The risk of pine twisting rust damage in young Scots pines: a multilevel logit model approach

Abstract

Factors affecting the 5- and 1-year risk of damage to a pine by pine twisting rust ( Melampsora pinitorqua) were analysed in young pine stands in North Karelia, eastern Finland. Multilevel logit models were developed for predicting the 5-year damage risk and the probability of infections and damage in a single year. Occurrence of pine twisting rust in pines was recorded as the number of new leadershoots with lesions caused by M. pinitorqua (1-year infections), the number of new leadershoots killed or bent by the pathogen (1-year damage) and as the number of pine stems damaged by the pathogen at least once during the 5 previous years (5-year damage). Stand and site characteristics were used as explanatory variables in the models. The residual variation in the data was divided into tree, sample plot and stand levels. A stand’s surface area was the only variable which explained the probability of infections and probability of damage in 1 year. In 1-year models, the residual variances were high both on the sample plot and on the stand levels. The three-level model for 5-year risk of pine twisting rust damage shows that the probability of damage in pines increases when the stand area and numbers of aspens and willows on the sample plot increase, or when the sample plot is situated at a lower elevation than the surrounding area. In addition, an increase in the total number of trees significantly decreases damage risk. In the model for 5-year damage risk, residual variances on sample plot and stand level are quite low compared to the 1-year risk models.