ENVISAT MERIS for country-wide estimates of forest and mountain vegetation in Sweden

Sammanfattning

Abstract – Reflectance-calibrated MERIS data were used to classify Swedish land cover. Three MERIS full-resolution scenes from the same date covered Sweden in one swath. Mountain vegetation was handled as a separate stratum. Clusters from unsupervised classification were identified using photo-interpreted 1-km2 samples from the new National Inventory of Landscapes in Sweden (NILS). Coarse classes fitting FAO’s LCCS scheme were classified. Discriminant analysis of classes and MERIS spectral responses identified bands 3, 5, 9 and 14 giving the best result for non-mountain vegetation, but for mountain vegetation, bands 5, 7, 9 and 14 were preferred. The MERIS Terrestrial Chlorophyll Index (MTCI) was investigated, but gave no advantage in classification. Comparing the classification to the Swedish Land Cover data gave 58% and 63% overall accuracy for mountain and non-mountain, respectively. There was no indication that the many narrow spectral bands in MERIS significantly improved the possibilities for classification in the boreal zone. Furthermore, problems with geometry and misalignment with water masks in the level-2 product were noted

Publicerad i

Utgivare: Proceedings, ISRSE Symposium

Konferens

31st International Symposium on Remote Sensing of Environment

SLU författare

• Reese, Heather

  • Institutionen för skoglig resurshushållning, Sveriges lantbruksuniversitet
    

• Joyce, Stephen

  • Institutionen för skoglig resurshushållning, Sveriges lantbruksuniversitet
    

• Olsson, Håkan

  • Institutionen för skoglig resurshushållning, Sveriges lantbruksuniversitet
    

UKÄ forskningsämne

Landskapsarkitektur
Miljö- och naturvårdsvetenskap

Permanent länk till denna sida (URI)

https://res.slu.se/id/publ/6207