The productivity of Scots pine (Pinus sylvestris L.) under changing climatic conditions in the southern part of Finland was studied by scenario analysis with a gap-type forest ecosystem model. Standard simulations with the model predicted an increased rate of growth and hence increased productivity as a result of climatic warming. The gap-type model was refined by introducing an overwintering sub-model describing the annual growth cycle, frost hardiness, and frost damage of the trees. Simulations with the refined gap-type model produced results conflicting with those of the standard simulation, i.e., drastically decreased productivity caused by mortality and growth-reducing damage due to premature dehardening in the changing climate. The overwintering sub-model was tested with frost hardiness data from Scots pine saplings growing at their natural site 1) under natural conditions and 2) under elevated temperature condition, both in open-top chambers. The model predicted the frost hardiness dynamics quite accurately for the natural conditions while underestimating the frost hardiness of the saplings for the elevated temperature conditions. These findings show that 1) the overwintering sub-model requires further development, and 2) the possible reduction of productivity caused by frost damage in a changing climate is less drastic than predicted in the scenario analysis. The results as a whole demonstrated the need to consider the overwintering of trees in scenario analysis carried out with ecosystem model for boreal conditions. More generally, the results revealed a problem that exists in scenario analysis with ecological models: the accuracy of a model in predicting the ecosystem functioning under present climatic condition does not guarantee the realism of the model, nor for this reason the accuracy for predicting the ecosystem functioning under changing climatic conditions. This finding calls for the continuous rigorous experimental testing of ecological models used for assessing the ecological implications of climatic change.
Two dynamic models predicting the development of frost hardiness of Finnish Scots pine (Pinus sylvestris L.) were tested with frost hardiness data obtained from trees growing in the natural conditions of Finland and from an experiment simulating the predicted climatic warming. The input variables were temperature in the first model, and temperature and night length in the second. The model parameters were fixed on the basis of previous independent studies. The results suggested that the model which included temperature and photoperiod as input variables was more accurate than the model using temperature as the only input variable to predict the development of frost hardiness in different environmental conditions. Further requirements for developing the frost hardiness models are discussed.