Abstract

Topic: Moose and forestry

Modelling the factors influencing moose damage in forest tree seedlings at several scales

Ari Nikula¹, Jyrki Pusenius², Joanne Demmler³, Tuomas Rajala³, Leena Ruha⁴, Antti Paasivaara⁴, Juho Matala⁵, Juha Heikkinen³

Natural Resources Institute Finland, Ounasjoentie 6, 96200 ROVANIEMI, Finland
Natural Resources Institute Finland, Yliopistokatu 6 B 80100 Joensuu
Natural Resources Institute Finland, Latokartanonkaari 9, 00790 Helsinki
Natural Resources Institute Finland, Paavo Havaksen tie 3, 90570 Oulu
Natural Resources Institute Finland, Yliopistokatu 6 B, 80100 Joensuu

Abstract
Moose resource selection varies by changing temporal and spatial criteria. Understanding these factors is essential for moose population management but also for predicting and preventing moose damage. We modelled the factors influencing moose damage in forest seedling stands by using information about local moose densities, forest landscape structure as well as the structure of the stands. Moose data were obtained from localized moose helicopter counts and forest landscape data were derived from the multi-source satellite imagery based National Forest Inventory dataset. Seedling stands for field inventory were selected by using spatially balanced sampling, which also considered variation in moose density. In each stand the number of browsed Scots pine seedlings, the degree of damage and several other stand structure parameters were registered. Inventories were made in Western-Finland (WF) and Southern-Lapland (SL) representing different vegetation zones and forest landscape structures. A total of 82 stands / 825 plots (WF) and 63 stands / 604 plots (SL) were inventoried. Plot and stand level information were augmented with habitat information with 2 km of each stand centroid. The number of browsed pines were modelled using a weighted generalized linear mixed model with random effects for each stand. We chose this method to adjust for stand variability and sampling variability of the moose densities. Incident rate ratios were calculated for each site at both stand and plot level. Finally, we cross-validated our models against equivalent Bayesian and zero-inflated generalized linear mixed models. For both areas local moose density was among the most important variables explaining damage. In WF the amount of pine plantations around stands increased damage probability and open areas reduced it. In SL the amount of mixed young thinning forests increased damage and pine dominated thinning forests reduced it. Of the stand-level variables site type, the amount of pines and deciduous trees explained damage best. Our results indicate that in winter, moose seek out areas with the most pine-dominated saplings, and of these, they choose saplings with the most pine and deciduous food available. Effective moose management should consider both regional and local resource availability to mitigate damage.