Materials

Abstract

The field sample of the ninth National Forest Inventory (NFI9) was measured from systematically located clusters of sample plots. The NFI9 sampling designs, cluster sizes and distances between clusters varied from 6 × 6 km in southernmost Finland to 10 × 10km in Lapland. In total, 67,264 field plots were measured on forestry land across the entire country. Stand level characteristics and tree level measurements were recorded from the field sample plots. In the MS-NFI, tree-level volumes were transformed to volumes per hectare using the basal area factor and the maximum radius of the plot. Images from Landsat 5 TM or Landsat 7 ETM+ sensors were the most suitable for the MS-NFI by virtue of the fairly large coverage area of each image. In total, 42 Landsat 5 TM or Landsat 7 ETM+ images and 6 IRS-1 C or D LISS-III images were used to cover the land area of Finland. Map information was used to separate forestry land from other land classes. Map data were also used to stratify the forestry land area and the corresponding field plots into a mineral soil stratum and a peatland soil stratum. The majority of the map data came from the topographic database of the National Land Survey but in some areas this map data had to be combined with separate, older map data. The accuracy of the MS-NFI9 land use map data was estimated using a confusion matrix between the map and field based land classes of the NFI field plot data. The percentage of correctly classified land classes was 92.2% and 96.3% for South Finland and North Finland, respectively. Digital municipal boundaries were used to delineate the basic computation units in the MS-NFI. A digital elevation model was used in two ways, for stratification on the basis of elevation data and for radiometric correction. Coarse scale digital forest variable maps were constructed to guide the selection of the nearest neighbours.