Abstract

The reindeer husbandry area in Sweden covers almost half of the country, with reindeer using mountainous areas in the summer period and forest areas in winter. Lichen is a major forage resource for reindeer and may constitute up to 80% of a reindeer’s winter diet. By integrating field data from the Swedish National Forest Inventory (NFI) and remote sensing images, large area maps indicating the presence and amount of lichen may be created. Previous classifications based only on original spectral data have not always produced satisfactory results due to, for example, the mixed spectral signature of the different cover types present, specifically the presence of lichen under tree canopies. Recent research has shown that preparation of artificial image bands (e.g. principal component, minimum noise fraction or vegetation index) has provided the basis for separability test for all investigated lichen classes and the following definition of best band combinations. The classification of under forest lichen cover can be improved by including artificial/processed image bands into the direction-sensitive statistical classification (Mahalanobis Distance), based on the training sets of NFI data. The aim of the present research is to estimate the fraction of lichen coverage and lichen biomass in the forest landscape by combining remotely sensed imagery, NFI data and ancillary data. The test area is located in northern Sweden and is covered by two SPOT-5 scenes registered on the same date (S5-054-217-J0 and S5-054-217J8, from 04/07/2005). Results of lichen and non-lichen areas discrimination using training data from 943 NFI plots show a classification accuracy of 83.73% for the lichen type class, with observed overall accuracy 80.88% (kappa=0.63). The estimation accuracy for lichen coverage and lichen biomass derived using the proposed method, was assessed by using an independent field dataset gathered in 2006 within one wintering area. The possibility of improving the classification by including additional ancillary data (e.g., soil maps, vegetation layers, and a DEM) will be further investigated and evaluated. The use of different classification methods and their usefulness in classifying ground lichen fraction will also be tested

Keywords

lichen mapping; supervised classification; Swedish NFI; SPOT-5

Published in

Conference

ForestSat 2007

UKÄ Subject classification

Environmental Sciences related to Agriculture and Land-use

Permanent link to this page (URI)

https://res.slu.se/id/publ/21674