Publikasjoner
NIBIOs ansatte publiserer flere hundre vitenskapelige artikler og forskningsrapporter hvert år. Her finner du referanser og lenker til publikasjoner og andre forsknings- og formidlingsaktiviteter. Samlingen oppdateres løpende med både nytt og historisk materiale. For mer informasjon om NIBIOs publikasjoner, besøk NIBIOs bibliotek.
2025
Forfattere
Marie-Claude Jutras-Perreault Fride Høistad Schei Julius Wold Terje Gobakken Hans Ole ØrkaSammendrag
Clear-cutting can resemble natural disturbances like forest fire, but key differences exist in biological legacy. One way to enhance similarity is by preserving structural features of old-forests, such as retention trees, within harvested areas. The latest Programme for the Endorsement of Forest Certification (PEFC) standards require not only the preservation of retention trees but also their mapping for centralized reporting. This study evaluates the accuracy of retention tree density and volume predictions using airborne laser scanning (ALS) data with low (2 pulses/m2) and high (~100 pulses/m2) pulse densities, with and without spectral data. We also assess the feasibility of large-area predictions with minimal field data by testing both in-situ and ex-situ sources. The study was conducted in a managed 1300 ha forest in southeast Norway. Three reference datasets were used: (1) 630 in-situ retention trees across 27 stands (for species and DBH predictions), (2) 1604 ex-situ sample trees (for DBH predictions), and (3) 150 ex-situ annotated segments (for species predictions). Retention trees were identified using an individual tree segmentation approach, using adaptive local maxima window size and applying an adaptative height threshold to filter regeneration. ALS at 2 pulses/m2 alone provided reliable total density and volume predictions, while adding spectral data improved species-specific predictions. Species predictions were relatively stable across data source (kappa=0.556 for in-situ, 0.519 for ex-situ), but DBH predictions were notably underpredicted with ex-situ data (RMSE=9.40 cm, MSD=-4.55 cm) compared to in-situ data (RMSE=8.84 cm, MSD=0.20 cm). Using adaptive segmentation methods enhances scalability. We recommend sampling ~40 in-situ retention trees to develop DBH-height models and delineating ex-situ annotated segments for species predictions. This approach balances accuracy and efficiency while enabling retrospective analysis using national ALS datasets and orthophotos.
Forfattere
Simon BergSammendrag
Root rot causes significant losses for Norwegian forestry. Mapping infected stumps and planting rot-resistant species around infected stumps could reduce future impacts. At 20 sites, root rot was mapped by adding specific assortments for rotten logs using a harvester that recorded tree locations with high accuracy. The optimal approach was considered detailed planning of planting a Norway spruce and Scotspine mix, using root rot information at tree positions. The average opportunity cost of business as usual(planting only Norway spruce) for the forest owner was 409 €/ha. Planting only Scots pine and detailed planning with rot information at harvester locations increased opportunity costs to 615–886 €/ha.Considering fertility variations reduced the opportunity cost to 408 €/ha, considering average rot at site level reduced it to 397 €/ha, considering rot information at harvester locations and coarse planning reduced it to 378 €/ha, and considering rot information at tree level and coarse planning reduced it to 268€/ha. The optimal approach is currently impractical, while coarse planning with rot information at tree locations is feasible. Costs for rot registration and multi-species planting, excluded due to high uncertainty, are likely covered by the increase of 141 €/ha in net present value.
Forfattere
Simon BergSammendrag
Root rot (Heterobasidion spp.) causes substantial losses for forest owners due to decreased wood quality in Norway spruce (Picea abies). Containing root rot spread in regeneration can be achieved by planting resistant species around infected stumps. However, detecting rotten trees remains challenging. In this study, ground truth data for root rot was collected by seven contractors by adding assortments for rotten pulpwood and cutoffs, with all energy wood assumed rotten. Root rot occurrence was estimated in two ways: (1) by developing Extreme Gradient Boosting (XGB) models from all data (XGB-only); and (2) trough binary classification for bucking patterns containing only rotten or healthy trees, followed by developing XGB models for remaining trees (combined). XGB models were developed nationwide and for two specific contractors. Classifications showed sensitivity of 83–87% (rot) and specificity of 95–99% (healthy).Whether nationwide, contractor-specific, XGB-only or combined classification was better varied by situation. Compared to prior studies, predictions from harvester data outperformed UAV images in classification but were surpassed by handheld camera images. Despite lower sensitivity compared to previous XGB applications, more rotten trees were detected than when using only energy wood as an indicator. As estimations are almost cost-free, the results may be acceptable.
Sammendrag
The European Union Deforestation Regulation (EUDR) mandates traceability of timber that makes up wood products from its harvest site to the end product to ensure sustainable wood sourcing. This study proposes a cost-effective, image-based method for tracing logs using alphabetic codes printed onto logs at the harvest site. These codes are detected and interpreted through a two-stage system leveraging deep learning models. The detection stage employs YOLOv8 to locate tracking codes in images of log piles. It is trained and evaluated on a dataset of 125 images, achieving an F1-score of 0.811 on unseen images. The recognition stage, trained on 1,020 images, uses YOLOv8 models to detect individual characters and their positions within each code. On a set of unseen images, the interpretation stage is able to identify 92.8% of the individual logs despite the limited quality of the printer and degradation of the codes due to stem wetness. Analysis indicates that errors predominantly arise in the character detection step. Compared to existing traceability approaches, this method is more cost-effective than RFID tags and attains higher accuracy than image-based biomarker tracking methods.
Sammendrag
NIBIOs eksperter på skog og kunstig intelligens (KI) trener opp datamodeller til å kjenne igjen enkelttrær i skogen. Utgangspunktet er data fra laserskanning. Jobben er enorm. Målet: Å gå fra bestandsskogbruk til forvaltning av skog på enkelttrenivå.
Sammendrag
Artikkelen tar for seg det økende problemet med rotråte i norsk granskog og spør om det er mulig å avle fram gran med bedre motstandskraft. Gjennom forskningsprosjektet Frisk Skog kombineres genetiske analyser, feltforsøk og data fra hogstmaskiner for å forstå variasjonen i råteresistens og utvikle mer presise verktøy for skogplanteforedling. Samtidig løftes praktiske råd for skogeiere, som bruk av Rotstop, skånsom drift og mer målrettet planting, som viktige tiltak for å redusere tap her og nå.
Sammendrag
Soil disturbance following forestry operations is influenced by multiple factors. Reducing disturbance requires placing strip and base roads in locations with minimal risk of disturbance. However, identifying these areas is a complex task. To address this, we have begun developing a forwarding risk index ranging from 1 to 100 that integrates several geographical information sources in the area around Oslo. This forwarding index seems to provide good estimates of areas with a higher risk of ground disturbance during forwarding operations at the sites used for development. With further development of geographical inputs, their combination into a risk index, and later on nationwide validation, the forwarding risk raster combined with a terrain map could improve the identification of suitable areas for forwarding trails. The risk raster was tested for path planning and performed well in areas with a low to moderate frequency of high-risk pixels but was less effective in areas with a high concentration of high-risk pixels. In these areas, an assessment of the potential ecological impact (erosion, sedimentation of streams, mobilisation of mercury, soil carbon impact, changes in hydrology, soil compaction) of ground disturbance is needed alongside the risk index to determine the least unsuitable trail locations.
Forfattere
Maria Åsnes Moan Stefano Puliti Rasmus Astrup Ole Martin Bollandsås Terje Gobakken Maciej Wielgosz Hans Ole Ørka Lennart NoordermeerSammendrag
The site index (SI) describes a site’s potential to produce wood volume. Accurate information on SI in young forests is essential for planning thinning operations and projecting future growth and yield. For tree species that form annual branch whorls, information on interwhorl distances along the stem may be used to determine the SI in young forests. Branch whorls, and consequently tree height growth trajectories, can be detected automatically using deep learning on very dense laser scanning data. In the current study, we demonstrate this approach in a case study in a young Norway spruce forest. We trained a pose estimation Convolutional Neural Network and detected branch whorls of 97 dominant trees in 54 plots scanned with mobile laser scanning data. We predicted SI determined from detected branch whorls in three different sections of each tree, selected in the stem height range between 2.5 and 8 m: all whorls, the lowest six whorls, and whorls selected with an automatic selection procedure. We compared the obtained SI to the SI determined from field-measured branch whorls. Obtained values of precision, recall, and F1 score for the branch whorl detection were 0.66, 0.58, and 0.62, respectively. Values of root mean square error and mean differences between reference and predicted SI ranged between 19.8%–20.9% and −3.6%–4.0%, respectively. Although the tested approach showed potential for SI determination in young forests, the obtained errors were large. This was due to detection errors and high sensitivity to small changes in height increment. These issues highlight the need for further research to improve branch whorl detection accuracy and address challenges associated with determining the SI in young forests.
Sammendrag
Formålet var å komme fram til anbefalte skogbehandlinger for et tematisk område, dvs lavlandet (< 500 moh) på Østlandet, gjennom en prosess med diskusjon i en gruppe interessenter. Vi har i prosjektet begrenset studieområdet til Fritzøe Skoger med tanke på at anbefalingene skulle kunne anvendes på hele den nevnte landsdelen.
Sammendrag
Statistikk fra Resultatkartleggingen for skogbruk/miljø og Landsskogtakseringen viser at en betydelig andel av sluttavvirkninger gjennomføres før skogen har nådd normal hogstmodenhetsalder (hogstklasse 5). Omfanget av slik tidlig hogst er særlig høyt i granskog på bedre boniteter. Dersom frisk skog avvirkes mens den løpende tilveksten fremdeles er høy vil dette medføre redusert produksjon av virke og lavere karbonopptak på arealene. Hensikten med dette arbeidet var å finne ut hvordan skogforholdene påvirker sannsynligheten for at et bestand avvirkes før det når hogstklasse 5. For å svare på problemstillingen har vi benyttet data fra Landsskogtakseringen og sammenstilt skoglige data registrert ved siste taksering før hogst for prøveflater der det utført snauhogst eller frøtrestillingshogst mellom år 2000 og 2022. Om lag 10 prosent av avvirkningen (areal) i skog som var yngre enn hogstklasse 5 er hogster knyttet til arealbruksendring (utbygging, oppdyrking o.a.). Ved arealbruksendring er det andre hensyn enn best mulig skogøkonomi som utløser beslutningen om hogst. Det står ofte relativt ung skog på de arealene som omdisponeres. Dette driver opp andelen ungskoghogst, men beslutninger gjort med dette som motiv er både vanskelig og kanskje lite relevant å vurdere i et skogfaglig perspektiv. Hovedfokuset i denne rapporten ligger derfor på hogster som gjennomføres på arealer hvor det fortsatt skal drives skogbruk. Vi sammenstiller relevant statistikk som beskriver skogtilstanden på alle avvirkede prøveflater i bartredominert skog uten arealbruksendring, slik den var registrert ved siste taksering før hogst. Vi undersøkte også om bestandsskader bidrar som en viktig årsaksfaktor til hogst. Dette på bakgrunn av at skogsskader og/eller bekymringer rundt skoghelse ble oppgitt som en viktig årsak til skogeiers beslutning om å avvirke i en tidligere undersøkelse. For ung barskog med alder tilsvarende 60-70% av hogstmodenhetsalder vil sannsynligheten for hogst omtrent dobles hvis bestandsskadeomfanget er ca. 15 %, mens for gammel barskog (50% over hogstmodenhetsalder) er denne forskjellen litt mindre. Selv om bestandsskader er en viktig bidragsyter til sannsynligheten for sluttavvirkning for både yngre og eldre skog, er det også en del frisk skog som avvirkes før bestandet kommer i hogstklasse 5. Prøveflatene i landsskogtakseringen takseres ikke hyppig nok til å avdekke sikkert om det har vært bestandsskader forut for en hogst. For å fastslå sikkert hvor stor andel av sluttavvirkningene som har vært utsatt for en forutgående bestandsskade, må andre metoder derfor benyttes.