We tested the sensitivity of a dynamic ecosystem model (LPJ-GUESS) to the representation of soil moisture and soil temperature and to uncertainties in the prediction of precipitation and air temperature. We linked the ecosystem model with an advanced hydrological model (JULES) and used its soil moisture and soil temperature as input into the ecosystem model. We analysed these sensitivities along a latitudinal gradient in northern Russia. Differences in soil temperature and soil moisture had only little influence on the vegetation carbon fluxes, whereas the soil carbon fluxes were very sensitive to the JULES soil estimations. The sensitivity changed with latitude, showing stronger influence in the more northern grid cell. The sensitivity of modelled responses of both soil carbon fluxes and vegetation carbon fluxes to uncertainties in soil temperature were high, as both soil and vegetation carbon fluxes were strongly impacted. In contrast, uncertainties in the estimation of the amount of precipitation had little influence on the soil or vegetation carbon fluxes. The high sensitivity of soil respiration to soil temperature and moisture suggests that we should strive for a better understanding and representation of soil processes in ecosystem models to improve the reliability of predictions of future ecosystem changes.

The dynamic vegetation model (LPJ-GUESS) is used to project transient impacts of changes in climate on vegetation of the Barents Region. We incorporate additional plant functional types, i.e. shrubs and defined different types of open ground vegetation, to improve the representation of arctic vegetation in the global model. We use future climate projections as well as control climate data for 1981-2000 from a regional climate model (REMO) that assumes a development of atmospheric CO(2)-concentration according to the B2-SRES scenario [IPCC, Climate Change 2001: The scientific basis. Contribution working group I to the Third assessment report of the IPCC. Cambridge University Press, Cambridge (2001)]. The model showed a generally good fit with observed data, both qualitatively when model outputs were compared to vegetation maps and quantitatively when compared with observations of biomass, NPP and LAI. The main discrepancy between the model output and observed vegetation is the overestimation of forest abundance for the northern parts of the Kola Peninsula that cannot be explained by climatic factors alone. Over the next hundred years, the model predicted an increase in boreal needle leaved evergreen forest, as extensions northwards and upwards in mountain areas, and as an increase in biomass, NPP and LAI. The model also projected that shade-intolerant broadleaved summergreen trees will be found further north and higher up in the mountain areas. Surprisingly, shrublands will decrease in extent as they are replaced by forest at their southern margins and restricted to areas high up in the mountains and to areas in northern Russia. Open ground vegetation will largely disappear in the Scandinavian mountains. Also counter-intuitively, tundra will increase in abundance due to the occupation of previously unvegetated areas in the northern part of the Barents Region. Spring greening will occur earlier and LAI will increase. Consequently, albedo will decrease both in summer and winter time, particularly in the Scandinavian mountains (by up to 18%). Although this positive feedback to climate could be offset to some extent by increased CO(2) drawdown from vegetation, increasing soil respiration results in NEE close to zero, so we cannot conclude to what extent or whether the Barents Region will become a source or a sink of CO(2).