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.
2024
Forfattere
Lars Michaloff NordbySammendrag
In this master thesis, the aim is to find out if it is possible to identify and segment out Juniper from terrestrial pictures taken during the fieldwork of NIBIO’s National Forest Inventory (NFI) project. The goal of the thesis has been to develop a machine learning model that can identify and localize juniper in the pictures. It is also studied if the model training could benefit from the addition of more data from an external dataset. Segmentation and classification are used to try to achieve these goals. The models are trained on datasets containing images from NIBIO and iNaturalist. The models have then been tested on pictures from NIBIO, and the validation parameters compared to find the best model. The model architectures used are from the YOLOv8 models, and this architecture has been used in both classification and segmentation tasks. The results show that all the segmentation models generally perform poorly, while the classification model with an adjustment in the confidence level delivers the best results. However, the results show that the segmentation models still have some potential to perform better by training on a sufficient dataset. The model trained on a combination dataset shows that using an external dataset could result in a model that is more accurate and confident in its predictions and shows signs of delivering a more robust model. The thesis concludes, therefore, that it is possible to develop a machine learning algorithm that can detect juniper in terrestrial pictures. Further work should be focused on gathering a sufficient dataset and trying out different model architectures.
Sammendrag
Prosjektbeskrivelse SUSDOCK-prosjektet
Sammendrag
Beskrivelse av SUSDOCK prosjektet
Forfattere
Tatiana F. Rittl Anne de Boer Reidun Pommeresche Anne-Kristin LøesSammendrag
Data paper
Sammendrag
Næringsstoffer og miljøtiltak i landbruket vil få mye oppmerksomhet på årets kornkonferanse. Blant foredragsholderne er NIBIO-forskerne Randi Berland Frøseth og Isabell Eischeid. De vil snakke om hvordan nitrogenet kan utnyttes bedre og om hvordan man kan unngå næringsstoffer på avveie i vann og vassdrag.
Forfattere
Mostafa Hoseini Helle Ross Gobakken Stephan Hoffmann Csongor Horvath Johannes Rahlf Jan Bjerketvedt Stefano Puliti Rasmus AstrupSammendrag
Det er ikke registrert sammendrag
Sammendrag
Large clear-cut areas as a consequence of drought and bark beetle infestations necessitate extensive replanting efforts in German forests, leading to an increased interest in efficient planting systems. In addition to manual planting, mechanized and semi-mechanized systems utilizing surplus forest machine capacities available after completion of salvage logging operations are likely required for timely reforestation of the clear-cut areas. A semi-mechanized system utilizing a standard forwarder with a grapple-actuated soil borer for both, the transport of planting material and the preparation of planting pits, combined with two workers carrying out manual planting, was investigated in a time-and-motion study. The frequency method was used after video recording of a planting operation that covered an area of approximately 1.2 hectares. A total of 815 alder saplings (Alnus glutinosa L.) with heights of 1.2–1.5 m were planted. Observed productivity was 93 saplings per system work hour (SWH). With additional placement of stakes for stabilizing the plants, the productivity decreased to 42 saplings per SWH. While directly comparable results were not found in the literature, available productivity figures of purely manual planting systems do not suggest an increased productivity of this semi-mechanized system. Considering ergonomics, however, forwarder utilization provides reduced workload not only in plant hole preparation but also with material transport and clearing of planting spots. Both the ergonomic aspects of the system and, in particular, the suitability of the soil borer for different soil textures should be further investigated.
Sammendrag
Background: Small-scale forests (woodlots) increasingly account for a greater proportion of the total annual harvest in New Zealand. There is limited information on the extent of infrastructure required to harvest a woodlot; road density (trafficable with log trucks), landing size, or the average harvest area that each landing typically services. Methods: This study quantified woodlot infrastructure averages and evaluated influencing factors. Using publicly available aerial imagery, roads and landings were mapped for a sample of 96 woodlots distributed across the country. Factors such as total harvest area, average terrain slope, length/width ratio, boundary complexity and extraction method were recorded and investigated for correlations. Results: The average road density was 25 m/ha, landing size was 3000 m2 and each landing was serviced on average 12.8 ha. Notably, 15 of the 96 woodlots had no internal infrastructure, with the harvest completed using roads and landings located outside of the woodlot boundary. Factors influencing road density were woodlot length/width ratio, average terrain slope and boundary complexity. Landing size was influenced by average terrain slope, woodlot length/width ratio, and woodlot area. Conclusion: The results provide a contemporary benchmark of the current infrastructure requirements when harvesting a small-scale forests in New Zealand. These may be used at a high level to infer the total annual infrastructure investment in New Zealand's woodlot estate and also project infrastructure requirements over the foreseeable future. Keywords: forest infrastucture, small-scale forestry
Sammendrag
Det er ikke registrert sammendrag
Sammendrag
Det er ikke registrert sammendrag