Transcript First Major Slide

Evaluation of Satellite-Derived Air-Sea
Flux Products Using Dropsonde Data
Gary A. Wick1 and Darren L. Jackson2
1NOAA
ESRL, Physical Sciences Division
2CIRES, University of Colorado
Outline
• Motivation/background
• Global Hawk observations
• Validation of Ta/Qa data
Motivation
• Desire data-based fields of turbulent air-sea
heat flux components
• Several products exist but validation still
limited
– OAFlux, HOAPS, J-OFURO
– Validation typically from ship and buoy data
• Performance in extreme environments largely
unknown
• Goal: Validate flux inputs and products in
more extreme environments and improve
uncertainty estimates
Wick and Jackson
ESA, October 2014
Satellite-Derived Flux Approach
• Turbulent flux calculation
– Bulk aerodynamic formula: QE=rCELvu(qs-qa)
– Requires transfer coefficient and mean atmospheric
variables
– Greatest variable uncertainty in near-surface air
temperature and specific humidity
• Near-surface air temperature and specific
humidity
– Regression based approach using microwave imager
and sounder data
– Trained on high-quality ship data
– Validated against ships and buoys
– Problems expected where significant precipitation
Wick and Jackson
ESA, October 2014
Multi-Sensor Retrievals of qa
and Ta
Can the combination of microwave sounder and imager data
provide improved estimates?
•
Most past qa retrievals
based solely on SSM/I
and indirect relationship
between surface values
and total column water
vapor
•
Jackson et al. (2006)
achieved improved
accuracy through
inclusion of sounder data
•
Profile information from
microwave sounder data
helps remove variability
in total column
measurements not
associated with the
surface
Figure courtesy P. Zuidema
Satellite-Derived Flux Approach
• Turbulent flux calculation
– Bulk aerodynamic formula: QE=rCELvu(qs-qa)
– Requires transfer coefficient and mean atmospheric
variables
– Greatest variable uncertainty in near-surface air
temperature and specific humidity
• Near-surface air temperature and specific
humidity
– Regression based approach using microwave imager
and sounder data
– Trained on high-quality ship data
– Validated against ships and buoys
– Problems expected where significant precipitation
Wick and Jackson
ESA, October 2014
Ta and Qa Retrievals
• Qa Retrievals
– Multi-channel SSM/I regression retrievals (Schluessel, Schulz,
Bentamy)
– Multi-channel regression retrieval (Jackson and Wick) combining
imager (SSM/I) and sounder (AMSU) observations
– Neural network retrievals (SSM/I – Roberts, AMSU – Shi)
– Retrieval errors range from 0.8 – 1.3 g/kg for these methods
• Ta Retrievals
– Multi-channel, multi-satellite, regression retrieval (Jackson and
Wick)
– Neural network retrievals (SSM/I – Roberts, AMSU – Shi)
– SST is a key parameter used in all Ta retrievals.
– Retrieval errors range from 1.0 – 1.5oC
• Flux products
– Need ~0.5 g/kg and 1.0oC accuracy to achieve 10 Wm-2 flux
accuracy
Wick and Jackson
ESA, October 2014
Multi-Satellite Regression Method
•
•
•
•
•
•
•
Construct grids of satellite
brightness temperatures over 3
hour time periods.
Training data sets includes 2580
collocated satellite observations
with ESRL and NOAA research
vessels
Use forward multiple linear
regression approach for both Qa
and Ta.
Qa = F (T252.8, T52.8, T19v, T22v, T37v)
Ta = F (T52.8, T53.6, T19v, T22v, T37v)
Quadratic 52.8 GHz term better
represents non-linear relationship
between lower tropospheric
temperature and Qa.
Validation accomplished with
ICOADS and SAMOS ship
observations
Wick and Jackson
ESA, October 2014
Endurance: 24-26 hours
Cruise Altitude: ~18.3 km
Cruise Speed: 620 km/hr
Gary Wick
Robbie Hood, Program Director
Relevant Global Hawk Missions
•
NASA Genesis and Rapid Intensification Processes
(2010)
•
•
•
NOAA Winter Storm Pacific and Atmospheric Rivers
(2011)
•
•
•
Dropsondes and microwave sounder
First operational dropsonde deployment, 177 sondes total
NASA Hurricane and Severe Storm Sentinel (2011
– 2014)
•
•
•
First overflights of hurricanes
Microwave sounder, Doppler radar, lightning sensor
Remote deployment of 2 Global Hawk aircraft
• Dropsondes, infrared sounder, cloud lidar
• Doppler radar, microwave sounder, surface winds
3 field seasons in Atlantic basin
NOAA Sensing Hazards with Operational
Unmanned Technology (2014 - 2017)
•
•
Tropical storms and Pacific high-impact weather
Dropsondes, microwave sounder, Doppler radar?
Wick and Jackson
ESA, October 2014
Global Hawk Flights in HS3
Wick and Jackson
HS3 Science Meeting, April 30, 2014
Winter Storms and Pacific Atmospheric
Rivers
•
WISPAR Feb-Mar 2011, NASA
Dryden Flight Research Center,
Edwards, CA
•
WISPAR flights were designed to:
•
•
Demonstrate the NOAA/NCAR
GH dropsonde system for NOAA
operations and research
•
Evaluate the capabilities of the
GH for operational observations
of atmospheric rivers (ARs),
winter storms, and remote Arctic
atmosphere
177 soundings performed on 3
high-altitude long-endurance
science flight
Wick and Jackson
ESA, October 2014
Global Hawk Dropsonde System
• Developed through partnership
between NOAA, NCAR, and
NASA
• Uses new Global Hawk sonde:
smaller and lighter than
standard dropsondes
• System has 88-sonde and 8channel capacity (track 8
sondes simultaneously)
• Sustained deployment rate of
every 2.5 min (~25 km)
• Capability to drop up to 8 sondes
every 1.5 min
• Automated telemetry frequency
selection
• Real-time data processing and
transmission
Wick and Jackson
ESA, October 2014
Sonde Specifications
•
•
•
•
Size: 4.5 cm dia. X 30.5 cm length
Mass: ~167 g
Fall rate: ~11 m/s at surface
Sensors based on Vaisala RS-92
radiosonde sensor module
•
•
•
•
•
Winds based on OEM GPS receiver and
position
•
•
•
•
Temperature: +60° to -90 ° C , 0.01 ° C
resolution
Humidity: 0 to 100%, 0.1% resolution
Pressure: 1080 to 3 mb, 0.01 mb resolution
2 Hz update rate
4 Hz update rate
Stable cone parachute design
Remote control of power on/off and
sonde release
Designed for extreme environmental
conditions
Wick and Jackson
ESA, October 2014
Collocation Approach
• 2011 and 2013 Global Hawk
dropsondes and NCAR hurricane
database (G-IV and P-3) for 20102011
• Matched dropsonde splash location
with 3-hr, 0.25 degree satellite grids
of Ta and qa
• Allow 1 cell separation (< 50 km)
and < 6 hr time difference
• Extracted single dropsonde
observation between 8-12 m above
surface
Wick and Jackson
F17 SSMIS March 10, 2011
ESA, October 2014
Global Hawk Results
•
•
•
•
Ta bias at lower temperatures
and rms difference similar to
regression results
Qa bias greatest for satellite
retrieval at the highest Qa
observations
Qa rms difference significantly
affected by bias and reduces to
0.8 g/kg with a line fit
Moist bias consistent with
WHOI study comparing Qa data
with buoy data and OAFlux
Wick and Jackson
ESA, October 2014
Results from All Aircraft
•
•
•
•
G-IV and P-3 data for tropical
storm environment
Qa bias similar to Global Hawk
sondes but scatter increased
Ta bias increases with addition
of NOAA data
Potential for greater number of
“in storm” observations from P-3
aircraft to influence increased
scatter
Wick and Jackson
ESA, October 2014
Tropical Storm Results by Aircraft
• XXX
Wick and Jackson
G-IV
P-3
ESA, October 2014
Tropical Storm Humberto
Global Hawk dropsonde
pattern shown with
satellite overlay of Qa
observations
Two sets of satellite
observations 12 hours
apart corresponding to
drop locations marked in
white
Wick and Jackson
ESA, October 2014
Tropical Storm Humberto
Ta time series shows satellite
and OAFlux capture
variations in temperature
field near tropical storm
Qa variability during late period
captured well. Bias seen in
both satellite and OAFlux
product during earlier period.
Wick and Jackson
ESA, October 2014
Conclusions
• Dropsondes a valuable validation source for satellitederived flux data
– Significant new data in extreme environments
• Agreement in Ta/Qa outside of hurricane environment
encouraging
• Significantly increased scatter as approach tropical
storm environment
– Qa moist (dry) bias in high (low) IWV environment. Wet bias
also seen in SAMOS ship and OAFlux buoy results
– Ta bias smaller for all conditions
• Satellite Ta and Qa variations in tropical storm
environment agree well with dropsonde observations
Wick and Jackson
ESA, October 2014
Water Vapor Dependence
•
•
•
Qa bias related to integrated
water vapor (IWV). Satellite Qa
is too dry for low IWV and too
wet for high IWV
Ta bias smaller but may be too
cold at low IWV
Should be able to correct bias
since satellite observations can
also are used to retrieve IWV
Wick and Jackson
ESA, October 2014
Ongoing Retrieval Developments
• Increased resolution of AMSU observations from 1.0o to 0.5o
resolution.
– 1.0o resolution used to accommodate lower resolution at scan limb
– New 0.5o grid uses higher resolution at nadir views and interpolates at
limb.
• Intercalibrated satellite data.
– CSU FCDR SSM/I and SSMIS data
– NOAA/NESDIS FCDR AMSU-A data
– Will provide more stability Ta and Qa products and mitigate artificial
trends and bias in the data set.
• A new validation source using dropsondes.