Mid-latitude mixed-phase stratocumulus clouds and their interactions with aerosols : 1 how ice processes affect microphysical , dynamic and thermodynamic development in 2 those clouds and interactions ?

53 54 Mid-latitude mixed-phase stratocumulus clouds and their interactions with aerosols remain 55 poorly understood. This study examines the roles of ice processes in those clouds and 56 interactions using a large-eddy simulation (LES) framework. Cloud mass becomes much 57 lower in the presence of ice processes and the Wegener-Bergeron-Findeisen (WBF) 58 mechanism in the mixed-phase clouds as compared to that in warm clouds. This is because 59 while the WBF mechanism enhances the evaporation of droplets, the low concentration of 60 aerosols as ice nuclei (IN) and cloud ice number concentration (CINC) prevent the efficient 61 deposition of water vapor whose mass is contributed by the evaporation. In the mixed62 phase clouds, the increasing concentration of aerosols that act as cloud condensation nuclei 63 (CCN) decreases cloud mass by increasing the evaporation of droplets through the WBF 64 mechanism and decreasing the intensity of updrafts. In contrast to this, in the warm clouds, 65 the absence of the WBF mechanism makes the increase in the evaporation of droplets 66 inefficient, eventually enabling cloud mass to increase with the increasing concentration of 67 aerosols as CCN. Here, the results show that when there is an increasing concentration of 68 aerosols that act as IN, the deposition of water vapor is more efficient than when there is 69 the increasing concentration of aerosols as CCN, which in turn enables cloud mass to 70 increase in the mixed-phase clouds. 71 72 73 74 75 76 77 78 79 80 81 82 83 https://doi.org/10.5194/acp-2020-1318 Preprint. Discussion started: 11 March 2021 c © Author(s) 2021. CC BY 4.0 License.

to follow the tri-modal log-normal distribution. Stated differently, the size distribution of 270 background aerosols as CCN in all parts of the domain during the whole simulation period 271 is assumed to follow size distribution parameters or the shape of distribution as shown in 272 points. This is based on a general difference in concentration between CCN and IN 282 (Pruppacher and Klett, 1978). 283 In clouds, aerosol sinks and sources control the evolution of aerosol size distribution. To examine effects of the aerosol advection on the observed stratocumulus clouds over the 309 Seoul area, the control run is repeated by removing the increase in aerosol concentrations 310 due to the aerosol advection. This repeated run is referred to as the low-aerosol run. In the 311 low-aerosol run, to remove the increase in aerosol concentrations, background aerosol 312 concentrations after 03 LST on January 12 th do not evolve with the aerosol advection but 313 is assumed to have background aerosol concentrations at 03 LST on January 12 th at every 314 time step and grid point only for the concentration of background aerosols acting as CCN. 315 Here, the time-and domain-averaged concentration of background aerosols as CCN after 316 03 LST on January 12 th in the low-aerosol run is lower than that in the control run by a 317 factor of ~3. It is notable that there are no differences in the concentration of background 318 aerosols acting as IN between the control and low-aerosol runs. This is to isolate effects of 319 CCN, which accounts for most of aerosols, on clouds from those effects of IN via 320 comparisons between the runs. Via the comparisons, we are able to identify how advection-321 induced increases in the concentration of aerosols acting as CCN affect clouds. The ratio 322 of the concentration of background aerosols as CCN at 03 LST on January 12 th to that after 323 03 LST on January 12 th varies among grid points and time steps, since the concentration 324 varies spatiotemporally throughout the simulation period in the control run. This means 325 that a factor by which the concentration of background aerosols as CCN varies after 03 326 LST on January 12 th between the control and low-aerosol runs is different for each of the 327 time steps and grid points. 328 To examine effects of the interplay between ice crystals and droplets on the adopted 329 system of stratocumulus clouds and its interactions with aerosols, the control and low-330 aerosol runs are repeated by removing ice processes. These repeated runs are referred to as 331 12 the control-noice and low-aerosol-noice runs. In the control-noice and low-aerosol-noice 332 runs, only aerosols as CCN, droplets (i.e., cloud liquid), raindrops and associated 333 microphysical processes (e.g., condensation and evaporation) exist, and aerosols as IN, all 334 solid hydrometeors (i.e., ice crystals, snow, graupel, and hail) and associated processes 335 (e.g., deposition and sublimation) are turned off, regardless of temperature. Via 336 comparisons between the control and control-noice runs, we aim to identify effects of the 337 interplay on the adopted system. Via comparisons between a pair of the control and low-338 aerosol runs and that of the control-noice and low-aerosol-noice runs, we aim to identify 339 effects of the interplay on interactions between the system and aerosols. Henceforth, the 340 pair of the control and low-aerosol runs is referred to as the ice runs, while the pair of the 341 control-noice and low-aerosol-noice runs is referred to as the noice runs. 342 To better understand findings in Section 4.1.1, which explain how the interplay between 343 ice crystals and droplets affects stratocumulus clouds, the control run is repeated by 344 increasing the concentration of background aerosols acting as IN by a factor of 10 and 100 345 at each time step and grid point. These repeated runs are detailed in Section 4.  Figure 3a shows the time series of the domain-averaged liquid-water path (LWP), ice-water 356 path (IWP) and water path (WP), which is the sum of LWP and IWP, for the control run, 357 and LWP for the control-noice run. Since in the control-noice run, there are no ice particles, 358 LWP acts as WP in the run. WP is higher in the control-noice run than in the control run 359 throughout the whole simulation period, although at the initial stage before 20:00 LST on 360 13 around 00:00 LST on January 13th when WP reaches its maximum value in each of the 363 runs ( Figure 3a). These differences decrease as time goes by after around 00:00 LST on 364 January 13th (Figure 3a). The time-and domain-averaged WP over the period between 00 365 LST (local solar time) on January 12th and 00 LST on January 14th is 18 g m -3 and 55 g 366 m -3 in the control and control-noice runs, respectively. Associated with this, the WP peak 367 value reaches 83 g m -3 in the control run, while the value reaches 230 g m -3 in the control-368 noice run (Figure 3a). Over most of the simulation period, IWP is greater than LWP in the 369 control run except for the period between ~22:00 LST on January 12th and ~01:00 LST on 370 January 13th (Figure 3a). In the control run, the time-and domain-averaged IWP and LWP 371 are 11 g m -3 and 7 g m -3 , respectively. Results here indicate that when solid and liquid 372 particles coexist, cloud mass, represented by WP, reduces a lot as compared to that when 373 liquid particles alone exist. To evaluate the control run, satellite and ground observations 374 can be utilized. In the case of the Moderate Resolution Imaging Spectroradiometer, one of 375 representative polar orbiting image sensors on board satellites, it passes the Seoul area only 376 at 10:30 am and 1:30 pm every day, hence, the sensor is not able to provide reliable data 377 that cover the whole simulation period. Multifunctional Transport Satellites (MTSAT), 378 which are geostationary satellites and available in the East Asia, do not provide reliable 379 data of LWP and IWP, although they provide comparatively reliable data of cloud fraction 380 and cloud-top height throughout the whole simulation period (Faller, 2005). Ground 381 observations provide data of cloud fraction and cloud-bottom height throughout the whole 382 simulation period. Hence, the simulated cloud fraction, and cloud-bottom and -top heights 383 are compared to those from the MTSAT and ground observations. The average cloud 384 cloud-bottom and -top heights between the control run and observations is ~ 10% and thus 389 the control run is considered performed reasonably well for these variables. 390 Condensation and deposition as phase-transition processes are the main sources of 391 cloud mass in the control run. Since in the control-noice run, there are no ice particles,

14
As seen in Figure 3b, condensation rates in the control-noice run are much higher than the 394 sum of condensation and deposition rates in the control run. Associated with this, there is 395 greater cloud mass in the control-noice run than in the control run, although deposition is 396 absent in the control-noice run. However, at the initial stage before 20:00 LST on January 397 12 th , differences between the sum in the control run and condensation rate in the control-398 noice run are not significant as compared to those after 20:00 LST on January 12 th ( Figure  399 3b). Hence, those differences become significant and increase as time progresses after the 400 initial stage. Those differences are greatest around 00:00 LST on January 13 th when the 401 sum in the control run or condensation rate in the control-noice run reaches its maximum 402 value. The differences decrease as time goes by after around 00:00 LST on January 13 th . 403 Condensation rate, deposition rate in the control run, and condensation rate in the control-404 noice run are similar to LWP, IWP in the control run, and LWP in the control-noice run, 405 respectively, in terms of their temporal evolutions (Figures 3a and 3b). This similarity 406 confirms that deposition and condensation are the main sources of IWP and LWP, 407 respectively, and control cloud mass. Thus, understanding the evolutions of condensation 408 and deposition is equivalent to understanding those of LWP and IWP, respectively. Hence, 409 in the following, to understand evolutions of cloud mass and its differences between the 410 control and control-noice runs, we analyze evolutions of condensation, deposition, and 411 their differences between the runs. 412 The qualitative nature of differences in WP, which represents cloud mass, over the 413 whole simulation period between the control and control-noice runs is initiated and 414 established during the initial stage of cloud development before 20:00 LST on January 12 th 415 (Figures 3a and 3b). Hence, to understand mechanisms that initiate differences in WP 416 between the control and control-noice runs, deposition, condensation and associated 417 variables are analyzed for the initial stage. Note that synoptic or environmental conditions 418 such as humidity and temperature are identical between the control and control-noice runs. 419 These conditions act as initial and boundary conditions for the simulations and thus initial 420 and boundary conditions are identical between the runs. Also, during the initial stage, 421 feedbacks between dynamics (e.g., updrafts) and microphysics just start to form and thus 422 are not fully established as compared to those feedbacks after the initial stage. This enables 423 us to perform analyses of deposition and condensation during the initial stage by reasonably excluding a large portion of complexity caused by those feedbacks. Hence, those analyses 425 during the initial stage can provide a clearer picture of either microphysical or dynamic 426 mechanisms that control differences in results between the runs. 427 During the initial stage before 20:00 LST on January 12 th , evaporation rates, averaged 428 over the cloud layer, are higher in the control run than in the control-noice run due to the 429 WBF mechanism which facilitates evaporation of droplets and deposition onto ice crystals 430 ( Figure 3c). As seen in Figure 3c, the cloud layer is between ~200 m and ~1.5 km in the 431 control run, while it is between ~200 m and ~2.5 km in the control-noice run. Associated 432 with more evaporation, droplets disappear more, leading to a situation where cloud droplet 433 number concentration (CDNC) starts to be lower in the control run during the initial stage 434 ( Figure 3d). Then, during the initial stage, the reduction in CDNC leads to a reduction in 435 condensation in the control run as compared to that in the control-noice run (Figure 3b). 436 Fewer droplets mean that there is a less integrated droplet surface area where condensation 437 occurs and this induces less condensation in the control run. However, aided by the WBF 438 mechanism, deposition is facilitated at the initial stage, and this leads to greater deposition 439 than condensation in the control run at the initial stage ( Figure 3b). This deposition is 440 inefficient and the subsequent increase in deposition is not sufficient, so, the sum of 441 condensation and deposition rates in the control run is slightly lower than condensation 442 rate in the control-noice run at the initial stage ( Figure 3b); this contributes to slightly lower 443 WP in the control run than in the control-noice run during the initial stage ( Figure 3a). 444 Hence, slightly greater latent heating, which is associated with condensation, in the control-445 noice run than that, which is associated with the sum of deposition and condensation, in 446 the control run develops during the initial stage. This leads to stronger feedbacks between 447 updrafts and latent heating in the control run than in the control-noice run after the initial 448 stage, which in turn result in much stronger updrafts after the initial stage in the control-449 noice run than in the control run. Due to these much stronger updrafts after the initial stage, 450 the time-and domain-averaged updrafts over the whole simulation period are also much 451 greater in the control-noice run than in the control run ( Figure 4a). The much stronger 452 updrafts after the initial stage produce much larger WP in the control-noice run than in the 453 control run after the initial stage ( Figure 3a).

16
The WBF mechanism indicates that the surplus water vapor produced by evaporation 455 acts as an additional source of deposition. If water vapor, including the surplus water vapor, 456 is efficiently deposited onto ice crystals, then the reduced cloud mass, due to the increased 457 evaporation and the subsequently reduced CDNC and condensation, can be efficiently 458 compensated by the additional gain of solid mass via deposition in the control run. This 459 would lead to much smaller differences in WP between the control and control-noice runs 460 than simulated. Here, we hypothesize that the inefficient deposition of water vapor is 461 related to much lower cloud ice number concentration (CINC) as compared to CDNC. As 462 seen in Figures 4b and 4c, CINC is ~ 2 orders of magnitude lower than CDNC. Hence, 463 while there is a comparatively large number of droplets that can potentially produce lots of 464 water vapor via evaporation, there is comparatively a small number of ice crystals and thus 465 a small integrated surface area of ice crystals where water vapor can be deposited in the 466 control run. It is hypothesized that this leads to a situation where water vapor, including 467 that from the evaporation of lots of droplets, is not able to find enough surface area of ice 468 crystals for efficient deposition, eventually leading to the inefficient deposition of the 469 surplus water vapor in the control run. 470 471 a. LWP and IWP frequency distributions 472 473 As seen in Figure 5a, the control-noice run has the lower (higher) WP cumulative frequency 474 for WP below (above) ~ 100 g m -2 than the control run at the last time step. This means 475 that the lower average WP in the control run is mainly due to a reduction in WP above 476 ~100 g m -2 in the control run. Through the WBF mechanism in the presence of ice particles, 477 liquid particles evaporate and condensation reduces. Hence, the LWP frequency reduces 478 substantially in the control run as compared to that in the control-noice run ( Figure 5b). 479 With this reduction, LWP above ~ 800 g m -2 disappears and there is in general two to three 480 orders of magnitude lower LWP frequency for LWP below ~ 800 g m -2 in the control run 481 than in the control-noice run ( Figure 5b). 482 As seen in Figure 5b, at the last time step, there is the presence of IWP frequency in 483 addition to the LWP frequency in the control run. Through the WBF mechanism, which 484 facilitates deposition, the IWP frequency is greater than the LWP frequency for IWP below ~ 200 g m -2 in the control run. Particularly for IWP below ~ 100 g m -2 , the IWP frequency 486 in the control run is greater than the LWP frequency in the control-noice run. This enables 487 the greater WP frequency in the control run than in the control-noice run for WP below ~ 488 100 g m -2 in spite of the lower LWP frequency below ~100 g m -2 in the control run (Figures 489 5a and 5b). However, the lower IWP frequency for IWP above ~ 100 g m -2 in the control 490 run than the LWP frequency for LWP above ~ 100 g m -2 in the control-noice run 491 contributes to the lower WP frequency for WP above ~ 100 g m -2 in the control run (Figures 492 5a and 5b). The lower WP frequency for WP above ~ 100 g m -2 in the control run is also 493 contributed by the lower LWP frequency for LWP above ~ 100 g m -2 in the control run 494 (Figures 5a and 5b). 495 leading to aerosol-induced decreases in the average WP between the ice runs. As seen in 561 Figure 8b, the WP frequency is greater particularly for WP < ~300 g m -2 , leading to the 562 higher average WP in the low-aerosol run than in the control run. As seen in Figure 8c, 563 particularly for WP below ~200 g m -2 , the IWP frequency increases, while the LWP 564 frequency decreases with increasing aerosols between the ice runs. The increase in the IWP 565 frequency is not able to outweigh the decrease in the LWP frequency, leading to aerosol-566 induced decreases in the average WP between the ice runs. Results here are contrary to 567 the conventional wisdom that increasing concentrations of aerosols as CCN tend to 568 increase WP in stratiform clouds (Albrecht, 1989). 569 Between the noice runs, there is an increase in LWP (i.e., WP) with the increasing 570 concentration of aerosols as CCN (Figure 8a). The greater LWP frequency, concentrated 571 in the LWP range between ~100 and ~600 g m -2 , leads to the greater average LWP or WP 572 in the control-noice run than in the low-aerosol-noice run (Figures 8b and 8c). The qualitative nature of aerosol-induced differences in deposition, IWP, condensation and 579 LWP over the whole simulation period between the ice runs is initiated and established 580 during the initial stage of cloud development before 20:00 LST on January 12 th (Figure 8a). 581 To understand mechanisms that control aerosol-induced differences in deposition and 582 condensation as a way of understanding mechanisms that control those differences in IWP 583 and LWP, the time series of deposition rate, condensation rate and associated variables in 584 each of the ice runs and differences in these variables between the ice runs is obtained for 585 the initial stage. Since this study focuses on these differences in the variables as a 586 representation of aerosol effects on clouds, in the following, the description of the 587 differences is given in more detail by involving both figures and text as compared to the 588 description of the variables in each of the ice runs, involving text only for the sake of 589 brevity. 590 591 i.
CDNC and its relation to condensation and evaporation 592 593 Evaporation and condensation rates are higher in the control run than in the low-aerosol 594 run throughout the initial stage and up to ~15:30 LST on January 12 th , respectively ( Figure  595 9a). Increases in evaporation tend to make more droplets disappear, while increases in 596 aerosol activation and condensation counteract the disappearance. The average CDNC over 597 grid points and time steps with non-zero CDNC is larger in the control run than in the low-598 aerosol run not only over the initial stage but also over the whole simulation period ( Figures  599   9a and 10a). This means that on average, the evaporatively-driven increases in the 600 disappearance of droplets is outweighed by the activation-and/or condensationally-601 enhanced counteraction particularly during the initial stage with increasing aerosol 602 concentrations between the ice runs. As marked by a green-dashed box in Figure 9a, there 603 are steady and rapid temporal increases in the CDNC differences between the ice runs over 604 a period from 12:50 to 13:20 LST on January 12 th . This is due to steady and rapid temporal 605 increases in CDNC, which are larger in the control run than in the low-aerosol run, over 606 the period (Figure 9a). More droplets or higher CDNC provides a larger integrated surface 607 area of droplets where evaporation and condensation of droplets occur, and thus acts as 608 more sources of evaporation and condensation. With steady and rapid temporal increases in CDNC as a source of evaporation and condensation, temporal increases in both 610 evaporation, which are linked to the WBF mechanism, and condensation show a jump (or 611 a surge or a rapid increase) in them for the period between 12:50 and 13:20 LST on January 612 12 th in each of the ice runs. This jump is higher associated with the larger temporal increase 613 in CDNC in the control run than in the low-aerosol run, which induces differences in each 614 of evaporation and condensation between the ice runs to jump, as marked by a red-dashed 615 box in Figure 9a, during the time period. 616 The jump in differences in condensation between the ice runs is not as high as that in 617 differences in evaporation between the ice runs ( Figure 9a). This situation accompanies the 618 fact that in each of the ice runs, the jump in evaporation is higher than that in condensation. 619 This means that differences in the jump between evaporation and condensation are greater 620 in the control run than in the low-aerosol run. Hence, evaporatively-driven jump in the 621 disappearance of droplets outweighs condensationally-driven jump in counteraction in 622 each of the ice runs and this outweighing is greater in the control run than in the low-aerosol 623 run during the period with the jumps. Due to this, the increasing temporal trend of CDNC 624 turns to its decreasing trend in each of the ice runs and this decreasing trend is larger in the 625 control run than in the low-aerosol run. This in turn turns the increasing temporal trend of 626 the CDNC differences between the ice runs to their decreasing trend around 13:30 LST on 627 January 12 th (Figure 9a). 628 The decreasing temporal trend of CDNC contributes to a decreasing temporal trend 629 of each evaporation and condensation, starting around 13:30 LST on January 12 th , by 630 reducing the integrated surface area of droplets in each of the ice runs. This decreasing 631 trend of each evaporation and condensation is larger associated with the larger decreasing 632 trend of CDNC in the control run than in the low-aerosol run. This induces the increasing 633 temporal trend of differences in each evaporation and condensation between the ice runs 634 to change into their decreasing temporal trend around 13:30 LST on January 12 th (Figure  635 9a). The decreasing trend of evaporation in each of the ice runs is smaller than that in 636 condensation. Associated with this, the decreasing trend of differences in evaporation 637 between the ice runs is smaller than that in condensation (Figure 9a). Stated differently, the 638 temporal reduction in evaporation in each of the ice runs and its differences between the and low-aerosol-noice runs, respectively. It is notable that the WBF-mechanism-induced 655 evaporation per unit volume of droplets is also strongly proportional to the surface-to-656 volume ratio of droplets (Pruppacher and Klett, 1978). Hence, between the ice runs, 657 enhanced evaporation efficiency by aerosol-induced increases in the surface-to-volume 658 ratio accompanies aerosol-enhanced WBF-mechanism-associated efficiency of 659 evaporation. 660 With the steady and rapid temporal increase in CDNC, there is a steady and rapid 661 temporal enhancement of the surface-to-volume ratio of droplets and evaporation 662 efficiency in each of the ice runs between 12:50 and 13:20 LST on January 12 th . Remember 663 that these increases in CDNC are larger in the control run than in the low-aerosol run. This 664 induces the greater temporal enhancement of the ratio and evaporation efficiency in the 665 control run than in the low-aerosol run. The temporal enhancement of the ratio and 666 evaporation efficiency accompanies temporally enhancing WBF-mechanism-related 667 efficiency of evaporation. This accompaniment boosts evaporation and enables the jump 668 in temporal increases in evaporation to be greater than that in condensation in each of the 669 ice runs. In association with the larger steady and rapid temporal increase in CDNC in the control run than in the low-aerosol run, the temporally enhancing WBF-mechanism-related 671 efficiency of evaporation and its boost on evaporation enhance with increasing aerosol 672 concentrations. This, in turn, enables greater aerosol-induced increases in evaporation than 673 in condensation or the greater jump in differences in evaporation between the ices runs 674 than that in condensation over the period between 12:50 and 13:20 LST on January 12 th 675 ( Figure 9a). 676

Even when both evaporation and condensation rates decrease with time in association 677
with the decreasing temporal trend of CDNC and the surface-to-volume ratio of droplets 678 over a period after 13:30 LST on January 12th during the initial stage in each of the ice 679 runs, evaporation (condensation) rates are maintained higher throughout the initial stage 680 (up to ~15:30 LST) in association with the higher CDNC and surface-to-volume ratio of 681 droplets in the control run than in the low-aerosol run (Figure 9a). The presence of the 682 WBF mechanism facilitates evaporation and this acts against the temporal decrease in 683 evaporation with time over the period in each of the ice runs. This counteraction by the 684 WBF mechanism reduces the temporal decrease in evaporation and enables evaporation to 685 reduce temporally to a less extent as compared to condensation in each of the ice runs for 686 the period. This accompanies the differences in the temporal reduction between 687 evaporation and condensation that are larger in the control run than in the low-aerosol run. 688 This, in turn, enables differences in evaporation between the ice runs to reduce to a less 689 extent as compared to those in condensation over the period (Figure 9a). Due to this, 690 differences (or aerosol-induced increases) in evaporation and associated aerosol-induced 691 increases in evaporation-driven negative buoyancy between the ice runs are higher than 692 those in condensation and condensation-driven positive buoyancy, respectively, for the 693 period ( Figure 9a). This induces the decreasing temporal trend of differences or aerosol-694 induced increases in updraft mass fluxes between the ice runs over the period (Figure 9a). 695 The decreasing temporal trend of aerosol-induced increases in updraft mass fluxes 696 eventually leads to lower updraft mass fluxes in the control run than in the low-aerosol run, 697 as represented by negative differences in updraft mass fluxes between the ice runs from 698 24 condensation between the ice runs from ~15:30 LST onwards during the initial stage 701 (Figure 9a). 702 The role of the WBF mechanism described in this section can be clearly seen by 703 comparing the ice runs in this section to the noice runs, with no WBF mechanism, detailed 704 in the following Section b. 705 706 2) Deposition and condensation 707 708 The difference in deposition between the ice runs is negligible and does not vary much 709 with time up to ~15:30 LST on January 12 th when the difference starts to show its 710 significant increase (Figure 9a). With the start of the decreasing temporal trend of 711 condensation around 13:30 LST on January 12 th , more water vapor, not used by 712 condensation, becomes available for deposition as compared to that before 13:30 LST on 713 January 12 th in each of the ice runs. Remember that this decreasing trend is greater in the 714 control run than in the low-aerosol run. Hence, from 13:30 LST on January 12 th onwards, 715 more water vapor is available for deposition in the control run than in the low-aerosol run. 716 This leads to the start of larger aerosol-induced increases in deposition between the ice runs 717 around 13:30 LST on January 12 th as compared to those increases before ~ 13:30 LST on 25 The increasing temporal trend of aerosol-induced increases in deposition is not able 731 to outweigh the increasing trend of aerosol-induced decreases in condensation between the 732 ice runs after ~ 15:30 LST on January 12 th (Figure 9a). As discussed in Section 4.1, due to 733 very low CINC as compared to CDNC, there is an insufficient integrated surface area of 734 ice crystals for the deposition of available water vapor in the control run. Remember that 735 there is no change in the background concentration of aerosols as IN between the ice runs. 736 Hence, as seen in Figure 9a, there are negligible differences in CINC between the ice runs, 737 although in comparison to the CINC differences, CDNC increases significantly with 738 increasing aerosols between the ice runs. Hence, the ratio of CINC to CDNC is lower in 739 the control run than in the low-aerosol run. This indicates that CINC per unit CDNC and 740 associated unit evaporation is lower in the control run. Hence, the available water vapor, 741 including that from droplet evaporation, has more difficulty in finding the surface area of 742 ice crystals for deposition in the control run. Remember that there is more available water 743 vapor for deposition, which increases deposition more in the control run than in the low-744 aerosol run, after ~ 13:30 LST on January 12 th as compared to that before ~ 13:30 LST on 745 January 12 th . However, the more difficulty in finding the surface area of ice crystals for 746 deposition makes the deposition of the more available water vapor less efficient in the 747 control run than in the low-aerosol run. This damps down the increase in deposition 748 particularly after ~ 13:30 LST on January 12 th in the control run. Then, aerosol-induced 749 increases in deposition are not large enough to overcome aerosol-induced decreases in 750 condensation in the control run particularly after ~ 15:30 LST on January 12 th (Figure 9a). 751 This in turn leads to the lower average WP in the control run than in the low-aerosol run 752 over the whole simulation period. 753 754 b. Noice runs 755 756 As between the ice runs, between the noice runs, the activation-and condensationally-757 enhanced counteraction outweighs the evaporation-induced decreases in CDNC, leading 758 to increases in CDNC with increasing aerosol concentrations (Figures 9a, 9b, and 10b). With temporal increases in CDNC, which are larger in the control-noice run than in 785 the low-aerosol-noice run, leading to those in CDNC differences between the noice runs, 786 there are temporal increases in condensation and evaporation, which are larger in the 787 control-noice run than in the low-aerosol-noice run, and thus in their differences between 788 the noice runs (Figure 9b). Associated with aerosol-induced smaller increases in 789 evaporation efficiency between the noice runs, aerosol-induced increases in condensation 790 are always greater than aerosol-induced increases in evaporation between the noice runs 791 during the initial stage (Figure 9b). This maintains aerosol-induced increases in updraft 792 dxdydz dV = mass fluxes between the noice runs and leads to aerosol-induced increases in WP between 793 the noice runs. In contrast to this, in the ice runs, after ~12:50 LST on January 12 th , aerosol-794 induced increases in condensation become lower than those in evaporation, leading to 795 aerosol-induced lower updrafts and WP (Figure 9a). This comparison between the ice and 796 noice runs confirms that the presence of ice processes and the associated WBF mechanism 797 plays a critical role in smaller aerosol-induced increases in condensation than in 798 evaporation in the ice runs. Figure  those runs than between the control and low-aerosol runs (Figures 9a and 9c). increases in the surface-to-volume ratio and the associated enhancement in the WBF-841 mechanism-related efficiency of evaporation are negligible as compared to those between 842 the control and low-aerosol runs. This contributes to aerosol-induced smaller increases in 843 evaporation between the control and IN-reduced runs than between the control and low-844 aerosol runs (Figures 9a and 9c). 845 Mainly due to the increase in evaporation, there is more negative buoyancy and 846 updraft mass fluxes start to reduce in the control run as compared to those in the IN-reduced 847 run around 12:50 LST on January 12 th (Figure 9c). Eventually, updraft mass fluxes in the 848 control run become smaller than those in the IN-reduced run around 15:50 LST on January 849 12 th (Figure 9c). This decrease occurs to a lesser extent mainly due to smaller aerosol-850 induced increases in evaporation between the control and IN-reduced runs than between 851 the control and low-aerosol runs (Figures 9a and 9c). Associated with weaker updrafts in 852 the control run, condensation in the control run becomes smaller than that in the IN-reduced 29 run around 15:50 LST on January 12 th but to a lesser degree as compared to that between 854 the control and low-aerosol runs (Figures 9a and 9c).  that increasing concentrations of aerosols as CCN increase cloud mass (Albrecht, 1989). 928 However, in contrast to this, this study shows that in the mixed-phase stratiform clouds, 929 the increasing concentration of aerosols as CCN can reduce cloud mass through the WBF 930 mechanism which involves efficient evaporation of droplets and inefficient deposition of 931 water vapor onto ice crystals. It is also shown that the increasing concentration of aerosols 932 as IN enhances cloud mass in contrast to roles of the increasing concentration of aerosols 933 as CCN in cloud mass. In addition, this study finds that the WBF mechanism reduces cloud 934 mass in the mixed-phase clouds as compared to that in warm clouds. Mid-latitude winter 935 stratiform clouds and high-latitude clouds such as the Arctic stratiform clouds frequently 936 involve ice particles as well as liquid particles and hence are affected by the WBF 937 mechanism and IN. As discussed in Stevens and Feingold (2009)