-
The large-scale synoptic pattern during RF0718 is illustrated by the 900-hPa geopotential heights obtained from ERA5 (Fig. 1a). As shown in Fig. 1a, the Azores High was located to the southwest of the ARM-ENA site, while the Icelandic Low was located on the northeastern side. Therefore, the prevailing wind during RF0718 was northwesterly. The vertical flight path is overlaid on the concurrent ground-based radar reflectivity (Fig. 1b). From ~ 0900 to 1100 UTC, the aircraft was ramping from near the sea surface up to the free troposphere. The aircraft had taken multiple horizontal flight legs accompanied by spirals when ascending or descending. Although the point-based radar reflectivity is not spatially matched with the cloud sampled by the aircraft, it provides the approximate aircraft locations, such as near the sea surface, sub-cloud, cloud-base, cloud-top and the free troposphere. This flight strategy was selected so that the aircraft sampled on one side follows the direction of the prevailing wind and the other side crosses the prevailing wind at similar height levels. Note that the flight legs collected between 1100–1200 UTC are not included in this study due to the lack of flight legs that collect sub-cloud aerosol samples and the lack of different levels of cloud samples.
Figure 1. (a) 900 hPa geopotential height (contour) and wind (arrow, color denotes wind speed) in a 50° × 30° domain surrounding the ENA, and the red star denotes the position of ARM-ENA site; (b) the ARM-ENA ground-based radar reflectivity (contour) overlayed by the ceilometer measured cloud base (black dot) and RF0718 aircraft vertical flight track (purple line) between 0830 to 1200 UTC.
To examine the horizontal structure of stratocumulus during RF0718, we plot Fig. 2 which shows the Meteosat-10 retrieved cloud optical depth (COD) over the 2° × 2° domain centered on the ARM ENA site at Graciosa Island. Combining the horizontal structures of COD from 900 to 1030 UTC (Fig. 2) and the Moderate Resolution Imaging Spectroradiometer (MODIS) real color image (not shown) during RF0718, we found that the study domain was dominated by a closed-cell marine stratocumulus clouds, which propagated from northwest to southeast. There were multiple bands of enhanced COD structures embedded in the stratocumulus cloud deck, and the evolution of the COD bands was generally along the direction of cloud propagation. Note that the aircraft horizontal flight path appears as an L-shaped pattern (purple lines on Fig. 2) with two directions: Leg-1 (northwest-southeast) is parallel to wind direction and cloud propagation, while Leg-2 (southwest-northeast) is perpendicular to wind direction and cloud propagation. Both legs consist of multiple flight segments at different altitudes, i.e., at each altitude or horizontal flight leg, the aircraft sampled the cloud (and/or aerosol) microphysical properties along both Leg-1 and Leg-2. These in situ measurements at each altitude are used to investigate the small-scale variation of cloud and drizzle properties in both horizontal and vertical directions. Furthermore, this flight strategy not only allows us to examine the marine stratocumulus cloud heterogeneity, but also interactions with sub-cloud aerosols, which will be discussed in the following sections.
Figure 2. Meteosat-9 measured Cloud Optical Depth over the 2° × 2° domain surrounding the RF0718 at (a) 0900 UTC; (b) 0930 UTC; (c) 1000 UTC; (d) 1030 UTC. The aircraft horizontal paths within each half hour are overlayed as purple lines with two flight directions: Leg-1 which is parallel to cloud propagation (fly from NW to SE) and Leg-2 which is perpendicular to cloud propagation (fly from SW to NE).
-
Figure 3 shows the vertical distributions of cloud and drizzle microphysical properties during the aircraft legs from the sub-cloud layer to near cloud top. The samples are classified into two categories following their directions: Leg-1 and Leg-2. Note that the upper four legs (from 600 to 1000 m) in Fig. 3 represent the aircraft measurements within the cloud layer, and the bottom legs (~400−500 m) denote the height levels below the cloud base. The
$ {N}_{\mathrm{c}} $ values for Leg-1 are consistently lower than those from Leg-2, specifically, the layer-mean$ {N}_{\mathrm{c}} $ for Leg-1 (~64$ \mathrm{c}{\mathrm{m}}^{-3} $ ) is about 40% less than that from Leg-2 (~ 107$ \mathrm{c}{\mathrm{m}}^{-3} $ ). For both legs, the$ {N}_{\mathrm{c}} $ values increased from the cloud base, maximized in the middle of the cloud, and decreased toward the cloud top (Fig. 3a). The$ {r}_{\mathrm{c}} $ and$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ for the Leg-2 side increased almost linearly from cloud-base to cloud-top (Figs. 3b and 3c), indicating that the cloud sampled at the Leg-2 side is approximately adiabatic. On the other hand, the cloud microphysical properties along the Leg-1 side of the cloud show lower$ {N}_{\mathrm{c}} $ and larger$ {r}_{\mathrm{c}} $ . The layer-mean values of$ {r}_{\mathrm{c}} $ are 11.5 μm and 10.1 μm for Leg-1 and Leg-2, respectively, with the differences in$ {N}_{\mathrm{c}} $ and$ {r}_{\mathrm{c}} $ having passed the 95% significance level. Furthermore, both the$ {N}_{\mathrm{c}} $ and$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ on the Leg-1 side show significant decreases near the cloud top. Note that the dashed lines in Fig. 3c denote the adiabatic LWC, which is calculated by$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{a}\mathrm{d}\left(z\right)}={\Gamma }_{\mathrm{a}\mathrm{d}}(z-{z}_{\mathrm{b}}) $ on the two sides, where$ {z}_{\mathrm{b}} $ is cloud base and$ {\Gamma }_{\mathrm{a}\mathrm{d}} $ denotes the linear increase of LWC with height under an ideal adiabatic condition (Wood, 2005). The cloud adiabaticity ($ {f}_{\mathrm{a}\mathrm{d}} $ ) is defined as the ratio of$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ to$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{a}\mathrm{d}} $ . The layer mean$ {f}_{\mathrm{a}\mathrm{d}} $ is 0.56 for Leg-1 and is 0.83 for Leg-2. The difference in$ {f}_{\mathrm{a}\mathrm{d}} $ indicates that the stratocumulus cloud for Leg-1 undergoes more sub-adiabatic processes, while the cloud layer for Leg-2 is closer to adiabatic. These results suggest that the cloud adiabaticities are different even within the same stratocumulus cloud deck. Such characteristics of inhomogeneous mixing on a small scale were also found in previous studies on the MBL stratocumulus (e.g., Pawlowska et al., 2000; Haman et al., 2007).Figure 3. Vertical profiles of (a) cloud-droplet number concentrations (
$ {N}_{\mathrm{c}} $ ); (b) cloud-droplet effective radii ($ {r}_{\mathrm{c}} $ ); (c) cloud liquid water content ($ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ ) with dashed lines denoting adiabatic LWC; (d) drizzle-drop number concentration ($ {N}_{\mathrm{d}} $ ); (e) drizzle mass median diameter ($ {D}_{\mathrm{m},\mathrm{d}} $ ) and (f) drizzle liquid water content ($ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{d}} $ ). Blue denotes sampling on Leg-1 side on L-shaped leg, and red denotes sampling on Leg-2 side on L-shaped leg. Dots show the mean values at each level, and the vertical bars from left to right represent 10%, 25%, 50%, 75%, and 90% values.Theoretically, the cloud sub-adiabaticity in the marine stratocumulus is often induced by the cloud droplet collision-coalescence and the cloud-top dry air entrainment processes (Wood, 2012; Wu et al., 2020b; Zheng et al., 2022). Notice that
$ {r}_{\mathrm{c}} $ values near the cloud top did not decrease like its$ {N}_{\mathrm{c}} $ and$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ counterparts because most of the small cloud droplets were either evaporated by the entrained dry air, or enlarged by the cloud droplet condensational growth and collision-coalescence, which means the cloud droplet is enlarged, but not large enough to be classified as drizzle droplet, so such large cloud droplet would contribute to$ {r}_{\mathrm{c}} $ , but not for$ {D}_{\mathrm{m},\mathrm{d}} $ . This argument is supported by the distributions of$ {N}_{\mathrm{c}} $ and$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ , which both show longer tails toward higher values. Note that the cloud microphysical properties profiles examined in this study are similar to the results of the MBL stratocumulus clouds over the eastern South Atlantic, where$ {r}_{\mathrm{c}} $ increases with height, and$ {N}_{\mathrm{c}} $ and$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ decrease with height in the upper part of cloud due to cloud-top entrainment (Diamond et al., 2018; Gupta et al., 2022).To reveal the causality of such cloud-top reductions of
$ {N}_{\mathrm{c}} $ ,$ {r}_{\mathrm{c}} $ , and$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ on the Leg-1 side, we first test the hypothesis of cloud-top entrainment because it typically causes the evaporation of cloud water and results in the elimination of small cloud droplets and shrinkage of large cloud droplets. The entrainment rates for the near cloud-top legs between Leg-1 and Leg-2 are compared following the method described in Albrecht et al. (2016). The cloud-top entrainment rate ($ {w}_{\mathrm{e}} $ ) within the stratocumulus can be estimated by:where the
$ {\sigma }_{\mathrm{w}} $ is the standard deviation of vertical velocities near cloud top legs, and$ {A}_{\mathrm{\sigma }} $ is the coefficient associated with the dissipation of the turbulence kinetic energy budget, which was empirically estimated in the study (taken as a value of 26 in this study as per Albrecht et al., 2016). The buoyancy Richardson number$ {R}_{\mathrm{i}\mathrm{\sigma }} $ at the near cloud-top leg can be calculated by:where
$ {\theta }_{0} $ is the reference potential temperature,$ \mathrm{\Delta }{\theta }_{\mathrm{v}} $ is the difference in virtual potential temperature across the MBL-top temperature inversion layer,$ h $ is the MBL depth, and$ {\sigma }_{\mathrm{w}}^{2} $ is the variance of vertical velocities. During RF0718, given the$ {\sigma }_{\mathrm{w}}^{2} $ of 0.115${\mathrm{m}}^{2}\;{\mathrm{s}}^{-2}$ (for Leg-1) and 0.103${\mathrm{m}}^{2}\;{\mathrm{s}}^{-2}$ (for Leg-2), the cloud-top entrainment rates are estimated to be 0.479 ± 0.063$\mathrm{c}\mathrm{m}\;{\mathrm{s}}^{-1}$ and 0.380 ± 0.055$ \mathrm{c}\mathrm{m}\;{\mathrm{s}}^{-1} $ for Leg-1 and Leg-2, respectively. The uncertainties of entrainment rates are estimated by propagating the uncertainties of$ \mathrm{\Delta }{\theta }_{\mathrm{v}} $ and$ h $ for the two sides. The difference in entrainment rates indicates the inhomogeneous mixing at the small scale, which likely is due to the small-scale variations of the thermodynamic conditions and the near-cloud-top turbulence (Haman et al., 2007; Hill et al., 2009; Lehmann et al., 2009; Gao et al., 2020). The entrainment rates fit in the general range (0.1–0.8 cm s−1) for marine stratocumulus (Nicholls and Turton, 1986; Faloona et al., 2005; Bretherton and Blossey, 2014).The results suggest that the cloud layer on the Leg-1 side of the cloud was undergoing ~25.9% stronger dry air entrainment than that on the Leg-2 side of the cloud based on their mean values. Based on entrainment theory only, the Leg-1 side of cloud-top would have lower
$ {N}_{\mathrm{c}} $ and$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ . However, on average, the cloud-top Leg-1$ {N}_{\mathrm{c}} $ (28$ \mathrm{c}{\mathrm{m}}^{-3} $ ) is ~71% less than Leg-2$ {N}_{\mathrm{c}} $ (95$ \text{c}{\mathrm{m}}^{-3} $ ), and Leg-1$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ (0.293$ \mathrm{g}{\mathrm{m}}^{-3} $ ) is ~57% less than Leg-2$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{c}} $ (0.682$ \mathrm{g}{\mathrm{m}}^{-3} $ ). Thus, such significant differences in cloud-top microphysics cannot be solely explained by the differences in entrainment rates and are more likely also the result of the combination of the entrainment effect and the collision-coalescence processes associated with the drizzle formation and development.The vertical profiles of drizzle drop number concentration (
$ {N}_{\mathrm{d}} $ ), mass median diameter ($ {D}_{\mathrm{m},\mathrm{d}} $ ), and liquid water content ($ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{d}} $ ) from below the cloud to the cloud-top are shown in Figs. 3d-3f. Both Leg-1 and Leg-2 sides show existing drizzle drops throughout the cloud-layer and below the cloud. From the cloud-top leg, the$ {N}_{\mathrm{d}} $ (0.58$ \mathrm{c}{\mathrm{m}}^{-3} $ ) and$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{d}} $ (0.233$ \mathrm{g}{\mathrm{m}}^{-3} $ ) on the Leg-1 side are much higher than the$ {N}_{\mathrm{d}} $ (0.12$ \mathrm{c}{\mathrm{m}}^{-3} $ ) and$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{d}} $ (0.087$ \mathrm{g}{\mathrm{m}}^{-3} $ ) on the Leg-2 side, whereas smaller$ {D}_{\mathrm{m},\mathrm{d}} $ (64.9 μm) is observed at Leg-1 than Leg-2 (93.4 μm). The mean ratios of$ {N}_{\mathrm{d}} $ to$ {N}_{\mathrm{c}} $ are 2% and 0.1%, respectively, for Leg-1 and Leg-2, suggesting that a much more effective collision-coalescence process occurred near the cloud top in the Leg-1 side of the cloud. The$ {r}_{\mathrm{c}} $ and$ {N}_{\mathrm{d}} $ distributions are both broader on the Leg-1 side, and their mean values are both larger than those on the Leg-2 side. These results suggest that more large cloud droplets near the cloud top experienced condensational growth and coalescence processes to form drizzle drops on the Leg-1 side, which results in more of the smaller drizzle drops (higher$ {N}_{\mathrm{d}} $ , smaller$ {D}_{\mathrm{m},\mathrm{d}} $ , more drizzle embryos) near the cloud top. Similar results were also found in other studies (e.g., Wood, 2006; Wu et al., 2020b; Dong et al., 2021). The enhanced drizzle formation process results in more drizzle embryos formed near the cloud top through the collision-coalescence processes.All four in-cloud aircraft legs show that the
$ {N}_{\mathrm{d}} $ and$ \mathrm{L}\mathrm{W}{\mathrm{C}}_{\mathrm{d}} $ on the Leg-1 side are higher than those on the Leg-2 side. The$ {D}_{\mathrm{m},\mathrm{d}} $ at Leg-1 are generally larger than the Leg-2 side with broader ranges of distributions, except near the cloud top. From the cloud top to the cloud base, the mean$ {N}_{\mathrm{d}} $ decreased ~94% (from 0.58 to 0.033$ \mathrm{c}{\mathrm{m}}^{-3} $ ) on the Leg-1 side, while the depletion of$ {N}_{\mathrm{d}} $ on the Leg-2 side is ~89% (from 0.12 to 0.013$ \mathrm{c}{\mathrm{m}}^{-3} $ ). The mean$ {D}_{\mathrm{m},\mathrm{d}} $ values on the Leg-1 side were consistently enlarged with a more than 137% increment at the cloud base relative to cloud top. While the$ {D}_{\mathrm{m},\mathrm{d}} $ on the Leg-2 side showed little variation from the cloud top to cloud center, the total increment at the cloud base is only ~41%. The higher$ {N}_{\mathrm{d}} $ depletion rate and larger$ {D}_{\mathrm{m},\mathrm{d}} $ growth rate for Leg-1 have demonstrated a more efficient self-collection process inside the cloud layer, especially in the lower part of the cloud. When drizzle drops fall from the cloud top, they grow by collecting cloud droplets and other drizzle drops, which shows that the collision-coalescence and drizzle self-collection processes become increasingly important and result in depleted$ {N}_{\mathrm{d}} $ and increased$ {D}_{\mathrm{m},\mathrm{d}} $ near the cloud base. Therefore, the enhanced drizzle processes on Leg-1 side are consistent with the larger CODs retrieved from the satellite on Leg-1 side because COD is proportional to the square of$ {r}_{\mathrm{c}} $ . -
Results and discussions from the last section show that the cloud on the Leg-1 side experienced a more effective collision-coalescence process. More cloud droplets were converted to drizzle via the collision-coalescence processes, and efficiently grew to larger drizzle drops near the cloud base. The turbulence on Legs -1 and -2 were examined to further investigate a possible mechanism for the enhanced collision-coalescence characteristic. The gradient Richardson Number (
$ {R}_{i} $ ) is often used to represent the thermal stability and turbulent condition of the environment (Garratt, 1994), which can be calculated from:The numerator represents the buoyancy production term of turbulence which is given by the change of virtual potential temperature across the atmospheric layer, and the denominator represents the shear production term of turbulence given by the change of horizontal wind components across the same layer. A positive
$ {R}_{i} $ denotes a stable atmospheric layer while a negative$ {R}_{i} $ denotes an unstable environment. In this study, the$ {R}_{i} $ for both Leg-1 and Leg-2 were estimated using a finite difference method for each aircraft horizontal leg in the cloud layer. The variables were interpolated to ensure comparable values of$ \partial z $ between the two sides. The$ {R}_{i} $ values, buoyancy and shear production terms of$ {R}_{i} $ , as well as the vertical velocity variances and absolute wind shears are listed in Table 1. As shown in column 8 of Table 1, for all four aircraft legs in the cloud, the$ {R}_{i} $ values on both Leg-1 and Leg-2 are negative and significantly lower than the critical value of 0.25, indicating that the MBL was unstable and turbulent.Leg Altitude Level Leg Sets Vertical
Velocity
Variance (${\mathrm{m} }^{2}\;{\mathrm{s} }^{-2}$)Wind Speed
Shear ($\mathrm{m}{\mathrm{s} }^{-1}\;{\mathrm{m} }^{-1}$)Wind Directional Shear
($\mathrm{d}\mathrm{e}\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}\;{\mathrm{m} }^{-1}$)Buoyancy Production
($ {\mathrm{s}}^{-2} $)Shear
Production
($ {\mathrm{s}}^{-2} $)Gradient
Richardson
Number, ${{R} }_{\mathrm{i} }$Near Cloud-Top Leg-1 0.115 0.0054 0.1137 −0.528 0.259 −2.03 Leg-2 0.103 0.0105 0.0607 −0.506 0.216 −2.34 Upper Mid-Cloud Leg-1 0.243 0.0142 0.0901 −0.678 0.320 −2.12 Leg-2 0.302 0.0082 0.0823 −0.665 0.188 −3.54 Lower Mid-Cloud Leg-1 0.260 0.0095 0.0582 −0.421 0.125 −3.37 Leg-2 0.222 0.0061 0.0514 −0.425 0.089 −4.74 Near Cloud-Base Leg-1 0.137 0.0096 0.0971 −0.557 0.240 −2.32 Leg-2 0.201 0.0065 0.0948 −0.543 0.178 −3.07 Table 1. Gradient Richardson number, Production terms, Vertical velocity variance, and Wind shears for PARL & PREP
Previous studies have shown that in-cloud turbulence can effectively enhance the drizzle production and growth processes by the turbulence-induced collision-coalescence processes (Feingold et al., 1996b; Pinsky et al., 2007). The negative
$ {R}_{i} $ on both sides indicate that cloud environments were favorable to form drizzle, which is also suggested by the comparable vertical velocity variances (column 3 of Table 1) on both sides. However, the differences in the drizzle microphysics between the two sides can be better explained by breaking down the$ {R}_{i} $ into separate buoyancy (column 6 of Table 1) and shear (column 7 of Table 1) terms. The buoyancy terms are negative for all legs, indicating positive buoyancy production of turbulence (since$ \partial {\theta }_{\mathrm{v}}/\partial z < 0 $ ), and the differences between both sides are small. The absolute values of the buoyancy term are always much larger than the shear term, which is common because buoyancy production is the primary contributor to the turbulence in MBL clouds (Nicholls, 1984; Nicholls and Leighton, 1986; Bretherton and Wyant, 1997).However, previous studies found that the shear production also plays an important role in the turbulence generation and drizzle evolution, especially in the marine stratocumulus-topped boundary layer (Brost et al., 1982; Magaritz-Ronen et al., 2016; Wu et al., 2017). The shear terms on the Leg-1 side are noticeably larger and have higher contribution percentages (~23%–32%) in the turbulence production than the Leg-2 side (~17%–29%). Such differences in the shear production terms are also supported by the generally stronger wind speed shear (column 4 of Table 1) and wind directional shear (column 5 of Table 1) on the Leg-1 side. On the one hand, the relatively stronger wind shear on the Leg-1 side, especially the upper part of the cloud, could promote drizzle formation by increasing the residence time of large cloud droplets and thus the chance of collision-coalescence among cloud droplets and/or drizzle drops (Magaritz-Ronen et al., 2016). On the other hand, strong wind shear effectively recirculates the drizzle drops in the middle and the lower part of the cloud, allowing them to grow larger by collecting smaller drizzle drops and large cloud droplets (Feingold et al., 1996b; Magaritz et al., 2009; Wu et al., 2017). Therefore, the strong wind shear enhances the efficiency of collision-coalescence processes, and results in the stronger drizzle evolution process on the Leg-1 side of the cloud. In contrast, the turbulence on the Leg-2 side is primarily driven by buoyancy, which is favorable to the updraft-dominant environment. Though buoyancy stimulates cloud-top drizzle formation, the lack of strong wind shear production limits the in-cloud efficiency of collision-coalescence and the recirculation of drizzle drops, which can explain the small variation of drizzle microphysical properties in the middle of the cloud.
Leg Altitude Level | Leg Sets | Vertical Velocity Variance (${\mathrm{m} }^{2}\;{\mathrm{s} }^{-2}$) | Wind Speed Shear ($\mathrm{m}{\mathrm{s} }^{-1}\;{\mathrm{m} }^{-1}$) | Wind Directional Shear ($\mathrm{d}\mathrm{e}\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}\;{\mathrm{m} }^{-1}$) | Buoyancy Production ($ {\mathrm{s}}^{-2} $) | Shear Production ($ {\mathrm{s}}^{-2} $) | Gradient Richardson Number, ${{R} }_{\mathrm{i} }$ |
Near Cloud-Top | Leg-1 | 0.115 | 0.0054 | 0.1137 | −0.528 | 0.259 | −2.03 |
Leg-2 | 0.103 | 0.0105 | 0.0607 | −0.506 | 0.216 | −2.34 | |
Upper Mid-Cloud | Leg-1 | 0.243 | 0.0142 | 0.0901 | −0.678 | 0.320 | −2.12 |
Leg-2 | 0.302 | 0.0082 | 0.0823 | −0.665 | 0.188 | −3.54 | |
Lower Mid-Cloud | Leg-1 | 0.260 | 0.0095 | 0.0582 | −0.421 | 0.125 | −3.37 |
Leg-2 | 0.222 | 0.0061 | 0.0514 | −0.425 | 0.089 | −4.74 | |
Near Cloud-Base | Leg-1 | 0.137 | 0.0096 | 0.0971 | −0.557 | 0.240 | −2.32 |
Leg-2 | 0.201 | 0.0065 | 0.0948 | −0.543 | 0.178 | −3.07 |