Temperature and precipitation scenarios for Norway: comparison of results from dynamical and empirical downscaling

Abstract

A scenario from the coupled atmosphere-ocean climate model ECHAM4/OPYC3 was downscaled by empirical and dynamical methods to show projected changes in temperature (T) and precipitation (R) in Norway under global warming. In the empirical models, large-scale T was applied as a predictor for T. For R, both T and sea-level pressure (SLP) were applied as predictors during most of the year, while only SLP was applied during summer. The dynamical model, HIRHAM, is a regional climate model based upon HIRLAM, but with physical parametrisations from ECHAM4/ OPYC3. Both approaches project from 1980-1999 to 2030-2049 an increase in annual mean temper- atures of between 1 and 2.5°C in various parts of the country. The projected warming is at a minimum along the southern coast, while greater warming is projected inland and in the north. Though the dif- ferences between the approaches are not statistically significant, empirical downscaling systemati- cally leads to a larger projected increase in annual mean temperature than dynamical downscaling does. The difference is at a maximum during winter and/or spring at localities exposed to tempera- ture inversions. Empirical downscaling projects larger winter warming in inland valleys than at more freely exposed localities, and thus implies a reduced intensity or frequency of winter inversions. It is argued that less favourable conditions for ground inversions are consistent with the future projection of increased winter wind speeds and reduced snow-cover. For precipitation, both downscaling approaches project a statistically significant increase in the west during autumn, and in the south during winter. The only significant difference between the results is that dynamical downscaling pro- jects increased summer precipitation in the southwest, while the empirically downscaled scenario shows no significant change. For summer precipitation the present empirical models do not include any predictor carrying the 'climate change signal', and thus the results from the dynamical down- scaling are probably more realistic concerning summer precipitation.