Hyperspectral and multispectral satellite data analysis of nitrogen in grazed and ungrazed tundra in Långfjället, Sweden

• Main Author: Dahlstedt, Emma
• Other Authors: University of Gothenburg/Department of Earth Sciences, Göteborgs universitet/Institutionen för geovetenskaper
• Format: Text
• Language: English
• Published: 2023
• Subjects: Hyperspectral imaging; Imaging spectroscopy; Nitrogen; Reindeer; Herbivory; Tundra; Grazing; reindeer husbandry
• Online Access: https://hdl.handle.net/2077/78263

Reindeer are a keystone species in the circumpolar north, where they hold great cultural and ecological importance. Continued climate change threatens the long-term sustainability of reindeer husbandry. Reindeer can affect the tundra landscape, influencing the soil nutrient cycling, including the soil nitrogen availability. Hyperspectral remote sensing of biochemical traits is a relatively new but promising field of study. This thesis aims to investigate if hyperspectral and multispectral satellite data could be used to differentiate between grazed and ungrazed tundra landscape, as well as to determine which hyperspectral wavelength bands are best suited to estimate nitrogen content over a tundra reindeer grazing site. A hyperspectral satellite, PRIMSA, and a multispectral satellite, Sentinel-2, were used over a heath shrub study site in Sweden. A visual comparison of the spectral curves from PRISMA and Sentinel-2 were used to differentiate between grazed and ungrazed tundra landscape. A variety of remote sensing analysis methods, i.e., PCA, RF regression, RF classification, MLR, and an unsupervised classification, as well as a standard C/N ratio analysis of 24 plant samples were used to get a final map over estimated nitrogen content. The result from this study supports using chlorophyll as a proxy for nitrogen content. From the MLR, all PRISMA wavelength bands below 680 nm (band 35) had R2 values >0.7. In particular, this study identified that band 1, at 405 nm, was best suited for determining nitrogen content over a tundra reindeer grazing site. Neither PRIMSA nor Sentinel-2 were definitively found to be better at differentiating between grazed and ungrazed rangeland. Although this project was limited by the small sample size of the reference data, it shows potential possibilities of what could be done with biochemical trait estimation via remote sensing. Contributing to the wider scientific field by giving a suggested method for estimating nitrogen content and by giving a suggestion for which ...