Chapter 4. Simulation of Dust – Cloud Interactions using the WRF Model
4.3.3. Effect of scavenge processes on dust deposition
To investigate the effect of rainout and washout on wet deposition of dust particles,
two sets of experiments were designed to compare their results with that of Exp-All.
In Exp-NWout, below-cloud (defined as grids below 1 km and the relative humility is
less than 0.98) dust particles cannot be collected by cloud drops or raindrops; in
Exp-NRout the dust particles cannot enter hydrometeors via heterogeneous nucleation,
activation, or in-cloud collision-collection by cloud drops or raindrops. The
comparison of the two experiments with the Exp-All could help clarify the relative
importance of rainout and washout. In addition, the Exp-Non, which ignored the
interactions between dust and cloud, can be used to double check the nonlinearity.
The near-surface dust concentration in each experiment has been shown in Figure
4.5. The vertical structure of dust distribution is displayed in Figure 4.17 which shows
that the near-surface dust particles have the highest concentration (domain-4 average)
around the 30thhour of simulation, at which times the concentrations decrease with
altitudes. But at other times the highest dust concentrations occurred above the
atmospheric boundary layer. Figure 4.18a shows the absence of either rainout or
washout has limited effect on surface dust concentration. The corresponding changes
in dust concentration for the Exp-NWout, EXP-NRout, and EXP-Non are +1%, +5%,
and +7%, respectively (Table 4.5). However, the differences in dust column loading
are much more obvious (Figure 4.18b). By turning off the rainout (i.e., Exp-NRout)
the column loading increased by 87%, but turning off the washout only lead to a 5%
increase (Table 4.5). So, rainout is much more important than washout in this
particular case. Note that rainout and washout do not seem to compete with each other.
When turning off both, the column loading of dust increased by 101%, which is
significantly higher than the sum of individual contributions (i.e., 87% and 5%). This
indicates that the effect of wet removal on the dust deposition processes is nonlinear,
because of the changes in cloud microphysical processes associated with the rainout
processes. .
The amount of dust deposition in the Exp-All is shown in Figure 4.19. Both the
dry deposition and the wet deposition have maxima at the upwind mountainous regions
near the eastern and northern part of Taiwan. The topography of Taiwan and the
transportation routes of dust are the primary causes for this phenomenon. The wet
deposition is mainly related to precipitation. The difference in arrival time between
the dusty air and the cloud system may cause the patterns of dust wet deposition and
surface precipitation to deviate. The domain averaged dust concentration in rainwater
is about 136 Pg m-3. The result of Exp-All (Figure 4.19c) showed the muddy rain near
Keelung was up to 400 μg m-3, and the maximum dust concentration in rainwater
occurred over eastern Taiwan where the high concentration of dust met the mild rain.
The time series in Figure 4.20a shows that the domain 4 dry deposition is insignificant
in the Exp-Non; while the largest amount of dry deposition is produced in the Exp-All,
which also has the highest integrated wet deposition. Such contrasts may be related to
the presence of the dust recycling mechanisms, in which the dust particle that was
collected by a hydrometeor would be return into the atmosphere when that hydrometeor
evaporated. Furthermore, these recycled dust particles have been resided in
hydrometeors and thus experienced higher gravitational sedimentation. Since there are
no dust-cloud interactions in the Exp-Non, the removal of dust particles would be
simply dominated by the size-related gravitational deposition. The larger particles
would have deposited much earlier near the source region over the continent (Figure
4.21a), while those small particles that reached the downstream have insufficient fall
speed for gravitational deposition onto the ground. Different from Exp-Non, upper
level dust in Exp-All can be scavenged into clouds and transported to lower levels at the
downstream area, then return to the atmosphere by recycling process and deposit via dry
deposition (Figure 4.21b). That’s why switching off the dust recycling mechanism in
the model (i.e. the Exp-NRcy), the resulted distribution of dry deposition (Figure 4.21c)
would be similar to that of the Exp-Non. The time series of the amount of wet
deposition (Figure 4.20b) would be the opposite of the amount of dust that remained in
the atmosphere (Figure 4.18b). The Exp-NWout has similar amount of wet deposition
with that of the Exp-All, while that of the Exp-NRout is lower. The result is consistent
with the aforementioned conclusion that rainout process plays a much more important
role than washout. On the other hand, the total wet deposition in the Exp-NWout is
higher than that of the Exp-All by 11%. This is possibly due to that fact that, without
the washout process, more dust can be retained in the upstream atmosphere which allow
more dust particles to be transported to the downstream (i.e. Domain 4) and caused
greater deposition. Otherwise, removing the rainout process, the amount of wet
deposition is dropped by 40%.
4.3.4 Effect of the recycling process
When a hydrometeor particle evaporated completely, the dust or solute resided in it
should be return to the atmosphere, which is called the aerosol recycling process. This
process is commonly ignored in meteorological models. In this section, a simulation
(Exp-NRcy) that does not allow dust recycling (i.e., assuming the dust in hydrometeors
would disappear when the hydrometeors evaporated completely) was conducted to
understand its effect. Figure 4.22 shows that surface dust concentrations do not
change much after turning off recycling process, but the column loadings decreased by
about 12% (also see Table 4.5). The decrease in dust concentration occurred mainly
around 3 km and 7 km (Figure 4.22c). In the end, the total dust deposition was
decreased by about 83% (Table 4.6). But the reduction is mainly in the dry deposition,
which makes the wet deposition ratio increased form 45% (in Exp-All) to 76% (in
Exp-NRcy).
Moreover, turning off the recycling process not only leads to underestimated
airborne dust concentration and wet deposition but also changes in precipitation. This
indicated that dust released from the evaporated hydrometeors into the atmosphere can
interact with the cloud again. So, this simulation truly showed a complete recycling
process. Figure 4.23 shows that cold cloud decreased at the earlier time (about the 12th
hour), while both clod cloud and warm cloud decreased at the later stage around the 38th
hour. The change in the cold cloud amount during the earlier times is likely due to a
reduction in airborne dust particles and thus their IN effects, whereas changes in both
the warm cloud and cold cloud amounts at the later stage might be due to either the
changes (slower) in precipitation efficiency or convective strength (mechanisms similar
to those mentioned in section 4.3.2.1). The changes in cloud properties lead to a
reduction in total precipitation by about 2.2% comparing to Exp-All. The reduction in
precipitation also feedback to the scavenging process and resulted in a reduction in wet
deposition.
4.4 Discussion
The purposes of this study are to investigate the impact of dust on cloud and
precipitation, and the corresponding effect of cloud on the wet deposition of dust
particles. Therefore it is necessary to choose a dust event accompanied with
precipitation within the same region for analyzing the two-way interactions.
Verification of the simulation results is necessary, and the satellite data used in this
study proved relatively good spatial coverage. However the variables that obtained
from satellite retrieval are not always match fully, in physical meaning, those of the
numerical models. Also, the information of the near-surface or low-level atmosphere
is usually insufficient from satellite observations. Therefore, the combination of the
high-level atmosphere data obtained by satellites and the surface observation data
should be a better verification method. Since the surface data (such as precipitation
and PM10 concentration) in Taiwan are more readily available, this study selected the
domain 4 to that covered the Taiwan for more detailed investigation. However, the
surface observations regarding to dust, especially the dry and wet deposition, are rather
limited in the vicinity of Taiwan. There are also difficulties in the modeling work.
Since Taiwan is located at far downstream of the source region of the East Asian dust, a
pre-run is required to provide sufficient time for the dust in the model to spin-up and
transport from the source region across the continent to Taiwan. However, the error of
the models would accumulate at the latter stage of the pre-run. For example, even
though the four-dimensional data assimilation with FNL data has been applied, the
location of the front at the end of the seven-day pre-run still deviated seriously from the
observed position. With a four-day pre-run, the dust concentration at Taiwan is too
low, because some of the atmospheric dust particle was produced in earlier dust storm
events. As a compromise, this study applied five-day pre-run in order to obtain a
better location of the front (compared with that of the seven-day pre-run). There are
still some discrepancies between the model simulations and the observation, such as the
underestimation of ice water path, the northwesterly shift of precipitation on Taiwan,
and the arrival time of dust to Taiwan mismatch the PM10 observations; but the
location of the front, the transportation route of dust from the source region to Taiwan,
and the precipitation amount, were acceptable and thus applicable to different sensitivity
tests for discussing the effects of relevant mechanisms.
Apart from discussing the impact of dust on clouds through different mechanisms
as mentioned in section 4.3, the corresponding effects of those mechanisms on dust
concentration are worthy of discussion. A sensitivity test is designed to investigate the
effect of different dust concentrations. The Exp-All.L and the Exp-All.H are similar to
the Exp-All, except that the dust intensities were adjusted by 0.35 and 3.5 times,
respectively. The results indicate that the near-surface dust concentration increases
almost linearly with the dust intensity (Figure 4.24a), and the amounts of cold-cloud
water paths in the Exp-All.L and the Exp-All.H are respectively changed by -6.2% and
2.7% (Table 4.7). Also there are slight decreases in warm-cloud water contents in both
experiments. This indicates that the ice-phase processes become stronger with
increasing amount of dust; while the changes in warm-clouds, although relatively minor,
correspond well with cold-cloud changes. Consequently, the amounts of precipitation
(Figure 4.24b) in the Exp-All.L and the Exp-All.H have changed by -2.3% and 0.5%,
respectively. The result shows that precipitation amount vary somewhat nonlinearly
with dust concentration. With 3.5 times dust concentration the Exp-All.H produced
only 0.5% increase in precipitation than in Exp-All, which may implied that there might
be a saturation effect for dust on precipitation.
Another discovery about the impact of dust on precipitation in the selected case is
that the amount of precipitation is reduced by only 3% when the effects of dust on
clouds are completely switched off. It indicates the ability of dust in affecting
precipitation is relatively insignificant under a strong convective system in which a
significant amount of cloud ice can be produced via homogeneous nucleation. Also,
the cloud system was rich in cloud water, which allows the ice particles to grow without
much competition effect, and also allows strong secondary ice production through the
Hallett-Mossop mechanism. Furthermore, the convection in the frontal system is
dominant by large-scale dynamics, and ice particle formation can only alter the
precipitation efficiency somewhat but cannot significantly influence the accumulated
precipitation especially for a cloud system with relatively long lifetime.
In the future, the impact of dust under different weather conditions are worthy of
further investigation, such as clouds with weaker convection and thus insignificant
homogeneous nucleation. For example, the simulation could focus on the high-latitude
regions, such as Japan, to investigate the effect of dust on precipitation formation
dominated by the ice-phase processes. In fact, in this particular frontal system,
switching off the dust-relevant mechanisms caused much stronger changes in the
precipitation (excluding precipitation from cumulus scheme) near the offshore of Japan
in domain 1 (Figure 4.25). The high amount of precipitation within that region (Figure
4.25b) and the fact that precipitation at high latitude is relatively sensitive to cold-cloud
processes might be the main causes for that phenomenon, and also make that region
desirable for discussing the impact of dust on cloud and precipitation. Unfortunately,
relevant information regarding to dust concentrations over the Japan area was difficult
to get.
Table 4.1: The processes considered in different experiments. V means turn on the process, and name in the brackets means the main effect in the experiments.
activation deposition
Table 4.2: The averaged water path of cold cloud, warm cloud and water vapor and accumulated precipitation during 2007/04/01 12UTC to 2007/04/03 12UTC in domain4. The unit of water path and precipitation are (g m-2) and (mm), respectively.
Cold cloud Warm cloud Water vapor acc. precipitation
All 176.27 519.40 36909 14.32
Table 4.3: The averaged water path of hydrometeors in different experiments during 2007/04/01 12UTC to 2007/04/03 12UTC in domain4. (Unit: g m-2)
ice snow graupel cloud rain
Table 4.4: Time and domain averaged net (out-going) long-wave radiative forcing for top of atmosphere and net show-wave radiative forcing for surface in different experiments during 2007/04/01 12UTC to 2007/04/03 12UTC in domain4. (Unit: W m-2)
TOA LW SFC SW
Table 4.5: Time and domain averaged dust surface concentration and total dust loading during 2007/04/01 12UTC to 2007/04/03 12UTC in domain4..
Sfc. Con.
Table 4.6: Accumulated domain averaged of dry, wet and total dust deposition and wet deposition ratio during 2007/04/01 12UTC to 2007/04/03 12UTC in domain4.
dry dep.
Table 4.7: The averaged water path of cold cloud and warm clod, and accumulated precipitation during 2007/04/01 12UTC to 2007/04/03 12UTC in domain4.
cold path (mg/m2)
warm path (mg/m2)
precipitation (mm)
All.L 165 516 13.99
All 176 519 14.32
All.H 181 511 14.39
Figure 4.1: Model domain setup for the numerical experiments.
Figure 4.2: Compare average ice water path (a) and liquid water path (b) on 2007/04/02 from MODIS with the simulated (experiment-All) average ice water path (c) and liquid water path (d) form 2007/04/01 12 UTC to 2007/04/02 12 UTC. Water path unit in g m-2.
Figure 4.3: Compare observed water contents from CloudSat and simulated results. Panel (a) shows the track of CloudSat on 2007/04/01, with segment 29 (17:48:31 UTC to 17:51:42 UTC) crossing the studied frontal system. Ice water and liquid water contents (unit: mg m-3) from track segment 29 of CloudSat are shown in (b) and (c); and the simulated ice water and cloud water contents (averaged during 17 UTC to 18 UTC) are shown in (d) and (e), respectively.
Figure 4.4: Accumulated precipitation during 2007/04/01 12 UTC to 2007/04/03 12 UTC from CWB measurement (a) and simulated result of Exp-All (b).
Figure 4.5: Simulated surface dust average concentration (unit: Pg m-3) during 2007/04/01 12 UTC to 2007/04/02 12 UTC (upper panels) and 2007/04/02 12UTC to 2007/04/03 12UTC (lower panels). The left and right panels are results from domain 1 and domain 4, respectively.
Figure 4.6: Time evolution of measured PM10 (black line) and simulated surface dust concentration (red line) at Wanli (a) and Guanyin (b). (unit: Pg m-3)
Figure 4.7: The simulated results of cloud ice (top left), snow (middle left), graupel (bottom left), cloud drops (top right), raindrops (middle right) and water vapor (bottom right) form Exp-All domain 4 during 2007/04/01 12Z to 2007/04/03 12Z. (unit: mg kg-1) The dashed lines indicate the domain mean temperature at from 0, -20 and -40qCġlevels.
Figure 4.8: Same as Figure 4.7 but for the Exp-Non experiment.
Figure 4.9: Domain 4 averaged precipitation from Exp-All (red), Exp-Nnuc (green), Exp-Nact (blue) and Exp-Non (black). (a) hourly precipitation rate (in mm/hr); and (b) accumulated precipitation (in mm).
Figure 4.10: Domain-4 average column water properties from Exp-All (red), Exp-Nnuc (green), Exp-Nact (blue) and Exp-Non (black): (a) cold cloud water paths (g m-2), (b) warm cloud water path (g m-2), and (c) cloud optical depth.
Figure 4.11: Domain and time average mixing ratio (mg kg-1) of cold cloud (a) and warm cloud (b) from Exp-All (red), Exp-Nnuc (green), Exp-Nact (blue) and Exp-Non (black) in domain 4.
Figure 4.12: Changes in accumulated surface precipitation comparing to the Exp-All run. Panels from left to right are the difference between Exp-Non - Exp-All, Exp-Nact – Exp-All, and Exp-Nnuc – Exp-All, respectively. Top panels are absolute values in mm and lower panels are percentages in %.
Figure 4.13: Domain 4 average TOA longwave radiation (a) and surface short-wave radiation (c) from Exp-All (red), Exp-Nnuc (green), Exp-Nact (blue) and Exp-Non (black). The differences comparing to Exp-All are shown in (b) for long-wave and (d) for short-wave radiation
Figure 4.14: Domain and time average mixing ratio (mg kg-1) of: (a) cold cloud and (b) warm cloud, from Exp-All (red), Exp-Ndep (green), Exp-Nimm (blue) and Exp-Non (black) in domain 4.
Figure 4.15: Domain average water paths (mg m-2) of: (a) cold cloud, and (b) warm cloud, from Exp-All (red), Exp-Ndep (green), Exp-Nimm (blue) and Exp-Non (black) in domain 4.
Figure 4.16: Domain-4 average hourly (a) and accumulated (b) precipitation from Exp-All (red), Exp-Ndep (green), Exp-Nimm (blue) and Exp-Non (black). (unit: mm)
Figure 4.17: Domain 4 averaged dust mass concentration in Exp-All (unit:Pg m-3).
Figure 4.18: Domain-4 average (a) dust surface mass concentration (unit: μg m-3) and (b) dust total loading (unit: mg m-2), from Exp-All (red), Exp-NRout (green), Exp-NWout (blue) and Exp-Non (black).
Figure 4.19: Accumulated dust deposition (unit: mg m-2) from Exp-All: (a) dry deposition, and (b) wet deposition, and the dust concentration (unit: Pg m-3) in surface rainwater (c) from Exp-All.
Figure 4.20: Domain-4 average dust deposition (mg m-2hr-1) at the surface from Exp-All (red line), Exp-NRout (green line), Exp-NWout (blue line) and Exp-Non (black line). (a) dry deposition; and (b) wet deposition.
Figure 4.21: 48-hour accumulated dust dry deposition (mg m-2) from: (a) Exp-All, (b) Exp-Non, and (c) Exp-NRcy
Figure 4.22: Domain-4 time averaged (a) dust surface mass concentration (in Pg m-3), (b) dust total loading (in mg m-2) and (c) dust mass time averaged vertical profile (in Pg m-3) from Exp-All (red line) and Exp-NRcy (light blue dash line).
Figure 4.23: Domain-4 averaged (a) cold cloud column water path (g m-2), (b) warm cloud column water path (g m-2) and (c) precipitation (mm hr-1) from Exp-All (red line) and Exp-NRcy (light blue dash line).
Figure 4.24: Domain-4 average dust surface mass concentration (unit: Pg m-3) (a) and accumulated precipitation (unit: mm) (b) from Exp-All.H (red line), Exp-All (green line) and Exp-All.L (blue line).
Figure 4.25: 48-hour accumulated precipitation (excluding those from cumulus parameterization) of Exp-All (a) and the difference between Exp-Non to Exp-All (b). (unit: mm)
Figure 4.A 1: Hydrometer mixing ratio in experiment-All, unit: mg kg-1.
Figure 4.A 2: Similar to Figure 4.A 1 but for experiment-Non, unit: mg kg-1.
Figure 4.A 3: Similar to Figure 4.A 1 but for experiment-Nact, unit: mg kg-1.
Figure 4.A 4: Similar to Figure 4.A 1 but for experiment-Nnuc, unit: mg kg-1.
Figure 4.A 5: Similar to Figure 4.A 1 but for experiment-Nimm, unit: mg kg-1.
Figure 4.A 6: Similar to Figure 4.A 1 but for experiment-Ndep, unit: mg kg-1.
Figure 4.A 7: Similar to Figure 4.A 1 but for experiment-NWout, unit: mg kg-1.
Figure 4.A 8: Similar to Figure 4.A 1 but for experiment-NRout, unit: mg kg-1
Figure 4.A 9: Similar to Figure 4.A1 but for experiment-NRcy, unit: mg kg-1.
Figure 4.A 10: Hydrometer number concentration in experiment-All, unit: cm-3.
Figure 4.A 11: Similar to Figure 4.A 10 but for experiment-Non, unit: cm-3.
Figure 4.A 12: Similar to Figure 4.A 10 but for experiment-Nact, unit: cm-3.
Figure 4.A 13: Similar to Figure 4.A 10 but for experiment-Nnuc, unit: cm-3.
Figure 4.A 14: Similar to Figure 4.A 10 but for experiment-Nimm, unit: cm-3.
Figure 4.A 15: Similar to Figure 4.A 10 but for experiment-Ndep, unit: cm-3.
Figure 4.A 16: Similar to Figure 4.A 10 but for experiment-NWout, unit: cm-3.
Figure 4.A 17: Similar to Figure 4.A 10 but for experiment-NRout, unit: cm-3.
Figure 4.A 18: Similar to Figure 4.A 10 but for experiment-NRcy, unit: cm-3.
Figure 4.A 19: Volume-mean size of hydrometeors in experiment-All, unit: μm.
Figure 4.A 20: Similar to Figure 4.A 19 but for experiment-Non, unit: μm.
Figure 4.A 21: Similar to Figure 4.A 19 but for experiment-Nact, unit: μm.
Figure 4.A 22: Similar to Figure 4.A 19 but for experiment-Nnuc, unit: μm.
Figure 4.A 23: Similar to Figure 4.A 19 but for experiment-Nimm, unit: μm.
Figure 4.A 24: Similar to Figure 4.A 19 but for experiment-Ndep, unit: μm.
Figure 4.A 25: Similar to Figure 4.A 19 but for experiment-NWout, unit: μm.
Figure 4.A 26: Similar to Figure 4.A 19 but for experiment-NRout, unit: μm.