OBSERVATIONS OF CIRRUS CLOUD EVOLUTION DURING THE TROPICAL WARM POOL INTERNATIONAL CLOUD

Transcription

1 OBSERVATIONS OF CIRRUS CLOUD EVOLUTION DURING THE TROPICAL WARM POOL INTERNATIONAL CLOUD EXPERIMENT by Elisabeth A. Cohen A thesis submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Master of Science Department of Meteorology The University of Utah December 2008

2 Copyright Elisabeth A. Cohen 2008 All Rights Reserved

3 THE UNIVERSITY OF UTAH GRADUATE SCH OOL SUPERVISORY COMMITTEE APPRO V AL of a thesis submitted by Elisabeth A. Cohen This thesis has been read by each member of the following supervisory committee and by majority vote has been found to be satisfactory. r 7 Chair: Gerald G. Mace 7 7 Timothy J. Garrett J.-:::: )- Chuntao Liu

4 THE U N J V E R S J T Y 0 F UTA H 'G R A D U ATE S C H 0 0 L FINAL READING APPROVAL To the Graduate Council of the University of Utah: J have read the thesis of Elisabeth A. Cohen in its final form and have found that (l) its fonnat, c itations, and bibliographic style are consistent and acceptable; (2) its illustrative materials including figures, tables, and charts are in place; and (3) the final manuscript is satisfactory to the supervisory committee and is ready for submission to The Gr aduate School. Date Gerald G. Mace Chair: Supervisory Committee Approved for the Major Department ChairlDean Approved for the Graduate Counci I David S, Chapman Dean of The Graduate School

5 ABSTRACT The evolution of cirrus from detrained convective outflow into large-scale persistent cloud systems is not well understood. At times, cirrus properties such as ice water content and particle size appear to evolve over periods of several tens of hours, while at other times the anvils are found to sublimate rapidly. To form a more complete understanding of these processes, the temporal evolution of the cloud systems can be followed using geostationary satellite imagery. Using data from the MTSAT-IR MTSAT-1R Satellite over tropical Australia, cirrus cloud systems that were studied by in situ aircraft and ground-based remote sensors are tracked backwards in time from the time they were observed until they reached a convective origin. This thesis presents results that demonstrate that it is possible to track cirrus clouds in the tropics using satellite imagery and model analyses as guidance. The sensitivity of observed cloud properties to the length of time since the cirrus are detrained from deep convection under various largescale meteorological conditions is examined. large-

6 TABLE OF CONTENTS ABSTRACT ,...iv CHAPTERS 1. INTRODUCTION Clouds and Convection in Earth's Climate Cirrus Formation Cirrus and Convection in Climate Models Tropical Warm Pool International Cloud Experiment METHODS Tracking Algorithm Summary Tracking Algorithm Steps Aircrafts and Instrumentation RESULTS Large Scale Meteorology during TWP-ICE Case Studies Discussion CONCLUSIONS REFERENCES

7 CHAPTER 1 INTRODUCTION This chapter introduces the motivation for this research, including the role of cirrus clouds and convection in Earth's climate. Then, the processes that create cirrus are explained, with a focus on cirrus originating from convection. Next, the occurrence and heights of cirrus are described, as well as cirrus properties such as; optical depth, ice water content and ice crystal habit. Subsequently, climate models and the parameterization of clouds and convection are explained. Finally, the Tropical Warm Pool International Experiment (TWP-ICE) is described and the weather regimes for the case studies are illustrated. The motivation for this research is to understand what processes maintain and dissipate cirrus clouds in the tropics. To understand these processes, the clouds are tracked back in time from a point of observation to a convective origin to determine the length of time since convection and the temporal evolution of the infrared (lr) (IR) brightness temperature (Tb). We refer to this time period as the longevity or age of the cirrus that is studied. Cirrus longevity is studied for cirrus over Manus and Nauru (Figure 1.1), two islands influenced by tropical convection. The frequency distributions of layer-mean radar reflectivity and Doppler velocity for different cirrus longevities are very similar for

8 2 Figure 1.1 Manus, Nauru, and Darwin are located in the tropical warm pool in the western Pacific Ocean (image courtesy of ARM program).

9 3 0.15, , 0.10 Ohr ~~ a.--t hr 0.05 O. 15 t--"' h-r----'~ 0.10 ~ 0.05 c ' ~ , g -, h... r ----t ll: 0.05 I :: :...;:L...4 O. 10 : -! h r =---> h -r ~--1 O ~ lo , , Ohr '-=---1-_3-h-r-~ > o c ~ 0.2 CT (]) a: =:'''''' _6-h-r--=--! =-,;iII:Iooo---l I-oe= :--=-:~---t RenectMty Reflectivity (d8z) (dbz) Velocity VelOclty (m <ms') S l) Figure 1.2 The frequency distributions of layer-mean radar reflectivity for different longevity are very similar at Manus and Nauru. Manus is represented by a dashed line and Nauru is represented by the solid line. Reprinted from Mace, G.G., M. Deng, B. Soden, and E. Zipser, 2006: Association of Tropical Cirrus in the km Layer with Deep Convective Sources: An Observational Study Combining Millimeter Radar Data and Satellite-Derived Trajectories. J. J Atmos. Sci., 63,

10 4 these two islands (Mace et al., ai., 2006; hereafter M06; Figure 1.2). As longevity increases, the variance of radar reflectivity decreases and the distribution of particle sizes is skewed to small particles. This suggests that cirrus can evolve over time into self-maintaining entities that have smaller particles than when the cirrus were part of convection. This also suggests that the larger particles fall out of the cirrus as time progresses or sublimation occurs. This research will examine individual cases of cirrus with convective origins sampled near Darwin, Australia (Figure 1.3) to identify processes that cause the cirrus to evolve. Darwin was chosen for the experiment location because it has Figure 1.3 Map of Australia with Darwin and the TWP-ICE ground-based area highlighted. Image courtesy of NASA.

11 5 an Atmospheric Radiation Measurement (ARM) site, an Australian Bureau of Meteorology forecast office and a heritage of studies that have focused on the Australian monsoon. Approximately 35% of the time cirrus clouds with bases above 10 km exist over Darwin (Mace et al., ai., 2008).We are interested in documenting the evolution of these cirrus from their early stages, in which they are primarily composed of particles that are injected from convective updrafts, to the point where the cirrus become self-maintaining entities that persist for long periods of time as found in M06. The goal of this study is to determine how cirrus clouds change over time by examining the Tb evolution and the composition of the cirrus when they are observed by in situ aircraft. Once the changes in the cloud properties have been determined, the processes that generate these changes can be better understood. The longevity of the TWP-ICE clouds sampled by in situ aircraft using a tracking algorithm is described in the Methods section of this thesis. The data we use from the aircraft include ice particle habits and sizes and ice water contents measured in situ by the Proteus aircraft. These data, along with the ages of the cirrus events and the temporal evolution of the IR Tbs, will help illuminate the processes that maintain and dissipate cirrus clouds Clouds and Convection in Earth's Climate Cirrus Clouds and Earth's Climate Basic conservation laws state that, given given a steady planetary temperature, the amount of energy received by the Earth must equal the amount it emits. The Earth's climate subsystems adjust to attain this energy balance and to equilibrate the latitudinal imbalance of absorbed solar energy, since the equatorial region absorbs more direct solar

12 6 radiation than the Polar Regions. The Earth's atmospheric circulations help distribute this energy by advecting temperature, momentum, and water vapor around the planet. These atmospheric circulations and the energy they transport determine where and when clouds form and how they evolve (Stephens, 2005). Clouds are known to play an important role in the maintenance of this energy balance. Clouds affect the amount of radiation that is absorbed, reflected, and transmitted by the earth-atmosphere system. The radiative energy balance at the top of Earth's atmosphere (TOA) is described in equation 1. t t,j, L +S -S = ~RToA (1) L'+S'-S = ARTOA (1) where S,j, S is the incoming shortwave radiation from the sun, S t is the reflected outgoing shortwave radiation, L t is the outgoing longwave radiation, and ~RTOA is the change in shortwave radiation, L is the outgoing longwave radiation, and ARTOA is the change in radiation at the top of the atmosphere. Clouds change the first three components of radiation at the top of the atmosphere. Clouds change the first three components of equation 1. equation 1. To determine how clouds impact this radiative balance, the Cloud Radiative To determine how clouds impact this radiative balance, the Cloud Radiative Effect (CRE) must be calculated. The difference between the radiation budget Effect (CRE) must be calculated. The difference between the radiation budget components for average conditions (~R) and cloud-free conditions (~Rclear) is the CRE. components for average conditions (AR) and cloud-free conditions (ARci ea r) is the CRE. ARTOA = AR-ARciear (2) (2) Radiation is measured at the TOA by instruments on satellites that receive radiation from earth; and using these data, the clear and the cloudy conditions of outgoing longwave

13 7 radiation, the net radiation, and the albedo of the planet can be compared. The absorbed radiation can be deduced by subtracting the incoming shortwave radiation from the outgoing longwave radiation.. Together, these measurements show that overall, clouds' radiative forcing cools the planet by approximately 17 W/m 2 (Harrison et ai., al., 1990). The change in net radiation at the top of the atmosphere that is produced by adding clouds into a clear atmosphere is approximated by Hartmann (1994; equation 3). ARTOA = -(So/4)Aa p + (oo) - ot (3) 4 Z c t where L\RTOA ARTOA is the change in radiation at the top of the atmosphere, So is the solar constant, L\u Aa p is the change in albedo due to clouds, L t clear( ciear( ) <X)) is the longwave radiation that is emitted from above cloud top to space, T Tz Zc c is the temperature of the cloud top, and ot4zc 4 z c is the amount of radiation emitted from the cloud top. Equation 3 shows that given a constant amount of cloud coverage the changes in albedo and cloud temperature are the two most important factors that determine the radiative effect of clouds. Clouds that produce the most cooling are highly reflective and warm at cloud top. High albedo clouds reflect more incoming radiation than most cloud-free columns (S t High albedo clouds reflect more incoming radiation than most cloud-free columns (S increases). When clouds are not present, the Earth's surface reflects some shortwave increases). When clouds are not present, the Earth's surface reflects some shortwave radiation to space. When clouds are present, they increase the albedo; this change is radiation to space. When clouds are present, they increase the albedo; this change is called the solar cloud radiative effect and it produces a cooling that occurs mostly at the called the solar cloud radiative effect and it produces a cooling that occurs mostly at the surface. surface. Low albedo and cold clouds produce the most atmospheric warming. Compared Low albedo and cold clouds produce the most atmospheric warming. Compared to high albedo clouds, low albedo clouds transmit most of the incoming solar radiation to to high albedo clouds, low albedo clouds transmit most of the incoming solar radiation to

14 8 the atmospheric layers below. Clouds with cold cloud tops also reduce the amount of outgoing radiation compared to clouds with warmer cloud tops (L t decreases). Many processes can change the distribution of energy in the climate system. To determine the processes that are the most influential in changing the climate, a measure called climate sensitivity is used. Climate sensitivity (/I.R) (XR) is defined as the relationship between the amount of climate forcing and the magnitude of the climate response. Mathematically, climate sensitivity is the ratio of the change of the surface temperature (dt s s) ) to a given a change in radiative forcing (dq). XR = dt s /dq (4) Given a constant large sensitivity, a large surface temperature change occurs when the radiative forcing is high. A constantly small sensitivity would imply that the climate system responds only weakly to the cloud radiative effect. The response of the climate system to an external forcing is described by climate feedbacks. Any variable that either directly or indirectly changes the radiative budget due to the increase in surface temperature would be considered a climate feedback (Bony et ai., al., 2006; hereafter referred to as B06). Mathematically this relationship is: AR = AQ +/AT. (5) where R is the radiation budget at TOA, i1q AQ is an external climate forcing (W/m 1m2), and/ f is the feedback parameter, which varies depending on the specific forcing (B06).

15 9 Surprisingly, the net radiative effect of clouds in the tropics generates a very small net radiative effect at the top of the atmosphere, which is due to highly reflective convective clouds producing a strong cooling effect of approximately -100 W/m 2 and cold cloud tops producing strong warming on the order of 100 W/m2,, so combined, the two approximately cancel (Harrison et ai., al., 1990). The atmosphere is warmed by the cold cloud tops and the large amount of water vapor beneath cloud base, which absorbs much of the down-welling infrared radiation emanating from cloud bases (Lohmann and Roeckner, 1995). At the surface, cooling occurs because tropical convection reduces the amount of shortwave radiation reaching the ground Convection, Tropical Anvils and Energy Transport Convection in the tropics transports heat upward, redistributes moisture, and thereby stabilizes the atmosphere. Each convective cloud type in the tropics can have a large effect on the net radiation in the tropics (Hartmann et ai., al., 2001). The large variation in net radiative effect is due mostly to the reflectivity in clouds, which depends largely on the cloud ice and/or water content. Hartmann et al. (2001) believe a cancelling from the positive and negative radiative forcing components is due to many cloud types adjusting through feedbacks to maintain the radiative balance, whereas Kiehl (1994) found this to be coincidental. Having discussed how convection clouds change the energy budget in the tropics, we next consider tropical anvils. The amount of heating and cooling due to anvils in the tropics is very roughly estimated (Frederick and Schumacher, 2008). More specifically, in 2 km thick tropical anvils, average column heating rates are on the order of 20 to 30

16 10 Klday K/day (Ackennan (Ackerman et ai., al., 1988; hereafter A88). The heating rates at cloud top and cloud bottom small to medium values ofiwc. For larger values ofiwc, strong cooling occurs at cloud top and strong warming wanning occurs at cloud base (A88). This strong wanning warming and cooling destabilizes the cloud layer, especially when the cloud layer is thin (A88). Garrett et ai. al. (2005; hereafter G05) argue against the proposed lifting mechanism of net heating which drives clouds vertically upward (A88; Lilly, 1988). Garrett et ai. al. (2005) observed an almost constant temperature and stratification of the anvil even though there were strong calculated radiative heating gradients. The Richardson values not in the range were turbulence would be expected and no vigorous updrafts were measured. In specific scenarios radiative heating can cause spreading (G05). The interaction between the radiative forcing at the cloud's vertical boundaries differ by 30 K to 200 Klday. K/day. The overall effect of cirrus is influenced by the cirrus duration. For example, the cirrus clouds that last longer increase the local albedo of for a longer period. Anvil cirrus tend to begin as thick clouds and thin with time. The transition from thick anvil to thin cirrus results in a change from net cooling at the TOA to net warming wanning (A88). Solar radiation is important in heating the cloud tops. The amount of heating that occurs in tropical anvils also depends on the ice water content (IWC). The average heating rates range from 6.3 Klday K/day to 18 Klday K/day for IWC values of.002 gm- gm" 3 and.1 gm- gm" 3, respectively (A88). The importance of solar heating compared to infrared heating increases with increasing IWC. Infrared heating occurs at all levels within an anvil for creates thin front-like boundaries that spread the cirrus laterally (G05).

17 Cirrus Formation Regional Formation Processes A variety of processes create cirrus or high ice crystal clouds. Most of these processes require a dynamic forcing mechanism that leads to adiabatic ascent or cooling of air. The range of spatial scales for forcing is large. For example, the forcings include large-scale atmospheric circulation, mesoscale circulations of larger storm systems, largescale lifting, gravity waves, and surface boundary conditions. Processes such as boundary layer stratification and wind shear also determine the cloud type (B06; Luo and Rossow, 2004; hereafter LR04). Using cloud top pressures from the International Satellite Cloud Climatology Project (lsccp) (ISCCP) data, Luo and Rossow (2004) find that cloud tops fall due to a complex interaction of rapid sedimentation to lower levels. The forcing varies not only in spatial scales, but also by location. Cirrus can form in a variety of ways in the mid-latitudes. For example, cirrus are often generated by synoptic lifting downstream of a baroclinic wave. Cirrus are also generated from fronts that span hundreds of kilometers that have synoptic-scale ascent rates of a few cm/s (Starr large- and Wylie, 1989; Mace et ai., al., 1995; M06). In the summer, cirrus often form through convective detrainment (Heymsfield, 1977; M06; Figure 1.4). Mountains can assist in generating cirrus because as air flows over mountain peaks, air is lifted and waves are generated. Cirrus can form in the ascending part of these waves. Mountain-generated cirrus are mostly mid-latitude phenomena that rarely reach the tropopause (Sassen et ai., al., 2001). According to satellite observations, CIrruS cirrus are most often located near the Intertropical Convergence Zone (ITCZ) (Wylie and Menzel, 1999; Mace et ai., al., in press)

18 12 and over the tropical Pacific warm pool. Cirrus clouds have a maximum coverage of more than 60% over the western Pacific warm pool (Mace et al., ai., in press; Figure 1.5), which is larger than at other latitudes (Figure 1.6). Optically thin laminar cirrus exist at the tropopause level in the tropics (McFarquhar et ai., al., 2000; Comstock et al., ai., 2002, G05). These high cirrus facilitate water vapor transport between the troposphere and stratosphere (Boehm et al., ai., 2000). This study will not focus on these high, very thin cirrus and instead the focus will be on the cirrus clouds generated from convection that typically are found in the km region of the tropical upper troposphere. Cirrus are usually located in the atmospheric column between 6 km and 17 km (Platt, (Piatt, 1998). The cirrus cloud top heights are influenced by temperature, vertical winds, and moisture contents. Ice clouds with bases above 6 km with a large vertical extent (thickness> > 3 km) are common when associated with deep convection outflows (Mace, 2007). Cirrus that originate from convection are the focus of this study. Cirrus clouds that originate in outflow from deep convective systems can transition from thick anvils to thin cirrus. In the tropics, dynamics are a main factor in determining these changes in cloud cover (B06). The fraction of cirrus that evolve from convection versus forming from in situ dynamics in the tropics is unknown. Heymsfield and Donner (1990) studied the occurrence of observing both cirrus and cumulonimbus at the same time. However, so far the number of cirrus generated by convection versus in situ processes is not known.

19 13 Figure 1.4 Convective storm with anvil. Photo credit Elisabeth Cohen

20 Anvil Formation In order to determine the importance of anvil blow-off as a cirrus CIrrus formation mechanism, the formation of anvils must first be examined. Anvils develop when thunderstorms send ice crystals high into the upper troposphere. The characteristics of the ice crystals vary depending on the strength of the updraft. Strong updrafts help loft the hydrometeors to the upper levels ofthe cloud. Since updraft speeds in convection can be greater than several meters per second, some of the hydrometeors grow by riming and then fall out as precipitation (Houze, 1993). The hydrometeors that do not fall out of the cloud can either advect into a stratiform cloud region or remain in the convective cell. The stratiform cloud regions have much weaker vertical motions than the convection regions. As hydrometeors slowly descend in the weak updrafts, the hydrometeors grow by vapor diffusion. They eventually melt and precipitate out as stratiform rain (Frederick and Schumacher, 2008). In both deep and weak convection, some hydrometeors can remain aloft to form anvil cirrus. When strong updrafts occur, precipitation size particles may loft to the level of neutral buoyancy and then move laterally to form anvil cloud (Folkins, 2002; Mullendore et al., ai., 2005). When weak updrafts occur, snow or small ice particles may remain aloft even when the convection or stratiform rain region has dissipated. Thunderstorms send ice crystals high into the upper troposphere. The characteristics of the ice crystals vary depending on the strength of the updraft. Strong updrafts help loft the hydrometeors to the upper levels of the cloud. Since updraft speeds in convection can be greater than several meters per second, some of the hydrometeors grow by riming and

21 15 Coverage Figure 1.5 Cirrus cloud occurrence from the CloudSat and Calipso Mission from km at cloud base using approximately one year of data. Image courtesy of Gerald Mace.

22 ,,-..., E ~ >-, 0.4 u [ """ 10 Ql... J..r:. ~ 0.3.~ L QJ LL Latitude Figure 1.6 Vertical occurrence frequencies of vertically averaged clouds for approximately one year from the CloudSat-Calipso data. Image courtesy of Gerald Mace.

23 17 then fall out as precipitation (Houze, 1993). The hydrometeors that do not fall out of the cloud can either advect into a stratiform cloud region or remain in the convective cell. The stratiform cloud regions have much weaker vertical motions than the convection regions. As hydrometeors slowly descend in the weak updrafts, the hydrometeors grow by vapor diffusion. They eventually melt and precipitate out as stratiform rain (Frederick and Schumacher, 2008). In both deep and weak convection, some hydrometeors can remain aloft to form anvil cirrus. When strong updrafts occur, precipitation size particles may loft to the level of neutral buoyancy and then move laterally to form anvil cloud (Folkins, 2002; Mullendore et ai., al., 2005). When weak updrafts occur, snow or small ice particles may remain aloft even when the convection or stratiform rain region has dissipated. An anvil's geometrical dimensions depend on factors such as upper-level wind shear, radiative interactions, upper-level moisture, and the duration and intensity of the initial convection (Frederick and Schumacher, 2008). Shear in the upper troposphere can differentially advect hydrometeors as well as increase their overall coverage Ice Crystal Habits The cloud motions mentioned earlier determine the environment in which the ice crystals grow. The supersaturation, temperature, and sometimes ice nuclei determine the ice crystal shape. The ice crystal shapes and sizes vary in cirrus clouds. Cirrus ice crystal shapes include bullet rosettes, plates, columns, capped columns and needles. There are often quasi-spheres, aggregates of the named shapes, and shapes that are unidentifiable (Figure 1.7).

24 18 Crystal habits are determined detennined by the slowest growing faces on the crystals which are illustrated in Figure 1.8. The faces (B) and (A) are the two slowest growing, whereas other less stable faces (triangular and pentagon faces) grow quickly and create the edges and comers corners of the crystal. The growth rate of the basal faces (B) along the C axis, relative to that of the growth of the prism faces (A), varies with supersaturation and temperature. As temperatures decrease, habits change from a plate to a needle, to a column, to a sector plate to a dendrite and then back to a sector plate, and a column when supersaturation is high with respect to ice. The changes occur at temperatures close to 269 K, 264 K, and 251 K. The changes that occur at 269 K and 264 K are more exact, whereas the transition around 251 K is more spread out. With low supersaturation values with respect to ice, the habit changes from a column to a plate close to 264 K and 251 K (Pruppacher and Klett, 1997). At colder temperatures columns and bullet rosettes grew. Columns formed when temperatures were between 243 K and 253 K and bullet rosettes formed fonned at temperatures between 243 K and 248 K. A combination of bullets, side planes and columns was also present in the temperature region from 243 K to 248 K. Temperature is the main factor determining detennining crystal habit, but the supersaturation is also important. For example, close to 258 K with increasing supersaturation, the habit changes from thick plates, to thin plates, and finally to dendrites. A bullet rosette can form fonn in conditions under which a frozen drop grows by vapor diffusion at 258 K dendritic branches form, fonn, and, since the temperature is so low, columns fonn form and the small amounts of available moisture bring these columns to a point. Ice crystal shape determines detennines the projected area of the crystal (Baumgardner et ai., al.,

25 19 (a) (b) = ~ (c) (f) (g) (h) (i) U) Figure 1.7 Example images from the cloud particle imager (CPI) of (a) small quasisphere, (b) medium quasi-sphere, (c) large quasi-sphere, (d) column, (e) plate, (f) bullet rosette, (g) aggregate of bullet rosettes, (h) aggregate of columns, (i) aggregate of plates, and G) (j) capped column. A small quasi-sphere is an ice crystal with a projected maximum dimension D < 50 l-lm. um. A medium quasi-sphere is classified as having 50 l-lm um < D < 100 l-lm urn and roundness R, which is the measured projected area over the area of the circumscribed circle with diameter D greater than.75. A large quasi sphere has ad a D > 100 l-lm um and R > 0.8 (images courtesy of Junshik lunshik Um Urn and Greg McFarquhar). quasi- C axis B A Figure 1.8 Stable crystal faces for an ice crystal. The faces B and A are the two slowest growing faces, whereas other less stable faces (triangular and pentagon faces) are quickly growing and create the edges and comers corners of the crystal.

26 ), which contributes to the amount of light reflected by cirrus clouds. The albedo of cirrus clouds whose crystal sizes are smaller than 50um 50j..lm (Heymsfield and McFarquhar, 1996; hereafter HM96; Yang et al, ai., 2001) and that are often in high concentrations (HM96; Garrett et al., ai., 2003) are sensitive to the projected area of the crystals (Baumgardner et ai., al., 2005). If the particle habits change, the albedo can be changed as much as 20% (Chepfer et al., ai., 2005) Particle Size Distribution Studying the size distribution of particles within a cirrus cloud helps to determine the processes that are occurring within it. In cirrus, size sorting occurs, with smaller particles in the top portion of the cloud and larger particles at the cloud base (Mace et ai., al., 1997; Smith et ai., al., 2001). More specifically, in the top level of the cloud, nucleation occurs, small ice crystals exist, and the relative humidity is high enough that ice nucleation and production occurs. Ice crystal particle sizes also depend on the source of uplift. In the layer beneath the top layer, ice supersaturation supports the growth of pristine ice crystals. sublimation occurs. The lowest layer of the cloud is the thinnest and is where Rounded amorphous crystals make up this layer and crystals sublimate due to ice subsaturation saturation (Deng, 2006). As cirrus clouds evolve, they often become semitransparent. This is due to low concentrations of relatively large ice crystals (Sassen, 2002). The sizes of ice crystals are often measured using an effective radius, which is equal to the total volume of the ice particles divided by the projected area of the ice particles. This includes the ratio of the 3 rd moment to the 2 nd moment of the particle size

27 21 distribution. The effective radius can be measured using the particle size distribution with assumption of ice particle habits or the IWC and extinction. Adjusting the effective radius of tropical high cirrus from 30!lm um to 10!lm um in interactive climate models corresponds to an increase in the magnitude of the surface shortwave and TOA cloud forcing by approximately 25% (McFarquhar et ai., al., 2003). This change in effective radius also increases upper tropospheric temperatures in the tropics by ~1 ~ 1 Kelvin (Kristjansson et ai., al., 2000). Thus, understanding the sizes of crystals in clouds, and what determines their sizes over time, will help answer questions as to how cirrus distribute energy in the tropics. The average particle size of ice crystals is debated. Crystal particles size length estimates range from approximately 1 to 8000!lm um (Dowling and Radke, 1990; Gardiner and Hallett, 1985; Arnott Amott et ai., al., 2000) Cirrus Optical Depth and Ice Water Content The crystal habit, IWC, and crystal size distribution are important in the calculation of the cirrus optical depth. The net global radiation is extremely sensitive to the optical depth of cirrus since higher optical depth clouds are more reflective. The optical depth is proportional to twice the cross-sectional area of the ice crystals. In a uniform cloud of thickness I1z, Az, the optical depth 'r x would be paz, Pl1z, where p P is the extinction coefficient at visible wavelengths. Optical depths of cirrus clouds range between less than 0.03 for subvisible cirrus (Sassen and Cho, 1992) to greater than 3.0 for thick cirrus. The optical depth of cumulonimbus anvils usually surpasses the traditional criteria for classification, and as

28 22 the clouds evolve, their optical thicknesses often decrease away from convection. The optical depth also depends on the ice water in the cloud or IWC. Since, IWC strongly relates to optical thickness, IWC helps determine cirrus shortwave radiative 3 3 properties. Ice water contents for thin cirrus are on the order of 10-3 gm- 3 (Heymsfield, properties. Ice water contents for thin cirrus are on the order of 10" gm" (Heymsfield, 1977). Average water content values of g m- 3 and concentrations of 0.03 cm- 3 are 1977). Average water content values of g m" 3 and concentrations of 0.03 cm" 3 are typical in cirrus (Dowling and Radke, 1990). Ice water content values in anvil cirrus typical in cirrus (Dowling and Radke, 1990). Ice water content values in anvil cirrus were observed to range between 10-1 gm- 3 within 1 km from the upwind edge of the anvil were observed to range between 10" gm" within 1 km from the upwind edge of the anvil to values close to 10-5 gm- 3 about 50 km from the upwind edge of the anvil (G05) 5 3 to values Ice close water to 10" content gm" about is fundamentally 50 km from the a upwind function edge of of temperature the anvil (G05) through the Clausius-Clapeyron Ice water content relationship. is fundamentally Ice water also a depends function on relative of temperature humidity, through wind speed the and Clausius-Clapeyron lapse rate (Heymsfield relationship. and Ice Donner, water 1990; also depends M06). Large-scale on relative humidity, vertical motion wind speed also impacts and lapse the rate IWC (Heymsfield of cirrus (M06). and Donner, For example, 1990; the M06). mean Large-scale IWC increases vertical (nearly motion doubles) also in impacts rising the air IWC as opposed of cirrus to a (M06). subsidence For example, regime (M06). the mean IWC increases (nearly doubles) in rising Changes air as opposed in IWC to change a subsidence the heating regime in tropical (M06). clouds. Increasing IWC in clouds increases Changes cloud top in IWC cooling change and the cloud heating base in warming tropical clouds. and small Increasing IWC amounts IWC in lead clouds to warming increases everywhere cloud top cooling the cloud and (A88). cloud base warming and small IWC amounts lead to warming everywhere in the cloud (A88) Decay Rates Deep convection can affect cirrus proprieties depending on the distance of the cirrus from the convective core, the environmental wind shear, and the moisture injected from the boundary layer (HM96; MH96). Deep convective systems often have large cirrus optical thicknesses and high cirrus tops (LR04). Both weak and strong convective systems have similar cirrus decay rates (LR04). About a day or so after cirrus leave a

29 23 convective region, the top of the cirrus generally stabilizes at around 300 hpa and the signs of the original convective strength are no longer apparent. After two days, there is no sign of the original convective strength in the cirrus optical thickness (LR04). On average, deep convection that produces anvils decays after 6-12 hours (LR04). The decay of deep convection is associated with the growth of both cirrostratus and cirrus. Cirrostratus then decays and cirrus continue to grow. The processes that dictate this evolution are not well understood. According to LR04 tropical cirrus systems last for ± 16 h. They often travel over distances of about km during their existence. The time it takes a single ice particle to move from convection to cirrostratus is about 6 hours and usually covers a distance of 200 km (LR04). The original individual ice particles within the anvil will likely fall out at approximately that time, so the duration of the system is longer than that of the original condensate Maintaining Cirrus Cirrus particles often do not last as long as the tropical cirrus systems they make up, demonstrating that some other process is replenishing the particles (LR04). Processes exist that maintain cirrus clouds after they leave a cumulonimbus anvil. One possible mechanism that might be related to maintaining the cirrus clouds is heating within the cirrus clouds. Infrared exchange heats tropical anvils, since the earth's surface in the tropics is warm and the temperature of cloud tops is cold. As mentioned earlier, A88 find that layer average heating rates in anvils that are 2 km thick are on the order of 20 to 30 K/day, KJday, and Garrett et al. (2005) find heating rates that are even higher. This heating is substantial and there must be an atmospheric response. A cirrus cloud at

30 24 15 km that is being heated by 30 K/day KJday would end up somewhere in the stratosphere if it rose to its equilibrium temperature. Obviously, this is not the case, and instead the heat is released by another process. For example, cloud layers from 200 to 600 hpa over the Malaysian region generated a net warming of 5 to 10 K/day KJ during the monsoon; Webster and Stephens (1980) suggest that this warming could partially maintain the cloud through lifting. Another mechanism that could maintain cirrus may be related to how the optical depth or thickness of clouds changes the stabilization of the atmosphere. For example, thinner clouds have stronger radiative destabilization since they have strong temperature gradients due to cool cloud tops and warm cloud bases. Thicker clouds do not see such strong gradients. When heating gradients do exist, they act to lift and spread the cloud. The lifting that occurs due to heating gradients can be cancelled by crystal settling (A88). An alternative explanation for cirrus maintenance can be found in Garrett et al. (2005) and Garrett et al. (2006), which describes that anvils may spread with time which can lead to the cirrus thinning. Luo and Rossow (2004) suggest that since cirrostratus rapidly decay downstream of convection after about km or 24 hours, a loss of cloud water mass occurs. Luo and Rossow (2004) note that it is likely that the cirrostratus that come from convection have much larger water mass and particle sizes than cirrus and so the cirrostratus have a larger sedimentation mass flux. At the same time, mesoscale circulations are providing water vapor from below. This water spreads over the expanding area, which is two to three times greater than the area of the convective towers (LR04).

31 Cirrus and Convection in Climate Models A global circulation model (GCM) is a predictive numerical model of the global atmosphere. Atmospheric variables such as temperature and winds are predicted. To predict these variables, GCMs solve the three primitive equations of motion, the conservation of momentum, thermal energy transfer and the continuity equation. The GCM solves these equations and predicts the variables for the entire Earth. The resolution and grid spacing in such models are coarse in both space and time. For example, the resolution of the Hadley Centre model, HadAM3, is latitude, longitude, and 19 vertical levels. The time step is 30 minutes and the HadAM3 has three sub-time-steps per time-step in the dynamics. The large spatial and temporal scale means that these models do not capture small-scale events such as the detrainment of cirrus clouds from convection. Although large-scale clouds are resolved in GCMs, the processes within them and the environmental properties around the clouds are not. Since these macrophysical and microphysical properties of cirrus clouds are not resolved with sufficient detail, a parameterization is carried out. A parameterization is an estimation of subgrid-scale properties so the small-scale properties and processes can work within the large-scale state (Randall et ai., al., 2003). Parameterizations are needed to reduce computational expense, to simplify complex processes that cannot be resolved properly, and for processes that are not understood well enough to be represented correctly in models. This study focuses on detrainment of cirrus clouds from convection. In global climate models, a convective parameterization (CP) is made to simplify the convective process. The parameterization does not capture all of the processes that occur during and

32 26 after convection. The main function of a convective parameterization is to redistribute moisture and temperature in a grid column. Since the focus in the CP scheme is on heating rates; precipitation is just a byproduct. Unfortunately, this means that the precipitation and moisture lost from convective clouds to the ground has not been a primary focus. Some models transport cloud water content from one area to another. For processes such as anvil detrainment from convection, the net transport of water away from convection is important. Without the movement of cloud water content, cloud properties cannot advect from one area to the next. Not transporting water correctly is detrimental to representing clouds. Accounting for water mass in clouds is difficult, especially because most models do not include processes that affect it. For example, the GISS GCM uses a mass flux cumulus parameterization that cannot incorporate changes in updraft strength to interact with cloud microphysics. Condensate is, therefore, not transported to the correct locations in the cloud, and the incorrect condensate amounts affect the radiative properties of the clouds and this moisture is not accurately converted into either precipitation or kept in the cloud. This, in turn, tum, affects the hydrologic cycle and, indirectly, convection in the models. The convective parameterization works in conjunction with a precipitation and cloud processes (PCP) parameterization. The PCP removes extra atmospheric moisture from forecasted temperature, wind, and moisture fields. Precipitation schemes that forecast mass fluxes of water make random assumptions about the amount of condensed water in updrafts that becomes precipitation since droplet size distributions are not computed in GCMs (Del Genio and Kovari, 2002). This makes the predictions of rain

33 27 and water vapor left in the cloud less accurate. The PCP process parameterizes grid-scale motions to produce latent heat from condensation, which affects wind and temperature structure. The PCP scheme also includes evaporative cooling underneath the cloud where rain evaporates into subsaturated layers. The PCP has more realistic large-scale convection when partnered with the convective parameterization since the convective parameterization increases the stability of the atmosphere. These schemes are embedded in GCMs. The mass fluxes and the heating and cooling from the PCP and CP schemes mentioned earlier are incorporated into the GCM and the temperature fields are adjusted. The GCMs that track moisture adjust their values depending on the predicted rainfall, and then the GCM continues with its computations for surrounding areas and times. Detrained cirrus clouds that become self-maintaining will not be accurately represented in GCMs without correctly tracking the amount of moisture that remains after a convective event. The predicted variables that are important for representing cirrus are temperature, moisture content, and winds. In addition to the PCP and CP parameterization schemes, other processes and variables that are important to cirrus formation are also parameterized in GCMs. These include emission from clouds, deep convection (warming), rain (cooling), reflection, condensation, snow, and turbulence. Within clouds, particle size, mixing ratio, concentration, and optical thickness are also parameterized in climate models. To parameterize optical thickness, models often use a cloud's altitude. High clouds are made less optically thick than low clouds in most models (Del Genio et al., ai., 1996). Variables such as condensate mixing ratios are averaged over grid boxes.

34 28 The inaccuracies in the representation of upper tropospheric clouds in GCMs is a major source of uncertainty for climate forecasts and is the largest uncertainty for climate change projections (Jakob, 2003; LR04). There are uncertainties in the available cloud data that lead to difficulties in quantifying quantifying the cloud errors in models (Zhang et ai., al., 2005). Tropical cirrus clouds exist in extreme conditions and at very high altitudes, which makes it even more difficult to collect data (LR04). Also, the viewing geometries of clouds for both surface and satellite observations blocks some clouds from being observed, and therefore they are missing in the data (Zhang et ai., al., 2005). Collecting and studying high quality data is important for understanding cirrus clouds. The TWP-ICE field campaign was designed to study detrained cirrus clouds and the convection that is associated with them. Studies such as TWP-ICE help to improve our understanding of cirrus clouds. For example, the TWP-ICE experiment results will help explain how moisture is moved away from a convective area. This understanding, in turn, can help modelers better represent this cloud process in global circulation models (GCMs). One goal of the TWP-ICE experiment is to understand the properties of detrained cirrus from convection. The physics of cirrus clouds is incorrectly parameterized in models and this prevents clear understanding of the impact of cirrus on climate change (Stephens et al.,1990). ai.,1990). The observations of cirrus will contribute to representing the variables and properties of cirrus correctly.

35 Tropical Warm Pool International Cloud Experiment Many field campaigns have studied cirrus. For example, The Tropical Ocean Global Atmosphere Coupled Ocean-Atmosphere Response Experiment (TOGA COARE) focused mostly on sea-air interactions and deep convection. During this experiment, they made observations that have been used to study anvil cirrus in the tropical western pacific (Ye, 2000). Measurements of the microphysical properties of tropical cirrus anvils (MH96) taken during the Central Equatorial Pacific Experiment (CEPEX) experiment clarified the association between cirrus microphysical properties and their relatively high albedo (HM96). The Cirrus Regional Study of Tropical Anvils and Cirrus Layers-Florida Area Cirrus Experiment (CRYSTAL-FACE) investigated tropical cirrus cloud physical properties and formation processes. Measurements of cirrus anvil properties were taken through as much of the cloud life cycle as possible to advance our knowledge of anvil cirrus evolution (Garrett et al., 2006). TWP-ICE took place in Darwin, Australia, from 19 January, 2006, to 14 February, 2006, during the Australian monsoon. This project was a collaboration between the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement Program (ARM) and the Australian Bureau of Meteorology (BOM) Australian Summer Monsoon TWP-ICE was scheduled for January and February of 2006 since these months are in the wet phase of the Australian monsoon (May and Jakob, 2004). The main feature of the monsoon in Darwin is the presence of westerly winds between 850 and 700 hpa. In Darwin, monsoon season usually occurs from December to March and falls

36 30 within the Darwin wet season, which is from 1 October to 30 April (Hastenrath, 1991). In northern Australia during the monsoon months average rainfall totals range between 400 and 600 mm (Frederick and Schumacher, 2008). West-to-northwesterly winds bring moist air from the Indian Ocean into the monsoon trough. This sparks convective activity over the region. In this region in January, the main global monsoon heat source is located over Northern Australia (May and Jakob, 2004). During the monsoon the diurnal cycle is relatively weak, with a nocturnal temperature maximum (Keenan et ai., al., 1989). During an active monsoon, convection is often widespread. The convection that occurs in Darwin during this time is mostly of maritime origin, with a large fetch over the water. The convection is typical of maritime convection since it is widespread, has complex organizational structure, and lacks the intensity and depth of typical continental or coastal convection (May and Jakob, 2004). When the winds change from a westerly direction, this is called a break in the monsoon (Figure 1.9). Break periods have strong diurnal cycles with late afternoon maximums in convection and the maximum convection is often associated with a sea breeze and other local circulations. The break periods often have lightning activity. Three different weather regimes were studied during the TWP-ICE experiment, an active monsoon, a relatively suppressed monsoon, and a break period. There were a few days of clear skies between the suppressed monsoon and the break. The event sequence was associated with a large-amplitude Madden-Julian Oscillation event. A mesoscale convective system (MCS) affected the TWP-ICE experiment and had an area-averaged rainfall accumulation of more than 45 mm (May et. ai., al., 2008).

37 31 'iii cu ~ bo cu "0 - c: 0 'B ~ C "0 c: '0 '\.\'\.~\~ I hpa hpa 500 hpa TWP-ICE Monsoon Figure 1.9 Wind shifts during the during the TWP-ICE experiment. The circled area is when the winds shifted direction and created a break period (image courtesy of Lori Chappell). During the large break period during TWP-ICE, intense convection was studied since there are still questions about the impact convection has on the near environment. The convection during the monsoon break is more representative of coastal convection (May and Jakob, 2004). During the break, convective storms above the Tiwi Islands, which are approximately 150 km in length and are immediately north of Darwin, are the research focus. The next chapter of this thesis describes the methods used for tracking the cirrus clouds using satellite imagery and describes the instrument used on the aircraft during the experiment. The Results section describes the weather regimes for each of the flights

38 32 during TWP-ICE and describes the particle size and habits sampled. The cloud track, time since convection and temperature evolution of the clouds will also be included in the Results.

39 CHAPTER 2 METHODS The purpose of this study is to understand how CIrruS cirrus clouds evolve from convection until they dissipate or become self maintaining. To study this, the source of the cirrus and the amount of time passing since the clouds sampled were associated with convection, or in other words, a cloud's age, are found. The age of clouds sampled during TWP-ICE varied. By comparing fresh or newly formed cirrus with cirrus that are aged or have been evolving for many hours, cirrus evolution can be studied. The tracking algorithm described here reveals the clouds' trajectories during their evolution. The first part of the Methods section broadly summarizes the entire tracking algorithm. The second section goes through an overview of each of the steps in the algorithm and the third section describes the assumptions made for this algorithm. The fourth section details each of the smaller steps taken in the algorithm. The final section describes the instruments and aircraft which were used to study the data in situ. 2.1 Tracking Algorithm Summary The tracking algorithm used here is based on the algorithm first introduced by Soden (1998; hereafter S98). To track upper tropospheric moisture, he followed areas of moisture through their evolution in time and space using satellite imagery. Soden's