Sensitivity of Precipitation and Structure of Typhoon Hato to Bulk and Explicit Spectral Bin Microphysics Schemes

0is study simulated the evolution of Typhoon Hato (2017) with the Weather Research and Forecasting model using three bulk schemes and one bin scheme. It was found that the track of the typhoon was insensitive to the microphysics scheme, whereas the degree of correspondence between the simulated precipitation and the cloud structure of the typhoon was closest to the observations when using the bin scheme. 0e different microphysical structure of the bin and three bulk schemes was reflected mainly in the cloud water and snow content. 0e three bulk schemes were found to produce more cloud water because the application of saturation adjustment condensed all the water vapor at the end of each time step. 0e production of more snow by the bin scheme could be attributed to several causes: (1) the calculations of cloud condensation nucleus size distributions and supersaturation at every grid point that cause small droplets to form at high levels, (2) different fall velocities of different sizes of particles that mean small particles remain at a significant height, (3) sufficient water vapor at high levels, and (4) smaller amounts of cloud water that reduce the rates of riming and conversion of snow to graupel. 0e distribution of hydrometeors affects the thermal and dynamical structure of the typhoon. 0e saturation adjustment hypothesis in the bulk schemes overestimates the condensate mass. 0us, the additional latent heat makes the typhoon structure warmer, which increases vertical velocity and enhances convective precipitation in the eyewall region.


Introduction
Recent rapid developments in computing technology and numerical modeling capability have enabled progress in simulation-based research using horizontal grid spacing with <5 km resolution.Because models with such fine resolution can resolve most convective clouds, explicit cloud microphysical schemes have replaced the cumulus convective parameterization schemes used previously in cloudresolving models [1].us, the explicit representation of microphysical processes is becoming increasingly important in numerical cloud-resolving models.Many microphysical parameterization schemes from bulk microphysics parameterization schemes (hereafter, bulk schemes) to spectral bin microphysics parameterization schemes (hereafter, SBM schemes) have been developed, improved, and applied to study the precipitation processes of typhoons [2].
Bulk schemes were developed earlier than SBM schemes, and they have been applied extensively in cloud-resolving models because of their advantageous computational efficiency and capability to reproduce observed features of clouds and precipitation in typhoons [3][4][5][6][7][8][9][10][11][12][13].For example, after satisfactory simulation of Typhoon Hagupit (2008) in terms of its track, intensity change, wind, and precipitation distribution by the Weather Research and Forecasting (WRF) model with the WSM6 bulk scheme, further research quantified every conversion rate and then analyzed the cloud microphysical processes, which revealed that the precipitation and structure of typhoons are influenced substantially by changes in microphysical processes [14,15].
In recent years, simulations have increasingly used SBM schemes instead of bulk schemes, especially following a demonstration of the SBM Hebrew University Cloud Model [16][17][18][19][20][21].By applying different microphysical schemes with various aerosol concentrations, the simulations of Hurricane Irene showed that storm intensity varied greatly and that only the SBM experiment with different aerosol concentrations could reproduce the observed temporal shift between the maximum wind and the minimum pressure [22].
In essence, the same microphysical processes are represented differently in bulk and SBM schemes.Bulk schemes calculate microphysical equations for several particle size distribution (PSD) moments, that is, n(m) with m being the particle mass, rather than for the PSDs of all types of hydrometeors.e k-th moment of the PSD can be defined as follows: in which k is usually taken as an integer.In contrast, SBM schemes do not require an a priori functional form of the PSDs.e primary differences between SBM and bulk schemes are that SBM schemes include parameterizations of various microphysical processes and interactions between particles that are more specific.rough dividing the particle distribution into a number of finite size or mass categories and then calculating the explicit microphysical equations, the representation of microphysical processes in SBM schemes is more reasonable than that in bulk schemes, which could improve simulation of precipitation.Furthermore, SBM schemes use identical equations and parameters to represent microphysical processes; therefore, any particular scheme could be applied without modification to different meteorological phenomena [2].
Although bulk schemes are used widely, they incorporate many unreasonable representations of microphysical processes that could cause errors in numerical simulations [2].For example, many simulations using bulk schemes overestimate the amount of rainfall.Furthermore, the various bulk schemes tend to differ considerably in terms of the number of interactions between microphysical particles, which means different simulations using different bulk schemes produce results with large dispersion [22].
Typhoons are important weather systems characterized by extreme winds and torrential rainfall.Cloud microphysical processes play a major role in the development and evolution of the cloud structure and precipitation of typhoons [10].Analysis of microphysical processes to investigate the sensitivities of typhoons or precipitation to different factors has become increasingly popular in recent years.Considerable research has been undertaken to verify and improve the efficacy of microphysical schemes through comparison and analysis of the effects of different schemes on the typhoon structure or precipitation [23][24][25][26][27][28].For example, some sensitivity analyses have been undertaken to identify the optimal combination of microphysical and convective schemes [29].However, many previous studies have focused on analysis of different bulk schemes despite the rapid development of the new approach using SBM schemes.In comparison with bulk schemes, the representation of microphysical processes in SBM schemes is more reasonable, which could lead to improved simulation of precipitation.us, it is valuable to study and compare the effects of bulk and SBM schemes on typhoon precipitation and structure.
In this study, we used the WRF model to conduct simulation experiments of Typhoon Hato (2017) at cloudresolving 3 km resolution.To examine the sensitivity of both the precipitation and the structure of typhoons to microphysical schemes, three bulk schemes (i.e., Lin, WDM6, and Morrison) and one bin scheme (SBM) were chosen.

Typhoon Development and
Experimental Design

Scheme Description and Experimental Design.
e Advanced Research WRF modeling system, whose solver is based on fully compressible nonhydrostatic equations, is a state-of-the-art atmospheric simulation system with terrainfollowing hydrostatic-pressure vertical coordinates.Arakawa C-grid staggering, the Runge-Kutta second-and thirdorder time integration options, and second-and sixth-order advection options (horizontal and vertical) are used in the Advanced Research WRF solver.It is portable and efficient on a single processor as well as being available for parallel computing platforms.
e cloud model has been applied widely in numerical simulations on various scales ranging from meters to thousands of kilometers, which demonstrates its capability and suitability for use in atmospheric research [30].
is study used WRF version 3.6.1 to simulate Typhoon Hato by performing 60 h integrations that were initialized at 1200 UTC on August 21, 2017, and ended at 0000 UTC on August 24, 2017.e initial 12 h period was treated as the model spin-up time.As shown in Figure 1, the simulation incorporated two nested domains: d01 and d02, Advances in Meteorology which had 9 and 3 km horizontal resolution [31] and corresponding time steps of 30 and 10 s, respectively.In the vertical, all simulations had 43 sigma layers with the model top at 50 hPa.For the initial and boundary conditions, we used the NCEP/NCAR global grid reanalysis dataset (NCEP-FNL; http://rda.ucar.edu/datasets/ds083.2), which is available with 1 °× 1 °spatial resolution and 6 h temporal resolution.None of the four experiments included assimilated observational data.Domain d01 used the Kain-Fritsch cumulus parameterization scheme [32][33][34]; however, no cumulus parameterization scheme was used for domain d02 because updrafts can be resolved sufficiently at 3 km resolution, resulting in explicit convective vertical transport [1].Long and shortwave radiation was computed using the Rapid Radiative Transfer Model for global climate models [35][36][37].Boundary layer processes were parameterized using the Yonsei University planetary boundary layer scheme [33,36,38].For the surface layer option, the revised MM5 Monin-Obukhov scheme was selected.e land surface option was set as the unified Noah land surface model.e scheme combination we used has been proved to be suitable for simulating typhoons [23,39].
In this study, all four experiments shared the same configuration shown above, but they incorporated different microphysical schemes.We chose three widely used bulk schemes (i.e., Lin, WDM6, and Morrison) for comparison with the SBM scheme, and each experiment was named in accordance with the microphysical parameterization scheme adopted.
e four microphysical schemes used have all been implemented in WRF model version 3.6.1,and they all contain six hydrometeors: water vapor, cloud water, rainwater, cloud ice, snow, and graupel.e Lin microphysics scheme [40][41][42] is a single-moment bulk scheme that outputs the particle mixing ratio.It has been developed by incorporating equations for snow in the bulk water microphysics scheme of Orville and Kopp [43].In the Lin scheme, snow can be generated both by the Bergeron process and contact freezing and by aggregation of cloud ice, following which it must then grow further to transform into hail.Cloud ice is allowed to coexist with cloud water in the temperature region of − 40 to 0 °C.Both the autoconversion of cloud water to form rain and the accretion of cloud water by rain are two major processes included for the formation of warm rain.e WDM6 scheme is a double moment developed by adding only double-moment warm-rain microphysics to the WSM6 scheme.e revised approach to the microphysical processes of ice has two distinguishing assumptions: (1) the number concentration of ice nuclei is a function of temperature and (2) the number concentration of ice crystals is a function of the amount of ice.Because of the indirect influence of the predicted number concentrations of cloud and rainwater on ice processes, it is an improvement on the single-moment scheme [42,44,45].e Morrison scheme is a complete double-moment scheme [46].e mass mixing ratios of five hydrometeor categories (cloud drops, ice, snow, rain drops, and graupel) are predicted, along with their total number concentrations.
e SBM schemes developed at the Hebrew University of Jerusalem (Israel) have been implemented in the WRF model.All the particles they contain are divided into 33 mass bins, and the equation system is solved for the size distribution functions of such particles.e full-SBM scheme has demonstrated strong universality in simulating various weather phenomena without requiring modification; however, its use remains restricted because of the enormous computational time required for the calculation of the equations.erefore, the fast-SBM scheme, which is based on the full-SBM scheme, was developed to improve the overall computational efficiency.By calculating all ice crystal and snow aggregates on one mass grid, the fast-SBM scheme reduces the computation time to less than 20% that of the full-SBM scheme, which means it can be run on standard PC clusters.
e fast-SBM scheme describes identical microphysical processes as the full-SBM scheme by solving kinetic equations with no a priori function, which means it retains all the advantages of an SBM scheme.Simulations of squall lines conducted using both fast-SBM and full-SBM schemes have shown they produce similar microphysical and dynamical structures and equivalent amounts of precipitation.

Comparison with Observations.
Figure 2 shows the observed track of Typhoon Hato, obtained from the China Meteorological Administration Shanghai Typhoon Institute (CMA; http://www.typhoon.org.cn/), and the results of the four simulations at 3 h intervals from 0000 UTC on August 22 to 0000 UTC on August 24, 2017.It can be seen that the results of the four experiments are similar and that the simulated tracks agreed well with the observations before the typhoon made landfall, reflecting the capability of the model in simulating the typhoon track.
e simulated times of landfall (at about 0900 UTC on August 23) lag behind the observed time by about 4 h.After the typhoon made landfall, the tracks varied a little, which may be caused by the landfall time lag and the influence of topography.Figure 3 shows the minimum sea level pressure and the maximum surface wind speed of the four simulations of Typhoon Hato together with the observed values.e simulated minimum sea level pressure and maximum surface wind speeds were in agreement with the observed values while three bulk experiments simulated a stronger typhoon than the SBM scheme.
e minimum sea level pressure values of Lin and Morrison at 0700 UTC on August 23 were 937.0 and 942.8 hPa, respectively, which were very close to the observed pressure (i.e., 940 hPa at 1000 UTC).After landfall, the observed and simulated intensities all weakened rapidly, but the simulated intensities weakened more slowly than the observed ones, which might be attributable to the delayed time of landfall.
Figure 4 shows the observed and simulated 24 h accumulated precipitation (0000 UTC on August 23 to 0000 UTC on August 24, 2017) of Typhoon Hato.Generally, the precipitation distribution in each of the experiments was similar to that observed.For example, the main precipitation event observed in western Guangdong Province was reproduced by all the experiments; however, the maximum rainfall amount of the Lin, WDM6, Morrison, and SBM experiments was 506.53, 408.05, 454.26, and 261.24 mm, respectively, while the observed value was 322.4 mm. e bulk schemes produced extremely torrential rainfall.Moreover, the area in which precipitation was >200 mm was obviously too large in the three bulk experiments, whereas it was closer to the observations in the SBM experiment.Figure 5 reveals that the three bulk schemes overestimated the rainfall in the eyewall area, which led to overestimation of the rainfall amount along the typhoon track.
From the above analysis, it was evident that the different microphysical schemes produced similar typhoon tracks as in previous studies but with different precipitation [6,48,49].e occurrence of precipitation is closely related to cloud microphysics; therefore, it is important to consider how the different cloud microphysical schemes might produce differences in precipitation.6 compares the fields of vertically integrated mixing ratios of cloud water in the four experiments at the time of landfall.e maximum vertically integrated mixing ratio of cloud water reached only 3.2 kg/ m 2 in the SBM experiment; however, it exceeded 8.3 kg/m 2 in all three bulk experiments (it even reached over 16.0 kg/ m 2 in WDM6).Furthermore, in the bulk experiments, cloud water was located across the entire range of the typhoon, and its areal coverage was wider than that in the SBM experiment.In bulk schemes, new nucleated droplets obey a gamma distribution similar to other cloud droplets [2].e formation of large amounts of cloud water might increase the mixing ratio of cloud water, although its existence could be unreasonable.Furthermore, the application of saturation adjustment in bulk schemes leads to condensation of all water at the end of each time step [2], which might form additional cloud water.Moreover, large cloud droplets could absorb more water vapor, which would decrease the supersaturation and hinder growth of the effective radius, preventing rainwater formation.Our simulated results showed there was less rainwater in the bulk schemes than in the SBM experiment (Table 1).

Cloud and Snow. Figure
As all hydrometeors are distributed in three dimensions, we plotted the vertical profiles of the hydrometeors by calculating the domain average within a 210 km radius of the typhoon center at the time of landfall (Figure 7).It can be seen that the SBM and bulk experiments differed most in terms of cloud water and snow.
e differences in the horizontal distribution and magnitude of cloud water have been discussed above.Figure 7 shows that little cloud water occurred above the height of 9 km in the SBM experiment, whereas significant cloud water content reached the height of 11 km in the bulk experiments (even reaching 13 km when using the Lin scheme).e overestimation of cloud water at such levels in the bulk experiments could increase the rates of riming and of the conversion of snow to graupel [22], leading to the lower snow content and larger graupel content.e snow content represented the greatest difference among all the hydrometeors between the bulk and SBM experiments, which can be seen in both vertical and horizontal distributions.In the SBM experiment, the snow content was considerable from 5 to 16 km; it was >0.7 g/kg between 7 and 12 km with a maximum of about 0.95 g/kg at the height of 9 km.Conversely, in all the bulk experiments, the maximum snow content was less than 0.5 g/kg.e horizontal distribution and magnitude of snow content are shown in Figure 8.
e maximum vertically integrated mixing ratio of snow in the SBM experiment was >24 kg/m 2 , whereas in the bulk experiments, it was about 8 kg/m 2 .Furthermore, the horizontal distribution of snow in the SBM experiment was obviously larger than that in the bulk experiments.e process of in-cloud nucleation in the SBM experiment, which determines the cloud condensation nucleus size distribution with supersaturation at every grid point, leads to penetration of greater numbers of small droplets to higher levels.e presence of these small droplets has an effect on ice microphysics [2], which could explain the large snow content.
e collision and aggregation of particles in all the bulk schemes used averaged values, and those particles obeyed a gamma distribution after each microphysical process; however, this could result in extremely unreasonable vertical distribution of the particles [2].Conversely, in the SBM scheme, these microphysical processes depended on the type and mass of particles, which lead to size sorting in the vertical direction that was considered a benchmark.From the vertical profiles, it can be seen that the content of solid particles in the SBM experiment was concentrated mainly above the height of 6 km and that the water vapor extended to a higher level than that in the bulk experiments.Table 1 shows that the ice water path (IWP) in the SBM experiment was larger than that in the bulk schemes.ese results indicate that the generation of solid particles might be caused by nucleation with subsequent condensational growth from water vapor.In the SBM experiment, the higher fall velocities of large particles in comparison with small particles meant large particles descended, whereas small particles remained floating at higher levels.In the presence of abundant water vapor, these small particles could be nucleated as ice nuclei.In contrast, the gamma distribution of particles in the bulk experiments produced minimal content of ice crystals above the cloud base.Furthermore, the averaged fall velocities of the particles introduced errors in the self-collision and aggregation processes [2] that could lead to miscalculation of the snow content.All these mechanisms, including the interaction of warm and ice microphysics, resulted in different distributions of hydrometeors, which could influence the thermal and dynamical structure of a typhoon, as well as its precipitation rate.

Precipitation Types.
To explore the reasons for the different distributions and magnitudes of precipitation associated with the different microphysical schemes, we calculated certain microphysical quantities and we separated the precipitation types.We followed the method of Sui et al. [50] to separate precipitation into convective, stratiform, and mixed types.First, we calculated the vertically integrated Advances in Meteorology mixing ratios of cloud hydrometeors.en, the sum of the vertically integrated mixing ratio of ice/water hydrometeors was calculated as the IWP/liquid water path (LWP).en, the cloud ratio, defined as the ratio of IWP to LWP, was used to evaluate the relative importance of ice and warm clouds.e grid was designed to represent convective rainfall when the corresponding cloud ratio was <0.2 or when the IWP value was >2.55 kg/m 2 , and to represent stratiform rainfall when the corresponding IWP value was <2.55 kg/m 2 and when the cloud ratio was >1.0; the remaining grids were regarded as mixed precipitation.
Table 1 shows that the grid number percentages of both stratiform-type and convective-type precipitation in the SBM experiment were larger than those in the three bulk   IWP and LWP are the ice water path and liquid water path, respectively; q r , q c , q i , q s , and q g are vertical integrations of rainwater, cloud water, cloud ice, snow, and graupel, respectively.Advances in Meteorology experiments both before and after typhoon landfall.

Advances in Meteorology
e spiral distribution at 0900 UTC on August 23, 2017, is plotted in Figure 9. Corresponding to the distribution of 24 h accumulated rainfall, the distribution of both convectivetype and stratiform-type precipitation in the SBM experiment was wider than that in the bulk experiments.It can be seen that the area of convective-type precipitation in the three bulk experiments was concentrated in the typhoon eyewall region, especially in the single-moment experiments (i.e., Lin), which caused extremely torrential rainfall in that area with a maximum value far greater than that observed.In the SBM experiment, the wider distribution of stratiformtype precipitation, which has a smaller precipitation rate relative to convective-type precipitation, produced a wider area of precipitation closer to the observations.According to the method adopted, when the IWP value of a grid was >2.55 kg/m 2 , the grid was designated as convective-type precipitation. is was because grids with large IWP values might have strong ascent.Stratiformtype precipitation was assigned to grids with a cloud ratio <1.0, meaning ice cloud processes were more important than warm cloud processes in a typhoon [50].e larger distributions of stratiform-type and convective-type precipitation in the SBM experiment than those found in the bulk experiments corresponded to the larger value and wider distribution of snow, indicating the importance of ice cloud processes in a typhoon.Table 1 shows that the IWP values in the SBM experiment before and after typhoon landfall were 3.73 and 3.65 kg/m 2 , respectively, whereas the IWP value in the bulk experiments was just below 2.9 kg/m 2 .
As mentioned above in relation to the SBM experiment, the calculation of the cloud condensation nucleus size distribution and supersaturation at every grid means the in-cloud nucleation processes transported small droplets to higher levels, and then the different fall velocities of the particles of different size produced appropriate sorting.Small ice crystals nucleated as ice nuclei that grew at such high levels make more solid particles, leading to an increased IWP value.In the bulk experiments, the spectrum of new nucleated droplets above the cloud base remained described by a gamma function, leading to unreasonable expression of large cloud droplets and preventing raindrop formation.us, there might be more cloud water but less rainwater in the bulk experiments in comparison with the SBM experiment.Furthermore, the cloud water at high levels in the bulk experiments had an effect on ice microphysical processes, leading to unreasonable representation of solid particles.Overall, these differences in hydrometeor content between the experiments not only affected the warm/ice microphysics but also changed the dynamical structure of the typhoon (as discussed below) and, ultimately, influenced the rate and type of precipitation.e different distributions of hydrometeors attributable to cloud microphysics could be expected to change the thermodynamics of a typhoon and produce a dynamical response.Figure 10 shows the axially symmetric vertical structure of temperature deviation and vertical velocity before the typhoon made landfall.It is evident that all experiments produced a warm core structure in the center of the typhoon, consistent with the facts, although the range and intensity of the warm core vary.e warm core structure in Lin and WDM6 experiments reached a height of about 8 km over the typhoon center, while in Morrison and SBM experiments, it was higher.e maximum temperature in each of the three bulk experiments was larger than 8.0 °C, but in the SBM experiment, it was only 6.8 °C.We considered the stronger warm core in the bulk experiments was attributable to the application of saturation adjustment in simulating a typhoon with large updrafts.In convective updrafts, especially in a typhoon, supersaturation could reach several percent because not all the water vapor condenses when supersaturation is greater than zero [2].Saturation adjustment assumes that supersaturation over water is forced to zero.As mentioned earlier, water vapor condensed to produce excess cloud water in bulk schemes, meaning it overestimated the condensate mass and released more latent heat, which is in accord with the simulation of structures with stronger warm cores.Furthermore, the warm core in the SBM experiment was higher, which had a substantial effect on ice microphysical processes.It was also more peripheral, indicating greater condensation that was reflected in the larger distributions of both stratiform-type and convective-type precipitation.10 Advances in Meteorology e greater the release of latent heat, the deeper the warm core structure and the larger the vertical velocity.From Figure 10, it can be seen that the vertical velocity in the typhoon eyewall is accompanied by the warm core.In all bulk experiments, the density of the contour lines illustrated the strong velocity gradient and reflected the rate of velocity increase.As vertical velocity increased, convection was strengthened and the rate of precipitation was enhanced.It can be seen from Figure 10 that the maximum vertical velocity in the SBM experiment was the smallest of all the simulations.It can also be seen that the largest value of vertical velocity in the bulk experiments was located within 50 km of the typhoon center, whereas in the SBM experiment, it had a relatively lower value and was more peripheral.Such a dynamic structure in the bulk experiments produced short-term heavy rainfall in the eyewall area.Overall, in comparison with the bulk experiments, we attributed the better simulation of precipitation in the SBM experiment to more reasonable representation of droplet growth by diffusion.

Conclusions
is study used the WRF model with 3 km resolution to simulate Typhoon Hato (2017) to test the sensitivity of the precipitation and the structure of the typhoon to bulk (Lin, WDM6, and Morrison) and explicit SBM schemes.
e results produced little evidence to suggest the different schemes could alter the track of the typhoon.However, it was evident that the rainfall simulated using the SBM scheme was closest to that observed.In comparison with observations, the magnitude of precipitation produced by all three bulk schemes was larger in the eyewall region and the area with accumulated precipitation of >200 mm was obviously too wide.Advances in Meteorology e horizontal and vertical distributions of hydrometeors showed that the cloud water content in the SBM experiment was smaller than that in the bulk experiments, whereas the converse was true for rainwater content. is was because of different nucleation and diffusion growth of the particles in the different schemes.e bulk experiments overestimated cloud water content because particles were forced to adopt a gamma distribution.e large cloud water content could absorb more water vapor, decreasing the supersaturation, which hindered the growth of the effective radius and prevented rainwater formation.In addition, in bulk schemes, droplet growth by diffusion was achieved by applying a saturation adjustment that condensed all the water vapor, which also contributed to overestimation of cloud water content.
Among all the hydrometeors, the vertically integrated mixing ratio of snow was found to represent the greatest difference between the bulk and SBM experiments.Snow content in the SBM experiment was obviously larger than that in the bulk experiments, as was the IWP.
is was because the SBM scheme calculated the cloud condensation nucleus size distribution together with supersaturation at every grid point.is caused small droplets to form at high levels, which could affect ice microphysical processes.Differential fall velocities can lead to size sorting that leaves small particles to act as ice nuclei at higher levels.Together with abundant water vapor, the nucleation and growth of small particles on ice nuclei can increase the snow content.In addition, smaller amounts of cloud water can decrease the rates of riming and of the conversion of snow to graupel.
rough determination of precipitation type, it was found that the grid number percentages of both stratiformtype and convective-type precipitation in the SBM experiment were larger than those in the three bulk experiments, indicating that ice cloud microphysical processes were important in typhoons.In addition, cross sections of temperature and vertical velocity showed that the SBM experiment produced the weakest, outermost, and highest warm core typhoon structure with the smallest vertical velocity.
e application of saturation adjustment in the bulk schemes during simulation of the typhoon overestimated the condensation mass and latent heat.Consequently, in comparison with the SBM experiment, the warmer thermodynamic structure in the bulk experiments produced stronger dynamics, and the larger vertical velocity caused stronger convective precipitation in the eyewall area.
In summary, the representation of microphysical processes in the different schemes changed the distribution of hydrometeors, and the interaction of all hydrometeors then affected the cloud structure and the rates of both warm and cold microphysical processes.Finally, the microphysical processes lead to the release of latent heat, which affected the vertical velocity and the precipitation rate.
e observation of cloud microphysical processes is extremely difficult, especially ice cloud processes.Because of the lack of adequate observational data, the representation of microphysical processes in SBM schemes still had certain shortcomings.Nevertheless, we have proven that the inherent insensitivity of bulk schemes to particle size means they always overestimate precipitation, whereas SBM schemes that describe microphysical processes more reasonably are better able to simulate the precipitation and cloud structure of a typhoon.Considering the rapid increase in computational capacity over recent years, SBM schemes may be adopted to realize optimal prediction of precipitation in future numerical simulations.

Data Availability
e NCAR Mesoscale and Microscale Meteorology Division provided the WRF model, which is available at http://www.mmm.ucar.edu/wrf/users.
e National Centers for Environmental Prediction Final Global Tropospheric Analysis (NCEP-FNL) data were taken from https://rda.ucar.edu/datasets/ds083.2. e data used to support the findings of this study are available from the corresponding author upon request.

Figure 1 :
Figure 1: Two nested domains used for the simulations-d01 and d02-with horizontal resolutions of 9 and 3 km, respectively.

Figure 2 :
Figure 2: ree-hourly track of Typhoon Hato obtained from the CMA (OBS) and simulated in the four experiments (Lin, WDM6, Morrison, and SBM) in domain d02.Period is from 0000 UTC on August 22, 2017, to 0000 UTC on August 24, 2017.

Figure 3 :
Figure 3: (a) Minimum sea level pressure (unit: hPa) and (b) maximum surface wind (unit: m/s) of Typhoon Hato obtained from the CMA (OBS) and simulated in the four experiments in domain d02.Period is from 0000 UTC on August 22, 2017, to 0000 UTC on August 24, 2017.

Figure 7 :
Figure 7: Vertical profiles of hydrometeors averaged over the domain within a 210 km radius of the center of Typhoon Hato at 0900 UTC on August 23, 2017, in the four simulations: (a) water vapor mixing ratio; (b) cloud water mixing ratio; (c) rainwater mixing ratio; (d) ice mixing ratio; (e) snow mixing ratio; (f ) graupel mixing ratio.

Table 1 :
Domain averages within a 210 km radius of the center of Typhoon Hato and time averages before and after landfall of important variables in the four simulations.s is the precipitation rate; convective, stratiform, and mixed are representative of the proportion of grids of each precipitation type to all precipitation grids; P