Where satellite observations meet climate models (in the ...ensembles-eu.metoffice.com/cmug/impres6/Pincus-ESA-CCI-CMUG-20… · Where satellite observations meet climate models

Where satellite observations meet climate models(in the atmosphere)

Robert PincusUniversity of Colorado

What do satellite observations have to do with climate?

CCI and similar efforts stress measurements of

important physical quantities (ECV) that are

consistent over time (CDR)

The working assumption is that retrievals of physical quantities are more useful than raw measurements

For clouds and aerosols (and likely composition) this is certainly true.

How are these data being used, and what interesting opportunities are there?

0°N

20°S

40°S

60°S

80°S

20°N

40°N

60°N

80°N

0°N

20°S

40°S

60°S

80°S

20°N

40°N

60°N

80°N

0°E20°W40°W60°W80°W100°W120°W140°W160°W 20°E 40°E 60°E 80°E 100°E 120°E 140°E 160°E


0°N

20°S

40°S

60°S

80°S

20°N

40°N

60°N

80°N

0°N

20°S

40°S

60°S

80°S

20°N

40°N

60°N

80°N



Total Aerosol Optical Depth at 550 nmSunday 13 March 2016 00UTC CAMS Forecast t+036 VT: Monday 14 March 2016 12UTC

0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.5 0.8 1 3 50

Satellite observations and climate state estimation

Operational aerosol forecasts are now routine



Models used for state estimation are used in other contexts

-3.0 -2.0 -1.0 0.0Aerosol radiative forcing (Wm-2)

0

1

2

3Pr

obab

ility

RFari

RFaci

RFari+aci

Nicolas Bellouin, Reading; update to Bellouin et al. 2013, 10.5194/acp-13-2045-2013

3/10/16, 4:09 PMAMWG Diagnostics Plots

Page 1 of 1http://webext.cgd.ucar.edu/FAMIP/gpci_cam5.1_cosp_1d_001/atm/gpci_cam5.1_cosp_1d_001-obs/

AMWG Diagnostics Package

gpci_cam5.1_cosp_1d_001Plots Created

Tue Aug 5 12:01:48 MDT 2014

Set Description1 Tables of ANN, DJF, JJA, global and regional means and RMSE.2 Line plots of annual implied northward transports.3 Line plots of DJF, JJA and ANN zonal means4 Vertical contour plots of DJF, JJA and ANN zonal means4a Vertical (XZ) contour plots of DJF, JJA and ANN meridionalmeans5 Horizontal contour plots of DJF, JJA and ANN means 6 Horizontal vector plots of DJF, JJA and ANN means 7 Polar contour and vector plots of DJF, JJA and ANN means8 Annual cycle contour plots of zonal means 9 Horizontal contour plots of DJF-JJA differences 10 Annual cycle line plots of global means11 Pacific annual cycle, Scatter plot plots12 Vertical profile plots from 17 selected stations13 Cloud simulators plots 14 Taylor Diagram plots 15 Annual Cycle at Select Stations plots 16 Budget Terms at Select Stations plots

WACCM Set Description 1 Vertical contour plots of DJF, MAM, JJA, SON and ANN zonalmeans (vertical log scale)

Chemistry Set Description1 Tables / Chemistry of ANN global budgets 2 Vertical Contour Plots contour plots of DJF, MAM, JJA, SON andANN zonal means 3 Ozone Climatology Comparisons Profiles, Seasonal Cycle and TaylorDiagram 4 Column O3 and CO lon/lat Comparisons to satellite data5 Vertical Profile Profiles Comparisons to NOAA Aircraftobservations6 Vertical Profile Profiles Comparisons to Emmons Aircraftclimatology 7 Surface observation Scatter Plot Comparisons to IMROVE

TABLES METRICS

Click on Plot Type

Comparison to models including evaluation

Evaluation including “metrics” became common for CMIP3

3

3

9

9

15

15

21

27

27

Standard deviation (W/m2)

Stan

dard

dev

iatio

n (W

/m2 )

Correlation

.4

.6

.8

.9

.99

21

Net CRE

4

4

8

8

12

12

16

16

20

20


Stan

dard

dev

iatio

n (W

/m2 )

Correlation

.4

.6

.8

.9

.99

LW CRE

6

6

12

12

18

18

24

24

30

30

36

36


Stan

dard

dev

iatio

n (W

/m2 )

SW CRE

3

3

9

9

15

15

21

21

27

27

Standard deviation (%)

Stan

dard

dev

iatio

n (%

)

Cloud frac.

0.3

0.3

0.9

0.9

1.5

1.5

2.1

2.1

2.7

2.7

3.3

3.3

Standard deviation (mm/day)

Stan

dard

dev

iatio

n (m

m/d

ay)

Precip.

Correlation

.4

.6

.8

.9

.99

Correlation

.4

.6

.8

.9

Correlation

.4

.6

.8

.9

20th C

AMIP

C32R1

2nd obs

ERA-40

Super-CAM

IPCC mean model

Pincus et al. 2008, 10.1029/2007jd009334

Routine evaluation becomes routine…

GMDD8, 7541–7661, 2015

ESMValTool (v1.0)

V. Eyring et al.

Title Page

Abstract Introduction

Conclusions References

Tables Figures

J I

J I

Back Close

Full Screen / Esc

Printer-friendly Version

Interactive Discussion

Discussion

Pa

per

|D

iscussion

Pa

per

|D

iscussion

Pa

per

|D

iscussion

Pa

per

|

Geosci. Model Dev. Discuss., 8, 7541–7661, 2015

www.geosci-model-dev-discuss.net/8/7541/2015/

doi:10.5194/gmdd-8-7541-2015

© Author(s) 2015. CC Attribution 3.0 License.

This discussion paper is/has been under review for the journal Geoscientific Model

Development (GMD). Please refer to the corresponding final paper in GMD if available.

ESMValTool (v1.0) – a communitydiagnostic and performance metrics toolfor routine evaluation of Earth SystemModels in CMIPV. Eyring1, M. Righi1, M. Evaldsson2, A. Lauer1, S. Wenzel1, C. Jones3,4,A. Anav5, O. Andrews6, I. Cionni7, E. L. Davin8, C. Deser9, C. Ehbrecht10,P. Friedlingstein5, P. Gleckler11, K.-D. Gottschaldt1, S. Hagemann12, M. Juckes13,S. Kindermann10, J. Krasting14, D. Kunert1, R. Levine4, A. Loew15,12, J. Mäkelä16,G. Martin4, E. Mason14,17, A. Phillips9, S. Read18, C. Rio19, R. Roehrig20,D. Senftleben1, A. Sterl21, L. H. van Ulft21, J. Walton4, S. Wang2, andK. D. Williams4

1Deutsches Zentrum für Luft- und Raumfahrt (DLR), Institut für Physik der Atmosphäre,Oberpfa◆enhofen, Germany2Swedish Meteorological and Hydrological Institute (SMHI), 60176 Norrköping, Sweden3University of Leeds, Leeds, UK4Met Oce Hadley Centre, Exeter, UK5University of Exeter, Exeter, UK

7541

… but can be misleading

“Here the progress that has been made in recent years is measured by comparing .. cloud properties [cloud amount, liquid water path, and cloud radiative forcing] … from the CMIP5 models with satellite observations and with results from comparable CMIP3 experiments. …the differences in the simulated cloud climatology from CMIP3 and CMIP5 are generally small, and there is very little to no improvement apparent in the tropical and subtropical regions in CMIP5.”

Lauer and Hamilton 2013, 10.1175/JCLI-D-12-00451.1

“… based on these biases in the annual mean, Taylor diagram metrics, and RMSE, there is virtually no progress in the simulation fidelity of [outgoing TOA radiation and surface solar] fluxes from CMIP3 to CMIP5. ..We hypothesize that at least a part of these persistent biases stem from the common global climate model practice of ignoring the effects of precipitating and/or convective core ice and liquid in their radiation calculations.”

Li et al. 2013, 10.1002/jgrd.50378

Klein et al. 2013, 10.1002/jgrd.50141

0 0.5 1 1.5 2

MODIS

i

QR

B

c4

C4

g2

G3N4

n3

N5P

h4h1

h3

H2

m4m3

M5

1

2

Total Cloudiness

Better Worse

i

QR

B

g2

G3N4

n3

N5

c4

C4P

h4h1h3

H2

m4 m3

M5

1

2

CFMIP1

CFMIP2

P(τ, pc)

i

QRB

g2

G3N4

n3

N5C4P

m4m3

M5

h4h1 h3

H2

1

2

SW CRE(τ, pc)

i

QRP

g2

G3 B N4

n3

N5 C4

h4h1 h3

H2

m4m3

M5

1

2

LW CRE(τ, pc)

Two big changes in the last decade





Climate Model Clouds Pseudo-Satellite Observations

COSP Processing

Climate Model Clouds Pseudo-Satellite Observations

COSP Processing

Observational proxies(i) — matching scales

0 0.2 0.4 0 0.05 0.15 0 2 4 x 10-3

Cloud fraction Liquid water Ice water

400

1000

800

600

Pre

ssur

e

10-6 10-4 10-2 10

1000

800

600

Ice water content

Liquid water content

g/m2

Pincus et al. 2006, 10.1175/MWR3257.1

Observational proxies(i) — matching scales

0 0.2 0.4 0 0.05 0.15 0 2 4 x 10-3

Cloud fraction Liquid water Ice water

400

1000

800

600

Pre

ssur

e

10-6 10-4 10-2 10

1000

800

600

Ice water content

Liquid water content

g/m2

Pincus et al. 2006, 10.1175/MWR3257.1

Simulators map the model description of clouds

into synthetic pixel-scale observations using rough approximations

and aggregate these in space and time as per the observational data sets

P =

Z ⌧=1

TOA

P (z)�c(z)dz

Observational proxies(ii) — a satellite’s-eye view

re(l,i)(z), ⌧(l,i)(z) or q(l,i)(z)

re = F�1(F (re(z)))⌧ =

Z sfc

TOA

�c(z)dz

pc =

Z �=1

TOA

p(z)�c(z)dz

Diagnostics from the CFMIP Observation Simulator Package were requested for CFMIP2/CMIP5 and have been revised for CFMIP3/CMIP6.

COSP facilitates the mapping of model state information to observations from passive (MISR, MODIS, ISCCP) and active (CloudSat, CALIPSO) platforms

Observations are produced for each data stream

Can be extended by adding new sensors (e.g. CLARA), analyses…

Most climate models have observation proxies for clouds

AFFILIATIONS: BODAS-SALCEDO, WEBB, AND JOHN—Met Office Hadley Centre, Exeter, United Kingdom; BONY, CHEPFER, AND DUFRESNE—Laboratoire de Météorologie Dynamique/L’Institut Pierre-Simon Laplace, Centre National de la Recherche Scientifique, Université Pierre et Marie Curie, Paris, France; KLEIN AND ZHANG—Program For Climate Model Diagnosis and Intercomparison, Lawrence Livermore National Laboratory, Livermore, California; MARCHAND— Joint Institute for the Study of the Atmosphere and Ocean, University of Washington, Seattle, Washington; HAYNES—School of Mathematical Sciences, Monash University, Clayton, Victoria, Australia; PINCUS—University of Colorado and NOAA/Earth System Research Laboratory, Boulder, ColoradoCORRESPONDING AUTHOR: Dr. A. Bodas-Salcedo, Met Office Hadley Centre, FitzRoy Road, Exeter EX1 3PB United KingdomE-mail: [email protected]

The abstract for this article can be found in this issue, following the table of contents.DOI:10.1175/2011BAMS2856.1

In final form 8 April 2011©2011 American Meteorological Society

By simulating the observations of multiple satellite instruments, COSP enables quantitative evaluation of clouds, humidity, and precipitation processes in diverse numerical models.

G eneral circulation models (GCMs) of the atmosphere, including those used for numerical weather prediction (NWP) and climate projec-

tions, operate with resolutions from a few kilometers to hundreds of kilometers. Many atmospheric pro-cesses, such as turbulence and microphysical process-es within clouds, operate at smaller scales and hence

cannot be resolved by current model resolutions. These processes are included by means of parameter-izations, which are semiempirical or statistical models that relate gridbox mean variables to these subgrid processes. For instance, some cloud parameterizations diagnose the amount of cloud condensate and the fraction of the grid box that a cloud occupies (cloud area fraction) as a function of the relative humidity (RH) of the grid box (Slingo 1980; Smith 1990). The formulation of these parameterizations is very im-portant for the model evolution because they modify the three-dimensional structure of temperature and humidity directly (e.g., condensation/evaporation) or indirectly by interacting with other parameteriza-tions (e.g., radiation) and the large-scale dynamics. Therefore, the evaluation of these parameterizations is crucial to improving our weather forecasts or in-creasing our confidence in climate projections.

Satellites have proven to be very helpful tools for this purpose because they provide global or near-global coverage, thereby giving a representative sample of all meteorological conditions. However, satellites do not measure directly those geophysical quantities of interest, such as the amount or phase of cloud condensate. They measure the intensity of radiation coming from a particular area and direc-tion in a particular wavelength range (radiances). The range of wavelengths covered by past and cur-rent systems spans several orders of magnitude, from

COSPSatellite simulation software for model assessment

BY A. BODAS-SALCEDO, M. J. WEBB, S. BONY, H. CHEPFER, J.-L. DUFRESNE, S. A. KLEIN, Y. ZHANG, R. MARCHAND, J. M. HAYNES, R. PINCUS, AND V. O. JOHN

1023AUGUST 2011AMERICAN METEOROLOGICAL SOCIETY |

Bodas-Salcedo et al. 2011, 10.1175/2011BAMS2856.1

Using proxies to pick apart correlations between aerosols and clouds

−0.4

0.0

0.4Δln(N )/Δln(τa)

∂ln(N )/∂ln(τa)

MODIS AM3 CAM5 ModelE2

Sensitiv

ity

after Ban-Weiss et al. 2014, 10.1002/2014JD021722

But there’s a lot the proxies can’t do…

We understand the sensitivity of our instruments

See, for example: GEWEX cloud assessment (10.1175/BAMS-D-12-00117.1)

Every observation has a model attached to it.

Our models for interpreting reflectance measurements use

simple forward models (e.g. one-dimensional radiative transfer) operating on

highly parameterized representations of clouds

A simple question. How much of the planet is cloudy?

ISCCP: 66% MODIS mask: 67%

8010 Cloud fraction (%)

Pincus et al. 2012, 10.1175/JCLI-D-11-00267.1

A simple question. How much of the planet is cloudy?

MODIS retrievals: 50%

ISCCP: 66% MODIS mask: 67%

8010 Cloud fraction (%)


minimum optical depth

clo

ud fraction (

%)

0.3 1.3 3.6 9.4 23 60

2.7

49.9

65.8

MODIS

ISCCP


On the limits of instrument simulators (i): partly-cloudy pixels

The largest differences in estimates of cloud fraction between MODIS and other data streams stems from the treatment of partly-cloudy pixels

Most (~50-85%) optically thin pixels are in fact partly-cloudy

This sensitivity can not be represented in observation proxies because they don’t produce cloudy pixels

But there are sensitivities we are only beginning to understand

Zhang and Platnick (2011), doi:10.1029/2011JD016216

On the limits of instrument simulators (ii): spectral dependence of re


Hints from observations

(optical thickness retrieved at different angles were rarely consistent;Liang et al. 2009, doi:10.1029/2008GL037124)




inspired modeling

(large-eddy simulation clouds, three-dimensional radiative transfer; Zhang et al 2012; 10.1029/2012JD017655)




inspired modeling


that led to understanding:

even fully cloudy pixels can be inhomogeneousreflection is reduced in such pixels by an amount depending on wavelength reduced reflection looks like absorption i.e. larger cloud drops




inspired modeling




i.e that drop size retrievals in inhomogeneous (i.e. most) pixels are based high




inspired modeling




i.e that drop size retrievals in inhomogeneous (i.e. most) pixels are based high

Like partly cloudy pixels, this isn’t treated in observation proxies, making comparisons of modeled and observed size uninformative

pers. comm., Frank Evans, University of Colorado

0.2

0.4

0.6

0.8

1.0

1.2

RM

S e

rro in ln(d

epth

)

..

..

. . ..

.

..

.

..

.. .

.

.

. .

..

.. . .

. ..

.

. .

.. .

.

500 m

1000 m

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Cloud Fraction

0

2

4

6

8

10

12

14

RM

S e

rror in e

ffective r

adiu

s (

(µ)

.

. .

.

. .

..

. ..

.. .

. . ..

.

..

. ..

.

. .

. ..

.. .

. . . .

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

.. . .

..

.

..

. . . . . ..

.

.

.. . .

..

..

. . . . . . ..

.

0.5 1.0 2 5 10 20 50 100

True Optical Depth

0

2

4

6

8

10

12 . .

.

.

.

..

..

..

.. . . .

. . ..

.

.

.

.

.. . .

..

. . . .

Being careful what we wish for

Making relevant data more useful is a good thing

Finding common ground between retrievals and models is informing modeling

But too great an emphasis on success as “use by climate modelers” can deemphasize other valuable uses…

… and implies certainty in our data sets that we know isn’t always warranted

Being careful what we do and say

Better than anyone the remote sensing community understands

the limits of the models we use andhow those limits impact our retrievals

We might be better served by devoting less energy to “products” and more to answering specific questions in context

Documents

Where satellite observations meet climate models (in the ...ensembles-eu.metoffice.com/cmug/impres6/Pincus-ESA-CCI-CMUG-20… · Where satellite observations meet climate models