Climate change in Northern Europe

Dr. Jouni Raisanen
Rossby Centre, SMHI
S-60176 Norrkvping
Sweden

Tel: +46-11-495 8501; Fax: +46-11-495 8001
Email: jouni.raisanen@smhi.se

From 16 Dec 2002:
Department of Physical Sciences, Division of Atmospheric
Sciences, P.O.Box 64 (Gustaf Hällströmin katu 2),
FIN-00014 University of Helsinki, Finland

Email: jouni.raisanen@helsinki.fi (works already - prefer over the old one!)

Background

Northern Europe is an area where many climate models suggest a rather strong response to increased greenhouse gas concentrations, in particular as regards wintertime temperatures (e.g., Raisanen 1994). On the other hand, the observed interannual and interdecadal variations of climate are also large in this region. Furthermore, the location near the western margin of the Eurasian continent and the importance of the Gulf Stream may make the climate in Northern Europe sensitive to changes in both the atmospheric and the oceanic circulation.

The Swedish regional Climate Modelling Program (SWECLIM; http://www.smhi.se/ sgn0106/rossby/index-e.htm), within which the proposed research would be conducted, has as one of its main aims the production of high-resolution climate scenarios for northern Europe by using a regional climate model that takes its boundary conditions from a global OAGCM. To get an idea of the associated uncertainties, it is of vital importance to also study the behavior of other OAGCMs in the same area.

Objectives

We aim at 1) documenting the similarities and differences between the response of different OAGCMs to increased CO2 in the Nordic area and 2) studying, in particular, the importance of internal climate variability in interpreting the model results. To the extent that the available data allow, we will also try to 3) achieve some physical understanding on the intermodel differences in the simulated climate changes. Apart from SWECLIM, the results are expected to be of interest for the climate modelling community in general.

Methodology

To document the intermodel similarities and differences in the simulated response to increasing CO2, all model results will be interpolated to a common grid representative of the typical resolution of the models. Following Raisanen (1997a;b), the intermodel agreement will be quantified in both area-averaged sense (e.g., area mean changes in temperature and precipitation in different models; pairwise centred and/or uncentred spatial cross correlations for different model pairs) and in individual grid boxes (ratio of the intermodel mean change to the intermodel standard deviation; total fraction of models that simulate a change to a given direction).

The amplitude of internal variability in the models will be estimated from the 80-year control simulations, using, for example, the standard deviation of 16 non-overlapping 5-year means. For each model, the statististical significance of the simulated climate changes against internal variability will be characterized (e.g.) by the standard t statistics. It will be studied 1) at which stage of the experiment, if any, the CO2-induced changes in different variables reach statistical significance in different models, and 2) how this depends on the spatial (individual grid boxes vs. area means) and temporal scale (monthly/seasonal/annual means) considered.

An implicit way for assessing the relative importance of internal variability for intermodel differences in the simulated climate changes is to repeat the quantification of intermodel agreement with different averaging periods. For example, the same statistics might be calculated using 5-, 10-, 20- and 40-year periods centred at year 60. Another, more explicit method, applied to one model pair by Raisanen (1997b;1998), is to repeat the statistical comparisons neglecting those grid boxes and models in which the significance of the simulated changes is weak. Furthermore, the variance between the simulated climate changes in different models may be compared with the expected value of the variance resulting from internal variability alone.

If all intermodel differences in the response to increased CO2 cannot be explained by internal variability, some clues of the possible other reasons may be gained by comparing the CO2-induced changes between different variables and with the control climates. For example, simulated temperature changes in northern Europe might be correlated between models with the simulated time mean change in the NAO pressure pattern or in the strength of the North Atlantic thermohaline circulation, or with temperatures in the control run.

Data requirements

This analysis requires, at least, the 80-year (control and greenhouse run) monthly time series of surface air temperature, precipitation and sea level pressure. For physical interpretation, the following 20-year mean fields given in the CMIP II data request list are also of interest: (I.a.i) snow cover, surface latent and sensible heat flux, net solar and IR radiation; (II.i) the North Atlantic meridional overturning streamfunction; (II.iv) sea ice thickness and concentration. Although our focus is regional, having global data seems beneficial, both for the physical interpretation and because some of the more mechanical calculations might be for comparison conducted for other regions of the world.

References

Raisanen, J., 1994: A comparison of the results of seven GCM experiments in northern Europe. Geohysica, 30, 3-30.

Raisanen, J., 1997a: Objective comparison of patterns of CO2-induced climate change in coupled GCM experiments. Clim. Dyn., 13, 197-211.

Raisanen, J., 1997b: Climate response to increasing CO2 and anthropogenic sulphate aerosols - comparison between two models. Report No. 46, Department of Meteorology, University of Helsinki, 80 pp. Also available on-line from http://www.meteo.helsinki.fi/Reports.html

Raisanen, J., 1998: Model differences and internal variability as causes of qualitative intermodel disagreement on anthropogenic climate changes. Submitted to Journal of Climate.