Precipitation bias in RegCM model simulations for South Caucasus Domain

To study the regional climate in Caucasus, GRENA performed simulations using the Regional Climate Model RegCM[1] v4.3. The model was applied for the period of 1990-2002 with ERA Interim[2] reanalyzes for model tuning and optimal configuration. It was configured for two land model options (bats and LCM) with different combination of skims. The nesting opportunity with horizontal grid mesh 10 km has been examined.

The study was focused on model fidelity of precipitation performance. The model output (SRF data) was compared to E-OBS[3] gridded data, as well as for nested domain point based (observed time series).

Carrying out such tasks requires high computational and storage resources and various programs & scripts for manipulating on different type formats and projections. An essential part of this study has been performed on GRENA GE-01-GRENA Grid site, by creating and testing most of the post processing scripts and programs. For long period climate simulations, the IICT–BAS Avitohol supercomputing system was used.

The variations in wet days number, also for precipitation sums are maximal during the summer. The model overestimates both parameters for all seasons and over the whole domain, but for other seasons it is in the range of 6-9 wet days. Each picture has orographic features, where differences are dominant and model tend to increase local convective processes.

For the further improvement of the model performance a more precise investigation of sub-grid scale processes, and better cross validation against more sources of observation, such as satellite data, will be benefiting. Bias and noise reduction in the model results can make it reliable for future prediction.



Biases in mean seasonal number of wet days (upper) and mean seasonal precipitation sums (bottom) (units: mm) for the mean of different  model configurations.