1. Introduction
One of the challenges to improve simulated precipitation lies in the elaboration of the cloud and precipitation processes, which are parameterized by a microphysics scheme (MPS) as well as a cumulus parameterization scheme (CPS) in a regional climate or mesoscale model. An MPS simulates the precipitation based on gridcell-mean (i.e., resolved) variables when the gridcell-mean relative humidity is greater than 100%. Meanwhile, a CPS simulates the precipitation depending on the parameterized subgrid-scale (i.e., unresolved) cumulus convections, which can be conceptualized in many different ways. A CPS has always been at the core of the efforts that are directed toward the improvement of model performance (Arakawa 2004) and has been considered to be one of the most challenging aspects of numerical atmospheric modeling (Arakawa et al. 2011; Hong and Dudhia 2012), because a proper representation of the subgrid-scale convective process is essential for simulations of severe weather and climate events. In addition, a CPS exhibits a large uncertainty in prediction and simulation of weather and climate. For these reasons, many weather and climate prediction centers have made tremendous efforts to develop and update their CPS in order to improve their precipitation forecast skills (e.g., Tiedtke 1989; Han and Pan 2011).
Some of these efforts result in numerous studies on the CPS intercomparison from both the numerical weather prediction and climate modeling points of view. Wang and Seaman (1997) compared the performance of four CPSs in simulating a heavy rainfall event over the Great Plains in the United States, in which they demonstrated that no single scheme could be clearly better than the others. Similar studies with long-term simulations have also been conducted (Gochis et al. 2002; Ratnam and Kumar 2005; Kang and Hong 2008; Yu et al. 2011). For example, Gochis et al. (2002) examined the sensitivity of the North American monsoon system to the CPS and found substantial differences in the simulated precipitation, surface climate, and atmospheric stability. Ratnam and Kumar (2005) also showed that the spatial distribution and amounts of the Indian summer monsoon precipitation could be sensitive to the selected CPS. Kang and Hong (2008) evaluated four different CPSs for the simulated East Asian summer climatology and demonstrated distinct differences in the intraseasonal variability of simulated rainfall as well as the interannual variability of seasonal rainfall. Yu et al. (2011) evaluated three different CPSs on the summer monsoon case over China in the Weather Research and Forecasting (WRF) Model by conducting 9-yr simulations from 2000 to 2009. All CPSs reasonably could simulate the summer monsoon precipitation by reproducing the observed north–south shift of the monsoon rain belt.
Meanwhile, previous studies concerning the effects of a horizontal resolution on regional climate show that a high-resolution simulation does not always promise improvement in the simulated climate (Giorgi and Marinucci 1996; Rauscher et al. 2010; Varghese et al. 2011; Sharma and Huang 2012; Chan et al. 2013). Giorgi and Marinucci (1996) showed that increasing model resolution from 200- to 50-km grid spacing could not affect the simulated synoptic system structure, but it could increase the precipitation amount over western Europe. They concluded that the closure, which is based on moisture convergence and buoyant energy release by clouds, could behave differently at the different resolution and result in the precipitation sensitivity to grid spacing. Rauscher et al. (2010) simulated regional climates over Europe and showed enhanced skills in both the spatial pattern and temporal evolution of precipitation at 25 km compared to 50 km during the summer months, but not in the winter. Varghese et al. (2011) found the best performance with a coarse horizontal model resolution in a simulation of the meteorological parameters and chemical species concentrations. Sharma and Huang (2012) showed that the simulation under relatively coarse grid spacing (12 km) could be closer to observation relative to the other simulations under 6- and 3-km grid spacing within a regional modeling framework over Arizona. In addition, Chan et al. (2013) could not find any clear evidence to show that the 1.5-km simulation is superior to the 12-km simulation from the 17-yr (1991–2007) regional climate simulations over England.
On the other hand, the effects of the horizontal resolution on the simulated regional climate and the performance of the CPSs under various spatial resolutions have not been evaluated in detail nor have been documented over the East Asian region. This is probably due to a couple of reasons, such as growing societal demands for high-resolution climate information in recent years (Leung et al. 2003). As the spatial resolution of atmospheric models increases, it is expected that the surface precipitation will be parameterized more by an MPS than by a CPS because more clouds and precipitation processes can be resolved in a grid box under a higher-resolution modeling frame. However, no systematic study about the behavior of CPS over the East Asian region with varying grid spacing has been conducted to our knowledge.
WRF can serve well as a test bed, because it can easily be configured for various spatial resolutions and different CPSs. As of April 2013, 11 different CPSs are implemented in WRF, version 3.5. Among them, the Kain–Fritsch (KF; Kain 2004; Kain and Fritsch 1990), Betts–Miller–Janjić (BMJ; Betts and Miller 1993; Janjić 1994), and Grell–Dévényi (GD; Grell and Dévényi 2002) have been evaluated widely for monsoon rainfall (Kumar et al. 2012; Srinivas et al. 2013) as well as severe weather events (Raju et al. 2011; Nasrollahi et al. 2012). The effects of the convection-induced forcing on cumulus momentum transport are included only in the simplified Arakawa–Schubert (SAS) (Pan and Wu 1995), Tiedtke (Tiedtke 1989; Zhang et al. 2011), and Zhang and McFarlane (Zhang and McFarlane 1995) CPSs in WRF, version 3.5. Zhang et al. (2011) showed that the Tiedtke scheme successfully captures the main features of the observed marine boundary layer clouds and cloud regime transition, compared to other CPSs in WRF, version 3.2.1. Over East Asia, WRF has been widely used for the simulation of heavy rainfall (Lee et al. 2010; Choi et al. 2011; Yu and Lee 2011; Jung and Lee 2013), heavy snowfall (Lee et al. 2011; Jung et al. 2012), typhoons (Ma et al. 2012), and regional/urban climate (Yang et al. 2012; Kim et al. 2013). It is noted that all these studies over East Asia employed the KF scheme. Meanwhile, the performance of the SAS scheme has not been elucidated in the WRF community even though the SAS scheme is a competitive CPS compared to others for the simulation of the precipitation and summer climatology over East Asia (Park and Hong 2007; Kang and Hong 2008; Koo and Hong 2010) using a regional spectral model (RSM) (Juang et al. 1997). Recently, Han and Pan (2011) have significantly modified the SAS scheme, which is now used in the Global Forecast System (GFS) in the National Centers for Environmental Prediction (NCEP) (GFS SAS). To achieve a better simulation of the convective processes across horizontal scales, we have made a further modification of the recent version of the GFS SAS scheme, which is described in section 2 together with the overview of the GFS SAS scheme.
The objective of this study is twofold: (i) to evaluate the performance of both the original GFS SAS CPS and the updated one in WRF on East Asian monsoon precipitation and (ii) to evaluate the response of four different CPSs including the KF, BMJ, and Grell–Freitas (GF), which is the updated version of the GD scheme, and the GFS SAS to the spatial resolutions used. Section 3 overviews the experimental setup. The results of those experiments and the concluding remarks are presented in sections 4 and 5, respectively.
2. The GFS SAS cumulus parameterization scheme
a. Overview of the GFS SAS scheme
The Arakawa and Schubert (1974) CPS was simplified by Grell (1993) by considering only a single cumulus updraft–downdraft couplet within a single grid cell, thus leading to the SAS scheme (Pan and Wu 1995). Subsequent revisions (Hong and Pan 1998; Han and Pan 2006, 2011) have been made to this scheme after its first implementation in the NCEP Medium-Range Forecast Model in 1993 (Pan and Wu 1995). More significant changes, including the convective momentum transport with the effect of the convection-induced pressure gradient force (Han and Pan 2006), were made and implemented in the GFS model in late July 2010 (Han and Pan 2011). Major modifications by Han and Pan (2011) include a larger cloud-base mass flux and higher cloud tops. These changes effectively eliminate some of the remaining instability in the atmospheric column that is responsible for the excessive grid-scale precipitation.
b. Modification to the GFS SAS scheme
We implemented the most recent version of the GFS SAS scheme (Han and Pan 2011) in WRF (the original SAS), which was released in April 2011. In this study, two key modifications are made (the new SAS), which include (i) the triggering conditions and (ii) the convective mass flux.
In the new SAS scheme, 
Figure 1 shows the fcr values as a function of 
Variable fcr with respect to ω at the cloud base. The black (red) line is for the original (new) SAS scheme. Black solid and dashed–dotted lines indicate the values of the original SAS scheme over land and ocean, respectively. Red solid, dotted, and dashed lines indicate the values of the new SAS scheme under 24-, 48-, and 96-km grid spacing, respectively.
Citation: Weather and Forecasting 29, 5; 10.1175/WAF-D-13-00143.1
3. Model setup
WRF, version 3.5 (Skamarock et al. 2008), is used in our study. The physics packages other than the CPS include the WRF single-moment 6-class microphysics scheme (WSM6) (Hong and Lim 2006), the unified Noah land surface model (Chen and Dudhia 2001), a simple cloud-interactive shortwave radiation scheme (Dudhia 1989), the Rapid Radiative Transfer Model (RRTM) longwave radiation scheme (Mlawer et al. 1997), and the Yonsei University PBL scheme (Hong et al. 2006) for vertical diffusion. For the CPS, the original and new SAS schemes described in section 2 are used to evaluate the effects of the modification introduced in the GFS SAS scheme. The initial and boundary conditions are generated every 6 h by the NCEP Final Analysis (FNL) data on 1° × 1° global grids (available online at http://rda.ucar.edu/datasets/ds083.2/). The low boundary for the model is based on the Optimum Interpolation Sea Surface Temperature (OISST) on a 1° × 1° grid. The weekly OISST is linearly interpolated in time to derive daily values during the integration period.
The model configuration consists of a single domain defined on a Lambert conformal projection, and the domain covers the East Asian region centered over the Korean Peninsula. The entire grid system has 27 vertical layers and the model top is located at 10 hPa. The analysis domain is shown in Fig. 2, which excludes the boundary region. The model integration is conducted from 21 May to 1 September 2006, during the East Asian monsoon period. We use the former 10 days as a spinup time and analyze the model results during the 3 months from 1 June to 1 September. We perform the simulations under the six different grid spacing in order to figure out the effects of the model resolution as well as the effects of the modified GFS SAS CPS on the simulated monsoon precipitation and radiative fluxes. Toward this aim, the simulations are conducted with grid sizes of 96, 60, 48, 36, 24, and 12 km and the schemes of the original and new SAS, KF, GF, and BMJ, respectively. The integration time step is reduced with a decrease in the grid size, that is, 450, 300, 240, 180, 100, and 40 s. To investigate the effects of a decrease in the grid size and the integration time step separately, additional experiments in which the integration time step is set as 40 s regardless of the grid size are conducted with the original and new SAS schemes. All of the simulations are run using a spectral nudging technique for the three largest wave fields of temperature and wind. As addressed by Correia et al. (2008), the calling CPS time step can change the simulated precipitation and strength of the gravity wave. To avoid the sensitivity issue of the selected CPSs with respect to the calling CPS time step, we set the CPS called every model time step for all simulations conducted in our study.
Spatial distribution of accumulated precipitation (mm) during the 3 months from June to August 2006 from (a) TMPA, (b) ORG, and (c) NEW. Differences in precipitation are also shown (d) between ORG and TMPA (shaded: ORG minus TMPA) and between NEW and ORG (contour: NEW minus ORG), (e) between ORG (ORG with 12-km minus ORG with 96-km grid spacing), and (f) between NEW (NEW with 12-km minus NEW with 96-km grid spacing). Contour lines in (e) and (f) indicate the differences of the subgrid-scale precipitation from the cumulus scheme. The interval of the contour lines is (±) 50, 100, 200, and 400 mm.
Citation: Weather and Forecasting 29, 5; 10.1175/WAF-D-13-00143.1
4. Results and discussion
a. Effects of the modification on the GFS SAS scheme
Accumulated precipitation from observations [Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis (TMPA) (Huffman et al. 2007)] and simulation with the original SAS scheme (ORG) are shown in Figs. 2a and 2b, respectively. During the summer of 2006, major precipitation bands are located over South Korea, the southern part of Japan along with the East Sea, and the southeastern part of China (Fig. 2a). ORG under the 96-km grid spacing largely underestimates the rainfall over land and the region where a major monsoon band is located, indicating a much weaker monsoon compared to the observation (Fig. 2b). However, the new SAS scheme (NEW) captures the monsoon band well in terms of its location and strength under 96-km grid spacing (Fig. 2c). The bias of ORG (see shaded areas in Fig. 2d) has a clear distinction between over land and over ocean, that is, negative bias over land and positive bias over ocean. This distinction is significantly reduced in NEW (see contour lines in Fig. 2d) and improvement over land is more significant than over ocean. Table 1 shows the statistical skill scores of the simulated surface precipitation and radiative properties from NEW and ORG. The pattern correlation (PC) (0.69 vs 0.72) and root-mean-square error (RMSE) (170.6 vs 163.4) of the simulated precipitation indicates improvement in NEW. Even though slight deterioration is revealed in the simulation of the shortwave flux at the surface with NEW, the simulated radiative fluxes are comparable between NEW and ORG.
Statistical skill scores of PC and RMSE for the simulated surface precipitation, shortwave radiative flux at the surface (RAD_SW_SFC), and longwave radiative flux at the top of atmosphere (RAD_LW_TOA) with 96-km grid spacing. The values in parentheses are calculated using the simulated results with 12-km grid spacing. The observation of TMPA for precipitation and Clouds and the Earth’s Radiant Energy System (CERES) for radiative fluxes data are used for the statistics calculation. The horizontal resolutions of TMPA and CERES data are 0.25° and 1.0°, respectively.
Next, we explore how the GFS SAS scheme behaves with finer grid spacing. For this purpose, another set of simulations is conducted with 12-km grid spacing for ORG and NEW. With higher resolution (i.e., 12 km), both ORG and NEW produce more intense precipitation centers with an increasing total amount of precipitation (Figs. 2e and 2f). ORG shows better performance in terms of the statistical skill cores of precipitation under a high-resolution platform (Table 1). Even though the slight deterioration of the RSME under the fine grid spacing is shown in NEW due to excessive precipitation over South Korea and the southwestern part of Japan, the pattern correlation of the simulated precipitation is improved under 12-km grid spacing. The simulated radiative properties generally exhibit better performance in the simulation with finer grid spacing for both NEW and ORG except for the RMSE of the shortwave flux at the surface with ORG (see Table 1). Interestingly, the difference in the simulated subgrid precipitation from ORG between the two experiments (ORG with 96- and 12-km grid spacing), drawn with contour lines as shown in Fig. 2e, is similar to that of the total precipitation. On the other hand, the difference in the simulated subgrid precipitation from NEW between the two experiments (NEW with 96- and 12-km grid spacing) clearly shows the opposite sign of that of the total precipitation (Fig. 2f). In other words, increasing precipitation in NEW with 12-km grid spacing is mainly deduced from the increasing precipitation from the MPS.
The convective rain ratios (CRR) from the TRMM 3A12 data and from the simulations are shown in Fig. 3. The classification of convective and stratiform precipitation monthly TRMM 3A12 data is done through a method developed by Hong et al. (1999) using microwave brightness temperature. However, we cannot classify the modeled convective and stratiform precipitation within regional climate model (RCM) framework as a same manner as the one used with the TRMM data or the cloud-resolving model simulation. In RCM framework, the deep convective precipitation process is not usually resolved within a grid, which is usually larger than 10-km grid spacing, and should be parameterized by a CPS. The CRR in RCM simulation is defined as the ratio of subgrid-scale precipitation (from a CPS) to total precipitation. Because of the different definition of the CRR in TRMM and the simulations, the comparison of the CRR is carried out only in a qualitative manner. TRMM data (Fig. 3a) shows that the CRR is relatively higher over land than over ocean—especially, the western region of China shows larger CRR than other East Asian regions. Compared to TRMM, ORG produces too much subgrid-scale rain over ocean, resulting in the excessive CRR over the southern part of analysis domain (Fig. 3b). As seen in Fig. 3c, NEW also simulates the large CRR over ocean, but the contrast of CRR between ocean and land is distinct [i.e., less (more) CRR over ocean (land)], which is also seen in TRMM.
CRR (%) averaged during the 3 months from June to August 2006 from (a) TMPA, (b) ORG, and (c) NEW. Differences between (d) the ORG (ORG with 12-km minus ORG with 96-km grid spacing) and (e) between the NEW (NEW with 12-km minus NEW with 96-km grid spacing).
Citation: Weather and Forecasting 29, 5; 10.1175/WAF-D-13-00143.1
In the original SAS scheme, convections can be more easily triggered with greater convective mass flux at the cloud base over ocean than over land (see Fig. 1). By adopting the same threshold value of vertical velocity for both ocean and land, NEW relatively reduces the CRR over ocean by suppressing convections, compared to ORG (Fig. 3c). With finer resolution (Fig. 3d), ORG exhibits a broad increase of CRR, especially over land. Meanwhile, NEW shows a substantial decrease of CRR over most of the analysis domain (Fig. 3e). As the finer model grid spacing is used, the clouds and precipitation processes can be simulated more explicitly by an MPS, leading to a decrease of CRR (Hong and Dudhia 2012). Convection triggering is easier and the updraft mass flux is larger in ORG due to stronger vertical velocities with finer resolution, which leads to an increase of CRR with finer resolution. Note that as discussed in section 2b, the threshold values of the vertical velocity in ORG are constant and will be much smaller than those in NEW in finer resolution (e.g., 12-km resolution). On the other hand, increased subgrid-scale rain over land with ORG under 12-km grid spacing reduces the negative bias in ORG with 96-km grid spacing (Figs. 2d and 2e) and results in enhanced statistical skill score for surface precipitation (Table 1).
Figure 4 shows the probability density function (PDF) of rain rate from observation and simulations. ORG suffers from a lack of intense rain and excessive light rain. Even with finer grid spacing, ORG does not sufficiently reproduce moderate- and high-intensity precipitation rates, compared to the observation. NEW simulates better PDF of rain rate both with finer and coarser grid spacing compared to ORG by reducing light rain rate and increasing heavy rain rate. Figure 4 also indicates that the simulation with the reduced integration time step helps to improve the PDF of rain rate (cf. dashed and dotted lines for both NEW and ORG experiments). Overall, the NEW experiment performs better than ORG in terms of precipitation pattern and PDF distribution, and a high resolution helps to improve the regional climate simulation for both ORG and NEW; yet, improvement in ORG could be deduced from a counterintuitive response of a cumulus parameterization that is increasing subgrid precipitation under a fine grid spacing.
PDF of the rain rate. (top) The gray line is for TMPA, and the black (red) line is for the ORG (NEW) experiment. (bottom) The green lines are the results from the GF scheme, the blue lines are from the BMJ scheme, and the purple lines are the results from the KF scheme. Solid (dotted) lines indicate the results from 12-km (96 km) grid spacing. Dashed lines in (top) indicate the results from the 96-km grid spacing but with the integration time step of 40 s, as for the 12-km grid spacing.
Citation: Weather and Forecasting 29, 5; 10.1175/WAF-D-13-00143.1
b. Effects of the spatial resolution on the simulated surface rainfall
To have a more comprehensive understanding of how the response of a CPS to varying spatial resolution affects the simulated monsoon rainfall over East Asia, four different CPSs of GF, KF, and BMJ, including the new and original SAS, are chosen for a sensitivity study and tested with six different grid sizes (e.g., 12, 24, 36, 48, 60, and 96 km). The GF scheme is based on a stochastic approach using ensemble and data assimilation techniques, originally implemented by Grell and Dévényi (2002). In the GF, the subgrid transport associated with deep and shallow convective transport is coupled to the GD scheme for the smooth transition to cloud-resolving scales based on Arakawa et al. (2011). Interaction of aerosols with convective clouds is also considered through the processes of autoconversion and evaporation of cloud droplets in the GF scheme (Grell and Freitas 2013). The KF, which is the same mass-flux-type scheme as the GF scheme, handles the impact of convections on the atmospheric fields using mathematical equations (Kain and Fritsch 1990). The KF scheme uses the Lagrangian parcel method to estimate any possible existing instability. Kain (2004) adds the modification in the KF scheme for the representation of convective updraft–downdraft and closure assumption. The BMJ scheme, which is fundamentally different from the KF and GF schemes, is a large-scale quasi-equilibrium scheme and its reference profiles are based upon a similar-sounding structure obtained from tropical convections (Betts and Miller 1993; Janjić 1994). In the BMJ scheme, the precipitation is calculated directly from the amount of latent heat released. These schemes and resolution of 10-km order have been widely used in previous regional climate studies (e.g., Kang and Hong 2008; Zanis et al. 2009; Wu et al. 2013).
Figure 5 (top) shows the relative ratio of the simulated convective mass flux from the original and the new SAS schemes under different resolutions. The convective mass flux amount generally decreases with decreasing grid spacing in the new SAS. However, the original SAS scheme simulates more convective mass flux with decreasing grid spacing. As discussed in section 4a, this indicates that as the grid spacing decreases, the updraft mass flux is larger due to stronger vertical velocities under the constant threshold values of the vertical velocity in the original SAS scheme. According to Eqs. (1) and (2), the convection triggering is more difficult in the new SAS scheme than in the original one due to the large magnitude of the threshold vertical velocities. However, Fig. 5 (bottom) shows that as the grid spacing decreases, the new SAS scheme triggers convection much more frequently than the original scheme, which displays similar frequency of triggering convection with different resolutions. This suggests that under the finer grid spacing, it takes a longer time to stabilize the environment in the new SAS scheme due to the smaller mass flux, resulting in more convection triggering. Thus, we can conclude that more (less) precipitation under the finer grid spacing in the original (new) SAS scheme is caused by more (less) intense convective mass flux than by the more (less) frequent occurrence of convections.
Relative ratio of the (top) convective mass flux at the cloud base and (bottom) frequency of triggering convections under different model resolutions. Simulated convective mass flux and frequency of triggering convections under 96-km grid spacing are considered to be a reference value of 1. Black (red) line is for the ORG (NEW) experiment.
Citation: Weather and Forecasting 29, 5; 10.1175/WAF-D-13-00143.1
The CRR from the original SAS scheme does not vary much with the different grid spacing (black line in Fig. 6, top left). Meanwhile, the new SAS scheme is able to simulate a smaller CRR with decreasing grid spacing through the modified triggering conditions and convective mass flux depending on the horizontal grid spacing. The other three experiments employing the other CPSs, except the GF, show linearly decreasing trends of the CRR as the grid spacing decreases from 96 to 12 km (Fig. 6, top left). Interestingly, the GF simulates a relatively constant CRR of approximately 38.8% with grid spacing of 96–24 km. However, an abrupt change in the CRR is shown when the model resolution goes into around the gray-zone resolution, at 12-km spacing.
(top left) CRR over the total rain, (top right) total rain amount, (bottom left) PC, and (bottom right) RMSE with the TMPA data under different model resolutions for each experiment. Black (red) lines are the results from the ORG (NEW) experiment. Green lines are the results from the GF scheme, the blue lines are from the BMJ scheme, and the purple lines are the results of the KF scheme. All of the quantities are averaged over the entire analyzed domain. In (top right), the dashed gray line indicates the observation value from TMPA. Dashed lines indicate the results from the NEW and ORG experiments with the integration time step of 40 s.
Citation: Weather and Forecasting 29, 5; 10.1175/WAF-D-13-00143.1
Even though the different types of CPSs simulate a wide range of the CRR from 70.2% to 22.7% at 12-km grid spacing, the total amount of precipitation does not vary significantly (Fig. 6, top right). Different CPSs tend to remove the atmospheric instability in a different way and to stabilize the large-scale environment. After a cumulus parameterization process by a CPS, the adjusted large-scale variables, such as temperature, wind, and moisture field, are used for the resolved cloud and precipitation processes, which are explicitly treated in an MPS. A comparable amount of surface precipitation is produced regardless of the CPS chosen, because of feedback between CPS and MPS. For the new SAS scheme, the simulated precipitation amount reaches the observed amount when the model is run on 36-km grid spacing. If the grid resolution is finer than 36-km grid spacing, then the simulated precipitation amount is higher than the observed amount. Meanwhile, the other CPSs show better performance in terms of the rain amount as the grid spacing becomes finer, producing more precipitation. The simulations with the integration time step of 40 s do not change the convective rain ratio under different grid spacings, but produce more precipitation under coarser grid spacing (60 and 96 km) in the original and the new SAS schemes (see dashed lines in Fig. 6, top). Thus, the increase in total precipitation at the finer grid spacing can be explained by the total effects of the decrease in the model resolution and the integration time step. The PDF of the rain rate shows improvement under fine (12 km) grid spacing rather than under coarse (96 km) grid spacing (see Fig. 4, bottom).
However, all of the high-resolution simulations do not always assure the improvement of the spatial distribution of the simulated precipitation, as shown in Fig. 6 (bottom). The performance of the GF and BMJ (the original SAS and KF) gets worse (better) as the grid spacing becomes finer. If the monsoon precipitation is captured well with relatively coarse grid spacing, then a high resolution does not further improve the model skill, which can be seen in simulations with the new SAS, GF, and BMJ. Reduced integration time step slightly changes the simulated precipitation skill, but its effect is smaller than the effect of choice of CPSs (dashed line in Fig. 6, bottom). Meanwhile, the GF shows the worst capability in capturing the monsoon precipitation over the East Asian region (green line in Fig. 6, bottom).
5. Concluding remarks
Precipitation is one of the key variables in both the scientific and practical senses. As computing power grows and localized climate prediction as well as projection is needed in societal decision-making processes, high-resolution simulation has become a more realistic choice. Even though the simulated seasonal-mean pattern of the precipitation and synoptic system structure are not affected by the spatial resolution (Rauscher et al. 2010; Giorgi and Marinucci 1996), high-resolution simulation is needed to capture short time-scale phenomena, such as daily-scale heat waves and heavy rainfall (e.g., Yuan and Liang 2011), and to reasonably represent the orographic precipitation (Chan et al. 2013). This poses an important challenge in the weather forecasting and climate modeling community. What is reasonable behavior of convective precipitation as the spatial resolution increases?
In this study, we introduce a modification of the GFS SAS in order to consider the model spatial resolution in triggering convections and the simulated convective mass flux. A series of simulations with different spatial resolutions and CPSs are performed. The new SAS scheme exhibits overall improvement in simulating precipitation during the East Asian summer monsoon season in comparison to the original SAS scheme. Contrary to our general expectations, the CRR in the original SAS shows almost no dependency on the spatial resolution, indicating that the constant weaker threshold values of the vertical velocity in the original SAS scheme make convection trigger easier and updraft mass flux larger with increasing spatial resolution. Meanwhile, the CRR in the new SAS decreases as the spatial resolution decreases, showing that more precipitation is produced by grid-resolved processes with the finer spatial resolution, as expected.
Interestingly, the performance of high-resolution modeling does not always increase as the spatial resolution becomes higher. The improvement of the PDF of rain rate by higher-resolution simulations, on the other hand, is robust regardless of the choice of CPS. However, the overall performance matrix, such as the RMSE and the pattern correlation for the spatial distribution of the simulated precipitation, does not monotonically increase with an increase in the spatial resolution. This suggests the need for further improvement of the current CPSs or a new design of the CPS. As one of the efforts in improving CPS, Arakawa et al. (2011) suggest a framework in unifying conventional cumulus parameterization and cloud-resolving approaches. In doing so, a continuous transition from one to the other can be guaranteed as model grid space becomes smaller and smaller.
Acknowledgments
The authors would like to express their gratitude to Dr. Samson Hagos for his valuable comments and to acknowledge support of computing resources from the KISTI Super Computing Center through the Strategic Support Program for Supercomputing Application Research (Grant KSC-2012-G3-07). A portion of the computation is performed using the Pacific Northwest National Laboratory (PNNL) Institutional Computing (PIC) at PNNL. This study is supported by the Office of Science of the U.S. Department of Energy as part of Science Biological and Environmental Research under a bilateral agreement with the China Ministry of Sciences and Technology on regional climate research and the Earth System Modeling program. The PNNL is operated for DOE by Battelle Memorial Institute under Contract DE-AC05-76RL01830. The second author is supported by the R&D project on the development of global numerical weather prediction systems of the Korea Institute of Atmospheric Prediction Systems (KIAPS), funded by the Korea Meteorological Administration (KMA). Insightful comments offered by the three anonymous reviewers are highly appreciated.
REFERENCES
Arakawa, A., 2004: The cumulus parameterization problem: Past, present, and future. J. Climate, 17, 2493–2525, doi:10.1175/1520-0442(2004)017<2493:RATCPP>2.0.CO;2.
Arakawa, A., and Schubert W. H. , 1974: Interaction of a cumulus cloud ensemble with the large-scale environment. Part I. J. Atmos. Sci., 31, 674–701, doi:10.1175/1520-0469(1974)031<0674:IOACCE>2.0.CO;2.
Arakawa, A., Jung J.-H. , and Wu C.-M. , 2011: Toward unification of the multiscale modeling of the atmosphere. Atmos. Chem. Phys., 11, 3731–3742, doi:10.5194/acp-11-3731-2011.
Betts, A. K., and Miller M. J. , 1993: The Betts–Miller scheme. The Representation of Cumulus Convection in Numerical Models, Meteor. Monogr., No. 46, Amer. Meteor. Soc., 107–121.
Chan, S. C., Kendon E. J. , Fowler H. J. , Blenkinsop S. , Ferro C. A. T. , and Stephenson D. B. , 2013: Does increasing the spatial resolution of a regional climate model improve the simulated daily precipitation? Climate Dyn., 41, 1475–1495, doi:10.1007/s00382-012-1568-9.
Chen, F., and Dudhia J. , 2001: Coupling and 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, doi:10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.
Choi, H.-Y., Ha J.-H. , Lee D.-K. , and Kuo Y.-H. , 2011: Analysis and simulation of mesoscale convective systems accompanying heavy rainfall: The goyang case. Asia-Pac. J. Atmos. Sci., 47, 265–279, doi:10.1007/s13143-011-0015-x.
Correia, J., Arritt R. W. , and Anderson C. J. , 2008: Idealized mesoscale convective system structure and propagation using convective parameterization. Mon. Wea. Rev., 136, 2422–2442, doi:10.1175/2007MWR2229.1.
Dudhia, J., 1989: Numerical study of convection observed during the Winter Monsoon Experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 3077–3107, doi:10.1175/1520-0469(1989)046<3077:NSOCOD>2.0.CO;2.
Giorgi, F., and Marinucci M. R. , 1996: An investigation of the sensitivity of simulated precipitation to the model resolution and its implications for climate studies. Mon. Wea. Rev., 124, 148–166, doi:10.1175/1520-0493(1996)124<0148:AIOTSO>2.0.CO;2.
Gochis, D. J., Shuttleworth W. J. , and Yang Z.-L. , 2002: Sensitivity of the modeled North American monsoon regional climate to convective parameterization. Mon. Wea. Rev., 130, 1282–1297, doi:10.1175/1520-0493(2002)130<1282:SOTMNA>2.0.CO;2.
Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterizations. Mon. Wea. Rev., 121, 764–787, doi:10.1175/1520-0493(1993)121<0764:PEOAUB>2.0.CO;2.
Grell, G. A., and Dévényi D. , 2002: A generalized approach to parameterizing convection combining ensemble and data assimilation techniques. Geophys. Res. Lett., 29, 1693, doi:10.1029/2002GL015311.
Grell, G. A., and Freitas S. R. , 2013: A scale and aerosol aware stochastic convective parameterization for weather and air quality modeling. Atmos. Chem. Phys. Discuss., 13, 23 845–23 893, doi:10.5194/acpd-13-23845-2013.
Han, J., and Pan H.-L. , 2006: Sensitivity of hurricane intensity forecast to convective momentum transport parameterization. Mon. Wea. Rev., 134, 664–674, doi:10.1175/MWR3090.1.
Han, J., and Pan H.-L. , 2011: Revision of convection and vertical diffusion schemes in the NCEP Global Forecast System. Wea. Forecasting, 26, 520–533, doi:10.1175/WAF-D-10-05038.1.
Hong, S.-Y., and Pan H.-L. , 1998: Convective trigger function for a mass flux cumulus parameterization scheme. Mon. Wea. Rev., 126, 2599–2620, doi:10.1175/1520-0493(1998)126<2599:CTFFAM>2.0.CO;2.
Hong, S.-Y., and Lim J.-O. J. , 2006: The WRF single-moment 6-class microphysics scheme (WSM6). J. Korean Meteor. Soc., 42, 129–151.
Hong, S.-Y., and Dudhia J. , 2012: Next-generation numerical weather prediction: Bridging parameterization, explicit clouds, and large eddies. Bull. Amer. Meteor. Soc., 93 (Suppl.), doi:10.1175/2011BAMS3224.1.
Hong, S.-Y., Noh Y. , and Dudhia J. , 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, doi:10.1175/MWR3199.1.
Hong, Y., Kummerow C. D. , and Olson W. S. , 1999: Separation of convective and stratiform precipitation using microwave brightness temperature. J. Appl. Meteor., 38, 1195–1213, doi:10.1175/1520-0450(1999)038<1195:SOCASP>2.0.CO;2.
Huffman, G. J., and Coauthors, 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 38–55, doi:10.1175/JHM560.1.
Janjić, Z. I., 1994: The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945, doi:10.1175/1520-0493(1994)122<0927:TSMECM>2.0.CO;2.
Juang, H.-M. H., Hong S.-Y. , and Kanamitsu M. , 1997: The NCEP Regional Spectral Model: An update. Bull. Amer. Meteor. Soc., 78, 2125–2143, doi:10.1175/1520-0477(1997)078<2125:TNRSMA>2.0.CO;2.
Jung, S.-H., Im E.-S. , and Han S.-O. , 2012: The effect of topography and sea surface temperature on heavy snowfall in the Yeongdong region: A case study with high resolution WRF simulation. Asia-Pac. J. Atmos. Sci., 48, 259–273, doi:10.1007/s13143-012-0026-2.
Jung, W., and Lee T.-Y. , 2013: Formation and evolution of mesoscale convective systems that brought the heavy rainfall over Seoul on September 21, 2010. Asia-Pac. J. Atmos. Sci., 49, 635–647, doi:10.1007/s13143-013-0056-4.
Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170–181, doi:10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.
Kain, J. S., and Fritsch J. M. , 1990: A one-dimensional entraining/detraining plume model and its application in convective parameterization. J. Atmos. Sci., 47, 2784–2802, doi:10.1175/1520-0469(1990)047<2784:AODEPM>2.0.CO;2.
Kanamitsu, M., Ebisuzaki W. , Woollen J. , Yang S.-K. , Hnilo J. , Fiorino M. , and Potter G. , 2002: NCEP–DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 1631–1643, doi:10.1175/BAMS-83-11-1631.
Kang, H.-S., and Hong S.-Y. , 2008: Sensitivity of the simulated East Asia summer monsoon climatology to four convective parameterization schemes. J. Geophys. Res., 113, D15119, doi:10.1029/2007JD009692.
Kim, D.-Y., Kim J.-Y. , and Kim J.-J. , 2013: Mesoscale simulations of multi-decadal variability in the wind resource over Korea. Asia-Pac. J. Atmos. Sci., 49, 182–192.
Koo, M.-S., and Hong S.-Y. , 2010: Diurnal variations of simulated precipitation over East Asia in two regional climate models. J. Geophys. Res., 115, D05105, doi:10.1029/2009JD012574.
Kumar, P., Shukla M. V. , Thapliyal P. K. , Bisht J. H. , and Pal P. K. , 2012: Evaluation of upper tropospheric humidity from NCEP analysis and WRF model forecast with Kalpana observation during Indian summer monsoon 2010. Meteor. Appl., 19, 152–160, doi:10.1002/met.1332.
Lee, D.-K., Eom D.-Y. , Kim J.-W. , and Lee J.-B. , 2010: High-resolution summer rainfall prediction in the JHWC real-time WRF system. Asia-Pac. J. Atmos. Sci., 46, 341–353, doi:10.1007/s13143-010-1003-2.
Lee, J. G., Kim S.-D. , and Kim Y.-J. , 2011: A trajectory study on the heavy snowfall phenomenon in Yeongdong region of Korea. Asia-Pac. J. Atmos. Sci., 47, 45–62, doi:10.1007/s13143-011-1004-9.
Leung, L. R., Mearns L. O. , Giorgi F. , and Wilby R. , 2003: Workshop on regional climate research: Needs and opportunities. Bull. Amer. Meteor. Soc., 84, 89–95, doi:10.1175/BAMS-84-1-89.
Ma, Z., Fei J. , Huang X. , and Cheng X. , 2012: Sensitivity of tropical cyclone intensity and structure to vertical resolution in WRF. Asia-Pac. J. Atmos. Sci., 48, 67–81, doi:10.1007/s13143-012-0007-5.
Mlawer, E. J., Taubman S. J. , Brown P. D. , Iacono M. J. , and Clough S. A. , 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663–16 682, doi:10.1029/97JD00237.
Nasrollahi, N., AghaKouchak A. , Li J. , Gao X. , Hsu K. , Sorooshian S. , 2012: Assessing the impacts of different WRF precipitation physics in hurricane simulations. Wea. Forecasting, 27, 1003–1016, doi:10.1175/WAF-D-10-05000.1.
Pan, H.-L., and Wu W.-S. , 1995: Implementing a mass flux convective parameterization package for the NMC Medium-Range Forecast Model. NMC Office Note 409, 40 pp.
Park, H., and Hong S.-Y. , 2007: An evaluation of a mass-flux cumulus parameterization scheme in the KMA Global Forecast System. J. Meteor. Soc. Japan, 85, 151–168, doi:10.2151/jmsj.85.151.
Raju, P. V. S., Jayaraman P. , and Mohanty U. C. , 2011: Sensitivity of physical parameterizations on prediction of tropical cyclone Nargis over the Bay of Bengal using WRF model. Meteor. Atmos. Phys., 113, 125–137, doi:10.1007/s00703-011-0151-y.
Ratnam, J. V., and Kumar K. K. , 2005: Sensitivity of the simulated monsoons of 1987 and 1988 to convective parameterization schemes in MM5. J. Climate, 18, 2724–2743, doi:10.1175/JCLI3390.1.
Rauscher, S. A., Coppola E. , Piani C. , and Giorgi F. , 2010: Resolution effects on regional climate model simulations of seasonal precipitation over Europe. Climate Dyn., 35, 685–711, doi:10.1007/s00382-009-0607-7.
Sharma, A., and Huang H.-P. , 2012: Regional climate simulation for Arizona: Impact of resolution on precipitation. Adv. Meteor., 2012, 505726, doi:10.1155/2012/505726.
Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp. [Available online at http://www.mmm.ucar.edu/wrf/users/docs/arw_v3_bw.pdf.]
Srinivas, C. V., Hariprasad D. , Bhaskar Rao D. V. , Anjaneyulu Y. , Baskaran R. , and Venkatraman B. , 2013: Simulation of the Indian summer monsoon regional climate using advanced research WRF model. Int. J. Climatol., 33, 1195–1210, doi:10.1002/joc.3505.
Tiedtke, M., 1989: A comprehensive mass flux scheme for cumulus parameterization in large-scale models. Mon. Wea. Rev., 117, 1779–1800, doi:10.1175/1520-0493(1989)117<1779:ACMFSF>2.0.CO;2.
Varghese, S., Langmann B. , Ceburnis D. , and O’Dowd C. D. , 2011: Effect of horizontal resolution on meteorology and air-quality prediction with a regional scale model. Atmos. Res., 101, 574–594, doi:10.1016/j.atmosres.2011.02.007.
Wang, W., and Seaman N. L. , 1997: A comparison study of convective parameterization schemes in a mesoscale model. Mon. Wea. Rev., 125, 252–278, doi:10.1175/1520-0493(1997)125<0252:ACSOCP>2.0.CO;2.
Wu, L., Su H. , and Jiang J. H. , 2013: Regional simulation of aerosol impacts on precipitation during the East Asian summer monsoon. J. Geophys. Res. Atmos., 118, 6454–6467.
Yang, B., Zhang Y. , and Qian Y. , 2012: Simulation of urban climate with high-resolution WRF model: A case study in Nanjing, China. Asia-Pac. J. Atmos. Sci., 48, 227–241, doi:10.1007/s13143-012-0023-5.
Yu, E., Wang H. , Gao Y. , and Sun J. , 2011: Impacts of cumulus convective parameterization schemes on summer monsoon precipitation simulation over China. Acta Meteor. Sin., 25, 581–592, doi:10.1007/s13351-011-0504-y.
Yu, X., and Lee T.-Y. , 2011: Role of convective parameterization in simulations of heavy precipitation systems at grey-zone resolutions—Case studies. Asia-Pac. J. Atmos. Sci., 49, 99–112, doi:10.1007/s13143-011-0001-3.
Yuan, X., and Liang X.-Z. , 2011: Improving cold season precipitation prediction by the nested CWRF-CFS system. Geophys. Res. Lett., 38, L02706, doi:10.1029/2010GL046104.
Zanis, P., Douvis C. , Kapsomenakis I. , Kioutsioukis I. , Melas D. , and Pal J. S. , 2009: A sensitivity study of the regional climate model (RegCM3) to the convective scheme with emphasis in central eastern and southeastern Europe. Theor. Appl. Climatol., 97, 327–337, doi:10.1007/s00704-008-0075-8.
Zhang, C., Wang Y. , and Hamilton K. , 2011: Improved representation of boundary layer clouds over the southeast Pacific in WRF-ARW using a modified Tiedtke cumulus parameterization scheme. Mon. Wea. Rev., 139, 3489–3513, doi:10.1175/MWR-D-10-05091.1.
Zhang, G. J., and McFarlane N. A. , 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model. Atmos.–Ocean, 33, 407–446, doi:10.1080/07055900.1995.9649539.
