SilviLaser 2011, Oct. 16-20, 2011 – Hobart, Australia 1 Using high density ALS data in plot level estimation of the defoliation by the Common pine sawfly Tuula Kantola 1 , Päivi Lyytikäinen-Saarenmaa 1 , Mikko Vastaranta 1 , Ville Kankare 1 , Xiaowei Yu 2 , Markus Holopainen 1 , Mervi Talvitie 1 , Svein Solberg 3 , Paula Puolakka 4 & Juha Hyyppä 2 1University of Helsinki, Finland, first.last@helsinki.fi 2Finnish Geodetic Institute, first.last@fgi.fi 3Norwegian Forest and Landscape Institute, Norway, first.last @skogoglandskap.no 4 Finnish Forest Research Institute, Vantaa, Finland, first.last@metla.fi Abstract The climate change has been related to the increase of forest insect damages in the boreal zone. The prediction of the changes in the distribution of insect-caused forest damages has become a topical issue. The common pine sawfly (Diprion pini L.) is regarded as a significant threat to boreal Scots pine (Pinus sylvestris L.) forests. Efficient and accurate methods are needed for monitoring and predicting changes in insect defoliation. In this study, the field work has been carried out in 2009 in Eastern Finland, where D. pini has caused considerable damage in managed Scots pine forests. Altogether 95 sampling plots were used in the analysis. A high density ALS data was acquired simultaneously with the field work. The aim of the present study was to test the accuracy of the plot level needle loss predictions determined from the area based and single tree ALS features separately. The Random Forest method (RF) was utilized in the estimation. The best classification accuracy for the test set was 67.4% (area based features). The best plot level accuracy using the tree-wise features was 60.6%, respectively. Keywords: ALS, Random Forest, defoliation, Diprion pini, forest disturbances 1. Introduction Evergreen coniferous forests dominate the landscape in Finland, covering about 76 % of the land. Forests have been profoundly altered by human activities and most of the massive old-growth forests have been replaced with younger, even-aged managed forests. Climate change and its effects in Finland may be the most serious environmental issue threatening the health of forests. Average annual temperatures have increased more in northern latitudes than the global average temperatures. Ecological balance of forests has been interrupted, causing wide pest damages in managed forests (Moore and Allard 2008). Substantial changes in patterns of forest disturbance have been observed to cover larger areas than ever before (e.g. Lyytikäinen-Saarenmaa and Tomppo 2002). Outbreaks of defoliating insects have increased sharply in the two past decades (Kantola et al. 2010, Karjalainen et al. 2010). There is an increasing need to map and monitor the area, estimate severity and detect spatial location of the hazard (Lyytikäinen-Saarenmaa et al. 2008, Karjalainen et al. 2010). The remote sensing (RS) methods have related differences in spectral responses to chlorosis, foliage reddening or foliage loss over time, aiming to interpret, classify or correlate to damage caused by pest insects. RS can produce data for large areas of remote, inaccessible forest lands quickly and with a higher cost-efficiency rate than ground surveys (Hall et al. 2007). Rapid development of airborne laser scanning (ALS) techniques has provided new perspectives to the forest inventory, as well as forest health survey. With the capability of directly measuring