Influences of ice particle model on ice cloud optical thickness retrieval

Influences of ice particle model on ice cloud optical

thickness retrievalZhibo (zippo) Zhang

03/29/2010ESSIC

Outline

• Background• Importance of ice cloud• Ice particle model and ice cloud retrieval

• Influence of ice particle model on t retrieval• Comparison of MODIS and POLDER ice t

retrieval• Influence on our understanding of ice cloud

seasonal variability

• Summary

Ice cloud: fun

Photo from Wiki

Ice cloud: important

Ice clouds are important, because• Cover large portion of the Earth’s surface • Radiative effects• Water vapor budget • Cloud feedbacks

ISCCP day-time ice cloud amount

Earth

Albedo Effect

GreenhouseEffect (dominant)

Ice cloud: not well understood

Duane Waliser et al. 2009 JGR

Satellite-base remote sensing of ice cloud properties

Satellite remote sensing

In-situ measurements Scattering model

microphysics

GCMs

Ice Particle Model

Ice particle model

• Size distribution • Shape distribution• Orientation• Inhomogeneity & surface roughness

Ice Particle model

Ice particle size

• Size matters• Cloud life time (e.g., Heymsfield 1972, Jensen et

al.1996)• Cloud reflectance, radiative forcing,

heating/cooling rate (e.g., Ackerman et al. 1988; Jensen et al. 1994 )

• Cloud feedback (e.g., Stephens et al. 1990)Hard to measureShattering of large particlesGardiner and Hallett 1985; Gayet et al. 1996Field et al. 2003;

Earth Observing Laboratory NCAR

Num

ber d

ensit

y

Particle Size50 µm µm mm

Ice particle shape

• Why shape also matters?

wavelength

wavelength

Aerosol

Ice particle

From Bryan BaumComplicacy of ice particle shape must be acceptable by scattering models

Capabilities of current scattering models

Ice particle orientation

Horizontally orientated

Randomly orientated

Images from www.atoptics.co.uk

Ice particle orientationHorizontally orientated

Image credit: CNES

Inhomogeneity and surface roughness

Yang et al. 2008 JAMC Yang et al. 2008 ITGRS

Ice particle model

• Size distribution • Shape distribution• Orientation• Inhomogeneity & surface roughness

Ice Particle model

So many things to consider…not surprising that ice particle models are usually different from one another

Ice particle models: MODIS C5

Baum et al. 2005 JAMC• More than 1000 PSDs• Complicate habit/shape distribution• Random orientation• Homogeneous and smooth

Ice particle model: MODIS C5

Baum et al. 2005 JAMC

Baum et al. 2005 JAMC

IWC from MODIS C5 ice particle mode is consistent with in situ measurement

Ice particle model:POLDER

Inhomogeneous Hexagonal Monocrystal

Courtesy ofJerome Riedi

• Constant size (30µm)• One habit only• Random orientation• Internal inclusion of air bubbles

Scattering signature consistent with POLDER observationC.-Labonnote et al. 2000 GRL

Scattering phase function Baum05 VS IHM

Comparison of MODIS and POLDER ice cloud retrieval

Motivation

• How are MODIS and POLDER ice cloud retrievals different?

• What is the role of ice particle model?

• Any implications for climate studies?

• Is it possible to build up a long-term ice cloud property dataset from multiple missions?

Zhang, Z.et al. 2009: Atmos. Chem. Phys., 9, 1-15. (www.atmos-chem-phys.net/9/1/2009/)

MODIS POLDER

Resolution 1km 20kmCloud effective radius Retrieved AssumedIce particle model Baum05 IHMDirectionality Single Up to 16

Case for comparisonAqua-MODIS granule on July 22, 2007 (UTC 18:45)

NASA Langley TC4 team

Flight track

GOES IR image

Flight track of TC4 mission

CollocationCollocation of Level-1 radiance data

Collocation of Level-2 cloud products

6km6k

m

POLDER full resolution pixel

MODIS 1km pixel

POLDER 20km downscale to 6kmMODIS 1km aggregated to 6km

6km

6km

POLDER full resolution pixel

MODIS t vs POLDER t

tPOLDER/tMODIS follows the log- normal distribution

tPOLDER is substantially smaller than tMODIS

For more than 80% pixels tPOLDER < tMODIS

For more than 50% pixelstPOLDER < tMODIS by more than

30%

Same clouds; different t?Why?

Main reason for the difference

• Difference in resolution (Plane parallel albedo bias) ✗

• Difference in effective radius treatment ✗• Difference in ice particle model✔

R ~ (1−g)τ τ retrieval ~ Robs / (1 − g)

τ POLDER

τ MODIS~

1 − gBaum05

1 − g IHM= 0.7126 (0.6827)(From data: 0.68)

Implications for ice SW CRF Zonal mean ice optical thickness vs month (2006)

Implications for ice SW CRF

Instantaneous Shortwave CRF (FSW)

Implications for ice SW CRF

Wrong ice particle model

Wrong t retrieval

Wrong g used

“Not so wrong” FSWFSW ~ R ~ (1−g)τ

retrieval

FSW computation

Error cancellation

Ice particle model and seasonal variation of t retrieval

τ cIHM

τ cBaum05

IHM model is used for MODIS retrievalBaum05 model is used for MODIS retrieval

Difference in higher-order moment of P11

Difference in g

Angular signature of ice cloud reflectance

Angular signature is mainly determined by single-scattering

SatelliteSingle-scatteringMultiple-scattering

MODIS angular sampling

winter θ0

θs

θ0

θs

summer

MODIS angular sampling vs season

winter

summer

Impact on seasonal variation of t retrieval

winter

summer

Assume IHM to be the truth

Summary• The t of ice clouds retrieved from POLDER is substantially

smaller than that from MODIS retrieval. • This difference is mostly attributed to the difference in ice

bulk scattering models used in MODIS and POLDER retrievals• If a wrong bulk scattering model is used in the retrieval

algorithm, the error in g factor may lead to overestimation or underestimation of t . However, this error in t retrieval is largely cancelled in FSW computation by the error in g factor.

• The error in higher-order moment of P11 may lead to artificial seasonal variation of t and this error can NOT be cancelled in FSW computation

Questions?