1 Introduction
Typhoons are strong cyclonic vortices formed over the tropical ocean with a warm core structure. Their occurrence, development, movement, and dissipation result from the interaction of various scales of atmospheric circulation systems. Numerical simulations combine the motion equations and physical parameterization equations of the atmosphere and ocean to simulate these interactions, and the atmosphere can be discretized into grids of varying sizes to analyze processes across different scales. Numerical models employ parameterization schemes to describe various physical processes when depicting atmospheric motion near the grid or sub-grid scales. Consequently, variations in grid resolutions and cloud microphysics parameterization schemes within typhoon precipitation simulations can lead to diverse simulation outcomes. By improving the precision of precipitation forecasts based on a reasonable combination of different resolutions and schemes, we can propose proactive response measures to mitigate secondary disasters caused by typhoon precipitation and minimize property and life losses.
Recent advancements in computer capabilities have led to a substantial increase in the grid resolution employed within numerical models (
Hendricks et al., 2016;
Yuk and Joh, 2019). Increasing the model’s horizontal resolution has the potential to improve forecasting performance while minimally impacting the simulation of large-scale circulation patterns (
Bacmeister et al., 2018;
Gettelman et al., 2018). Therefore, studying the impact of grid resolution on typhoon simulation is crucial for improving typhoon prediction. With improvements in high-performance computing power, high-resolution numerical models are now able to explicitly simulate complex cloud and precipitation systems. In numerical weather prediction (NWP), cloud microphysical parameterization schemes significantly impact the simulation of spatial distribution, intensity, and duration of precipitation (
Mohan et al., 2018;
Otieno et al., 2019;
Li et al., 2020;
Xu et al., 2023). However, there remains some uncertainty in the microphysical parameterization schemes of NWP models regarding the processes of water droplets and ice crystal formation, growth, and deposition. Several studies have conducted numerical experiments and analyzed observations to investigate differences in cloud microphysical schemes. As microwave technology possesses the capability to detect the microphysical characteristics of precipitation systems, the Tropical Rainfall Measuring Mission (TRMM) satellite employs the Passive Microwave Imager (TMI) and the Precipitation Radar (PR) to perform quantitative assessments of tropical precipitation properties. These instruments provide valuable information on precipitation amounts and vertical profiles of microphysical parameters (
Kim et al., 2013). Multiple investigations have emphasized the necessity of carefully selecting physics parameterization schemes to obtain reliable precipitation forecasts across different regions and climatic conditions (
Gómez-Navarro et al., 2015;
Khain et al., 2016;
Tian et al., 2017).
High-precision numerical simulation studies of typhoons primarily utilize high-resolution finite regional models, such as Weather Research and Forecasting (WRF) model (
Shen et al., 2016,
2024;
Xu et al., 2024). These models use initial fields obtained from either dynamical downscaling of coarser-resolution global model forecasts or from reanalysis data (
Cha and Wang, 2013). However, this approach can introduce artificial errors due to boundary resolution changes when transitioning from coarse to high-resolution grids (
Park et al., 2014;
Hashimoto et al., 2016), potentially leading to inadequate representation of large-scale forcing within the regional model grids (
Cha et al., 2011). While global models offer the advantage of reducing the lateral boundary issue faced by regional models and provide a broader view of large-scale circulation, the high computational demands of high-resolution global numerical models limit their capacity for precise simulation of small- and medium-scale physical processes of typhoons (
Laprise et al., 2012). The Model for Prediction Across Scales for Atmosphere (MPAS-A, hereafter MPAS) is a global model with variable-resolution, which is well-suited for regional modeling at a scale that captures convective processes (
Skamarock et al., 2012). MPAS model, with its unstructured grids, facilitates a smooth transition in grid spacing from a coarse quasi-uniform resolution to a high-resolution refinement in specific regions (
Ringler et al., 2008). This flexibility enables the selection of high-resolution grid regions based on the area of interest, while utilizing a coarse resolution grid for the remaining regions, thus improving computational efficiency (
Lui et al., 2020). Simulations with a standard variable-resolution grid have demonstrated performance comparable to those using a quasi-uniform grid with the same fine resolution (
Sakaguchi et al., 2015). This approach improves the resolution of fine-scale features in the regions of interest and mitigates issues related to lateral boundaries when forecasting typhoon systems at the convection-permitting scale.
Davis et al. (2016) evaluated the effectiveness of the real-time forecasts of the MPAS model and operational forecasts from the NCEP Global Forecast System (GFS) in the eastern Pacific using a 15 km uniform grid and 60−15 km variable resolution grid, confirming the value of variable resolution grids for typhoon forecasting. In recent years, studies using the MPAS model for typhoon simulation in the north-west Pacific have focused on how Taiwan’s mountainous topography, grid resolution, cumulus convection, and cloud microphysical parameterization schemes affect typhoon tracks and intensities (
Huang et al., 2017,
2019,
2022a,
2022b,
2022c;
Gao et al., 2019).
Zhao et al. (2019) employed the MPAS model to simulate extreme rainfall events during the East Asian Meiyu season in China, analyzing the impact of model grid resolution and parameterization schemes for convective clouds and cloud microphysics on heavy precipitation. The results revealed that the MPAS model’s sensitivity to extreme precipitation at convection-permitting scale depends strongly on the chosen cloud microphysics scheme, influencing both the distribution and intensity of heavy rainfall events.
Based on the available information, there have been limited research involving non-hydrostatic global variable-resolution atmospheric models for predicting the typhoon-related precipitation. Conducting research is essential to understand the impact of different cloud microphysics parameterization schemes and grid resolutions within the MPAS model for simulating typhoon precipitation in China. This research will provide valuable insights for the application of the MPAS model in typhoon forecasting in the north-west Pacific region. The distinct goals of this study are outlined as follows: 1) to examine the performance of MPAS in forecasting the spatial distribution and intensity of precipitation in a typhoon case with different variable-resolution grids. 2) to investigate how grid resolution and cloud microphysics schemes affect the variability of typhoon precipitation. The remaining sections of this study are organized as follows: Section 2 offers a detailed description of the model, experimental design, and the observation and reanalysis data used. Section 3 investigates the mechanisms responsible for the variations in typhoon precipitation simulations through the analysis of the thermal dynamic field and hydrometeor mixing ratios. The study concludes with a summary and discussion in Section 4.
2 Methodology and data
2.1 Models and experiments
2.1.1 MPAS model
MPAS model uses unstructured spherical centroidal Voronoi tessellation (SCVT) generation algorithms to create global quasi-uniform resolution meshes and variable-resolution meshes using a single scalar density function. This approach avoids the singularity and over-density issues commonly encountered in polar regions with conventional latitude and longitude grids, as illustrated in Fig. 1(a) (
Kramer et al., 2020). The variable-resolution meshes enable a gradual progression from coarser to finer resolution as shown in Fig. 1(b), providing smoother transitions compared to the abrupt transitions typically observed in the nesting techniques utilized by conventional regional models. MPAS provides a selection of coordinate systems, including a terrain-following coordinate system and a hybrid terrain-following coordinate system that incorporates geometric height. The hybrid coordinate exhibits the lowest error near steep slopes (
Klemp, 2011). MPAS utilizes a third-order Runge-Kutta method and explicit time-splitting technique for time-integration scheme (
Wicker and Skamarock, 2002), similar to the approach employed by the WRF model. The dynamical core and physical parameterization scheme of the MPAS model are very similar to those of WRF, with the physical parameterization scheme primarily derived from the existing parameterization scheme in the WRF model. Currently, MPAS is equipped with two physics scheme suites: a mesoscale suite and a convective suite. For the convective parameterizations, it offers the Kain-Fritsch scheme (
Kain, 2004), the New Tiedtke scheme (
Zhang et al., 2011), and a revised version of scale-aware Grell-Freitas scheme (
Grell and Freitas., 2014). The model includes two cloud microphysics schemes, namely the WSM6 scheme (
Hong and Lim, 2006) and the Thompson scheme (
Thompson et al., 2004). The model incorporates two planetary boundary layer (PBL) parameterization schemes: the MYNN scheme (
Nakanishi and Niino, 2004) and the YSU scheme (
Hong et al., 2006). In addition to these, the model offers other physical parameterizations, such as the Noah land surface scheme (
Chen and Dudhia, 2001), and the RRTMG scheme to handle longwave and shortwave radiative processes (
Iacono et al., 2008).
2.1.2 Numerical experiments
Typhoon Fitow originated in the North-west Pacific Ocean near the northern coast of Palau and was officially classified as the 22nd tropical cyclone on September 29, 2013. It gradually moved north-westward and intensified, reaching tropical storm strength. At 2100 UTC on October 2, it further strengthened and became a typhoon. By 0900 UTC on October 4, the typhoon experienced rapid intensification and reached the status of a severe typhoon. It then changed its direction to the west-north-west and arrived at Fujian Province, making landfall precisely at 1715 UTC on October 6. The central sea level pressure (SLP) was recorded at 955 hPa, with peak wind speeds reaching 42 m/s. The typhoon rapidly lost intensity upon moving inland and dissipated at 0100 UTC on October 7. Fitow was the strongest typhoon to hit China during the October period from 1949 to 2013. Its impact included heavy rains and strong winds, causing significant damage such as urban waterlogging, flooding disasters in Zhejiang Province (Fig. 1, marked as ZJ in Fig. 2) from October 6 to 8, 2013 (
Yu et al., 2014;
Bao et al., 2015;
Xu and Du, 2015;
Lou and Li, 2016;
Xu and Li, 2017). Figure 2 displays the best track of Fitow (left) and Danas (right) from tcdata.typhoon.org.cn (
Ying et al., 2014;
Lu et al., 2021).
In this study, six experiments were conducted to simulate typhoon precipitation using three grid configurations and two cloud microphysics schemes. The grid settings used were as follows. 1) A 24 km quasi-uniform grid, which is similar to the coarse mesh area in Fig. 1(b) and contains 1024002 horizontal grids, is not shown in this article. 2) A variable-resolution mesh ranging from 60 km to 10 km, with a maximum resolution of 10 km centered at 120°E, 30°N and containing 999426 horizontal grids (as shown in Fig. 3(a)). 3) A variable-resolution mesh ranging from 60 km to 3 km, with the higher resolution of 3 km centered at 120°E, 30°N and containing 835586 horizontal grids (as shown in Fig. 3(b)). For the cloud microphysics schemes, the WSM6 and Thompson schemes were selected, and their characteristics are presented in Table 1. Both schemes include six hydrometeor species: water vapor, cloud water, rain, cloud ice, snow, and graupel. The WSM6 scheme determines only the hydrometeor mixing ratio, while the Thompson scheme incorporates the raindrop number concentration as an additional predictor variable. Additional physical parameterizations employed in the study include the KF cumulus scheme, the Noah land surface model, the YSU planetary boundary layer scheme, the Noah surface layer, and the RRTMG scheme for both long- and short-wave radiation (
Iacono et al., 2008). For the initialization of static fields, land use data from the Moderate Resolution Imaging Spectroradiometer (MODIS) and topographical data from the United States Geological Survey (USGS) were utilized. All experiments commenced at 0000 UTC on 6 October, 2013, using the 0.5° × 0.5° Global Forecast System (GFS) data produced by the National Centers for Environmental Prediction (NCEP), and were integrated over a 36-h period.
2.2 Observation data
The precipitation validation data are released by the China Meteorological Data Center, with a spatial resolution of 0.1° × 0.1°, covering the range from 70°E to140°E and from 15°N to 60°N. Hourly precipitation fusion products are generated using probability density matching and optimal interpolation data fusion algorithms. These products are derived from observed data from a network of 30000 Automatic Weather Stations (AWS) across China and from the CPC MORPHing technique (CMORPH) precipitation data, which are provided by the National Oceanic and Atmospheric Administration (NOAA) in the United States (
Xu et al., 2016). The CMORPH product has a spatial resolution of 8 km and a temporal resolution of 0.5 h. This precipitation verification data set has undergone evaluation in mainland China and has been utilized for simulating and evaluating multiple typhoon events (
Yu et al., 2009).
The TMI is a precipitation measurement instrument carried on the TRMM satellite. It operates within a scanning range from 38°N to 38°S, completing approximately 16 orbits per day. The scanning bandwidth is 750 km, and it achieves a horizontal resolution of 5.1 km × 5.1 km. In the vertical direction, TMI has a total of 28 layers, with a vertical resolution of 0.5 km below 10 km, increasing to 1 km above 10 km, and extending up to 18 km. TMI effectively captures low-frequency emission signals and high-frequency scattering signals, providing valuable information on the integrated information of water vapor condensate across each layer of the atmospheric column. The TRMM/TMI 2A12 product offers data on surface instantaneous precipitation intensity, total perceptible water, three-dimensional latent heat structure, and vertical profiles of hydrometeors at a pixel resolution. The hydrometeor profiles within the 2A12 product encompass cloud water, rainwater, cloud ice, snow, and hail. These profiles are derived from TMI brightness temperature data and utilize the microphysical processes of cloud modes in the GODDARD Profiling Algorithm (GPROF2008) to classify hydrometeors in precipitation cloud systems (
Liu and Fu, 2007;
Liu and Moncrieff, 2007). By analyzing the spatial distribution structure of water vapor condensates, valuable insights into the release or absorption of latent heat in the atmosphere can be obtained (
Ma and Duan, 2005;
Yao et al., 2014).
In our study, the CMORPH 0.5-h precipitation estimates were aggregated into 6-hourly and 24-h accumulated data. To evaluate the model predictions against gridded data, the unstructured outputs of MPAS-A were interpolated onto the same regular grids using barycentric interpolation. Additionally, the ERA5 reanalysis data set from the European Centre for Medium-Range Weather Forecasts (ECMWF) was utilized to validate atmospheric circulation patterns. Moreover, the model’s performance in simulating hydrometeor mixing ratios was assessed using TMI retrievals.
3 Results
3.1 Spatial distribution of the typhoon rainfall
Figure 4(a) displays the spatial distribution of observed 24-h accumulated precipitation from 1200 UTC on October 6, 2013 to 1200 UTC on October 7, 2013 in Zhejiang Province, China. A robust north-west-south-east precipitation band is evident in the northern part of Zhejiang Province, accompanied by a strong precipitation area in the southern coastal region and a precipitation center in the middle of the eastern coastal area. Figures 4(b) and 4(c) illustrate the simulated spatial distribution of 24-h accumulated precipitation during the same period using the WSM6 and Thompson schemes, respectively, on a 24 km quasi-uniform grid. The simulation results from both schemes are similar but differ significantly from the observations. The observed strong precipitation band in the north-west-south-east direction is generally weaker in the simulations, and its coverage area extends to the south compared to the actual observations. Additionally, the simulated results fail to capture the presence of a strong precipitation center in the southern coastal area of Zhejiang Province.
Figure 5 shows the spatial distribution of 24-h accumulated precipitation during the corresponding period, simulated using the WSM6 and Thompson schemes on a 60−10 km variable mesh. Both schemes exhibit similar spatial distribution and intensity of simulated precipitation. They are capable of reproducing the heavy precipitation center in the southern coastal area of Zhejiang Province, although the simulated heavy precipitation area is larger than observed. However, the precipitation intensity in northern Zhejiang Province is weaker, and no strong precipitation belt is observed from north-west to south-east in the simulated precipitation area.
Figure 6 illustrates both the observed and simulated precipitation distribution at 6-h intervals, as well as the total 24-h precipitation distribution for the WSM6 and Thompson schemes on a 60−3 km variable mesh. The figure is organized as follows: The first row (a1−d1) shows the observed cumulative precipitation every 6 h starting from 1200 UTC on October 6, 2013. Panel (e1) represents the spatial distribution of the observed 24-h accumulated precipitation. The second row (a2−d2) shows the 6-h accumulated precipitation simulated by the WSM6 scheme on the 60−3 km variable grid, and (e2) depicts the spatial distribution of the 24-h accumulated precipitation simulated by the WSM6 scheme. The third row follows the same format as the second row but uses the Thompson scheme for simulation. Comparing the simulated cumulative precipitation for each 6-h period reveals the following results. The first period (a1−a3) reveals that both simulations exhibit similar precipitation intensity along the coastal area but with an overestimation in the central part and fails to simulate the heavy precipitation in the southern coastal area. The Thompson scheme simulates weaker precipitation intensity compared to the WSM6 scheme. During the second period (b1−b3), the WSM6 scheme can simulate the north-west-south-east strong precipitation band in the northern Zhejiang Province, but the area is narrower than observed, and the strong precipitation center in the southern coastal area is shifted further north. Conversely, the Thompson scheme fails to simulate this band, instead shifting it toward the region’s center. In the third period (c1−c3), both schemes simulate the north-west-south-east strong precipitation band. However, the simulated strong precipitation area is narrower and shifted further south compared to observations, and neither scheme successfully simulates the strong precipitation center in the southern coastal area. During the fourth period (d1−d3), neither scheme manages to simulate the north-west-south-east strong precipitation band or the strong precipitation area in the southern coastal area. Overall, the Thompson scheme produces lower cumulative precipitation than the WSM6 scheme for each 6-h period. The WSM6 scheme more closely simulates the intensity and spatial distribution of the 24-h accumulated precipitation, making it closer to observations compared to the Thompson scheme. The precipitation process is mainly concentrated in the first 12 h, from 1800 UTC on October 6 to 0600 UTC on October 7. The subsequent analyses will focus on this time period.
The precipitation distribution simulated using two parameterization schemes at three different grid resolutions reveals that the WSM6 scheme outperforms the Thompson scheme at the same resolution. Additionally, as the grid resolution increases, the precipitation distribution and intensity become closer to the observations.
3.2 Thermal and dynamic field analysis
Convective available potential energy (CAPE) is a thermodynamic variable that indicates the strength of convective and precipitation systems, reflecting the maximum strength of updrafts that may develop. Storm-relative helicity (SRH) describes the degree of rotation (horizontal vorticity) and strength of the airflow along the direction of motion of the environmental wind field, representing characteristics of the atmospheric motion field. A combination of CAPE and SRH values is often used to assess the risk of severe convective weather. Strong convective weather can occur either in an environment with low SRH (SRH < 200 m
2/s
2) and high CAPE (CAPE > 2000 J/kg), or in the opposite environment (CAPE < 1500 J/kg, SRH > 200 m
2/s
2) (
Yang et al., 2006). The analysis period in Fig. 7 corresponds to Fig. 6, panels (a−d) are the CAPE values simulated at 0600 UTC on October 6, while panels (a1−d1) are those simulated at 0000 UTC on October 7. Figure 7 displays the simulated CAPE values for the WSM6 and Thompson schemes on a 60−3 km mesh, represented in panels (a and a1) and (b and b1), respectively. Panels (c and c1) and (d and d1) show the CAPE values for the WSM6 and Thompson schemes on a 60−10 km mesh, respectively. The simulated CAPE values at 0000 UTC on October 7 are larger than those at 1800 UTC on October 6, the range of CAPE values over the ocean (1000−2500 J/kg) is significantly higher than the range of values over land (less than 800 J/kg). The CAPE values of the 60−3 km mesh on the ocean are larger than those for the 60−10 km mesh, while the situation is reversed on the land area.
Figures 8 and 9 depict the maximum updraft helicity and the vertically integrated water vapor, respectively, using the same simulation time and experimental configuration as Fig. 7. Figure 8 demonstrates that the fine resolutions applied by the two cloud microphysics schemes result in higher simulated SRH values compared to the coarse resolutions. Additionally, Fig. 9 illustrates that areas of high values in vertically integrated water vapor are more widely distributed over the ocean with fine grid resolutions than with coarse resolutions. Based on the conclusions of
Yang et al. (2006), regions with low CAPE values combined with relatively high SRH values have an increased likelihood of severe convective weather. It can be observed that the areas with low CAPE values along the east coast in Fig. 7 overlap with the regions of high helicity values in Figs. 8(a and a1) and 8(b and b1) of, corresponding to the areas of heavy rain simulated on the 60−3 km grid in Figs. 6(a2 and a3) and 6(b2 and b3). Furthermore, there is a good correspondence with the areas of high precipitable water values in columns (a and a1) and (b and b1) of Fig. 9.
Figure 10 shows the nonadiabatic heating resulting from microphysical processes when the WSM6 and Thompson schemes are configured with 60−3 km and 60−10 km grid resolutions, respectively, at 1800 UTC on October 6, 2013. The regions with strong nonadiabatic heating rates are significant, as they indicate areas of intense upward motion and convective activity within the atmosphere. Figures 10(a) and 10(c) show the spatial distribution of nonadiabatic heating, averaged over the 1−8 km height range, for the WSM6 and Thompson schemes, respectively, on a 60−3 km grid resolution. It is evident that the intensity of nonadiabatic heating and cooling is higher in the WSM6 scheme compared to the Thompson scheme. Conversely, Figs. 10(b) and 10(d) present the spatial distribution of nonadiabatic heating for the corresponding microphysical schemes on a 60−10 km grid resolution. Within the same microphysical scheme, the cooling center of coarse grid is both smaller and situated further south compared to the fine grid. Moreover, the heating area of the coarse grid is less extensive than that of the fine grid. Consequently, the fine grid exhibits stronger precipitation intensity compared to the coarse grid, and the region of strong nonadiabatic heating rate corresponds to the area of heavy precipitation.
Figure 11 shows the vertical cross-sections of divergence, vorticity, and vertical velocity along the center of the main rain band, which is characterized by strong precipitation at 29.5°N. The parameters are simulated using the Thompson and WSM6 schemes with 60−3 km grid resolutions, alongside reanalysis data at 1800 UTC on October 6. Figure 11(a) shows the convergence zone from the ERA5 reanalysis data distributed between the surface and 750 hPa. Figure 11(b) shows the positive vorticity zone, primarily located between 650 hPa and 800 hPa as well as near the surface, with an intensity level of up to 10−3 s−1. Figure 11(c) presents the vertical velocity, indicating dominant sinking motion below the precipitation area, with rising motion confined to the surface to 800 hPa between 119°E and 120°E. The divergence and vorticity fields simulated by the Thompson and WSM6 schemes (Figs. 11(a1) and 11(b1) and 11(a2) and 11(b2)) are both approximately one order of magnitude larger than the ERA5 reanalysis data, with the Thompson scheme showing larger divergence, vorticity, and vertical velocity than the WSM6 scheme. In Figs. 11(a2) and 11(b2), the divergence and vorticity simulated by the WSM6 scheme are low in the area east of 120°E, corresponding to the area of strong surface precipitation. The simulated intensity in the low-value areas is relatively close to those of the ERA5 reanalysis data, despite the overall simulation intensity being one order of magnitude larger. The intensity and area of the upward motion simulated by the WSM6 scheme (Fig. 11(c2)) are closer to the ERA5 reanalysis data. It is evident that the intensity of divergence, vorticity, and vertical velocity simulated by the WSM6 scheme more closely matches the ERA5 reanalysis data, and the difference in dynamic conditions between the two microphysical schemes contributes to the variations in the precipitation simulation.
The thermal and dynamic field analysis highlights significant differences between the WSM6 and Thompson microphysics schemes at various grid resolutions. The WSM6 scheme consistently produces higher CAPE values at finer resolutions, aligning more closely with observations. Higher SRH values at finer resolutions suggest a more accurate representation of atmospheric rotation and airflow. Enhanced vertically integrated water vapor and nonadiabatic heating rates in the WSM6 scheme further support intense convective activity. Vertical cross-sections indicate that the WSM6 scheme’s simulations are more consistent with ERA5 reanalysis data, demonstrating a superior ability to simulate dynamic conditions and precipitation distribution.
3.3 Distribution characteristics of hydrometeor mixing ratio
At 0000 UTC on October 6, TRMM/TMI captured the complete typhoon cloud system. The TMI 2A12 data from this time were selected for validation to compare the distribution characteristics of hydrometeor mixing ratios simulated by two physical parameterization schemes at different grid resolutions. In Fig. 12, panels (a1) and (b1) present the spatial distribution of the vertically integrated content of liquid phase particles and ice phase particles retrieved by TMI observations, respectively. The black boxes in the figure indicate the areas where data are missing, with one located over Taiwan and the other over an island to the east of Taiwan. According to Figs. 12(a1) and 12(b1), the vertically integrated content of liquid phase particles in the typhoon center and eyewall area is slightly higher than that of ice phase particles. In the outer cloud region, the ice phase particle content is high, while the liquid phase particle content is minimal. In Fig. 12, panels (a2) and (a3) display the spatial distribution of the vertically integrated content of liquid phase particles simulated by WSM6 scheme and Thompson scheme, respectively, using a 60−10 km grid in MPAS. Similarly, Figs. 12(b2) and 12(b3) show the spatial distribution of the vertically integrated content of ice phase particles under the same experimental configuration. The ice and liquid phase particle content and spatial distribution simulated by the two schemes show significant differences compared to the TMI observations: the simulated contents in the typhoon area are lower than the observation retrievals. However, both schemes overestimate the concentration of ice phase particles in the outer regions of the typhoon north of 28°N, as indicated by the black frame in Fig. 12.
The simulation settings in Figs. 12(a4), 12(a5), 12(b4), 12(b5) remain the same as in Figs. 12(a2), 12(a3), 12(b2), 12(b3), with the exception of the grid resolution, which has been refined to 60−3 km. Specifically, panels (a4) and (a5) correspond to (a2) and (a3), while panels (b4) and (b5) correspond to (b2) and (b3). With the grid refinement, the simulated hydrometeor content distribution more closely matches the TMI retrievals. However, the simulated content of liquid and ice phase particles from both schemes still exhibits falsely elevated values in the typhoon cloud belt on the north side of 28°N. The WSM6 scheme simulates liquid water content in the typhoon cloud area similarly to the Thompson scheme, and both schemes show similar simulations for the typhoon center and eye wall (as shown in Figs. 12(a4) and 12(a5)). When evaluating the distribution of ice phase particles, the WSM6 scheme indicates lower content in the typhoon center but higher content in the eye wall and outer cloud region compared to the Thompson scheme (as shown in Figs. 12(b4) and 12(b5)). Additionally, the vertically integrated content of ice and liquid phase particles simulated with a 60−3 km grid is larger compared to that simulated with a 60−10 km grid. The ice phase particles distribution pattern simulated by the WSM6 scheme using a 60−3 km grid most closely matches the TMI retrievals. In the region with elevated ice particle content north side of 28°N, the simulation of the WSM6 scheme using a 60−3 km grid is lower than that using a 60−10 km grid, effectively reducing the extent of the overestimated value area (as shown in Figs. 12(b2) and 12(b4)). Furthermore, in the northeast of the typhoon outer cloud band, the ice phase particles content high value area simulated by the WSM6 scheme with a 60−3 km mesh, which closely matches the TMI retrievals, is more extensive than that using a 60−10 km grid. In conclusion, the ice phase particle content simulated by the Thompson scheme is lower than that of the WSM6 scheme. Moreover, the impact of refining the grid resolution on typhoon precipitation is more significant than the choice of microphysics scheme.
Analyzing cloud microphysical processes, especially the distribution of hydrometeors, offers valuable insights into the mechanisms of typhoon rainfall. In this study, clouds within a 150-km radius of the typhoon center are classified into convective and stratiform types. Vertical profiles of hydrometeors are calculated for both cloud types, and the cloud microphysical processes simulated by the typhoon Fitow are analyzed. For the classification of cloud types observed by TRMM/TMI, the precipitation echo intensity and precipitation rate profiles obtained from the PR rain radar 2A25 product can be utilized. The PR radar provides cross-track scans of 49 pixels per orbit, with a horizontal resolution of 4.5 km and a vertical resolution of 250 m. Pixels with radar reflectivity greater than 39 dBZ are classified as convective, while those with reflectivity between 15 dBZ and 39 dBZ are identified as stratiform clouds (
Xu et al., 2017). Similarly, the typhoon cloud types simulated by MPAS are classified into convective and stratiform clouds based on the simulated radar reflectivity echo intensity, employing the same thresholding approach.
Figure 13 illustrates the horizontally averaged vertical profiles of five types of hydrometeor mixing ratios for convective and stratiform clouds at 0000 UTC on October 6, 2013, over the typhoon center area. Panels (a1) and (b1), retrieved by TRMM/TMI, display convective and stratiform clouds, respectively, while other panels are simulated by the MPAS model. Rainwater and cloud water mixing ratios are mainly concentrated below 6 km, while snow mixing ratio is the lowest. The graupel mixing ratio reaches its peak close to 7 km and is distributed between 5 and 9 km in both convective and stratiform clouds. The ice mixing ratio is distributed between 5 and 15 km, spanning the widest altitude range. In the vertical direction, ice particles are distributed at the highest altitudes, succeeded by graupel particles. The simulated hydrometeor mixing ratio profiles using a 60−10 km grid are shown for the WSM6 in Fig. 13(a2) of convective clouds and in Fig. 13(b2) of stratiform clouds, and for Thompson in Fig. 13(a3) of convective clouds and in Fig. 13(b3) of stratiform clouds. Panels (a4) and (b4) correspond to (a2) and (b2), respectively, and (a5) and (b5) correspond to (a3) and (b3), respectively, except for the grid resolution, which has been refined to 60−3 km.
The simulated snow mixing ratio profiles are significantly greater than those retrieved by TRMM/TMI, particularly for the Thompson scheme, whereas the cloud ice mixing ratio is notably lower, with the exception of the WSM6 scheme’s simulation in stratiform clouds (Figs. 13(b2) and 13(b4)), which closely matches the retrieved values. For both the WSM6 and Thompson schemes, increasing the resolution from 60−10 km to 60−3 km enhances the concentrations of the five hydrometeor mixing ratios. However, in the WSM6 scheme, the ice mixing ratio in stratiform clouds decreases (as shown in Figs. 13(b2) and 13(b4)). The rainwater mixing ratio in both stratiform and convective clouds simulated by the WSM6 scheme with the 60−3 km grid closely matches the TMI retrievals, while the simulated rainwater and cloud water mixing ratio are both a bit larger than the TMI retrieval values (as shown in Figs. 13(a4) and 13(b4)). The simulated graupel mixing ratio with the Thompson scheme is lower than that simulated by the WSM6 scheme. The graupel mixing ratio increases with the resolution refining to 60−3 km grid (shown in Figs. 13(a5) and 13(b5)), and the simulated graupel particle content and the altitude of its peak concentration align closely with TMI retrievals.
The preceding analysis suggests that, with equivalent grid resolutions, the simulations of rainwater mixing ratios are largely consistent across different schemes. The most significant disparities emerge in the depiction of solid hydrometeors, such as cloud ice, snow, and graupel. The enhancement of grid resolution brings the ice mixing ratio in stratiform clouds and rain water mixing ratio in convective clouds simulated by the WSM6 scheme closer to retrievals. These findings indicate that the higher grid resolutions facilitate the production of rain water mixing ratios in convective clouds under a uniform microphysical parameterization, and along with the ice mixing ratio in the stratiform clouds, have a pronounced impact on precipitation.
4 Conclusions and summary
This study employs three different resolutions and two cloud microphysics parameterization schemes within the MPAS model to simulate the 24-h precipitation process associated with Typhoon Fitow. By analyzing the thermo and dynamic fields, the research explores how model resolution and physical parameterization affect the spatial distribution of precipitation. Additionally, it investigates the vertical distribution of hydrometeors and their influence on precipitation in the simulation results of various cloud physics parameterization schemes. The main conclusions are as follows.
1) Simulation of typhoon precipitation distribution reveals significant differences exist in the amount and spatial distribution of typhoon precipitation, even when employing the same physical parameterization scheme but varying mesh resolutions. The 60−3 km grid, which is the highest resolution used, aligns most closely with observed precipitation intensity distributions. At the same resolution, the WSM6 and Thompson schemes produce similar typhoon precipitation intensities and spatial distributions, although the Thompson scheme exhibits slightly weaker rainfall. The most accurate simulation of this typhoon’s precipitation was achieved with the 60−3 km resolution grid in conjunction with the WSM6 scheme.
2) Examination of the thermal and dynamic fields suggests that the combination of helicity, CAPE, and total column precipitable water on the 60−3 km grid can more precisely pinpoint areas of intense convection and heavy precipitation compared to coarser grids. By comparing the simulated hydrometeor profiles from the MPAS model, using both the WSM6 and Thompson schemes on 60−10 km and 60−3 km grids, with the TRMM/TMI 2A12 satellite products, it is clear that the difference in hydrometeor content within convective clouds is more pronounced between different grid resolutions than in stratiform clouds, even when the same physical parameterization scheme is applied. Moreover, at the same grid resolution, the disparity in ice phase particle content between the two schemes far exceeds the difference in liquid phase particle content, with the ice particle content having a more significant impact on precipitation intensity.
In summary, this study conducted six experiments using the MPAS model to investigate the influence mechanism of cloud physical parameterization schemes and grid resolution selection on the precipitation simulation effect of Typhoon Fitow. The findings indicate that the impact of high-resolution grids on precipitation simulation effects is more significant than the choice of microphysical parameterization schemes. Therefore, the development of physical parameterization schemes compatible with high-resolution numerical models is of great importance for enhancing the performance of precipitation forecasts. The MPAS model, with its high-resolution variable-resolution grid, can be effectively used for typhoon precipitation simulation research. While the research results indicate that variable-resolution models with high-resolution grids can effectively simulate typhoon rainfall, it is essential to acknowledge certain limitations that should be addressed in future studies. First, the performance of physical parameterization schemes is case-dependent. Conclusions based on single-case experiments may lack generality, necessitating further studies across multiple cases. Furthermore, the impact of horizontal resolution on precipitation forecasting is complex. It involves various factors such as initial ensemble diffusion, different model physical parameters, and the amplitude and phase errors of model predictions in strong interactions with the environment, requiring detailed evaluation on the role of horizontal resolution in rainfall prediction. These concerns will be explored in forthcoming studies.
PDF
(12098KB)
