Testing of the K-Nearest Neighbour Method to Estimate and Map Forest Biomass and Cover Type in Fragmented Forests

The k-nearest neighbour (kNN) algorithm is commonly used in Scandinavia and North America for the mapping and estimation of forest variables for relatively large and continuous forest areas. The characteristics of the Danish forests differ from those in areas where the intensive kNN test have been carried out. Forests in Denmark are highly fragmented and consist of small and isolated forest patches. The current forest cover is app. 12% of the land area. The paper describes and evaluates the use of the kNN method for mapping basic forest variables in Denmark. The kNN algorithm is tested in local level assessment of biomass, and cover type by combining Landsat ETM+ satellite data and sample-based forest inventory data from the first 5 year cycle of the Danish National Forest Inventory. The paper highlights and discusses the challenges that the fragmented nature of the forest sets on the application of the kNN algorithm.

BASTRUP-BIRK Annemarie;  PEKKARINEN Anssi;  REITHMAIER Lucia;  KVIST JOHANNSEN Vivian; 

2009-01-07

UMR TETIS

JRC43746

http://forestsat.teledetection.fr/index.php?option=com_frontpage&Itemid=1,    https://publications.jrc.ec.europa.eu/repository/handle/JRC43746,