Journal of Geophysical Research: Atmospheres

Effects of bulk ice microphysics on the simulated monsoonal precipitation over east Asia

Authors


Abstract

[1] This study examines the effects of bulk ice microphysical processes on the simulation of monsoonal precipitation in summer over east Asia, centered over Korea. The mixed phase microphysics scheme of the WRF-Single-Moment-MicroPhysics class 5 (WSM5) is implemented into the fifth-generation Pennsylvania State University/National Center for Atmospheric Research Mesoscale Model (MM5). The performance of the WSM5 scheme is compared to that of Reisner's mixed phase scheme, that is, the MM5-Single-moment-Microphysics class 5 (MSM5). Together with looking at the impact of ice microphysics, the importance of the sedimentation of falling ice crystals on simulated rainfall events is investigated. The same sensitivity experiments are extended to a 2-month-long simulation of the east Asian summer monsoon. It was found that the new microphysics in the WSM5 scheme produces a more realistic vertical distribution of condensates. For a locally developed heavy rainfall event over Korea, the impact of revised ice microphysics is significant. The WSM5 scheme simulates more (less) precipitation in the south (north), compared with that of the MSM5 scheme, by stabilizing the air columns in the rainfall area, leading to a better agreement with the observed precipitation. By contrast, ice sedimentation becomes more important for those cases of heavy rainfall, associated with a mobile surface cyclone system, accomplished by suppressing large-scale bias through a realistic ice cloud/radiation feedback. The sedimentation of cloud ice is found to be crucial to the successful simulation of monsoonal precipitation and large-scale features within the east Asian summer monsoon.

1. Introduction

[2] Recently, Hong et al. [2004, hereinafter referred to as HDC] have suggested a revised ice microphysical process in order to overcome the deficiencies in widely used bulk parameterization schemes of clouds and precipitation in mesoscale models. They significantly modified the ice microphysical processes utilized by Rutledge and Hobbs [1983, 1984] and Lin et al. [1983]. Their improved microphysics includes processes such as number concentration, accretion, and ice nucleation of cloud ice. The most distinguishing point among these is that of practically representing ice microphysics process through making the assumption of ice nuclei number concentrations as a function of temperature, and a new and separate assumption of ice crystal number concentrations as a function of ice amount (Table 1).

Table 1. Major Components of the Ice Cloud Processes Suggested by Hong et al. [2004]
ComponentValue
Number concentration of cloud iceNI = cqI)d
Ice nuclei numberNI0 = 103 exp[0.1(T0T)]
Sedimentation of cloud iceVI (ms−1) = 1.49 × 104D1.31
Intercept parameter for snowN0S (m−4) = 2 × 106 exp{0.12(T0T)}

[3] HDC implemented the revised microphysics onto the WRF model and evaluated the performance of the proposed scheme for 2D idealized thunderstorms and 3D heavy rainfall events over Korea. HDC found that the modifications introduced to microphysical processes play a role in significantly reducing cloud ice and increasing snow at colder temperatures, while slightly increasing cloud ice and decreasing snow at warmer temperatures. They stressed that, together with the sedimentation of ice, this new microphysics produces a significant improvement in high-cloud amount, surface precipitation, and large-scale mean temperature through a better representation of ice cloud/radiation feedback.

[4] This study is an extension of HDC's work, further examining the performance of their approach. Originally, tests of their revised ice microphysics were limited in that the proposed approach was evaluated for the simulation of a single heavy rainfall event with a simple ice microphysics scheme that did not allow for the mixed phase. Also, the 45-km resolution was too coarse to resolve mesocale features. In this study, their revised ice microphysics scheme is implemented into the MM5 and the results were compared with those of the existing microphysics scheme. Two events were selected: heavy rainfall that occurred on 23–25 June 1997 and a band type of heavy precipitation over Korea on 14–15 July 2001. Additionally, regional climate experiments were executed to test the performance of new ice microphysical processes and their stability in terms of long-range forecast. Section 2 describes a synoptic overview of the selected heavy rainfall events and the synoptic settings of the summer of 2002 that were associated with the flooding over Korea. In section 3, the numerical experiments conducted in this study are described, with their results being discussed in section 4. A summary and concluding remarks appear in the final section.

2. Selected Heavy Rainfall Events

2.1. Heavy Rainfall Over Korea During 14–15 July 2001 (Case 1)

[5] A significant amount of precipitation was recorded in Korea on 15 July 2001, with a local maximum of about 371.5 mm near Seoul (Figure 1a). Rainfall data are based on the observed station data over South Korea, which has 73 stations. Most of the rainfall was observed during the 12-hour period from 1200 UTC 14 to 0000 UTC 15 July 2001, and the maximum rainfall intensity was 99.5 mmh−1 (Figure 1b). At 0000 UTC 14 July 2001 (not shown), there was a low-pressure system centered at the Yangtze River, with a warm front to the east and a cold front to the southwest, while north of it, high-pressure systems extended from Manchuria to Siberia. Meanwhile, a subtropical high lay on Japan. After 12 hours (Figure 1c), the low-pressure system traveled northeastward over the Yellow Sea. This was 3 hours prior to the onset of heavy rainfall over Seoul, Korea. During the heavy precipitation, high-pressure systems located on the northern side of the peninsula prevented the monsoon front from moving northward, tying it up over Korea. Associated with this low-pressure system, convective activity was visible both to the south of the cold front and to the north of the warm front (Figure 1d).

Figure 1.

(a) Observed 24-hour accumulated precipitation (mm) valid at 0000 UTC 15 July 2001, (b) radar image of rain rate at 1700 UTC 14 July, (c) surface weather map analysis, and (d) satellite image at 1200 UTC 14 July 2001. Shading in Figure 1a indicates where the value is over 90 (mm). All data are provided by the Korea Meteorological Administration (KMA).

[6] In association with significant rainfall over Korea, it can be seen that a southerly low-level jet (LLJ) brought moisture northward to a stationary monsoon front associated with a cyclonic circulation centered over the low system and an anticyclonic circulation along the western flank of a subtropical high at 850 mb (Figure 2a). As the surface low-pressure system moved northeastward, the LLJ south of the monsoon front intensified. The intensification of anticyclone circulation associated with a high-pressure system in northern China helped strengthen the frontal circulation and caused heavy rainfall. The 500 mb analysis indicated the intensification of baroclinicity in China in relation to the development of a surface low-pressure system (Figure 2b). A 500 mb temperature trough was located over northern China, west of the heavy precipitation region over Korea. The 250 mb analysis circulation was also dynamically favorable to the heavy precipitation mechanism (Figure 2c). The heavy precipitation region was located south of the exit of the upper level jet.

Figure 2.

(a) Analyzed 850 mb wind vectors, geopotential height (solid line, m), (b) 500 mb wind vectors, geopotential height (solid line, m) and temperature (dashed line, °C) and (c) 250 mb wind vectors, geopotential height (solid line, m) and temperature (dashed line, °C) at 1200 UTC 14 July 2001. Shading denotes the areas that mixing ratio is greater than 13 gkg−1 (Figure 2a), wind speed greater than 20 ms−1 (Figure 2b) and 40 ms−1 (Figure 2c). The NCEP-GDAS data were interpolated onto the 45-km model grid through the MM5 analysis system.

[7] Hong [2004] selected this case for identifying differences in mechanisms responsible for heavy rainfall occurring over geographically different regions. From the modeling studies with different precipitation physics, Hong [2004] demonstrated that the removal of the convective instability by the cumulus parameterization scheme is an essential process for heavy rainfall over the US, whereas it plays an insignificant role in reproducing heavy rainfall over Korea.

2.2. Heavy Rainfall Over Korea During 23–25 June 1997 (Case 2)

[8] This event is the same case studied by HDC. A significant amount of precipitation was recorded in Korea on 25 July 1997 (Figure 3a), with a local maximum of about 170 mm in the west-central part of South Korea, and another of about 90 mm at the southeastern flank of the peninsula. The maximum intensity was about 100 mm/day at the western flank of central Korea (Figure 3b). Most of the rainfall was observed during the 12-hour period of 0000–1200 UTC 25 June 1997. Before the onset of heavy precipitation at 0000 UTC 25, moderate rain was observed at the southwestern part of Korea, with less than 40 mm between 1200 UTC 24 and 0000 UTC 25 June 1997. Associated with the heavy rainfall, there was a low-pressure system centered over the Yellow Sea, with a warm front to the east and a cold front to the southwest, which was embedded within a monsoon circulation (Figure 3c). Associated with this mobile low-pressure system, convective activity was visible both to the south of the cold front and to the north of the warm front (Figure 3d).

Figure 3.

(a) Twenty-four-hour accumulated precipitation (mm) over the Korean peninsula, valid at 1200 UTC 25 June 1997, (b) radar image of rain rate at 0300 UTC and (c) surface analyses, and (d) satellite image at 0000 UTC 25 June 1997.

[9] From the upper level analyses for case 2 (not shown), many similarities were found, including a southerly low-level jet (LLJ) bringing moisture northward, an anticyclonic circulation along the western flank of a subtropical high, a thermal ridge over the peninsula, a trough to the west of the surface low system, and an upper level jet dynamically favorable for upward motion over the precipitation region. These LLJ and upper level patterns satisfied the synoptic circulation conditions favorable for developing heavy precipitation over the Korean peninsula [Uccellini and Johnson, 1979; Lee et al., 1998; Chen et al., 1999; Sun and Lee, 2002].

[10] Although the large-scale patterns associated with the two heavy rainfall events are similar at a mature stage, there are differences in some respects. First, the intensity of precipitation is stronger and its distribution is more localized for case 1 (cf. Figures 1a and 1b versus Figures 3a and 3b). Also, the precipitation for case 1 was activated in conjunction with the stagnant monsoonal front elongating from west to east, whereas case 2 was associated with a mobile low-pressure system traveling northeastward (cf. Figure 1c versus Figure 3c). Twelve hours prior to the onset of heavy precipitation over Korea, case 1 indicates that the monsoon front was located in southern Korea at 0000 UTC 14 July 2001, whereas the center of the low system was located in south China at 1200 UTC 24 June 1997, according to case 2. In summary, the heavy precipitation in case 1 was locally developed when a stagnant monsoon front was activated over the Korean peninsula; thus the precipitation pattern shows a band type. On the other hand, in case 2, precipitation over Korea was recorded as a mobile low-pressure system was moving from south China to the Korean peninsula and precipitation activity was relatively weak and widespread.

2.3. Synoptic Features of July and August 2002

[11] The typical summer over Korea is characterized by heavy rainfall, mainly connected with the Changma front in June and July (and/or sometimes concerned with typhoon activity) and a hot spell from late July to mid-August. However, there was an abnormally heavy rain period in the summer of 2002, with significant rain being recorded over the whole nation during the 12 days from 4 to 15 August (Figure 4). The area mean precipitation amount in August 2002 reached about 400 mm, or 30% of the annual total.

Figure 4.

Daily precipitation (mm) averaged over South Korea for the summer of 2002. Shaded areas indicate Changma period (19 June to 24 July) and an abnormal heavy rainfall period (4–15 August). All data are provided by the Korea Meteorological Administration (KMA).

[12] Precipitation anomalies show more rainfall than normal extending from south China to northern Japan and through the Korean peninsula, whereas less rainfall than normal was present to the north and south of the positively anomalous regions (Figure 5a). Positive anomalies in south China were determined to be due to a period of unusual rainfall in July, whereas anomalies over Korea and northern Japan were related to flooding in August (not shown). At upper levels (Figure 5b), two positive height anomalies were visible at 500 mb, with the one over the Mongolian region being associated with a strong ridge and the other, to the south of Japan, exhibiting anomalously strong subtropical highs over the northwestern Pacific. Between these two positive anomalies, relatively weak negatively anomalous regions stretched from south China to northern Japan and across the Korean peninsula. At 850 mb, low-level convergence zones extended from south China northeastward. Again, areas of convergence over south China were associated with flooding in July. The convergence zones over Korea indicate a large amount of moisture transporting toward Korea and northern Japan, forming unstable synoptic conditions over the Korean peninsula.

Figure 5.

(a) Precipitation anomalies (mm) from Xie and Arkin [1996] and (b) 500 mb height (lines, m) and 850 mb wind (vector) anomalies from the NCEP-NCAR reanalysis for July to August of 2002. The climatology is averaged for 1979–2003. Solid (dotted) lines indicate positive (negative) anomalies.

3. Numerical Experimentation Setup

3.1. Model

[13] The MM5 (Fifth-Generation Pennsylvania State University/National Center for Atmospheric Research (PSU/NCAR) Mesoscale Modeling System version 3.5 [Grell et al., 1994]) is used in this study. The MM5 is a 3D nonhydrostatic, nonlinear primitive equation model using Cartesian coordinates in the horizontal and terrain-following sigma coordinates in the vertical. This includes prognostic equations for pressure, the three wind components, temperature, and water vapor mixing ratio.

[14] The Kain-Fritsch cumulus parameterization scheme [Kain and Fritsch, 1993] is selected to account for subgrid-scale precipitation processes. A nonlocal vertical diffusion scheme is used to calculate the vertical fluxes of sensible heat, latent heat, and momentum [Hong and Pan, 1996]. A simple shortwave radiation scheme [Dudhia, 1989], and a multiband longwave radiation package are also used [Mlawer et al., 1997].

3.2. Model Setup for the Two Heavy Rainfall Events

[15] The model configuration consists of a nested domain configuration defined in Lambert conformal space (Figure 6). The inner grid (domain 2, 100 × 100), with a resolution of 15 km, is nested by a 45-km grid model (domain 1, 80 × 80) by one-way interaction. Both grid systems have 23 vertical levels and the model top is located at 50 mb. Initial and boundary conditions are based on the global analysis and prediction system that is routinely produced at the NCEP (National Centers for Environmental Prediction) Global Data Assimilation System (GDAS).

Figure 6.

Model terrain contoured every 100 m for 45-km domain (D01). The inner box (D02) designates the 15-km domain. Terrain heights greater than 1000 m are shaded.

[16] The mixed phase microphysics scheme of HDC, the WRF-Single-moment-Microphysics class 5 (WSM5), was implemented into the MM5. The performance of the new scheme is compared with that of Reisner's mixed phase scheme, the MM5-Single-moment-Microphysics class 5 (MSM5). In this mixed phase scheme, there are five prognostic equations, corresponding to: water vapor, cloud water, cloud ice, rain, and snow. Three experiments were carried out. The MSM5 experiment employs the existing mixed microphysical process in the MM5 [Reisner et al., 1998], whereas the WSM5 experiment includes the ice microphysics suggested by HDC. The NOVI experiment, excluding ice sedimentation in the WSM5 experiment, investigates the importance of the sedimentation of falling ice crystals. Another experiment excluding the ice sedimentation in the MSM5 scheme resulted in the same impact sensitivity in the WSM5 scheme, and will not be further discussed in this paper. For case 1, the experiments are carried out over 24 hours, running from 0000 UTC 14 to 0000 UTC 15 July 2001. The experiments run for 48 hours, from 1200 UTC 23 to 1200 UTC 25 June 1997, in case 2.

3.3. Model Setup for the Regional Climate Experiments

[17] A general circulation model (GCM) is the ultimate testbed for evaluating cloud parameterization schemes. However, running a global model is very time consuming. Also, the impact of physics changes in GCM is difficult to be understood since interaction between the physical processes in 3D model framework is highly nonlinear. As an alternative, a single-column model (SCM) is a useful testbed [Randall et al., 1996] since it is more controllable and faster, but has an uncertainty in the horizontal condensate advection. For highly advective conditions, a SCM may not be a suitable testbed for cloud parameterizations [Ghan et al., 2000]. Therefore, in this study, we chose a regional climate model (RCM) as our testbed. The RCM can internally generate the appropriate dynamical forcings, including condensate advection for all grid columns set well away from the lateral domain boundaries.

[18] A regional climate run experimental setup is the same as for the case experiments except for the use of a 45-km-resolution single domain and initial and boundary conditions with the NCEP-NCAR reanalysis [Kalnay et al., 1996]. The simulations were executed from 0000 UTC 1 July to 1800 UTC 31 August 2002, which is forced by the reanalysis data that are available at 6-hour intervals. Large-scale forcing from the reanalysis data is limited in the buffer zone of the lateral boundary as in the case experiments.

4. Results

[19] For the heavy rainfall case experiments, this discussion will focus on the 15-km experiments since the 45-km run shows a similar sensitivity.

4.1. Heavy Rainfall Over Korea During 14–15 July 2001

[20] Figure 7 shows the predicted 24-hour accumulated precipitation valid at 0000 UTC 15 July 2001. The model with the WSM5 scheme captured the long banded feature of the observed rainfall extending from southwest to northeast across the center part of the Korean peninsula (see Figures 1b and 1d), but the maximum amount was underestimated (see Figure 1a). The observed maximum precipitation was 371.5 mm, whereas the WSM5 scheme produced the maximum at about 140 mm. The location of the simulated rainfall band also shifted to the north by about 100 km. Note that the portion of the simulated precipitation is mostly due to the grid-resolvable precipitation physics. In the WSM5 experiment, the convective parameterization (implicit) scheme and grid-resolvable microphysics (explicit) scheme are responsible for 0.3 mm/day and 17.3 mm/day, respectively, in terms of the domain-averaged 24-hour accumulated precipitation. This issue will be further discussed in the analysis of the regional climate run (section 4.3).

Figure 7.

Twenty-four-hour accumulated rainfall (mm) ending at 0000 UTC 15 July 2001 obtained from the (a) WSM5 experiment and the differences, (b) WSM5 minus MSM5, and (c) WSM5 minus NOVI experiments. Shaded area in Figure 7a means that rainfall (mm) is over 90 (mm).

[21] In the MSM5 case, the local maximum increases to 180 mm, but the rainband was displaced further to the north (Figure 7b). It can be seen that the WSM5 experiment produced more (less) rainfall to the south (north) compared to the MSM5 experiment, resulting in a wider rainband. The effect caused by ice sedimentation also shifts the rainband southward but this impact is negligible compared with changes in the microphysics (Figure 7c).

[22] Figures 8a–8c compare the vertical profiles of condensates and precipitation mixing ratios averaged over the heavy rainfall region centered over Korea. Figure 8d was deduced from a TRMM observation [Del Genio and Kovari, 2002]. This observed profile is just a reference for corresponding species since the retrieved profiles from Del Genio and Kovari depend upon the cloud model used in the retrieval algorithm. Despite this uncertainty in verification, we can say that the WSM5 experiment produces more realistic hydrometeor profiles, qualitatively agreeing with those from the TRMM observation, than the results derived from the MSM5 scheme. For example, a mixture of cloud ice and water exists between 300 mb to 400 mb in the MSM5 experiment, whereas the layer extends from 300 mb to 600 mb in the WSM5 experiment. Also, the level of maximum cloud water from the MSM5 experiment, at 400 mb, is relatively too high to the freezing level at 600 mb in the WSM5 experiment. Another distinct feature is the distribution of snow, whereby the maximum amount is lower in the WSM5 than in the MSM5 experiment. The amount of snow nearly doubled by introducing the new ice microphysics, producing a result closer to that of the TRMM observation.

Figure 8.

Vertical distributions of area averaged (33°N–41°N, 123°E–131°E) water species at 1800 UTC 14 July 2001 obtained from the (a) WSM5, (b) MSM5, and (c) NOVI experiments, and (d) TRMM observation [from Del Genio and Kovari, 2002]. Solid lines are for rain, dotted lines are for snow, dashed lines are for cloud water, and dot-dashed lines are for cloud ice.

[23] These unrealistic distributions of hydrometeors in the MSM5 experiment care rooted in the characteristics of ice microphysics, especially the number concentration of ice nuclei, a function of temperature [Fletcher, 1962]. As HDC pointed out, the ice number concentration is extremely high at colder temperatures in the Fletcher function, leading to a fast ice generation process at temperatures below −40°C and nearly no generation above it. Moreover, the conversion of cloud ice to snow is fast enough to remove existing ice particles at temperatures warmer than −30°C in the MSM5, which leads to the absence of cloud ice between the 400 and 500 mb layers. Thus too much cloud water is generated in that layer through condensation created by the removal of existing supersaturated water vapors in the MSM5 experiment.

[24] The distribution of the condensates in the NOVI experiment is very similar to that from the WSM5 case, except that the ice amount is much larger at colder temperatures and the snow amount is slightly smaller at warmer temperatures. In the NOVI experiment, cloud ice in the upper troposphere does not fall, resulting in the suppression of the accretion process by snow. It is noted that having more cloud ice at 200 mb with the NOVI experiment resembles the profile of a TRMM observation. Nevertheless, the inclusion of ice sedimentation was found to be crucial to the bulk parameterization of clouds and precipitation, as shown by HDC. This issue will be further discussed in the next section.

[25] Figure 9 shows the time series of the differences in temperature (°C) and water vapor (gkg−1) between the MSM5 and WSM5 experiments. The WSM5 run tends to stabilize the troposphere, compared to the results of the MSM5 scheme (Figure 9a). More heating by the WSM5 scheme at 400 mb and cooling below that level seem to be due to the fact that the WSM5 scheme increased the ice phase species around 600–300 mb, inducing an enhanced longwave heating. Cooling near the 600 mb level in the lower troposphere seems to have been due to enhanced melting into more ice phase particles sediment. This stabilizing effect of the large-scale environment induces a reduction of the local maximum precipitation and its widespread horizontal distribution through the introduction of new ice microphysics in the WSM5 scheme (see Figure 7).

Figure 9.

Pressure-time cross sections of differences in (a) temperature (K) and (b) water vapor (gkg−1) between the WSM5 and MSM5 experiments (WSM5 minus MSM5), averaged over heavy rainfall region (33–41.5°N, 131–141°E).

[26] It is apparent that the WSM5 experiment produces a lesser amount of water vapor in the upper troposphere and near the surface than the MSM5 experiment (Figure 9b). The drier upper troposphere in the WSM5 run reflects more consumed water vapor being converted to the ice phase. Drying near the surface and moistening above it in the WSM5 experiment cannot be explained straightforward in the three-dimensional framework since each microphysics term interacts with one another and the resulting hydrometeors change the large-scale environment.

4.2. Heavy Rainfall Over Korea During 23–25 June 1997

[27] Figure 10 shows the predicted 24-hour accumulated precipitation at 1200 UTC 25 June 1997. Overall, the WSM5 experiment successfully reproduced the rainfall in terms of position and pattern, in conjunction with the development of a low-pressure system traveling from south China (Figure 10a). The portions of the simulated precipitation due to the convective parameterization (implicit) scheme and grid-resolvable microphysics (explicit) scheme are 2.8 mm/day and 31.2 mm/day, respectively, in terms of the domain-averaged 24-hour accumulated precipitation. For this heavy rainfall event, the impact of cloud ice sedimentation is larger than the changes in ice microphysics (cf. Figures 10b and 10c).

Figure 10.

Twenty-four-hour accumulated rainfall (mm) ending at 1200 UTC 25 June 1997, obtained from the (a) WSM5 experiment, and the differences (b) WSM5 minus MSM5 and (c) WSM5 minus NOVI experiments.

[28] The relative sensitivity of the simulated precipitation to the sedimentation of cloud ice for these two cases is better shown by a comparison with the large-scale environment (Figure 11). Overall, the temperature and moisture fields from all the experiments show a similar deviation except for a distinct warm bias in the upper troposphere resulting from the NOVI experiment, indicating the importance of ice sedimentation. Enhanced longwave radiation heating due to excessive cirrus clouds caused a warm bias in the upper troposphere when ice sedimentation is excluded, as shown by HDC. It is noted that changes in the microphysics does not significantly affect the large-scale fields in terms of the averaged domain temperature. These sensitivities occur qualitatively in the same fashion for both cases, but they are quantitatively different. The impact of ice sedimentation is more significant for case 2 than for case 1. A possible reason for this can be attributed to the different synoptic environments causing heavy rainfall for each case. That is, the precipitation in case 2 is associated with a mobile low-pressure system traveling from China, whereas that for case 1 was locally activated on a stagnant frontal band. Thus the warm bias due to excessive cirrus clouds is more significant for case 2 than case 1. The impact of ice microphysics itself is most likely limited to the area of heavy precipitation. Since the precipitation over Korea is localized for case 1, the improvement of microphysics is more significant for that case.

Figure 11.

Vertical profiles of the difference in temperature from the analysis for the MSM5 (solid line), WSM5 (dotted line) and NOVI (long-dashed line) experiments, averaged over the whole domain for (a) case 1 during 1200 UTC 14 July to 0000 UTC 15 July 2001 and for (b) case 2 during 1200 UTC 24 June to 1200 UTC 25 June 1997.

[29] Meanwhile, note that the resulting precipitation from the WSM5 scheme in Figure 10a is not as good as the simulated precipitation in HDC (see Figure 9b of HDC). Preliminary studies showed that these differences are largely due to a more accurate dynamical frame in the WRF than in the MM5 model. As an example, Ham et al. [2005] showed that given an identical model setup for both WRF and MM5, WRF improves the simulation of heavy rainfall over east Asia, which is beyond of the scope in this study. For case 1, the modeled precipitation using the WRF model was better than the result in Figure 7a (not shown). The WRF model reduced the amount of precipitation on the mountain ridge and placed precipitation cores upwind toward what was observed.

4.3. Regional Climate Run for July and August of 2002

[30] Figure 12 compares the 2-month accumulated simulated precipitation during the period from 0000 UTC 1 July to 1800 UTC 31 August 2002. Observed precipitation in Figure 12a is based on the global telecommunication system (GTS) gauge-based analyses of daily precipitation over global land with the horizontal resolution of 0.5 ° (http://islscp2.sesda.com/ISLSCP2_1/html_pages/groups/hyd/gts_precip_daily_xdeg.html). The WSM5 experiment reproduced the observed precipitation fairly well in terms of the distribution and position of monsoonal rainfall (cf. Figures 12a and 12b). Two peaks, one at the southern tip and the other at the eastern flank of the Korean peninsula, are well simulated, even though the southern peak is displaced to the southwest by about 100 km. Another peak at the eastern flank and associated with a mountain ridge is well reproduced. The MSM5 run shifts the rainband at the southern tip farther to the north than that of the WSM5 experiment, with the localized precipitation at the eastern flank not as organized as that from the WSM5 experiment. The NOVI experiment hardly captures the southern peak of the rainfall. The spatial pattern correlation coefficient was computed between simulated and observed precipitation on the MM5 grids for each experiment. The coefficients over the land points obtained from WSM5, MSM5, and NOVI experiments in Figure 12 are 0.40, 0.32, and 0.19, respectively. For precipitation over the model domain including oceans, the Global Precipitation Climatology Project (GPCP) daily precipitation with a resolution of 1° [Huffman et al., 2001] was interpolated onto the model grid. The coefficients for the entire model domain are 0.37, 0.34, and 0.21, respectively.

Figure 12.

(a) Observed accumulated precipitation (mm) during the period from 0000 UTC 1 July to 0000 UTC 31 August 2002 and the corresponding simulation results from the (b) WSM5, (c) MSM5, and (d) NOVI experiments. Observed precipitation in Figure 12a is based on the station observation over land areas. Shadings in Figure 12b–12d denote the implicit precipitation due to the convective parameterization scheme.

[31] Note that the portion of the simulated precipitation over the Korean region is mostly explicitly resolved by the grid-resolvable precipitation physics. In Figure 12, it is seen that simulated precipitation by the convective parameterization (implicit) scheme appears in heavy precipitation region, but the amount less than 150 mm, whereas the total amount of precipitation for each experiment ranges up to 850 mm. In terms of the domain-averaged amount of precipitation, the portion of explicit rain is twice larger than that of implicit rain over the Korean region (Table 2). It is opposite in the analysis of the precipitation averaged over the entire model domain. Much of above-normal precipitation over south China in Figure 5a was implicitly resolved by the convective parameterization scheme.

Table 2. Daily Mean Precipitation (mm) for July and August 2002, Averaged Over the Korean Region (32–40°N, 122–131°E) and Over the Model Domain (Parentheses)
ExperimentImplicit RainExplicit RainTotal
WSM52.02 (3.98)4.28 (1.83)6.30 (5.81)
MSM52.05 (4.02)4.19 (1.74)6.24 (5.76)
NOVI1.91 (3.73)4.56 (2.00)6.47 (5.73)

[32] The simulated 300 mb temperature fields explain the relative consequence of microphysical processes and cloud ice sedimentation (Figure 13a). The model tends to produce slightly warm biases irrespective of the choice of microphysics schemes, but generates a pronounced warming during the simulation period in the NOVI experiment. This warm bias is due to excessive amounts of ice in the absence of ice sedimentation (Figure 13b). From the sensitivity experiments of cloud-radiation interaction, HDC explained that excessive longwave radiation heating, due to the trapping caused by excessive ice clouds above, causes the warm bias. A reduction of shortwave heating in the troposphere due to cirrus shading produces an effect opposite to that of the longwave, but has a much smaller impact.

Figure 13.

Domain-averaged (a) 300 mb temperature bias from analysis and (b) the volume averaged mixing ratio of cloud ice time series from WSM5 (solid line), MSM5 (dotted line) and NOVI (green solid line) experiments.

[33] In the GCM community, the implementation of sophisticated microphysics for grid-resolvable precipitation physics is primary used for realistic representation of the cloud-radiation feedback [e.g., Fowler et al., 1996]. The importance of the sedimentation of cloud ice has also been pointed out in short-range forecasts [Manning and Davis, 1997; Wang, 2001]. Our results indicated that this sedimentation of cloud ice becomes crucial in long-range forecasts, including seasonal forecasts and climate simulations, for suppressing the large-scale bias through an improved cloud-radiation feedback.

5. Conclusion

[34] This study examines the impact of ice microphysical processes on the simulation of heavy rainfall over Korea, embedded within the east Asian monsoonal circulation system. The mixed phase microphysics scheme of Hong et al. [2004], the WRF-Single-Moment-MicroPhysics class 5 (WSM5), is implemented into the fifth-generation Pennsylvania State University/National Center for Atmospheric Research Mesoscale Model (MM5). The performance of the WSM5 scheme is compared with that of Reisner's mixed phase scheme [Reisner et al., 1998], the MM5-Single-moment-Microphysics class 5 (MSM5). Two heavy rainfall events are selected: one a locally organized band-type mesoscale system on 14–15 July 2001 (case 1) and the other associated with a synoptically evolving convective system that occurred on 23–25 June 1997 (case 2). The impact of ice microphysics and sedimentation of ice particles is further examined for a regional climate-modeling framework driven by an observed large-scale forcing during the period from July to August 2002.

[35] For the event on 14–15 July 2001 (case 1), all the experiments capture the long banded feature of the observed heavy rainfall that is distributed from the southwest to northeast over the center of Korea; however, the maximum rainfall was underestimated. The new microphysics scheme, the WSM5, tends to cause the rainband to shift southward, growing closer to observations. The new scheme produces more ice phase species around the 600–300 mb layer than the existing scheme (MSM5) does, leading to an increased stability. As a result, the new microphysics scheme creates a more widespread simulation and shifts the precipitation southward. The impact of ice sedimentation is negligible for this case. While the impact of the new microphysics and cloud ice sedimentation on the second case (23–25 June 1997) is qualitatively similar to that of the case on 14–15 July 2001, the effect of cloud ice sedimentation is more significant than the changes in ice microphysics.

[36] A possible reason for this different sensitivity is found to be due to the different synoptic environments causing the heavy rainfall. That is, the precipitation for case 2 is associated with a mobile low-pressure system traveling from China, whereas that of case 1 was locally activated on a stagnant frontal band. The impact of ice microphysics itself is most likely limited to the area of heavy precipitation. Thus the improvement of microphysics is more significant for the heavy rainfall event that developed locally over the Korean peninsula (case 1). By contrast, the impact of ice sedimentation plays an important role in the simulation of heavy rainfall associated with a mobile surface cyclone (case 2). A warm bias due to excessive cirrus clouds is distinct when ice sedimentation is not taken into account. An incorrect representation of cloud-radiation interaction is found to cause large-scale bias, with the resulting impact possibly proving significant in long-range predictions, such as seasonal forecasts and climate simulations.

[37] Recently, Hong [2004] revealed that the removal of the convective instability by the cumulus parameterization scheme is an essential process for heavy rainfall event over the US, whereas it plays an insignificant role in reproducing heavy rainfall over Korea. He further showed that it is evident that summertime climatology over Korea is characterized by stronger baroclinicity. Climatologically, the Korean peninsula is characterized as thermodynamically neutral in contrast to large convective available potential energy (CAPE) over the US. Major findings of Hong [2004] support the overall importance of cloud ice microphysics in the explicit scheme to monsoonal precipitation identified in this study.

Acknowledgments

[38] The authors would like to express their gratitude to Korea Meteorological Administration (KMA) for providing the satellite images and observed precipitation analyses. This study is supported by the Korean Ministry of Science and Technology through National Research Laboratory Program, KOSEF through SRC Program and by KMA through the R&D Project. Comments by two anonymous reviewers are appreciated for improving the original manuscript.

Ancillary