SIMULATION OF ATMOSPHERIC STATES FOR THE CASE OF YEONG-GWANG STORM SURGE ON 31 MARCH 2007 : MODEL COMPARISON BETWEEN MM5, WRF AND COAMPS JEONG-WOOK LEE 1 ; KYUNG-JA HA 1* ; KI-YOUNG HEO 1 ; KWANG-SOON PARK 2 ; KI-CHEON JUN 2 1 Division of Earth Environmental System, Pusan National University Busan, Republic of Korea e-mail: jwl1101@pusan.ac.kr ; corresponding author, kjha@pusan.ac.kr * ; kyheo21@pusan.ac.kr 2 Coastal Engineering Research Division, Korea Ocean Research & Development Institute Ansan, Republic of Korea e-mail: kspark@kordi.re.kr; kcjun@kordi.re.kr ABSTRACT The highly qualified information for the sea level pressure and sea surface wind stress is required to predict the storm surge in the Korean Peninsula. The storm surge on 31 March 2007 at Yeong- Gwang was very unusual in the presence over the western coast and in the cause of the usual mesoscale cyclone development. In the present study, we attempt to produce the reliable surface wind, and sea level pressure using the optimal physical parameterizations for the high wind condition in the different els such as MM5, WRF, and COAMPS. To select the optimal physics combination for the high wind, we designed three experiments (EXP1, EXP2, and EXP3) with different parameterization combinations (Eta PBL and Betts-Miller cumulus, Eta PBL and Grell cumulus, and MRF PBL and Kain-Fritsch cumulus) for MM5 and WRF. As a result, the EXP2 and EXP3 showed the good performance in terms of the magnitude and the evolution of the cyclone, respectively. The COAMPS el simulated well the evolution but for a smaller magnitude. The WRF and MM5 with EXP2 physics have an early evolution and a delayed maximum magnitude, respectively. As cause responsible for the meso-β scale cyclone development, the warm advection, moisture flux and its convergence, and convective instability were explored by MM5 and WRF with EXP2. Keywords: Storm surge, sea level pressure, sea surface wind, MM5, WRF, COAMPS 1. INTRODUCTION On 31 March 2007, the storm surge struck the Yeong-Gwang in the cause of the mesoscale cyclone development. The invaded storm surge occurs with the high tide
simultaneously, the sea-surface height was recorded above the high water level culminating 703 cm, which is corresponding to the additional sea-surface height of approximately 200 cm from the height of Yeong-Gwang high tide (Fig. 1) at 1631UTC 30 March 2007. Wave is supposed to be generated by superposition and amplification because of the same speed of cyclone and wave formed by the cyclone (Seo at al., 2007). Sea level height at Yeong-Gwang FIGURE 1 TEMPORAL VARIATION OF SEA LEVEL HEIGHT OBTAINED BY OBSERVATION AT YEONG-GWANG. Figure 2 shows the surface weather chart on 30 March 2007. On 09 UTC 30 March 2007, small scale cyclone and anticyclone were located at Korean Peninsula (Fig. 2a). The quickly generated cyclone as shown in black box of the figure 2b on 15 UTC 30 March 2007 was dissipated and the pattern of air pressure moved eastward on 21 UTC 30 March 2007 (Fig. 2c). (a) 09 UTC 30 Mar. 2007 (b) 15 UTC 30 Mar. 2007 (c) 21 UTC 30 Mar. 2007 FIGURE 2 SURFACE WEATHER CHART ON (A) 09 UTC 30 MARCH 2007, (B) 15 UTC 30 MARCH 2007, AND (C) 21 UTC 30 MARCH 2007.
Generally, the storm surge frequently occurred due to the typhoon over the East or South coasts in the Korean Peninsula. However, this case is exceptional one generated by meso-β scale cyclone with center pressure of 1010hPa at Yeong-Gwang near the western Coast of Korea (Fig. 2b). To predict the storm surge, accurate sea level pressure and sea surface wind are required. Very few studies are found in the literature (Seo and Chang, 2003; Kim et al., 2006) of sea surface wind simulated by regional els near the coast of Korea. Seo and Chang (2003) have analyzed the characteristics of the monthly mean sea surface winds and wind waves near the Korean marginal seas in the 2002 year on the basis of prediction results of the sea surface winds from MM5/KMA el. Kim et al. (2006) carried out sensitivity experiment of sea surface wind based on PBL schemes (Medium-range Forecast, MRF and Mellor-Yamada- Janjic, MYJ) and dynamic frame of MM5 and WRF. In this study, we select the optimal physics combination for the high wind (Davis and Simon, 2001; Ivanov and Palamarchuk, 2007) in Fifth-generation Mesoscale Model (MM5) and Weather Research and Forecasting el (WRF) and examine the results between MM5, WRF and The Coupled Ocean/Atmosphere Mesoscale Prediction System (COAMPS) for the time evolution of surface wind and sea level pressure associated with cyclone. Finally, we try to investigate the cause of the cyclone development. 2. MODELS AND EXPERIMENTAL DESIGN MM5 version 3.7, WRF version 2.2 and COAMPS version 3.1 are used in this study. The nested domains over the center of 35.0 N and 129.0 E are covered with 115 (in longitude) ⅹ 124 (in latitude) grids in the 27 km mesh and 175 (in longitude) ⅹ238 (in latitude) grids in the 9 km mesh systems, respectively. Initial and boundary conditions are based on the NCEP Final Analysis (FNL) for MM5 and WRF and come from NOGAPS and several satellite data of GOES, QuikSCAT, SSMI, and so on for COAMPS. The simulations were executed from 00 UTC 30 March 2007 to 00 UTC 31 March 2007. Except for the COAMPS el, three sets of experiments are carried out to select the physics options which can accurately simulate the magnitude and the evolution of the cyclone for the high wind condition (Table 1). EXP1 and EXP2 are selected from the best physical parameterizations for tropical cyclone prediction (Davis and Simon, 2001) and EXP3 comes from the best optimal parameterization scheme sets for the atmospheric variables such as temperature, geopotential height (Ivanov and Palamarchuk, 2007). The selected physics schemes are applied to the MM5 and WRF el and then the results are compared with the output of COAMPS el. Betts-Miller (Betts and Miller, 1993) scheme is the most popular for tropical systems, but it produces few downdrafts and has a tendency to spinup vortices too easily. With weak tropical cyclones there can be numerous small-scale cyclonic spinups away from the storm center for Grell scheme (Grell, 1993). Grell is the only MM5 scheme that is routinely run near
10 km grid spacing (Davis and Simon, 2001). The Kain-Fritsch scheme (Kain and Fritsch, 1993) has not been run extensively in the tropics. The MRF PBL scheme (Hong and Pan, 1996) tends to produce boundary layers that are too deep and too dry outside the eye wall of mature hurricanes (Braun and Tao, 2000) and may be adequate if storms are weak. TABLE 1 - EXPERIMENTAL DESIGNS FOR PHYSICS OPTIONS. Physics Options EXP1 EXP2 EXP3 PBL Scheme Eta Eta MRF Cumulus Scheme Betts-Miller Grell Kain-Fritsch Explicit Moisture Scheme Mixed phase Surface Scheme Five-Layer Soil Radiation Scheme RRTM longwave scheme 3. EXPERIMENT RESULT (a) Optimal combination of parameterization To select the optimal physics combination for the high wind, we executed three experiments with different parameterization. (b) MM5 EXP1 (c) MM5 EXP2 (d) MM5 EXP3 (a) QuikSCAT (e) WRF EXP1 (f) WRF EXP2 (g) WRF EXP3 FIGURE 3 HORIZONTAL DISTRIBUTION OF SEA SURFACE WIND OBTAINED BY (A) QUIKSCAT AND SIMULATED BY (B) MM5 EXP1, (C) MM5 EXP2, (D) MM5 EXP3, (E) WRF EXP1, (F) WRF EXP2 AND (G) WRF EXP3 ON 09 UTC 30 MARCH 2007.
Figure 3 shows the horizontal distribution of wind speed at 10 m height on 09 UTC 30 March 2007. Figure 3(a) is ervation from the QuikSCAT surface wind speed and Figs. 3(b)-3(g) show the simulated sea surface wind speed using MM5 and WRF with different physics for Table 1. Observation displays strong wind region in the western part of Yellow Sea, South Sea and East Sea. All experiments simulate the strong wind speed region in the western part of Yellow Sea and East Sea while only WRF EXP1 and WRF EXP2 simulate the strong wind speed region in the South Sea. (b) MM5 EXP1 (c) MM5 EXP2 (d) MM5 EXP3 (a) QuikSCAT (e) WRF EXP1 (f) WRF EXP2 (g) WRF EXP3 FIGURE 4 SAME AS FIGURE 3 EXCEPT FOR 21 UTC 30 MARCH 2007. Figure 4 shows the horizontal distribution of wind speed on 21 UTC 30 March 2007. Observation displays strong wind region in the East China Sea, Yellow Sea and East Sea. MM5 simulations show the stronger wind over the northern part of Yellow Sea, compared to the ervation. Overall, wind speed simulated by WRF is stronger than MM5. All experiments simulate the strong wind speed region in the East China Sea except for the MM5 EXP1. In general, EXP2 and EXP3 show better distributions compared to the EXP1, even though the speeds in the western Korean Peninsula are underestimated. For the evaluation of horizontal distribution of simulations, statistics of the BIAS (mean error), RMSE (Root Mean Square Error), SI (Scatter Index), and CORR (Correlation Coefficient) are used (given in Table 2). Where N is the total number of grid point except for the land and missing data. W ( W ) is the wind speed of el (ervation) and
W ( W ) is the mean wind speed of el (ervation) for the area. Table 2. The equations for validation scores. Validation scores BIAS RMSE 1 N 1 N Equation ( W W ) ( W W ) 2 SI CORR 1 N 1 1 2 ( W W BIAS ) N 1 N 1 ( W ( W W W W ) 2 )( W 1 N 1 W ( W ) W ) 2 Table 3 shows the result of statistics over the whole domain on 09 UTC 30 March 2007 and 21 UTC 30 March 2007. TABLE 3 - STATISTICS OF THE BIAS, RMSE, SI, AND CORRELATION OF SIMULATED WIND SPEED BY MM5 AND WRF MODEL OVER THE WHOLE DOMAIN ON 09 UTC 30 MARCH 2007 and 21 UTC 30 March 2007. Model EXP BIAS RMSE SI CORR EXP1-0.64 1.60 0.28 0.78 09 UTC 30 MM5 EXP2 EXP3-0.49-0.64 1.54 1.50 0.28 0.26 0.79 0.81 March EXP1-0.20 1.64 0.31 0.71 2007 WRF EXP2 EXP3-0.47-0.57 1.60 1.59 0.30 0.29 0.76 0.77 21 UTC 30 March 2007 MM5 WRF EXP1 EXP2 EXP3 EXP1 EXP2 EXP3 0.41 0.55 0.16 0.15 0.59 0.44 2.67 3.11 2.87 2.61 2.64 2.65 0.37 0.43 0.40 0.37 0.36 0.37 0.44 0.35 0.40 0.53 0.56 0.51 All experiments have negative values on 09 UTC 30 March, while all experiments have positive values on 21 UTC 30 March for the BIAS. It means that wind speeds of els
are underestimated on 09 UTC and overestimated on 21 UTC. RMSE and SI have the lower values and CORR has the higher value for the EXP3 than other experiments on 09 UTC 30 March 2007. EXP1 and EXP2 have the better value in terms of the SI and CORR on 21 UTC 30 March 2007 for the MM5 and WRF, respectively. Overall, EXP2 showed the good performance in terms of the distribution and the magnitude of wind speed, especially in the case of strong wind. (b) Model comparison To examine the simulations of cyclone development, three runs of MM5 with EXP2, WRF with EXP2, COAMPS are compared. Horizontal distribution of wind speed and sea level pressure simulated by three els on 16 UTC 30 March 2007 is shown in Fig. 5. Models simulate the cyclone in the southern part of western Korean Peninsula, and the result shows the anticyclone in the north of the cyclone as shown in black box of the figure. The WRF well simulates the developed meso-β scale cyclone whereas COAMPS shows the cyclone westward from the south-western part of Korean Peninsula. The distribution of wind speed is well corresponding to that of the sea level pressure. (a) MM5 EXP2 (b) WRF EXP2 (c) COAMPS FIGURE 5 HORIZONTAL DISTRIBUTION OF WIND SPEED AND SEA LEVEL PRESSURE SIMULATED BY (A) MM5 EXP2, (B) WRF EXP2 AND (C) COAMPS AT 16 UTC 30 MARCH 2007.
Figure 6 displays the time series of wind speed and sea level pressure obtained by ervation and simulations by MM5, WRF and COAMPS el. The most strong wind speed occurs at the 16 UTC 30 March. Before that maximum peak, wind speed shows weak peak on 04 UTC and 08 UTC 30 March and it falls to the zero at 12 UTC 30. Maximum wind speed arrives at the time of 16 UTC 30 March, showing the rapid development with about 12m/s speed for five hours. The COAMPS el simulated well the evolution but for a smaller magnitude. The WRF and MM5 with EXP2 physics have an early evolution and a delayed maximum magnitude, respectively. As time goes, simulated sea level pressures show the stronger second peak after the first peak, compared to that of ervation. (a) Wind speed (10 m) (b) Sea level pressure FIGURE 6 TEMPORAL VARIATION OF (A) WIND SPEED AND (B) SEA LEVEL PRESSURE OBTAINED BY OBSERVATION AND SIMULATIONS WITH MM5, WRF, AND COAMPS AT YEONG-GWANG. 4. THE CAUSE OF THE CYCLONE DEVELOPMENT We compared the atmospheric state to investigate the cause of the cyclone development. Low-level warm air advection, convective instability, low-level moisture flux and its convergence and longitude-height cross-section of temperature were investigated.
Figure 7 shows the horizontal distribution of temperature advection at 900hPa. Figure 7(a) and Fig. 7b are analysis data from the NCEP Final Analysis (FNL) and Figs. 7(c)-7(h) show the simulated temperature advection using MM5, WRF and COAMPS. To measure the temperature advection, we used Eq. (1) v T (1) [ⅹ10-3 K s -1 ] FIGURE 7 - HORIZONTAL DISTRIBUTION OF WARM AIR ADVECTION (ⅹ10-3 K s -1 ) AT 900hPa SIMULATED BY (A) ANALYSIS DATA ON 06 UTC 30 MARCH 2007, (B) ANALYSIS DATA ON 12 UTC 30 MARCH 2007, (C) MM5, (D) WRF AND (E) COAMPS ON 06 UTC 30 MARCH 2007 AND (F) MM5, (G) WRF AND (H) COAMPS ON 12 UTC 30 MARCH. The region of warm advection over 3.0ⅹ10-3 K s -1 is shaded. Analysis data showed that warm advection region is situated in the Eastern part of the China with the core (116 E, 35 N) on 06 UTC 30 March 2007 (Fig. 7a). As time goes on, the region of warm advection moves eastward and extends to the broad area. On 12 UTC 30 March 2007, the warm
advection arrived in the Korean Peninsula and the core is located at 120.5 E, 35.5 N (Fig. 7b). All els well simulated the region of warm advection and the core compared with the analysis data. On 06 UTC 30 March 2007, COAMPS el showed that the warm advection region is located at more eastward to the Korean Peninsula and the core is weaker than MM5 and WRF. It is proved that WRF simulated broad region of warm advection over 9 ⅹ10-3 K s - 1 compared with MM5 and COAMPS in the Yellow Sea on 12 UTC 30 March 2007. To measure the convective instability, the gradient between equivalent potential temperature ( θ e ) at 900 hpa and θ e at 850 hpa are calculated (Fig. 8). Where the positive value means stable and the negative value means unstable. [ K ] FIGURE 8 - SAME AS FIGURE 7 EXCEPT FOR CONVECTIVE INSTABILITY (DIFFERENCE BETWEEN EQUIVALENT POTENTIAL TEMPERATURE ( θ e ) at 850 hpa AND θ e at 900 hpa, K). Analysis data showed that stable region is located from Eastern part of China to Yellow Sea and instable region is located on the Southeastern part of China and southwestern part of Korean Peninsula on 06 UTC 30 March 2007 (Fig. 8a). As time goes, the pattern
moves north-eastward and stable region is destabilized in the Yellow Sea (Fig. 8b). All els well simulated the stable region while all els had the large value of convective instability over -10K compared with the analysis data, especially in COAMPS el. Convective instability fields occur at the western Korean peninsula, which is corresponding to the area of downstream of warm advections. And the warm air advection destabilizes the low layer from 900hPa to 850hPa. [ ms -1 ] [ⅹ10-6 s -1 ] FIGURE 9 - SAME AS FIGURE 7 EXCEPT FOR MOISTURE FLUX (VECTOR, m s -1 ) AND ITS CONVERGENCE (SHADED, ⅹ10-6 s -1 ). Figure 9 shows the horizontal distribution of moisture flux and its convergence at 900hPa. Moisture flux and its convergence are calculated by v q and ( q v), respectively. The region of moisture convergence over 2.0ⅹ10-6 s -1 is shaded. Analysis data showed that the convergence of moisture flux region is situated in the Eastern part of the China on 06 UTC 30 March 2007 (Fig. 9a). As time goes, the region of moisture convergence
moves north-eastward and extends to the broad area. On 12 UTC 30 March 2007, the moisture flux is transported north-eastward and the convergence of moisture flux region is generated over the Yellow Sea (Fig. 9b). All els well simulated the moisture flux and moisture convergence region compared with the analysis data. The strong moisture convergence merges to the warm air advection area. Low-level warm advection destabilizes the low-level, and then moisture flux contributes to enhance the development of the cyclone. It has been shown that MM5 and WRF simulated broad region of moisture convergence over 6.0ⅹ10-6 s -1 compared with the COAMPS. FIGURE 10 - LONGITUDE-HEIGHT CROSS-SECTION OF TEMPERATURE ( C) AT 35.25N FROM 116E TO 130E SIMULATED BY (A) ANALYSIS DATA ON 06 UTC 30 MARCH 2007, (B) ANALYSIS DATA ON 12 UTC 30 MARCH 2007, (C) MM5, (D) WRF AND (E) COAMPS ON 06 UTC 30 MARCH 2007 AND (F) MM5, (G) WRF AND (H) COAMPS ON 12 UTC 30 MARCH. We also compared the longitude-height cross-section of temperature at 35.25 N from 116 E to 130 E to analysis the cause of the cyclone development (Fig. 10). Almost all of warm cores are existed at 900 hpa. There are two warm cores (116 E-124 E and 126 E-130 E) on 06 UTC 30 March 2007 and the cores are merged and developed on 12 UTC 30 March [ C ]
2007 in analysis data. All els well simulated the warm core compared with the analysis data, especially in COAMPS el. On 06 UTC 30 March 2007, all els had the core from 116 E to 125 E and COAMPS had another core from 126 E to 127.5 E. As time goes, the core was developed and extended eastward. 5. DISCUSSION Yeong-Gwang case on 31 March 2007 is very exceptional one that generated by meso-β scale cyclone with center pressure of 1010hPa. To predict the storm surge, accurate sea level pressure and sea surface wind are required. In this study, the atmospheric conditions of the storm surge case are simulated according to the different el of MM5, WRF and COAMPS and physics options. And we selected the physics options as the optimal parameterizations for the high wind condition and investigated the cause of the cyclone development. At first, EXP1 (Eta PBL and Betts-Miller cumulus), EXP2 (Eta PBL and Grell cumulus), EXP3(MRF PBL and Kain-Fritsch cumulus) with MM5 and WRF were designed for the optimal combination of physics. EXP2 and EXP3 show better distributions of strong wind speed region compared to the EXP1. According to the statistics for the evaluation of horizontal distribution of simulations, EXP1 and EXP2 have the better value for MM5 and WRF, respectively. Although there is a somewhat difficulty in selecting the optimal physics options, EXP2 showed the good performance in terms of the distribution and the magnitude of wind speed. It has been shown that different els make much of a difference on the magnitude and the evolution of the cyclone. The results for three runs of MM5 with EXP2, WRF with EXP2 and COAMPS showed that the distribution of wind speed is well corresponding to that of the sea level pressure. The WRF and MM5 with EXP2 physics have an early evolution and a delayed maximum magnitude respectively while COAMPS el simulated well the evolution but for a smaller magnitude. The cause of the cyclone development is considered by multiple components of warm advection, moisture flux and its convergence, and convective instability. The results of the numerical simulations performed in this study showed that WRF simulated well the developed meso-β scale cyclone. WRF simulation is stronger than MM5 and COAMPS in the western Korean peninsula in terms of warm advection and moisture convergence. Lowlevel warm advection destabilized the low-level, and then moisture flux contributes to enhance the development of the cyclone.. Acknowledgement This work was supported by Top-Brand project of KORDI and the Brain Korea 21 Project in 2006/7.
REFERENCES 1. A. K. Betts and M. J. Miller, The representation of cumulus convection in numerical els of the atmosphere, The Betts-Miller scheme, 246pp, 1993, Amer. Meteor. Soc.. 2. C. A. Davis and S. Low-Nam, The NCAR-AFWA Tropical Cyclone Bogussing Scheme, A Report Prepared for the Air Force Weather Agency (AFWA), National Center for Atmospheric Research Boulder, Colorado, 2001, pp. 13. 3. G. A. Grell, Prognostic evaluation of assumptions used by cumulus parameterizations, Mon. Wea. Rev. 121 (1993), 764-787. 4. J. S. Kain and J. M. Fritsch, The representation of cumulus convection in numerical els, Convective parameterization for mesoscale els: The Kain-Fritsch scheme, Eds. K.A. Emanuel and D.J. Raymond, Eds., Amer. Meteor. Soc., 1993, pp. 246. 5. J.-W. Seo and Y.-S. Chang, Characteristics of the Monthly Mean Sea Surface Winds and Wind Waves near the Korean Marginal Seas in the 2002 Year Computed Using MM5/KMA and WAVEWATCH-Ш el, Journal of the Korean Society of Oceanography, 8(2003), 262-273. 6. S.-Y. Hong and H.-L. Pan, Nocturnal boundary layer vertical diffusion a medium-range forecast el, Mon. Wea. Rev. 124 (1996), 2322-2339. 7. S. A. Braun and W.-K. Tao, Sensitivity of High-Resolution Simulations of Hurricane Bob (1991) to Planetary Boundary Layer Parameterizations, Mon. Wea. Rev. 128 (2000), 3941-3961. 8. S.-N. Seo, S.-G. Kim, K.-S. Park, and Y.-J. Yeon, Analysis of Storm Surges over Southwest Coasts of Korea in the year 2007, Proc. The Korean Association of Ocean Science and Technology Societies Joint Meeting, Seoul, 2007, pp. 2290-2293. 9. S. Ivanov and Y. Palamarchuk, Systematic Error of Parameterization Schemes in the MM5 Model, Proceedings of The 3rd WGNE Workshop on Systematic Errors in Climate and NWP Models, 12-16 Februrary, San Francisco, 2007, pp. 49. 10. Y.-K. Kim, J.-H. Jeong, J.-H Bae, I.-B Oh, J.-H Kweon and J.-W Seo, Improvement in the Simulation of Sea Surface Wind over the Complex Coastal Area Using WRF Model, Journal of the Korean Society for atmospheric Environment, 22(2006), 309-323.