Mixed-phase layer clouds
Chris Westbrook and Andrew Barrett
Thanks to Anthony Illingworth, Robin Hogan, Andrew
Heymsfield and all at the Chilbolton Observatory
What is a mixed-phase cloud?
• Cloud below 0°C where liquid water droplets and ice crystals
co-exist
• Interesting because Wegener, Bergeron & Findeison showed that
this situation is unstable
• Difference in vapour pressure between liquid water and ice
surfaces means droplets evaporate, and ice crystals grow at their
expense
• Apparent implication is that mixed-phase clouds should be rare
• Reality: mixed-phase clouds can persist for hours, even days
Observations
• Millimetre radar and near-infrared lidar together
• Powerful technique:
– liquid droplets very reflective to lidar
– ice particles dominate radar
ice crystal virga
ice crystals
cloud top ≈ -12°C
red stripe = base of liquid layer
Mechanisms for nucleating ice
deposition
immersion
condensation
contact
• Know that pure water droplets freeze spontaneously at ≈-37°C
• In warmer clouds, need an aerosol particle to form ice
cooling, time cooling, time
diffusion
Mechanisms for nucleating ice
deposition immersion condensation contact cooling, time cooling, time diffusion immediate freezingJUST NEED RH
iceTO BE HIGH ENOUGH
THESE ONES NEED A LIQUID WATER CLOUD
• Know that pure water droplets freeze spontaneously at ≈-37°C
• In warmer clouds, need an aerosol particle to form ice
So how important is liquid water for
forming ice in the atmosphere?
Use radar and lidar observations over 4 years to identify ice cloud layers – how many have liquid water at the top?
Find that in clouds > -22C almost all ice clouds have a liquid water top
Tells you droplets are needed to nucleate ice in these clouds – deposition nucleation is not important *because don’t see any ice clouds
without liquid in them]
Fraction of liquid-topped ice clouds drops off at colder temperatures:
50% at -27C
- increased deposition nucleation activity, or Bergeron-Findeison process?
0% at -37C [homogeneous freezing]
Implications:
• Cloud scheme should not form ice without liquid
water content for temperatures > -22°C
• Means simulation of supercooled liquid is important
if we want to successfully model the ice phase in
mid-level clouds
• Very frequent occurrence of liquid water at top of
mid-level clouds is important for radiation: optical
depth of liquid cloud is ≈10x that of ice cloud for
given water content
Focus of rest of talk
• Will concentrate on persistent thin mixed-phase
layer clouds such as Ac
• Typical structure: thin supercooled liquid layer at
top, ice crystals nucleated in layer, growing in it,
and falling below for ≈ 1km
• Relatively idealised setting to investigate basic
physics of how ice is formed
• Important for radiation
[see also Hogan et al 2003, QJRMS]
Key questions
• How much ice is nucleated in the liquid
water layers?
• How does that ice evolve and fall out?
• And how does the supercooled liquid
Microphysical structure
• Example from 18 May 2008: persistent Ac with virga
Liquid water dominates optical properties.
What did the model predict?
Colours are ice Contours are liquid
Cloud fraction much too low Cloud structure is wrong – no liquid at top
Microphysical structure
• Enormous reflection from ice crystals if lidar beam exactly vertical • Caused by mirror reflections from oriented plate-like ice crystals
Microphysical structure
Confirmed by polarisation radar pointing at 45°
- Horizontally polarised return is much stronger - Oriented pristine crystals dominate
General picture:
<ZDR> profiles from 6
persistent Ac clouds during May 2008 – same signature Images from 2DS cloud probe in flight through an Ac cloud, top -13C
Conclusion:
• Vapour growth dominates in these
clouds which are vapour-rich, but
which have low liquid water path
and are geometrically thin
• Typical habits are planar types
(plates, dendrites, stellars etc) –
reflects the typical temperature
range for persistent supercooled
layers ≈-10 to -20°C
Dynamics
Shallow mixed layer in top 500m
Radar spectral width – measure of turbulence
Blue values = still air
Dynamics
• PDF of vertical velocity of liquid
droplets at top (tracers for air
motion) from Doppler lidar
• Mean close to zero, negatively
skewed
• Narrow intense downdrafts
surrounded by broad weaker
updrafts
• Indicates overturning is driven
from top-down by radiative
cooling
Fluxes of ice, vapour and liquid
• Flight over Chilbolton 18 Feb 2009
• Very persistent layer of Ac, top -13C, lasted over
site for > 1 day
• Flight took place over 4 hour period in the
afternoon
Liquid water and θ
e
profiles
• Liquid water profile ~ adiabatic
• Droplet concentration constant 50/cc
• 25g/m
2liquid water path
equiv. potential temperature profile:
500m deep well mixed layer,
turbulence: well-mixed layer 500m deep at cloud top stable below
ice crystals falling 0.5-1m/s
Supercooled-top cloud, ice virga1km deep
Flux of ice crystals
12:20 UTC 15:30 UTC 15:20 UTC 11:20 UTC 12:40 UTCConclusion:
Flux ≈ 50/m
2/second
= 1.810
5/m
2/hour
… and it goes on for hours …
• Measure size spectra over 100km legs using CIP probe
• Calculate ∫n(D) v(D) dD
• Some uncertainty on v(D) so try different relationships and use spread as error bar
Ice nuclei budget
Flux of ice out layer = 50/m
2/second
According to De Mott et al (2010) concentration of ice nuclei at this temperature is ≈ 0.5/litre
Total depth of well-mixed layer is 500m, so at this rate available ice nuclei are completely depleted in about 1 hr.
But we sampled in situ for 4 hours, and flux at end was similar to that at the start… In fact, the radar observations show production of ice continuing for at least 24 hrs! Entrainment of fresh IN?
• to keep a steady supply of IN at this rate, would need entrainment velocity of 5cm/s (large) – no sign of cloud top rise
• RH is only 7% above cloud top. Mix in this air at 5cm/s and cloud quickly evaporates!
These long-lived supercooled clouds seem to be able to maintain a
steady production of ice over many hours: why aren’t IN depleted?
• Suggest a time-dependent freezing mechanism
•Most studies assume ice nuclei are sparse and efficient – once cooled to critical
temperature, freezing is immediate
• Suggest also have droplets which contain more inefficient ice nuclei, which freeze
randomly and slowly over time
• MODEL IMPLICATION: Could parameterise a steady flux of ice crystals as a
function of temperature?
•Met Office cloud scheme already sort-of simulates nucleation like this – automatic
replenishment of ice nuclei if all ice has fallen out
• CRMs with coupled cloud & aerosol– may be artificially simulating depletion of ice
nuclei in long lived mixed-phase clouds
Vapour / liquid water budget
Can estimate growth of ice at expense of liquid water:
dm/dt capacitance supersaturation
• know water saturated environment, temperature
• capacitance for planar habits is weak function of shape
≈ 0.3 maximum dimension, to within 20% or so (Westbrook et al 2008, JAS) • Now integrate over CIP size spectrum, and over supercooled layer depth:
• dLWP/dt ≈ 2g/m
2/hr
(vs 25g/m
2measured liquid water path)
• Complete glaciation takes 12 hours
• Bergeron-Findeison process in these clouds is slow
- could be
offset by a weak net radiative cooling of cloud layer (≈1K/day)
LW cooling
- destabilises top 500m of cloud - small net cooling to liquid layer
overturning
- well mixed layer
mixed-phase layer
500m
500m
dry air: evaporation of ice
Schematic diagram of mixed-phase Ac
stable layer
stable, potentially warm, often very dry air
ice virga
How can we successfully
simulated these clouds in a GCM?
• Investigate case studies using 1D model forced
by ERA interim
• Edwards & Slingo radiation
• Non-local mixing scheme
• MetO cloud microphysics, but can be altered
• Idea is to test sensitivities – what do we need to
do to get these clouds to come out right?
Effect of resolution on persistence of liquid water
Black contours are liquid water, colours are ice water content
• High res simulation: less ice near cloud top, persistent liquid layer
• Low res simulation: more ice water at top, cloud glaciates in ≈ 1.5 hrs
• Reason: gradient of IWC near top due to sedimentation is not resolved =
too much ice in grid box = big vapour flux from liquid to ice.
• Note +ve feedback: more liquid → more LW cooling → more liquid → ...
Need liquid to have a chance to get going for radiation to do its bit
How high does the vertical
resolution need to be?
• Only get close to convergence once Δz<100m
• Typical grid spacing in mid-troposphere: 400m ECMWF; 250m MetO UK4
Parameterise for low res GCMs
Parameterise ice, temperature & humidity profiles in 500m grid box Calculate mean grid box process rates from that parameterised profile Leads to more liquid, less ice growth
Effect of the size spectrum
• Growth rate integrated over size spectrum to get dIWC/dt (and hence - dLWC/dt) • Usually represented as an exponential in models
Concentration = N0 exp (-λD)
N0 diagnosed from temp
Con ce n tra tio n (lo g sca le) Particle size
Slope λ determined from IWC
Warmer temperatures N0 lower, meaning fewer, bigger particles for given IWC
In MetO model N0(T) is fixed function.
Have calculated N0 from large in-situ data set – see what happens when use these
Effect on ice growth and fall out
For low IWC clouds (eg Ac)
growth of ice is 2x bigger
than it ought to be
Effect on ice growth and fall out
For low IWC fall out rate is
half as fast as it should be
Net effect – ice sucks out
vapour too fast and sticks
Fix: fit N0 = f(T,IWC)
Long term evaluation of model: CloudNet
• Major failing is inability to fill the grid box – cloud
fraction is too low in mid-level clouds
• All model resolutions equally poor
• Problem with sub-grid humidity PDF?
• Not unique to Met Office!
34 34 Extending European Profiling Network 4 to 7 stations;
Launch of FP7 ‘Actris’ May 11: + 3 more sites in D.
Will soon be able to test model clouds over wide range of conditions
eg. supercooled Sc in Finland, very clean air at Mace Head in W. Ireland, etc.