Error Correction of Daily Temperature and Precipitation from Regional Climate Simulations in Europe and the Effects on Climate Change Signals
Abstract
State-of-the-art regional climate models (RCMs) have shown their capability to reproduce mesoscale and even finer climate variability satisfactorily. However, considerable differences between model results and observational data remain, due to scale discrepancies and model errors. This limits the direct utilization of RCM results in climate change impact studies. Besides continuous climate model improvement, empirical-statistical post-processing approaches (model output statistics) offer an immediate pathway to mitigate these model problems and to provide better input data for climate change impact assessments. Among various statistical approaches, quantile mapping (QM) represents one powerful non-parametric technique to post-process RCM outputs. In this study, results from a transient regional climate simulation (period: 1951 to 2050; general circulation model: HadCM3; emission scenario: A1B; RCM: CLM) with horizontal grid spacing of 25 km is error corrected for entire Europe based on the E-OBS European daily gridded observational dataset (http://ensembles-eu.org). Firstly, the performance of QM for correcting daily temperature and precipitation for long-term simulations is evaluated in a decadal cross-validation framework between 1961 and 2000 and the error characteristics are discussed. In the case of precipitation amount a frequency adaptation tool is presented which deals with rare situations where the probability for non-precipitation days is lower in the observations than in the model. Secondly, the issue of generating new extremes in future scenarios is raised. For this purpose, the ERA-40 reanalysis driven hindcast is used to assure best possible temporal correlation between observations and model output. The hindcast is split such that the independent validation period contains observed extremes outside the range of the calibration period. Two extrapolation schemes at the tails of the calibrated correction functions are tested and compared to the simple mapping on the calibration extremes. Finally, the impact of QM on the climate change signal (2021-2050 minus 1971-2000 from the transient simulation) is analyzed for monthly means as well as monthly extreme parameters according to spatial patterns as well as annual cycles of the climate change signals. It is demonstrated that QM reduces RCM errors by one order of magnitude independent of region or season considered. Additionally, it is shown that, if new extremes outside the calibration range occur, QM using an extrapolation of the error correction function shows more reliable results concerning extremes than using a simple mapping to the extremes of the historical calibration period. Regarding the impacts of QM on the climate change signal, it can be concluded that if variables feature a distinct trend in combination with intensity-dependent error characteristics, QM can change the mean climate change signal by more than 50% as well as the respective annual cycles. The ENSEMBLES data used in this work was funded by the EU FP6 Integrated Project ENSEMBLES (Contract number 505539) whose support is gratefully acknowledged.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2010
- Bibcode:
- 2010AGUFMGC51A0732T
- Keywords:
-
- 0550 COMPUTATIONAL GEOPHYSICS / Model verification and validation;
- 1616 GLOBAL CHANGE / Climate variability;
- 1637 GLOBAL CHANGE / Regional climate change