-
Since 2010, routine observations of aerosols, clouds, and precipitation have been carried out by using several ground-based lidar systems together with a sun photometer at our site on the campus of Wuhan University (Wang et al., 2020; He et al., 2021a, b, c; Yin et al., 2021). In this study, a 532 nm polarization lidar was the main instrument used to observe cirrus clouds and the optical properties of their related dust layers. In addition, a water vapor Raman lidar with a 355-nm laser was employed to derive the water vapor mixing ratio (WVMR) within the cirrus cloud (Wu and Yi, 2017). Combining the lidar-derived WVMR with temperature and pressure data from radiosondes, the relative humidity with respect to water (RHw) and ice (RHi) can be derived.
-
The polarization lidar was the main instrument used to observe the dust aerosols and cirrus clouds. This lidar system began routine operation in 2010 and was already introduced in detail by Kong and Yi (2015). The outgoing laser is 532-nm linearly polarized light with a polarization purity better than 10000:1, pulse energy of ~120 mJ, a pulse width of 6 ns, and a repetition rate of 20 Hz. The receiving unit employs a 300-mm aperture Cassegrain telescope with a field-of-view (FOV) of ~1 mrad to collect the atmospheric backscattering light. Then the received light is separated into parallel and perpendicular polarized components by using double cascaded cubic polarizing beamsplitters enabling sufficient suppression of crosstalk between two receiving channels. Hamamatsu 5783P Photomultiplier tubes (PMTs) were used as the detector with a two-channel Licel TR40-160 as the signal digitizer. The gain ratio between two orthogonally polarized channels is calibrated using the Δ90° method (Freudenthaler et al., 2009). The lidar-derived volume depolarization lidar
$ {\delta }_{\mathrm{v}} $ indicates the nonspherical shape of backscattering objects to effectively distinguish dust particles and ice crystals (Wang et al., 2020) from the liquid droplets and other types of spherical aerosols.The aerosol extinction coefficient
$ \alpha $ and backscatter coefficient$ \beta $ were calculated with the method of Fernald (1984). The lidar ratio was set to 45 sr for the background/dust aerosols (He et al., 2021a). Considering$ {\delta }_{\mathrm{v}} $ value is contributed by both atmospheric molecules and aerosols and thus depends on the ratio of molecular to particulate backscatter, it is necessary to extract the individual contribution of aerosol (i.e., particle depolarization ratio$ {\delta }_{\mathrm{p}} $ ) to depict its microphysical properties independent of the aerosol loadings. The particle depolarization ratio can be calculated using the following equation (Freudenthaler et al., 2009):where the molecular depolarization ratio
$ {\delta }_{\mathrm{m}} $ , which depends on the specification of the narrowband filter used in the receiving channels, is 0.004 for our lidar system and$ R $ is aerosol backscattering ratio. Note that the assumption of constant lidar ratio in the Fernald method would lead to errors in retrieved$ \beta $ , which would be subsequently propagated to$ {\delta }_{\mathrm{p}} $ . Particle depolarization ratio is strongly sensitive to the nonsphericity of particles, while lidar ratio is highly sensitive to the absorption characteristics of particles (Wiegner et al., 2009). Lidar ratio and particle depolarization ratio are controlled by different primary factors; some secondary factors (i.e., particle size) may influence them both and result in errors in retrieved$ {\delta }_{\mathrm{p}} $ when using a single-wavelength polarization lidar.The backscatter coefficient for the dust component
$ {\beta }_{\mathrm{d}} $ can be extracted with the following equation (Tesche et al., 2009; Mamouri and Ansmann, 2014):where the dust particle depolarization ratio
$ {\delta }_{\mathrm{d}}=0.31 $ and the non-dust particle depolarization ratio$ {\delta }_{\mathrm{n}\mathrm{d}}=0.05 $ (Sakai et al., 2010; Mamouri and Ansmann, 2014). The dust backscatter coefficient$ {\beta }_{\mathrm{d}} $ was then multiplied by the dust lidar ratio 45 sr to obtain the dust extinction coefficient$ {\alpha }_{\mathrm{d}} $ , which was then input to the POLIPHON method as described in section 2.4. The typical lidar ratio at 532 nm for Asian dust was independently measured to be 45±7 sr (Hu et al., 2020) and 47±4 sr (Peng et al., 2021) by pure rotational Raman lidar (Ångström-relationship assumption can be ignored theoretically), suggesting the assumed dust lidar ratio (45 sr) is reasonable in this study and leads to acceptable relative uncertainties (as listed in Table 1).Parameter Uncertainty Dust backscatter coefficient $ {\beta }_{\mathrm{d}} $ 10% Dust extinction coefficient $ {\alpha }_{\mathrm{d}} $ 20% Volume depolarization ratio $ {\delta }_{\mathrm{v}} $ 5% Particle depolarization ratio $ {\delta }_{\mathrm{p}} $ 5%–10% Dust mass concentration ${M }_{\mathrm{d} }$ 20%–30% Dust surface area concentration $ {S}_{\mathrm{d}} $ 30%–40% INP number concentration $ {n}_{\mathrm{I}\mathrm{N}\mathrm{P}} $ 50%–500% Table 1. Estimated uncertainties in the dust-related optical parameters and cloud-relevant properties (Kafle and Coulter, 2013; Ansmann et al., 2019b).
-
A Raman lidar system emits a 355 nm laser beam with a power of ~6 W and detects the Raman backscatter from water vapor at 407 nm and nitrogen molecules at 387 nm by using a 200-mm aperture Cassegrain telescope. The WVMR can be derived using the method from Whiteman et al. (1992), with the calibration constant determined by using the WVMR data from radiosondes launched at 0000 UTC or 1200 UTC from the Wuhan Weather Station (~24 km from our site) (Wu and Yi, 2017). To ensure a sufficient signal-to-noise ratio (SNR) at cirrus cloud altitudes, hourly integrated data were applied to derive the WVMR in this study. If the telescope is upgraded with an aperture of 600 mm and the laser is upgraded with a power of 9 W, integrated time can be substantially promoted from one hour to five minutes which is more favorable for realizing the process-level observation of cirrus cloud evolution. A similar scheme was used by Leblanc et al. (2012) for the long-term monitoring of water vapor at the upper troposphere and lower stratosphere.
Combining the lidar-derived WVMR profile with the atmospheric pressure and temperature profiles from radiosonde measurements, RHw and RHi profiles can be calculated (Murphy and Koop, 2005). The related calculations are given below. The relationship between the vapor pressure
$ E $ and WVMR can be given by (Bolton, 1980):where r denotes the WVMR measured by Raman lidar and
$ P $ is the atmospheric pressure provided by radiosonde measurement. The saturated watervapor pressure with respect to liquid water $ {E}_{\mathrm{w}} $ and ice water$ {E}_{\mathrm{i}} $ can be calculated, respectively, using the empirical formulas as below (Murray et al., 1967; Bolton, 1980):where T is the temperature provided by radiosonde measurement. Therefore, RHw and RHi can be obtained, respectively, by the following.
-
ERA5 is the fifth-generation atmospheric reanalysis product produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) and was released publicly in 2016 (Hersbach et al., 2020). It can provide hourly datasets of atmospheric parameters (e.g., relative humidity, temperature, three components of wind speed, horizontal divergence, fraction of cloud cover, ozone mass mixing ratio, and cloud water content) on 37 pressure levels (from 1000 hPa to 1 hPa) with an enhanced horizontal resolution of 31 km (0.25°) from 1979 onwards. In this study, the grid cell containing Wuhan (30.5°N, 114.4°E) was selected to provide the necessary meteorological parameters, including RHw, temperature (T), the eastward (U) and northward (V) components of wind speed, and vertical velocity (W). Benefiting from the one-hour temporal resolution, ERA5 data can cover the time intervals between the launches of the radiosondes (12 h). ERA5 data are employed to provide the meteorological parameters in this study. Due to the insufficient vertical resolution of ERA5, linear interpolation was conducted on the temperature profile to obtain the required cloud top temperature.
A single cirrus cloud usually exists in lidar FOVs for only tens of minutes, which is less than the one-hour temporal resolution of ERA5. Therefore, it is assumed in this work that the one-hour meteorological parameters from ERA5 that overlap in time with the presence of a cirrus cloud are representative of the meteorological conditions during the entire lifetime of the cirrus cloud. For the water vapor condition (RH), the influences of waves or turbulence on the variation of RH values (Kärcher et al., 2014) over the short time scale are not considered. Since most of the cases lasted for more than one hour, attributing to the presence of a series of adjacent cirrus clouds, it is substantially reasonable to employ one-hour RH as the vapor condition during the cloud lifetime. In addition, in-cloud peak RH is defined as the maximum RH value within the vertical extension of the cirrus cloud.
-
The dust mass concentration
$ {M}_{\mathrm{d}} $ , dust surface area concentration$ {S}_{\mathrm{d}} $ , and dust-related INP concentration$ {n}_{\mathrm{I}\mathrm{N}\mathrm{P}} $ can be retrieved by the POLIPHON method (Mamouri and Ansmann, 2014, 2015; Ansmann et al., 2019a; Marinou et al., 2019). Dust particles pertinent to cirrus clouds appear in the upper troposphere and hence are less influenced by anthropogenic aerosols near the surface. Therefore, we used the pure-dust conversion factors (${c}_{\mathrm{v},\mathrm{ }\mathrm{d}}$ and${c}_{\mathrm{s},\mathrm{ }\mathrm{d}}$ ) obtained at the Lanzhou SACOL site by Ansmann et al. (2019b). All the uncertainties for the derived parameters are listed in Table 1. The applied conversion factors and constants are given in Table 2.Parameter value Unit Reference $ {c}_{\mathrm{v},\mathrm{d}} $ 0.77×10−12 Mm (Ansmann et al., 2019b) $ {c}_{\mathrm{s},\mathrm{d}} $ 3.1×10−12 Mm m2 cm−3 (Ansmann et al., 2019b) Dust lidar ratio LR 45 sr (He et al., 2021a) Dust density $ {\mathrm{\rho }}_{\mathrm{d}} $ 2.6 g cm−3 (Wagner et al., 2009) Table 2. Applied values of the conversion parameters and constants required in the POLIPHON retrieval.
The specific calculation processes of dust mass concentration
$ {M}_{\mathrm{d}} $ and dust-related INP concentration$ {n}_{\mathrm{I}\mathrm{N}\mathrm{P}} $ are given below. The dust mass concentration$ {M}_{\mathrm{d}} $ is retrieved using the following equation:where
$ {\rho }_{\mathrm{d}} $ is the dust particle density,$ {\alpha }_{\mathrm{d}} $ is the dust extinction coefficient (see section 2.1), and$ {c}_{\mathrm{v},\mathrm{d}} $ is the extinction-to-volume conversion factor. At the cirrus cloud altitudes, we only consider INP concentration for deposition nucleation mode and pore condensation and freezing using the parameterization scheme U17-D (Ullrich et al., 2017). The ice nucleation active surface site (INAS) density$ {n}_{\mathrm{s}} $ can be retrieved with the following equation:where
$ {S}_{\mathrm{i}} $ is ice saturation ratio and$ T $ is the temperature, both obtained from the measurement of the most recently launched radiosonde. The surface area concentration of dust$ {S}_{\mathrm{d}} $ can be obtained by:where
$ {c}_{\mathrm{s},\mathrm{d}} $ is the extinction-to-surface area conversion factor (Ansmann et al., 2019b). Finally, we can calculate the INPC for deposition freezing mode$ {n}_{\mathrm{I}\mathrm{N}\mathrm{P}} $ with the following equation (Marinou et al., 2019): -
From December 2012 to February 2018, 22 cases of cirrus clouds related to transported dust layers were identified based on polarization lidar observations over Wuhan, China. Water vapor Raman lidar began to operate in November 2017, thus lidar-derived WVMR and RH data were available for only a few cases. The criteria for cirrus clouds were described in detail by Wang et al., (2020). The criteria for an aloft dust layer are as follows: (i)
$ {\delta }_{\mathrm{p}} $ value is larger than 0.1 throughout the layer; (ii) layer thickness exceeds 0.3 km; (iii) layer base is above the planetary boundary layer (He et al., 2021a). As long as any edge (top or base) of a cirrus cloud layer was in contact with an edge of a dust layer for >10 min, we determine that the formation or evolution of this cirrus cloud layer was affected by the dust layer (i.e., the cirrus cloud is related to the dust layer). Note that a single bulk cloud or a series of very adjacent clouds near to the same dust layer would be considered a single case. Finally, all the selected cases were checked individually by eye. In the following, we first provide two typical case studies (both with lidar-derived WVMR, temperature, and RH data available) and then some brief statistics. -
Dust aerosols were observed to intrude into Wuhan on 12 January 2018. Figures 1b and 1d show the time-height contour plots (1-min/30-m resolution) of the range-corrected signal and volume depolarization ratio
$ {\delta }_{\mathrm{v}} $ during 2030–2400 LST (Local Standard Time = UTC + 8 h) measured by the polarization lidar. The corresponding height profiles of the RHw and RHi, temperature, horizontal wind speed, and vertical velocity obtained from ERA5 reanalysis data during 2200–2300 LST are shown in Figs. 1a and 1c. Two dust aerosol layers were identified by the volume depolarization ratio (Fig. 1d). The lower layer located from the surface to approximately 3.6 km showed relatively small$ {\delta }_{\mathrm{v}} $ values of < 0.1, while the upper layer that appeared at altitudes of 7–9 km had larger$ {\delta }_{\mathrm{v}} $ values up to 0.16.Figure 1. Time-height contour plots (1-min/30-m resolution) of (b) range-corrected signal and (d) volume depolarization ratio measured by a 532 nm polarization lidar on 12 January 2018 during 2030–2400 LST. Height profiles of (a) temperature and relative humidity (RH with respect to water and ice respectively), and (c) wind speed of the U (westerly), V (southerly), and W (vertical velocity) components obtained from ECMWF ERA5 reanalysis data during 2200–2300 LST. The horizontal dashed lines indicate cloud top level with a temperature of –47.8°C. The vertical dotted line indicates 100% relative humidity. Several cirrus clouds (8.6–9.3 km) appeared on the top of the dust layer.
As seen from plots of the volume depolarization ratio, three optically transparent cirrus clouds, with
$ {\delta }_{\mathrm{v}} $ values larger than 0.2, began to appear on top of the dust layer (above ~8.7 km) at ~2100 LST. Each cirrus cloud existed for less than half an hour, which is very different from stratiform midlevel cloud layers that typically have longer durations (He et al., 2021a). The cloud top temperature was –47.8°C as denoted by the horizontal dashed lines. The base of the first cirrus cloud (2100–2130 LST) was embedded in the dust layer below, hinting that the ice crystals within the cloud were probably related to heterogeneous nucleation triggered by dust particles. In general, only heterogeneous nucleation can take place at temperatures ranging from –36°C to 0°C. DeMott et al. (2003) found that homogeneous freezing requires a somewhat higher ice supersaturation as temperature decreases. Therefore, even at temperatures below –36°C, homogeneous nucleation cannot be initiated unless the RHi is strongly supersaturated up to 150% (Koop et al., 2000; Ansmann et al., 2019a). To confirm the specific nucleating regime in this case, the RHi condition within the cirrus cloud needs to be checked further.Figure 2 shows the profiles of RHw, RHi, WVMR, saturated WVMR, and temperature derived by the combination of water vapor Raman lidar and radiosonde measurements during 2200–2300 LST. As a comparison, the simultaneous RHw and RHi profiles obtained from the observation of water vapor Raman lidar are also given here. As seen in Fig. 2a, the RHw and RHi values from lidar measurements and ERA5 data agree well with each other at altitudes below 8.0 km. Above 8.0 km, both lidar measurements and ERA5 data show an enhancement of RH (both RHw and RHi), which is attributed to the occurrence of cirrus clouds. The RH values from lidar measurements are consistently smaller than those from ERA5 data. Nevertheless, the peak values of RHi provided by the two methods both exceed 150% (i.e., 183.2% from lidar and 157.4% from ERA5) at the altitudes where clouds are located (8.6–9.8 km), indicating that homogeneous nucleation may also participate in ice formation. Consequently, we can conclude that the ice-nucleating pattern here is the competition between heterogeneous and homogeneous nucleation. This situation was found only three times during the whole observation period, as seen in subsection 3.3. In addition, it is found that the lidar-derived peak RHi during 2000–2100 LST, before cirrus clouds appeared in the lidar FOV, is as high as 133.5% (not shown here). This provides a great RHi condition for the possible initiation of subsequent homogeneous nucleation. As a result of competition, the homogeneous nucleation process is generally suppressed because heterogeneous nucleation advances the onset of ice nucleation and consumes the available water vapor (Kärcher and Lohmann, 2003; Barahona and Nenes, 2009). Based on an investigation using the Community Atmospheric Model version 5 (CAM5), Liu et al. (2012) reported that dust-related heterogeneous nucleation reduces the occurrence of homogeneous nucleation and thus the ice crystal number concentration for upper atmospheric cirrus clouds in the Northern Hemisphere.
Figure 2. Profiles of (a) relative humidity (RH with respect to water and ice) derived by lidar and ERA5 reanalysis data and (b) water vapor mixing ratio and saturated water vapor mixing ratio derived by lidar during 2100–2200 LST, and (c) temperature measured by radiosonde launched at 2000 LST on 12 January 2018.
The dust-related INPC, which determines the significance of heterogeneous nucleation (Kärcher and Lohmann, 2003), can be obtained quantitatively by the POLIPHON method (Mamouri and Ansmann, 2014, 2015). Figure 3 shows profiles of the total (dust + non-dust) and dust extinction coefficient, total (dust + non-dust) and dust backscatter coefficient, volume depolarization ratio
$ {\delta }_{\mathrm{v}} $ , and particle depolarization ratio$ {\delta }_{\mathrm{p}} $ for a cloud-free period during 2000–2030 LST measured by polarization lidar. Considering that the vertical distribution of the dust layer at 7–9 km was uniform over time, we assume that the dust optical properties during these 30 minutes are representative of the entire event. As seen in Fig. 4a, dust extinction predominantly contributes to the total aerosol extinction at altitudes of 7.0–9.8 km with a mean of 10.4 Mm−1, indicating the dominance of pure dust particles within this layer. Moreover, the presence of a pure dust layer is also verified by$ {\delta }_{\mathrm{p}} $ values larger than 0.3 throughout the layer, as shown in Fig. 3d (Sakai et al., 2010).Figure 3. Profiles of (a) dust and total (dust + non-dust) extinction coefficient, (b) dust and total (dust + non-dust) backscatter coefficient, (c) volume depolarization ratio
$ {\delta }_{\mathrm{v}} $ , and (d) particle depolarization ratio$ {\delta }_{\mathrm{p}} $ derived by the 532 nm polarization lidar during 2000–2030 LST on 12 January 2018. Horizontal error bars denote the relative errors of each parameter.Figure 4. Profiles of (a) dust mass concentration
$ {M}_{\mathrm{d}} $ , (b) dust surface area concentration$ {S}_{\mathrm{d}} $ , and (c) ice-nucleating particle concentration derived by the POLIPHON method using the U17-D parameterization scheme during 2000–2030 LST on 12 January 2018. The horizontal dashed lines denote the temperature height levels of –35°C and –47.8°C (cloud top temperature).$ {S}_{\mathrm{i}} $ represents ice saturation ratio.Figures 4a and 4b show the concentration profiles of the dust mass and surface area during the same period as Fig. 3. The layer-averaged dust mass concentration is 20.8 μg m−3 with a peak of 49.5 μg m−3. The layer-averaged surface area concentration is 3.2 × 10−11 m2 cm−3. Considering that the dust layer mostly appeared above the –35°C isotherm and liquid water was not observed before cirrus formation, deposition nucleation and pore condensation and freezing mechanism (PCF) should be exclusively responsible for the ice crystal formation, rather than immersion/condensation and contact nucleation (Marcolli, 2014; Campbell and Christenson, 2018; David et al., 2019). In Fig. 4c, the mean (maximum) INPCs within the dust layer are estimated to be 0.02 L−1 (0.1 L−1), 0.47 L−1 (2.3 L−1), and 4.7 L−1 (27.0 L−1) for three assumed constant ice saturation ratio (
$ {S}_{\mathrm{i}} $ ) values of 1.15, 1.25, and 1.35, respectively, based on the U17-D parameterization scheme (Ullrich et al., 2017). Note that Ullrich et al. (2017) also mentioned PCF is the responsible mechanism for part of their parametrizations, especially in the temperature range between –43°C and –23°C. The maximum INPC values in this case are nearly on the same order of magnitude as those of 0.23 L−1, 0.94 L−1, and 9.3 L−1 (respectively corresponding to$ {S}_{i} $ values of 1.15, 1.25, and 1.35) observed in a Saharan-dust-related thin cirrus case over Limassol (34.7°N, 33.0°E), Cyprus, which was generally consistent with the ice crystal concentrations in the adjacent regions (Ansmann et al., 2019a). The INPC profile calculated with the$ {S}_{\mathrm{i}} $ profile (light blue in Fig. 2a) measured by water vapor Raman lidar is shown by the gray line in Fig. 4c. Attributing to the significant enhancement of RHi at altitudes of 9–10 km, the actual INPC around the cloud top reached an order of magnitude of 103–104 L−1. -
Another dust-intrusion event was captured by ground-based lidars over Wuhan on 7 February 2018. Figures 5b and 5d show the time-height contour plots (1-min/30-m resolution) of the range-corrected signal and volume depolarization ratio
$ {\delta }_{\mathrm{v}} $ during 0000–0200 LST measured by polarization lidar. The corresponding height profiles of the RHw, RHi, temperature, horizontal wind speed, and vertical velocity obtained from ERA5 reanalysis data during 0000–0100 LST are shown in Figs. 5a and 5c. Similar to the first case, double-layer dust plumes appeared at altitudes of 0–4.8 km and 7–10 km, respectively, as seen from the volume depolarization ratio of approximately 0.1 (Fig. 5d).Figure 5. Time-height contour plots (1 min by 30 m resolution) of (b) range-corrected signal and (d) volume depolarization ratio measured by a 532 nm polarization lidar for 0000–0200 LST 7 February 2018. Height profiles of (a) temperature and relative humidity (RH with respect to water and ice, respectively), and (c) wind speed of the U (westerly), V (southerly), and W (vertical velocity) components obtained from ECMWF ERA5 reanalysis data during 0000–0100 LST. The horizontal dashed lines indicate cloud top level with a temperature of –51.9°C. The vertical dotted line indicates 100% relative humidity. Several cirrus clouds (9.0–10.1 km) appeared embedded in the upper part of the dust layer.
Above 9 km, several short-lived cirrus clouds were observed to embed in the upper dust layer (7–10 km) since 0012 LST (Fig. 5d). The cloud top temperature was –51.9°C, as denoted by the horizontal dashed lines. Within approximately 80 minutes, each cloud top overlapped with the top edges of the dust layer. Therefore, the ice formation within these cirrus clouds is considered to be relevant to the dust-related heterogeneous nucleation. To verify whether homogeneous nucleation was involved, the RHi profiles during 0000–0100 LST obtained from both lidar detection and ERA5 reanalysis data are given in Fig. 6a. The corresponding profiles of the water vapor mixing ratio from water vapor Raman lidar and temperature from radiosonde are shown in Figs. 6b and 6c. The simultaneous RHi values from lidar measurements and ERA5 data are in accordance with each other well below 7.5 km. An RHi enhancement appeared above 8 km, reflecting the presence of cirrus clouds. Different from the first case, the peak ice saturation ratio (i.e., 124.2% from lidar and 133.9% from ERA5) at cloud altitudes (9–10 km) does not exceed 150%, which is insufficient to initiate homogeneous nucleation, indicating that only heterogeneous nucleation is likely to be responsible for in-cloud ice formation. Before the cirrus clouds appeared in the lidar FOV, the lidar-derived peak RHi during 2300–2400 LST on 6 February was only 74.2% (not shown here), which is much less than that of 133.5%, as observed in the first case, indicating that the RHi condition for the subsequent occurrence of homogeneous nucleation is insufficient. In addition, the invariably low vertical velocity (approximately 0 m s−1, see Fig. 5c) throughout the vertical extension of the cloud region also supports the probable absence of homogeneous nucleation, since a strong vertical velocity, such as >1 m s−1, is essential to the onset of homogeneous freezing (Spice et al., 1999). However, there is still a possibility that homogeneous nucleation had previously occurred and was subsequently inhibited by concomitant heterogeneous nucleation with the consumption of water vapor and thus the decreased RHi to be <150%. This situation was frequently observed during the whole observation period, as discussed in subsection 3.3.
Figure 6. Profiles of (a) relative humidity (RH with respect to water and ice) derived by lidar and ERA5 reanalysis data and (b) water vapor mixing ratio and saturated water vapor mixing ratio derived by lidar during 0000–0100 LST on 7 February 2018, and (c) temperature profile measured by radiosonde launched at 2000 LST on 6 February 2018.
Figure 7 shows profiles of the total (dust + non-dust) and dust extinction coefficient, total (dust + non-dust) and dust backscatter coefficient, volume depolarization ratio, and particle depolarization ratio for a cloud-free period during 0140–0150 LST measured by polarization lidar. It is assumed that this period depicts the representative optical properties of dust aerosols for this case. As seen in Fig. 7a, dust extinction contributes most of the total aerosol extinction at altitudes of 7–10 km, with a mean of 6.1 Mm−1. The
$ {\delta }_{\mathrm{p}} $ values invariably exceed 0.3 above 7.7 km, suggesting that the upper part of the dust layer is composed of pure dust particles, while$ {\delta }_{\mathrm{p}} $ values decrease to 0.2–0.3 at 7–7.8 km (see Fig. 7d), revealing that mixed dust (i.e., dust particles mixed with other urban aerosols that usually show a$ {\delta }_{\mathrm{p}} $ value of <0.3) is dominant in the lower part of the dust layer (Mamouri and Ansmann, 2014, 2015; He et al., 2021b; Yin et al., 2021).Figure 7. Profiles of (a) dust and total (dust + non-dust) extinction coefficient, (b) dust and total (dust + non-dust) backscatter coefficient, (c) volume depolarization ratio
$ {\delta }_{\mathrm{v}} $ , and (d) particle depolarization ratio$ {\delta }_{\mathrm{p}} $ derived by the 532 nm polarization lidar during 2000–2030 LST on 7 February 2018. Horizontal error bars denote the relative errors of each parameter.Figure 8 shows the concentration profiles of the dust mass, surface area, and ice nucleating particles during the same period as Fig. 9. The layer-averaged dust mass concentration is 12.3 μg m−3, with a peak of 17.5 μg m−3 appearing at ~8.6 km. The layer-averaged surface area concentrations are 1.6 × 10−11 m2 cm−3. For the study of heterogeneous nucleation within cirrus clouds, the possibility of immersion/condensation freezing and contact freezing can be completely excluded because liquid water was not observed previously, as seen from the volume depolarization ratio in Fig. 5b (Hoffmann et al., 2013). At cirrus cloud altitudes (9–10 km), dust particles probably initiate ice nucleation through deposition freezing (Ansmann et al., 2019a). Note that preactivation can be another process responsible for ice crystal formation here, because pore ice and water can survive even below ice saturation and may subsequently remerge once the ice saturation ratio exceeds 100% again (Marcolli, 2017, 2020). Deposition nucleation was considered in the INPC calculation using the parameterization scheme U17-D (see Fig. 8c). The mean (maximum) INPCs above 7.5 km are estimated to be 0.8 L−1 (2.6 L−1), 27.2 L−1 (102.3 L−1), and 374.9 L−1 (1484.9 L−1) for three assumed constant ice saturation ratio (
$ {S}_{i} $ ) values of 1.15, 1.25, and 1.35, respectively, based on the U17-D parameterization scheme (Ullrich et al., 2017). The INPC profile calculated with the$ {S}_{\mathrm{i}} $ profile (light blue in Fig. 6a) measured by water vapor Raman lidar is shown by the gray line in Fig. 8c. The actual INPCs generally varied between the INPC values calculated from the fixed$ {S}_{\mathrm{i}} $ of 1.15 and 1.35.Figure 8. Profiles of (a) dust mass concentration
$ {M}_{\mathrm{d}} $ , (b) dust surface area concentration$ {S}_{\mathrm{d}} $ , and (c) ice-nucleating particle concentration derived by the POLIPHON method during 0140–0150 LST on 7 February 2018. The horizontal dashed lines denote the temperature height levels of –35°C and –51.9°C (cloud top temperature). The INPC profiles obtained by the U17-D parameterization schemes are given.$ {S}_{\mathrm{i}} $ represents ice saturation ratio.Figure 9. Overview of cloud top temperature and in-cloud peak RHi (obtained by ERA5 reanalysis data) of cirrus clouds for each case. The red circles (CTT < –36°C and also RHi ≥ 150%) indicate the cases showing competition between heterogeneous and homogeneous nucleation. The dark blue circles (100% < RHi < 150%) indicate that only heterogeneous nucleation takes place.
-
During the period from December 2012 to February 2018, 22 cases of cirrus clouds (a single bulk cloud or a series of very adjacent clouds) relevant to transported dust layers were identified over Wuhan, China. All the events occurred in spring and winter when Wuhan was frequently intruded by dust aerosols. In addition to the two case studies shown above, the remaining 20 occasions were also examined. The detailed statistics, including the local time, cloud top height and temperature, dust extinction coefficient, dust mass concentration, volume depolarization ratio, particle depolarization ratio, in-cloud peak RHw and RHi, maximum INPCs for three assumed constant
$ {S}_{\mathrm{i}} $ , and actual upper limited INPC (calculated with the CTT, cloud top$ {S}_{\mathrm{d}} $ , and in-cloud peak RHi) are listed in Table 3.Date Time (LST) CTH
(km)CTT
(°C)Peak RHw (%) Peak RHi (%) VDR PDR Dust Ext. (Mm−1) Dust mass (μg m−3) U17D
(L−1) (Si=1.15)U17D
(L−1) (Si=1.25)U17D
(L−1) (Si=1.35)Peak INPC (L−1) 20121231 0730−1100 9.9 −46.3 100.3 153.9 0.04 0.37 4.31 8.63 0.024 0.571 5.903 4139.180 20130114 0527−0720 8.1 −37.3 101.9 138.8 0.03 0.29 3.39 6.79 0.001 0.013 0.085 0.051 20131224 0510−0737 9.3 −45.3 96.2 140.2 0.05 0.37 5.36 10.72 0.008 0.132 1.044 NAN 20131225 0445−0800 9.2 −43.5 97.0 140.1 0.18 0.42 42.80 85.69 0.004 0.050 0.324 19.333 20141220 1627−1718 7.7 −38.5 76.1 107.5 0.05 0.23 10.24 20.51 0.016 0.352 3.405 NAN 20150312 2102−2207 8.8 −42.8 88.1 120.2 0.07 0.24 15.15 30.32 0.001 0.010 0.059 0.045 20150322 0405−0650 8.6 −36.1 98.8 141.3 0.02 0.21 2.80 5.61 0.001 0.011 0.075 0.278 20160311 0100−0210 8.5 −39.4 80.7 111.2 0.10 0.35 21.36 42.76 0.003 0.047 0.373 0.105 20160328 2050−2315 9.1 −36.4 102.1 154.0 0.07 0.31 11.26 22.55 NAN 0.003 0.019 NAN 20161128 2055−2400 9.4 −41.5 91.4 140.8 0.03 0.28 3.51 7.03 0.544 17.639 224.957 5.215 20170413 0810−1000 8.9 −34.0 91.6 129.3 0.03 0.25 4.78 9.57 0.001 0.013 0.079 0.016 20170414 0220−0445 9.0 −38.1 82.8 123.3 0.03 0.23 4.51 9.02 0.008 0.150 1.319 0.038 20170423 0545−0757 7.8 −30.4 92.7 122.8 0.05 NAN 2.91 5.83 NAN NAN NAN NAN 20170427 2335−2357 9.7 −39.7 81.2 109.4 0.03 0.30 3.57 7.16 0.032 0.795 8.413 NAN 20170428 0000−0620 9.0 −33.4 95.8 129.2 0.03 0.32 4.52 9.06 NAN NAN NAN NAN 1746−1800 9.4 −38.5 78.3 116.1 0.03 0.28 2.11 4.23 0.089 2.574 30.228 0.042 20170517 2220−2308 10.0 −42.7 88.4 130.6 0.03 0.30 3.13 6.26 0.168 5.074 61.567 2.521 20171209 2340−2400 8.3 −30.4 93.0 127.9 0.05 0.24 7.27 14.56 NAN NAN NAN NAN 20171210 0300−0455 8.1 −34.4 93.1 126.0 0.09 0.34 16.10 32.24 NAN NAN NAN NAN 20180109 0557−0714 7.6 −36.1 83.6 111.4 0.03 0.18 4.86 9.73 0.001 0.007 0.041 NAN 20180112 2055−2328 9.7 −47.8 100.0 157.4 0.07 0.32 10.40 20.82 0.081 2.315 26.966 18739.800 20180207 0013−0132 10.2 −51.9 83.3 133.9 0.05 0.33 6.13 12.27 2.647 102.314 1484.930 6576.980 Table 3. Information on 22 cirrus cloud layer cases with dust-related ice nucleation during December 2012–February 2018 at Wuhan (30.5°N, 114.4°E) obtained from polarization lidar observations and ECMWF ERA5 reanalysis data. Cloud top height (CTH) and the dust-layer-averaged values of dust extinction coefficient (Dust Ext.), dust mass concentration (Dust mass), volume depolarization ratio (VDR), and particle depolarization ratio (PDR) are calculated based on the polarization lidar data. Cloud top temperature (CTT), and respective in-cloud peak RHw and RHi are given by ERA5 reanalysis data.
$ {S}_{\mathrm{i}} $ represents the ice saturation ratio. The maximum INPCs within the dust layer for three assumed constant$ {S}_{\mathrm{i}} $ (1.15, 1.25, and 1.35) are calculated by the POLIPHON method using U17-D parameterization schemes. Peak INPC is calculated with the CTT, cloud top$ {S}_{\mathrm{d}} $ , and in-cloud peak RHi, indicating the upper limit INPC in the actual atmosphere.The studied cirrus clouds were mostly located on top of the aloft dust layers. The dust-layer-averaged values of the dust extinction coefficient, dust mass concentration, volume depolarization ratio, and particle depolarization ratio were 2.1–42.8 Mm−1, 4.2–85.7 µg m−3, 0.03–0.18, and 0.18–0.42, respectively. In total, 11 cases showed a
$ {\delta }_{\mathrm{p}} $ value of ≥ 0.3, indicating the presence of pure dust particles (Sakai et al., 2010). Kojima et al. (2006) confirmed that these uncoated dust particles can still exist in the free troposphere after thousands of kilometers of transport, which can be of particular importance for cirrus cloud formation in the atmosphere. The INPC values mostly varied between the order of magnitude of 10−3 L−1 and 102 L−1. DeMott et al. (2003) also measured a similar cirrus-active INPC of < 30 L−1 in western America. These INPC values are consistent with the typical ice crystal number concentration of 1–100 L−1 caused by heterogeneous freezing (Cziczo et al., 2013).Figure 9 shows the overview of cloud top temperature (CTT) and in-cloud peak RHi obtained by ERA5 reanalysis data for each cirrus cloud case. It should be mentioned that the lidar-derived RHi data are available only for the last five cases listed in Table 3. We compared the RHi profiles from lidar measurements and ERA5 data and found that they generally agree with each other well. Hence, the RHi values derived from ERA5 data were used in the statistical study (Immler et al., 2008; Gamage et al., 2020). The RH peak, indicating the occurrence of cloud, from ERA5 and Raman lidar appeared at almost the same altitudes with similar values. Benefiting from the much higher vertical resolution, Raman lidar observation showed more details of RH vertical variation compared with ERA5. For cases with CTT < –36°C and RHi ≥ 150%, we infer that there is competition between heterogeneous and homogeneous nucleation within the cirrus cloud (Cziczo et al., 2013). The respective proportion (or importance) of these two nucleating regimes depends on the velocity of updraft within the cloud (Kärcher and Lohmann, 2003). Considering the large amount of dust INP supplied here, a stronger updraft on the order of 1 m s−1 is necessary to favor the transition from heterogeneous to homogeneous freezing (Dierens, 2003). However, the necessary high-temporal-resolution vertical velocity data (e.g., from Doppler lidar) are absent and should be a focus of future research. Similarly, using the RHi measured during the INCA campaign, Haag et al. (2003) also reported that in-situ cirrus clouds are probably formed from the combination of heterogeneous and homogeneous nucleation in midlatitude regions of the Northern Hemisphere. Once RHi reduces to the range of 100%–150%, we infer that only heterogeneous nucleation takes place in cirrus cloud formation. As a result, the observed events can be classified into two categories corresponding to the two case studies: (1) category A (3 cases): competition between heterogeneous and homogeneous nucleation (red circles in Fig. 9); (2) category B (19 cases): only heterogeneous nucleation takes place (dark blue circles in Fig. 9).
To evaluate the possible differences in lidar-derived dust optical properties between the two case types, a Mann-Whitney U test was performed for dust extinction coefficient and particle depolarization ratio (Mann and Whitney, 1947). For dust extinction coefficient, both U1 for category A (=37) and U2 for category B (=20) are larger than the critical value U0.05 (=7). For particle depolarization ratio, both U1 for category A (=41) and U2 for category B (=13) are also larger than the critical value U0.05 (=7). Therefore, there are no significant differences in the dust extinction coefficient and particle depolarization ratio between the two categories. This suggests that the optical properties of the dust aerosols in the vicinity of the investigated cirrus clouds are possibly not responsible for the distinction between the two case types. Hence, we conjecture that ambient ice saturation ratio and temperature should be the major factors leading to the two case types, and sufficient INPCs were provided in all the cases for catalyzing heterogeneous nucleation during the dust intrusion period.
Overall, the significance of heterogeneous nucleation within the cirrus cloud triggered by transported dust particles was verified in the upper troposphere over Wuhan. Although accompanying cloud top temperatures were low (–51.9°C to –30.4°C), heterogeneous nucleation can still predominantly contribute to the total ice nucleation and even completely inhibit the homogeneous nucleation process. This is in accordance with the earlier finding that if the upper-tropospheric INPCs reach up to the same magnitude as the lower troposphere and the vertical velocity is as weak as < 0.3 m s−1, homogeneous nucleation is almost negligible (DeMott et al., 1994; Spice et al., 1999). Based on the analysis of the cirrus residue composition, relative humidity, and cirrus particle concentration measurements, Froyd et al. (2013) inferred that heterogeneous nucleation is a dominant cirrus formation mechanism for midlatitude, subtropical, and tropical regions. This effect can strongly influence the microphysical properties of cirrus clouds because heterogeneous nucleation generally leads to a low number concentration of ice crystals compared with homogeneous nucleation (Froyd et al., 2010) and thus induces a net cloud forcing of –0.4 W m−2 (Liu et al., 2012) or –2.0 W m−2 (Lohmann et al., 2008) as estimated by the atmosphere model. In addition, these dust-related cirrus clouds were also reported to show modified optical and microphysical properties, including decreases in cloud optical depth, cloud ice water path, and cloud effective particle radius (Huang et al., 2006; Wang et al., 2015; Pan et al., 2019).
Parameter | Uncertainty |
Dust backscatter coefficient $ {\beta }_{\mathrm{d}} $ | 10% |
Dust extinction coefficient $ {\alpha }_{\mathrm{d}} $ | 20% |
Volume depolarization ratio $ {\delta }_{\mathrm{v}} $ | 5% |
Particle depolarization ratio $ {\delta }_{\mathrm{p}} $ | 5%–10% |
Dust mass concentration ${M }_{\mathrm{d} }$ | 20%–30% |
Dust surface area concentration $ {S}_{\mathrm{d}} $ | 30%–40% |
INP number concentration $ {n}_{\mathrm{I}\mathrm{N}\mathrm{P}} $ | 50%–500% |