The Pennsylvania State University. The Graduate School. College of Earth and Mineral Sciences

Transcription

1 The Pennsylvania State University The Graduate School College of Earth and Mineral Sciences EFFECTS OF ICE NUCLEI CONCENTRATIONS, ICE NUCLEATION MECHANISMS AND ICE CRYSTAL HABITS ON THE DYNAMICS AND MICROPHYSICS OF ARCTIC MIXED-PHASE CLOUDS A Dissertation in Meteorology by Muge Komurcu 2011 Muge Komurcu Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy August 2011

2 The dissertation of Muge Komurcu was reviewed and approved* by the following: Jerry Y. Harrington Associate Professor of Meteorology Dissertation Advisor Chair of Committee Johannes Verlinde Associate Professor of Meteorology Marcelo Chamecki Assistant Professor of Meteorology Andrew M. Carleton Professor of Geography William H. Brune Professor of Meteorology Head of the Department of Meteorology *Signatures are on file in the Graduate School

3 iii ABSTRACT There is a significant warming in the Arctic that is evident in both observations and in the future climate predictions. The Arctic warming is greater than any other region on Earth, however, the degree of warming is inconsistent among the climate models even for the same emission scenarios. Clouds, especially low-level clouds, are a prevailing feature of the Arctic atmosphere. They strongly affect the surface radiative and energy budgets, which make them a key component of the Arctic climate. Recent inter-comparison studies using regional climate models show that models are incapable of reproducing the supercooled liquid water observed in clouds during the cold season. Large discrepancies exist in the partitioning of phase between ice and liquid water among different models. It is currently thought that these discrepancies are due to the uncertainties in ice nuclei concentrations, ice nucleation, and ice crystal habits used in models. Predicting these physical processes controls the partitioning between liquid and ice, and hence the impact of mixed-phase clouds on the surface energy budget. There is a need to improve model cloud predictions in the Arctic, however, the microphysical uncertainties mentioned above are tied directly to the cloud dynamics that help maintain persistent mixed-phase clouds. Therefore, this dissertation analyzes and inter-compares the impacts of different ice nuclei concentrations, ice nucleation mechanisms and ice crystal habits on mixedphase cloud dynamics. Separate simulations using different ice nuclei concentrations, ice nucleation mechanisms, and crystal habits are performed. It is found that the choice of habits in models alters the water paths and cloud dynamics strongly. Next, the relative

4 iv importance of and interactions among the processes that influence the dynamics of the cloud, such as the radiative cooling at cloud top, and the ice precipitation induced cloudbase stabilization are investigated. To examine these processes in detail, sensitivity studies are performed by fixing the radiative cooling, and the diabatic influences of ice precipitation. In addition, simulations with increasing ice nuclei concentrations, different nucleation mechanisms, and crystal habits are repeated with surface fluxes and largescale forcing included. The influence of surface fluxes is important as it can compensate for the water mass that is lost through ice precipitation if the ice precipitation is weak. Surface fluxes can also lead to the coupling of the liquid cloud layer with the sub-cloud layer. The cloud-base stability is diminished with the inclusion of the surface fluxes, and the effect of entrainment is enhanced. Sensitivity tests are also repeated with the added surface fluxes. Using the results of the sensitivity analysis, a ratio identifying the decoupling of the cloud and subcloud layers is generated, and also with the sensitivity analysis cloud dynamic and microphysical interactions within Arctic mixed-phase clouds are explained.

5 v TABLE OF CONTENTS LIST OF FIGURES vii LIST OF TABLES xvi ACKNOWLEDGEMENTS xvii Chapter I Introduction 1.1 The role of Arctic clouds in global climate change and prediction 1.2 Uncertainties in the numerical prediction of Arctic mixed-phase clouds Uncertainties related to ice nuclei concentrations and ice nucleation mechanisms Uncertainties related to ice crystal habits Impacts of the uncertainties in numerical prediction of Arctic clouds on cloud evolution and dynamics 1.3 Summary of the current knowledge of the dynamics of liquid clouds 1.4 Dynamics of Arctic mixed-phase clouds 1.5 Goals of this dissertation Chapter II Case Description, Numerical Model and Methodology 2.1 Case description 2.2 Description of the numerical model Parameterization of the ice nucleation mechanisms Parameterization of the ice crystal habits and ice growth Model setup 2.3 Methodology Chapter III Simulations without Surface Fluxes 3.1 Simulation results with increasing ice nuclei concentrations 3.2 Simulation results using different ice nucleation mechanisms 3.3 Simulation results using different ice crystal habits 3.4 Summary and discussion

6 vi Chapter IV Simulations with Surface Fluxes and Large Scale Forcing 4.1 Simulations using different ice nuclei concentrations 4.2 Simulations using different ice nucleation mechanisms 4.3 Simulations using different ice crystal habits 4.4 Summary and discussion Chapter V Crystal Habit Influence on Ice Growth and Cloud Dynamics 5.1 Analysis of habit influence 5.2 Influence of incorporating surface fluxes and large scale forcing 5.3 Summary Chapter VI Sensitivity Tests 6.1 Simulation setup Fixing the radiatve cooling Fixing the ice precipitation induced stabilization 6.2 Results of the sensitivity studies Effect of radiative cooling Effect of ice production and precipitation 6.3 Summary 6.4 Analysis of the simulations in chapter Influences of incorporating the surface fluxes in the simulations 6.6 Quantitative analysis of decoupling of the cloud and subcloud layers Applicability of the ratio Chapter VII Conclusions and Future Work 130 REFERENCES 137

7 vii LIST OF FIGURES Figure 1.1: Conceptual diagram of liquid stratiform cloud dynamics (Shubert et al., 1979). Figure 2.1: Analysis map of surface temperature, mean sea level pressure, and wind barbs for October 9 at 1200 UTC (Verlinde et al, 2007). Figure 2.2: Measurements of temperature, water content, mean droplet diameter and concentration from different measurement devices given in red, blue and black at 2145 UTC on October over Oliktok point, located on the North Slope of Alaska (Verlinde et al., 2007). Figure 2.3: Ice crystal growth habits (Libbrecht, 2005). Figure 2.4: Profiles of water and vapor mixing ratios and potential temperature for the initialization of the model simulations. Actual sounding taken at Barrow on 1700 UTC 9 October 2004 is also given (Klein et al., 2009). Figure 3.1 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with increasing IN concentrations of a factor of 1, 10, 25 and 50 times the original Arctic IN concentrations. Figure 3.2 Domain-averaged profile of the sum of pristine ice and snow concentrations (L -1 ) at the start of the 6 th hour given for a factor of 1 (black), 10 (blue), 25 (green), and 50 (red) times original Arctic IN concentrations Figure 3.3 Mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour given for original Arctic IN concentration (black), and a

8 viii factor of 50 times original Arctic IN concentration (red). Figure 3.4 (a) Potential temperature profiles at the start of the sixth hour of the simulations for Arctic IN concentration (black) and a factor of 50 times Arctic IN concentration (red). Initial profile is given by magenta. (b) Times series of ice precipitation at the surface (g/kg) for Arctic IN concentration (black) and a factor of 50 times Arctic IN concentration (red). Figure 3.5 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w in m 2 /s 2, (b) turbulence kinetic energy (m 2 /s 2 ), (c) buoyancy production of TKE (m 2 /s 3 ) and (d) shear production (m 2 /s 3 ) for original Arctic IN concentration (black), and a factor of 50 times original Arctic IN concentration (red). Figure 3.6 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with different ice nucleation mechanisms: Depositioncondensation nucleation (green), evaporation freezing (blue), evaporation IN (purple), liquid simulation with no precipitation (black) and glaciation simulation (red). Simulation with evaporation Freezing is performed with 10-9 of the evaporating droplets are assumed to be freezing. In evaporation IN simulation, 10-7 of the evaporating droplets are assumed to leave behind an IN. Glaciation simulation is performed with increasing the IN concentrations to 100 times the original Arctic concentrations. Liquid simulation is performed with ice phase and precipitation turned off. Figure 3.7 (a) Potential temperature profiles at the start of the sixth hour of the simulations for simulations with different ice nucleation mechanisms. Initial profile is given by magenta. (b) Times series of ice precipitation at the surface (g/kg) for simulations using different nucleation mechanisms. Figure 3.8 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with ifferent ice nucleation mechanisms and habits: Deposition-condensation nucleation (solid), evaporation freezing (dashed), evaporation IN (dashed dotted), reds are for simulations with hexagonal plates, and blues are for dendrites

9 Figure 3.9 (a) Potential temperature profiles at the start of the sixth hour of the simulations for simulations with hexagonal plates (red) and dendrites (blue) for deposition condensation nucleation. (b) Times series of ice precipitation at the surface (g/kg) for simulations using hexagonal plates (red) and dendrites (blue) for deposition condensation nucleation. 52 ix Figure 4.1 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with a factor of 1, 10, 25, and 50 times the original Arctic IN concentrations. Figure 4.2 Domain-averaged profiles of (a) the sum of pristine ice and snow concentrations and (b) aggregate concentrations (L -1 ) at the start of the 6 th hour given for a factor of 1 (black), 10 (blue), 25 (green), and 50 (red) times original Arctic IN concentrations. Figure 4.3 Mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour given for original Arctic IN concentration (black), and a factor of 50 times original Arctic IN concentration (red). Figure 4.4 (a) Potential temperature profiles at the start of the sixth hour of the simulations (b) Times series of ice precipitation at the surface (g/kg) for Arctic IN concentration (black) and a factor of 50 times Arctic IN concentration (red). Figure 4.5 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w in m 2 /s 2, (b) turbulence kinetic energy (m 2 /s 2 ) for original Arctic IN concentration (red), and a factor of 50 times original Arctic IN concentration (green). Figure 4.6 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with deposition-condensation nucleation (red) and evaporation IN with a fraction of 10-7 (blue). Figure 4.7 Domain-averaged profile of (a) the sum of pristine ice and snow concentrations and (b) aggregate concentrations (L -1 ) at the start of the 6 th hour given evaporation IN (blue), and

10 x deposition-condensation nucleation (red). Figure 4.8 (a) Potential temperature profiles at the start of the sixth hour of the simulations (b) Times series of ice precipitation at the surface (g/kg) for simulations with depositioncondensationn nucleation (red), and evaporation IN (blue). Figure 4.9 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with deposition-condensation nucleation using hexagonal plates (red) and dendrites (blue). Figure 4.10 Domain averaged profile of (a) the sum of pristine ice and snow concentrations (L -1 ) and (b) aggregate concentrations (L -1 ) at the start of the 6 th hour given for the depositioncondensation nucleation with hexagonal plates (red), dendrites (blue). Figure 4.11 (a) Potential temperature profiles at the start of the sixth hour of the simulations (b) Times series of ice precipitation at the surface (g/kg) for simulations using hexagonal plates (red) and dendrites (blue) for deposition-condensation nucleation. Figure 5.1 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from simulations with deposition condensation nucleation assuming dendrites with spherical fall speeds (black), spheres with dendrite fall speeds (blue), dendrites (green), and spheres (red). Figure 5.2 Domain-averaged profiles of (a) the sum of the pristine ice and snow concentrations (b) aggregate concentrations (L -1 ) at the start of the 6 th hour. Figure 5.3 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour

11 Figure 5.4 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). Figure 5.5 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. Figure 5.6 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) profile of potential temperature (K) and (d) time series of ice precipitation (g/kg). Figure 5.7 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. Figure 5.8 Domain-averaged (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) buoyancy production of TKE (m 2 /s 3 ), (c) potential temperature profile (K), and (d) time series of ice precipitation at the surface (g/kg) at the start of the 6 th hour. Figure 5.9 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. Figure 5.10 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). Figure 5.11 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from simulations with deposition condensation nucleation assuming dendrites with spherical fall speeds (black), spheres with dendrite fall speeds (blue), dendrites (green), and spheres (red) xi

12 Figure 5.12 Domain-averaged profiles of (a) the sum of the pristine ice and snow concentrations (b) aggregate concentrations (L -1 ) at the start of the 6 th hour. Figure 5.13 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. Figure 5.14 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). Figure 5.15 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th. Figure 5.16 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). Figure 5.17 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. Figure 5.18 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). Figure 5.19 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th. Figure 5.20 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg) xii

13 Figure 6.1 Time series of integrated in-cloud radiative cooling for simulations with (a) 30 W/m 2 (b) 130 W/m 2. Figure 6.2 (a) Profile of the actual mean change in θ il (b) A sample profile of the adjusted change in θ il based on a chosen cloud top and cloud-base. Figure 6.3 Domain-averaged profile of u u in time for constant latent in-cloud heating and below cloud cooling with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m 2. Figure 6.4 Domain-averaged profile of w w in time for constant latent in-cloud heating and below cloud cooling with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m 2. Figure 6.5 Domain-averaged profile of liquid water mixing ratio (g/kg) in time for constant latent in-cloud heating and below cloud cooling with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m 2. Figure 6.6 Domain-averaged profile of u u in time for strong cloud-base stabilization with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m Figure 6.7 Domain-averaged profile of w w in time for strong cloud-base stabilization with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m Figure 6.8. Domain-averaged profile of liquid water mixing ratio (g/kg) for strong stabilization in time for constant latent in-cloud heating and below cloud cooling with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m 2. Figure 6.9 Average TKE for the sixth hour of the simulations by keeping radiative cooling and cloud-base stabilization fixed, red is for strong cloud-base stabilization, blue is for weak cloudbase stabilization xiii

14 Figure 6.10 Domain-averaged profile of liquid water mixing ratio (g/kg) in time for (a) IN simulation and (b) 50IN simulation. Figure 6.11 Domain-averaged profile of w w in time for (a) IN simulation and (b) 50IN simulation. Figure 6.12 Domain-averaged profile of liquid water mixing ratio (g/kg) in time for simulations with (a) spheres and (b) dendrites. Figure 6.13 Domain-averaged profile of w w in time for simulations with (a) spheres and (b) dendrites. Figure 6.14 Domain-averaged profile of w w in time with a quarter of the observed surface fluxes and constant radiative cooling of 30 W/m 2 for (a) weak stabilization (b) strong stabilization. Figure 6.15 Domain-averaged profile of u u in time with a quarter of the observed surface fluxes and constant radiative cooling of 30 W/m 2 for (a) weak stabilization (b) strong stabilization. Figure 6.16 Domain-averaged profile of liquid water mixing ratio (g/kg) in time with a quarter of the observed surface fluxes and constant radiative cooling of 30 W/m 2 for (a) weak stabilization (b) strong stabilization. Figure 6.17 Domain-averaged profile of liquid water mixing ratio (g/kg) in time with surface fluxes of 130 W/m 2 and constant radiative cooling of 130 W/m 2 for (a) weak stabilization (b) strong stabilization. Figure 6.18 Domain-averaged profile of w w in time with surface fluxes of 130 W/m 2 and constant radiative cooling of 130 W/m 2 for (a) weak stabilization (b) strong stabilization xiv

15 Figure 6.19 Domain-averaged profile of u u in time with with surface fluxes of 130 W/m 2 and constant radiative cooling of 130 W/m 2 for (a) weak stabilization (b) strong stabilization. Figure 6.20 Hourly averaged magnitudes of the vertical circulations for the simulation with constant 50 W/m 2 surface fluxes, 50 W/m 2 radiative cooling, and a strong cloud-base stabilization. Figure 6.21 Time series of the ratio of the magnitude of the vertical circulations in-cloud to below cloud for weak (blue) and strong (red) cloud-base stabilization (a) 3 to 12 simulation hours (b) a close up to the 5 to 10 simulation hours. Figure 6.22 Hourly averaged magnitudes of the vertical circulations for the simulation with constant 100 W/m 2 surface fluxes, 100 W/m 2 radiative cooling, and a strong cloud-base stabilization. Figure 6.23 Time series of the ratio of the magnitude of the vertical circulations in-cloud to below cloud for weak (blue) and strong (red) cloud-base stabilization (a) 3 to 12 simulation hours (b) a close up to the 5 to 8 simulation hours. Figure 7.1 Illustration of the influence of cloud-top radiative cooling on cloud dynamics. Figure 7.2 Illustration of the influence of diabatic influences of precipitation, weak and strong cloud-base stabilization on cloud dynamics xv

16 xvi LIST OF TABLES Table 2.1: Ice crystal habit parameterization Table 6.1: Magnitude of the radiative cooling and ice precipitation for IN and 50 IN simulations Table 6.2: Magnitude of the radiative cooling and ice precipitation for simulations with spheres and dendrites

17 xvii ACKNOWLEDGEMENTS First of all, I would like to thank my academic advisor Jerry Y. Harrington for his incredible mentoring, continuous help and support. I will be forever grateful to have the opportunity to work within the research group of Drs. Jerry Y. Harrington, Johannes Verlinde and Eugene Clothiaux. This opportunity helped me gain many qualities that I will pass on to my students in the future. I would also like to thank my committee, Dr. Marcelo Chamecki and Dr. Andrew Carleton for their comments, and contributions, which helped me gain a broader perspective. Furthermore, I am indefinitely indebted to Dr. Alexander Elkov Avramov, who kindly helped me get started with my studies and always provided very supportive guidance. I would like to thank Dr. Jonathan L. Petters for his help with RAMS and for always providing honest advice on graduate school related issues. I would like to acknowledge my dearest friends, Drs. Maria Herrman, Maria Cazorla, Martha Butler, and Fangxing Fan, for their incredible friendship and support during my studies and our graduate life together. Moreover, I would also like to thank my officemates Chengzhu Zhang and Guo Yu for always being cheerful and bringing sunshine to even the most stressful days.

18 xviii I would like to acknowledge my family, my late grandfathers army Colonel Recep Celebi, and police lieutenant Muhittin Komurcu, my late grandmothers Emine Celebi and Nadire Komurcu, who have taught me to be kind, truthful, honest, determined and brave at every second of my life. Furthermore, I would also like to thank my late father Ismail Komurcu, who has given me the sense of humor, especially the humor I needed to see in situations of life when I tended to take things too seriously, and most importantly for a can do attitude. I would like to thank my mother Sibel Komurcu, for putting up with me throughout the stressful graduate school life, and for her support and continuous prayers, without which I could not have succeeded. Finally, I would like to thank my husband Zikri Bayraktar, who is the perfect example of hard work, integrity and honesty, for making my life meaningful and happy. Before I conclude, I would like to finally acknowledge an incredible person, whose life has helped me develop my character and my choices: the founder of the modern Turkish Republic, Mustafa Kemal Ataturk, for teaching me to be determined, to work hard for my ideals, and to not rest until accomplishment.

19 CHAPTER I INTRODUCTION Clouds play an essential role in the radiation budget of the atmosphere through their absorptive, emissive, and reflective properties. Shortwave radiation is weakly absorbed and strongly reflected by clouds. Longwave radiation, on the other hand, is strongly absorbed when cloud liquid water paths are high. With increased absorption, the amount of energy that escapes to space is reduced, which makes clouds a major contributor to the surface energy budget (Cotton and Anthes, 1989, p.161). In the Arctic, many unusual cloud types are observed throughout the year. In summer, clouds systems can exist in multiple cloud layers. In the transition seasons between summer and winter mixed-phase clouds are commonly observed. Highly stable low-level ice clouds, and clear sky ice precipitation occur frequently during winter. Winter also brings plume clouds originating from the cracks in sea ice (Curry et al., 1996). Clouds occur routinely in the Arctic atmosphere, as a result they can have a strong influence on the radiative component of the surface energy budget (Pinto, 1998). The radiative properties of clouds are determined by their microphysical, vertical and horizontal structure. It is important to analyze Arctic clouds during the transition and winter seasons, since these clouds are most dominantly mixed-phase (includes both ice and liquid water) with significant amounts of supercooled liquid water. Because Arctic mixed-phase clouds can persist for long periods of time, cloud liquid phase can affect the radiation and surface energy budgets considerably. Cold

20 2 season Arctic mixed-phase clouds can exist either as a single cloud layer or as multiple layers of clouds (e.g. Pinto, 1998). Single cloud layers are more frequently observed compared to the multiple layer clouds (Klein et al., 2009). In this chapter, the importance of Arctic Clouds in terms of global climate and prediction of the future climate change will be explained. Uncertainties in model predictions of Arctic clouds related to cloud microphysics will be introduced, and the current knowledge of cloud dynamics will be summarized. At the end of the chapter, the goal of this study will be presented. 1.1 The Role of Arctic Clouds in Global Climate Change and Prediction The importance of Arctic regions, and Arctic clouds emerged when climate model results revealed that doubled CO 2 concentrations lead to an increased warming in the Arctic as the sea ice retreats (Curry et al., 1996). An increasing trend in surface air temperatures in the Arctic is evident in both observations (Hansen et al., 2010), and in the model predictions of the future Arctic (ACIA, 2004). Vavrus et al. (2010) investigate the future Arctic using a climate model coupled with the underlying surface, to evaluate how the Arctic will respond to a warming and to understand the mechanisms responsible for it. The model uses observed concentrations of greenhouse gases, solar and volcanic forcing from 1870 to CO 2 concentrations are then forced to rise to 720 ppm until 2100, and aerosol concentrations to rise until mid 2020s after which they are forced to decline. They

21 3 find that cloudiness especially in the form of low-clouds is enhanced mostly during the autumn season, when evaporation is increased due to an amplified sea ice melting. During this time, the increased liquid water content of the clouds leads to a positive feedback in the melting of the sea ice, where enhanced cloud radiative forcing at the surface seems to prevent the formation of new sea ice and accelerates the loss of existing sea ice. A potential relationship between sea-ice and cloud cover is also seen in the observations. During the summer of 2007, sea ice extent was very low, marked by a minimum in September of the period, which coincides with an increase in total cloud cover that was observed in the Arctic during the autumn of 2007 (Levinson and Lawrimore, 2008). Levinson and Lawrimore (2008) suggest that although this increase could be linked to the synoptic situation, such as the existence of a low-pressure system along with the relatively higher surface temperatures, it could also be an indication of a potential link between the reduced sea-ice extent and the increased cloudiness. During the autumn of 2007, low-level clouds tended to show a lower cloud-base, and an increase in the amount of mid-level clouds occurred, which could have been confused with the increased vertical extent of the low clouds in response to the reduced atmospheric stability over melted sea ice (Levinson and Lawrimore, 2008). The reduction in stability near the surface is apparent in the composite of vertical temperature profile at times when sea ice extent is one standard deviation below its mean (at times of intense melting) during autumn (September, October and November) for the period (Schweiger et al., 2008).

22 4 The reduced surface stability over melted sea ice can increase the cloudiness. Increased cloudiness results in increased downwelling long wave radiation, which can potentially increase surface sea-ice melting or slow down the growth of lost sea ice (Francis and Hunter, 2006). Using a regression analysis, Francis and Hunter (2006) show that the variances in maximum sea ice retreat anomalies before the onset of the anomalies are primarily driven by the changes in downwelling long wave forcing from 1979 to Other parameters are also investigated in the regression analysis such as the winds, and the advection in the surface layer. Analysis of the forcing related to these parameters also reveals the important contributions to the melting and the thinning of sea-ice through advection caused by large-scale circulations (such as a positive phase of the North Atlantic Oscillation (NAO)) and the transport of warm and moist air from southern latitudes. It is also suggested that when ice is thinned through the effects of advection and winds, the sea-ice surface is more vulnerable to changes in other parameters in the energy budget such as the long-wave forcing. Francis et al. (2009) put forward that the reduction in stability through warming of the surface during the summers of increased sea ice melting also affect the large-scale circulation patterns, and produce conditions similar to a neutral or negative NAO phase following winters. This circumstance arises from the fact that at lower sea ice extents geopotential height surfaces are increased, leading to a weaker temperature gradient between mid-latitudes and the poles. All of the studies mentioned hint at a potential relationship between sea ice extent and cloud cover, however, with current data and

23 5 models, it is unclear whether and how the Arctic sea ice and clouds feedback on each other, and whether and how the feedback will link to or modify the lower frequency variability. 1.2 Uncertainties in the Numerical Prediction of Arctic Mixed-Phase Clouds Even though both observations and models reveal a warming in the Arctic, future climate predictions with different global climate models using even the same greenhouse gas emission scenario exhibit strikingly different amounts of warming (ACIA, 2004). In addition, both regional climate and weather models produce results that are highly inconsistent among them, and that do not match the observations in the Arctic (Curry et al. 1996, Klein et al. 2009). Intercomparison of the results of regional climate model simulations in the Arctic reveal that the model errors in the downwelling longwave radiation are negative, and scatter in the magnitude of the error among models is large during the cold season (Tjernstrom et al., 2008). A comparison of model predicted and observed liquid water paths show a lack of liquid water in model predictions, which is the reason for the large negative errors in downwelling longwave radiation. Curry et al. (1996) suggest that the parameterizations used in climate models are not suitable for the Arctic climate, therefore new parameterizations that involve the processes taking place specifically in the Arctic should be developed.

24 6 Due to the relative lack of observations in the Arctic, our understanding of physical and dynamical processes taking place is limited, which also limits the creation of parameterizations specific to the Arctic. For instance, few publications exist in the literature that attempt to understand Arctic cloud-scale and mesoscale dynamics (e.g. Harrington and Olsson, 2001). In an attempt to increase knowledge of mixed-phase clouds, which frequently occur in the Arctic atmosphere during the cold season, the Mixed Phase Arctic Cloud Experiment (MPACE) was undertaken during September to October While many cloud cases were observed, two representative observational periods, Case A, and Case B are of interest here. Both cases are of persistent mixed-phase clouds. Multiple layers occur in Case A, but only a single layer occurs in Case B. Recent intercomparison model studies of both Case A and Case B show that model results do not match the observations (Klein et al., 2009; Morrison et al., 2009), and the biggest spread in model results occurs in the simulated water contents. One possible way to overcome the mismatch between observations and modeled water contents is to include more sophisticated cloud microphysics parameterizations. Although model simulations generally tend to improve when more detailed cloud microphysics schemes are used, some models with detailed microphysics still produce poor results (Klein et al., 2009). This is consistent with the previous intercomparison study by Tjernstrom et al. (2008), where even the models with more detailed microphysics did not do well in predicting cloud water during the cold season in the Arctic. These results imply that in order to improve model predictions in the Arctic, it is important to identify the processes responsible for mixed-phase cloud development and maintenance.

25 1.2.1 Uncertainties Related to Ice Nuclei Concentrations and Ice Nucleation Mechanisms 7 Prenni et al. (2007) propose that a reason for the mismatch between observed and modeled cloud water contents could be the treatment of aerosols, specifically ice nuclei, and ice nucleation parameterizations in models. They were able to attain better agreement with observations when they limited the ice nuclei (IN) activation parameterization of Meyers et al. (1992) based on Arctic IN observations, and when they included ice nuclei advection, diffusion, and depletion through precipitation. Furthermore, observations reveal that at times ice crystal number concentrations are much greater than the number concentrations of ice forming nuclei (Hobbs, 1969). This problem is an outstanding one in cloud physics, and remains so to this day. The classical heterogeneous nucleation mechanisms (i.e. depositional, condensational, immersion, and contact freezing) are not always able to simultaneously account for ice concentrations and water paths. Therefore, other innovative mechanisms have been put forward to explain the higher ice crystal concentrations. These mechanisms include ice splinter production during riming, which is thought to occur between -5 and -7 C, fragmentation of ice crystals (Hobbs and Rangno, 1985), and some relatively controversial nucleation mechanisms such as evaporation IN (Rosinki and Morgan, 1991), and evaporation freezing (Cotton and Field, 2002). These same theories have also been incorporated into models to allow for a better match with observed water paths. For example, by using evaporation IN and evaporation freezing,

26 Fridlind et al. (2007) overcome the mismatch between the modeled ice concentrations and the observed values for the MPACE cases Uncertainties Related to Ice Crystal Habits Recent findings by Avramov and Harrington (2010) show that the choice of ice crystal habit in models also leads to large differences in the partitioning of phase between liquid water and ice. According to their study, large differences in water paths are simulated depending on the general habit assumed in the model such as plates, columns, and spheres. Large differences also exist among the simulated water paths within a certain habit category. These differences arise because habits are represented in models through mass-dimensional, and velocity-dimensional power laws. Crystal habits determine the fall velocity of ice crystals, as a result they also determine the in-cloud residence time of an ice crystal. The more a crystal stays in cloud, the more it can grow depleting the liquid. Moreover, habits also determine the mass growth rate of ice crystals through these mass and velocity parameterizations. The combined impact of simplified habits on simulated clouds is to strongly alter the predicted water paths.

27 1.2.3 Impacts of the Uncertainties in Numerical Prediction of Arctic Clouds on Cloud Evolution and Dynamics 9 Although many ice nucleation mechanisms including evaporation IN and evaporation freezing have been widely used in modeling studies, their impacts on the dynamics of clouds have not yet been explored. These impacts could be essential in the following way. Each ice nucleation mechanism leads to ice formation under certain conditions. For instance, evaporation freezing leads to ice formation when cloud droplets are evaporating. Consequently, ice formation by evaporating drops would occur preferentially in downdrafts and at the entrainment interface, which could alter the cloud dynamics through latent heating effects on buoyancy (Rosinki and Morgan, 1991). Because the equilibrium vapor pressure over ice is smaller than equilibrium vapor pressure over liquid water at temperatures below 0 C, ice grows while liquid is evaporating (Bergeron process). Since different ice nucleation mechanisms produce different numbers of ice crystals, liquid water contents vary with different nucleation mechanisms altering the latent heating impacts. Cloud radiative properties can also change correspondingly due to the change in the water content of the clouds, thus altering the cloud top radiative cooling that drives buoyant downdrafts. Through these complex interactions amongst latent heating, buoyancy, radiation, and the number of ice crystals formed, the preferred region of ice formation may cause cloud evolution and dynamics to change in comparison to other nucleation mechanisms.

28 10 Crystal habits may also impact the dynamics of simulated clouds. Because different habits possess different growth rates, the buoyant effects of different habits could be substantial. There are many processes acting in the mixed-phase clouds that affect their dynamics, such as entrainment, latent heating, and radiation. Entrainment can bring warmer and drier air from above cloud top into the cloud layer, which may lead to the evaporation of drops at cloud top, and thus a latent cooling (Schubert et al., 1979). Moreover, latent heating in-cloud and cooling below cloud through ice precipitation can result in the stabilization of the cloud layer (e.g., Harrington et al., 1999). As a result, latent heating processes affect the buoyancy in clouds both directly and indirectly. In addition, radiative cooling due to the existence of liquid at cloud top produces positively buoyant parcels that sink, leading to vertical motions, and mixing, therefore generating Turbulence Kinetic Energy (TKE) (e.g. Pinto, 1998). All these processes are dependent on the amounts of liquid water and ice in some manner. Furthermore, the amounts of liquid and ice in clouds are strictly affected by the amount of ice nuclei, the ice nucleation mechanism, and the ice crystal habit (Avramov and Harrington, 2010). Therefore it is important to understand how changes in these mechanisms influence cloud microphysical processes to alter cloud dynamics and evolution. In order to understand cloud dynamics, a summary on the current knowledge on liquid clouds will be presented next followed by mixed-phase cloud dynamics.

29 Summary of the Current Knowledge of the Dynamics of Liquid Clouds Understanding the dynamics of clouds has been a challenging area of study for several decades. Numerous types of models have been developed to explain and predict the various processes acting within liquid stratocumulus clouds. One example is the mixed layer models. Although mixed layer models are relatively simple, and their applicability is restricted to well mixed layers, they are useful in examining fundamental interactions. One should note that some crucial mechanisms are ignored due to the simplified nature of these models (Cotton and Anthes, 1989, p.328). For example, these models do not resolve turbulent fluxes explicitly; instead they specify fluxes at boundaries and make assumptions about the flux profiles in the boundary layer. Moreover, simplifying assumptions are also made in the turbulent kinetic energy (TKE) equation, and turbulence closure is obtained through relating buoyant fluxes in the mixed layer to entrainment velocity (Nicholls, 1984). The classical paper on mixed-layer models is Lilly s (1968) study, in which he developed a mixed layer model to examine a stratocumulus layer under a strong subsidence inversion. Lilly s paper is significant because he provided a relatively simple set of equations for modeling the effects of cloud dynamics using an energy balance and entrainment assumptions for closure. Schubert (1976) re-developed Lilly s equations, improving the entrainment closure and a more detailed radiation scheme. Schubert et al. (1979) explains the basic mixed-layer dynamics of stratocumulus as encapsulated in Figure 1.1.

30 12 Figure 1.1 Conceptual diagram of liquid stratiform cloud dynamics (Shubert et al., 1979). Conceptually, imagine that rising air condenses at the Lifting Condensation Level (LCL), and latent heating occurs. This warming adds buoyancy to the air (w'θ'>0). When the cloudy air reaches the inversion, its momentum is redirected horizontally through continuity. Some of the rising air, however, proceeds into the inversion because of its inertia. This process can cause the warmer and drier air in the inversion layer to entrain into the cloud layer. This warm and dry air leads to evaporation of the surrounding cloud droplets, and latent cooling takes place. If this evaporative cooling is greater than the entrainment warming, a positive heat flux (w'θ'>0) occurs causing parcels to sink. In addition, radiation leads to a cooling in this region, so that the net effect of all processes at cloud top can be a positive heat flux (w'θ'>0). As the cool air moves downward, adiabatic compression causes it to become warmer than its surroundings even though cloud droplets evaporate. Because entrainment dries the parcel, the downward moving air reaches its cloud-base at a higher altitude than the cloud-base of an updraft.

31 13 Consequently, the downdraft is potentially warmer than the air in the subcloud below. This causes a piling up of air which builds high pressure perturbation. The potentially warmer downdraft air is then forced to move down by the high pressure resulting in a negative heat flux (w'θ'<0). Here, turbulent kinetic energy (TKE) is consumed as the cloud layer must do work against the subcloud layer. Similarly, near cloud-base where latent heating is taking place in the updrafts, a negative pressure perturbation occurs. This low pressure causes air at the surface to move upward (w'θ'>0). The proposed circulation mechanism is self-maintaining, therefore a cloud can maintain itself for a considerable amount of time. Schubert s explanation is a well-established one, though it ignores the possible decoupling of the cloud and subcloud layers, and it does not consider the effects of drizzle. Decoupling of cloud and subcloud layers is commonly observed in nature (e.g. Turton and Nicholls, 1987). When warmer air is forced down into a cooler layer, TKE is consumed, because mechanical work must be done. This TKE consumption can only take place when sufficient energy production occurs in the cloud layer. If the production of TKE is scarce, then the increased pressure, as hypothesized by Schubert, cannot be maintained by circulations that extend from the surface to cloud top. Therefore the cloud and subcloud layers tend to separate with a slight temperature inversion forming, at times, between the two layers (Turton and Nicholls, 1987). Stevens et al. (1998) investigated the effect of drizzle on the dynamical structure of stratocumulus using a large eddy simulation (LES) model. They found that drizzle stratifies the boundary layer

32 14 in the following manner. The formation of drizzle-sized drops, and their sedimentation leads to latent warming of the cloud layer. As drizzle precipitates, it evaporates in the subcloud layer, which leads to a cooling and moistening below cloud-base. As a result, a colder subcloud layer underlies a warmer cloud layer, and a stable stratification is obtained. The effect of drizzle formation therefore reduces the TKE by inhibiting the turbulent mixing between the cloud and subcloud layers. In their case, the cooling and moistening of the subcloud layer could also lead to enhanced surface sensible heat fluxes, which promotes cumulus convection. This then could cause the cloud and subcloud layers to re-couple. They claim, however, that if surface temperatures were colder, this effect would possibly not have occurred, and decoupling would have taken place. 1.4 Dynamics of Arctic Mixed-Phase Clouds In the case of mixed-phase clouds where liquid and ice coexist, there are additional processes acting that may affect the dynamics. For example, at temperatures below 0 C the equilibrium vapor pressure over ice is less than the equilibrium vapor pressure over liquid water. Therefore, liquid water evaporates while ice grows and this mechanism is known as the Bergeron process (e.g. Pruppacher and Klett, 1997, p.6). Depending on the conditions, this mechanism may deplete the liquid causing cloud dissipation through ice precipitation (e.g. Harrington et al., 1999). Because liquid water dominates the cloud top radiative cooling (e.g. Pinto, 1998, Harrington and Olsson, 2001), the loss of mass through ice precipitation reduces cloud-top radiative cooling rates and leads to weaker

33 15 cloud-scale circulations (e.g. Harrington et al., 1999). Ice particles can grow large enough by vapor diffusion to rapidly fall out of the layer leading to a net latent heating of the cloud. Just as in the case of drizzle from liquid-only clouds, falling ice can sublimate in the subcloud layer, leading to a cooling (e.g. Pinto, 1998) which can cause decoupling. The cooling and moistening effect of ice precipitation is much greater compared to liquid-only clouds (e.g. Harrington and Olsson, 2001). Moreover, ice crystal habits are also important, as they determine the growth rates and the fall speeds of ice crystals, and in turn the in-cloud residence times (e.g. Harrington and Olsson, 2001). Based on observations of autumnal Arctic mixed-phase clouds, Pinto (1998) suggests that a strong radiative cooling that provides liquid water formation, along with IN removal through ice precipitation and the lack of IN sources could allow for persistent clouds. If this is true, then the concentration and type of the ice nuclei are of prime importance for mixed-phase clouds. In order to examine the effects of increasing IN concentrations, Harrington et al. (1999) performed simulations by cooling a summertime Arctic stratus sounding by 5ºC and 10ºC to obtain mixed-phase clouds. They found that at lower temperatures, a mixed-phase cloud could be maintained only with very low IN concentrations. For the colder simulation, the cloud top attained the temperature where the Bergeron process maximizes (T ~ -15ºC). As a result, liquid water was strongly depleted, yielding a much lower radiative cooling. The reduced radiative cooling means a reduction in buoyant energy that

34 16 can drive and maintain vertical circulations. In addition, the increased precipitation due to strong crystal growth produced a latent warming of the cloud layer and cooling of the subcloud layer. The net result was to stabilize the cloud layer shutting off the circulations and therefore the moisture supply to the cloud layer. Because ice growth also depends on the crystal habit, the sensitivity to ice crystal habit was also investigated. Ice spheres were used to contrast with the plates used in the previous 10ºC-cooled simulation. Spherical ice grows more slowly and rapidly precipitates, consequently reducing the latent heating of the cloud layer. Thus cloud stabilization with respect to the subcloud layer is reduced. Using spherical ice allowed liquid to be maintained in the cloud layer, and radiative cooling dominated the circulations. As a consequence, a colder mixed-phase cloud cannot be maintained unless a reduction in ice formation and growth occurred with decreasing IN concentrations, or ice crystals precipitate rapidly. Harrington and Olsson (2001) examined the impacts of IN concentrations on surface forced mixed-phase Arctic clouds. They compared simulations of liquid-only clouds with mixed-phase simulations using different IN concentrations. Moreover, they developed a method where IN concentrations are depleted as IN are activated. In comparison with the liquid only case, they demonstrated that TKE in the mixed-phase cloud was lower, and that large reductions in TKE occurred at times when there was significant ice precipitation. Sensitivity tests with surface fluxes held constant showed that oscillations in TKE and ice water path (IWP) were reduced significantly. This result suggested that ice precipitation caused a feedback with the surface fluxes, which produced the

35 17 oscillations in TKE. It seems that ice precipitation in mixed-phase clouds influences the vertical motions primarily through stabilization by precipitation in a similar manner to Stevens et al. (1998). The process occurs in the following manner: Ice formation mostly occurs in the updrafts, and latent heating occurs there because ice precipitation out of the updrafts causes water contents to be reduced. Therefore less evaporation occurs in downdrafts leading to less evaporational cooling, which causes a net heating of the cloud layer and stable stratification. This stratification reduces the strength of the vertical motions, and convective activity. Air that converges in the downdraft regions is then forced to sink mechanically, consuming TKE. The reduction in TKE also results in the reduction in the momentum fluxes to surface, and surface winds. The temperature difference between the air ocean interface is also reduced, all of which result in the reduction of surface sensible heat fluxes, and in turn in the TKE. Reductions in IN concentrations without IN depletion causes liquid water paths to increase because of the reduction in the number of ice crystals formed, which slows the Bergeron process. However, in time, the cloud slowly glaciates. When IN are depleted through precipitation, fewer ice crystals, and less ice precipitation occur. Therefore boundary layer convection is not suppressed as much and mixed-phase clouds persist for the duration of the simulations. All of these studies suggest that IN concentrations, ice formation mechanisms and crystal habits are critical to evolution of the cloud structure and dynamics. Additionally, surface

36 fluxes and the supply of moisture are also important for cloud dynamics and the maintenance of the cloud Goals of This Dissertation Despite the common use of ice nuclei concentrations, nucleation mechanisms and crystal habits in models, the effects of these parameters on cloud dynamics are not known, and their possible feedbacks with cloud processes have not yet been examined. The goal of this dissertation is to explore how each of these mechanisms affects cloud structure and cloud dynamics to produce long-lived mixed-phase clouds in the Arctic. Another major goal of this dissertation is to identify the relative degree of significance of the physical processes that maintain these clouds, such as radiation, latent heating, surface fluxes, and cloud top entrainment, and their feedbacks to dynamics under different ice nuclei concentrations, ice nucleation mechanisms and ice crystal habit settings. Consequently, the main questions that seek answers in this study are: - How do Arctic Mixed-Phase Clouds operate? - What are the feedbacks between cloud microphysics and cloud dynamics that lead to persistent mixed-phase clouds? Because at Arctic temperatures ice grows at the expense of liquid water, the amount of

37 19 liquid is restricted by the ice formation. Therefore, to answer these questions, one needs to know how changes in microphysics (such as the amount of liquid water) due to changes in ice crystal concentrations, ice nucleation mechanisms, and ice crystal habits alter cloud dynamics. Sensitivity studies are used to ascertain the relative importance of the processes leading to mixed-phase cloud persistence. To explore these ideas, the Regional Atmospheric Modeling System developed at Colorado State University hereafter (RAMS), is used in the large eddy simulation (LES) mode. The single layer mixed-phase cloud case, Case B, of the Mixed Phase Arctic Cloud Experiment (MPACE) is the focus of this study. This dissertation is organized as follows. The methods and analysis used in this study are explained in Chapter 2. In Chapter 3, analysis of the simulations with increasing IN concentrations, different nucleation mechanisms and crystal habits is presented without the addition of surface fluxes or large scale forcing. The simulations incorporating surface fluxes and large scale forcing with different ice nuclei concentrations, nucleation mechanisms, and crystal shapes are discussed in Chapter 4. Chapter 5 is devoted to the analysis of the sensitivity of mixed-phase cloud microphysics and dynamics to ice crystal habit. In Chapter 6, sensitivity of the mixed-phase cloud dynamics to radiative cooling and ice precipitation induced cloud-base stabilization is investigated. Conclusions and future work are given in Chapter 7.

38 20 With this study, the mixed-phase cloud dynamics and microphysics are explored, and a conceptual model of Arctic mixed-phase clouds is presented. Improving our understanding of the cloud microphysical and dynamical processes is not only crucial for accurate modeling of Arctic clouds, but is also significant for the accurate estimates of global energy budgets in future models.

39 21 CHAPTER II CASE DESCRIPTION, NUMERICAL MODEL AND METHODOLOGY This chapter provides a thorough description of the observed Arctic Mixed-Phase cloud case that is used to initialize the numerical model simulations, the numerical model, the methods of analysis and the setup of the numerical simulations. 2.1 Case Description Over the past ten years, an increasing number of observations have been made in the Arctic atmosphere to help evaluate models and to reduce the scatter among model results. One of these observational experiments is the Mixed Phase Arctic Cloud Experiment (MPACE), which took place during October of MPACE had many observational case periods containing mixed-phase clouds, for example, during Case A multiple layers of mixed phase clouds were detected. During Case B, a single layer mixed-phase cloud was observed. This study will concentrate on the single layer case (Case B), because cloud dynamic processes in even single layer clouds are not well understood (e.g. Shupe et al., 2008). The focus of the study is from 17UTC on 9 October to 5UTC on 10 October

40 , when mixed-phase clouds occurred expansively. During this time, a high pressure center was located to the north of Alaska, which brought flow from the sea ice to the open ocean, leading to strong surface heat fluxes and enhanced cloud production. A surface analysis map showing a low-pressure center and a developing high-pressure center over the open ocean is given in Figure 2.1. Figure 2.1 Analysis map of surface temperature, mean sea level pressure, and wind barbs for October 9 at 1200 UTC (Verlinde et al, 2007). Because of the enhanced surface forcing through surface fluxes, the boundary layer was well mixed with depth varying between m. The mixed-phase cloud was capped with a 2K inversion, and a dry, cloudless atmosphere. The mixed-phase cloud had more liquid water (about 115 g/m 2 ) than ice water content (about 7 g/m 2 ). Surface fluxes were obtained from the ECMWF reanalysis, and were W/m 2 for sensible heat flux and W/m 2 for latent heat flux. In addition, ice precipitation occurred continuously

41 and in bursts below cloud-base, and reached the surface (Klein et al, 2009). Examples of the cloud properties for a few hours after Case B are given in Figure Figure 2.2 Measurements of temperature, water content, mean droplet diameter and concentration from different measurement devices given in red, blue and black at 2145 UTC on October over Oliktok point, located on the North Slope of Alaska (Verlinde et al., 2007). The observed cloud-base was around 800m, while cloud top was around m during 9-11 October Streaks of ice precipitation and/or drizzle were observed in the cloud layer, while ice precipitation occurred below cloud-base. Cloud top temperature was around C with a liquid water content of 0.36 g/m 3 at cloud top. Although cloud properties were similar during this time, the number concentration varied significantly among different days, and even during the flights of a particular day suggesting a role of synoptic influence such as the weakening or strengthening of the pressure systems

42 24 (Verlinde et al., 2007). 2.2 Description of the Numerical Model The regional Atmospheric Modeling System developed at Colorado State University (RAMS) is used for modeling purposes. This model has eight categories of water; vapor, cloud droplets, rain, pristine ice, snow, graupel, hail, and aggregates. Pristine ice constitutes small ice crystals, and is only allowed to grow through vapor deposition. Snow, which is larger pristine ice crystals, can grow through vapor deposition as well as some moderate riming. Aggregates form through the collection of pristine ice, snow and other aggregates. Aside from pristine ice, snow and aggregates, which are all assumed to be completely ice, graupel and hail are mixed phase particles (include both ice and liquid). All hydrometeor categories are represented as complete gamma distribution functions (Walko et al., 1995). For each category, mixing ratio and number concentration are predicted, except for cloud droplets where a constant number concentration is assumed and only the mixing ratio is predicted Parameterization of the Ice Nucleation Mechanisms Ice crystal formation can take place both homogeneously, and heterogeneously. Homogeneous nucleation is the process where a liquid water droplet freezes spontaneously when temperatures fall below -40 C (Young, 1974). Although the cases

43 25 simulated in this study have much higher temperatures, homogeneous freezing nucleation is included in RAMS and its parameterization is based on the work of Demott et al. (1994). Heterogeneous nucleation requires the aid of a medium to initiate ice crystal formation (an aerosol), named as an ice forming nucleus (IN). Heterogeneous nucleation mechanisms are typically categorized into four main classes: deposition nucleation, condensation nucleation, contact freezing, and immersion freezing. Deposition nucleation occurs when an ice crystal forms as water vapor deposits directly as ice on a solid ice nucleus. In contrast, condensation nucleation occurs when vapor condenses as liquid on an ice nucleus and then freezes (Young, 1974). As expected from their definitions, in nature it is hard to differentiate between the deposition and condensation nucleation mechanisms (Meyers et al., 1992) though there is evidence for it (e.g. Pruppacher and Klett, 1997, pg. 315). Parameterization of deposition-condensation nucleation is based on Meyers et al. (1992), where a certain number of IN are activated based upon the supersaturation: N exp( a bs ) (2.1) id In equation 2.1, N id is the number of active ice nuclei per liter, S i is the supersaturation with respect to ice, and a and b are empirical coefficients. The coefficients a and b were initially derived from observations in mid-latitudes, where IN concentrations are much larger compared to Arctic IN concentrations. Prenni et al. (2007) suggested that the Meyers parameterization should be adjusted for the Arctic calculations, using the IN i

44 measurements taken in the Arctic. Therefore, in this study, coefficients a and b are 26 adjusted to match the observed Arctic IN concentrations of 0.15L -1 at 1% of supersaturation. Contact freezing takes place when a supercooled liquid water droplet comes into contact with an ice nucleus (Young, 1974). Mechanistically, contact freezing involves the contact of ice nuclei by a droplet through Brownian motion, diffusiophoresis and thermophoresis (Young, 1974). Parameterization of contact freezing in RAMS follows Cotton et al. (1986), where the contact IN activation is based on the Meyers et al. (1992): N exp m n( T ) (2.2) ic In equation 2.2, N ic is the number of IN available for contact nucleation, and T c is the cloud droplet temperature, where m and n are empirical coefficients. Although there are no observations of contact IN in the Arctic, in application of the contact nucleation parameterization in RAMS for simulations presented in this work, coefficients m and n are also adjusted assuming that contact IN concentrations are reduced similarly to deposition condensation IN to account for Arctic conditions. c Another mechanism similar to contact nucleation is immersion freezing whereby the freezing takes place on an ice nucleus immersed inside the droplet. The process takes a longer time than contact freezing (Young, 1974) and is dependent on the mass of the droplet and the degree of supercooling (Bigg, 1953). It is represented by the following equation:

45 dn f ( m, t) m Nw( m, t) Aexp[ B( T 0 T)] (2.3) dt w Here, N f is the number of frozen droplets with a certain mass, and A and B are empirical coefficients. 27 In addition to the classical nucleation mechanisms presented above, two additional mechanisms have been proposed, namely evaporation freezing (Cotton and Field, 2002) and evaporation IN (Rosinki and Morgan, 1991). These mechanisms have been incorporated into models in part because simulations with classical nucleation mechanisms alone do not match the observed values of ice concentrations. In fact, observed active IN concentrations are sometimes much lower than the observed ice concentrations. Therefore, the aforementioned novel processes have been proposed as a way to produce higher ice concentrations (e.g. Rosinki and Morgan, 1991). Evaporation IN postulates that when droplets in a cloud evaporate, they leave their cloud condensation nuclei (CCN) in a different chemical form due to aqueous reactions. In a small number of instances the new substance becomes insoluble, or partially insoluble in water, and may act as an ice nucleus (Rosinki and Morgan, 1991). Hence drop evaporation may lead directly to IN formation. The measurements of Rosinki and Morgan (1991) suggest that this happens to one in 10 4 to 10 7 cases of drop evaporation. A somewhat similar proposed mechanism is evaporation freezing. This mechanism suggests droplets freeze during the evaporation process (Cotton and Field, 2002). Unlike evaporation IN, no mechanism has been postulated for evaporation freezing, and in fact there is evidence that the results of Cotton and Field (2002) may have been misinterpreted (Field et al., 2004; Korolev et al.,

46 ). Both evaporation IN and evaporation freezing are parameterized in RAMS by moving a fraction of the concentration of evaporated droplets to the number of the ice nuclei in case of evaporation IN, or to pristine ice crystals in case of evaporation freezing. Although the fraction has applicable upper and lower limits, its value is usually adjusted to match the observations. Evaporation freezing has subsequently been discredited. The amount of available ice nuclei is also prognosed. If there are already a certain number of active ice nuclei, then the remaining portion of the predicted ice nuclei are activated. The reason for prognosis of IN is due to the significantly fast reduction of the liquid water path if IN are not prognosed, which does not match the observations, and does not produce persistent mixed-phase clouds Parameterization of the Ice Crystal Habits and Ice Growth There are numerous ice crystal habits in nature. The basic shapes can be described in terms of the primary habits, which are defined in terms of the aspect ratio (prism face length divided by the basal face length). These habits can be categorized generally as columns when aspect ratio (φ) is greater than one, plates when φ is less than one, or isometric shapes when φ is unity (Young, 1993). The primary habit is dependent on the air temperature, and secondary habit depends on the environmental supersaturation as shown in Figure 2.3. In reality, crystal habit and its evolution are very complex microscale phenomena that involve the formation of defects, or nucleation sites on the crystal surface, the rate of vapor diffusion to the surface, and the rate of incorporation of the

47 29 diffused vapor to the surface, all of which depend on both temperature and supersaturation and even the size of the crystal, as well as the surface characteristic (i.e. whether there is a quasi-liquid cloud layer on the surface or not) (e.g. Frank (1949), Kuroda and Lacmann (1982), and Wettlaufer (1999)). Figure 2.3 Ice crystal growth habits (Libbrecht, 2005) In simple terms, the strong temperature dependence of the primary habits is related to surface kinetics, which causes the different growth rates of the basal and prism faces. The difference in growth rates arises from the different deposition coefficients on the faces and the different vapor gradients on the faces, however why the deposition coefficients and vapor gradients differ for each face is unknown. The supersaturation dependence of the secondary habits, on the other hand, is a consequence of the different water vapor

48 30 density gradients along the basal and prism faces of the ice crystal. For example, a column has its maximum vapor density gradient along the basal faces, where the curvature is greatest. A plate, however, has a maximum vapor density gradient along the prism faces (e.g. Sheridan et al., 2009). These vapor gradients are a result of both the shape of an ice crystal, which alters the diffusive transfer of vapor to the faces, and the different deposition coefficients of the basal and prism faces, which control the vapor uptake on each face. In nature, various ice crystal habits may coexist, and aspect ratios change in time as a cloud evolves. In RAMS, however, ice crystal habit can be fixed at one habit during the entire simulation, or the growth characteristic can evolve spontaneously with temperature. Habits are represented in RAMS, and in most models through mass and velocity dimensional power laws. Empirically derived power-law relationships relate the mass and fall velocities of ice crystals to the maximum dimension of the crystal. Power law relationships are obtained for various observed ice crystal habits based mainly on the observations taken in a few cloud cases (Mitchell, 1996). m D v t m D v v m (2.4) Here α m, and m v, v are empirically derived constants, and are different for each crystal shape. These representations are significant because the mass growth rate of individual ice crystals depends on the coefficients in the mass-dimensional relationships (e.g. Harrington et al., 1995). Also, the influence of habits on the fall velocity of an ice crystal determines the duration an ice crystal will remain in the liquid-portion of the

49 cloud and the Bergeron process. The ice crystal growth rate follows Walko et al. (1995): 31 dmi dt 2 D f ( ) (2.5) Re va vsh In equation 2.5, the first term on the left hand side is the growth rate of a single ice crystal, D is the maximum dimension of the crystal, is the diffusivity, f Re is the ventilation coefficient, va is the ambient vapor density, and vsh is the equilibrium vapor density over ice. The ventilation coefficient is represented as follows: vd t fre Vk 1/2 S (2.6) where S is the shape parameter, v t is the terminal velocity of the crystal, and V k is the kinematic viscosity. Physically, the shape parameter is similar to the capacitance in electromagnetics, which is the ability of a particle to attract electrons. Here, it is a representation of the vapor growth characteristic of the crystal based on diffusion of the vapor mass flux over the crystal, which is related to its assumed shape. Although in nature shape can change in time based on environmental conditions, shape parameter is kept constant in models for ease of calculation Model Setup The model is set up the same as in the studies of Klein et al. (2009), with 168 grid points in the horizontal direction and 100 grid points in the vertical direction. The horizontal grid spacing is 60 m, and the vertical grid spacing is 30 m. Lateral boundary conditions

50 32 are cyclic, while the top boundary is a rigid lid with a Rayleigh damping layer and bottom boundary is the ocean surface with constant surface fluxes at a temperature of K. The model time step is 1 s, with a simulation duration of 12 hours. Radiation is calculated every 20 s using the Fu and Liou radiation scheme (Fu and Liou, 1992) in the infrared. Solar radiation is not included, because solar influence is negligible in the Arctic in October. Subgrid scale turbulence is parameterized using the model of Deardorff (1980). The model is initialized with the sounding from Klein et al. (2009) given in Figure 2.4. Initially there is no ice. Figure 2.4 Profiles of water and vapor mixing ratios and potential temperature for the initialization of the model simulations. Actual sounding taken at Barrow on 1700 UTC 9 October 2004 is also given (Klein et al., 2009). Simulations performed in this study are conducted in two dimensions (Eddy Resolving Model or ERM). Since 3D simulations require major computational power and time, it is more suitable to identify the interactions between microphysics and dynamics in 2D. It is important to note that 2D simulations have an upscale energy cascade, in other words, the

51 33 energy is not dissipated at smaller scales, but rather it enhances the larger scales of motion. This is due to the fact that vortex stretching, which allows energy transfers to smaller scales, cannot happen with only two dimensions. This brings caution to the 2D results of this study in the following manner; any perturbation in dynamics due to a change in microphysics will be enhanced, and may not exist in the real 3D world. 2.3 Methodology In this section, the methods and analysis used in each chapter of this study are presented. In Chapter 3, simulations are performed using the model setup explained in the previous section, however, surface fluxes and large-scale forcing are not included. Simulations in Chapter 3 are carried out to investigate how changing ice nuclei concentrations, ice nucleation mechanisms and crystal habits influence cloud microphysics and dynamics. Therefore, separate investigations are done with increasing IN concentrations, with different nucleation mechanisms and different crystal shapes. The crystal shapes used in this study are spheres, hexagonal plates and dendrites. The shape parameter, mass and velocity dimensional relationships that represent a specific crystal shape in the numerical model are given in Table 2.1.

52 34 Table 2.1 Ice crystal habit parameterization CRYSTAL HABITS Shape Parameter S Mass Dimensional Relationship m m D m Velocity Dimensional Relationship v v D v m m v v Spheres Hexagonal Plates Dendrites In Chapter 4, the simulations are set up as explained in the previous section, but in contrast to the simulations in Chapter 3, surface fluxes and large scale forcing are included. Surface fluxes are kept constant W/m 2 for sensible heat fluxes and W/m 2 for latent heat fluxes, which are the ECMWF values during the time period of MPACE Case B. Large-scale forcing, namely the omega, temperature tendency, water vapor mixing ratio tendency, and potential temperature tendency, is also obtained from ECMWF model data for the ocean region close to North Slope of Alaska. A more detailed analysis of habit influence is presented in Chapter 5, where influences of ice crystal mass growth rate and fall speed on cloud microphysics and dynamics are investigated. The fall speed relationship determines how long an ice crystal can stay in the cloud depleting the liquid water through vapor depositional growth. The mass relationship determines how large an ice crystal can grow through vapor diffusion for a given increment in mass. The effects of these relationships are interrelated because growth depends on both vapor diffusion, and the time an ice crystal spends in cloud. Therefore, it is essential to understand the role of these relationships in more detail and

53 35 analyze how the changes in each dimensional relationship affect cloud microphysics and dynamics. Sensitivity studies are done by keeping the mass dimensional relationship of spheres, but exchanging the fall speed dimensional relationship of spheres with the fall speed dimensional relationship of dendrites. The same setup is repeated with dendrites. These simulations are performed both with and without surface fluxes and large scale forcing, and the results are compared. In Chapter 6, sensitivity tests are performed to understand the relative impacts of radiative cooling and ice precipitation induced cloud-base stabilization on cloud dynamics and microphysics. For this purpose, the cloud-integrated amount of the radiative cooling is kept fixed in time, and simulations are performed with the ice microphysics turned off. The implied effect of ice formation is applied through a fixed change in ice-liquid-water-potential temperature, which is obtained from the simulations in Chapter 3. The results of these tests are then used to explain the microphysics and dynamics of the simulations presented in Chapters 3 and 4 to understand the persistent Arctic mixed-phase clouds. Finally, Chapter 7 presents the conclusions of this dissertation, along with suggested future work.

54 36 CHAPTER III SIMULATIONS WITHOUT SURFACE FLUXES The dynamics of liquid clouds have been widely studied (see Cotton et al., 2010, p. 328, and references therein), however the complex interactions between microphysics and dynamics that help produce persistent mixed-phase clouds are not well understood. In this chapter, using the current parameterizations of ice nuclei activation (and ice nuclei concentrations), ice nucleation mechanisms and ice crystal shapes, cloud microphysical and dynamical interactions are investigated. The aim of this chapter is to understand the dynamic response to microphysical changes related to different factors such as increasing ice nuclei concentrations, different ice nucleation mechanisms and assuming differing crystal habits. The interactions between microphysics and dynamics arising from these factors are important as these interactions are partially responsible for the persistence of the clouds, and in turn influence the surface and atmospheric energy budgets. Simulations presented in this chapter are of isolated mixed-phase clouds and therefore do not include surface fluxes and large-scale forcing. This method isolates the interactions among ice nuclei concentrations, nucleation and habits with cloud dynamics and the persistence of mixed-phase clouds.

55 37 Because dynamical-microphysical interactions are complex, results with increasing ice nuclei (IN) concentrations and a single habit choice (hexagonal plates) will be given first. In a broad sense, this analysis allows one to understand the cloud microphysics and dynamic influence of increasing anthropogenic emissions that could lead to greater IN concentrations, or of changing IN concentrations such as is thought to occur naturally with the seasons. Next, simulation results involving different ice nucleation mechanisms with a single habit choice (hexagonal plates) will be presented. This method allows for the isolation of nucleation effects on the formation and evolution of mixed-phase clouds. The choice of habit is somewhat arbitrary; although the obtained water paths are different in magnitude for different crystal habits, same physical assessments are valid for different habits such as dendrites or spheres (not shown). Next, simulation results with different habits for specific ice nucleation mechanisms will be presented, which allows for the examination of the effects of different habit choices on mixed-phase cloud formation and evolution. 3.1 Simulation Results with Increasing Ice Nuclei Concentrations In the Arctic, due to the scarcity of measurements, IN concentrations are not as well documented as at other locations around the globe. Measuring IN has always posed a considerable challenge, and because of this even the basic nucleation mechanisms operating in cold clouds are not well understood. Nevertheless, cloud properties are sensitive to the number of IN present (e.g. Prenni et al., 2007, Avramov and Harrington,

56 ). For the prediction of climate change and the future Arctic climate, it is important to understand how Arctic clouds will react to possibly increased anthropogenic aerosol emissions. However, clouds are complex systems in which changes in growth processes, such as ice concentrations through nucleation, may alter the dynamics that support the cloud system itself. It is, therefore, of interest to analyze how cloud properties and cloud dynamics change with increasing IN concentrations. In this section, ice crystals are assumed to be hexagonal plates, and a series of simulations are performed with increasing IN concentrations by a factor of 1, 10, 25 and 50 times the original Arctic IN concentration using the initial conditions from MPACE Case B as explained in Chapter 2. These simulations will be referred to as Arctic IN, 10IN, 25IN, and 50IN from here on. Homogeneous nucleation, deposition-condensation nucleation, contact nucleation, and secondary ice production through Hallet-Mossop process are active for these simulations. As IN concentrations are increased more ice crystals are produced, increasing ice vapor growth, and therefore, the depletion of the liquid water. Ice growth at the expense of liquid is known as the Bergeron process (Pruppacher and Klett, 1997, p6). The reduction in the liquid water path and the increase in the ice water path with increasing IN concentrations is evident in Figure 3.1a and b. Figure 3.2 shows the domain-averaged profiles of the sum of the pristine ice and snow concentrations instantaneously at the 6 th simulation hour. This figure confirms that the ice formation increases with increasing IN concentrations. The secondary peak in ice concentrations close to surface is due to the secondary ice nucleation that occurs between 3 and 8 C. Separate simulations that do not

57 include ice multiplication through secondary ice nucleation show that its influence on cloud dynamics can be ignored. 39 Figure 3.1 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with increasing IN concentrations of a factor of 1, 10, 25, and 50 times the original Arctic IN concentrations. To analyze the cloud microphysics and dynamics interactions in detail, vertical profiles of domain-averaged cloud liquid water, pristine ice, snow and aggregate mixing ratios are given in Figure 3.3 at the 6 th simulation hour. Within the liquid cloud layer, pristine ice formation is greater for the 50IN case due to the increased number of available IN (Figure 3.3b). Once formed, ice grows depleting the liquid water, and as a result the liquid layer is much thinner for the 50IN case (Figure 3.3a), and ice growth as snow is much larger (Figure 3.3c). As ice crystals grow large, they also get bigger via collection

58 40 (aggregate formation, Figure 3.3d). Figure 3.2 Domain-averaged profile of the sum of pristine ice and snow concentrations (L -1 ) at the start of the 6 th hour given for a factor of 1 (black), 10 (blue), 25 (green), and 50 (red) times original Arctic IN concentrations. To assess the impact of changes in the cloud radiative forcing on cloud dynamics, cloudintegrated radiative cooling is plotted in Figure 3.1d. This quantity is a measure of the total, vertically integrated infrared cooling in the liquid cloud layer. The thinner clouds obtained with increasing IN concentrations (Figure 3.1a) lead to less infrared radiative cooling. Cloud radiative forcing influences cloud dynamics through buoyancy. Radiative cooling of air at cloud top produces buoyant parcels that sink thus generating vertical motion. As a result less radiative cooling means less buoyant production of turbulence kinetic energy (TKE), and weaker circulations (Figure 3.1c). Although it is clear that increasing IN concentrations deplete the liquid water more rapidly changing the dynamics through a reduction in radiative cooling, there are

59 41 additional processes contributing to the changes in dynamics. These include processes such as the cloud-base stabilization, and entrainment. However, how all processes operate and feedback to cloud dynamics have not been examined with regard to predicted ice habit and cloud longevity, and will be addressed in the future chapters of this dissertation. Figure 3.3 Mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour given for original Arctic IN concentration (black), and a factor of 50 times original Arctic IN concentration (red). Enhanced ice formation and growth lead to increased ice precipitation as IN concentrations are increased (Figure 3.4b). Aside from the reduced radiative cooling due to thinning of the clouds with increasing IN concentrations, enhanced precipitation also influences the TKE. Ice formation and growth cause latent heating and as ice precipitates from the liquid layer some of it sublimates yielding a latent cooling below cloud-base.

60 42 These latent heating and cooling processes can lead to TKE consumption through thermal stratification near cloud-base created through the latent heating in-cloud and cooling below cloud-base. The stratification of potentially warmer air overlying potentially colder air provides a stable layer, which can prevent the circulations from reaching the surface, thus reducing the strength of the circulations and yielding a reduction in TKE. This stabilization increases with IN concentration. To show this effect, vertical profiles of domain-averaged potential temperature (θ) are plotted for Arctic IN and 50IN cases at the sixth hour of the simulations (Figure 3.4a). The profiles show that the regions of ice formation and growth from about 750m to 1500 m (Figure 3.3b and c) yields greater warming with 50IN simulation compared to the simulation using the original Arctic IN concentrations, and a greater cooling below about 500m where some of the snow and aggregates sublimate (Figure 3.3c and d). A reduction in magnitude of the vertical motion and the TKE due to the concomitant reduction in radiative cooling and increased stabilization with increasing IN concentrations are evident in the instantaneous vertical profiles given in Figure 3.5a and b for the start of the 6 th hour of the simulations. Although the latent heating of the cloud layer is larger in 50IN case, the reduction in moisture supply (due to stronger stabilization) and radiative cooling produce less vigorous buoyancy production for this case (Figure 3.5c). In these simulations, shear production does not contribute to the TKE except at the surface because the mean wind does not change with height (Figure 3.5d).

61 43 Figure 3.4 (a) Potential temperature profiles at the start of the sixth hour of the simulations for Arctic IN concentration (black) and a factor of 50 times Arctic IN concentration (red). Initial profile is given by magenta. (b) Times series of ice precipitation at the surface (g/kg) for Arctic IN concentration (black) and a factor of 50 times Arctic IN concentration (red). Figure 3.5 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w in m 2 /s 2, (b) turbulence kinetic energy (m 2 /s 2 ), (c) buoyancy production of TKE (m 2 /s 3 ) and (d) shear production (m 2 /s 3 ) for original Arctic IN concentration (black), and a factor of 50 times original Arctic IN concentration (red).

62 Simulation Results Using Different Ice Nucleation Mechanisms Several ice nucleation mechanisms have been proposed for the formation of ice in mixedphase clouds. These include the classical nucleation processes and some alternative nucleation mechanisms such as the evaporation IN (Rosinki and Morgan, 1991) and evaporation freezing (Cotton and Field, 2002). The impacts of these nucleation mechanisms on cloud dynamics have not been studied. Because different ice nucleation mechanisms may produce different numbers of ice crystals, liquid water contents will vary with different nucleation mechanisms altering the radiative and latent heating impacts. Changes in-cloud radiative properties due to changing liquid water mass alter the cloud top radiative cooling that drives buoyant downdrafts. Through these complex interactions amongst latent heating, buoyancy, radiation, and the number of ice crystals formed, the change in the preferred region of ice formation may cause cloud evolution, and dynamics to change in comparison to other nucleation mechanisms. Nucleation mechanisms investigated here are deposition-condensation nucleation, evaporation IN and evaporation freezing. Contact freezing is not investigated, as this mechanism does not produce any ice crystals when the case is simulated. For observations of this case, it was not possible to measure contact IN, it is also not possible to detect the nucleation mechanisms separately. Figures 3.6a and b show the liquid and ice water paths (LWP and IWP) obtained using different ice nucleation mechanisms. A simulation using an all-liquid cloud is also shown

63 45 as it provides an upper bound to the results. To provide a lower bound a glaciation simulation, where IN concentrations are increased by 100 times the original value, is also included. Simulations with deposition condensation nucleation, evaporation freezing and evaporation IN are also plotted. Note that a fraction of 10-9 for evaporation freezing means that 10-9 of the evaporating droplets freeze and become pristine ice crystals. A persistent mixed-phase cloud is obtained with this nucleated fraction for evaporation freezing. A persistent mixed-phase cloud is also obtained for deposition-condensation using IN concentrations reduced to Arctic values. A fraction of 10-7 for evaporation IN means that 10-7 of the evaporating droplets become active deposition-condensation IN. Evaporation IN with a fraction of 10-7 also produces a mixed-phase cloud, but with less liquid water content and more ice content compared to the evaporation freezing simulation. A huge scatter in the water paths with different nucleation mechanisms is obvious. Fridlind et al. (2007) used the evaporation freezing and evaporation IN in their simulations in addition to the classical nucleation mechanisms to simultaneously match the observed ice water contents and ice concentrations. (Studies with other nucleation mechanisms are not able to cause both the IWP and the concentrations to match in-situ observations.)the fraction of evaporating droplets contributing to IN or ice that Fridlind et al. (2007) used are 10-4 and 10-5, which are much larger than used here. The reasons for the reduced fractions in our study are two fold. First the surface fluxes are not active in these simulations, so there is no surface-based moisture source available to resupply the

64 46 cloud with the water mass lost through ice precipitation. Second, ice production in relatively comparable ranges with each different nucleation mechanism is sought for a consistent analysis of the differences in dynamics due to changing nucleation mechanisms. Nucleation mechanisms cannot only lead to differences in concentrations, but also to differences in the spatial locations of the nucleation events. While differences in dynamics due to differences in concentrations can be explored with a single nucleation mechanism, it is not possible to examine the spatial impacts of a given nucleation mechanism unless the liquid water paths are similar. As the ice production and growth increase depending on the assumed fraction of evaporating droplets contributing to IN or ice crystals, liquid water paths are reduced more with evaporation IN or evaporation freezing. As the fraction of evaporation IN or evaporation freezing increases, more of the evaporating droplets contribute to increasing the number concentration of ice nuclei for the former, and ice crystals for the latter. As a result, the depletion of the liquid phase is faster for the higher fractions of evaporation IN or evaporation freezing. Therefore, it is not surprising that total glaciation of the cloud occurs with relatively high fractions of evaporation IN or evaporation freezing. The results are consistent with Harrington et al., (1999) and Harrington and Olsson (2001), where a more rapid Bergeron process is obtained by increasing ice crystal concentrations. This effect is similar to increasing the concentrations of IN in clouds as presented in the previous section. Therefore, similar influences of radiative cooling and precipitation on cloud dynamics as with increasing IN concentrations are expected with increasing

65 47 fractions of ice nucleation and will be examined further here. Figure 3.6 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with different ice nucleation mechanisms: Deposition-condensation nucleation (green), evaporation freezing (blue), evaporation IN (purple), liquid simulation with no precipitation (black) and glaciation simulation (red). Simulation with evaporation Freezing is performed with 10-9 of the evaporating droplets are assumed to be freezing. In evaporation IN simulation, 10-7 of the evaporating droplets are assumed to leave behind an IN. Glaciation simulation is performed with increasing the IN concentrations to 100 times the original Arctic concentrations. Liquid simulation is performed with ice phase and precipitation turned off. For instance, the average TKE is plotted for each nucleation simulation in Figure 3.6c. TKE is strongly affected by the nucleation mechanism: TKE is greatest when the cloud is composed of the liquid phase only, and generally decreases when the ice phase is present. As the mass fraction of the ice phase increases, TKE tends to decrease. The reasons for this reduction in TKE are similar to the reasons for the suppression of TKE with

66 48 increasing IN concentrations in Harrington and Olsson (2001): with increased fractions of evaporation IN or evaporation freezing, there is more ice formation, and hence precipitation (Figure 3.7b). In fact, the peaks in IWP in Figure 3.6b are due to rapid ice production and precipitation cycles that take place until almost all of the cloud liquid is depleted. Therefore, the peaks in IWPs (Figure 3.6b) coincide with the peaks in the surface ice precipitation (Figure 3.7b). In addition, as ice precipitation increases (Figure 3.7b), more water mass is removed from the system, reducing the radiative cooling (Figure 3.6d) and in turn the radiatively driven circulations (Figure 3.6c). Precipitation also tends to suppress TKE through the stabilization of the cloud and subcloud layers similar to drizzle from liquid stratocumulus (Stevens et al., 1998). When the stabilization is strong, the cloud layer has to do work against it to maintain the circulations, which sometimes leads to a decoupling of the cloud layer from layers below. This stabilization effect is shown in Figure 3.7a, where the domain-averaged vertical potential temperature (θ) profile is plotted for the 6 th hour of the simulations. It is apparent that θ is decreasing below cloud where precipitating ice is sublimating and therefore cooling the layer, whereas it is increasing at regions of ice formation and growth. In the all-liquid simulation, there is the heating of the cloud layer through entrainment because precipitation is deactivated in this case. In contrast, the glaciation simulation shows intense latent warming overlying intense latent cooling. This stabilization appears as an important factor in causing the TKE to decrease in simulations that include ice formation and growth, and becomes more significant as the ice growth and precipitation increases (Figures 3.6c and 3.7a).

67 49 Figure 3.7 (a) Potential temperature profiles at the start of the sixth hour of the simulations for simulations with different ice nucleation mechanisms. Initial profile is given by magenta. (b) Times series of ice precipitation at the surface (g/kg) for simulations using different nucleation mechanisms. 3.3 Simulation Results Using Different Ice Crystal Habits In numerical models that include ice microphysics, a single crystal habit is generally assumed for an entire simulation, and many times the crystal type is determined solely based on the cloud temperature. Equivalent density spheres are commonly used ice crystal shapes for ease of calculations with both the primary and the secondary habits approximated through a reduced density (e.g. Fridlind et al., 2007). In most models, habits are represented simply using mass and velocity dimensional relationships, which relate the maximum dimension of a crystal to its mass or velocity as discussed in Chapter 2. These are simplified relationships that ignore the essentials of crystal growth such as the surface kinetic effects or even predicting two axis dimensions. Because the mass and fall-speed relationships determine the in-cloud lifetime of a crystal, they are closely tied to the growth process. In this section, the effects of assuming different crystal habits on

68 50 cloud properties and cloud dynamic evolution are examined. Simulations using dendrites as the ice crystal shape are compared to simulations using hexagonal plates. Because dendrites grow quickly, but fall slowly, they provide a strong contrast to hexagonal plate crystals (e.g. Avramov and Harrington, 2010), which grow slower, and fall faster for a given mass. For the simulations presented in this section, separate simulations with deposition condensation nucleation, evaporation IN and evaporation freezing are performed and they are repeated using hexagonal plates, spheres and dendrites. The influence of ice crystal habit on LWP and IWP are shown in Figures 3.8a and b, respectively. Although these two crystal habits are both plate-like, a large difference between the water paths is obtained. Liquid water depletion and ice production are very rapid when dendrites are assumed in the simulation; dendrites grow much faster and remain longer in the cloud because of their shape. These results are consistent with Avramov and Harrington (2010), where diverse water paths were obtained for the same initial conditions but with different habit choices. Because parameterized habit strongly influences growth and precipitation, an influence on cloud dynamics should be expected. As Figure 3.8c shows, simulations with hexagonal plates have greater TKE compared to the simulations with dendrites. Note that while nucleation certainly influences growth and therefore dynamics, habit also has a substantial impact on TKE. This result is important because many studies assume that ice nucleation dominates the microphysical impact on mixed-phase clouds (e.g. Harrington and Olsson, 2001; Fridlind et al., 2007).

69 51 Although increasing the IN concentrations, changing nucleation mechanisms, and using different crystal shapes are all different processes, they all can elicit a strong dynamic response. This influence is through the consumption of liquid water and its link to cloudtop radiative cooling, and ice precipitation and its link to cloud-base stabilization. Figure 3.8d shows the cloud-integrated longwave cooling for simulations with deposition condensation nucleation assuming dendrites and hexagonal plates. The reduced LWP of the simulation with dendrites produces much less radiative cooling compared to hexagonal plates. In addition, Figure 3.9b shows that ice precipitation at the surface is also larger for dendrites. Therefore stabilization of the cloud with respect to subcloud layer (Figure 3.9a) is much greater in the simulation with dendrites. The more substantial stabilization in combination with the reduced cloud top radiative cooling lead to a significant drop in TKE (Figure 3.8c). These results are consistent with Harrington et al. (1999), where oblate spheroids lead to increased ice growth, slower fall speeds and weaker circulations. Consequently, greater radiative cooling, and smaller ice precipitation are the reasons for the larger TKE when hexagonal plates are assumed instead of dendrites. The choice of habits impacts the simulated water paths, and the evolution of the cloud. Results presented in this section suggest that the choice of habit is as important as ice nucleation in determining the dynamics of mixed-phase clouds.

70 52 Figure 3.8 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with different ice nucleation mechanisms and habits: Depositioncondensation nucleation (solid), evaporation freezing (dashed), evaporation IN (dashed dotted), reds are for simulations with hexagonal plates, and blues are for dendrites. Figure 3.9 (a) Potential temperature profiles at the start of the sixth hour of the simulations for simulations with hexagonal plates (red) and dendrites (blue) for deposition condensation nucleation. (b) Times series of ice precipitation at the surface (g/kg) for simulations using hexagonal plates (red) and dendrites (blue) for deposition condensation nucleation.

71 Summary and Discussion The results of this chapter show that ice nuclei concentrations, ice nucleation mechanisms and ice crystal habits have significant impacts on cloud structure and dynamics. Moreover, it appears that the ice crystal habit assumed in a cloud model has an impact on cloud properties and dynamics, which has been ignored by previous studies. In this chapter, it is shown that both cloud-base stabilization and radiative cooling influence cloud dynamics. Ice growth takes place at the expense of liquid water drops. In time, as ice precipitates, this mass is removed from the system. As liquid water is depleted, radiative cooling is reduced, which means less buoyant production of TKE. Ice precipitation also produces a stabilization, which forces circulation strengths to cease. Noticeably, the effects of radiative cooling and ice precipitation are convolved together. In order to understand how this mixed-phase cloud system works, it is important to isolate the individual effects and the conditions under what each process dominates in changing cloud properties and the TKE. Based on the results of this chapter one can conclude that for the simulations of the cases where surface fluxes and surface forcing are not contributing, the choice of habit is significant as it alters the partitioning of cloud water mass and, therefore affects cloud dynamics. Consequently, results suggest that mixed-phase studies attempting to match observations (e.g. Fridlind et al. 2007) likely need to consider the effects of the crystal

72 54 shape. Additionally, IN concentrations and the process through which ice is produced, are critical for cloud evolution and dynamics. Although Avramov and Harrington (2010) investigated the change in cloud water paths, here the links between changing cloud microphysical properties and cloud dynamics are explored. This exploration will continue with sensitivity tests to changes in radiative cooling and ice precipitation in the following chapters. Surface forcing in the form of latent and sensible heat fluxes from the surface are also important as they can influence cloud microphysics and dynamics by providing the moisture and buoyancy that help maintain the cloud. In this chapter, surface fluxes were not considered: as a result the simulations presented in this chapter do not compare with observations of Case B, where surface fluxes play a role. The observed values of the surface fluxes are included in the simulations that will be presented in the next chapter.

73 55 CHAPTER IV SIMULATIONS WITH SURFACE FLUXES AND LARGE-SCALE FORCING In the previous chapter, the simulations presented did not include surface fluxes and large-scale forcing. The reason for this approach was to isolate the impacts of ice nuclei (IN) concentrations, nucleation mechanisms and habits on the growth and evolution of mixed-phase clouds with the only dynamic response being that internal to the system (i.e. eddies driven by cloud-scale processes). The reason for doing this was two-fold: first, examining the system with only internal dynamic processes avoids added complications associated with other forcings. Second, many times Arctic clouds exist in environments where cloud processes are relatively isolated from surface effects, such as when inversions exist below cloud-base or when the clouds persist over the sea-ice, and weak synoptic flow occurs. Thus, simulations without any moisture and heat sources from the surface or large-scale dynamic effects help to not only understand relatively isolated Arctic cloud cases but also allows for the better separation of the internal cloud processes from other effects. The current chapter analyzes how the inclusion of surface latent and sensible heat fluxes and large scale forcing modify the results of the previous chapter.

74 56 For ease of comparison, this chapter is organized similarly to the previous chapter. The effects of IN concentrations will be presented first. Next, simulation results using different ice nucleation mechanisms will be provided, followed by the results of simulations with different ice crystal habits. 4.1 Simulations Using Different Ice Nuclei Concentrations In this section, the effects of increasing IN concentrations on cloud macroscale properties and dynamics are investigated. It was shown in Chapter 3 that increasing IN concentrations influences cloud lifetime through changes in cloud water paths and cloud dynamics when surface and large-scale forcing are not included. Ultimately, it is the interaction between dynamics, microphysics and the larger-scale environment that determine the lifetime of the cloud. Weaker circulations, for example, may shut off moisture supply to the cloud layer from the surface and in the face of constant ice precipitation this may cause the cloud to dissipate. If cloud-scale circulations are coupled to the surface, then surface sources can provide moisture and thermal energy to the cloud and help maintain circulations. As mentioned before, it is also important to analyze microphysics-dynamics interactions to determine how clouds might act with increased aerosol emissions as seems likely to be true of future climates. Figure 4.1a and b show the liquid and ice water paths obtained by increasing IN concentrations for a factor of 1, 10, 25 and 50 times the original Arctic IN concentration.

75 57 These simulations will be referred to as IN, 10IN, 25IN and 50IN from here on. Similar to the simulations in Chapter 3, as IN concentrations are increased more ice crystals are produced and ice growth intensifies, therefore, more liquid water is depleted through the Bergeron process (Pruppacher and Klett, 1997, p.6). Figure 4.2 shows the domainaveraged profiles of the (a) the sum of the pristine ice and snow, and (b) the aggregate concentrations instantaneously at the 6 th simulation hour. It is evident that total ice concentrations increase with increasing ice nuclei concentrations. The more numerous ice crystals formed with increasing IN concentrations grow rapidly to large sizes, depleting the liquid water. Because there are many crystals depleting the available liquid in the 50IN simulation, the crystals do not grow to as larger sizes as in the IN simulation. The maximum snow diameter at the 6 th simulation hour obtained with the 50IN simulation is 2.4 mm, whereas it is 10 mm with the IN simulation. The smaller crystals have slower fall speeds and reside for longer periods of time in the liquid portion of the cloud therefore depleting more liquid water. At the same time, higher concentrations lead to greater collection due to the increase in the probability of collisions, which allow for more aggregate formation. The smaller snow sizes also lead to smaller aggregate sizes in the 50 IN simulation (maximum diameter of 2mm at 6 th simulation hour) compared to the IN simulation (maximum diameter of 8mm). Comparing Figure 4.2 (a) with Figure 3.2, the secondary layer of ice growth around m disappears with the inclusion of surface fluxes and large scale forcing. This is because surface fluxes provide stonger mixing and reduce the sublimational cooling: potential temperature is indeed larger at these levels compared to the same simulations without surface fluxes.

76 58 Figure 4.1 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with a factor of 1, 10, 25, and 50 times the original Arctic IN concentrations. Figure 4.2 Domain-averaged profiles of (a) the sum of pristine ice and snow concentrations and (b) aggregate concentrations (L -1 ) at the start of the 6 th hour given for a factor of 1 (black), 10 (blue), 25 (green), and 50 (red) times original Arctic IN concentrations.

77 59 The vertical extent of the liquid cloud layer is increased for both simulations with surface fluxes, which provide moisture and thermal energy from the surface (Figure 4.3a versus Figure 3.3a). Investigating the horizontal structure at this and other simulation hours (not shown) indicate that there is less pristine ice formation, but more snow and aggregate formation when surface fluxes are included (Figure 4.3b, c, and d versus 3.3b, c, and d). This result is physically reasonable, because once ice forms, the increased moisture will promote rapid growth to larger sizes such as snow, and contribute to aggregate formation. Figure 4.3 Mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour given for original Arctic IN concentration (black), and a factor of 50 times original Arctic IN concentration (red). Figure 4.4a shows the potential temperature profile for the IN and 50IN simulations, and the location of the liquid cloud layers are given in Figure 4.3a. Below the liquid cloud-

78 60 base (around 1000m), the boundary layer is well mixed when surface fluxes and large scale forcing are included (compared to Figure 3.4a). The mixing effect can be deduced from the shift in the location of the maximum in magnitude of the vertical circulations toward the surface (Figure 4.5a), and the potential temperature profile (Figure 4.4a). Despite the enhanced ice precipitation obtained with the 50IN simulation (Figure 4.4b), the boundary layer is cooled in this case (Figure 4.4a), which highlights the effect of entrainment. The magnitude of the vertical circulations and TKE are stronger in the IN than the 50IN simulation due to greater liquid water amounts (Figure 4.5a and b). The reduced liquid water path of the 50IN simulation results in much less radiative cooling (Figure 4.1d), which contributes to the reduction in TKE for this simulation. The magnitudes of the TKE and vertical circulations are much larger compared to the simulations without surface fluxes and large scale forcing. Figure 4.4 (a) Potential temperature profiles at the start of the sixth hour of the simulations (b) Times series of ice precipitation at the surface (g/kg) for Arctic IN concentration (black) and a factor of 50 times Arctic IN concentration (red).

79 61 The boundary layer is deeper in the IN simulation, which is due to the stronger radiative cooling that drives stronger circulations and promotes mixing in of warmer and drier air from above the cloud top inversion. This process results in a warmer boundary layer in the IN simulation compared to the 50IN simulation. Figure 4.5 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w in m 2 /s 2, (b) turbulence kinetic energy (m 2 /s 2 ) for original Arctic IN concentration (red), and a factor of 50 times original Arctic IN concentration (green). 4.2 Simulations Using Different Ice Nucleation Mechanisms In this section, the simulations presented in Chapter 3 with different nucleation mechanisms will be compared to the same simulations including surface sensible and latent heat fluxes and large-scale forcing. In Chapter 3, the evaporation IN simulation with a fraction of 10-7 had a greater ice production rate, much reduced liquid water paths, and in turn a lower TKE compared to the deposition-condensation nucleation simulation. The simulations with deposition-condensation nucleation and evaporation IN are chosen for this analysis because the two simulations yielded the largest difference in Chapter 3,

80 where different nucleation mechanisms were compared. The differences in cloud dynamics with the addition of surface fluxes will be explored here. 62 Figure 4.6a and b show the liquid and ice water paths obtained through repeating simulations with evaporation IN (fraction of 10-7 ) and deposition-condensation nucleation. Although the ice production and growth are larger with evaporation IN (Figure 4.7), the inclusion of the surface fluxes provide sufficient moisture to the cloud layer so that the liquid water paths and, therefore, the cloud-integrated radiative cooling (Figure 4.6d) and the TKE (Figure 4.6c) are similar in both simulations. After the 9 th hour, toward the end of the simulations, the TKE decreases more for evaporation IN (Figure 4.6c), which coincides with an increase in ice water paths (Figure 4.6b) and ice precipitation (Figure 4.8b). The somewhat greater ice formation and growth with the evaporation IN simulation deplete the cloud liquid water slightly more compared to deposition-condensation nucleation simulation (Figure 4.6a). However, this depletion in liquid water content is relatively weak, and the vertical extent of the liquid cloud layer is deeper than the simulation without surface fluxes (not shown) because of the supply of moisture from the surface.

81 63 Figure 4.6 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with deposition-condensation nucleation (red) and evaporation IN with a fraction of 10-7 (blue). Figure 4.7 Domain-averaged profile of (a) the sum of pristine ice and snow concentrations and (b) aggregate concentrations (L -1 ) at the start of the 6 th hour given evaporation IN (blue), and deposition-condensation Nucleation (red).

82 64 Figure 4.8 (a) Potential temperature profiles at the start of the sixth hour of the simulations (b) Times series of ice precipitation at the surface (g/kg) for simulations with depositioncondensation nucleation (red), and evaporation IN (blue). Turbulence kinetic energy (Figure 4.6c), the magnitude of the vertical circulations, and buoyancy production (not shown) are therefore similar for both the depositioncondensation nucleation and evaporation IN simulations. It is important to note that the magnitude of surface fluxes is essential for keeping a persistent cloud layer with strong circulations. Surface fluxes can break down the cloud-base stabilization induced by ice precipitation, which acts to suppress the TKE as well as reducing the magnitude of the radiative cooling, which acts to increase the TKE. 4.3 Simulations Using Different Ice Crystal Habits In Chapter 3, it was shown that simulations with dendrites produce much lower liquid water paths and TKE compared to hexagonal plates and spheres. The inclusion of a surface forcing and large-scale forcing on cloud dynamics will be explored in this section

83 65 using different crystal habits. Figure 4.9a and b show the liquid and ice water paths obtained with simulations using dendrites and hexagonal plates. In contrast to the results obtained using different nucleation mechanisms in the previous section, the water paths and TKE (4.9c) still show a considerable difference associated with the different habits assumed. Liquid water path and TKE are larger for hexagonal plates as they consume less of the cloud liquid water through the Bergeron process compared to the simulation with dendrites. Dendritic growth is more rapid leading to lower liquid water content (Figure 4.9a) and therefore less cloud-integrated radiative cooling (Figure 4.9d). The reduction in cloud-top radiative cooling causes a concomitant reduction in the strength of the cloud circulations. The sum of pristine ice and snow concentrations increase when dendrites are the asssumed habit compared to when plates are the assumed habit (Figure 4.10a). As dendrites have lower fall speeds, they remain longer in the liquid portion of the cloud as compared to plates, therefore depleting the liquid water while growing to larger sizes.

84 66 Figure 4.9 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with deposition-condensation nucleation using hexagonal plates (red) and dendrites (blue). Figure 4.10 Domain-averaged profile of (a) the sum of pristine ice and snow concentrations (L -1 ) and (b) aggregate concentrations (L -1 ) at the start of the 6 th hour given for the depositioncondensation nucleation with hexagonal plates (red), dendrites (blue).

85 67 Rapid ice growth and precipitation (Figure 4.11b) in simulations with dendrites consumes the cloud liquid water more strongly compared to the simulation with hexagonal plates. With the inclusion of surface fluxes, the depletion in liquid water content is slightly stronger for dendrites, and is reduced for the hexagonal plates compared to simulations without surface fluxes (Figure 4.9a vs. Figure 3.8a). This result suggests that supply of moisture through surface fluxes can compensate for the lost water mass if the loss is weak as in the case of the weakly precipitating hexagonal plates. Below the liquid cloud, the boundary layer is again well mixed due to the inclusion of the surface fluxes (Figure 4.11a). Overall, the simulation with dendrites is cooler than the simulation with hexagonal plates despite the increased ice growth and precipitation. The deeper boundary layer (Figure 4.11a) and stronger circulations (Figure 4.9c) in the simulations with hexagonal plates compared to dendrites allow for greater entrainment rates leading to a warmer boundary layer. Weaker entrainment is due to the reduced TKE (Figure 4.9c). However the reduction in TKE is not as strong since surface fluxes provide a continuous supply of water vapor, and thermal energy.

86 68 Figure 4.11 (a) Potential temperature profiles at the start of the sixth hour of the simulations (b) Times series of ice precipitation at the surface (g/kg) for simulations using hexagonal plates (red) and dendrites (blue) for deposition-condensation nucleation. Consequently, simulations using different habits still result in different cloud properties and energetics when surface fluxes and large scale forcing are included (results that are reminiscent of the studies by Harrington and Olsson, 2001). Dendrites grow so quickly and fall so slowly that even strong surface vapor sources cannot compensate for the mass loss through ice growth and precipitation. These results indicate that the way habits are predicted may have substantial consequences for both the microphysical and the dynamic evolution of the cloud layer. Comparing the vertical profiles of the potential temperature of different habits with that of simulations without surface fluxes, it is clear that the magnitude of the fluxes is important, as they contribute to the reduction in vertical stratification and the decoupling of the cloud from the surface.

87 Summary and Discussion During MPACE Case B, the observed range is around 120 g m -2 for the liquid water path, and 15 g m -2 for ice water path (Klein et al., 2009), while the ice concentrations are around 1-10 L -1 (McFarquhar et al., 2007). Simulations presented in this chapter produce slightly lower liquid water paths compared to observations, whereas the ice water paths and ice concentrations are significantly under-predicted. Simulations presented here are idealized simulations that are aimed at investigating the influences of different parameters such as ice nuclei concentrations, ice nucleation mechanisms and crystal habits on the cloud microphysics and dynamics, therefore, all nucleation mechanisms are not active at simulations presented here; as a result a match with observations is not expected. However, the under-prediction of ice contents is a common problem with numerical studies of the Arctic (e.g. Fridlind et al., 2007, and Klein et al., 2009). While nucleation mechanisms do not produce larger differences on liquid water paths, and TKE when surface fluxes and large scale forcing are included, habits impact the cloud water paths and dynamics relatively strongly. Increasing IN concentrations also impact cloud properties and dynamics. The degree of impact is related to the magnitude of ice formation and growth, comparing Figures 4.2a, 4.7a and 4.10a, it is evident that when sum of the pristine ice and snow concentrations increase, cloud properties and dynamics are influenced more compared to the simulations that produce less of the pristine ice and snow concentrations. The influence of increasing ice concentrations on

88 cloud dynamics occur due to the reduction in radiative cooling caused by the decrease in liquid water paths during ice growth. 70 Furthermore, with the addition of surface fluxes and large-scale forcing, the cloud properties and energetics alter compared to Chapter 3. The clouds can retain more liquid water as the fluxes provide moisture to the cloud layer. Entrainment also becomes more significant when the fluxes from the surface are included. The magnitude of surface fluxes is quite important, as the stabilization effect can dominate and shut off the supply of moisture to the cloud by reducing the turbulence kinetic energy. For that reason, the effect of surface fluxes will also be included in the analysis of the effects of radiative cooling and ice precipitation induced stabilization on cloud dynamics in the following chapters. To isolate the effects of radiative cooling and stabilization, sensitivity tests with fixing these processes will be performed first and then surface fluxes will be added to the analysis for a more detailed investigation.

89 71 CHAPTER V CRYSTAL HABIT INFLUENCE ON ICE GROWTH and CLOUD DYNAMICS It was shown in Chapter 3 that the approximation of crystal habits has a profound influence on cloud water paths and dynamics. Because each crystal habit is represented in models by specific mass dimensional and velocity dimensional power laws, the assumed habit in cloud models controls both mass growth rate and the fall velocity of the crystals. In this chapter, an investigation will be made toward understanding the influence of the mass and the fall speed dimensional relationships of crystals on cloud properties and dynamics. Whereas Avramov and Harrington (2010) examined the consequences of approximating crystal habits for predicted mass partitioning in simulated mixed-phase clouds, this work examines the possible impacts of such approximations on cloud dynamics. For this purpose, and following Avramov and Harrington (2010), simulations are carried out by exchanging the power law relationship representing the mass of dendrites with that of spheres and hexagonal plates. For instance, simulations are performed using the mass dimensional relationship of dendrites with the fall speed relationship of spheres, and the mass dimensional relationship of spheres with the fall speed relationship of dendrites. Studies like these allow for examination of whether the habit impacts on dynamics described in earlier chapters are due more to vapor growth or

90 72 removal by sedimentation. These studies will not provide a complete separation of growth from sedimentation effects because both impacts are intertwined: fall velocities depend on the size of the ice particles that in turn depend on the growth rate. Moreover, the total amount of liquid converted to ice through the Bergeron process depends on the amount of time that ice crystals remain within the supercooled liquid portion of the cloud, however, that in itself is determined by both the vertical motion strength and the fall speed. Nevertheless, these sensitivity studies should allow for at least a partial separation of the two commensurate effects. Indeed, Avramov and Harrington (2010) were able to determine that vapor growth is likely more important than sedimentation in predicting the impacts of liquid water depletion by ice growth, especially at low ice concentrations. However, both physical processes were shown to be important for liquid maintenance and cloud lifetime. 5.1 Analysis of Habit Influence To assess the role of assumed mass and velocity dimensional power laws, simulations are performed exchanging the velocity dimensional power law of one crystal habit with the other. In most model studies, spheres are assumed for simplicity, and for this reason simulation results with spheres and dendrites will be presented first. Simulation results comparing dendrites and hexagonal plates are similar to the results presented in this section, therefore only comparison between dendrites and spheres will be shown. Simulations in this section are performed using deposition-condensation nucleation only.

91 73 The resulting water paths are plotted in Figure 5.1a and b. As shown in previous chapters, simulations with dendritic fall speed and mass dimensional power laws produce the most ice (Figure 5.1b), depleting the liquid water more (Figure 5.1a) and significantly reducing the TKE (Figure 5.1c). When dendrites fall with spherical speeds, the TKE and liquid water paths are the highest, and ice water paths are the lowest of all four simulations. Because now these dendrites fall faster, the crystals end up having less time for growth, in turn depleting the liquid water least. Similarly when spheres fall with dendrites fall speeds, liquid water depletion increases compared to the spherical simulation. These results suggest the importance of fall speeds in preserving the liquid water and maintaining cloud circulations. The results are consistent with Avramov and Harrington (2010), where they performed a similar test using a coarser grid model and found that using the fall speed of a compact crystal such as a sphere leads to retention of more liquid water in simulations with dendritic growth rates. It is also important to note that, compared to dendrites (green), spheres with dendritic fall speeds lead to greater liquid water retention and radiative cooling along with larger TKE, while having relatively comparable ice water paths. This result suggests the importance of the mass dimensional relationship in maintaining a mixed-phase cloud in terms of its dynamics, while preserving both ice and liquid water. Combining these results, it is clear that both fall speed and mass dimensional relationships are important for the persistence of a mixed-phase cloud, its microphysics and dynamics. Fall speeds determine the incloud lifetime of the crystal, and in turn how much time it has for growth before

92 74 precipitating to the surface, whereas mass relationship determines how large a crystal can become while depleting the liquid water through vapor depositional growth for a given amount of time spent within the cloud. Figure 5.1 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with deposition condensation nucleation assuming dendrites with spherical fall speeds (black), spheres with dendrite fall speeds (blue), dendrites (green), and spheres (red). Because the liquid water paths in all simulations yield nearly the same results, the radiative cooling is the same for all simulations except for that with dendrites, where liquid water is depleted most strongly.

93 75 A comparison of the sum of pristine ice and snow concentrations and aggregates are given in Figure 5.2a and b, respectively. The snow and pristine ice concentrations are greatest for the simulation of spheres with dendritic fall speeds, and lowest for dendrites with spherical fall speeds. Aggregate formation, on the other hand, is largest for dendrites, followed by the spheres with dendritic fall speeds, and lowest for spheres. Figure 5.2 Domain-averaged profiles of (a) the sum of the pristine ice and snow concentrations (b) aggregate concentrations (L -1 ) at the start of the 6 th hour. When spheres fall with dendrites fall speeds, because spheres now fall slowly, and the slower falls speeds allow more time for growth, the amount of snow produced is more comparable to the simulation with dendrites (Figure 5.3c). For dendrites, pristine ice mixing ratio is lower because pristine ice once formed is rapidly converted to snow, due to their mass dimensional relationship that allows for a faster vapor growth rate (Figure 5.3b). In total, the snow and ice concentrations are greater for spheres with dendritic fall speeds, while aggregate concentrations are larger for dendrites (Figure 5.2 a and b). Therefore, snow and aggregate sizes are much larger (about a factor of 10 for largest sizes) in the simulation with dendrites compared to the simulation of spheres with

94 76 dendritic fall speeds. Aggregate formation is a collection process, which depends on the area of impact and the relative fall speeds of the particles. Therefore, it is no surprise that dendrites have higher aggregate mass mixing ratios (Figure 5.3d) and concentrations (Figure 5.2b). Cloud liquid water is depleted more with dendrites due to their faster growth rate and slow fall speeds (Figure 5.3a). Figure 5.3 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. Because more liquid is retained in simulations of spheres with dendritic fall speeds, the magnitude of the vertical circulations (Figure 5.4a) and TKE are larger for this simulation (Figure 5.4b). With dendrites, the cloud-base stabilization is much stronger due (Figure

95 5.4c) to a more intense ice growth and sublimation of the much larger precipitation amounts (Figure 5.4d). 77 Figure 5.4 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). The simulations of spheres with dendritic fall speeds are also compared to the simulations with spheres. When spheres fall with dendritic fall speeds, they fall slower compared to actual spheres. The effect of the slower fall speeds is to provide longer time-periods for growth, and the impact can be seen in the mass mixing ratio profiles (Figure 5.5). It is evident that spheres with dendritic fall speeds have more snow and aggregate mixing ratio. At the same time, snow, ice and aggregate concentrations are larger for simulation of spheres with dendritic fall speeds (Figure 5.2). Both snow and aggregate particle sizes

96 78 are larger for the simulation of spheres with dendritic fall speeds. Large aggregate concentrations are physically reasonable because the slower fall speed, along with the larger particle sizes, would allow for more aggregation events. Overall, both the aggregate concentrations and mass mixing ratios are quite small. Hence the ice precipitation is similar in both cases (Figure 5.6d), and depletion of liquid water is not changed (Figure 5.1a). As a result, neither radiative cooling (Figure 5.1d) nor the cloudbase stabilization (Figure 5.6c) is different in both cases, resulting in similar TKE (Figure 5.6b). Magnitudes of vertical circulations are slightly smaller for spheres at this time, due to a slightly more intense ice precipitation, perhaps producing more stabilization. Figure 5.5 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour.

97 79 Figure 5.6 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) profile of potential temperature (K) and (d) time series of ice precipitation (g/kg). The simulation of dendrites with spherical fall speeds is compared to the simulation with spheres. In simulation with spheres, pristine ice and snow mixing ratios are larger (Figure 5.7b and c), and the concentrations of pristine ice and snow are also slightly larger. The aggregate mixing ratio (Figure 5.7d) is larger for dendrites with spherical fall speeds. Particle sizes of snow and aggregates are similar (~0.5mm, not shown), but only slightly larger for the simulation of dendrites with spherical fall speeds near the cloud top. Dendrites with spherical fall speeds have a stronger growth rate compared to spheres, and fall out faster; as a result they spend less time in the liquid portion of the cloud. Therefore, this simulation produces less ice mass growth of ice compared to spheres,

98 80 where ice precipitation is slightly stronger. However, because the overall ice concentrations, and mixing ratios are small in both simulations, and ice precipitation is only slightly larger for spheres, there is not a strong influence on the depletion of the liquid water and both simulations yield the same cloud liquid water (Figure 5.7a). Due to the similar amount of liquid water, and the limited amount of ice formation and growth in these two simulations, both the radiative cooling (Figure 5.1d) and the cloudbase stabilization at base are similar (Figure 5.8c). Therefore, the TKE (Figure 5.8b) and the magnitude of the vertical circulations (Figure 5.8a) also end up yielding the same results. Figure 5.7 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour.

99 81 Figure 5.8 Domain-averaged (a) resolved w w in (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature profile (K), and (d) time series of ice precipitation at the surface (g/kg) at the start of the 6 th hour. The simulation with dendrites is also compared to the simulation of dendrites with spherical fall speeds. The cloud water is depleted at a much more rapid rate with dendrites (Figure 5.9a). When dendrites fall with spherical fall speeds, crystals do not have as much time for growth, as they fall much faster compared to the simulation with dendrites. Because they fall slowly, dendrites have more time for growth, yielding larger mass mixing ratios of snow (Figure 5.9c). Both snow and aggregate sizes are much larger for the simulation with dendrites (a maximum snow diameter of 10 mm for dendrites compared to 1.2 mm maximum snow diameter of dendrites with spherical fall speeds, while aggregate sizes are more than a factor of 10 larger for dendrites).

100 82 Figure 5.9 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. Because liquid water depletion is much less, radiative cooling (Figure 5.1d) is more intense in the simulation of dendrites with spherical fall speeds, in comparison to the dendrites. Additionally, the larger growth of dendritic forms and their slower fall yield a stronger stabilization because the larger crystal masses have a longer time to sublimate in the relatively dry environment below cloud-base (Figure 5.10c). As a result of the reduced radiative cooling and stronger stabilization, the dendritic simulation produces reduced vertical circulation strengths (Figure 5.10a), and therefore a much lower TKE (Figure 5.10b).

101 83 Figure 5.10 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). 5.2 Influence of Incorporating Surface Fluxes and Large Scale Forcing The previous section described simulations without any moisture and heat sources from the surface or large-scale dynamical effects. This allowed for the investigation of the isolated influences of mass and velocity dimensional relationships used in models. The current section analyzes how the inclusion of surface latent and sensible heat fluxes, and large scale forcing modify the results of the previous section.

102 Although in general the results seem similar to those in the previous section, differences exist in terms of magnitudes (Figure 5.11). 84 Figure 5.11 Domain-averaged and vertically integrated (a) liquid water path (g/m 2 ), (b) ice water path (g/m 2 ), (c) turbulence kinetic energy (m 2 /s 2 ), (d) long-wave radiative cooling (W/m 2 ) obtained from the simulations with deposition condensation nucleation assuming dendrites with spherical fall speeds (black), spheres with dendrite fall speeds (blue), dendrites (green), and spheres (red). With the supply of thermal and moisture fluxes, dendrites produce much stronger ice growth compared to the previous section. The liquid water path depletion by dendritic growth is reduced with the addition of thermal and moisture sources, and this is true despite the enhanced ice production. The reduced liquid water path caused by dendrites yields the least radiative cooling among the simulations here, and compared to the simulations in the previous section without surface fluxes and large scale forcing. The

103 85 combination of enhanced ice production and reduced radiative cooling causes the dendritic simulation to yield the least TKE. The magnitudes of TKE for all simulations are greater compared to the simulations without surface fluxes and large-scale forcing. Magnitudes of both the total ice and snow concentrations, and aggregate concentrations are the largest for dendrites (Figure 5.12a and b). The concentrations of snow and ice are lowest for the simulation of dendrites with spherical fall speeds, whereas aggregate concentrations are lowest for spheres. Figure 5.12 Domain-averaged profiles of (a) the sum of the pristine ice and snow concentrations (b) aggregate concentrations (L -1 ) at the start of the 6 th hour. When spheres with dendritic fall speeds are compared to dendrites, dendrites produce much more snow and aggregate mass (Figure 5.13c and d), while spheres with dendritic fall speeds produce more pristine ice mass (Figure 5.13b). The simulation with dendrites produces much larger snow and aggregates compared to the simulation of spheres with dendritic fall speeds, just like in the simulations without surface fluxes. Dendrites grow

104 86 quickly and deplete the liquid water rapidly (Figure 5.13a), which produces a much lower radiative cooling at cloud top (Figure 5.11d). Contrary to the simulations with dendrites, where liquid water is more intensely depleted, comparing Figure 5.13a to Figure 5.3a, cloud liquid water is actually increased for the simulation of spheres with dendritic fall speeds when surface fluxes are included. This could be due to a slower reduction of the liquid water path, in turn a slower reduction in radiative cooling, which could still promote liquid water formation as the reduction is slow. In addition, the added moisture through the inclusion of the surface fluxes also promotes liquid water formation. The continuous supply of moisture through surface fluxes allows the cloud water contents to increase for simulation of spheres with dendritic fall speeds, despite the more intense ice precipitation compared to the simulation without surface fluxes. The reduced liquid water content in the simulation of spheres with dendritic fall speeds leads to a reduction in the radiative cooling (Figure 5.11d), which reduces the TKE (5.14b).

105 87 Figure 5.13 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. The boundary layer is well mixed below the liquid cloud layer when surface fluxes are included (Figure 5.14c). Despite the intense ice growth and precipitation (Figure 5.14d), the boundary layer is cooled in the simulations with dendrites. The warmer boundary layer in the simulation of spheres with dendritic fall speeds is due to a stronger TKE, which produces a deeper boundary layer height, and greater entrainment of the warmer and drier air from above the boundary layer.

106 88 Figure 5.14 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). The simulation of spheres with dendritic fall speeds are also compared to the simulation of spheres. Spheres with dendritic fall speeds fall more slowly and have more time for ice growth compared to spheres, which are removed rapidly from the liquid layer. Therefore, snow and aggregate mixing ratios are larger for the simulation of spheres with dendritic fall speeds (Figure 5.15c and d). Snow and aggregate sizes are also larger for the simulation of spheres with dendritic fall speeds when compared to the simulation of spheres. The greater ice growth depletes liquid water more rapidly in the case of spheres with dendritic fall speeds. This produces more ice precipitation when compared to the simulation with spheres (Figure 5.16d).

107 89 Figure 5.15 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. Despite the greater ice growth rates in the simulation of spheres with dendritic fall speeds and increased ice water paths (Figure 5.11b), the cloud liquid water is only slightly smaller compared to the simulation of spheres (Figure 5.15a and Figure 5.11a). This result is reasonable because although the ice precipitation is larger compared to the simulation with spheres, it is still quite small for spheres with dendritic fall speeds (Figure 5.16d), which means there is not a significant removal of liquid water. Consequently, surface fluxes of moisture are able to resupply the water mass lost through precipitation.

108 90 Figure 5.16 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). As a result of small ice precipitation, boundary layers in both simulations have the same structure (Figure 5.16c). The TKE (Figure 5.16b) and the magnitude of the vertical circulations (Figure 5.16a) are also similar due to the similar amount of radiative cooling. The simulation of dendrites with spherical fall speeds are compared to simulation of spheres. Spherical snow and aggregate sizes are smaller compared to dendrites with spherical fall speeds. Compared to spheres, dendrites grow quickly and precipitate rapidly due to faster spherical fall speeds (Figure 5.18d). Because the ice growth is small in both simulations, the same amount of liquid mass is retained (Figure 5.11a and Figure 5.17a), yielding similar cloud-top radiative cooling rates in both simulations (Figure

109 d). Figure 5.17 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. The simulation of dendrites with spherical fall speeds and spheres produce similar ice precipitation rates compared to the simulation with spheres (Figure 5.18d). However, the simulation of dendrites with spherical fall speeds has larger peaks in ice precipitation time series associated mainly with their mass dimensional relationship, which allows for a faster growth rate and the supply of moisture through surface fluxes. Compared to the simulations without surface fluxes, ice precipitation is now similar in both cases, and this is mainly because the surface fluxes provide a moisture supply to the cloud along with thermal energy, which enhances cloud circulations through greater buoyancy. The overall

110 effect of surface fluxes is to maintain the supercooled liquid, enabling dendrites to stay in the cloud layer and grow to larger sizes. 92 Figure 5.18 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). The simulation of dendrites with spherical fall speeds is also compared to the simulation of dendrites. When dendrites precipitate with spherical fall speeds, they are removed faster from the liquid cloud, having less time for growth. Therefore, in the simulation with dendrites, crystals grow fast (Figure 5.19c) and fall slowly producing more ice precipitation and larger precipitation sizes (Figure 5.20d), thereby removing more liquid water (Figure 5.19a). In the simulation with dendrites, both snow and aggregate particles

111 93 attain larger sizes compared to dendrites with spherical fall speeds. The reduction in liquid water results in a reduction in the radiative cooling (Figure 5.11d), which yields a smaller TKE for the simulation with dendrites (Figure 5.11c and Figure 5.20b). Figure 5.19 Domain-averaged profiles of mass mixing ratio given in (g/kg) of (a) cloud water, (b) pristine ice, (c) snow and (d) aggregates at the start of the 6 th hour. Despite the more intense ice precipitation in the simulation with dendrites, the boundary layer is cooler compared to the simulation of dendrites with spherical fall speeds. The larger liquid water paths lead to stronger radiative cooling, which increases TKE. Surface fluxes also enhance the circulations to yield a deeper boundary layer. The increased TKE and the larger boundary layer depth for the simulation of dendrites with spherical fall

112 speeds lead to the entrainment of warmer and drier air from above the cloud top, which yields a warmer boundary layer for this case. 94 Figure 5.20 Domain-averaged energetics at the start of the 6 th hour, (a) resolved w w (m 2 /s 2 ), (b) turbulence kinetic energy (m 2 /s 2 ), (c) potential temperature (K) and (d) time series of ice precipitation (g/kg). 5.3 Summary Although the internal microphysics and boundary layer dynamics may be different, the general results obtained with the simulations incorporating surface fluxes and without surface fluxes are very similar and can be summarized as follows. Dendrites can produce similar cloud water paths and TKE as spheres, only if they fall with the spherical fall

113 95 speeds. Physically this makes sense, because spherical fall speeds are greater, leaving crystals less time for growth within the supercooled liquid portion of the cloud and, therefore, depleting less liquid water. When spheres fall with dendritic fall speeds, not only the liquid water paths but also the ice water paths remain relatively high. An ice water path almost comparable to dendrites, and liquid water path comparable to spheres can then be obtained. This is also physically correct, when spheres fall with dendritic fall speeds, spheres fall slower. Hence, these crystals can stay long in the cloud, producing ice, and depleting some liquid water, but not as intense of a depletion as of dendrites. These results are consistent with those of Avramov and Harrington (2010). The combination of a compact crystal, such as a sphere, with a slow fall velocity (such as the velocity relationship for dendrites) could retain both liquid water and ice. Because a relatively large amount of liquid water remains in cloud, radiative cooling can enhance circulations, the circulation strengths are not suppressed as much as in the case of dendrites, and the cloud can sustain itself against collapse. Therefore, mass and velocity dimensional relationships both contribute significantly to the cloud properties and cloud dynamics. The simulations comparing spheres and dendrites were also repeated with two different hexagonal plates, and the obtained results were the same, which confirms that these results are robust. For simulations without surface fluxes, dendrites produce the most intense cloud-base stabilization, and deplete the cloud liquid water most. As a result, the TKE is lowest for

114 96 dendrites. When spheres fall with dendritic fall speeds, they produce more rapid ice growth compared to the simulations with spheres and with dendrites having spherical fall speeds. Because ice precipitation and cloud-base stabilization are not as strong as dendrites, and more liquid water remains in cloud resulting in a larger radiative cooling, the simulation of spheres with dendritic fall speeds has a larger TKE compared to dendrites. When hourly averaged vertical profiles of w w are investigated for all simulations without surface fluxes, the w w shows a similar structure for all simulations, with a maximum around 0.3 m 2 /s 2 except for dendrites. For dendrites, the maximum is much smaller (0.1 m 2 /s 2 ), and there is a secondary maximum with very weak circulations around 400m. This arises from the enhanced cloud-base stability for this simulation, and suggests a possible decoupling of the cloud layer from the subcloud layer. When surface fluxes are included in the simulations, the cloud-base stabilization seen in the potential temperature profile disappears due to mixing. Because of the stronger circulations and the added moisture supply through the surface fluxes, the liquid water lost by precipitation can be replaced and cloud can persist for all simulations except for dendrites, where ice precipitation is most rapid. As circulations are stronger with the addition of surface fluxes for all simulations where relatively large amounts of liquid water remain in cloud, the entrainment of warmer and drier air from above cloud top, which arises from the increased TKE, becomes a limiting factor for TKE.

115 97 Simulations with surface fluxes do not compare well with the observed values of liquid and ice water paths during MPACE Case B. The observed range is around 120 g/m 2 for the liquid water path, and 15 g/m 2 for ice water path (Klein et al., 2009). Although the liquid water paths are not drastically different from observations, the ice water paths are much smaller for all simulations. Even for the simulation with dendrites, where the largest ice water path is obtained, the ice water path is still smaller compared to observations. The ice concentrations are around 1-10L -1 in observations of MPACE Case B (McFarquhar et al., 2007), however, it is much smaller in these simulations. It is important to remember that the simulations presented in this chapter are idealized simulations to examine the habit influence on cloud microphysics and dynamics, therefore, not all nucleation mechanisms are activated in these simulations; as a result a match with observations is not expected. In this and the previous chapters, the radiative cooling and ice precipitation induced cloud-base stability are shown to be important contributors to TKE. In the following chapter, sensitivity tests with constant radiative cooling and ice precipitation will be performed to understand the role of these mechanisms on cloud dynamics. Simulations will be repeated with surface fluxes to assess the surface flux influence on mixed-phase cloud dynamics and maintaining mixed-phase clouds.

116 98 CHAPTER VI SENSITIVITY TESTS In the previous chapters, similar to Harrington et al. (1999), both radiative cooling and ice precipitation induced stabilization have been shown to be important contributors to cloud dynamics and evolution. It is also shown that changing crystal habits, ice nucleation mechanisms and increasing ice nuclei concentrations alter these processes, which in turn alter cloud dynamics. The influences of radiative cooling and ice precipitation on cloud dynamics are inter-related, and take place in the following manner. Radiative cooling at cloud top occurs due to the presence of liquid water, and it is the main driver of the circulations when surface fluxes are deactivated. Ice precipitation removes a portion of the liquid water from the cloud, resulting in the reduction of radiative cooling. A reduced radiative cooling yields a reduction in the magnitude and vertical extent of the cloud circulations. Ice precipitation also has a diabatic influence that affects the cloud circulations. When there is no surface forcing, the ice precipitation produces stabilization at cloud-base, which reduces the strength of the circulations and confines them to the cloud layer. The amount of ice precipitation differs with different crystal habits, nucleation mechanisms, and ice nuclei concentrations. Evidently, it is important to separate the indirect and direct influence of the ice precipitation to be able to understand the isolated influences of radiative cooling and ice precipitation on cloud

117 99 dynamics. In this chapter, the relative influences of both radiative cooling and diabatic influences of ice precipitation on cloud dynamics will be investigated, and an analysis into how these processes change, and cooperate for simulations with different crystal habits, nucleation mechanisms and ice nuclei concentrations will be made. Sensitivity tests are also repeated with the incorporation of surface fluxes in the simulations, where the influence of increasing surface fluxes on cloud dynamics is examined. Based on the results of the sensitivity tests, a quantitative analysis is performed to identify the coupling and decoupling of the cloud layers. This chapter explains the setup of the sensitivity simulations and provides the results of the sensitivity tests. 6.1 Simulation Setup In these sensitivity simulations, ice processes are turned off. As a result, a cloud that consists of only liquid water is developed, and liquid phase is not allowed to precipitate. The effect of ice formation, growth and precipitation is then applied through an average change in the ice-liquid-water potential temperature profile. The cloud integrated amount of the radiative cooling is also fixed throughout the simulations. In this way, both ice precipitation induced latent heating/cooling effects, and radiative cooling are kept constant in sensitivity simulations. The method for fixing these processes of the simulations will be given in this section.

118 Fixing the Radiative Cooling In order to isolate the effect caused by the total radiative cooling of the cloud layer, the radiative cooling is fixed in the sensitivity tests. This is achieved in the following manner. The model calculates the radiative cooling at each grid point every 20 seconds, the radiative time step. Then, at each radiative time step, the integrated radiative cooling (domain-averaged and vertically integrated amount of radiative cooling) is calculated for the liquid cloud layer starting from the 3 rd hour of the simulation. A fraction for a fixed cooling is obtained by weighting this integrated value to a fixed number (such as 30 or 100 W/m 2 ). Finally, multiplying the in-cloud radiative cooling at each grid point by this fraction, the desired amount of constant integrated cooling in-cloud (between 30 or 130 W/m 2 ) is obtained. Then the integrated radiative cooling timeseries is a fixed value starting from the third hour of the simulations. Figure 6.1a and b show these time series for two simulations with a weak and strong radiative cooling of 30 and 130 W/m 2 integrated radiative cooling simulations, respectively. The fixing of the radiative cooling is applied starting at the 3 rd hour of the simulation.

119 101 Figure 6.1 Time series of integrated in-cloud radiative cooling for simulations with (a) 30 W/m 2 (b) 130 W/m Fixing the Ice Precipitation Induced Stabilization A similar approach to the one described above is applied to the diabatic heating terms in order to fix the net heating and cooling effects of ice precipitation. To do so, the average change in the ice-liquid water potential temperature is computed starting at the 3rd hour of the actual mixed-phase cloud simulations presented in Chapter 3. These profiles provide an estimate of the range of diabatic influences of ice growth on cloud-base stabilization. The fixed profiles obtained from the standard mixed-phase simulations are then applied in the following way. The integrated in-cloud latent heating, and integrated below-cloud latent cooling from standard mixed-phase cloud simulations are calculated first using the in-cloud and below cloud-base profiles of the average change in ice-liquid water potential temperature. The obtained profile is then applied to the sensitivity simulations at each time step. Because the cloud layer changes in time as the simulation

120 102 evolves, at each time step, the depth of the liquid portion of the cloud layer is calculated. Then, the latent heating is applied assuming a sinusoidal function from cloud-base to cloud top with the constraint that the integrated amount of total heating is always constant and equal to the in-cloud heating obtained from the mean profile. Because sublimation causes a net cooling below cloud-base, a linear cooling is applied downwards for 4 vertical grid points with the cooling being constant below those four points. The integrated total cooling of this ideal profile is constrained to match the average value derived from the standard mixed-phase simulations. The profile of the mean change in ice-liquid water potential temperature ( il ) is shown in Figure 6.2a. A sample profile of the adjusted change in il based on the limits of a chosen cloud top and base is given in Figure 6.2b. Using an idealized profile, the impact of precipitation on cloud-base stabilization can be controlled. For instance, when precipitation is weak, a small net warming of the liquid cloud and cooling of the subcloud region are expected. However, when ice precipitation is strong, a greater net warming of the liquid cloud and cooling of the subcloud (the region below the liquid cloud-base) are expected. Consequently, the effects of strong precipitation and cloud-base stabilization can be emulated by multiplying the mean change in the il profile by a factor. For this factor, a range from 1 to 4 has been used in this study, which realistically covers the range of the diabatic effects of precipitation. Then, the new profiles are calculated using the sinusoidal method based on the cloud-base and top in the same way as the standard cloud-base stabilization explained above. The fixed profile is applied starting at the 3 rd hour of the simulations.

121 103 Figure 6.2 (a) Profile of the actual mean change in θ il (b) A sample profile of the adjusted change in θ il based on a chosen cloud top and cloud-base. 6.2 Results of the Sensitivity Studies Using these idealized profiles, the relative strength of cloud top longwave cooling, and the net diabatic effects of precipitation on cloud-base stabilization can easily be controlled. This method is applied in this section to better understand the relative importance of cloud top radiative cooling and ice precipitation on the dynamics of mixedphase clouds.

122 104 It is important to note that the influence of entrainment still exists in the sensitivity simulations. Therefore it is not possible to attribute the differences in circulation strengths to radiative cooling or cloud-base stabilization alone. Additionally, the constant latent heating and cooling influences, and the constant radiative cooling are applied to the cloud layer as it changes in extent. As the cloud layer thins, the heating and cooling influences that are applied to the cloud layer could be outside the realistic range of possible effects. Nevertheless, the sensitivity simulations still produce beneficial results because they allow for the separation of impacts that are rather hard to isolate, and also provide a framework for the kind of dynamic effects that can occur Effect of Radiative Cooling Here the effect of increasing the total amount of radiative cooling is examined by varying the integrated in-cloud radiative cooling from 30 W/m 2 to 130 W/m 2. To isolate the radiative effects, the unmodified cloud-base stabilization profile derived from the standard simulations, which is referred to as standard cloud-base stabilization, is used. To understand the cloud dynamics response to increasing radiative cooling, the magnitude of the horizontal (Figure 6.3) and vertical components (Figure 6.4) of the circulations are investigated. As radiative cooling is increased, circulations become much stronger and this affects the extent of the cloud layer (Figure 6.5). Although circulation depth is limited below the liquid base by cloud-base stabilization, circulations can extend below cloud-base to some degree as radiative cooling is increased. This result is important

123 105 because it gives an insight into the magnitude of radiative cooling required to break down the cloud-base stabilization and recouple the cloud and subcloud layers. If the layers can be recoupled, then moisture and aerosol can be resupplied to the layer from the surface, which could either increase or decrease the layer s longevity. For instance, if recoupling supplies more vapor, then precipitation-drying of the layer will be reduced. However, if more ice nucleating aerosols are added to the layer during re-coupling, then precipitation could be enhanced potentially causing cloud collapse. Figure 6.3 Domain-averaged profile of u u in time for constant latent in-cloud heating and below cloud cooling with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m 2.

124 106 Figure 6.4 Domain-averaged profile of w w in time for constant latent in-cloud heating and below cloud cooling with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m 2. Figure 6.5 Domain-averaged profile of liquid water mixing ratio (g/kg) in time for constant latent in-cloud heating and below cloud cooling with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m 2.

125 Effect of Ice Production and Precipitation The results in the previous subsection were based on diabatic precipitation effects derived from the standard simulation. In order to explore the effect of ice formation and precipitation on cloud dynamics, a stronger precipitation induced cloud-base stabilization is applied, as explained in section This is done to emulate the diabatic effects that increasing precipitation rates may have on the net stabilization of the cloud layer, and hence on the dynamics. Naturally, then, the result of increasing the idealized precipitation rate is strong cloud-base stabilization. This implies that significant in-cloud ice production, which causes net latent heating of the liquid layer, and significant subcloud ice precipitation, which leads to a latent cooling, together produce a much stronger cloudbase stabilization. The effects of increasing the net cloud-base stabilization on the strength of the circulations are seen in Figure 6.6 for the horizontal component of the winds, and Figure 6.7 for the vertical component of the winds. Comparing Figures 6.3 and 6.4 with Figures 6.6 and 6.7, circulation strengths are reduced significantly as the cloud-base stabilization is increased. Figure 6.8 shows the cloud layer for both weak and strong radiative cooling simulations.

126 108 Figure 6.6 Domain-averaged profile of u u in time for strong cloud-base stabilization with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m 2. Figure 6.7 Domain-averaged profile of w w in time for strong cloud-base stabilization with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m 2.

127 109 These results show that with strong stabilization, it is harder for downdrafts to penetrate below the cloud-base, and circulations are restricted to the cloud layer. This is consistent with Stevens et al. (1998), where the drizzle-induced stabilization reduces the vertical extent of the downdrafts in liquid stratocumulus clouds. Stabilization suppresses deeper mixing and prevents the radiatively driven convection from reaching the surface. Although the simulations presented here are constrained to produce a non-precipitating liquid cloud, the effect of ice precipitation is forced using the fixed profiles of the change in il. These results also imply that even when the magnitude of radiative cooling is increased, the penetration below cloud-base is small and quite limited in the case of strong precipitation induced cloud-base stabilization. Harrington and Olsson (2001) also showed that in mixed-phase cloud simulations, the strength of the circulations is reduced significantly at times of strong ice precipitation, and the reduction is much stronger compared to liquid clouds due to a much stronger stability below cloud-base.

128 110 Figure 6.8. Domain-averaged profile of liquid water mixing ratio (g/kg) for strong stabilization for constant latent in-cloud heating and below cloud cooling with constant radiative cooling of (a) 30 W/m 2 and (b) 130 W/m Summary Because microphysics is closely tied to cloud dynamics, the solitary impacts of radiative cooling and cloud-base stabilization by fixing these processes are investigated so that their impacts remain constant, in an energetic sense, throughout the simulation. Figure 6.9 summarizes the effect of the various simulations performed keeping the radiative cooling and cloud-base stabilization constant on the magnitude of the TKE averaged over the 6 th hour of each simulation.

129 111 Figure 6.9 Average TKE for the sixth hour of the simulations by keeping radiative cooling and cloud-base stabilization fixed, red is for strong cloud-base stabilization, blue is for weak cloudbase stabilization. Increasing radiative cooling can allow circulations to extend below cloud-base. However, ice production and precipitation from the cloud layer are also important as precipitation defines the degree of stabilization at cloud-base. It is harder for motion to penetrate below cloud-base when the stabilization is strong. Similar to the drizzling stratocumulus simulations of Stevens et al (1998), in the mixed-phase cloud simulations presented here, a strong cloud-base stabilization acts as a barrier to mixing, and limits the circulations to the cloud layer. This limitation may lead to a decoupling of the cloud layer from the surface. In their study, the feedback between surface fluxes and moistening through sublimation of ice below cloud-base eventually acts to produce cumulus-like convection

130 that re-couples the layers. However, surface fluxes are not imposed in simulations presented here, and the surface is at a much lower temperature for this case Analysis of the Simulations in Chapter 3 In this section, the simulations with increasing IN concentrations, and different crystal habits presented in Chapter 3 are analyzed based on the sensitivity simulations of this chapter. The simulations with different ice nucleation mechanisms are not included, as the changes in results were small for the chosen fractions of evaporation IN and evaporation freezing compared to deposition-condensation nucleation. Simulations previously referred to as IN and 50IN in Chapter 3 are analyzed first. In Chapter 3, it was shown that the radiative cooling is larger for the IN simulation, where more liquid water is retained in the cloud. In addition, the precipitation is much smaller for the IN simulation. The radiative cooling and precipitation are compared quantitatively and magnitudes are given in Table 6.1. There is an order of magnitude change in precipitation for the two simulations. The 50IN simulation has much larger precipitation, which produces a stronger stabilization. The average change in ice-liquid water potential temperature due to ice processes over the 3 to 6 hours of the simulations is the same for both simulations. This average change is also the same as the profile of change in iceliquid water potential temperature used for the sensitivity tests with weak cloud-base stabilization. Investigating the change in the liquid cloud layer (Figure 6.10), and w w

131 113 (Figure 6.11), it is evident that with increasing radiative cooling, as in the sensitivity tests for increasing radiative cooling (Figures 6.4 (a) and (b)), the circulation strengths increase, and can extend further below the liquid cloud layer. Table 6.1 Magnitude of the radiative cooling and ice precipitation for IN and 50IN simulations. Average Radiative Cooling (W/m 2 ) Precipitation at Surface (3h to 6h average) (g/kg) Precipitation at Surface (3h-12h average) (g/kg) IN x x IN x x 10-3 The average change in ice-liquid water potential temperature is the same for the IN and 50IN simulations. However, the time evolution of the change in ice liquid water potential temperature is stronger in the 50IN simulation (not shown) compared to the mean profile used in the sensitivity test with weak cloud-base stabilization, and to the IN simulation. This is due to the more intense ice precipitation obtained with the 50IN simulation. Ice precipitation is an order of magnitude larger for the 50IN simulation (Table 6.1). Therefore, the increased stabilization should also be considered, as in the investigation of the sensitivity to cloud-base stabilization in Figure 6.7. The increased stabilization limits the vertical extent of the cloud layer and the moisture supply to the cloud layer, in addition to reducing liquid water through a more intense ice precipitation. As a result, the liquid cloud layer is smaller in vertical extent (Figure 6.10), and the liquid water content is lower for the 50IN simulation (Figure 6.11).

132 114 Figure 6.10 Domain-averaged profile of liquid water mixing ratio (g/kg) in time for (a) IN simulation and (b) 50IN simulation. Figure 6.11 Domain-averaged profile of w w in time for (a) IN simulation and (b) 50IN simulation.

133 115 For the 50IN simulation, the maximum of the magnitude of the vertical circulations is much less than 0.1 m 2 /s 2, while for IN simulation it is around 0.35 m 2 /s 2. According to Shupe et al. (2008), the observed vertical motion for the persistent cloud layer at MPACE is about 0.4 m/s (then the magnitude is 0.16 m 2 /s 2 ). This is exceeded for the IN simulation, where most of the liquid water is retained, and is much weaker for the 50IN simulation, where cloud liquid water is rapidly depleted. The same consequences of increasing ice nuclei concentrations occur for the simulations with different crystal habits. Table 6.2 lists the magnitudes of the radiative cooling and ice precipitation for the 12-hour simulations with spheres and dendrites. As discussed in Chapter 3, dendrites consume more liquid water through ice formation and growth. This consumption of liquid water reduces the radiative cooling more for the simulation with dendrites. Table 6.2 Magnitude of the radiative cooling and ice precipitation for simulation with spheres and dendrites. Average Radiative Cooling (W/m 2 ) Precipitation at Surface (3h to 6h average) (g/kg) Precipitation at Surface (3h-12h average) (g/kg) Spheres x x 10-3 Dendrites x x 10-3 The average change in the ice-liquid water potential temperature over the 3-6h of the simulations is the same for both spheres and dendrites. This average change is also the same as the profile of change in ice-liquid water potential temperature used for the

134 116 sensitivity tests with weak cloud-base stabilization. Investigating the change in the liquid cloud layer (Figure 6.12) and w w (Figure 6.13), it is evident that with increasing radiative cooling, circulation strengths increase, and circulations can extend further below the liquid cloud layer. This result is similar to the sensitivity tests for increasing radiative cooling (Figures 6.4a and b). Figure 6.12 Domain-averaged profile of liquid water mixing ratio (g/kg) in time for simulations with (a) spheres and (b) dendrites.

135 117 Figure 6.13 Domain-averaged profile of w w in time for simulations with (a) spheres and (b) dendrites. Although the average change in ice-liquid water potential temperature is the same for the simulations with dendrites and spheres, the time evolution is distinctly different. The change in ice-liquid-water potential temperature is stronger in the simulation with dendrites (not shown) compared to the mean change used in the sensitivity tests to weak cloud-base stabilization and the IN simulation, due to more intense ice precipitation. The temporal average of the precipitation for the simulation with dendrites is two orders of magnitude larger than for the simulation with spheres (Table 6.2). Therefore, the increased stabilization should also be considered for this case, as in the investigation of the sensitivity to cloud-base stabilization in Figure 6.7. Increased stabilization is caused by more intense ice precipitation, which is typically associated with a reduction in liquid

136 118 water. Consequently, the vertical extent of the cloud layer is limited, and the moisture supply to the cloud layer is weaker. As a result, the liquid cloud layer is deeper in extent (Figure 6.12) and stronger in magnitude for the simulation with spheres, compared to the simulation with dendrites (Figure 6.13). Vertical circulation strengths are weak for dendrites, and hourly averaged profiles (not shown) indicate that the maximum is less than 0.1 m 2 /s 2. For spheres, on the other hand, the maximum is about 0.3 m 2 /s 2. The observed vertical velocities required to stop the cloud layer from collapsing for the observed cloud from MPACE Case B was around 0.4 m/s (Shupe et al., 2008). This is exceeded for the simulation with spheres but is much weaker for dendrites, where cloud liquid is depleted rapidly. The observed case had maximum vertical velocities of 2.5 m/s, which are much larger compared to the simulations presented here. However, simulations presented in this section did not include surface fluxes that caused an underestimation of the strength of the vertical motions. Nevertheless, it is important to keep in mind that the purpose of the simulations that neglect surface fluxes is to better understand the systems that are driven primarily by processes internal to the cloud system. 6.5 Influence of Incorporating the Surface Fluxes in the Simulations Depending on the magnitude of the surface fluxes, the cloud circulations can be enhanced, and moisture can be provided to the cloud from the layers beneath. To

137 119 understand the lifetime of a cloud, it is important to understand how surface fluxes influence the dynamics of the cloud layer, and whether these fluxes enable coupling of the cloud layer with the surface. Therefore, this section analyzes the effect of increasing surface fluxes on the dynamics of the cloud layer. Similar to the analysis in section 6.1, the effect of increasing the cloud-base stabilization and increasing the integrated in-cloud radiative cooling from 30 W/m 2 to 130 W/m 2 are investigated here for different magnitudes of surface latent and sensible heat fluxes. First, simulations with constant surface fluxes of a quarter of the observed MPACE values are investigated. This value is chosen to understand the influence of a weak surface forcing on cloud dynamics. To understand the cloud dynamics response to increasing cloud-base stabilization, the magnitudes of the vertical (Figure 6.14) and horizontal components (Figure 6.15) of the circulations are analyzed for simulations with constant radiative cooling of 30 W/m 2. As the stabilization is increased, circulations become much weaker. In addition, for the stronger stabilization, the cloud layer shows a two-layer structure (Figure 6.16 (b)). This means that for stronger stabilization, cloud layer is decoupled from the subcloud layer. A similar structure is also obtained when radiative cooling is increased to 50 W/m 2, with two liquid cloud layers when cloud-base stabilization is strong (not shown). The second cloud layer close to the surface occurs because of the moisture supply of the surface fluxes, and due the relatively cooler boundary layer obtained with weaker surface fluxes (not shown).

138 120 When radiative cooling is increased to 130 W/m 2, the extent of the liquid cloud layer and the amount of liquid water are the same for both weak and strong cloud-base stabilization simulations (Figure 6.17). The magnitudes of the vertical (Figure 6.18) and horizontal circulations (Figure 6.19) are almost the same, with slightly larger magnitudes for weaker cloud-base stabilization. Also, the two cloud layers are no longer present because the cloud and subcloud layers are coupled with increased circulations provided by the enhanced surface fluxes (Figure 6.18), and the boundary layer is warmer (not shown). Figure 6.14 Domain-averaged profile of w w in time with a quarter of the observed surface fluxes and constant radiative cooling of 30 W/m 2 for (a) weak stabilization (b) strong stabilization.

139 121 Figure 6.15 Domain-averaged profile of u u in time with a quarter of the observed surface fluxes and constant radiative cooling of 30 W/m 2 for (a) weak stabilization (b) strong stabilization. Figure 6.16 Domain-averaged profile of liquid water mixing ratio (g/kg) in time with a quarter of the observed surface fluxes and constant radiative cooling of 30 W/m 2 for (a) weak stabilization (b) strong stabilization.

140 122 Figure 6.17 Domain-averaged profile of liquid water mixing ratio (g/kg) in time with surface fluxes of 130 W/m 2 and constant radiative cooling of 130 W/m 2 for (a) weak stabilization (b) strong stabilization. Figure 6.18 Domain-averaged profile of w w in time with surface fluxes of 130 W/m 2 and constant radiative cooling of 130 W/m 2 for (a) weak stabilization (b) strong stabilization.

141 123 Figure 6.19 Domain-averaged profile of u u in time with surface fluxes of 130 W/m 2 and constant radiative cooling of 130 W/m 2 for (a) weak stabilization (b) strong stabilization. When surface fluxes are increased to 50, 75, 100 and 130 W/m 2 with separate simulations, there is no longer a two-layered liquid cloud for strong cloud-base stabilization at the smaller radiative cooling of 50 W/m 2. However, at times with different combinations of the cloud radiative cooling, cloud-base stabilization and surface fluxes, the cloud layer is strongly decoupled from the subcloud layer and produces two maxima in the vertical. For example, when the magnitude of the integrated radiative cooling is smaller than or relatively close to the surface fluxes and the cloud-base stabilization is strong, the circulations tend to maximize at two vertical levels, which indicates decoupling. The decoupling can occur throughout a simulation, or can take place at certain times during a particular simulation. In order to identify these times,

142 when a distinct change in the vertical structure of the circulations from one maximum to two maxima is observed, a quantitative analysis is made Quantitative Analysis of the Decoupling of the Cloud and Subcloud Layers Although there have been other studies that developed a method for the identification of the decoupling, such as the critical ratio of Turton and Nicholls (1987), which was later updated by Stevens (2000), this critical ratio is based on the ratio of the integral of the positive buoyancy in the cloud to the negative buoyancy below. This critical ratio seems to work for studies of marine stratocumulus, but it does not provide a good measure for decoupling of Arctic Mixed-Phase clouds. The quantitative analysis of the decoupling developed here is based on the ratio of the magnitude of the circulations inside the liquid cloud layer to the magnitude of the circulations below the liquid cloud layer. The reasoning behind this method is that radiative cooling, cloud-base stabilization, and surface fluxes influence the vertical motions through buoyancy. However, when strong decoupling occurs, the vertical TKE (w w ) in each layer is not strongly linked. Therefore, the coupled and decoupled layers may have differences in the in-cloud and below-cloud ratios of w w. As an example of the applicability of this ratio, two cases will be investigated, one where the cloud layer is decoupled from the surface for the entire simulation period, and another case where decoupling occurs at certain times during a particular simulation.

143 125 For the simulation with constant 50 W/m 2 surface fluxes, 50 W/m 2 integrated radiative cooling, and a strong cloud-base stabilization, the cloud is entirely decoupled from the surface (Figure 6.20), while the same simulation repeated with weak cloud-base stabilization has one maximum in vertical circulations. In Figure 6.21, it is evident that there is a difference in the magnitude of the ratio for coupled and decoupled cases. The decoupled case has a distinctly lower ratio at all times compared to the coupled case. The fixed radiative cooling and cloud-base stabilization are applied starting at the 3 rd hour of the simulations. At hours 5 and later, the profiles of w w have two maxima, indicating that the layer has decoupled. Notice that for times greater than hour 4, the vertical motion shows a distinct separation between two simulations, indicating that decoupling occurred. Figure 6.20 Hourly averaged magnitudes of the vertical circulations for the simulation with constant 50 W/m 2 surface fluxes, 50 W/m 2 radiative cooling, and a strong cloud-base stabilization.

144 126 Figure 6.21 Time series of the ratio of the magnitude of the vertical circulations in-cloud to below cloud for weak (blue) and strong (red) cloud-base stabilization (a) 3 to 12 simulation hours (b) a close up to the 5 to 10 simulation hours. Decoupling could also happen at certain times of a simulation, such as for the simulation with constant 100 W/m 2 surface fluxes, 100 W/m 2 integrated radiative cooling, and a strong cloud-base stabilization (Figure 6.22). The transition from a single maximum to double maxima takes place from the 6 th to 8 th hours of the simulation. With weaker stabilization, this case shows a single maximum in the vertical structure of the magnitude of the vertical circulations for all simulation hours. In Figure 6.23, it is evident that the ratio is the same for all other hours, except for when decoupling takes place from the 6 th to 8 th hours of the strong stabilization simulation. During times of decoupling, the ratio decreases significantly.

145 127 Figure 6.22 Hourly averaged magnitudes of the vertical circulations for the simulation with constant 100 W/m 2 surface fluxes, 100 W/m 2 radiative cooling, and a strong cloud-base stabilization. Figure 6.23 Time series of the ratio of the magnitude of the vertical circulations in-cloud to below cloud for weak (blue) and strong (red) cloud-base stabilization (a) 3 to 12 simulation hours (b) a close up to the 5 to 8 simulation hours.

146 Applicability of the Ratio The ratio for determining the coupled and decoupled conditions is also applied to the simulations without surface fluxes, presented in Chapter 3. The average of the time series of the ratio shows that the ratio is an order of magnitude larger for the IN simulation compared to the 50IN simulation, where ice precipitation produces intense cloud-base stabilization. In addition, the mean ratio is also an order of magnitude larger for the simulation with spheres compared to the simulation with dendrites, where cloud-base stabilization is more intense. On the other hand, when the simulations of spheres with dendritic fall speeds and dendrites with spherical fall speeds are compared, as in the first section of Chapter 5, it is found that the ratio is almost the same. This is because the cloud-base stabilization is similar in both cases. It is the combination of a faster growth rate with slower fall speeds that produces a strong stabilization (see Figure 5.4c, and Figure 5.10c). For simulations with surface fluxes as presented in Chapter 4, the average ratio is again greater for the IN simulation compared to the 50IN simulation, however the difference is small. Similar to these results, the simulation with dendrites has a slightly lower ratio compared to the simulation with spheres when surface fluxes are included. The method for identifying decoupling seems to work for simulations presented in this study. Although the ratio is larger for the simulation with spheres, it does not necessarily

147 129 mean that the simulation with spheres is coupled to the surface. It is evident that it is not (Figure 6.13). Rather, it is a measure of the extent and magnitude of the circulations for that simulation being larger compared to dendrites, which could help sustain the cloud longer by bringing in moisture from the layers below.

148 130 CHAPTER VII CONCLUSIONS and FUTURE WORK Mixed-phase clouds are a dominant feature of the Arctic atmosphere, and they can persist for days. These clouds influence the Arctic surface energy budget, and likely affect the melting of sea ice through their radiative effects. To determine their influence on the Arctic energy budget, it is important to understand how these clouds operate and how they maintain themselves against collapse due to ice precipitation. Understanding of Arctic mixed-phase clouds, and the related physical and dynamical processes, is limited due to the relative lack of observations in this area. To increase the knowledge on Arctic mixed-phase clouds, an observational experiment, MPACE was undertaken, which gave rise to many modeling studies. These studies have shown that there are large uncertainties related to the number of ice nuclei available, mechanisms of ice nucleation, crystal habits, and the representation of these in numerical models. The goal of this dissertation is, therefore, to explore how each of these uncertainties along with the surface fluxes affect cloud structure and cloud dynamics to produce long-lived mixed-phase clouds. Changes in ice nuclei concentrations, ice nucleation mechanisms and ice crystal habits cause changes in ice growth, which alter cloud dynamics. Without the inclusion of

149 131 surface fluxes and large-scale forcing, the number of ice crystals formed, and the rate of ice growth affect the liquid water paths through the vapor depositional growth at the expense of liquid droplets (Bergeron process). Larger ice growth rates lead to smaller liquid water paths. This depletion of liquid water causes a reduction in the radiative cooling at cloud top. Because radiative cooling produces negatively buoyant parcels that sink, circulation strengths decrease when radiative cooling is reduced. The influence of radiative cooling is illustrated schematically in Figure 7.1. When the ice growth rate increases, there is stronger ice precipitation. The precipitation induces a latent warming of the cloud layer due to the formation and growth of ice. As the ice particles attain larger sizes, they are removed from the cloud layer and sublimate at the drier levels below cloud-base. The sublimation process produces a latent cooling below cloud-base and, when combined with the latent warming in the cloud layer, lead to stabilization near cloud-base. The strength of the stabilization is determined by the degree of the ice growth and precipitation. The effect of this stabilization, depending on its magnitude, is to act as a barrier for circulations generated by cloud-top radiative cooling to penetrate below the cloud-base. When the vertical extent of the circulations is limited to the cloud layer, the cloud could collapse because liquid is depleted through ice precipitation but is not replenished by fluxes of vapor from the surface. The influence of cloud-base stabilization is illustrated in a schematic presented as Figure 7.2.

150 132 Figure 7.1 Illustration of the influence of cloud-top radiative cooling on cloud dynamics. Figure 7.2 Illustration of the influence of diabatic influences of precipitation, weak and strong cloud-base stabilization on cloud dynamics.