Publications
NIBIOs employees contribute to several hundred scientific articles and research reports every year. You can browse or search in our collection which contains references and links to these publications as well as other research and dissemination activities. The collection is continously updated with new and historical material.
2024
Abstract
Point clouds from laser scanning are a standard data source in forest research, but the lack of clear and robust workflows has limited the translation of experimental knowledge into operational tools. We present a novel machine learning-based workflow that uses individual-tree point clouds from drone laser scanning to predict wood quality parameters. Unlike object reconstruction methods, our approach is based on simple metrics computed on vertical slices that summarize information of point distances, angles, and geometric attributes of the empty space between and around the points. Our models use these slice metrics as predictors and achieve high accuracy for predicting diameter of the largest branch per whorl (DLBw) and stem diameter at different heights (DSi) from survey-grade drone laser scans. We show that our models are also robust and accurate when tested on suboptimal versions of the data generated by reductions in the number of points or emulations of suboptimal segmentation scenarios. Our approach provides a simple, clear, and scalable solution that can be adapted to different situations and has the potential to revolutionize forest management and monitoring
Abstract
Climate change poses significant challenges to forests, increasing the occurrence and severity of disturbances such as windthrows, wildfires, pests, and diseases. Effective decision-making tools are crucial for mitigating forest damages and enhancing resilience under changing climatic conditions. We explore the role of forest decision support systems (DSS) in addressing these challenges and present their potential in mitigating future climate-induced damages to forests. We examine the key challenges associated with integrating disturbances and their interactions into decision support systems, using as motivating example the decision support systems for the high-resolution Norwegian forest resources map SR16.
Authors
Elias GarvikAbstract
The Norse built their hulls from oak boards hewn from cleaved or ‘rived’ logs. We know this through artworks and texts as well as archaeological finds describing the ships as well as their construction from the latter half of the early-medieval period. However, we know summarily little when it comes to the pre-production of hewn boards. I ask, - How could the Norse have possibly cleaved? - Which properties in white oak affect cleaving? - Which simple truths can be found of cleaving. Which direction of cleaving is best for oak? This is done by showcasing the differences in cleaving from the top, side, and bottom of a series of white oak logs. Namely why cleaving from each direction fails. To summarize the results. Cleaving from the top shines in cases where there is a large difference in the diameter at the top and bottom of the log. Cleaving from the side can act as a mitigator in the case of knots or to pierce directly through a knot or burl. Cleaving from the top or bottom falls more naturally in the cases where the pith is off-centre. With a log bent like an arch, cleaving from the side fell more naturally. The direction of the cleave ought to be thought of less as an immutable truth and more of as another tool to extract as many boards from a log as possible. We found that no direction of cleaving was inherently better in all scenarios. It is a case of risk-mitigation, which benefits from taking a thorough look at the log and making an informed choice. In cases where the angle of the twisting fibres of a log is significant, all cleaving directions struggle and fail, in cases where there are few natural faults, cleaving from all sides succeed.
Authors
Elise Lyng MonssenAbstract
Wildlife-train collisions are a widespread issue, leading to significant costs to society, reduced animal welfare and potential impacts on species' population sizes. Understanding how ungulates respond to oncoming trains and what affects their detectability is crucial for developing effective mitigation strategies. Building upon the work of Bhardwaj et al. (2022), this study investigates how various factors influence flight behaviour in fallow deer (Dama dama), roe deer (Capreolus capreolus) and moose (Alces alces), as well as wildlife detection from the train driver's perspective. By utilizing films from driver-activated scare systems (DASS) mounted on the front windshield of trains on Norwegian and Swedish railways, this study offers insights into both wildlife reactions and the viewpoints of train drivers. Based on approximately 1000 observations of ungulates, each individual encounter with trains was analysed in relation to railway features, train speed, the train driver’s use of typhoon warning, the ungulates’ location within the terrain and biotic and abiotic factors in the surrounding landscape of the railway. Detection distance and flight initiation distance (FID) was calculated based on the video footage. The results showed that with increasing distance from the railway tracks, ungulates in total were less likely to flee from an approaching train and more likely to increase their FID. Generally, there was a higher probability of flight when the typhoon was used to warn the ungulates, but the effect of this signal was not significant in any of the separate analyses on each species alone. There was also a higher likelihood of flight during dusk/dawn. Flight behaviour of fallow deer was significantly influenced by herd size, with a greater likelihood of flight in smaller groups separated from a large herd. In contrast to roe deer and fallow deer, moose were significantly affected by train speed, with flight likelihood and FID decreasing with increasing speed. All species showed a significantly higher probability of flight across track with closer distance to the track, and in areas where the rail embankment was covered by vegetation with exception of roe deer. Additionally, ungulates in total had a higher probability of an early flight, before detected by the train driver, when visibility was obstructed by vegetation, terrain and curvature of the railway. Lastly, detection distance from the train driver’s perspective was significantly obstructed by vegetation along the railroad embankment. Further studies are suggested on a warning system that could be activated before the animals are detected by the train driver to increase animals’ FID, alongside the implementation of night vision cameras for improved detection of animals during nocturnal hours.
Authors
Maria Åsnes MoanAbstract
Bonitet er definert som overhøyden ved en gitt referansealder og brukes for å beskrive skogens potensial for å produsere tømmer. Bonitering, altså å bestemme boniteten, har utviklet seg fra feltbaserte metoder til metoder som benytter seg av punktskydata. Punktskydata fra flybåren laserskanning og bildematching har blitt brukt til å bonitere med den direkte og høydedifferensielle metoden. Disse metodene bruker overhøydeutviklingen over en kjent periode til å bonitere ved bruk av punktskydata fra minst to tidspunkt, dvs. multitemporale punktskydata. Den direkte metoden boniterer ved bruk av en prediksjonsmodell for bonitet med forklaringsvariabler beregnet fra punktskydata. Den høydedifferensielle metoden finner boniteten der forventet overhøydeutvikling passer best med den predikerte overhøydeutviklingen. Denne avhandlingen tok sikte på å forbedre metodene for bonitering med punktskydata og består av fire studier. Forstyrrelser kan gjøre et område uegnet for bonitering med punktskydata. Tidligere har man definert egnethet som områder uten negativ utvikling i overhøyde eller biomasse, men dette betyr ikke nødvendigvis at overhøydeutviklingen er uforstyrret. Den første studien i denne avhandlingen klassifiserte egnethet med variabler fra multitemporale laserdata. Egnethet var definert basert på feltregistrerte forstyrrelser hos dominerende trær. Resultatene viste at egnethet kunne klassifiseres med multitemporale laserdata, selv om definisjonene av egnethet i den studien var konservative ettersom ett dødt dominerende tre var nok til at prøveflaten ble klassifisert som uegnet. En tidsserie med laserdata kan forbedre nøyaktigheten til boniteringen sammenlignet med å bruke laserdata fra to påfølgende laserskanninger. Dette er fordi overhøydeutviklingen for en lengre periode vil bli representert. Den andre studien i denne avhandlingen brukte laserdata fra tre tidspunkt for å bonitere med den direkte og høydedifferensielle metoden. Prediksjonsfeilene var ikke statistisk signifikant forskjellig når man brukte laserdata fra hele tidsserien sammenlignet med å bruke laserdata fra to påfølgende tidspunkt, dvs. enten første og andre eller andre og tredje tidspunkt. Imidlertid økte andelen av området som var egnet for bonitering når hvilket som helst delsett av påfølgende tidspunkter i tidsserien kunne brukes, noe som ga en økt fleksibilitet til å unngå å perioder der det hadde vært en forstyrrelse. «Value of improved information» kan brukes til å vurdere nytteverdien av forskjellige boniteringsmetoder når beslutningstakeren har ulike mål for skogforvaltningen. Den tredje studien brukte stokastisk programmering for å beregne «value of improved information» ved bruk av den direkte og høydedifferensielle metoden med enten multitemporale laserdata eller laserdata og påfølgende bildematchingsdata. Resultatene viste at «value of improved information» var nærmest null og dermed best for den høydedifferensielle metoden i vårt studieområde. Den høydedifferensielle metoden kan potensielt brukes til å bonitere i ungskog. Den fjerde studien detekterte posisjonen til kvistkranser fra veldig tette punktskydata med en dyplæringsmodell og brukte detekterte kvistkranser til å bonitere med den høydedifferensielle metoden. Dette ga en «root mean square error» mellom 19,85 og 20,87%. En utfordring med bonitering i ungskog er at bonitetskurvene er brattere for lave aldre sammenlignet med høyere aldre, noe som resulterer i at feil i deteksjonen av kvistkranser har større konsekvenser for bestemmelsen av bonitet i ungskog. Denne avhandlingen har tatt for seg noen utfordringer og muligheter for bonitering med punktskydata. Likevel viste den første og fjerde studien at det er behov for mer forskning på hvordan man best kan definere egnethet og bonitere i ungskog.
Abstract
Livestock production systems with ruminants are generally associated with high enteric methane emissions and thus a high carbon footprint, causing these systems to be challenged when it comes to what products to eat and wear in a sustainable future.
Abstract
No abstract has been registered
Authors
Petri Widsten Satu Salo Marc Borrega Anna Kalliola Lauri Tuominen Melissa Agustin Jenni Rahikainen Elisa Spönla Andreas TreuAbstract
No abstract has been registered
Abstract
No abstract has been registered
Abstract
Forests play a major role in the mitigation of avalanche risk in Norway, but the regulations surrounding the management of “protection forests” are still being worked out. To promote protection forest management, avalanche hazard indication maps for Norway have been produced with the automated mapping tool NAKSIN in a way that makes it possible to quantity the effects of the current forests in a spatially explicit way. NAKSIN makes use of published relations for forest effects on snow properties and uses national models of forest characteristics to estimate the effects on release probability and runout given local climate and topography. The forest properties contain parameters that are directly measured (canopy cover), and properties that are predicted (tree diameter, number of trees) with approximately 70% precision according to ground truth data. NAKSIN uses these forest properties in long chains of models, comprising of both mechanistic and empirical elements, some of which are iterated over timesteps during avalanche flow. This means that errors could be propagated throughout those model chains in unexpected ways. The aim of this study was to conduct a sensitivity analysis to examine the effects of errors in the forest data for hazard mapping in a relevant case study region in fjordic western Norway. We examined hazard maps produced using 95% prediction errors for tree diameter and the number of trees per hectare to determine if these would dramatically affect the hazard zones. These hazard maps focused on runout properties as common release areas were implied for avalanches through a common forest canopy cover percentage applied across the two extreme scenarios. Across the entire region, the hazard zones were generally stable with respect to potential errors in the forest data, suggesting the approach is robust and the braking effect of forest is not overstated. There was one exception, where the prediction errors could reduce the forest braking function to negligible. This exception was easy to identify from the difference in hazard zones and the process allows us to consider where more precise measurements of forests could be required in areas with high consequences. The implications of various approaches to estimate forest leaf area index, and how this might impact on release probability are illustrated to further consider this in the next steps of this research.