1. The National Environmental Agency of Georgia L. Megrelidze, N. Kutaladze, Kh. Kokosadze NWP Local Area Models’ Failure in Simulation of Eastern Invasion of Cold Air Masses in Georgia NWP Application in Meteorology 2009

2. Evaluation of NWP LAM skills for simulation of synoptic situations related to eastern invasion of air masses in Georgia Case of 05.11.2009 Global and mesoscale model performances gave different outputs: EMS NMM and HRM, which run operationally, 3 days ahead predicted rainfall over the most of part of Georgia whereas global models (GFS, ARPEGE) outputs as well observations were different. Objective

3. Main types of synoptic processes and concerning weather conditions on the South Caucasus territory are following:  western type  eastern type  anticyclonic situation  frontal activities from the South The capture of eastern and southern processes is the most difficult issue for NWP models, when main difficulties raised during the prediction of spatial-temporal distribution of precipitation fields.

5. Data and methods Observation (only surface: Temperature, Precipitation); Satellite images; Analysis maps (surface, 500, 700, 500/1000 mb) Global models: GFS, ARPEGE LAMs:  Environmental Modeling System Non-hydrostatic Mesoscale Model (EMS NMM) - initialized directly from the National Centers for Environmental Prediction Global Forecast System (GFS) with resolution 12 km and 31 vertical levels (operational);  High Resolution Model (HRM) - initialized from GME with resolution 14 km and 40 vertical levels (operational);  WRF ARW - initialized from GFS with resolution 15 km (research) and 27 vertical levels;  No: Data assimilation  No: Objective verification

6. Data and methods  Outer domain over the Caucasus region with 167x117 points in the north-south and east-west directions respectively  Nested domain fixed in space over Georgia and contained 145x115 points in the north-south and east-west directions respectively on a 5-km mesh The forecasts were integrated beginning at 00 UTC 72, 48 and 24 hours ahead of invasion using:  1-way (no feedback) and  2-way nesting runs, also with  three-dimensional analysis nudging for both domains, with identical nudging coefficients for u, v, theta and q. but for both domains within the PBL the analysis nudging were turned off for temperature and moisture fields and within the lowest 10 model layers for wind

7.  Microphysics: WSM 3 (5) - class simple ice scheme with ice and snow processes suitable for mesoscale grid sizes.;  Cumulus Parameterization : Kain-Fritsch (new Eta) scheme: Deep and shallow sub-grid scheme using a mass flux approach with downdrafts and CAPE removal time scale Data and methods

8. 5 th of November  In surface layers - cold air masses propagation from European territory to the southward direction and their invasion in South Caucasus, mostly from the East.  In upper levels of atmosphere (700, 500 mb) - south-western currents  It was not pure eastern type, but with influence from the south-west, characterized with rainfall also in West Georgia

9. WRF ARW: mean sea level pressure based 00Z 03/11/2009 T+48 VT 00z05nov2009 Surface analysis map for 00z05nov2009500mb analysis map for 00z05nov2009 WRF ARW: 500 mb geopotential height based 00Z 03/11/2009 T+48 VT 00z05nov2009

10. Fig. 2b. GFS: Relative humidity at 850 mb based 00Z 03/11/2009 T+48 VT 00z05nov2009 Fig. 2a. Eumetsat 0 degree MPE image for 00z05nov2009 GFS vs. Satellite

11. .ARW_outer domain: 12-h precipitation based 00Z 03/11/2009 T+48 VT 18z04nov-06z05nov2009 ARW_inner domain: 12-h precipitation based 00Z 03/11/2009 T+48 VT 18z04nov-06z05nov2009 ARW_outer domain: 12-h precipitation based 00Z 04/11/2009 T+24 VT 18z04nov-06z05nov2009 ARW_inner domain: 12-h precipitation based 00Z 04/11/2009 T+24 VT 18z04nov-06z05nov2009 WRF ARW vs. Observation Observation: 12-h precipitation 18z04nov- 06z05nov2009

12. Conclusions  precipitation was forecasted 12 hours earlier and overestimated  capture of factual situation WRF ARW could only 24 hours ahead  for 3-day forecast nesting does not improve precipitation forecast: overestimation was increased  for 2-day forecast there was some improvement using nesting: more detailed precipitation distribution  relatively better results (among all simulations) have been obtained using analysis nudging