17th Conference on Probablity and Statistics in the Atmospheric Sciences

6.3

Evaluation of GCM/RCM by the classified synoptic systems' approach

Isabella Osetinsky, Tel Aviv University, Tel Aviv, Israel; and P. Alpert

The evaluation of future climatic predictions from General/Regional Circulation Models (GCM/RCMs) initially requires the evaluation of their predictions for past periods. Consideration of the GCM output that is confined by a certain region as well as RCM output, allows the comparison of the modeled past climate with the well-established real regional climate reconstructed from the reanalysis data. The daily synoptic systems classification, developed recently for the Eastern Mediterranean (EM) region and based on the NCEP reanalysis data, has now been used for evaluation of the ECHAM4/OPYC3 modeled EM past climate. The model evaluation has been performed during 1950-2000.

The considered model past climate corresponds to the output of ECHAM4/OPYC3 model. The model-running period till 1990 was based on the observed CO2 and other greenhouse gases (GHG) increase, and since 1990 - on the GHG SRES (Special Report on Emissions Scenarios) scenario B2, where dynamics of technological change continue along the historical trends. The model originates from the European Centre for Medium-Range Weather Forecasts (ECMWF). The subsequent coupled global model ECHAM4/OPYC3 was developed in co-operation between the Max-Planck-Institute for Meteorology (MPI) and Deutsches Klimarechenzentrum (DKRZ) in Hamburg, Germany. The ECHAM4/OPYC3 climate data for the EM region were provided by MPI.

The real past EM climate has been described as the combination of the daily synoptic systems. The EM region was defined within 30E-40E / 27.5N-37.5N. Each EM daily synoptic system describes concisely the bulk of meteorological information – geopotential height H, temperature T, and two wind components U and V - on all grid points of the EM region. The EM synoptic classification includes the following big groups of the synoptic systems: autumn/winter Red Sea Troughs (RST), winter rainy Mediterranean Lows, spring hot and dusty Sharav Lows, summer Persian Troughs, and highs during a year.

The EM daily synoptic systems have been reconstructed using the NCEP Reanalysis data at 12Z. The data were provided by the NOAA-CIRES Climate Diagnostics Center, Boulder, Colorado, USA, from their Web site at http://www.cdc.noaa.gov/. The reconstruction has been carried out using the modified Discriminant Analysis method. This method helps to classify automatically the large dataset by use of the experts’ classification of the limited but representative dataset cut. The latter is a training set for the following automatic classification. The experts classified the set of the NCEP 12Z sea-level pressure daily maps of the EM region for 1985 and DJF of 1991/92 into the EM synoptic system classes. The corresponding H, T, U, V daily data has been used as the training set for automatic classification of both reanalysis and model data. For the compatibility with the NCEP reanalysis training set, the ECHAM4/OPYC3 model output obtained with a 2.8-degree grid was regridded to the NCEP 2.5-degree grid.

It was found that by averaging over 1950-2000, the ECHAM4/OPYC3 model predicts the annual frequencies for each synoptic system quite similar to the reanalysis data. However, the model does not depict the multi-decadal monotonic trends seen in the reanalysis data. For instance, the reanalysis RST annual frequencies almost monotonically increased from 50 to 100 (i.e. doubling) during 1966 - 2000. This fits the fact of drying EM climate, but the model doesn’t reflect this trend.

The months with correctly predicted daily synoptic systems differ for the different periods. For example, the highest peak of the RST monthly frequencies usually occurs in November. The year-by-year changes of the November’s RST frequencies are predicted well during 1950 - the late 1960s; but from the late 1960s through 2000 the model and the reanalysis November’s trends are in the opposite directions. The second peak of the RST monthly frequencies usually occurs in January-February. For these months, the model and the reanalysis year-by-year changes of RST frequencies are in the opposite directions over 1960-1980; but during 1980-2000 these frequencies are predicted well.

extended abstract  Extended Abstract (376K)

Session 6, Statistical Climatology (Room 602/603)
Thursday, 15 January 2004, 8:30 AM-11:30 AM, Room 602/603

Previous paper  Next paper

Browse or search entire meeting

AMS Home Page