Skip to main content
SLU publication database (SLUpub)

Research article2012Peer reviewed

Restricted imputation for improving spatial consistency in landscape level data for forest scenario analysis

Barth, Andreas; Lind, Torgny; Ståhl, Göran


Today, forest scenario analyses often require spatially comprehensive data. This is due to increasing needs to provide broad-scale descriptions of the consequences of different forest management regimes, encompassing issues with a clear spatial dimension such as biodiversity and recreation. Thus, the composition and spatial configuration of the forest landscape have to be accurately assessed. Today, such data are commonly provided by combining field plot data with remote sensing using imputation schemes such as the k Nearest Neighbour (kNN). Often, such algorithms are executed plot by plot, or stand by stand, and there is no final check that overall landscape composition and configuration are correctly captured. As a consequence, the result from scenario analyses may be severely erroneous; for example, kNN data often exhibits a tendency towards mean values, which may result in misleading predicted outputs of goods and services over time. The objective of this paper was to provide a method for improving composition and spatial configuration in landscape level data when applying imputation schemes such as kNN. The methodological framework comprises two steps: (i) composition at landscape level is preserved by a restricted imputation technique; and (ii) composition and spatial configuration within forest stands is improved by rearranging imputed sample-plots between stands. The method was demonstrated in a case study that showed promising results. (c) 2011 Elsevier B.V. All rights reserved.


Forest scenario analysis; National forest inventories; Remote sensing data

Published in

Forest Ecology and Management
2012, Volume: 272, pages: 61-68