Predictive state-space modelling of lemming population outbreaks on the Fennoscandian tundra: Are determinants of spatial variation in outbreak amplitude temporally consistent?

Abstract

Lemmings are famous for their spectacular population cycles that causes waves of biomass through the arctic tundra. Both climate variability and the interaction with the sympatric grey-sided vole have been shown to effect lemming outbreaks. However little is known about the transferability of these effects between peaks. I analyzed the spatial variability using snap-trapping data from two consecutive lemming outbreaks, sampled at 98-109 sites on the Fennoscandian tundra in the period from 2004 to 2013. I estimated the interaction between lemming and grey-sided vole, and the sensitivity of lemmings to climate variability as well as the temporal consistency of these effects. Effects were estimated using hierarchical state-space models, where the observation error was modeled using a removal model. My results suggested a positive effect of altitude on lemming abundance in 3 out of 4 seasons. In line with an earlier study, a mutualistic interaction between lemmings and the sympatric grey-sided vole was indicated for the winter of 2006/07, an effect likely driven indirectly by shared predators. However, I found that this interaction was neither consistent between seasons (winter and summer) nor between the two consecutive peaks. Therefore, determinants of lemming peaks, especially the interaction with grey-sided vole, have poor temporal transferability. I propose this to be due to the large temporal variability in snow properties in addition to the temporally long spanning arctic winter where little is known about both the predator and the lemmings. I also discuss how monitoring data could be improved to provide better efficiency of statistical models aimed at estimating predictors of lemming population dynamics.

Key word: Lemmus lemmus; Myodes rufocanus; population dynamics; apparent interactions; temporal transferability; detection probability.