NASA Langley Research Center / Atmospheric Sciences Comparison of RUC Cloud Analyses and Forecasts with Satellite-derived Cloud Properties Bill Smith , Patrick Minnis NASA LaRC Stan Benjamin, Steve Weygandt NOAA/ESRL/GSD/AMD
Mar 13, 2016
NASA Langley Research Center / Atmospheric Sciences
Comparison of RUC Cloud Analyses and Forecasts with
Satellite-derived Cloud Properties
Bill Smith, Patrick MinnisNASA LaRC
Stan Benjamin, Steve WeygandtNOAA/ESRL/GSD/AMD
NASA Langley Research Center / Atmospheric Sciences
Outline
• Recent satellite validation (GOES Cloud height, water path)
• Statistical Comparisons (several months)– Cloud Amount– Cloud Height– Cloud Thickness– Cloud Water Path
• Summary
Goal: Perform statistical comparison between RUC and LaRC satellite-derived cloud properties to help guide the assimilation of satellite cloud parameters and provide baseline for future model evaluations.
NASA Langley Research Center / Atmospheric Sciences
Satellite Data• NASA LaRC Real-Time 8-km CONUS cloud products
derived from GOES-E and GOES-W (Cloud top phase, height, water path)
• CLOUDSAT cloud water path
• NOAA NESDIS Operational Cloud top Height Product (CO2-slicing)
NASA Langley Research Center / Atmospheric Sciences
RUC Data
Dataset 1 (4 months):• Feb 12 - Apr 27, 2007; May 10 - July 4, 2007• Analyses, 1-hr, 3-hr and 6-hr forecasts valid at 18 UTC• Obtained to test Model analyses and forecasts with GOES
Note: Jan 8, 2008 RUC crisis change implemented for improved cloud/vis assimilation.
Dataset 2 (55 days):• March 17 - May 28, 2008 (18, 19, 20, 21, 22 UTC Analyses)• No Forecasts• Obtained for GOES and Cloudsat Comparisons over CONUS
NASA Langley Research Center / Atmospheric Sciences
RUC Cloud AssimilationUse multiple data types to modify cloud,hydrometeor, and moisture fields:
-- NESDIS GOES cloud-top pressure/temp(implemented in 2002 w/ RUC20)
-- Surface METAR (clouds, weather, visibility)(2005 w/ RUC13)
• Construct 3-d logical arrays (YES/NO/UNKNOWN)for clouds and precipitation from all info
• Clear/build moisture, cloud, precipitation fields• Safeguards for known problems (marine stratus,
convective clouds, snow, nocturnal inversion)
NASA Langley Research Center / Atmospheric Sciences
Satellite Validation
NASA Langley Research Center / Atmospheric Sciences
LaRC and NESDIS Cloud Top Height Comparisons with ARM ARSCL data
Smith et al. 2008 (GRL)
5-yr datasetSingle-layer clouds
X - CO2 slicingLaRC
Satellite-derived Cloud Top HeightBias Errors (km)
NASA Langley Research Center / Atmospheric Sciences
Satellite Liquid Water Path Validation
LWP
(MO
DIS
)
Dong et al., JGR, 2007
• Unbiased over wide range of LWP (up to 500 g/m2)• Excellent correlation • Instantaneous Uncertainty ~30%• Note mean value ~150 g/m2
NASA Langley Research Center / Atmospheric Sciences
Cloud Ice Water Path (From Waliser et al. 2008)Cloud Ice Water Path (From Waliser et al. 2008)
LaRC CERES (analogous to LaRC GOES) compares well with CloudSat IWP
NASA Langley Research Center / Atmospheric Sciences
Instantaneous matched GOES-CloudSat Water Path1930 UTC, 1 March 2007
010002000300040005000
19.4919.519.5119.5219.5319.5419.55UTC
Wat
er P
ath
(g/m
2)CloudSat TWPGOES IWP
CloudSat Reflectivity
NASA Langley Research Center / Atmospheric Sciences
GOES Cloud Water Path Comparisons with Cloudsat
Dec 2006 – May 2007
• Excellent agreement between GOES and Cloudsat monthly mean total water path for high thick clouds.
• Thin ice cloud IWP also in good agreement
• LWP from Cloudsat is unreliable
Low Cloud LWP Hi Thin Cloud TWP Hi Thick Cloud TWP
Ocean Land Ocean Land Ocean Land Mean (CloudSat) 252 117 61 64 617 573
Mean (GOES) 167 200 56 65 637 731 Bias 85 -83 5 -1 -20 -157 rms 199 200 68 120 586 751
StdErr of Fit 125 165 31 35 581 661 R2 0.10 0.02 0.3 0.2 0.5 0.3
Npts 594 848 1086 1018 2843 4225
Statistical Summary for All Months
NASA Langley Research Center / Atmospheric Sciences
GOES/RUC Cloud Parameter Comparisons
NASA Langley Research Center / Atmospheric Sciences
Matching ProceduresSpatial Matching: GOES 8-km pixel products averaged to RUC 20 km
grid.
Temporal Matching: 1745 UTC GOES within 15 minutes of RUC analyses and forecasts with valid times at 18 UTC.
RUC grid box either OVC or CLR. GOES is OVC, CLR or PC
RUC Total Water Path: Computed from 5 species level mixing ratios (cloud water, cloud ice, rain water, snow, graupel).
RUC Cloud Phase: Only analyzed in the top layer.
Comparisons shown here are stratified by cloud phase (I.e. clouds with either liquid or ice water tops
NASA Langley Research Center / Atmospheric Sciences
GOES INFRARED GOES VISIBLE
GOES CLOUD TOP PHASE RUC CLOUD TOP PHASE (0-hr fcst)
Example Cloud Top Phase Comparison - May 8, 2008 (1800 UTC)
Ice
Liquid
Mixed
SLW
Clear
NASA Langley Research Center / Atmospheric Sciences
2007 Cloud/Clear Frequency Comparison (%)
GOES RUC 0-hr RUC 1-hr RUC 3-hr RUC 6-hr
CLEAR 33 43 45 47 47
PC 19 NA NA NA NA
OVC 48 57 56 53 53
ICE 18 40 43 37 25
LIQUID 21 12 12 14 16
NASA Langley Research Center / Atmospheric Sciences
2007 Cloud/Clear Frequency Comparison (%)
GOES RUC 0-hr RUC 1-hr RUC 3-hr RUC 6-hr
CLEAR 33 43 45 47 47
PC 19 NA NA NA NA
OVC 48 57 56 53 53
ICE 18 40 43 37 25
LIQUID 21 12 12 14 16
NASA Langley Research Center / Atmospheric Sciences
RUC/GOES CLOUD FREQUENCY DIFFERENCES (%)
FOR LAND and OCEAN
RUC ANALYSIS minus GOES RUC ANALYSIS minus GOES
Land Ocean
All cloudsIce
Liquid
322
-10
1019-4
Land Ocean
All cloudsIce
Liquid
323
-10
1015
0
2007 dataset 2008 dataset
RUC has too much high-level ice cloud (land & ocean) and too little low-level liquid cloud (over land) relative to GOES.
NASA Langley Research Center / Atmospheric Sciences
RUC Cloud Frequencies with and without NESDIS Satellite Data
Compared to LaRC GOES
RUC GOES
All cloudsIce
Liquid
7968
5
773329
RUC GOES
All cloudsIce
Liquid
543810
441617
(Lat > 50˚ N) (Lat ≤ 50˚ N)
Low-level liquid cloud frequency difference largest over Canada where no satellite data are used. Large ice cloud frequency differences in both regions.
No NESDIS Data With NESDIS Data
NASA Langley Research Center / Atmospheric Sciences
Distribution of Ice Cloud Frequency for GOES and RUC (2008)
Relative Frequency (%) ofTotal Water Path for
Clouds with Ice Phase Tops
2008 Test Dataset
0
2
4
6
8
10
0
2
4
6
8
10
0 3000+25 50 100 300 500200 750 1000 20001500
TWP (gm-2)
Rel
ativ
e Fr
eque
ncy
(%)
All
GOES
RUC Analysis
35% of the ice cloud difference is found in the lowest bin
Other possible reasons for discrepancy:
• Cloud phase differences (mixed phase regime)• Poor GOES phase ID (multilayer systems)
NASA Langley Research Center / Atmospheric Sciences
RUC CLEAR/CLOUD FREQUENCIES FOR GOES 100% OVERCAST REGIONS
RUC CLOUD %
RUC CLEAR %
ANAL 1-hr 3-hr 6-hr0
20
40
60
80
100
Freq
uenc
y
Forecast Period
13
87
19
81
22
78
23
77
ANAL0
20
40
60
80
100
Freq
uenc
y
Forecast Period
10
90
20082007
NASA Langley Research Center / Atmospheric Sciences
RUC CLEAR/CLOUD FREQUENCIES FOR GOES 100% CLEAR REGIONS
20082007
RUC CLOUD %
RUC CLEAR %
RUC LIQUID %
RUC ICE %
ANAL 1-hr 3-hr 6-hr0
20
40
60
80
100
Freq
uenc
y
Forecast Period
83
1812 5
78
2216 6
80
2013
7
79
21
11 9
ANAL0
20
40
60
80
100
Freq
uenc
y
Forecast Period
83
1711 5
NASA Langley Research Center / Atmospheric Sciences
RUC CLOUD FREQUENCIES FOR GOES OVERCAST ICE CLOUD REGIONS
20082007
ANAL 1-hr 3-hr 6-hr0
20
40
60
80
100
Freq
uenc
y
Forecast Period
89
2 3 6
88
2 1 9
82
4 1
13
78
51
16
ANAL0
20
40
60
80
100
Freq
uenc
y
Forecast Period
88
2 6 4
RUC CLOUD FREQUENCY
CLEARLIQUID CLOUDS
ICE CLOUDS
MIXED PHASE
NASA Langley Research Center / Atmospheric Sciences
RUC CLOUD FREQUENCIES FOR GOES OVERCAST LIQUID CLOUD REGIONS
20082007
RUC CLOUD FREQUENCY
CLEARLIQUID CLOUDS
ICE CLOUDS
MIXED PHASE
ANAL 1-hr 3-hr 6-hr0
20
40
60
80
100
Freq
uenc
y
Forecast Period
3339
1117
3535
2
28 29
40
3
28 26
43
4
27
ANAL0
20
40
60
80
100
Freq
uenc
y
Forecast Period
31
47
102
NASA Langley Research Center / Atmospheric Sciences
Water Path Comparison Nov 8, 2008 (1800 UTC)
GOES WP
GOES minus RUC WP
RUC WP
GOES Cloud Phase
NASA Langley Research Center / Atmospheric Sciences
Liquid Water Cloud Parameter Comparison
Ocean
Land
WP (gm-2) Zt (km) Z (km) Dataset
GOES RUC %diff GOES RUC BIAS GOES RUC BIAS 2008 Analysis 193 133 -31 2.9 2.6 -0.3 1.0 1.2 0.2 2007 Analysis 174 65 -63 3.0 2.1 -0.9 1.0 0.7 -0.3 2007 1-hr fcst 183 65 -64 3.0 1.9 -1.1 1.0 0.7 -0.3 2007 3-hr fcst 189 90 -52 3.0 2.0 -1.0 1.0 0.8 -0.2 2007 6-hr fcst 192 93 -52 3.0 1.9 -1.1 1.0 0.8 -0.2
WP (gm-2) Zt (km) Z (km) Dataset
GOES RUC %diff GOES RUC BIAS GOES RUC BIAS 2008 Analysis 109 79 -28 1.8 1.2 -0.6 0.8 0.6 -0.2 2007 Analysis 105 57 -46 1.6 0.8 -0.8 0.8 0.4 -0.4 2007 1-hr fcst 107 59 -45 1.6 0.7 -0.9 0.8 0.4 -0.4 2007 3-hr fcst 108 67 -38 1.6 0.8 -0.8 0.8 0.5 -0.3 2007 6-hr fcst 107 70 -35 1.6 0.8 -0.8 0.8 0.5 -0.3
• GOES LWP >> RUC LWP, 2008 RUC compares better• LaRC GOES cloud top heights and thicknesses exceed RUC values• Accounting for cloud thickness differences would yield better LWP agreement over Ocean but not Land.
NASA Langley Research Center / Atmospheric Sciences
Ice Cloud Parameter Comparison
Ocean
Land
• GOES IWP >> RUC IWP• LaRC GOES and RUC cloud top heights agree for Analyses• RUC forecasted cloud top height exceeds GOES by over 0.5 km• GOES cloud thickness likely underestimated (poor Deep Convective estimates)• Land/Ocean differences generally consistent
WP (gm-2) Zt (km) Z (km) Dataset
GOES RUC %diff GOES RUC BIAS GOES RUC BIAS 2008 Analysis 473 328 -31 8.5 8.3 -0.2 4.2 5.3 1.1 2007 Analysis 416 254 -39 7.9 7.9 0.0 3.4 4.5 1.1 2007 1-hr fcst 418 260 -38 7.9 8.5 0.6 3.4 5.0 1.6 2007 3-hr fcst 428 275 -36 7.8 8.6 0.8 3.5 5.0 1.5 2007 6-hr fcst 430 251 -42 7.8 8.4 0.6 3.5 4.7 1.2
WP (gm-2) Zt (km) Z (km) Dataset
GOES RUC %diff GOES RUC BIAS GOES RUC BIAS 2008 Analysis 350 214 -39 9.3 9.3 0.0 3.7 4.8 1.1 2007 Analysis 355 184 -48 8.9 8.9 0.0 3.0 4.3 1.2 2007 1-hr fcst 361 191 -47 8.8 9.4 0.6 3.0 4.6 1.6 2007 3-hr fcst 377 204 -46 8.8 9.4 0.6 3.1 4.6 1.5 2007 6-hr fcst 376 190 -49 8.8 9.2 0.4 3.1 4.3 1.2
NASA Langley Research Center / Atmospheric Sciences
Summary• The RUC is one of the more advanced NWP models
worldwide that assimilates cloud observations.
• Comparison of RUC cloud parameters with similar parameters derived from GOES indicate the RUC assimilation works reasonably well.
• However, significant differences have been identified that exceed the satellite-derived uncertainties which implies that assimilation of LaRC GOES-derived cloud products could lead to improved cloud analyses and forecasts in the model.
• Results from this study are being used to guide the assimilation of satellite-derived cloud parameters and also provide a baseline for future model evaluations.
NASA Langley Research Center / Atmospheric Sciences
Extra slides
NASA Langley Research Center / Atmospheric Sciences
Regional Liquid Water Path Comparison (g/m2)
Atlantic Pacific Gulf of Mex0
20
40
60
80
100
120
140
160
180
LWP
(g/m
**2)
GOES
RUC-Analysis
2008
Poor agreement over the Atlantic but good
over Pacific and Gulf of Mexico
NASA Langley Research Center / Atmospheric Sciences
Regional Ice Water Path Comparison (g/m2)
2008
RUC does track GOES regional differences.
Atlantic Pacific Gulf of Mex0
50100150200250300350400450500
IWP
(g/m
**2)
GOES
RUC-Analysis
NASA Langley Research Center / Atmospheric Sciences
Cloud Ice Water Path (From Waliser et al. 2008)Cloud Ice Water Path (From Waliser et al. 2008)
CERES-MODIS most similar to CloudSat values
CloudSat
CERES-MODIS ISCCP
MODIS MYD06