Clouds of different optical depth t should be treated differently

RadOn : Retrieval of microphysical and radiative properties of ice clouds from Doppler cloud radar observations

J. Delanoë and A. Protat IPSL / CETP

Clouds of different optical depth should be treated differently

Lidar-Radiometer for very thin clouds (not detected by radar) Radar-Lidar / Radar-Radiometer for < 3 Radar / Radar-Radiometer / Dual-Wavelength Radar for> 3

2

Illustration for the need of different methodsIllustration for the need of different methods

Radar Z

Lidar RadarLidar

Radar+Lidar

Existing radar methods : Matrosov Z+V (2002) and Hogan IWC-Z-T (2005)

3

The two measurements of a Doppler cloud radarThe two measurements of a Doppler cloud radar

Reflectivity factor

Doppler velocity

For a vertically-pointing cloud radar :

These measurements are related to N(D)

418

510 ( )

2e bscZ N D dDKw

( )

( )d t

v N D dDbscV V w wN D dDbsc

4

The ice cloud propertiesThe ice cloud properties

32.10 ( ) ( )N D A D dD

4

3

( )

( )m

N D D dDD

N D D dD

3( ) ( )6wIWC D N D D dD

re IWC /

The ice cloud properties are also related to N(D)

5

The normalized particle size distributionThe normalized particle size distribution

High variability in ice clouds

Scaling the PSD so that it does not depend on IWC, Dm

N(D) = No* F (Deq/Dm)

Delanoë et al. (JGR, 2005) : shape F can be approximated by a single analytical form for all ice clouds (<10% error)

The unknowns to get cloud properties : No* and Dm

Then Re = ( (7) / 2 (6) ) Dm IWC = No* w Dm

4 / 44

= (3/2) IWC / Re =dz

The idea in RadOn and Matrosov (2002) is to get it

from the two radar measurements Z and VD

Mean spectra for all experiments

CLARE 98, CARL 99, EUCREX, ARM IOP, FASTEX, CEPEX, CRYSTALFACE

The normalized particle size distributionThe normalized particle size distribution

Analytical formulation

IWC error=f(T)

5% 10% 1 dB

Z error=f(T)

0.5 dB

Z Doppler velocityVD=VT+w

Most representative density-diameter and area-diameter relationships

Dm (VT, (D), A(D))

N0* =f(Dm,Z)

IWC, , re

VT –Z statistical relationships or mean VD : VT retrieval

Principle of the radar retrieval methodPrinciple of the radar retrieval method

8

Principle of the radar retrieval method Principle of the radar retrieval method

First step : Retrieval of VT from (VD , Z)

Hypothesis : for a long enough time span <w> << <VT>Error = synoptic ascent / descent (typically 5 cms-1)

A VT - Z relationship is derived for each cloud

V D w V T

scatter = w contribution

9

Principle of the radar retrieval method Principle of the radar retrieval method First step : Retrieval of VT from (VD , Z)Alternative approach : 20-minutes means (Matrosov 2002)Improvement : Running means over 20 minutes (resolution)

V D w V T

IWCRW

IWCVTZ

VTZ

RW

10

Z Doppler velocityVD=VT+w

VT –Z statistical relationships or mean VD : VT retrieval

Dm (VT, (D), A(D))

N0* =f(Dm,Z)

IWC, , re

Most representative density-diameter and area-diameter relationships

Principle of the radar retrieval methodPrinciple of the radar retrieval method

Principle of the radar retrieval methodPrinciple of the radar retrieval method

Second step : Retrieval of most representative (D),A(D) relationships

Using the micro in-situ database and theoretical v(D) = f((D),A(D))for different ice particle shapes and habits

we have computed synthetic VT-Z relationships

For each cloud, we compare the synthetic and radar-derived VT-Z relationships the set of (D),A(D) relationships

that minimises the difference is retained

12

Z Doppler velocityVD=VT+w

Most representative density-diameter and area-diameter relationships

N0* =f(Dm,Z)

IWC, , re

Principle of the radar retrieval methodPrinciple of the radar retrieval method

Dm (VT, (D), A(D))

VT –Z statistical relationships or mean VD : VT retrieval

Third step : Dm retrieval from VT , (D), A(D)

Knowing (D) and A(D) and using an analytical form for the

normalised PSD shape F, there is a direct relation between VT and Dm

( / )( )

( / )iBbsc m

t m i mibsc m

v F D D dDV D AD

F D D dD

Vt=f(Dm) Dm=f(Vt)

- Vt radar

Principle of the radar retrieval methodPrinciple of the radar retrieval method

14

Z Doppler velocityVD=VT+w

Most representative density-diameter and area-diameter relationships

IWC, , re

Principle of the radar retrieval methodPrinciple of the radar retrieval method

Dm (VT, (D), A(D))

VT –Z statistical relationships or mean VD : VT retrieval

N0* =f(Dm,Z)

15

Fourth step : N0* retrieval from Dm and Z

In Mie regime there is a direct expression that relates N0*, Dm and Z

Principle of the radar retrieval methodPrinciple of the radar retrieval method

25 18 1

*0 4

10( / )e m bsc

wN Z F D D dD

K

Database: CLARE 98, CARL 99, EUCREX, ARM IOP, FASTEX, CEPEX, CRYSTAL-FACE

We use (D)=0.00556(D in cm)-1.1, A(D)=/4D², radar at 95GHz.

RadOnIWC-Z-T

(2005)

Matrosov

(2002)

Biais % % Biais % % Biais % %

IWC -0.2 17.2 9 60 25 75

-3.7 19 102 - -

re 5.2 10.5 - - - -

Compute Vt, Z, IWC, and re from the in-situ data, with A(D) and (D) constant

RadOnHogan IWC-Z-TMatrosov Z+V

IWC, , re retrieved

Vt + Z micro

IWC, , re micro

Similar study for other A(D) / (D) Error estimates are comparable

Evaluation of RadOn using the Evaluation of RadOn using the in-situ database in-situ database

Global error analysis

IWC 95GHz

A=0,5D2 A=0,2D1,8 A=0,05D1,4A=π/4D2 A=0,2D2-40

-30

-20

-10

0

10

20

30

lois d'aire-diamètre

% -1,4-1,1-0,8-0,5

17

Evaluation of RadOn using the IPSL Ra-Li methodEvaluation of RadOn using the IPSL Ra-Li method27 Chilbolton clouds selected for intercomparisons

9 cases : good 6 cases : bias 7 cases : Mie effect5 cases : bias + Mie effect

These differences in performance are due Mie scattering not in Ra-Li method.The 9 good cases : RadOn density retrieval close to Ra-Li (Brown-Francis 1995).

IWCIWC IWC

We restrict to the 9 good cases

IWC

18

Evaluation of RadOn using optical depth from lidarEvaluation of RadOn using optical depth from lidarOptical depth from lidar can be obtained from difference in molecular return

Comparisons with RadOn optical depths and those lidar cases

Limitations : Can only be done when radar and lidar thicknesses comparable + lidar traverses entirely + no occurrence of SLW case study approach only

OK

OK

Thin SLW layer

Overall : when good conditions errors < 0.1, fractional error +15% / -25%

This method works for these radar frequencies : 3, 10, 35, 95 GHzYields very encouraging results :

-17%<errIWC<+17%,

-22.5%<err<+15%,

-5.5%<errre<+15.5%

During CloudNet this method allowed (see talk this afternoon):

Statistics of (D) / A(D) from CloudNet radars Climatology of European ice cloud properties Evaluation of the representation of clouds

in the CloudNet NWP models

Available to all ground-based remote sensing sites (Matlab code)

Conclusions and perspectivesConclusions and perspectives

20

3 march 2003: prefrontal cloud

Vt=96.3Z0.177857

A(D)=0.2D1.6

(D)=0.0156D-1

Aggregates

Z

Vd

21

14 april 2003: Thick ice cloud

Vt=66.44Z0.189598

A(D)=0.5D1.8

(D)=0.0132D-0.9

Aggregates

22

15 april 2003: thin cirrus

Vt=57.954Z0.184944

A(D)=/4D1.8

(D)=0.0318D-0.8

Dl=170µm, up to this diameter solid ice