Asia-Pacific J. Atmos. Sci., 48(3), 259-273, 2012 DOI:10.1007/s13143-012-0026-2

The Effect of Topography and Sea Surface Temperature on Heavy Snowfall in the Yeongdong Region: A Case Study with High Resolution WRF Simulation

Sun-Hee Jung, Eun-Soon Im, and Sang-Ok Han National Institute of Meteorological Research, Korea Meteorological Administration, Korea

(Manuscript received 4 November 2011; revised 5 April 2012; accepted 6 April 2012) © The Korean Meteorological Society and Springer 2012

Abstract: An analysis of the heavy snowfall that occurred on 11-14 to traffic flow, agriculture, and fishery. The total snow amount February 2011 in the Yeongdong region along the eastern coast is in Donghae (DH) was 134.7 cm (11-14 Feb.), and the 24-hour presented. Relevant characteristics based on observation and model accumulated fresh snow in Gangneung (GN) was 77.7 cm (11 simulations are discussed with a focus on the times of maximum Feb.), which is the highest value recorded since observations snowfall in Gangneung (GN) and Daegwallyong (DG). This event was considered part of the typical snowfall pattern that frequently started in 1911. This event, considered the heaviest snowfall in a occurs in the Yeongdong region due to the prevailing northeasterly century on South Korea’s east coast, paralyzed the community flow. The control simulation using the high resolution Weather and caused widespread chaos. Hundreds of houses collapsed Research and Forecasting (WRF) model (1 km × 1 km) showed under the weight of the snow while hundreds of motorists were reasonable performance in capturing the spatial distribution and stranded in deep drifts. The cost of the damage was expected temporal evolution of precipitation. The area of precipitation maxima to run to approximately US 65,000,000 dollars (http://www. appeared to propagate from the plain coastal region further into the safekorea.go.kr). inland mountainous region, in relation to the location of convergence zone. In addition, a series of sensitivity experiments were performed According to the brief press release by the Korea Meteor- to investigate the effect of topography and sea surface temperature ological Administration, several typical factors contributed to (SST) on the formation of heavy snowfall. The change of topography this unprecedented event. First, a well developed Siberian High tended to modulate the topographically induced mechanical flow, expanding over the East Sea formed a synoptic pressure pattern and thereby modify the precipitation distribution, which highlights that blew the cold northeasterly winds onto the eastern coastal the importance of an elaborate representation of the topography. On region. Secondly, as this continental cold air mass persistently the other hand, the sensitivity experiment to prescribe positive advected and passed over the relatively warm sea surface, it (negative) SST forcing shows the enhanced (suppressed) precipi- tation amount due to the change of the sensible and latent heat was rapidly modified by large oceanic heat and moisture fluxes. fluxes, which enhanced the vertical instability. In addition, the Low pressure located in the southeastern Sea of Japan played a Key words: Heavy snowfall, topography and SST effect, WRF role in keeping the persistent northeasterly flow in the eastern simulation coastal region from blocking the closed meso-scale Low developing at the eastern Sea of the Korean peninsula. These 1. Introduction factors combined to provide a favorable condition for extended heavy snowfall. The Yeongdong region frequently suffers various severe Several previous studies have examined heavy snowfalls in weather events such as heavy precipitation and downslope the Yeongdong region based on both observations and windstorm, mostly due to the combined effect of its steep numerical model experiments. Most of these studies indicated mountain slopes (Taebaek mountain range) and close proximity that the orographic lifting due to the Taebaek mountain range to the ocean (East Sea) (Chung et al., 2004; Kim et al., 2005; and the abundant moisture and heat from the East Sea were the Kim and Chung, 2006; Lee et al., 2006; Han and Lee 2007; causes of the more frequent and heavier snowfall compared to Lee and Kim 2008b; Chang et al., 2009; Lee and In, 2009). Of other regions. Lee and Kim (2008a) postulated topographic particular concern is the heavy snowfall during the winter effect as the key factor in the formation of heavy snowfall in season because of its frequent occurrence and negative impacts the Yeongdong region through an experiment that removed the on the ecosystem and economy. Recently, exceptionally heavy topography over the Taebaek Mountains. Lee and Lee (1994) snowfall occurred on 11-14 February 2011 in the Yeongdong also showed that the height of the topography significantly region along the eastern coast, resulting in tremendous damage affects the amount of snowfall. On the other hand, Ahn and Cho (1998) highlighted the importance of sea surface tempera- ture (SST) in the simulation of heavy snowfall events in the Corresponding Author: Eun-Soon Im, 401 Education Service Center, Gangnueng-Wonju National University, 7, Jukheon-gil, Gangneung- Yeongdong region based on a meso-scale model experiment. si, Gangwon-do 201-702, Korea. In this study, we attempt to simulate the snowfall event that E-mail: [email protected] occurred on 11-14 February 2011 in the Yeongdong region 260 ASIA-PACIFIC JOURNAL OF ATMOSPHERIC SCIENCES

Fig. 1. The model domain and topography (upper panels), and vertical transects of the surface elevation along the line between points A and B (lower panels) used for the CONT (a, c) and EXP_T1 (b, d) simulations. Topography is represented with shading based on scale at right of the (b). using the high resolution Weather Research and Forecasting Table 1. Summary of numerical experiments. (WRF) system (1 km × 1 km). The model results are evaluated Experiment Topography SST by comparison with station observations in terms of spatial distribution and temporal evolution. Assessing the model sys- CONT 30sec. resolution data set Default tem’s capability of capturing observed features can provide EXP_T1 10-min. resolution data set Default some confidence in interpreting the following sensitivity experi- EXP_S1 Same as CONT + 2 K ments. To better understand the effect of the topography and EXP_S2 Same as CONT − 2K SST on the formation of heavy snowfall, a series of ex- periments are performed. In the sensitivity experiment for EXP_S3 Same as CONT No horizontal gradient (SST = constant) topography effect, the topography is smoothed by using 10-min United States Geological Survey (USGS) dataset (EXP_T1) while the control experiment (CONT) uses 30-sec data, resultant snowfall intensity. resulting in different topographic features despite the same In section 2 we briefly describe the model configuration and resolution of the two experiments (See Fig. 1). In the sensitivity experiment design. The synoptic condition and the relevant experiment investigating the SST effect, SST is uniformly characteristics of the snowfall event are explained in section 3. increased (decreased) by 2 K over the ocean areas within the The results for the control (CONT) and four kinds of sensitivity interior domain when interpolating the initial and lateral experiment (EXP_T1 and EXP_S1-3) were then validated and boundary conditions (EXP_S1 & EXP_S2). Additionally, a compared in section 4. Finally, the summary and discussion sensitivity experiment is performed to prescribe the SST as a are presented in section 5. constant in order to examine the effect of horizontal gradient of the SST distribution (EXP_S3, See Table 1). By comparison 2. Model configuration and experiment design with CONT, EXP_T1 present an ideal of the topography effect while EXP_S1-3 reveals the influence of SST condition on the The numerical model used in this study is the WRF (version 31 August 2012 Sun-Hee Jung et al. 261

3.2.1) described by Skamarock et al. (2008). The WRF model Iacono et al., 2008), Goddard shortwave radiation scheme (Tao is a next-generation meso-scale numerical weather prediction et al., 1989), the Double-Moment (WDM) 6-class microphysics system designed to serve both operational forecasting and scheme (Lim and Hong, 2010), and none cumulus parameter- atmospheric research needs. The Advanced Research WRF ization scheme. The WDM 6-class scheme consists of six solver developed at the National Center for Atmospheric Re- hydrometeors: vapor, cloud water, cloud ice, rain, snow, and search was used for the dynamic core, which is a fully com- graupel. This scheme is considered the most suitable for cloud- pressible and non-hydrostatic model. The physical parameter- resolving grid. Through various sensitivity experiments, we izations employed in this simulation include the 5-layer thermal determined the optimal selection and combination among the diffusion land-surface model (Chen and Dudhia. 2001), the variety of physical parameterizations. Yonsei University (YSU) planetary boundary layer scheme Figure 1a shows the model domain and topography for the (Hong et al., 2006), the Rapid Radiative Transfer Model CONT experiment. The domain focused on the eastern part of (RRTMG) longwave radiation scheme (Malwer et al., 1997; Korean peninsula where the snowfall event selected in this

Fig. 2. Spatial distribution of SST used in (a) CONT, (b) EXP_S1, (c) EXP_S2 and the difference field between CONT and (d) EXP_S3. 262 ASIA-PACIFIC JOURNAL OF ATMOSPHERIC SCIENCES study was concentrated. The horizontal resolution was 1 km × SST affects the simulation of the vertical and horizontal 1 km with 400 × 400 grid points while 31 vertical levels were meteorological fields and the resultant snowfall amount. employed up to 50 hPa. We also carefully selected the domain The initial and time-dependent lateral boundary conditions area through several sensitivity tests since the simulation were interpolated using the Korea Local Analysis and Predic- results may have been affected by the domain size and location tion System (KLAPS) with a horizontal resolution of 5 km × of the lateral boundaries (Seth and Giorgi, 1998). This domain 5 km at 1-hour intervals (Hwang et al., 2011). The integration was determined as a suitable compromise between the accuracy spanned from 0000 UTC on 10 February to 0000 UTC on 15 and computational efficiency. February, 2011, when the record-breaking snowfall event hit In addition to the CONT simulation, a series of sensitivity the target region. experiments were performed and major features of the experi- ments are summarized in Table 1. The first sensitivity 3. Observational synoptic condition and general charac- experiment (EXP_T1) used smoothed topography from the 10- teristics min resolution dataset produced by USGS while CONT used 30-sec data to setup the topography. Since topography exerts a We begin our analysis with a discussion on the synoptic strong dynamic forcing, accurate representation plays a key overview based on the observation. Figure 3 shows the surface role in intensifying the vertical motion and resultant snowfall. weather chart at 1500 UTC 11 February (a) and 0000 UTC (b) Both CONT (Fig. 1a) and EXP1_T1 (Fig. 1b) topography 12 February 2011 when the snowfall reached the peak in reasonably represented the main mountainous feature with the Gangneung (GN) and Daegwallyong (DG), respectively (Figs. most prominent ranges extending from north to south along the 8a and 8c). A well developed Siberian High was dominant eastern coastal regions (Taebaek Mountains). However, signifi- across the huge continent of East Asia while a Low pressure cant finer-scale details were only captured by the CONT accompanied with a front was located over the southeastern Sea topography (a). Therefore, the orographic features were quite of Japan. Such a synoptic map, the so-called “Northern High different despite the same resolution (1 km × 1 km) of the two Southern Low”, is a typical pattern to maintain northeasterly experiments. Compared to the vertical cross-section of topog- over the Yeongdong region (Lee et al., 2011). Although the raphy (Figs. 1c and 1d), this characteristic was more clearly general patterns between two synoptic maps appear to be revealed and further emphasized the necessity of detailed similar, several detailed features are different. Compared to orographic features in a topographically diverse region. Note 1500 UTC 11 February, the Siberian High extended eastward that even the topography in the 1-km resolution of CONT (Fig. and a more intense pressure gradient was found over the 1c) did not fully resolve the fluctuating feature of elevation, as Korean peninsula at 0000 UTC 12 February. The isobars at observed in the 30-sec resolution global dataset produced by 0000 UTC were also tilted rather perpendicularly over the East USGS (not shown). However, this is currently considered the Sea, which is directly related to the wind direction (See Fig. 9). state-of-the-art resolution for numerical simulation. The Low pressure around the southeastern Sea of Japan was The second type of sensitivity experiment was performed by deepening and moving toward the Western Pacific. In spite of varying the SST condition. We prescribed the SST over the the eastward propagation of this Low, the meso-scale Low ocean areas uniformly by adding and subtracting 2 K from the developing over the eastern part of the East Sea appeared to be SST used for the CONT experiment when interpolating the blocked. This closed Low was a main factor that maintained initial and boundary conditions. Based on the variation of SST the strong and prevailing northeasterly in the Yeongdong in February during the period 1981-2012 (not shown), the SST region (Park et al., 2009), and thus caused the long-lasting mostly varied in the range of −1 K to 1 K which implied that snowfall. Since the prevailing northeasterly can reach further abnormally warm (cold) SST was above (below) 2 K. Figure 2 inland, the observed area of the snowfall maxima moved to presents the spatial distribution of SST used as the initial inland mountainous region (e.g., DG) from the eastern coastal condition for the CONT (a), EXP_S1 (b) and EXP_S2 (c) region (e.g., GN and DH). experiments. We also performed the sensitivity experiment to Figure 4 shows observational surface wind fields based on prescribe the SST as constant, which is the area-averaged value station data corresponding to the synoptic patterns shown in over the ocean area. In case of the initial condition, EXP_S3 Fig. 3. Since the distances between the stations were much prescribed the SST as 280 K without any spatial gradient. In coarser than those of model grid and only several observational that case, the SST of the north area was higher than that of stations were included in the model domain, it was rather CONT, while the SST of the south area was lower than that of difficult to compare this with simulated wind field (Fig. 9) for CONT region (Fig. 2d). SST is an important factor because the accurate quantitative validation. However, observational wind ocean is the source of moisture and heat for snowfall formation. fields can provide the reference for the qualitative behavior of Higher SST can modify the vertical fluxes of heat and mois- the change of wind direction in the Yeongdong region. The ture in the atmospheric boundary and produce convection (Cha change of wind direction at 1500 UTC 11 February (a) and et al., 2011), which increases the intensity of the snowfall. 0000 UTC 12 February (b) 2011 well explains the movement Interpretation of the simulations with different SST conditions of the maximum snowfall. The northwesterly that appeared in promises to increase our understanding of how higher (lower) GN was caused by the northeasterly that was blocked due to 31 August 2012 Sun-Hee Jung et al. 263

Fig. 3. Synoptic surface weather chart at 1500 UTC 11 February (upper) and 0000 UTC 12 February (lower) 2011. 264 ASIA-PACIFIC JOURNAL OF ATMOSPHERIC SCIENCES

Fig. 4. Surface wind field from observation valid at (a) 1500 UTC 11 February and (b) 0000 UTC 12 February 2011.

Fig. 5. 850 hPa equivalent potential temperature (θe) at (a) 0600 UTC, (b) 1500 UTC 11 February, and (c) 0000 UTC 12 February 2011. Blue and red arrows schematically represent the cold and warm advections, respectively. The red dotted line is the reference line of 276 K. the mountain barrier when the maximum snowfall occurred temperature at 1500 UTC 11 February became more dense along the plain coastal region (Fig. 4a). However, after 9 hours (Fig. 5b). The wedge form of equivalent potential temperature (Fig. 4b), the prevailing northeasterly in the Yeongdong region (especially 276 K) gradually propagated to the east coast over supported the movement of the snowfall maximum in further time. Eventually, the reference line of 276 K became located inland. toward southeastern part of the Korea peninsula at 0000 UTC The intensification of snowfall at 1500 UTC 11 February 12 February (Fig. 5c). This implies that the encounter area and 0000 UTC 12 February can also be explained by the field between the cold air advected from the north and the warm and of 850 hPa equivalent potential temperature derived from humid air advected from the East Sea moved to along the east KLAPS which are used as initial and lateral boundary con- coast and thereby provided a favorable condition for heavy ditions (Fig. 5). Compared to 9-hour before (0600 UTC 11 snowfall. February, Fig. 5a), the gradient of the equivalent potential Figure 6 presents the spatial distribution of 3-hour accumu- 31 August 2012 Sun-Hee Jung et al. 265

Fig. 6. Spatial distribution of the 3-hour (1300-1500 UTC and 2200-2400 UTC 11 Feb) and 4-day (11-14 Feb) accumulated precipitation derived from observations. Precipitation is represented with shading based on the scale at right of (c). Here, GN, DH, and DG indicate the observational stations at Gangneung, Donghae, and Daegwallyong, respectively. lated precipitation from 1300 to 1500 UTC 11 February (a), 4. Results and 2200 to 2400 UTC 11 February (b), and 4-day accumu- lated precipitation during 11-14 February 2011 (c). By compari- a. Validation of the control experiment son with (a) and (b), the precipitation band evolved over time, moving to the maximum location from GN to DG, which Figure 7 presents the spatial distribution of 3-hour accumu- corresponded to the synoptic condition shown in Fig. 3. The lated precipitation at 1300 to 1500 UTC 11 February (a), and distribution of total accumulated precipitation showed a gradient 2200 to 2400 UTC 11 February (b), and 4-day accumulated pattern, with decreasing amount of precipitation towards inland. precipitation during 11-14 February 2011 (c) derived from the This event broke several records. The 24-hour accumulated CONT simulation. These model results were qualitatively in fresh snow amounts in GN observational station were 77.7 cm good agreement with the observed estimates as shown in Fig. and 49 cm on 11 and 12 February, respectively, which were the 6. For the accumulated pattern during 1300 to 1500 UTC 11 first and second highest since records began there back in February, the model simulated more precipitation at GN than 1911. In particular, the new first rank 77.7 cm snowfall was at DG, showing an intensive precipitation band along the coast much higher than the previous highest of 48.5 cm (12 Dec. sea. However, after 9 hours, the maximum location of pre- 2008). The 24-hour accumulated fresh snow (11 Feb.) in DH cipitation moved to DG, in line with the observed distribution. observational station at 70.2 cm was also the highest ever Also captured is the gradient pattern of precipitation amount recorded. The records of maximum depth of snow in GN (12 such as less precipitation toward inland for the 4-day total Feb.) and DH (14 Feb.) were also all-time records. accumulated precipitation during 11-14 February. In order to provide a more quantitative measure of the per-

Fig. 7. Same as Fig. 6, except for the CONT simulation. 266 ASIA-PACIFIC JOURNAL OF ATMOSPHERIC SCIENCES

Fig. 8. Time-series of 3-hour accumulated precipitation from 11 to 14 February 2011 derived from the observation and CONT simulation at Gangneung (GN; upper), Donghae (DH; middle), and Daegwallyong (DG; lower). formance of the model, we examined the temporal evolution of and simulated estimates. For GN and DH, the model tended to the three stations hit by the heavy snowfall. Figure 8 presents underestimate the maximum amount. In particular, the maxi- the time-series of every 3-hour accumulated precipitation GN, mum amount of GN was only approximately half of the DH, and DG stations, derived from the observations and observation. On the other hand, the model exactly captured the CONT simulation. We compared simulated precipitation with timing and amount of the maximum at DG, while a large error individual station values using the grid points closest to the occurred in the second maximum when the observation stations. The relatively high model resolution (1 km) justified showed very weak intensity. the comparison between station data (Im et al., 2008) and To explain the synoptic condition to derive this movement of closest grid point model data. The temporal evolution exhibited the precipitation band, we examined the surface wind and a bimodal structure having two peaks with different maximum streamline at the same time of the observed surface weather amounts, regardless of the stations. Generally, the model chart, as shown in Fig. 3 from the CONT simulation (Fig. 9). showed good phase coherence with the observed variation. The The prevailing winds (Figs. 9a and 9c) tended to be approxi- correlation coefficients at GN, DH, and DG were 0.70, 0.72, mately parallel to the isobars presented in the observed surface and 0.70, respectively. The model demonstrated the capability weather chart. When the snowfall peaked along the Yeongdong to capture the timing of maximum precipitation despite the coastal area (e.g., GN and DH), rather than the ridge of the presence of quantitative discrepancies between the observed mountain (DG) (1500 UTC 11 February, Fig. 9a), the north- 31 August 2012 Sun-Hee Jung et al. 267

Fig. 9. Surface wind fields (left) and streamline (right) at 1500 UTC 11 February (upper panels) and 0000 UTC 12 February (lower panels) 2011 from the CONT simulation. easterly appeared to be blocked due to the mountain barrier magnitude. The magnitude of the prevailing northeasterly and the wind direction turned to northwesterly along the foot became intense approximately perpendicular to the Yeongdong of the mountains. This cold northwesterly flow then merged coast, so that it reached far toward inland. This difference with the relatively warm and moist northeasterly or easterly underlying the synoptic background was attributed to the advecting from further offshore areas. This collision induced change of the area in which the snowfall maxima occurred. heavy snowfall along the coast by providing a favorable condition for the convergence (Alestalo et al., 1985; Yeh and b. Analysis of the sensitivity experiment for topography effect Chen., 2002). Indeed, the streamline at the same time (Fig. 9b) clearly reveals the discontinuity line, implying the convergence Figure 10 presents the 4-day total accumulated precipitation zone along the coastal area. Moving to the peak time at DG (11-14 Feb.) simulated by EXP_T1 (a) and their differences (Figs. 9c and 9d), the discontinuity line stepped forward further from the CONT simulation (b). By comparison, although the inland, possibly due to the change of wind direction and area-averaged amount of precipitation (over the whole domain) 268 ASIA-PACIFIC JOURNAL OF ATMOSPHERIC SCIENCES

Fig. 10. Spatial distribution of (a) the 4-day (11-14 Feb) total accumulated precipitation derived from the EXP_T1 simulations, and (b) their differences from the CONT simulation. Precipitation is represented with shading based on scale at the right of (a) and (b). Here, GN, DH, and DG indicate the observational stations at Gangneung, Donghae, and Daegwallyong, respectively.

Fig. 11. Vertical cross-section of the u-w vector derived from the CONT (upper), and EXP_T1 (lower) simulations at 1500 UTC 11 February (left) and 0000 UTC 12 February (right) 2011. 31 August 2012 Sun-Hee Jung et al. 269

Fig. 12. Spatial distribution of the divergence field derived from the CONT (upper), EXP_T1(lower) simulations at 1500 UTC 11 February (left) and 0000 UTC 12 February (right) 2011. Convergence (red) and divergence (blue) are represented with shading based on the scale at right. derived from EXP_T1 was mostly similar to that of CONT 2011 when the snowfall reached the peak in GN and DG, (CONT = 25.4 mm and EXP_T1 = 26.1 mm), the difference respectively. At the peak time in GN, upward and downward field revealed that the spatial details were rather different. The flows along the mountain ridge and valley are found over EXP_T1 with its smoothed topography tended to simulate a mountains in CONT. In particular, on the windward of the more widespread distribution of the precipitation, but with lower Taebaek Mountains, the vertical flows lifting along the steep amounts of localized maximum compared to CONT. Since the mountain slope appears to be descent due to blocking effect, peak of the moderate slope of the EXP_T1 topography was forming the cold air dome on the foot of the mountains. This located further inland (See Fig. 1d), the northeasterly could cold air encounters the warm and moist air flow from the reach relatively far from offshore and deliver more moisture ocean, leading to convergence. On the other hand, EXP_T1 do inland. This implies that the change of topography tended to not develop such a vertical circulation and the flows are modulate the topographically induced mechanical flow, resulting extremely weak over mountains. There is few orographic in the modification of the precipitation distribution. Therefore, lifting on the windward slope of the mountains. It suggests that the precipitation was enhanced around the mountainous area the representation of realistic topography can play a critical role and the corresponding reduction was aligned nearly parallel to to control the vertical structure of the wind, directly affecting the coastal area. the resultant snowfall simulation. Moving to the peak time in For more detailed analysis, we provide the vertical cross- DG, the magnitude of the upward motion at the steep mountain sections of the u-w vector (Fig. 11) along the line passing slope became stronger, and thus propagated further inland. The through the GN and DG stations (See Figs. 1c and 1d) at 1500 stronger orographic lifting going over the mountain induced UTC 11 February (a, c) and 0000 UTC 12 February (b, d) heavy snowfall in DG station. EXP_T1 also showed an 270 ASIA-PACIFIC JOURNAL OF ATMOSPHERIC SCIENCES upward motion along the mountain slope, but the magnitude experiment (a, b, and c) and their differences from the CONT was relatively weak and the pattern appeared to be artificial. simulation (d, e, and f). The spatial distribution and the amount This figure again confirmed the importance of an elaborate of precipitation were directly correlated with the SST forcing representation of the topography by means of the modification in the positive and negative ways. Positive forcing (increasing of the topographically induced mechanical flow. SST) enhanced the precipitation while negative forcing (de- Figure 12 shows the same quantities as Fig. 11 except for the creasing SST) suppressed the precipitation. The difference spatial distribution of the convergence and divergence fields. fields clearly represent the spatial distribution of the precipita- Convergence (divergence) is taken as a proxy of the strength tion enhancement or suppression regions compared to the of dynamical distribution. A comparison between the peak CONT simulation. In the case of EXP_S1 (EXP_S2), the time in GN and DG revealed the westward propagation of the higher (lower) SST prescribed over the East Sea effectively convergence zone toward inland. This behavior matched the produced more (less) precipitation, showing a dominant vertical cross-section of the u-w vector reasonably well. For positive (negative) sign. On the other hand, EXP_S3 exhibited CONT, the spatial distribution of the convergence and diver- a mixed pattern. Consistently with the SST difference (Fig. gence over the land area were quite inhomogeneous, reflecting 2d), the EXP_S3 simulated more precipitation over the north the topographical signal. On the contrary, complex regional area with higher SST than that of CONT and less precipitation variations mostly disappeared in the EXP_T1 pattern. over the south area with lower SST than that of CONT. Increasing SST tends to enhance the precipitation as c. Analysis of the sensitivity experiment for SST effect considerable moisture and heat flux are supplied (Kang and Ahn, 2008; Cha et al., 2011). Therefore, the results shown in Figure 13 presents the 4-day total accumulated precipitation Fig. 13 were somewhat expected. To verify this point, we (11-14 Feb.) simulated by three kinds of SST sensitivity investigated the thermodynamical effect due to the change of

Fig. 13. Spatial distribution of the 4-day (11-14 Feb) total accumulated precipitation derived from the EXP_S1, EXP_S2, and EXP_S3 simulations (a, b, and c), and their differences from the CONT simulation (d, e, and f). Precipitation is represented with shading based on the scale at the right of (c) and (f). Here, GN, DH, and DG indicate the observational stations at Gangneung, Donghae, and Daegwallyong, respectively. 31 August 2012 Sun-Hee Jung et al. 271

Fig. 14. Spatial distribution of the time-averaged sensible heat flux (a, c, e, and g) and latent heat flux (b, d, f, and h) from the CONT, EXP_S1, EXP_S2 and EXP_S3 simulations of the 4-day (11-14 Feb) total accumulated precipitation. Sensible heat flux and latent heat flux are represented with shading based on the scale at the right of (a) and (b). the moisture and heat fluxes. Figure 14 displays the spatial SST distribution (Fig. 2), the regions with higher sensible and distribution of the sensible and latent heat flux derived from latent fluxes generally coincided with the SST pattern. The CONT and three SST sensitivity experiment. Compared to the sensible and latent heat flux were similar, which was attributed 272 ASIA-PACIFIC JOURNAL OF ATMOSPHERIC SCIENCES to the SST pattern. In the experiment with SST gradient adding (subtracting) 2 K and EXP_S3 prescribed SST as (CONT, EXP_S1 and EXP_S2), rather zonal gradient pre- constant by removing the spatial gradient as the initial and dominated with a maximum over warmer ocean area, whereas boundary conditions; all the other conditions were held EXP_S3 tended to have a northward gradient with the maxi- identical. A comparison between CONT and the two types of mum around the north boundary area. Both maximums sensitivity simulations revealed the role of topography and SST appeared to occur in the region with a larger difference between in the formation of heavy snowfall. EXP_T1 with unrealisti- SST and the upper air temperature. For instant, in EXP_S3 cally smoothed topography did not accurately simulate the with the constant SST fields, a more favorable thermal envir- vertical circulation with repeated upward and downward motion onment for cold air modification occurred in the north areas along the mountain ridges and valleys. Both the resolution and where the upper airflows tended to reach earlier due to the the orographic feature critically affected the formation of heavy weakening of the upward transfers of moisture and heat as the snowfall by modifying the topographically induced mechanical cold upper air passed by the relatively warm ocean and gained flows, which changed the spatial distribution of snowfall. The energy. Therefore, the SST condition can thermodynamically SST sensitivity experiment clearly showed that the higher modify the heat fluxes, thereby modulating the main source of (lower) SST tended to enhance (suppress) the precipitation energy for the formation of heavy snowfall. amount because the SST condition could modulate the inten- sity of sensible and latent heat fluxes that acted as the source of 5. Summary and discussion energy for the formation of heavy snowfall. Except for the topography and SST, many factors determine In this study, we investigated the synoptic overview and the snowfall in the Yeongdong region, such as air mass relevant characteristics of the unprecedented heavy snowfall modification, strength of the northeasterly, and the difference that occurred on 11-14 February 2011 in the Yeongdong region. between air temperature and SST. We will examine the detailed This event can be explained within the framework of the typical effects applying to a variety of snowfall events in a future type of Yeongdong heavy snowfall. When the northeasterly work, which will generate more robust results. Such a study winds caused by the extension of the Siberian High passed will enhance our understanding of the mechanism of the heavy through the relatively warm sea surface, they brought abundant snowfall event. Eventually, we will develop effective strategies heat and moisture fluxes from the ocean. Their subsequent for preventing or reducing the damage caused by heavy encounter with the cold and dry air flows that had been turned snowfall. down due to the effect of mountain blocking gave rise to a convergence zone. Furthermore, the closed meso-scale Low Acknowledgements. This work was supported by a grant (code that developed over the eastern part of the East Sea maintained No. 3100-3136-442) funded by the National Institute of the strong and prevailing northeasterly in the Yeongdong Meteorological Research (NIMR), the Korea Meteorological region, which produced a favorable condition for long lasting Administration (KMA). heavy snowfall. Interestingly, the area of snowfall maxima tended to move from the plain coastal area to the inland REFERENCES mountainous area. It is related to the magnitude and direction of the northeasterly, which affected the location of the con- Ahn, J. B., J. H. Oh, and E. H. Cho, 1998: A mesoscale atmosphere/ocean vergence zone. coupled model experiment for a heavy snowfall event in Korean 34 To simulate this snowfall event and extend our understanding peninsula. J. Korean Meteor. Soc., , 652-663. (In Korean with English abstract) of the key factors for the formation of heavy snowfall, we used × Alestalo, M., and H. Savijärvi, 1985: Mesoscale circulations in a the WRF numerical model with a high resolution (1 km hydrostatic model: Coastal convergence and orographic lifting. Tellus, 1 km) with a focus on the Yeongdong region. We performed 37A, 156-162. one control experiment (CONT) and two types of sensitivity Cha, Y.-M., H.-W. Lee, and S.-H. Lee, 2011: Impacts of the High- experiment to examine the topography and SST effect for the Resolution Sea Surface Temperature Distribution on Modeled Snowfall heavy snowfall (EXP_T1, and EXP_S1, S2, S3). We first Formation over the Yellow Sea during a Cold-Air Outbreak. Wea. 26 validated the CONT simulation against the station observations. Forecasting, , 487-503. Chang, H., J. Franczyk, and C. Kim, 2009: What is responsible for The CONT simulation reproduced the spatial distribution and increasing flood risks? The case of Gangwon Province, Korea. Nat. temporal evolution of the precipitation very well. In particular, Hazards, 48, 339-354. the movement of the snowfall maximum from the coastal Chen, F., and J. Dudhia, 2001: Coupling and advanced Land surface- region (e.g., GN and DH) to the Taebaek Mountains (DG) was hydrology model with the Penn State-NCAR MM5 modeling system. well captured in spite of the quantitative discrepancy. Given Part: Model implementation and sensitivity. Mon. Wea. Rev., 129, 569- this validation of the model performance, we performed two 585. types of additional sensitivity experiment, one for topography Chung, G.-B., J.-Y. Kim, and T.-Y. Kwon, 2004: Characteristics of Lower- Tropospheric Wind Related with Winter Precipitation in the Yeongdong effect (EXP_T1) and one for SST effect (EXP_S1, S2, S3). Region. J. Korean Meteor. Soc., 40, 369-496. (In Korean with English EXP_T1 used the smoothed topography derived from 10-min abstract) USGS dataset while EXP_S1(S2) prescribed SST uniformly by Han, S.-H., and J.-G. Lee, 2007: Numerical Simulation Study on the 31 August 2012 Sun-Hee Jung et al. 273

Sensitivity of WRF Model in the Wind Field to the Steepness of a heavy snowfall event in the Yeongdong coastal area in relation to the Mountain Slopes. Atmosphere, 17, 349-364. (In Korean with English northeasterly. Atmosphere, 18, 339-354. (In Korean with English abstract) abstract) Hong, S.-Y., Y. Noah, and J. Dudhia, 2006: A new vertical diffusion ______, and S. R. In, 2009: A numerical sensitivity experiment of the package with an explicit treatment of entrainment processes. Mon. Wea. downslope windstorm over the Yeongdong region in relation to the Rev., 134, 2318-2341. inversion layer of temperature. Atmosphere, 19, 331-344. (In Korean Hwang, Y.-J., J.-C. Ha, Y.-H. Kim, K.-H. Kim, E.-H. Jeon, and D.-E. with English abstract) Chang, 2011: Observing System Experiments Using KLAPS and ______, S-.D. Kim, and Y. J. Kim, 2011: A Trajectory Study on the Heavy 3DVAR for the Upper-Air Observations over the South and West sea Snowfall Phenomenon in Yeongdong Region of Korea. Asia-Pacific J. during ProbeX-2009. Atmosphere, 21, 1-16. (In Korean with English Atmos. Sci., 47, 45-62. abstract) Lee, J.-S., T.-Y. Kwon, and D.-R. Kim, 2006: Statistical Verification of Iacono, M. J., J. Delamere, E. Mlawer, M. Shephard, S. Clough, and W. Precipitation Forecasts from MM5 for Heavy Snowfall Events in Collins, 2008: Radiative forcing by long-lived greenhouse gases: Yeongdong Region. Atmosphere, 16, 124-139. (In Korean with English Calculations with the AER radiative transfer models. J. Geophys. Res., abstract) 113, D13103, doi:10.1029/2008JD009944. Lim, K.-S.S., and S.-Y., Hong, 2010: Development of an effective double- Im, E.-S., W.J. Gutowski Tr., and F. Giorgi, 2008: Consistent changes in moment cloud microphysics scheme with prognostic Cloud Condensa- twenty-first century daily precipitation from regional climate simulations tion Nuclei (CCN) for weather and climate models. Mon. Wea. Rev., for Korea using two convection parameterizations. Geophys. Res. Lett., 138, 1587-1612. 35, L14706, doi:10.1029/2008GL034126. Malwer, E. J., S. J. Taubman, P. D. Brown, M. j. Iacono, and S. A. Clough, Kang, S.-D., and J.-B. Ahn, 2008: Numerical Study on the Formation and 1997: Radiative transfer for on-homogeneous atmospheres: RRTM, a Maintenance Mechanisms of Cloud Street in the East Sea during Cold validated correlated-k model for the longwave. J. Geophys. Res., 102, Air Outbreak. Asia-Pacific J. Atmos. Sci., 44, 105-119. 16663-16682. Kim, J.-E., T.-Y. Kwon, and B. Y. Lee, 2005: Characteristics of Sensible Park, J.-H., K.-E. Kim, and B.-K. Heo, 2009: Comparison of Development Heat and Latent Heat Fluxes over the East Sea Related with Yeongdong Mechanisms of Two Heavy Snowfall Events Occurred in Yeongnam Heavy Snowfall Events. Oce. Pol. Res., 27, 237-250. (In Korean with and Yeongdong Regions of the Korean Peninsula. Atmosphere, 19, 9- English abstract) 36. (In Korean with English abstract) Kim, J.-H., and I.-U. Chung, 2006: Study on Mechanisms and Orographic Seth, A., and F. Giorgi, 1998: The effects of domain choice on summer Effect for the Springtime Downslope Windstorm over the Yeongdong precipitation simulation and sensitivity in a regional climate model. J. Region. Atmosphere, 16, 67-83. (In Korean with English abstract) Climate, 11, 2698-2712. Lee, H., and T. Y. Lee, 1994: The Governing Factors for Heavy Snowfalls Skamarock, W. C., and Coauthors, 2008: A Description of the Advanced in Youngdong Area. J. Korean Meteor. Soc., 30, 197-218. (In Korean Research WRF Version 3. NCAR Tech. Note NCAR/TN-475+STR, with English abstract) 125 pp. Lee, J. G., and Y. J. Kim, 2008a: A numerical case study examining the Tao, W.-K., J. Simpson, and M. McCumber, 1989: An ice-water saturation orographic effect of the Taebaek mountains on snowfall distribution adjustment, Mon. Wea. Rev., 117, 231-235. over the Yeongdong Area. Atmosphere, 18, 367-386. (In Korean with Yeh, H., and Y. Chen, 2002: The role of offshore convergence on coastal English abstract) rainfall during TAMEX IOP 3. Mon. Wea. Rev., 130, 2709-2730. ______, and ______, 2008b: A numerical simulation study using WRF of