution near-infrared (NIR) method for combined imaging of the physical and chemical properties of stem sections from Populus trees. Pyrolysis-GC/MS data was used for sensitive and spatially resolved calibration of wood chemistry while SilviScan™ analyses provided reference data for wood physical properties with 25 μm resolution for wood density and 0.2-2.0 mm for microfibril angle (MFA). NIR prediction models were trained and calibrated on material from both field- and greenhouse-grown trees. Thus, the method was developed for NIR imaging of stem samples as small as 4 mm in diameter with an image resolution of 0.03 mm for small-diameter samples and 0.5 mm for samples with multiple annual rings. The NIR model performance, tested against data not used in the training set, reached the coefficient of determination (Rpred2) values for wood density and MFA of 0.60 and 0.72, respectively. The NIR models for wood chemistry showed Rpred2 values of 0.78 and 0.77 for carbohydrates and lignin, respectively. Models for the G-, S- and H-type lignin had Rpred2 values between 0.58 and 0.86. In addition, we developed a prediction model for the determination of tension wood distribution. According to this model, tension wood was frequently observed in young greenhouse samples, which might explain the higher variation found in the chemical and physical properties of wood in greenhouse-grown compared to field-grown trees. The study also demonstrated that NIR-model estimations in image format can capture spatial variations that are not detectable in bulk analyses of wood properties. Examples of the method applied to greenhouse-grown trees highlight the efforts to develop NIR models with good prediction accuracies based on high-resolution data.

Background: Forest geneticists typically use provenances to account for population differences in their improvement schemes; however, the historical records of the imported materials might not be very precise or well-aligned with the genetic clusters derived from advanced molecular techniques. The main objective of this study was to assess the impact of marker-based population structure on genetic parameter estimates related to growth and wood properties and their trade-offs in Norway spruce, by either incorporating it as a fixed effect (model-A) or excluding it entirely from the analysis (model-B). Results: Our results indicate that models incorporating population structure significantly reduce estimates of additive genetic variance, resulting in substantial reduction of narrow-sense heritability. However, these models considerably improve prediction accuracies. This was particularly significant for growth and solid-wood properties, which showed to have the highest population genetic differentiation (QST) among the studied traits. Additionally, although the pattern of correlations remained similar across the models, their magnitude was slightly lower for models that included population structure as a fixed effect. This suggests that selection, consistently performed within populations, might be less affected by unfavourable genetic correlations compared to mass selection conducted without pedigree restrictions. Conclusion: We conclude that the results of models properly accounting for population structure are more accurate and less biased compared to those neglecting this effect. This might have practical implications for breeders and forest managers where, decisions based on imprecise selections can pose a high risk to economic efficiency.

In this project the possibility to detect resin pockets on the surface and inside wood products with near-infrared (NIR) spectroscopy were investigated. This project is a laboratory study, where samples provided by IKEA were scanned at the wood and fiber analysis laboratory at RISE.

Different samples with and without resin pockets were scanned with a NIR line camera. Linear partial least squares regression models were built to determine the presence of these defects. Models were then applied on samples outside the calibration data set to validate how well they can predict resin pockets hidden below the surface of new samples.

The measurement method could not detect resin pockets below the wood surface. Samples containing resin pockets beneath the surface would, however, often have generally elevated resin content levels at the surface that was not visible to the naked eye, but identifiable with the NIR-camera. Consequently, the method could help to identify wood components that may give problems in the lacquering process or may result in product reclamations due to resin bleeding at a later stage.

The current distribution and population structure of many species were, to a large extent, shaped by cycles of isolation in glacial refugia and subsequent population expansions. Isolation in and postglacial expansion through heterogeneous environments led to either neutral or adaptive divergence. Norway spruce is no exception, and its current distribution is the consequence of a constant interplay between evolutionary and demographic processes. We investigated population differentiation and adaptation of Norway spruce for juvenile growth, diameter of the stem, wood density, and tracheid traits at breast height. Data from 4461 phenotyped and genotyped Norway spruce from 396 half-sib families in two progeny tests were used to test for divergent selection in the framework of QST vs. FST. We show that the macroscopic resultant trait (stem diameter), unlike its microscopic components (tracheid dimensions) and juvenile growth, was under divergent selection that predated the Last Glacial Maximum. Altogether, the current variation in these phenotypic traits in Norway spruce is better explained by local adaptation to ancestral environments than to current ones, where populations were partly preadapted, mainly through growth-related traits. © 2022 The Authors. 

Background: Wood represents the majority of the biomass on land and constitutes a renewable source of biofuels and other bioproducts. However, wood is recalcitrant to bioconversion, raising a need for feedstock improvement in production of, for instance, biofuels. We investigated the properties of wood that affect bioconversion, as well as the underlying genetics, to help identify superior tree feedstocks for biorefining. Results: We recorded 65 wood-related and growth traits in a population of 113 natural aspen genotypes from Sweden (https://doi.org/10.5061/dryad.gtht76hrd). These traits included three growth and field performance traits, 20 traits for wood chemical composition, 17 traits for wood anatomy and structure, and 25 wood saccharification traits as indicators of bioconversion potential. Glucose release after saccharification with acidic pretreatment correlated positively with tree stem height and diameter and the carbohydrate content of the wood, and negatively with the content of lignin and the hemicellulose sugar units. Most of these traits displayed extensive natural variation within the aspen population and high broad-sense heritability, supporting their potential in genetic improvement of feedstocks towards improved bioconversion. Finally, a genome-wide association study (GWAS) revealed 13 genetic loci for saccharification yield (on a whole-tree-biomass basis), with six of them intersecting with associations for either height or stem diameter of the trees. Conclusions: The simple growth traits of stem height and diameter were identified as good predictors of wood saccharification yield in aspen trees. GWAS elucidated the underlying genetics, revealing putative genetic markers for bioconversion of bioenergy tree feedstocks. © 2023, The Author(s).

The Circular Economy (CE) can be implemented with different strategies. In this analysis, it is assumed that there is a desire to find a CE introduction strategy that leads to as rapid and environmentally friendly increase of resource-efficiency as possible. It is further assumed that an introduction strategy needs to take full advantage of other resource-efficiency concepts, societal trends, and developments. It is observed that the ongoing Digitalization and Servitization trends have been found to significantly affect each other, and as Digitalization impacts the possibility to make operations more resource-efficient, and Servitization means a change in the business model, it is assumed that those trends affect the incentives and possibilities for companies to develop Circular Solutions (CS) suitable for the CE, and that it is relevant to ask how. The purpose of this is to contribute to the development of a resource-efficiency concept that synthesizes Digitalization and Servitization, to contribute with suggestions on how to efficiently introduce the CE. It is asked; What measures may take advantage of the Digitalization and Servitization trends, and efficiently facilitate introduction of the CE? To find the answer a theory synthesis conceptual approach is used, and it is analyzed how Digitalization and Servitization trends can contribute to increasing resource-efficiency, and what implications these trends may have on a strategy for the development of CS and the introduction of the CE. The findings indicate that companies interested in enacting resource-efficiency measures and being important actors in the future CE, should have their main focus on business models that incentivizes development of CS, rather than on techniques that enables the development of CS. This further indicates that an efficient CE strategy should prioritize policies that give actors responsible for procurement and purchasing, incentives to 'servitize' their purchasing strategy. © 2022 The authors and IOS Press.

In this project the possibility to determine different wood species and detect defects on wood cross-sections on logs with hyperspectral near-infrared camera was investigated. This project was a laboratory study where cross-sections of logs were scanned at the wood and fibre analysis laboratory at RISE with near-infrared instrument. Wood samples from different origins and species, with different defects were scanned with a hyperspectral near-infrared camera. Classification models were developed to characterise and classify the different logs. Spruce and pine samples were collected from different sawmills, a group of these samples contained fungal growth. The defects within the collected logs varied and some samples contained decay fungi such as sap- and heart rot as well as non-destructive blue stain. Classifications models to distinguish between different wood species were developed, as well classification models to differentiate between healthy wood and the fungal attack were also developed.

Background: Genomic selection (GS) or genomic prediction is considered as a promising approach to accelerate tree breeding and increase genetic gain by shortening breeding cycle, but the efforts to develop routines for operational breeding are so far limited. We investigated the predictive ability (PA) of GS based on 484 progeny trees from 62 half-sib families in Norway spruce (Picea abies (L.) Karst.) for wood density, modulus of elasticity (MOE) and microfibril angle (MFA) measured with SilviScan, as well as for measurements on standing trees by Pilodyn and Hitman instruments. Results: GS predictive abilities were comparable with those based on pedigree-based prediction. Marker-based PAs were generally 25-30% higher for traits density, MFA and MOE measured with SilviScan than for their respective standing tree-based method which measured with Pilodyn and Hitman. Prediction accuracy (PC) of the standing tree-based methods were similar or even higher than increment core-based method. 78-95% of the maximal PAs of density, MFA and MOE obtained from coring to the pith at high age were reached by using data possible to obtain by drilling 3-5 rings towards the pith at tree age 10-12. Conclusions: This study indicates standing tree-based measurements is a cost-effective alternative method for GS. PA of GS methods were comparable with those pedigree-based prediction. The highest PAs were reached with at least 80-90% of the dataset used as training set. Selection for trait density could be conducted at an earlier age than for MFA and MOE. Operational breeding can also be optimized by training the model at an earlier age or using 3 to 5 outermost rings at tree age 10 to 12 years, thereby shortening the cycle and reducing the impact on the tree.