Aerosol-Cloud-Precipitation Interactions in warm clouds
Graham Feingold
NOAA Earth System Research Laboratory Boulder, Colorado USA
Keck Ins tute for Space Studies Geoengineering Workshop
May 2011 Acknowledgements
• Keck Ins tute of Space Sciences • Present and former group members and colleagues – Hailong Wang, Huiwen Xue, Hongli Jiang, Adrian Hill, Armin Sorooshian – Ilan Koren, Phil Rasch, Bjorn Stevens, Patrick Chuang, Jen Small, Shouping Wang Aerosol Effects on Clouds
Shiptracks
(Courtesy, NASA)
Ri s
Stevens et al. 2005 Outline • Simple constructs that may have outlived their usefulness – Albedo effect – Life me effect – nth indirect effect • An cipated and unan cipated responses
• Correla on vs. Causality
• Self-organizing systems
• Bistability and the role of Aerosol
• Numerical Ship track experiments Some simple cloud physics
• Clouds are dynamical en es and the characteris cs of the cloud field depend on the meteorological regime – To first order, cloud frac on and depth or LWP dictate albedo
• More aerosol à more drops (almost always) – But drop size response is unpredictable
d N
! Nd ! Na Drop concentra on,
Ramanathan et al. 2001 Aerosol concentra on, Na Some simple cloud physics
Small droplets don’t collide Collector drop radius easily with large droplets
2 2 K(x, y) ∝ (rx + ry )Vx −Vy E(x, y) Collected drop radius, µm Wallace and Hobbs 2006 E(x,y) = collec on efficiency
• Smaller drops are less likely to collide and coalesce; need ~ 20 µm radius drops to ini ate the process Some simple cloud physics Clean Polluted
-3 N = 300 cm-3 Na = 50 cm a t=0 t=0 r=20 µm r=20 µm
t=10 min
t=10 min
Drop radius, µm
• Smaller drops are less likely to collide and coalesce; need ~ 20 µm radius drops to ini ate the process Spatial/Temporal scales
• Involves complexity in both aerosol and clouds • Range of spa al scales – Aerosol par cles 10s – 1000s nanometres – Cloud drops/ice par cles: µm – cm – Cloud scales: ~ 10 m – 103 km 4:00 pm EDT, July 23, 2002 • Range of temporal scales – Ac va on process (aerosolà droplet): < sec – Time to generate precipita on ~ 20 min – Cloud systems: days • Coupled System – Scale interac ons
10 km 10 km 0 Some pitfalls
Correla on vs Causa on
Most satellite studies show correla ons between cloud and aerosol Increasingly, model reanalysis is being used to separate aerosol from meteorological influences
Ice cream causes shark a acks Satellite remote-sensing Breon et al. 2002 Shark Attacks Attacks Shark
Dropradius “slope”=0.05 – 0.08
Ice Cream Sales Aerosol Index Some pitfalls contd..
Remote sensing of aerosol-cloud interac ons from space is hard!
- Adjacency
- “Cloud contamina on”
- 3-D effects
Photo credit: Interna onal Space Sta on
Some pitfalls contd..
Models are imperfect!
- Range of scales is a huge challenge
4:00 pm EDT, July 23, 2002 - “All models are wrong: some are useful”
George Box
10 km 10 km 0 What controls cloud albedo?
1/3 5/6 ! ~ Nd LWP A ~ ! (for low )
• Cloud Op cal depth is 2.5 x more sensi ve to changes in LWP
than changes in Nd
• LWP has a much stronger control on cloud op cal depth than
Nd
A useful metric
Albedo Suscep bility Nd =const Which clouds are most sensi ve to
increases in Nd (vis-à-vis albedo)? dA/dNd
dA A(1! A) 0.5 1 S = = A 0 dN 3N d LWP,... d ! Nd ! Na
" 5 d ln LWP d lnk % S0!,LWP,... = S0! $1+ + ' # 2 d ln Nd d ln Nd &
LWP response breadth response Platnick and Twomey 1994 Note: breadth k !1/ Feingold and Siebert 2008
Influences of aerosol on distribution breadth d lnk d ln Nd Not accoun ng for breadth effects Accoun ng for breadth effects Non-precipita ng clouds* Xue and Feingold, 2004
1/3 At constant LWP: τ /τ0 = (Nd / Nd0)
• Non precipita ng clouds: breadth effects reduce albedo suscep bilty • Precipita ng clouds: breadth effects enhance albedo suscep bilty
*For precipita ng clouds see Feingold et al. JGR 1997 How does LWP respond to aerosol?
Sign of the change is uncertain! " 5 d ln LWP % S0!,LWP,... = S0! $1+ +..' # 2 d ln Nd &
Δ ln LWP β = Δ ln Nd
Han et al. 2002: ISCPP Data
Shiptracks shiptrack control re is smaller in ship tracks
Cloud op cal depth is # ~ constant
re LWP tends to decrease
# Reduced Increased LWP LWP
Figure 4. Average optical depths and droplet effective radii for contaminated (solid line) and uncontaminated (dashed line) pixels takenCoakley from 30-km and Walsh, 2002 track segments. The number of track segments, means and standard deviations of the distributions are given.
Figure 8. " !ln# / !ln Re for the 30-km ship track segments in which the average change in droplet effective radius, uncontaminated " contaminated, was greater than 2 µm.
" !ln# / !ln Re = 1 if the cloud liquid water amount remains constant. The number of segments, means, and estimates of two standard deviations for the ensemble means are given.
Global low cloud satellite measurements and NCEP reanalysis r stability e Drop size decreases with aerosol
Aerosol Index
LWP LWP decreases with aerosol
Aerosol Index
Cloud Albedo Albedo ~ constant (compe ng effects)
Aerosol Index Matsui et al. 2006 D15205 LEBSOCK ET AL.: AEROSOL INDIRECT EFFECTS D15205
the albedo enhancement is indeed diminished in the lower static stability regime presumably because the LWP de- crease is larger in these environments than in the high stability regime. This independent result adds credence to the observations that clouds in unstable environments tend to suffer decreases in LWP not observed in stable environ- ments. The analysis has been performed with both the CERES TOA cloudy sky albedo and a cloud top albedo derived using the MODIS cloud products (t, re)andthetwo stream approximation. Both estimates of cloud albedo lead to the same conclusions, therefore the following results show the CERES albedo because it provides an independent estimate of this quantity. Figure 4 demonstrates that the two products are well correlated. The high bias of the MODIS estimate results from the sensor resolution differences and the fact that MODIS is a cloud top estimate as opposed to Figure 4. Comparison of the CERES TOA albedo with the TOA estimate of CERES. the cloud top albedo derived from the MODIS cloud [19]Althoughthepreviousresultssuggestthatthecon- products using a two stream radiative transfer model. stant liquid water path approximation is not valid on the global scale, it is still of interest to examine the constant MODIS cloud properties and the CloudSat rain flag and liquid water path case. To do this cloud properties are may contain some clear sky. In fact, on average, the CERES examined as a function of LWP. For example, the CERES pixels used in this analysis contain 19.2% clear sky. To cloud albedo is plotted against LWP for four stability and account for this fact and to eliminate any relationship aerosol regimes in Figure 5A. This plot demonstrates that between cloud fraction and AI the cloudy sky albedo is for a constant LWP the increase in albedo is greater for high derived from the TOA flux. First dividing the observed LWP clouds than it is for low LWP clouds. This result might upwelling TOA flux by the calculated incident TOA flux be slightly counter-intuitive, as one would expect a satura- derives the all sky albedo. Second, the clear sky component tion of the cloud albedo for high LWP. An explanation for of the flux is removed using the CERES estimate of clear this phenomenon can be found in Figure 5B, which shows sky area within the CERES field of view via, re as a function of LWP for the same stability and aerosol regimes as in Figure 5A. The sensitivity of the effective radius to AI increases with LWP and this in turn results in aall 1 f aclr acld À ðÞÀ ; 4 an increased sensitivity of the optical depth to AI. It can be ¼ f ð Þ predicted that the sensitivity of re to aerosol should increase with increasing LWP by considering the LWP of a vertically where f is the MODIS cloud fraction within the CERES homogenous cloud with a mean volume radius rv,number footprint and a is the albedo. concentration N,andthicknessH, [18]ThetrendoftheCEREScloudyskyalbedowithAI is given in Figure 1E and again the magnitude of the trends 4 3 are shown in Table 2. These independent estimates of LWP pr rv NH: 5 ¼ 3 l ð Þ albedo agree wellMeteorological with the previous results, indicating context that for aerosol influences
Figure 5. Trends ofα the = A = albedo (A) CERES cloudy sky albedo and (B) MODIS re with the AMSR-E LWP. The delineation between cleancld and dirty is given by AI = 0.1. The stable curve represents all LTSS between 18 and 21 K. The unstable curve represents all LTSS between 12 and 15 K. Lebsock et al. 2008 A-Train data 6of12 Modeling of Aerosol effects on Stratocumulus in LES
Models help determine causality
Pollu on elevates cloud tops LWP decreases with increasing aerosol:
Evapora on-entrainment feedback polluted
clean
height 100/cc 1000/cc Bulk microphysics
LWC Also, Xue and Feingold 2006 Wang, Wang, Feingold, 2003 for cumulus Modeling of Aerosol effects on Stratocumulus in LES
Rela vely moist above inversion
Very dry above inversion
dLWP >0 ? dNd <0 Humidity of air above stratocumulus is important
Ackerman et al. 2004 Modeling of Aerosol effects on Stratocumulus in LES
Fine resolu on FIRE-I Scu, non precipita ng
Low aerosol
- Coarse resolu on has lower LWP High aerosol
- Reduc on in LWP with Δx = Δy = 20 m; Δz = 10 m increasing aerosol
Coarse resolu on - Effect of representa on of Low aerosol mixing (homogeneous vs. inhomogeneous; black vs grey) is small
High aerosol
Δx = Δy = 40 m; Δz = 20 m
Hill et al. 2009 Cumulus Clouds Cumulus Clouds Height, m
12 km 12 km
Houston, GoMACCS 2006: Mike Hubbel adiaba c • Large varia on in cloud depth • Significant entrainment
• * Subadiaba c liquid water Height, m • Cloud Frac on ~ 10% • Frequent in trade wind regime Adiaba c ra o • Precipita on common *Warner et al. 1955 Lu et al. 2008 -3 clean (Na = 25 cm ) Time series -3 polluted (Na = 2000 cm )
LWP_cloud (g m-2) LWP_domain (g m-2) Height, m cloud frac on cloud top height (m)
12 km 12 km
TKE (kg s-2) surface drizzle rate (mm day-1)
precip suppression
Time (h) Time (h) Xue and Feingold, JAS 2006 Model of a Cumulus Cloud Contours: buoyancy Arrows: wind vectors
Posi ve buoyancy
nega ve buoyancy
Zhao and Aus n 2005 LES output
! = vor city B = buoyancy The link between smaller droplets, faster evaporation and dynamics: Turbulent flux and buoyancy flux
Downdra regions
Turbulent flux Buoyancy flux evapora ve cooling is larger for polluted clouds clean polluted evidence of enhanced evapora ve cooling 2 -2 w' ' (K m s-1 ) w'w' (m s ) θv & higher turbulence due to aerosol: Smaller drops Xue and Feingold, JAS 2006 evaporate faster buoyancy vor city dB/dz
Con nental Cumulus
Entrainment: dB/dz>0; Detrainment: dB/dz<0
Trade Cumulus Polluted Clean Normalized Cloud Depth
Con nental Cumulus (single 2-D cloud simula ons; Tel Aviv Univ.)
Jiang et al., GRL 2006 Evidence from Observa ons: GoMACCS (Houston 2006)
Observa ons Model nega ve posi ve
clean Normalized height clean polluted
Normalized height polluted Jiang et al. 2006
buoyancy
Small et al., GRL 2009 Aerosol Effects on Cloud Lifetime
Liquid water path: snapshot 3000 m
Track life me of individual clouds y-direc on
-3000 m 3000 m x-direc on Aerosol Effects on Cloud Lifetime Based on analysis of 100s of individual clouds using object oriented cloud tracking tools
Jiang, Xue, Teller, Feingold, Levin: GRL 2006 Why? Competing Aerosol effects on Cloud Microphysics
- Small droplets do not coalesce efficientlyà less rain
vs.
- Small droplets evaporate faster than large ones dr S Ratio of timescales for evaporation (clean vs polluted) ! may be a factor of 5-10 dt r
Beware of simple constructs
More aerosol à more drops à less coalescenceà less rain à higher LWP à higher cloud frac onà longer life me ? ? ? A non monotonic response…
Precipita on suppression Enhanced evapora on increases cloud frac on due to smaller drops ATEX decreases cloud frac on Xue, Feingold, Stevens, 2008 Cloud frac on
Aerosol concentra on, cm-3
- Microphysical feedbacks complicate the simple monotonic response - Rain, LWP, cloud frac on and life me responses are not necessarily connected Precipitating Warm Clouds ! !"#$%&'()*+',-.(%'($/)$#-0%*1-(2$-3)4%*5-(*-3$#*#4',$- "##
*90 Macrophysics vs Microphysics !"#$
0 3 • Measurements show that Rainrate R#! ~ H /N or )/ R ~ LWP1.5/N #. !"#% ()*+,)- "#$%
!
• Rain produc on is 1.5 -2.5 x more sensi ve to changes in LWP ' than changes in N !"#& • H or LWP is a much stronger control of rainrate!"#!% than N!" #!$ !4&$ !4&$ #. 7 !4&$ 123 )5)'&'( )6*+,)- )- 0 8 – For accre on, R is independent of N *:0
-1 !"#$ 0 #! )/ #. , mm d ()*+,)- #% "#$% !" ! ' Rainrate
!"#& "4! !4" 7 7 7 '(( )5)'&'()*- )- 0 H3 N-1, m6 Van Zanten et al. 2005; Wood and Brenguier, 2008 !"#$%&'()*('+,-./".$&01' $%&'(%$!)&*+!)+5#)(6!+%,-.&%+%/0-!,/$!$&*12%0!3*/3%/0&,4
0'*/5!7'4(5!$%&'(%$!)&*+!-,+12%-!/*0!,))%30%$!67!+'8'/9!*&!$&*12%0!-3,(%/9'/9:!;3?@! ;-A.,&%-<5!,/$!+%,/! $&*12%0!3*/3%/0&,0'*/!745!$%&'(%$!)&*+!-.&),3%!&%+*0%!-%/-'/9:!;$
Precipita on suscep bility Parcel model Which clouds are most sensi ve to increases in Nd (vis-à-vis precip)? 1.1 d ln R R0! = " d ln Nd
1.0 AMSR-E R CloudSat R
0.8 ) AI LES of TradeCu
)/dln( 0.6 R 0.6 = -dln(
0.4 o o ' S R 600 1000 0.2 -2 A-Train LWP, g m 0.0 600 800 1000 -2 Feingold and Siebert 2008; AMSR-E LWP (g m ) Sorooshian et al., 2009 Parcel model Extrapola on to Tropical Oceans
Large eddy simulation 1.10 Cloud Parcel Model: Adiabatic Calculate expected precipita on Sub-adiabatic
) 1.00 d reduc on in a given region N 0.90 )/dln( R 0.80 = -dln( =
o ’ S Input: LWP and aerosol perturba on 0 0.70 R LES (TradeCu) 0.60
200 400 600 800 1000 1200 -2 LWP (g m )
25 600 500 LWP m (g 400 0 300
200 -2 )
Latitude (deg)Latitude 100 -25
25 0.30 AI: σ
0.20 /mean 0 0.10 0.00 Latitude (deg)Latitude -25 -160 -120 -80 -40 0 40 80 120 160 Longitude (deg) Sorooshian et al. 2009 Where on Earth might aerosol reduce precipita on?
25 -20
-15 Δ
R
0 -10 (%) -5 Latitude Latitude (deg) -25 0 -160 -120 -80 -40 0 40 80 120 160 Longitude (deg) -2.6% -28.3% -8.9% -18.5%
What about the ABSOLUTE reduc on in precipita on?
25 -0.10 Δ
-0.08 R -0.06 (mm h 0 -0.04 -1
-0.02 ) Latitude (deg) Latitude -25 0.00 -160 -120 -80 -40 0 40 80 120 160 Longitude (deg)
Note: Ro may be biased in certain regions characterized by persistent above-cloud aerosol layers (CALIPSO can help with this). Sorooshian et al. 2009 D15205 LEBSOCK ET AL.: AEROSOL INDIRECT EFFECTS D15205
D15205 LEBSOCK ET AL.: AEROSOL INDIRECT EFFECTS D15205
Albedo Susceptibility Precipita ng clouds Non-precipita ng clouds = A = albedo cld α
Figure 6. Global relationships between AI and nonprecipitating, transitional, and precipitating cloud parameters. Circles represent the mean values and error bars show" the5 standardd ln L deviation.WP d Thelnk LWP% estimate is for the cloud component' d ln only.A The reflectivity is the CPR vertical cloud mean reflectivity factor. S = S0!,LWP,... = S0! $1+ + ' 0 2 d ln N d ln N d ln Nd # d d &
Lebsock et al. 2008 LWP response breadth response a Table 3. Slope of Linear Fit Between Cloud Parameters and log10(AI) Figure 6. Global relationships between AI and nonprecipitating, transitional, and precipitating cloud Parameterparameters. CirclesUnits represent the meanNonprecipitating values and error barsTransitional show the standardPrecipitating deviation. The LWPAll Clouds re estimate is for the cloudmicron component only. The2.25 reflectivity is the CPR2.04 vertical cloud mean1.37 reflectivity factor. 2.32 t none À0.90 À2.37 À2.01 À0.67 LWP (AMSR-E) g/m2 0.29 21.32 41.73 0.30 Cloudy sky albedo none 0.007 0.027 0.040 0.008 Reflectivity dBZ 0.57 1.14 0.69 0.37 À À À À aComparison of non-precipitating, transitional and precipitating clouds.
a Table 3. Slope of Linear Fit Between Cloud Parameters and log10(AI) 8of12 Parameter Units Nonprecipitating Transitional Precipitating All Clouds re micron 2.25 2.04 1.37 2.32 t none À0.90 À2.37 À2.01 À0.67 LWP (AMSR-E) g/m2 0.29 21.32 41.73 0.30 Cloudy sky albedo none 0.007 0.027 0.040 0.008 Reflectivity dBZ 0.57 1.14 0.69 0.37 À À À À aComparison of non-precipitating, transitional and precipitating clouds.
8of12 A picture seems to be emerging…
A non monotonic response…
Precipita ng clouds Non-precipita ng clouds LWP Cloud frac on
Aerosol concentra on
Based on Wang et al. 2003, Ackerman et al. 2004, Lu and Seinfeld 2005, Matsui et al. 2006, Xue et al. 2008, Lebsock et al. 2008, Wang et al. 2011 The global picture
D15205 LEBSOCK ET AL.: AEROSOL INDIRECT EFFECTS D15205 dA
d log10 AI
Strongest in: extratropics, subtropical stratus winter strongly capped MBL regions less apt to show decreases in LWP Lebsock et al. 2008
Figure 9. Regional seasonal maps of the slope of a linear fit between the cloudy sky albedo and log10(AI). All red and blue regions are statistically significant at the 95% level.
10 of 12 Self-organization in the aerosol-cloud-precipitation system Mesoscale Cellular Convection
MODIS image, South East Pacific - Radia ve forcing of the ocean-cloud system - Fundamental understanding of closed/open-cell evolu on MODIS image, South East Pacific Self-Organization in a Bowl of Soup
Warm currents rise; Cold surface currents sink; Opposite movements cannot take place at the same me without self-organiza on;
Cellular structures emerge; Rayleigh-Benard cells
Spontaneous crea on of globally coherent pa erns out of local interac ons
Convec onal and sedimenta on dissipa ve pa erns of Miso soup Tsuneo Okubo. Colloid Polym Sci (2009) 287:167–178 System Equilibria
Stable Equilibria Non-drizzling regime A ractors
dCCN/dt A Drizzling regime
B
CCN, m-3 Lorenz, 1954
Baker and Charlson, 1990 Large eddy simulation of open and closed-cells
Feingold, Koren, Wang, Xue, Brewer (2010)
See also Stevens et al. 2005; Savic-Jovcic and Stevens 2008; Xue et al. 2008; Wang and Feingold 2009 Aerosol/drizzle selects the state Albedo Albedo
Closed-cell Albedo ~ 0.6 high aerosol (non- precipita ng)
Onset of WRF Model drizzle + 2-moment results in (i) Aerosol µphysics; “selects” the transi on state of the system 60 km domain; to open-cell (same meteorology) Δx = Δy = 300 m z = 30 m convec on Δ DYCOMS-II (ii) The stable state oscillates
Open-cell low aerosol Albedo ~ 0.2 (precipita ng)
Garay et al. 2004, MISR Wang and Feingold, 2009a Buffering
Buffering: different paths to a specific end buffer the system against disrup ons to any par cular path
buffer A ractors
Stable states A and B are stable and self-sustaining A
Small perturba ons strengthen B the resilience of the state Lorenz, 1954
Stevens and Feingold, 2009 Precipitation and convergence patterns
Clean: 65 cm-3 Moderate: 150 cm-3 Polluted: 500 cm-3
Near-surface ver cal velocity Red: Updra s/surface convergence DYCOMS-II Model simula ons Blue: Downdra s/surface divergence Black contours: Drizzle Wang and Feingold (2009a) See also Willis and Deardorff 1979 Agee, 1984; Gemini V image Global Order from Local Interactions
Y-shaped surface convergence zone is region favoured for new convec on
Precipita on is ini ated
Downdra s, opening of cell
Surface divergence
Red: Updra s Blue: Downdra s/precipita on 200-m ver cal velocity from t = 6:15 to 9:15 Cells compete or cooperate while interac ng with their shared physical environment Open Cells: Surface updrafts
Red: Updra s Feingold, Koren, Wang, Xue, Brewer (2010) Blue: Downdra s/precipita on Y-shaped surface convergence zone is region favored for new convec on
Precipita on is ini ated
Downdra s, opening of cell
Surface divergence Y-convergence
Open Cells: Surface updrafts
Red: Updra s Feingold, Koren, Wang, Xue, Brewer (2010) Blue: Downdra s/precipita on Y-shaped surface convergence zone is region favored for new convec on
Precipita on is ini ated
Downdra s, opening of cell
Surface divergence Divergence
Synchronization: Oscilla ons in Precipita on
3 cases: DYCOMS ATEX VOCALS
Hovmuller diagram
Shi in rain “grid”
Colored contours: rain Contours: updra
Feingold, Koren, Wang, Xue, Brewer (2010) Synchronization
Feingold, Koren, Wang, Xue, Brewer (2010) How important is the Aerosol? Stability of the open-cell state: Role of Aerosol
VOCALS October 28th Natural aerosol sources: - Nuclea on of new par cles - Surface (wind-driven) produc on
-2
No aerosol source LWP, g m runaway (un-buffered) reduc on in drop concentra on à 06 12 18 00 06 00 Collapse of cloud Local me, h
Wang et al. 2010 ACP
See also Ackerman et al. (1993) Kazil et al. ACPD (2011) New Particle Formation in the Clean Marine Boundary Layer
Accumula on mode number conc m µ Pockets of open cells: Scavenging of accumula on mode;
New par cle forma on Par cle diameter, (from gasphase) Day in November 2003 Tomlinson et al. 2007 Influence of Giant CCN on precipitation in Stratocumulus
Liquid water path
(1/litre)
Precipita on, mm
- Significant increase in precip due to ~1/litre Giant CCN - More par cles does not always mean smaller drops
Feingold et al. 1999 Influence of Giant Nuclei on µphysics and Cloud Albedo
Drop size No giant CCN
Drop Number
No giant CCN
With Giant CCN
Cloud Albedo Significant reduc on in cloud albedo due to ~1/litre Giant CCN Feingold et al. 1999 Strong aerosol source: shiptrack
65 cm-3 150 cm-3
Contours: rain Shading: ship par cles - Mesoscale circula on transverse to track - Strengthens LWP in track - Clearing on either side of track Wang and Feingold, 2009b The Shiptrack Shadow
Credit : Jeff Schmaltz NASA/GSFC
ship track Unpredictable!
62 Clearing on either side of shiptracks
Ship track
shiptrack
Wang and Feingold, 2009b
W. Porch et al. Apollo-Soyuz, July 1975 Massive aerosol source: Self organising systems are resilient to change Massive aerosol perturba on Polluted case
Clean case with massive aerosol perturba on (65 à 300 cm-3)
-Thin “anvil cloud” lacks dynamical support A - Cells remain open B
- a certain amount of random perturba on will facilitate rather than hinder self-organiza on - possible implica ons for geoengineering
Stevens and Feingold 2009 Wang and Feingold, 2009b Main Themes/Conclusions
• Aerosol-Cloud Interac ons are seldom simple; – even shiptracks are complex
• Aerosol-Cloud-Precipita on System is o en self-organizing or buffered
– Iden fying mes when it is not would appear to be most frui ul – Aerosol influences may trigger changes in self-organiza on that are much more complex/interes ng than the original microphysical perturba on
• Geoengineering ship track experiments may yield unintended consequences – Ship track “shadows” due to mesoscale circula ons