Environmental characteristics of primary outbreak foci of the pine-tree lappet Dendrolimus pini (Lepidoptera, Lasiocampidae)

Dendrolimus pini is a severe pest of Scots pine forests in many European countries. The area and frequency of its outbreaks in Poland has increased since 1990s. Outbreak probability is predicted to increase in yet non-outbreak parts of Europe due to climate change. Several studies have focused on the effect of meteorological factors favorable for D. pini outbreaks. However, tree stand, soil, site, and topographic characteristics have been rarely considered, particularly at a fine spatial level (subcompartment, which delimits an individual stand). Therefore, our study aimed to determine the characteristics of D. pini primary outbreak foci (POF). Topographic variables were calculated from a digital terrain model. All the other variables originated from forest inventory and are available to foresters at the stand level for daily management. Random Forest (RF) procedure was applied to analyze two data sets: with and without topographic variables. The analysis revealed that the most important variables common for both models were soil type/subtype and diameter-to-age ratio, with poor Albic Podzols (Ochric) and Albic Brunic Arenosols (Protospodic) and low values (0.27–0.38) of diameter-to-age ratio characterizing POF. The other most important factors in the RF model with all variables were three topographic variables indicating that flat and hilly terrains are typical for POF. In the RF model without topographic variables, additional important variables were site moisture (dry and fresh sites) and site class (I–V). Out of other variables, stand age and tree density also deserve consideration, with older (over 60 years) and denser (above 0.8) stands being more predisposed to the D. pini outbreaks. For pine stands growing on poor soils and in sites with water shortages, stand density seems to be the only factor that foresters can modify to improve conditions for tree growth. The relatively high overall accuracy of RF models with and without topographic variables (0.75 and 0.68, respectively, based on training and test data subsamples) and their high sensitivity in particular (0.94 for both models) make them a useful and easy-to-implement tool for detecting and mapping outbreak foci of D. pini and subsequent planning a distribution of monitoring plots.