1. Introduction
High-resolution numerical models, often referred to as convection-permitting models, play an important role in the prediction of severe weather phenomena. While the horizontal resolution of global weather prediction models has gradually been increasing, limited area models (LAMs), such as the Weather Research and Forecasting (WRF) Model (Skamarock et al. 2008), are still used in research, operational numerical weather prediction (NWP), and regional climate studies. LAMs are favored because of their ability to provide high-resolution numerical predictions with comparatively lower computational resource demands. Unlike global models, LAMs focus their calculations on specific regions of interest by relying on lateral boundary conditions obtained from large-scale NWP models. However, this transition between different resolutions and physics can introduce unrealistic numerical errors at the boundaries (e.g., Harris and Durran 2010; Park et al. 2014). Nonetheless, LAMs remain essential for significant computational time and resource savings compared with those of global models with equivalent resolutions.
The Model for Prediction Across Scales (MPAS) (Skamarock et al. 2012) is conspicuous among global NWP models due to its use of centroid Voronoi meshes. The MPAS incorporates variable-resolution meshes, enabling smooth transitions from coarser to finer resolutions. This design aimed to mitigate numerical errors at the boundaries between various grid resolutions. Numerous studies have attested to the advantages of MPAS variable-resolution meshes (e.g., Park et al. 2014; Zhao et al. 2019; Cheng et al. 2023), which contribute to the computational efficiency of high-resolution simulations for regions of interest compared to that of the global models using uniform grids or meshes (e.g., Kramer et al. 2020). Despite these advantages, the computational costs of the global MPAS remain higher than those of LAMs, prompting the development of a regional version, known as regional MPAS (rMPAS), to address the need for efficient computational costs in high-resolution numerical simulations (Skamarock et al. 2018). Skamarock et al. (2018) demonstrated the capability of rMPAS to accurately reproduce the structure of a squall line and its surrounding environment, highlighting its promise for high-resolution regional forecasting.
The rMPAS and WRF Models share many dynamic aspects and physical processes. However, compared to those of the WRF Model, the rMPAS has prominent differences in both its horizontal and vertical grid structures. Specifically, the WRF Model relies on a map projection technique to transform the simulation domain into a uniform horizontal grid, whereas rMPAS constructs its simulation domain using uniform- or variable-resolution meshes extracted from those of the global MPAS model on a spherical surface. Furthermore, the WRF Model employs a Cartesian coordinate system in both the zonal and meridional directions, facilitating the straightforward determination of horizontal flux order, which is typically set to the fifth order by default. Contrastingly, rMPAS, with its unstructured mesh, has only third- and fourth-order horizontal flux operators. Another distinction is the vertical coordinate system. Klemp (2011) introduced a terrain-following coordinate system with smoothed vertical coordinate surfaces, which effectively mitigates artificial vertical motions often associated with basic or hybrid terrain-following coordinates. This smoothed terrain-following (STF) coordinate, designed for height-based vertical coordinates, facilitates the construction of smoothed vertical coordinate surfaces over complex topography. However, creating smoothed vertical coordinate surfaces using a pressure coordinate is more challenging. Choi and Klemp (2021) demonstrated a smoothed hybrid sigma-pressure coordinate system including the WRF reference state. However, in this study, this WRF Model option was not used.
To the best of our knowledge, there is a notable lack of studies validating the predictive performance of rMPAS through high-resolution numerical experiments in the East Asia region, particularly focusing on numerical differences discussed above. Given that physical parameterizations are shared between WRF and MPAS, our analysis focuses on differences in their numerical methodologies. Therefore, this study aimed to evaluate the rMPAS in predicting snowfall episodes associated with complex terrain during the International Collaborative Experiments conducted during the Pyeongchang 2018 Olympics and Winter Paralympics Games (ICE-POP 2018) field campaign by comparing the results with those from the WRF Model using the same physics parameterizations.
The Pyeongchang region, shown in Fig. 1c, along the eastern coast of the Korean Peninsula, located between the Taebaek Mountains to the west and the East Sea (also known as the Sea of Japan) to the east, experiences frequent heavy snowfall episodes when winds from the East Sea interact with the steep terrain of the Taebaek Mountains (e.g., Park et al. 2019a; Kim et al. 2021). The ICE-POP 2018 field campaign focused on this region and gathered diverse observational datasets with high spatiotemporal resolutions (e.g., Gehring et al. 2020; Kim et al. 2021), which enhance our understanding of the mechanisms driving heavy snowfall in the Pyeongchang region and provide valuable verification data for high-resolution numerical experiments (e.g., Lim et al. 2020; Ko et al. 2022). Among these, the Wind Synthesis System using Doppler Measurements (WISSDOM) data, created from a high-resolution observation network during the ICE-POP 2018 field campaign, are particularly beneficial for validating high-resolution numerical model results (e.g., Tsai et al. 2018), offering a three-dimensional high-resolution wind field from the surface to an altitude of 10 km.
(a) Cell spacing in the 15–3-km variable-resolution global mesh. (b) Regional domain of rMPAS (black box) and WRF (red box) with terrain height for rMPAS shaded. The zoomed-in area of the East Asia region in (b) shows a cell spacing of less than 3.1 km, outlined by the dark blue line. The dashed red line in (b) indicates the location of the cross sections shown in Fig. 2. (c) Terrain height in the Pyeongchang area for the WISSDOM domain, zoomed into the green box from (b). The red (coastal) and blue (mountain) triangles in (c) indicate ASOS observation site locations, and the thick black dashed line indicates the Taebaek Mountains.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
Here, we evaluated rMPAS simulations of heavy snowfall episodes over complex terrain during the ICE-POP 2018 field campaign and compared them with those of the results from the WRF Model and observations. We selected three prominent heavy snowfall episodes forthe high-resolution numerical experiments. The subsequent sections detail our observational dataset and model configuration (section 2), verify the simulated meteorological fields of rMPAS against the WRF Model results and observations (section 3), examine and discuss the horizontal and vertical motion characteristics in complex terrain regions simulated by rMPAS and WRF (section 4), and conclude with a summary and discussion (section 5).
2. Data and experimental setup
a. Data
The performance of both the rMPAS and WRF Models were assessed by analyzing three heavy snowfall case studies. Given the coarse spatiotemporal resolution of available snowfall observations and the absence of corresponding variables in the model outputs, we utilized precipitation data from both observations and model results to evaluate the snowfall event simulations. The verification dataset consists of hourly precipitation data collected from 613 automatic weather stations (AWSs), which were interpolated using Barnes-type iterative correction objective analysis. The 10-m wind data obtained from eight Automated Surface Observing System (ASOS) stations in the Yeongdong region were also used for the verification of the model results (Fig. 1c). We employed hourly data from the fifth major global reanalysis produced by European Centre for Medium-Range Weather Forecasts (ERA5; Hersbach et al. 2020) with a horizontal resolution of 0.25° to verify the synoptic environment during these heavy snowfall events. ERA5 data were used not only for verification but also as the source for initial and boundary conditions in both the rMPAS and WRF Models, with lateral boundary conditions updated at 1-h intervals.
The composite radar reflectivity field was used to verify the general features of precipitation patterns over the Korean Peninsula. The hybrid surface rainfall (HSR) method was employed to generate individual reflectivity maps at the lowest height (Kwon et al. 2015), and HSR reflectivity maps from 10 operational S-band dual-polarimetric radars (Oh et al. 2020) were averaged. The radar beamwidth is 0.95°, with the volume coverage pattern including nine short-range (240 km) and one long-range (480 km) sweep every 5 min, respectively (see Table 1 of Oh et al. 2020). Although the radar network is quite dense, the composite is affected by beam blocking over mountainous regions as shown in the study domain (Fig. 1c). Additionally, the reflectivity and precipitation rate can be underestimated due to the shallowness and steepness of snowstorm, combined with radar beam geometry and terrain. This underestimation is mitigated by further adding three X-band and two S-band dual-polarimetric radars during ICE-POP 2018 (Fig. S1 in the online supplemental material).
To further evaluate the simulated wind fields of both the rMPAS and WRF Models, we used a three-dimensional wind field obtained from WISSDOM. It generates this wind field by integrating wind data from various sources, including radar Doppler velocity measurements, soundings, model forecasts, surface AWS, and other observational data. The methodology involves a variational-based approach aimed at minimizing the cost function. The cost function comprises eight terms: 1) the difference between observed Doppler velocities and those derived from true winds, 2) the difference between background and true winds, 3) the anelastic continuity equation, 4) the vertical vorticity equation, 5) a Laplacian smoothing filter, 6) the difference between the true wind fields and the sounding observations, 7) the discrepancy between the true wind fields and ASOS data, and 8) the misfit between the true winds and the local reanalysis dataset (Tsai et al. 2023). Additionally, the immersed boundary method (IBM) is used to enhance the representation of terrain (Liou et al. 2012). This significantly improves the accuracy of retrieved winds in the lower layer over the complex terrain such as the current study domain. Liou et al. (2012) demonstrated that the root-mean-square error (RMSE) of horizontal wind and vertical air velocity is under 0.25 m s−1. The bias of horizontal wind speed was almost negligible above 1.1 km MSL (above mean sea level) within the study domain, and the median value of the vertical air velocity error ranged from −0.2 to 0.6 m s−1 at heights between 0.8 and 2.0 km MSL (Tsai et al. 2023). A comprehensive description of the methodology of WISSDOM is provided by Liou et al. (2012); Liou and Chang (2009); Liou et al. (2014); and Tsai et al. (2018, 2023). High-temporal-frequency radar observations allow WISSDOM to generate three-dimensional wind fields at 10-min intervals. The primary WISSDOM domain, shown in Fig. 1c, covers a 150 × 150 km2 area within the mideastern Korean Peninsula. The horizontal grid size within this domain is 1 km, with a vertical resolution of 0.25 km extending from the surface to an altitude of 10 km. Although the WISSDOM data are confined to a specific region within the field campaign, their high-resolution three-dimensional wind field is indispensable for understanding the characteristics of horizontal and vertical motions, particularly in the vicinity of complex terrain features.
b. Model configuration
Here, we utilized rMPAS and WRF version 7.0 and 4.1, respectively. The rMPAS domain was derived from a global variable-resolution mesh with horizontal cell spacing ranging from 15 to 3 km, specifically designed for East Asia. As depicted in Figs. 1a and 1b, the rMPAS domain has nearly uniform 3-km cell spacing, with only a few cells deviating from this resolution at the corners. Conversely, we constructed a WRF grid with a 3-km single domain using a Lambert conformal map projection with true latitudes of 30° and 60°N (Fig. 1b). We attempted to align the simulation domains of rMPAS and WRF to the greatest extent possible (Fig. 1b). In terms of vertical levels, both models utilized 94 layers, extending up to 28 km in rMPAS and 12 hPa in WRF Models, respectively. The vertical resolution was set at 300 m for most layers, with the exception of the lower and upper regions for both the rMPAS and WRF Models. Both the models used a hybrid vertical coordinate, setting the flattening level parameter for Zh (Klemp 2011) at 14 km in rMPAS and configuring ηc (Park et al. 2019b) to 0.2 in WRF. Both the rMPAS and WRF Models were initialized without data assimilation for CASE1 at 0000 UTC 27 February 2018, for CASE2 at 1200 UTC 3 March 2018, and for CASE3 at 1200 UTC 14 March 2018, as indicated in Table 1. Although the available physics options in rMPAS are limited compared to those in WRF, we configured both models using identical physical parameterization schemes. These schemes encompassed the WRF single-moment 6-class cloud microphysics (Hong and Lim 2006), the scale-aware Grell–Freitas cumulus scheme (Grell and Freitas 2014; Fowler et al. 2016), the unified Noah land surface model (Chen and Dudhia 2001), the Yonsei University planetary boundary layer scheme (Hong et al. 2006), the Monin–Obukhov surface layer scheme (Fairall et al. 2003), and the Rapid Radiative Transfer Model for general circulation models longwave and shortwave radiation schemes (Iacono et al. 2008; Morcrette et al. 2008). In the control simulation, we chose not to apply a gravity wave drag scheme. This decision was based on the assessment that a 3-km horizontal grid resolution was adequate for the explicit resolution of mesoscale orographic drag associated with gravity waves.
Forecast and analysis periods of the cases during the ICE-POP 2018 field campaign.
c. Sensitivity experiments
Description of the sensitivity experiments for the rMPAS.
Description of the sensitivity experiments for the WRF.
Vertical cross section of coordinate surfaces along the red line in Fig. 1b for (a) rMPAS and (b) WRF. rMPAS data are interpolated from the native mesh to the WRF grid.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
3. Verification of simulated results
During the ICE-POP 2018 field campaign, the eastern coastal region of the Korean Peninsula experienced significant heavy snowfall episodes (Gehring et al. 2020; Kim et al. 2021). From this set of events, we chose to analyze three specific heavy snowfall cases (CASES 1 and 2: warm low, CASE3: air–sea interaction by Kim et al. 2021) to assess the simulations conducted using rMPAS and WRF. This section emphasizes on the verification of the rMPAS simulation results achieved by contrasting them with those of the WRF and observational data. Our verification process encompassed precipitation, the synoptic environment, and radar reflectivity.
a. Synoptic environment
CASE1, which occurred from 27 to 28 February 2018, marked a significant heavy snowfall event, resulting in a maximum snow accumulation of 41.2 cm in Daegwallyeong (located in the mountainous region, as depicted in Fig. 1c) and 9.5 cm in Gangneung (situated in the coastal area shown in Fig. 1c). Figures 3a–c depict the geopotential height and wind vectors at 925 hPa obtained from the ERA5 reanalysis. In CASE1, at 0600 UTC 28 February 2018, the synoptic environment revealed the development of a cyclone in eastern China, which subsequently moved northeastward, approaching the Korean Peninsula (Fig. 3a). The cyclone center was positioned in the ocean southwest of the Korean Peninsula. As a result, the eastern coastal region of the Korean Peninsula, located northeast of the cyclone, was influenced by warm and humid southeasterly winds from the East Sea. Under these conditions, heavy snowfall occurred as the southeasterly winds encountered the steep terrain of the Taebaek Mountains. This synoptic setup is consistent with previous studies in which heavy snowfall in the eastern coastal region resulted from extratropical cyclones passing through the Korean Peninsula. Heavy snowfall events are relatively frequent and often lead to substantial snow accumulation (e.g., Cheong et al. 2006).
Geopotential height (m; solid line) and wind vector (m s−1; arrow) at 925 hPa obtained from ERA5 reanalysis for (a) 0600 UTC 28 Feb 2018 (CASE1), (b) 1200 UTC 4 Mar 2018 (CASE2), and (c) 1200 UTC 15 Mar 2018 (CASE3). Simulated geopotential height (m; solid line) and wind vector (m s−1; arrow) at 925 hPa in the (d)–(f) rMPAS and (g)–(i) WRF experiments for (left) CASE1, (center) CASE2, and (right) CASE3. Gray shadings in (d)–(i) represent areas with terrain heights exceeding 600 m MSL.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
CASE2 featured a significant heavy snowfall event occurring from 3 to 5 March 2018, resulting in substantial snow accumulations of 21.3 and 8.0 cm in Daegwallyeong and Gangneung, respectively. The synoptic environment in CASE2 showed the Siberian High located northwest of the Korean Peninsula, while a cyclone moved from eastern China toward the Korean Peninsula, as displayed in Fig. 3b. Similar to that in CASE1, heavy snowfall in CASE2 occurred when southeasterly winds induced by the advancing cyclone blew into the eastern coastal region of the Korean Peninsula. CASE3 exhibited a distinct synoptic environment compared with that of CASE1 and CASE2. The Siberian High was located northwest of the Korean Peninsula at 1200 UTC 15 March 2018 (Fig. 3c). Additionally, a high-pressure system existed over the ocean southeast of the Korean Peninsula. A trough positioned between the two high pressure systems extended from the center of a synoptic cyclone situated on the Kamchatka Peninsula to the East Sea. Consequently, the eastern coastal region of the Korean Peninsula experienced northeasterly winds due to the anticyclonic circulation of the Siberian High. The environmental conditions in CASE3 were similar to those of the air–sea interaction mechanisms described in previous studies (e.g., Yeo and Chang 2018), wherein cold air passing over a warm sea induced snowfall along the coastline.
Figures 3d–i show the simulated 925-hPa meteorological fields from the rMPAS and WRF Models. These figures display the forecasting fields at 30 h for CASE1 and 24 h for CASE2 and CASE3, all from the initial time of the model simulation. In CASE1, all experiments adequately captured the intensity and location of the cyclone near the southwestern Korean Peninsula compared to those in the ERA5 reanalysis (Figs. 3d,g). Furthermore, the southeasterly flow generated by the cyclonic circulation was well represented in the eastern coastal region of the Korean Peninsula. Although the rMPAS experiment (Fig. 3d) exhibited a slightly weaker simulation of the cyclone center than that in the experiments using the WRF Model (Fig. 3g), although the difference was not critical. The simulated geopotential height and wind patterns around the Korean Peninsula for CASE2 closely resemble those of the ERA5 reanalysis, as illustrated in Figs. 3e and 3h. Particularly, the cyclonic circulation around the Korean Peninsula and associated easterly winds blowing toward the eastern coastal region were accurately reproduced. Both the rMPAS and WRF Models simulated weaker cyclones over the Yellow Sea than those indicated in the ERA5 reanalysis (i.e., the 720-m contour in Fig. 3b). In CASE3, the rMPAS and WRF Models effectively reproduced the synoptic environment, as displayed in Figs. 3f and 3i, respectively. The models captured a trough extending southeastward from the Kamchatka Peninsula. Nevertheless, the 780-m contour over the East China Sea differed from those that in ERA5 because the models reproduced a weaker high pressure system located southeast of the domain.
b. Precipitation
The left column of Fig. 4 compares observed and simulated accumulated precipitation for CASE1. Between 0300 UTC and 1500 UTC 28 February 2018, AWS observations (indicated by black dots) showed over 40 mm of precipitation in the eastern coastal region, east of the mountain range. Additionally, Daegwallyeong, situated at the top of the mountain range (Fig. 4a), recorded more than 30 mm of precipitation. Both the rMPAS and WRF Models successfully captured the intense snowfall in the eastern coastal region, generating similar amounts of precipitation. However, both models overestimated the total accumulated precipitation compared to that in the AWS observations (Figs. 4d,g). In CASE2, from 0900 UTC 4 March to 0900 UTC 5 March 2018, accumulated precipitation was primarily concentrated along the eastern coastal line, with Daegwallyeong showing lower precipitation than that in CASE1 (Fig. 4b). The simulated patterns of accumulated precipitation in both rMPAS and WRF closely matched with that in the observed patterns. No significant differences were observed between the rMPAS and WRF results, although both models overestimated precipitation in the eastern coastal region, consistent with the trends from CASE1 (Figs. 4e,h). CASE3 presented a heavy snowfall event in a distinct synoptic environment compared to that of the other cases. Nevertheless, it exhibited similarities in significant winds blowing from the East Sea toward the Taebaek Mountains. From 0300 UTC 15 March to 0000 UTC 16 March 2018, observed accumulated precipitation was concentrated on the northeastern coastline, with lower amounts than those in other cases (Fig. 4c). The rMPAS and WRF Models reproduced the observed precipitation pattern for the northeastern coastline but tended to overestimate the accumulated precipitation amounts upstream of the Taebaek Mountains along the coastline (Figs. 4f,i).
Accumulated precipitation amount (mm; shaded) during (left) 0300 UTC 28 Feb–1500 UTC 28 Feb 2018 (CASE1), (middle) 0900 UTC 4 Mar–0900 UTC 5 Mar 2018 (CASE2), and (right) 0300 UTC 15 Mar–0000 UTC 16 Mar 2018 (CASE3). The accumulated precipitation amount for each case from (a)–(c) AWS, (d)–(f) rMPAS, and (g)–(i) WRF. The black dots in the top panel indicate the location of the AWS station.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
Both the rMPAS and WRF Models effectively simulated the horizontal distribution of snowfall, showing remarkably similar precipitation patterns compared to those of the observations. The overestimation of accumulated precipitation in the eastern coastal region of the Korean Peninsula is evident (Fig. 4). This overestimation is detailed in Table 4, which presents the mean and maximum values of accumulated precipitation for each case. In all cases, both rMPAS and WRF Models overestimated precipitation compared to that in AWS observations, both in mean and maximum values. While some variations exist between cases, the discrepancy in accumulated precipitation between the two models is not significant. Ko et al. (2022) reported similar findings, noting that WRF simulations tended to overestimate precipitation for most snowfall events during the ICE-POP 2018 field campaign. They suggested that this positive bias may be reduced by adjusting the melting efficiency in the microphysics parameterization scheme. Given the scope of this study, additional sensitivity tests beyond the numerical aspects of these models were not conducted.
Mean and maximum values of the accumulated rainfall amount (mm) at AWS observation sites (black dots at the top of Fig. 4) during 0300 UTC 28 Feb–1500 UTC 28 Feb 2018 (CASE1), 0900 UTC 4 Mar–0900 UTC 5 Mar 2018 (CASE2), and 0300 UTC 15 Mar–0000 UTC 16 Mar 2018 (CASE3). The RMSE of rMPAS and WRF is calculated based on these observations.
c. Radar reflectivity
Figure 5 depicts the horizontal distribution of observed radar reflectivity from the Korea Meteorological Administration, along with the simulated radar reflectivity from the rMPAS and WRF Models. For CASE1, regions of convection were observed, accompanied by the cyclonic circulation of the propagating cyclone to the west of the Korean Peninsula (Fig. 3a). The observed radar reflectivity showed convection prevailing in the midwestern, eastern, and southeastern coastal regions of the Korean Peninsula (Fig. 5a). Among these regions, the southeastern Korean Peninsula and the eastern coastal region exhibited the strongest radar reflectivity and weaker reflectivity, respectively. However, high mountains in the region can induce heavy snowfall in the eastern coastal area despite the weaker radar reflectivity (Fig. 4). Figures 5d and 5g show that both models consistently overestimate radar reflectivity compared to that in the observations. The simulated radar reflectivity for heavy snowfall in the eastern coastal region aligns with the observed west–east rainband structure but with higher intensity. Figures 5b and 5c, representing CASE2 and CASE3, respectively, show stronger radar reflectivity in the southern coastal region due to strong southwesterly winds, while it is weaker in the eastern coastal region due to milder easterly winds. Nevertheless, mountains can still cause significant snowfall in the eastern coastal region even with weak radar reflectivity. The radar reflectivity for CASE2 and CASE3 exhibited a horizontal pattern similar to that in the observations (Figs. 5e,f,h,i), which is consistent with CASE1. However, both models also overestimated radar reflectivity in the eastern coastal region for CASE2 and CASE3.
(a)–(c) Composite radar reflectivity (dBZ; shaded) obtained from KMA for (a) 0600 UTC 28 Feb 2018 (CASE1), (b) 1500 UTC 4 Mar 2018 (CASE2), and (c) 1200 UTC 15 Mar 2018 (CASE3). Simulated radar reflectivity in the (d)–(f) rMPAS and (g)–(i) WRF for the (left) CASE1, (center) CASE2, and (right) CASE3.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
4. Characteristics of the simulated low-level wind field over the complex terrain
Here, we compare the characteristic features of simulated wind fields over regions with complex terrain using data from the rMPAS and WRF Models and compare them with those of the observations from the ASOS and WISSDOM datasets. Our statistical analyses were based on 1-h interval datasets from observations and model results during the available WISSDOM data periods (Table 1): 2300 UTC 27 February–2300 UTC 28 February 2018 for CASE1, 0800 UTC 4 March–1000 UTC 5 March 2018 for CASE2, and 1900 UTC 14 March–0000 UTC 16 March 2018 for CASE3. Additionally, we conducted sensitivity experiments related to the numerical aspects of the models, as described in section 2c, to gain insights into the influence of the dynamics of the models on horizontal and vertical motions over complex terrain (Tables 2 and 3). To assess the simulated horizontal winds from the rMPAS, we employed three analytical approaches: examining the kinetic energy spectra of the model at the lower-, mid-, and upper-troposphere levels; comparing the 10-m wind speeds with those of the observations obtained from ground-level observation stations; and evaluating the probability density function (PDF) in the lower troposphere using WISSDOM data.
a. Kinetic energy spectra
To gain insights into the dynamics of the rMPAS and compare it with those of WRF, an analysis of the kinetic energy spectra was conducted (Fig. 6). These spectra were calculated following the procedures outlined by Durran et al. (2017), utilizing two-dimensional Fourier transforms at each model level. To facilitate the spectral analysis, the results from the unstructured rMPAS grid were interpolated onto the WRF grid. Subsequently, we computed the average kinetic energy spectra for each level at 1-h intervals throughout the analysis period (Table 1). To denote the averaged kinetic energy spectra for different levels of the atmosphere, we used the terms “lower,” “middle,” and “upper” to refer to 10 lower-tropospheric levels (model levels 6–15, with an approximate mean altitude of 2 km), 10 midtropospheric levels (model levels 16–25, with an approximate mean altitude of 5 km), and 10 upper-tropospheric levels (model levels 26–35, with an approximate mean altitude of 11 km), respectively. Calculations were performed along the model levels, avoiding the use of constant isobaric or height levels due to the numerous missing values in the lower tropospheric levels, which were mainly attributed to the high topography in the western part of the WRF domain (Fig. 1b).
Kinetic energy spectra for rMPAS (black line) and WRF (red line) at (a) lower-, (b) mid-, and (c) upper-tropospheric levels. The kinetic energy spectra for sensitivity experiments of (d)–(f) rMPAS and (g)–(i) WRF (sensitivity experiments are summarized in Tables 2 and 3). All values represent the average at 1-h intervals and are calculated within the WRF domain. Pink dashed lines indicate wavelengths of 4Δx, 8Δx, and 12Δx, while gray dashed slopes represent the k−5/3 and k−3 slopes.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
The computed kinetic energy spectra revealed that both the rMPAS and WRF Models exhibited the characteristic k−5/3 wavenumber dependence, typical of mesoscale regions (Figs. 6a–c). Both models also displayed an effective resolution, defined as the scale at which the model spectrum begins to decay from the k−5/3 slope, occurring at approximately 7 to 8 times the grid spacing (Δx), consistent with the findings of Skamarock (2004). Despite rMPAS employing fourth-order diffusion for momentum variables and WRF utilizing a fifth-order horizontal flux operator with sixth-order implicit diffusion, the slope of the rMPAS kinetic energy spectra closely resembled that of WRF.
In the rMPAS sensitivity experiments (Figs. 6d–f), the rMPAS-HYP4 experiments revealed significant differences in the kinetic energy spectra at wavenumbers less than 12Δx in all layers. The rMPAS-HYP4 experiments exhibited a reduction in the small-scale kinetic energy, which was primarily attributed to numerical diffusion. The rMPAS-beta1.0 experiment also showed diminished kinetic energy for small-scale motions, although the difference was not substantial compared to that of the rMPAS-HYP4 experiment. In the rMPAS-noSTF experiment, no differences in the slopes of the kinetic energy spectra were observed, even when the vertical coordinate surface was smoothed, including at lower levels. The WRF-3rd experiment exhibited prominent differences from those of the fifth-order WRF experiment at scales less than 12Δx (Figs. 6g–i). These differences arise from the use of the third-order horizontal flux operator (accompanied by the fourth-order implicit diffusion) in the WRF, which introduces more diffusion at smaller scales than that by the fifth-order horizontal flux operator (associated with the sixth-order implicit diffusion). Furthermore, the WRF-6diff experiment illustrated the impact of the sixth-order explicit numerical diffusion, leading to a rapid decline in the slope of the kinetic energy spectra at scales of less than 8Δx.
b. Horizontal wind speed
Figure 7 illustrates the relationship between 10-m wind speeds (y axis) from the model results for the mountainous (left) and coastal (right) regions, with ASOS wind speeds plotted on the x axis. The simulated 10-m winds were interpolated to the location of each station and averaged over the four stations in the mountainous (blue triangles in Fig. 1c) and coastal (red triangles in Fig. 1c) regions. The hourly wind speeds from the ASOS stations and models are represented by scatterplots, and the solid lines represent the regression lines between the ASOS data and the results of each model.
Relationship between 10-m wind speeds from ASOS and model results for (a) the mountainous and (b) coastal regions. Black and red dots represent the 10-m wind speeds of rMPAS and WRF, respectively. Regression lines between ASOS and model results are depicted as black (rMPAS) and red (WRF) solid lines. Sensitivity experiments for (c),(d) rMPAS, and (e),(f) WRF.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
Both rMPAS and WRF tend to overestimate observed wind speeds in the mountainous region (Fig. 7a). Contrastingly, in the coastal region, the overestimation of simulated wind speeds was not as significant as that in the mountainous region (Fig. 7b). Due to limitations in representing the complex terrain effects on surface winds, the disparity in wind speed between the observations and the model was more pronounced in the mountainous region compared to that in the coastal region. Additionally, a difference in the slope of the regression line between the rMPAS and WRF Models was evident in the coastal region. The WRF Model overestimates 10-m wind speeds across the wind speed range, with this overestimation tendency being more prominent at lower wind speeds. Contrastingly, the rMPAS model tends to overestimate and underestimate at low and high wind speeds, respectively.
Figures 7c–f show the results of the sensitivity experiments on the simulation of 10-m wind speeds to examine the effects associated with the numerical models. The results revealed no apparent differences in the rMPAS-HYP4, rMPAS-noSTF, WRF-3rd, and WRF-6diff experiments compared to those in the control experiments in both the mountainous and coastal regions. This indicates that the changes in the numeric of the model in the rMPAS and WRF Models did not significantly influence wind speeds near the surface.
Figure 8 displays the PDF for the horizontal wind speeds at altitudes of 1500, 3000, and 5000 m MSL in all cases. Due to spatial limitations of WISSDOM data, the PDF was calculated within the WISSDOM domain, as depicted in Fig. 1c. At an altitude of 1500 m, WISSDOM indicated a maximum PDF for wind speeds within the range of 6–10 m s−1, whereas both the rMPAS and WRF Models exhibited a maximum PDF at a wind speed of 12 m s−1 (Fig. 8a). The rMPAS and WRF Models exhibited lower PDF magnitudes in the weak wind range and higher magnitudes in the strong wind range compared to WISSDOM at 1500 m. This indicates that the models tended to overestimate the wind speeds at an altitude of 1500 m compared to that in the observations. However, as the altitude increased, the tendency of the models to overestimate horizontal wind speeds diminished. Figures 8b and 8c reveal no significant differences in the PDF between WISSDOM data and the models at altitudes of 3000 and 5000 m. Similarly, sensitivity experiments were conducted to examine the effects of model dynamics on the horizontal wind speed in the upper layers for both the rMPAS and WRF Models (Figs. 8d–i). The results indicated that the strength of the wind speed in the lower and midtropospheric layers did not differ significantly with changes in the numerical model. These findings align with the results of the sensitivity experiments for 10-m wind speeds, as shown in the scatterplots.
PDF of horizontal wind speed (m s−1) for WISSDOM (gray dashed line), rMPAS (black line), and WRF (red line) at (a) 1500-, (b) 3000-, and (c) 5000-m altitudes. Results for (d)–(f) rMPAS and (g)–(i) WRF sensitivity experiments, as well as (j)–(l) with and without the SSOD scheme (summarized in Tables 2 and 3). Values represent the 1-h interval average, calculated within the WISSDOM domain.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
Earlier research suggested that applying SSOD can lead to a reduction in 10-m wind speeds (e.g., Jiménez and Dudhia 2012; Lim et al. 2019). Figures 8j–l present the PDFs of the horizontal wind speed for rMPAS and WRF, both with and without SSOD. Similar to the results from the dynamic sensitivity experiments of the model, the PDF shows that the application of SSOD does not result in significant differences in the horizontal velocity over complex terrain when compared to those in the control experiments. Even with the inclusion of SSOD, the model persistently overestimated horizontal wind speeds in the lower tropospheric layer when compared to that in the observational data.
However, when we specifically examined the simulation of 10-m wind speeds, the rMPAS-SSOD and WRF-SSOD experiments showed a reduction in wind speed in the mountainous region (Fig. 9). The efficacy of such schemes is commonly assessed by comparing simulated 10-m wind speeds with those of the observed values. Our results indicate that the application of SSOD effectively reduced 10-m wind speeds near the surface; however, this improvement did not extend to the upper tropospheric layers. Despite the inclusion of SSOD, the ongoing challenge of both the rMPAS and WRF Models persistently overestimating low-level wind speeds remains unresolved.
Relationship between 10-m wind speeds from ASOS and rMPAS for (a) the mountainous and (b) coastal regions. Black and red dots represent the 10-m wind speeds of rMPAS and rMPAS-SSOD results, respectively. Regression lines between ASOS and model results are depicted as black (rMPAS) and red (rMPAS-SSOD) solid lines. (c),(d) Sensitivity experiments for WRF.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
c. Vertical motion over complex terrain
Figure 10 shows the horizontal distribution of both the observed and simulated vertical velocities at an altitude of 1500 m MSL. Across all cases, the intricate patterns in the vertical velocity are a consequence of interactions with the mountainous terrain. The western part of the mountains, when compared to that of the eastern coastal region of the Korean Peninsula (Figs. 10a–c), exhibited particularly complex small-scale vertical motions. One intriguing observation arises when comparing the model results with those of the observational data, specifically concerning the intensity of the vertical motion. Generally, the rMPAS and WRF experiments simulated stronger vertical velocities over complex terrain than those observed in WISSDOM.
Vertical velocity (m s−1; shaded) at 1.5 km for (left) 0300 UTC 28 Feb 2018 (CASE1), (center) 1200 UTC 4 Mar 2018 (CASE2), and (right) 1200 UTC 5 Mar 2018 (CASE3). The vertical velocity of each case from (a)–(c) WISSDOM, (d)–(f) rMPAS, and (g)–(i) WRF.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
Table 5 shows the mean, standard deviation, minimum, and maximum of the vertical velocity for WISSDOM, rMPAS, and WRF at 1500 and 5000 m MSL. These values are calculated within the WISSDOM domain for the analysis periods in Table 1. At 1500 m MSL, the standard deviation of the vertical velocity was higher in rMPAS and WRF than that in WISSDOM. The magnitude of the minimum and maximum of the vertical velocities was also significantly stronger in the models compared with those in the observations. This indicates that the intensity of the vertical motion was much stronger in all model simulations, as shown in Fig. 10. Although the rMPAS experiments showed relatively weaker vertical velocities than those of the WRF based on the standard deviation and minimum/maximum of the vertical velocity, a consistent trend of overestimation of vertical velocity compared to that in observations was observed in both models. At 5000 m MSL, the standard deviation of the vertical velocity remained higher in the model experiments than that in WISSDOM. However, the discrepancy in minimum and maximum velocities between WISSDOM and the models was reduced compared to that at 1500 m MSL.
Mean, standard deviation, min, and max values of the vertical velocity for WISSDOM, rMPAS, and WRF. All values are calculated within the WISSDOM domain for the analysis periods listed in Table 1.
Figure 11 illustrates the PDF for the square of the vertical velocity at altitudes of 1500, 3000, and 5000 m MSL for all cases. As shown in Figs. 11a–c, it is evident that the probability density of vertical velocity observed in WISSDOM was significantly lower than that depicted in the model results. This suggests that the intensity of the vertical velocity in WISSDOM was weaker than that in the model experiments, which is consistent with the patterns observed in the vertical velocity fields shown in Fig. 10 and Table 5, where the model simulations depicted stronger vertical velocities. The difference in the vertical velocity at an altitude of 1500 m MSL between WISSDOM and the models decreased as the altitude increased. This suggests that the most significant difference in the vertical velocity between the observations and models occurred in the lower layer, and this difference tended to diminish with increasing altitude.
PDF of the (a)–(c) vertical wind speed (m s−1) for WISSDOM (gray dashed line), rMPAS (black line), and WRF (red line) at (left) 1500, (center) 3000, and (right) 5000 m MSL. Results for (d)–(f) rMPAS and (g)–(i) WRF sensitivity experiments, as well as (j)–(l) with and without the SSOD scheme. The black and red dashed lines represent the results of the rMPAS-SSOD and WRF-SSOD experiments, respectively. All values represent the average of 1-h intervals and are calculated within the WISSDOM domain.
Citation: Monthly Weather Review 153, 3; 10.1175/MWR-D-24-0053.1
While prominent differences in the magnitudes of observed and modeled vertical velocities were evident, a discernible contrast in the intensity of simulated vertical velocities emerged between the rMPAS and WRF Models. In line with the characteristics illustrated in Fig. 10, the WRF Model consistently simulated slightly higher vertical velocities than those in the rMPAS model. This difference diminished with increasing altitude. As shown in Figs. 11d–f, our sensitivity experiments, which included adjustments to the numerical parameters of rMPAS, such as increasing the coefficient of hyperdiffusion (i.e., rMPAS-HYP4), modifying β values in the horizontal flux operator (i.e., rMPAS-beta1.0), and not using the smoothed vertical level option (i.e., rMPAS-noSTF), did not yield significant differences in the probability density of vertical velocity in the lower troposphere. Contrary to our expectations, the omission of the smoothed vertical-level option in rMPAS, which was anticipated to affect the representation of small-scale terrain features in low-level winds, did not substantially alter the tendency of the model to overestimate low-level vertical motion. These findings were consistent across various altitudes, including 1500, 3000, and 5000 m. The analysis underscores a persistent trend in both models, rMPAS and WRF, to overestimate the low-level vertical motion compared with that in the observations. This overestimation feature persisted in various sensitivity experiments, including those that did not employ the smoothed vertical-level option (Figs. 11g–i). In Figs. 11j–l, the PDFs of the vertical wind speed for rMPAS and WRF are presented, both with and without SSOD. Similar to those of the results for horizontal wind speed, applying SSOD does not lead to significant differences in vertical velocity compared to those in the control experiments. The model consistently overestimates vertical motion in the lower tropospheric layer when compared to that of the observational data.
We have observed that the rMPAS tended to overestimate both horizontal and vertical wind speeds compared to that of the observational data, a trend that was also present in the WRF Model. This overestimation was most pronounced in the lower troposphere and gradually decreased toward the midtroposphere. The overestimation of precipitation and the strong bias in low-level winds, both horizontal and vertical, may be interrelated. This potential relationship, which was not investigated in this study, should be addressed in future research. While the simulated results of the model were significantly influenced by numerical aspects, particularly in terms of kinetic energy and numerical diffusion, these factors did not substantially impact the local wind patterns.
5. Summary and conclusions
The rMPAS model simulations were evaluated for the ICE-POP 2018 field campaign, focusing on their capacity to simulate heavy snowfall episodes. We compared the rMPAS simulations with outputs from the WRF Model and observational data. Three significant heavy snowfall events were selected for in-depth examination.
Initially, we validated the accuracy of rMPAS simulations by comparing both precipitation patterns and associated synoptic environments. Our comparisons with WRF simulations and observational data demonstrated that both the rMPAS and WRF Models effectively simulated the synoptic conditions of heavy snowfall events during the winter season on the Korean Peninsula. However, both models exhibited a tendency to overestimate precipitation in the eastern mountainous region and had errors in the alignment and position of rainfall features compared with those of the observations. The rMPAS consistently achieved performance levels comparable to those of the WRF Model, affirming its utility in weather forecasting in these complex scenarios.
To gain a deeper understanding of the atmospheric model dynamical cores, we analyzed the kinetic energy spectra of rMPAS and WRF. Despite the distinct approaches, our analysis revealed prominent similarities between the kinetic energy spectra of rMPAS and WRF, with WRF employing a fifth-order horizontal flux compared to that of the rMPAS third-order flux, especially at smaller spatial scales. Sensitivity experiments showed that increasing the coefficient of numerical diffusion in rMPAS (particularly in rMPAS-HYP4 and rMPAS-beta1.0) resulted in significant variations in the kinetic energy spectra, reducing kinetic energy at smaller scales (approximately 12Δx). This finding highlights the role of numerical diffusion in shaping the dynamic characteristics of rMPAS.
Our extensive evaluation of horizontal and vertical winds, compared with those of WRF and various observational datasets, offers valuable insights. The analysis of horizontal wind speed across both rMPAS and WRF Models displayed a consistent overestimation trend, which is most pronounced at the surface and in the lower troposphere. This tendency to overestimate decreased with increasing altitude. A similar pattern was observed in the analysis of vertical winds, where both models invariably simulated stronger vertical velocities over complex terrains than that observed in the data. Although the rMPAS showed slightly lower vertical velocities relative to WRF, both models exhibited a consistent trend of overestimation. The magnitude of this discrepancy, though less pronounced in vertical velocities, was evident. Sensitivity experiments did not fully address this persistent overestimation. Furthermore, the analysis highlighted regional differences in wind speed overestimation, with a more prominent bias in mountainous regions compared to that in the coastal regions. The application of the SSOD scheme mitigated the overestimation of horizontal wind speeds near the surface in mountainous regions, though this effect was not observed at higher altitudes, particularly above 1500 m MSL.
Conclusively, our comprehensive evaluation of the rMPAS model under winter conditions in the Korean Peninsula confirmed its usefulness in predicting heavy snowfall events, particularly in the eastern coastal region. Although the model’s performance often aligns with or occasionally exceeds that of WRF, it also demonstrates its capability in this specific meteorological context. Areas requiring further investigation were identified, prominently the consistent overestimation of low-level wind speeds and the tendency to overestimate precipitation. Additional research during the summer season on the Korean Peninsula is suggested to develop a complete understanding of the rMPAS performance under various conditions. Future studies should prioritize investigating the underlying causes of these biases, particularly in mountainous terrains, and exploring potential interrelationships between precipitation overestimation and low-level wind biases. Such investigations are crucial for developing strategies to enhance the accuracy of NWP models in regions with complex topographical features.
Acknowledgments.
(U. SHIN) This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (Grant: 2021R1I1A1A0105197511). (S. H. P.) This work was also supported by the National Research Foundation of Korea (NRF) Grant funded by the Korean Government (MEST: 2021R1A2C1012433). Computational funding was also supported by the National Supercomputing Center (KSC-2021-CRE-0079). The authors greatly appreciate the participation of the World Weather Research Program Research Development Project and Forecast Demonstration Project, International Collaborative Experiments for Pyeongchang 2018 Olympic and Paralympics Winter Games (ICE-POP 2018) hosted by the Korea Meteorological Administration (KMA). We thank Kwonil Kim at Kyungpook National University and Sang-Won Joo, Yong-Hee Lee, Kwang-Deuk Ahn, Namwon Kim, and Seung-Bo Choi at KMA for their support during the ICE-POP 2018 field campaign. Author Skamarock was supported at the National Center for Atmospheric Research (NCAR), which is a major facility sponsored by the NSF under Cooperative Agreement 1852977.
Data availability statement.
The ERA5 data can be obtained from the Copernicus Climate Data Store (https://cds.climate.copernicus.eu). The precipitation data from the Automatic Weather Stations (AWS) operated by KMA are available at https://data.kma.go.kr/data/grnd/selectAwsRltmList.do?pgmNo=56. The Weather Research and Forecasting (WRF) model, version 4.1, and the Model for Prediction Across the Scales (MPAS), version 7.0, are available at https://github.com/wrf-model/WRF/releases/tag/v4.1 and https://github.com/MPAS-Dev/MPAS-Model/releases/tag/v7.0, respectively. The scripts for running rMPAS and WRF Models, along with the static file for the East Asia region of MPAS, are available at https://doi.org/10.5281/zenodo.14087443.
REFERENCES
Chen, F., and J. Dudhia, 2001: Coupling an advanced land surface–hydrology model with the Penn state–NCAR MM5 modeling system. Part I: Model implementation and sensitivity. Mon. Wea. Rev., 129, 569–585, https://doi.org/10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.
Cheng, Y., P. Hui, D. Liu, J. Fang, S. Wang, and J. Tang, 2023: MPAS-A variable-resolution simulations for summer monsoon over China: Comparison between global and regional configuration. J. Geophys. Res. Atmos., 128, e2022JD037541, https://doi.org/10.1029/2022JD037541.
Cheong, S.-H., K.-Y. Byun, and T.-Y. Lee, 2006: Classification of snowfalls over the Korean Peninsula based on developing mechanism. Atmosphere, 16, 33–48.
Choi, S.-J., and J. B. Klemp, 2021: A new hybrid sigma-pressure vertical coordinate with smoothed coordinate surfaces. Mon. Wea. Rev., 149, 4077–4089, https://doi.org/10.1175/MWR-D-21-0086.1.
Durran, D., J. A. Weyn, and M. Q. Menchaca, 2017: Practical considerations for computing dimensional spectra from gridded data. Mon. Wea. Rev., 145, 3901–3910, https://doi.org/10.1175/MWR-D-17-0056.1.
Fairall, C. W., E. F. Bradley, J. E. Hare, A. A. Grachev, and J. B. Edson, 2003: Bulk parameterization of air–sea fluxes: Updates and verification for the COARE algorithm. J. Climate, 16, 571–591, https://doi.org/10.1175/1520-0442(2003)016<0571:BPOASF>2.0.CO;2.
Fowler, L. D., W. C. Skamarock, G. A. Grell, S. R. Freitas, and M. G. Duda, 2016: Analyzing the Grell–Freitas convection scheme from hydrostatic to nonhydrostatic scales within a global model. Mon. Wea. Rev., 144, 2285–2306, https://doi.org/10.1175/MWR-D-15-0311.1.
Gehring, J., A. Ferrone, A.-C. Billault–Roux, N. Besic, K. D. Ahn, G. Lee, and A. Berne, 2020: Radar and ground-level measurements of precipitation collected by the École Polytechnique Fédérale de Lausanne during the International Collaborative Experiments for PyeongChang 2018 Olympic and Paralympic winter games. Earth Syst. Sci. Data, 13, 417–433, https://doi.org/10.5194/essd-2020-134.
Grell, G. A., and S. R. Freitas, 2014: A scale and aerosol aware stochastic convective parameterization for weather and air quality modeling. Atmos. Chem. Phys., 14, 5233–5250, https://doi.org/10.5194/acp-14-5233-2014.
Harris, L. M., and D. R. Durran, 2010: An idealized comparison of one-way and two-way grid nesting. Mon. Wea. Rev., 138, 2174–2187, https://doi.org/10.1175/2010MWR3080.1.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hong, S.-Y., and J.-O. J. Lim, 2006: The WRF Single-Moment 6-class microphysics scheme (WSM6). Asia-Pac. J. Atmos. Sci., 42, 129–151.
Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, https://doi.org/10.1175/MWR3199.1.
Hundsdorfer, W., B. Koren, M. vanLoon, and J. G. Verwer, 1995: A positive finite-difference advection scheme. J. Comput. Phys., 117, 35–46, https://doi.org/10.1006/jcph.1995.1042.
Iacono, M. J., J. S. Delamere, E. J. Mlawer, M. W. Shephard, S. A. Clough, and W. D. Collins, 2008: Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res., 113, D13103, https://doi.org/10.1029/2008JD009944.
Jiménez, P. A., and J. Dudhia, 2012: Improving the representation of resolved and unresolved topographic effects on surface wind in the WRF model. J. Appl. Meteor. Climatol., 51, 300–316, https://doi.org/10.1175/JAMC-D-11-084.1.
Kim, K., W. Bang, E.-C. Chang, F. J. Tapiador, C.-L. Tsai, E. Jung, and G. Lee, 2021: Impact of wind pattern and complex topography on snow microphysics during international collaborative experiment for Pyeongchang 2018 Olympic and Paralympic winter games (ICE-POP 2018). Atmos. Chem. Phys., 21, 11 955–11 978, https://doi.org/10.5194/acp-21-11955-2021.
Klemp, J. B., 2011: A terrain-following coordinate with smoothed coordinate surfaces. Mon. Wea. Rev., 139, 2163–2169, https://doi.org/10.1175/MWR-D-10-05046.1.
Knievel, J. C., G. H. Bryan, and J. P. Hacker, 2007: Explicit numerical diffusion in the WRF model. Mon. Wea. Rev., 135, 3808–3824, https://doi.org/10.1175/2007MWR2100.1.
Ko, J., K.-S. Lim, K. Kim, G. Lee, G. Thompson, and A. Berne, 2022: Simulated microphysical properties of winter storms from bulk-type microphysics schemes and their evaluation in the Weather Research and Forecasting (v4.1.3) model during the ICE-POP 2018 field campaign. Geosci. Model Dev., 15, 4529–4553, https://doi.org/10.5194/gmd-15-4529-2022.
Kramer, M., D. Heinzeller, H. Hartmann, W. van den Berg, and G.-J. Steeneveld, 2020: Assessment of MPAS variable resolution simulations in the grey-zone of convection against WRF model results and observations: An MPAS feasibility study of three extreme weather events in Europe. Climate Dyn., 55, 253–276, https://doi.org/10.1007/s00382-018-4562-z.
Kwon, S., G. Lee, and G. Kim, 2015: Rainfall estimation from an operational S-band dual-polarization radar: Effect of radar calibration. J. Meteor. Soc. Japan, 93, 65–79, https://doi.org/10.2151/jmsj.2015-005.
Lim, K.-S. S., J.-M. Lim, H. H. Shin, J. Hong, Y.-Y. Ji, and W. Lee, 2019: Impacts of subgrid-scale orography parameterization on simulated atmospheric fields over Korea using a high-resolution atmospheric forecast model. Meteor. Atmos. Phys., 131, 975–985, https://doi.org/10.1007/s00703-018-0615-4.
Lim, K.-S. S., E.-C. Chang, R. Sun, K. Kim, F. J. Tapiador, and G. Lee, 2020: Evaluation of simulated winter precipitation using WRF-ARW during the ICE-POP 2018 field campaign. Wea. Forecasting, 35, 2199–2213, https://doi.org/10.1175/WAF-D-19-0236.1.
Liou, Y.-C., and Y.-J. Chang, 2009: A variational multiple–Doppler radar three-dimensional wind synthesis method and its impacts on thermodynamic retrieval. Mon. Wea. Rev., 137, 3992–4010, https://doi.org/10.1175/2009MWR2980.1.
Liou, Y.-C., S.-F. Chang, and J. Sun, 2012: An application of the immersed boundary method for recovering the three-dimensional wind fields over complex terrain using multiple-Doppler radar data. Mon. Wea. Rev., 140, 1603–1619, https://doi.org/10.1175/MWR-D-11-00151.1.
Liou, Y.-C., J.-L. Chiou, W.-H. Chen, and H.-Y. Yu, 2014: Improving the model convective storm quantitative precipitation nowcasting by assimilating state variables retrieved from multiple-Doppler radar observations. Mon. Wea. Rev., 142, 4017–4035, https://doi.org/10.1175/MWR-D-13-00315.1.
Morcrette, J.-J., H. W. Barker, J. N. S. Cole, M. J. Iacono, and R. Pincus, 2008: Impact of a new radiation package, McRad, in the ECMWF Integrated Forecasting System. Mon. Wea. Rev., 136, 4773–4798, https://doi.org/10.1175/2008MWR2363.1.
Oh, Y.-A., H.-L. Kim, and M.-K. Suk, 2020: Clutter elimination algorithm for non-precipitation echo of radar data considering meteorological and observational properties in polarimetric measurements. Remote Sens., 12, 3790, https://doi.org/10.3390/rs12223790.
Park, H.-H., J. Lee, E.-C. Chang, and M. Joh, 2019a: High-resolution simulation of snowfall over the Korean eastern coastal region using WRF model: Sensitivity to domain nesting-down strategy. Asia-Pac. J. Atmos. Sci., 55, 493–506, https://doi.org/10.1007/s13143-019-00108-x.
Park, S.-H., J. B. Klemp, and W. C. Skamarock, 2014: A comparison of mesh refinement in the global MPAS-A and WRF models using an idealized normal-mode baroclinic wave simulation. Mon. Wea. Rev., 142, 3614–3634, https://doi.org/10.1175/MWR-D-14-00004.1.
Park, S.-H., J. B. Klemp, and J.-H. Kim, 2019b: Hybrid mass coordinate in WRF-ARW and its impact on upper-level turbulence forecasting. Mon. Wea. Rev., 147, 971–985, https://doi.org/10.1175/MWR-D-18-0334.1.
Skamarock, W. C., 2004: Evaluating mesoscale NWP models using kinetic energy spectra. Mon. Wea. Rev., 132, 3019–3032, https://doi.org/10.1175/MWR2830.1.
Skamarock, W. C., and J. B. Klemp, 2008: A time-split nonhydrostatic atmospheric model for weather research and forecasting applications. J. Comput. Phys., 227, 3465–3485, https://doi.org/10.1016/j.jcp.2007.01.037.
Skamarock, W. C., and A. Gassmann, 2011: Conservative transport schemes for spherical geodesic grids: High-order flux operators for ODE-based time integration. Mon. Wea. Rev., 139, 2962–2975, https://doi.org/10.1175/MWR-D-10-05056.1.
Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., https://doi.org/10.5065/D68S4MVH.
Skamarock, W. C., J. B. Klemp, M. G. Duda, L. D. Fowler, S.-H. Park, and T. D. Ringler, 2012: A multiscale nonhydrostatic atmospheric model using centroidal Voronoi tesselations and C-Grid staggering. Mon. Wea. Rev., 140, 3090–3105, https://doi.org/10.1175/MWR-D-11-00215.1.
Skamarock, W. C., M. G. Duda, S. Ha, and S.-H. Park, 2018: Limited-area atmospheric modeling using an unstructured mesh. Mon. Wea. Rev., 146, 3445–3460, https://doi.org/10.1175/MWR-D-18-0155.1.
Tsai, C. L., K. Kim, Y.-C. Liou, G. Lee, and C.-K. Yu, 2018: Impacts of topography on airflow and precipitation in the Pyeongchang area seen from multiple-Doppler radar observations. Mon. Wea. Rev., 146, 3401–3424, https://doi.org/10.1175/MWR-D-17-0394.1.
Tsai, C.-L., K. Kim, Y.-C. Liou, and G. Lee, 2023: High-resolution 3D winds derived from a modified WISSDOM synthesis scheme using multiple Doppler lidars and observations. Atmos. Meas. Tech., 16, 845–869, https://doi.org/10.5194/amt-16-845-2023.
Wicker, L. J., and W. C. Skamarock, 2002: Time-splitting methods for elastic models using forward time schemes. Mon. Wea. Rev., 130, 2088–2097, https://doi.org/10.1175/1520-0493(2002)130<2088:TSMFEM>2.0.CO;2.
Xue, M., 2000: High-order monotonic numerical diffusion and smoothing. Mon. Wea. Rev., 128, 2853–2864, https://doi.org/10.1175/1520-0493(2000)128<2853:HOMNDA>2.0.CO;2.
Yeo, N., and E.-C. Chang, 2018: Impact of the land-sea temperature contrast on the snowfall structure over the western coastal region of the Korean Peninsula. Asia-Pac. J. Atmos. Sci., 54, 253–263, https://doi.org/10.1007/s13143-018-0011-5.
Zhao, C., and Coauthors, 2019: Modeling extreme precipitation over East China with a global variable-resolution modeling framework (MPASv5.2): Impacts of resolution and physics. Geosci. Model Dev., 12, 2707–2726, https://doi.org/10.5194/gmd-12-2707-2019.
