Thursday 30 March 2017
Validation

Validation of regional climate models means to integrate a model over a control period and to compare its results to observational datasets. Hereinafter the most important validation characteristics of our models are presented for the period of 1961-1990.

Regarding the annual and seasonal mean temperature (Fig. 1), REMO simulation driven by ECHAM5/MPI_OM coupled global model provides the best results. Differences between simulated values and CRU data over the Carpathian Basin are less then 1°C. On the other hand, REMO driven by ERA-40 (i.e., by quasi-"perfect" boundary conditions) strongly overestimates the annual and seasonal mean temperature. This characteristic points to the probable fact that consistency between the boundary conditions and the regional model is less adequate in the re-analysis-forced simulations than in the ones forced by global models. For ALADIN-Climate, mean temperature is underestimated in most cases (therefore, ALADIN-Climate is rather a cold model).

Over Hungary all the ALADIN-Climate experiments and the GCM-driven REMO overestimate the annual mean precipitation. The ERA-40-driven REMO is characterized by overestimation of precipitation over the Northern part of Hungary and underestimation over the Southern one. As far as seasonal means are concerned, simulations give different results, but all of them enhance the precipitation in spring. In autumn simulations are generally wetter than the reality, and the best results are obtained in winter.

For both examined parameters ALADIN-Climate model performs better for the large domain with lower resolution than for the smaller and higher resolution one, indicating that the latter ALADIN-Climate domain is most probably too small (the domain boundaries are over mountainous areas, which can cause spurious noises originated from the boundary relaxation zone). Recognizing this problem we are going to carry out some sensitivity tests for the ALADIN-Climate model in order to find the optimal model domain for future integrations.

As a summary, model validation results indicate that the behaviour of the models is complex; they perform better for certain parameters, however, quite erroneously for other ones making the decision on model development extremely difficult.

1. ábra 

Figure 1: Observed and simulated monthly mean temperature values (°C) averaged
over Hungary for the period of 1961-1990