cloud fraction: can it be defined and measured? …cloud fraction, anyway? clouds have a strong...
TRANSCRIPT
CLOUD FRACTION: CAN IT BE DEFINEDAND MEASURED? AND IF WE KNEW IT
WOULD IT BE OF ANY USE TO US?Stephen E. Schwartz
Upton NY USA
www.bnl.gov/envsci/schwartz
Cloud Properties, Observations, and their Uncertainties
CLOUD FRACTION: CAN IT BE DEFINED AND MEASURED? AND IF WE KNEW IT
WOULD IT BE OF ANY USE TO US?
CONCLUSIONSNo.No.No.
I come to bury cloud fraction, not to praise it.- Shakespeare, 1599
WHAT IS A CLOUD?AMS Glossary of Meteorology (2000)
A visible aggregate of minute water droplets and/or ice particles in the atmosphere above the earth’s surface.Total cloud cover: Fraction of the sky hidden by all visible clouds.
Clothiaux, Barker, & Korolev (2005)Surprisingly, and in spite of the fact that we deal with clouds on a daily basis, to date there is no universal definition of a cloud. . . . Ultimately, the definition of a cloud depends on the threshold sensitivity of the instruments used.
Ramanathan, JGR (ERBE, 1988)Cloud cover is a loosely defined term.
Potter Stewart (U.S. Supreme Court, 1964)I shall not today attempt further to define it, but I know it when I see it.
WHY DO WE WANT TO KNOW CLOUD FRACTION, ANYWAY?
Clouds have a strong impact on Earth’s radiation budget: -45 W m-2 shortwave; +30 W m-2 longwave.
Slight change in cloud fraction could augment or offset greenhouse gas induced warming – cloud feedbacks.
Getting cloud fraction “right” is an evaluation criterion for global climate models.
Warren, Hahn, London, Chervin, Jenne
Domain ObservationsMillions
Cloud cover%
Land 116 52.4Ocean 43.3 64.8Global 159 61.2
CLOUD FRACTION BY MULTIPLE METHODS 2 Surface, 3 satellite methods at U.S. Southern Great Plains; 10 years data
0.30.40.50.60.7
1.0˚ PATMOS!X0 6 12 18 24
ARSCL (Lidar, Radar)TSI (Total Sky Imager)
0.5˚ GOES1.0˚ PATMOS!X
2.5˚ ISCCP D1
2 4 6 8 10 12
98 00 02 04 06 08 10Year
All!S
ky C
loud F
racti
on
Local Hour Month
0.30.40.50.60.7
Wu, Liu, Jensen, Toto, Foster & Long, JGR, in review
Different methods yield substantial systematic differences in the mean. Error of 0.1 in cloud fraction is ~ 9 W m-2 in shortwave, 6 W m-2 in
longwave.
MULTIPLE APPROACHES TO DETERMINING CLOUD FRACTION
TSI (FOV 100°)
Ground
Satellite
Modified from Wu, Liu, Jensen, Toto, Foster & Long, JGR, in review
Domain 55-140 kmon a side
Total Sky Imager
Although different approaches yield different instantaneous, local CF, they would be expected to yield the same average CF.
MULTIPLE APPROACHES TO DETERMINING CLOUD FRACTION
TSI (FOV 100°)
Cloud Radars (FOV 0.2°; 3.5 mrad)
Ground
Satellite
Modified from Wu, Liu, Jensen, Toto, Foster & Long, JGR, in review
Cloud Lidars (FOV 80 µrad)
Domain 55-140 kmon a side
ARSCLActive Remote Sensing of CLouds
Total Sky Imager
Although different approaches yield different instantaneous, local CF, they would be expected to yield the same average CF.
REASONS FOR DIFFERENCES IN MEASURED CLOUD FRACTION
Trivial
Mismatch of spatial and/or temporal domain.
View angle – sidewall effect – cloud aspect ratio.
Intrinsic
Spatial resolution. 6% 25% 100%
Threshold.
1
0Clou
d Fr
actio
n
0.450.400.350.300.25Threshold
COMPARISON OF CLOUD FRACTION BY DIFFERENT METHODS
Hourly cloud fraction at SGP by multiple methods, May, 2009 ARSCL 1.0
0.8
0.6
0.4
0.2
0.01.00.80.60.40.20.0
1.0
0.8
0.6
0.4
0.2
0.01.00.80.60.40.20.0
TSI
GOES
ActiveRemoteSensing
by Radar,Lidar
TotalSky
Imager
Satellite, 0.5˚
1:1 line
Comparison plots show some skill but substantial differences.
CORRELATION IS DOMINATED BY ONES AND ZEROES
Hourly cloud fraction at SGP by ARSCL AND GOES, May, 2009
1.0
0.8
0.6
0.4
0.2
0.01.00.80.60.40.20.0
CF GOES
1.0
0.8
0.6
0.4
0.2
0.01.00.80.60.40.20.0
CF GOES
All points, May, 2009 Points within 5% of 0 or 1 in both data sets excluded
r2 = 0.78 r2 = 0.44
CF
ARSC
L
CF
ARSC
LLeast squares fit
Excluding all-cloud and no-cloud scenes reduces variance accounted for by the regression from 78% to 44%.
n = 1346 n = 620
0
Cou
nts
10 CF GOES
0
Cou
nts
10ARSCL CF
TIME SERIES OF CLOUD FRACTION BY MULTIPLE METHODS
ARM SGP site (north central OK) May 13, 2009
Date, May, 2009, UTC
1
0.5
0
Clo
ud F
ract
ion
12005/13/09
00005/14/09
ARSCL
TSI GOES
Substantial variation among methods.
TIME SERIES OF CLOUD FRACTION BY MULTIPLE METHODS
ARM SGP site (north central OK) May 13, 2009
Date, May, 2009, UTC
1
0.5
0
Clo
ud F
ract
ion
12005/13/09
00005/14/09
ARSCL
TSI GOES
TSI 30 s
141600 141630 141700
Substantial variation among methods.Substantial fluctuation in TSI images taken at 30-second intervals.
TOTAL SKY IMAGES AND CLOUD MASKS FROM TSI ALGORITHM
ARM SGP site (north central OK) May 13, 2009, 1416-1417141600 141630 141700
141600 141630 141700
TSI threshold misses thin visible cloudsSubstantial changes at 30-s intervals as clouds are blown by wind.
CLOUD RADIATIVE EFFECTDependence on shortwave optical depth and cloud-top temperature
24-Hour average CRE, north central Oklahoma, at equinox
-200
-100
0
100
0.01 0.1 1 10 100Cloud shortwave optical depth
Shortwave
-200
-100
0
100
200
0 02 04 06 08 001-10
-5
0
5
10
m W ,A
OT ta tceffE evitaidaR duol
C-2
20.0 40.0 60.0 80.0 01.00-60
-40
-20
0
20
40
60
0.0 2.0 4.0 6.0 8.0 0.1-200
-100
0
100
200
0 2 4 6 8 01
Cloud shortwave optical depthCRE is initially linear in optical depth, saturating at high optical depth.
CLOUD RADIATIVE EFFECTDependence on shortwave optical depth and cloud-top temperature
24-Hour average CRE, north central Oklahoma, at equinox
-200
-100
0
100
0.01 0.1 1 10 100Cloud shortwave optical depth
Shortwave
-200
-100
0
100
200
100806040200-10
-5
0
5
10
m W ,A
OT ta tceffE evitaidaR duol
C-2
0.100.080.060.040.020-60
-40
-20
0
20
40
60
1.00.80.60.40.20.0-200
-100
0
100
200
1086420
Cloud shortwave optical depth
Disk of sun no longer visible100 g m ⇔-3 100 μm LWP ⇔ 5 12-μm radius drops
Thin clouds
Thin clouds
Even clouds of very low optical depth exert substantial radiative effect.
PERSISTENT VERY THIN CIRRUS AT MIDLATITUDE SITE
Kienast-Sjögren et al., 9th Int. Symp. on Tropospheric Profiling, 2012
Optical depth of cirrus layer estimated from lidar return as 0.003 to 0.004.
OPTICALLY THIN CLOUDS CAN BE PREVALENT IN TROPICS
Subvisible cirrus detected by lidar from space, DJF 0.01 ≤ τ ≤ 0.03
Clou
d Fr
actio
n, %
Martins Noel & Chepfer, JGR, 2011
CLOUD RADIATIVE EFFECTDependence on shortwave optical depth and cloud-top temperature
24-Hour average CRE, north central Oklahoma, at equinox
-200
-100
0
100
0.01 0.1 1 10 100Cloud shortwave optical depth
Shortwave Tct 280 260 240 220 K Longwave
-200
-100
0
100
200
0 02 04 06 08 001-10
-5
0
5
10
m W ,A
OT ta tceffE evitaidaR duol
C-2
20.0 40.0 60.0 80.0 01.00-60
-40
-20
0
20
40
60
0.0 2.0 4.0 6.0 8.0 0.1-200
-100
0
100
200
0 2 4 6 8 01
Cloud shortwave optical depthLongwave CRE also initially linear; saturates; depends on cloud-top temp.
CLOUD RADIATIVE EFFECTDependence on shortwave optical depth and cloud-top temperature
24-Hour average CRE, north central Oklahoma, at equinox
-200
-100
0
100
0.01 0.1 1 10 100Cloud shortwave optical depth
Shortwave Tct 280 260 240 220 K Longwave Net
-200
-100
0
100
200
0 02 04 06 08 001-10
-5
0
5
10
m W ,A
OT ta tceffE evitaidaR duol
C-2
20.0 40.0 60.0 80.0 01.00-60
-40
-20
0
20
40
60
0.0 2.0 4.0 6.0 8.0 0.1-200
-100
0
100
200
0 2 4 6 8 01
Cloud shortwave optical depthNet CRE depends on optical depth and cloud-top temperature even in sign. Knowledge of cloud fraction tells you nothing about the cloud radiative effect.
Shortwave Tct 280 260 240 220 K Longwave Net
A FOOL’S ERRANDThreshold photometric determination of cloud fraction
Natural color photo
What is the cloud fraction?
A FOOL’S ERRANDThreshold photometric determination of cloud fractionNatural color photo Red/(Red + Blue)
Examine ratio Red/(Red + Blue), common cloud discrimination technique.
A FOOL’S ERRANDThreshold photometric determination of cloud fraction
0.30, 86%Threshold, Cloud Fraction
Natural color photo Red/(Red + Blue)
Apply cloud mask with threshold R/(R+B) 0.30. Cloud fraction 86%. Threshold is too low.
A FOOL’S ERRANDThreshold photometric determination of cloud fraction
0.40, 35%0.30, 86%Threshold, Cloud Fraction
Natural color photo Red/(Red + Blue)
Try threshold 0.40. Cloud fraction is 35%. That threshold is too high.
A FOOL’S ERRANDThreshold photometric determination of cloud fraction
0.35, 63% 0.40, 35%0.30, 86%Threshold, Cloud Fraction
Natural color photo Red/(Red + Blue)
Try threshold 0.35. Cloud fraction is 63%.There is no “just right”. False positives and false negatives.
A FOOL’S ERRANDThreshold photometric determination of cloud fraction
40x103
30
20
10
0
slexiP
1.0
0.8
0.6
0.4
0.2
0.0
noitcarF duolC
0.450.400.350.300.25
0.35, 63% 0.40, 35%0.30, 86%Threshold, Cloud Fraction
Natural color photo Red/(Red + Blue)
Examination of cloud fraction as function of Red/(Red + Blue) threshold. There is no unique threshold.
R/(R+B)
CLOUD FRACTION: CAN IT BE DEFINEDAND MEASURED? AND IF WE KNEW IT
WOULD IT BE OF ANY USE TO US?
CONCLUSIONSNo!No!No!
Cloud fraction is dead! Do not resuscitate.