1. Introduction
An accurate description of microphysical processes is important in improving the characterization of precipitation and the horizontal and vertical structures of cloud systems (as well as their evolution with time) in climate models [
1,
2]. The terminal velocity of hydrometeors is one of the most important parameters affecting precipitation [
3], and it varies with their size and shape. If a group of hydrometeors (such as those located in the same voxel of model grids) can be taken as an object, we can calculate their mean terminal velocity. Mean terminal velocity can reflect the sedimentation of the gravity center of the hydrometeor group [
4,
5].
The magnitudes of the terminal velocity affect the relative motion among the hydrometeors [
6], thus changing the collision frequency among different types or sizes of hydrometeors, the precipitation rate, as well as macro- and microphysical processes of cloud systems [
7]. Ouyang [
8] found that different parameterization schemes of terminal velocity affect the vertical distribution of hydrometeors in clouds. Ice particles play an important role in the formation of precipitation in convective clouds [
9]. An increase in terminal velocity of graupel particles led to the enhanced drag force of hydrometeors. Xu et al. [
10] found that smaller terminal velocities of graupel and hail would lead to an increase in upward-moving graupel. Therefore, reducing terminal velocities of graupel and hail at mid- and lower tropospheric levels would increase sources of upward-moving graupel and hail and reduce local maximum rainfall intensity. Wang [
11] found that larger terminal velocity of snow and hail used in bulk microphysical schemes lead to the overestimation of precipitation intensity. Lou et al. [
12] found that the modified mean terminal velocity of graupel affected the collision of graupel with other hydrometeors, as well as graupel melting. Franklin et al. [
13] found that the increase in the terminal velocity of graupel increased the rain in the precipitation center of a tropical cyclone and decreased the coverage of precipitation. Wang and Qiao [
14] discussed the influence of terminal velocities of rain, snow, and hail on precipitation and found that the maximum precipitation intensity is more sensitive to the terminal velocity of rain, while the precipitation range is more sensitive to terminal velocity of ice particles such as snow and hail. Therefore, the change of mean terminal velocity not only affects the cloud and precipitation-related microphysical processes but also affects cloud dynamics.
The larger the difference in terminal velocity among hydrometeors, the larger the collision efficiency [
15]. As such, accurate calculation of terminal velocity plays a significant role in predicting the number and mass concentration of hydrometeors in cloud microphysical schemes. Several climate models, such as the Community Earth System Model [
16,
17], employ the two-moment Morrison scheme [
18], which assumes the hydrometeors obey the gamma distribution function. They calculate the mass-weighted mean terminal velocity (
Vm) and number-weighted mean terminal velocity (
Vn) based on the shape parameter (
α), slope parameter (
β), and the empirical coefficient (
Q,
k). The two-moment bulk scheme generally calculates both
Vm and
Vn. Compared with the single-moment bulk scheme, which predicts only
Vm (considering only hydrometeor mass-weighted mean terminal velocity), the two-moment scheme can simultaneously consider the settlement of both large and small particles, which can better reflect the effect of size sorting [
19]. The
Vm is larger than
Vn [
20], in that
Vm is mostly affected by those hydrometeors with large sizes, whereas
Vn is mostly affected by those small hydrometeors with larger numbers [
6,
21].
The shape parameter
α of the gamma distribution is usually fixed when calculating the mean terminal velocity of hydrometeors, varying from 1 to 6 [
1,
19,
22,
23,
24]. However, the size of the hydrometeors change with time and location. Hence, setting α to a fixed constant cannot provide an accurate calculation of the mean terminal velocity [
25].
Moreover, the mean terminal velocity of the hydrometeors in a cloud may vary under different background aerosol concentrations. Aerosols can be activated as cloud condensation nuclei to participate in the formation of cloud droplets. The number of cloud droplets increases with the initial aerosol concentration. Due to the competition of water vapor during condensational growth [
26], the effective radius of cloud droplets decreases, and the cloud droplet size distribution becomes narrower, thus affecting the size distribution of other types of hydrometeors [
27]. Aerosols can affect the mean terminal velocity of hydrometeors through the above process, and then influence the movement trend of hydrometeors with updraft (mobility). Koren et al. [
5] proposed that aerosol effects on cloud droplet mobility are one of the most important factors to promote the early development of convection. Koren et al. [
28] found that the change in aerosol concentration affected the mass-weighted mean terminal velocity of precipitation particles. Under the premise that aerosols promotes convection, the higher the background aerosol concentration is, the greater the drop mass-weighted mean terminal velocity in the mature stage of cloud development. The different aerosol concentrations not only change the number concentration of cloud droplets but also have a non-negligible impact on the spectrum shape of cloud droplets [
27]. In most bulk schemes, the premise of calculating the mean terminal velocity is that the hydrometeors obey a gamma distribution, and the parameters of the gamma distribution determine the parameterization of the mean terminal velocity [
28], which can directly affect the forecasting ability of precipitation by a numerical model [
25].
The volume-mean diameter (
Dv) can be used to represent the mean volume or mass of the hydrometeors [
24]. Zou et al. [
29] and Seifert [
30] fitted the dispersion of different hydrometeors and droplet spectrum shape parameters, respectively. Based on this previous research, this study aims to analyze the relationship of
Dv and
Vm (
Vn) to establish parameterizations of
Vm and
Vn and explore their sensitivity to the background aerosol concentration.
3. Results and Discussion
3.1. Relationship between Vm and Dv
Modern bulk microphysics in mesoscale models and climate models typically uses the mass-weighted mean terminal velocity of different hydrometeors for the description of sedimentation [
18]. In this paper, the hydrometeors are assumed as spherical particles, and the equivalent volume-mean diameter is calculated using their mass and density. For each selected voxel as listed in
Table 1, the mass-weighted mean terminal velocity and volume-mean diameter of the group of particles within the given voxel were calculated for all types of hydrometeors. Then the joint probability density function of the relationship between
Vm and
Dv are presented in
Figure 1. A larger volume-mean diameter means a greater mass of the hydrometeors, therefore greater mass-weighted mean terminal velocity. It shows that there is a positive correlation between mass-weighted mean terminal velocity and volume-mean diameter (
Figure 1) for all types of hydrometeors. The correlation coefficients R are greater than 0.8.
The relationship between the mass-weighted mean terminal velocity and the volume-mean diameter is closely related to the growth of cloud droplets and raindrops. The mass-weighted mean terminal velocity is more susceptible to large-sized particles, thus the larger the cloud drop size is, the larger the mass-weighted mean terminal velocity of the hydrometeors. The increased aerosol concentration leads to a higher number concentration of cloud droplets with reduced effective radius, and a lower occurrence of collisions [
32]. It can be seen in
Figure 1 that the maximum value of the volume-mean diameter of the cloud droplet decreases with enhanced aerosol loading. In the polluted condition, the number concentration of cloud droplets with smaller sizes tends to be higher than that in clean conditions. The number of large droplets in the polluted cloud is also greater than that in the clean cloud (
Figure 2). The mass-weighted mean terminal velocity is more affected by large-sized droplets, so
Vm is larger for a given volume-mean diameter (the slope of the fitting curve is larger).
The
Vm of the ice-phase hydrometeors is also positively associated with
Dv. The more common the large ice crystals in the particle swarm are, the greater the mass-weighted mean terminal velocity and the larger the volume-mean diameter [
21]. The ice crystals can collide with each other and grow into the snow, and they can also rime as the supercooled liquid falls and forms graupel. Snow and graupel increase with the development of the cloud system, and the volume-mean diameter also gradually increases. The existence of large ice particles will accelerate the process of aggregation and riming, then the large particles will become larger, resulting in greater mass-weighted mean terminal velocity. In the polluted cloud, affected by rising airflow, more small cloud droplets can be carried upward by the updrafts and reach high altitudes above 0 °C, where they can freeze into small ice crystals [
32]. Therefore, the size distribution of ice crystals shifts to smaller sizes, changing the mass-weighted mean terminal velocities of ice crystals, snow, and graupel.
According to Equation (10), when the shape parameter is a fixed value, the relationship between
Vm and
Dv should be
, where a is the constant value and
k is the effective terminal velocity parameter of the corresponding falling particle. Because the shape parameters are not fixed values, we consider fitting the function using the formula
. The intercept (c) appears in the fitting equation, probably because the viscous flow dominates in regions of smaller particle size [
21]. The values of coefficients a, b, and c are shown in
Figure 1, and all fitting functions passed a two-tailed
t-test with a significance level of 95%. The size distribution of different hydrometeors changes with different aerosol loading, with more large hydrometeors appearing in more polluted clouds, which is crucial for calculating
Vm. The slope of the
Vm parameterization changes accordingly, resulting in different mass-weighted mean terminal velocities (
Figure 1). Hence, it is necessary to establish specific parameterization under different aerosol conditions to improve the accuracy of bulk microphysics.
3.2. Relationship between Vn and Dv
Figure 3 shows the joint probability density distribution of the relationship between volume-mean diameter and number-weighted mean terminal velocity, which also displays a positive correlation. Because both
Dv and
Vn are number-weighted terms, there is a strong correlation between them, with the correlation coefficient R greater than 0.9. Based on the derived Equation (11),
is used for fitting, which is similar to that for mass-weighted mean terminal velocity. The values of coefficients a and b are shown in
Figure 3, and all the fitting functions passed the two-tailed
t-test with a significance level of 95%.
Compared with Vm, the number-weighted mean terminal velocity is more sensitive with smaller-sized hydrometeors, hence the values of Vm are greater than that of Vn. During the development of a convective cloud, the smaller-sized cloud droplets gradually grow through condensation, and the large-sized particles lead to the larger Dv and Vn of cloud drops, so Vn of the cloud drops is positively correlated with Dv. Small-sized raindrops also grow with the collisions, but the collection is slower and less efficient than for large-sized raindrops, so the slope of the Vm parameterization is greater than that of Vn. Vn of ice-phase hydrometeors also has a positive correlation with Dv, and small ice crystals are mainly formed by the freezing of small cloud droplets carried up by updraft. Both volume-mean diameter and number-weighted mean terminal velocity increase with condensation. Similar to raindrops, the collection of smaller-sized snow and the sedimentation of small graupels is slower than that of larger-sized ones, so the slope of the Vn parameterization for ice particles is smaller relative to that of Vm.
The parameterization of
Vn is also different under different background aerosol concentrations. Unlike
Vm, which is affected mainly by an increased number of large particles,
Vn is susceptible to a much larger number of small particles (
Figure 2). With enhanced aerosol loading, the number concentration of small cloud droplets increases, resulting in the lower number-weighted mean terminal velocity, and the slope of
Vn parameterization is smaller than that in clean air conditions. Ice crystals are affected by aerosols indirectly, and more small-sized ice crystals are formed due to the freezing of small cloud droplets. Consequently, the
Vn of ice crystals is smaller in polluted conditions for a given volume-mean diameter. According to the size distribution of the ice hydrometeors, it can be found that the proportion of the number of small particles increased more, so the
Vn parameterization results of snow and graupel decrease compared with the clean air situation. It is found that the
Vn of different hydrometeors is affected mostly by the increasing proportion of small particles under polluted conditions, hence the slope of parameterization decreases.
3.3. Comparison of Different Parameterized Schemes
In many climate models, the gamma distribution shape parameter is calculated as a constant value for the mass-weighted mean terminal velocity and the number-weighted mean terminal velocity with Equation (10) and Equation (12), respectively. However, Vm and Vn cannot be reasonably set to the fixed constants. Therefore, we compared the above fitting results with the commonly used parameterizations of the mass-weighted mean and number-weighted mean terminal velocities when the shape parameter is set to a fixed value. This provides a reference for the selection of the shape parameter values in those models.
Overall, the mass-weighted mean terminal velocity is larger than the number-weighted mean terminal velocity, which is consistent with the expected results. For the fixed-shape parameterization tests, both
Vm and
Vn increase with the
Dv, and
Vm decreases and
Vn increases with shape parameter values. According to
Figure 4, it can be found that the arc trend of the curve of the number-weighted mean terminal velocity is consistent with the ones when
α is constant, so
Vn can be reasonably described if the appropriate constant is selected. However, the mass-weighted mean terminal velocity is different between our fitting curve and that calculated from a constant
α, especially for the cloud and rain, hence
Vm is less reliable than
Vn when
α is assumed to be constant.
Except for cloud droplets, parameterizations of other hydrometeors calculated using fixed α value will overestimate Vm of small volume-mean diameter particles, which may make small particles fall faster and reduce their relative speed, resulting in lower collision efficiency of large particles, delaying the onset of precipitation. For cloud droplets, parameterization of Vm using a fixed α value will lead to overestimation for the entire size distribution, which is more significant for the droplets with larger volume-mean diameter. This can increase the collection efficiency of large cloud droplets.
To further explore the impact of the new parametrization of mean terminal velocity on the vertical structure of the microphysics, the convective system was simulated using the WRF model with the default Morrison scheme (Control) and the modified Morrison scheme with new parameterization of
Vm and
Vn (ParamV), respectively (
Figure 5). The default Morrison scheme calculates
Vm and
Vn with shape parameter α = 3 of the cloud droplets and α = 1 of the other hydrometeor species. Considering that the simulation area is around the Marshall Islands in a tropical region, the ParamV test uses the terminal velocity parameterizations obtained under clean condition. Based on the discussion in
Figure 4, we can expect that the ParamV experiment predicts reduced
Vm and increased
Vn.
The prediction of the mass concentration and the number concentration of different hydrometeors in the Morrison scheme is determined jointly by the advection, the turbulent diffusion, and the source and sink of the microphysical process [
18]:
Among them, the mass- and number-weighted mean terminal velocity affects the mass concentration and the number concentration by directly affecting sedimentation, and , and indirectly impacting microphysical processes.
According to
Figure 5, it can be found that due to the decreasing
Vm of the raindrops, sedimentation is inhibited, resulting in an increased mass concentration of the raindrops. The
Vn of raindrops increases, which reduces the concentrations at different altitudes. Snow can exist from 4 km at lower levels to 15 km in the upper troposphere. With decreased
Vm in the ParamV experiment, the settlement slows, resulting in the mass concentration of snow increasing at higher altitudes (8–15 km), and only changing slightly at lower altitudes (0–8 km). Snow sedimentation is faster with increased
Vn, resulting in smaller number concentrations. A similar trend can be found for graupels. At higher altitudes (11–15 km), the increase of
Vn for graupel may enhance the collection of ice crystals, leading to a reduced number concentration of ice crystals. The results show that using the modified parametric scheme of
Vm and
Vn increases the average mass concentration of raindrops, snow, and graupel, and the total amount of all precipitation particles in the whole cloud area increases accordingly, which is consistent with the research results of Tao et al. [
33].
The sedimentation of cloud droplets and ice crystals is slow due to their low terminal velocities (<<1 m/s). Therefore, cloud droplets and ice crystals are mainly affected by the changing microphysical processes due to different
Vm and
Vn parameterizations (an indirect mechanism). Due to the complexity of cloud microphysical processes, the changes in number and mass concentration of cloud droplets and ice crystals are not discussed here further, and only that of precipitating particles, such as raindrops, snow, and graupel, are presented in
Figure 5.
4. Conclusions
The WRF-SBM simulations of a convective system over the Marshall Islands were used to explore the relationship between the volume-mean diameter and the mean terminal velocity of different hydrometeors, and the mass-weighted and number-weighted mean terminal velocity parameterizations were established under different background aerosol concentration conditions.
It was found that the mean terminal velocity of hydrometeors and the volume-mean diameter showed a positive correlation for different aerosol scenarios, and the correlation coefficient was greater than 0.8. The Vm was affected mostly by the large hydrometeors, so Vm was much larger than Vn for a given volume-mean diameter. The larger volume-mean diameter of the hydrometeors increased the collection efficiency of large particles in the cloud, which further increased the slope of Vm.
Because the aerosol concentration affects the mean terminal velocity of hydrometeors through microphysical processes, the parameterizations under the clean and polluted conditions are established separately. Under polluted conditions, enhanced aerosol loading increases the concentration of both small and large hydrometeors. The Vm is mainly determined by the increased number of large particles, and more aerosols lead to larger Vm. The Vn is affected mostly by the increase of small particles, and Vn is smaller in polluted conditions.
The Vm and Vn parameterizations usually use fixed shape parameters. It is found that using a fixed shape parameter generally underestimates the Vm of the hydrometeors when the volume-mean diameter is small. With a newly established parameterization of Vm and Vn, we find that the changes in Vm and Vn directly affect sedimentation, and indirectly affect microphysical processes. For precipitating hydrometeors, such as raindrops, snow, and graupels, the mass concentration increases due to the reduced Vm which inhibits sedimentation, whereas the number concentration decreases with increased Vn. The differences in the mean terminal velocity parameterization scheme impact the vertical distribution of different hydrometeors. The parametric schemes obtained in this study can be used to improve sedimentation in bulk microphysics, which provides a reference for other climate models and can improve precipitation forecasts.
Due to the complexity of the microphysical processes, the responses of cloud droplets and ice crystal number concentration and mass concentration are not the same. Moreover, the influence of the mean terminal velocity parameterization scheme under different background aerosol concentration conditions on cloud microphysical processes and precipitation warrants further investigation.