DIGITALNA ARHIVA ŠUMARSKOG LISTA
prilagođeno pretraživanje po punom tekstu
ŠUMARSKI LIST 3-4/2011 str. 26 <-- 26 --> PDF |
L. Kutnar, A. Kobler: PREDICTION OF FORESTVEGETATION SHIFT DUE TO DIFFERENT ... Šumarski list br. 3–4, CXXXV (2011), 113-126 Climate change scenarios –Scenariji klimatskih promjena For simulation of future vegetation states, the existing climate-change predictions for Slovenia have been used (Bergant 2007,Kutnar etal. 2009).To estimate the future temperature and precipitation conditions in different regions of Slovenia by the end of the 21st century, empirical downscaling was used to project the results of General Circulation Model (GCMs) simulations with four different models (CSIRO/Mk2, UKMO/HadCM3, DOE-NCAR/PCM in MPI-DMI/ECHAM4-OPYC3) to five selected locations in Slovenia (Ljubljana, Novo Mesto, Murska Sobota, Rateče-Planica in Bilje) (Bergant 2007). A combination of empirical orthogonal function analysis together with a partial least squares regression was used to develop empirical models based on local observations and NCEP/NCAR reanalysis in the large scale. Spatial model – Within the present forest area, a model was constructed, linking the vegetation type to the climate factors, the relief and the soil at the spatial level of 100×100 m quadrants, in order to provide the model- based predictions of potential vegetation distribution in case of climate warming.The relationship was gleaned with the data mining tool SEE5 (www.rulequest.com) from the empirical data (training dataset).The training data consisted of equal numbers of randomly sampled records for each vegetation type. Each record consisted of the current vegetation type at a particular 100×100 m quadrant, followed by the corresponding climate data (average monthly and yearly temperature, precipitation and evapotranspiration values for the 1970–2000 period), relief data (elevation, terrain slope, terrain exposition), and soil data (FAO soil type). The model constructed with SEE5 took the form of a decision tree. The accuracy of the model was estimated to be 71% (at the level of 13 vegetation types) with 10-fold cross-validation on training data. The cross-validation returns As GCM simulations are commonly based on a limited number of emission scenarios, in this case SRESA2 and B2, local projections were additionally scaled to other marker SRES scenarios (A1Fl,A1T,A1B).The results of projections indicate the strongest warming in summer(3.5 °C to 8°C) followed by winter (3.5°C to 7°C), spring (2.5°C to 6°C), and autumn (2.5°C to 4°C) (Bergant 2007). No significant change in precipitation amounts is expected in spring and autumn, while in summer a decrease in precipitation (-20%) and in winter an increase (+30%) is expected. For the simulation of changes of potential forest vegetation, the existing climate-change predictions for Slovenia (Bergant 2007) have been used to create three different scenarios. Prostorni model similar accuracy values as the validation using an independent control sample. Using the model and the existing predictions of the likely future climate (Bergant 2007), we predicted the shift of the forest vegetation in Slovenia for the years 2040, 2070 and 2100 under three climatic scenarios: the mean scenario (median predicted temperature T, median predicted precipitation R, median predicted evapotranspiration E), the pessimistic scenario (maxT,min R, max E), and the optimistic scenario (minT, max R, min E). For comparisons of the predicted values to the present values, we used the potential present values, i.e. modelled present values, and not the real present values. Therefore, the differences between the present and the predicted values were less influenced by the errors of the model. Since the empirical model is only valid within the present forest area, it cannot predict change of the forest area due to climate warming. Therefore, our predictions of vegetation change were only made within the present confines of the forests. RESULTS – Rezultati Taking into consideration the future climate changes (defined by three different climate scenarios: the mean scenario, the pessimistic scenario and the optimistic scenario), the simulation of the future potential forest vegetation showed significant changes of vegetation-type shares in Slovenia. By using all three climate scenarios in an empirical GIS model, the simulations showed the alteration of spatial pattern of 13 vegetation types (groups of similar forest communities) under impacts of climate vary considerably (Figure 2). The mesic forest vegetation may be adversely af fected by such changing environmental conditions. The decrease of the share of currently prevailing beech vegetation types, e.g. groups of Acidophilic Fagus sylvatica forests (14.2%), of Submontane Fagus sylvaticaforests (13.0%), and (Alti-) montaneFagus sylvatica forest in (Pre-)Dinaric region (11.2%), could be expected (Table 1). By the year 2100, the share ofAcidophilicFagus sylvaticaforests might be decreased to range between 0.0% (pessimistic scenario) and 5.0% (optimistic scenario); and the share of Submontane Fagus sylvatica forests might be contracted to range between 0.0% and 4.0% (Table 1). The constant decreasing of (Alti-)montane Fagus sylvatica forest in the (Pre-)Dinaric region, among which Dinaric fir-beech forests (Abieti-Fagetum dinaricum, sin. Omphalodo-Fa ge tum) prevail, has been forecast (Graph 1). |