GCM-based regional temperature and precipitation change estimates for Europe under four SRES scenarios applying a super-ensemble pattern-scaling method |
| |
| Authors: | Kimmo Ruosteenoja Heikki Tuomenvirta Kirsti Jylhä |
| |
| Institution: | 1.Finnish Meteorological Institute,Helsinki,Finland |
| |
| Abstract: | Seasonal GCM-based temperature and precipitation projections for the end of the 21st century are presented for five European regions; projections are compared with corresponding estimates given by the PRUDENCE RCMs. For most of the six global GCMs studied, only responses to the SRES A2 and B2 forcing scenarios are available. To formulate projections for the A1FI and B1 forcing scenarios, a super-ensemble pattern-scaling technique has been developed. This method uses linear regression to represent the relationship between the local GCM-simulated response and the global mean temperature change simulated by a simple climate model. The method has several advantages: e.g., the noise caused by internal variability is reduced, and the information provided by GCM runs performed with various forcing scenarios is utilized effectively. The super-ensemble method proved especially useful when only one A2 and one B2 simulation is available for an individual GCM. Next, 95% probability intervals were constructed for regional temperature and precipitation change, separately for the four forcing scenarios, by fitting a normal distribution to the set of projections calculated by the GCMs. For the high-end of the A1FI uncertainty interval, temperature increases close to 10°C could be expected in the southern European summer and northern European winter. Conversely, the low-end warming estimates for the B1 scenario are ~ 1°C. The uncertainty intervals of precipitation change are quite broad, but the mean estimate is one of a marked increase in the north in winter and a drastic reduction in the south in summer. In the RCM simulations driven by a single global model, the spread of the temperature and precipitation projections tends to be smaller than that in the GCM simulations, but it is possible to reduce this disparity by employing several driving models for all RCMs. In the present suite of simulations, the difference between the mean GCM and RCM projections is fairly small, regardless of the number or driving models applied. |
| |
| Keywords: | |
| 本文献已被 SpringerLink 等数据库收录！ |
|
