Assimilation and Evaluation of MISR Cloud Tracked Winds with GEOS-5

Operational Data Assimilation System

Junjie Liu1 and Kevin Mueller1

1. Jet Propulsion Lab (JPL), Caltech

With special thanks to Joe Stassi (GMAO), Dan Holdaway (GMAO), Meta Sienkiewicz (GMAO), Will McCarty (GMAO), Ron Gelaro (GMAO), Nancy Baker (NRL), David Diner (JPL), and Earl Hansen (JPL)

©2015 . California Institute of Technology. Government sponsorship acknowledged.

Outline

• Brief description of current winds in GEOS-5, MISR, motivation and objective.

• GEOS5 assimilation experiments• Results• Summary, and ongoing work and future plans

Current wind observation coverage still have large gaps

plot courtesy of Dr. Dagmar Merkova

AMVs assimilated at 00Z20 August, 2010

• Sparse observation coverage over the mid latitudes in each Hemisphere.• Low density wind observations over low levels.

4

The Multi-Angle Imaging SpectroRadiometer (MISR) retrieves along-track and cross-track

winds

• On EOS Terra platform. Mission Lifetime: 1999 -> ~2020+• Sun-synchronous, polar orbit (10:30am)• MISR NRT winds (under 2.5 hours) became publicly available since September 1, 2014• Two Camera Retrieval

– Cross-track cloud motion, ∆t = 50 seconds between two cameras. – Cloud Top Height (along-track error 90 m/ms-1)

• Three Camera Retrieval:– Full cloud motion vector, ∆t = 200 seconds

Along-tra

ckCross-track

Up

∆t ~ 7 minutes

Cross-track

Along

-trac

k

MISR wind coverage within 6 hours

5

MISR retrieves cloud height along with winds

• Height and motion retrieved in single 7 minute satellite overpass• Height measured through parallax• Along-track motion differentiated from parallax using 3rd camera

• Sensitive to camera pointing and correspondence accuracy• Yields correlated error between retrieved height and along-track motion

Two angles allow for geometric retrieval of

feature height based on pattern matching MISR

imagery (below)

Feature height and motion are retrieved

using triplets of MISR cameras

Motivation and Objective

• MISR winds has been shown to have positive impact on the forecast error reduction in NAVEM

Optimally assimilate and evaluate the impact of MISR winds on GEOS-5 operational data assimilation system

Plots courtesy of Dr. Nancy Baker (NRL)

MISR winds

MISR winds

Per observation impact (10-6J/kg) Percent reduction per observation type

0.7

6.9

GEOS-5 assimilation experiments

• GEOS-5 AGCM + GSI analysis (~0.5°, L72)• Time period: Dec 16, 2009 – Feb 16, 2010• Data thinning on 200km×100hPa box• Observation errors: 3.0 m/s• Quality control: o-f<20 m/s• Assimilate u and v components• Observation impact on 24-hour forecast error

reduction: moist energy norm, calculate every 6 hours

MISR winds has reduced 24-hour forecast errors in GEOS-5 DAS

• ~3000 MISR observations assimilated in each 6-hour assimilation cycle. • Per observation impact is the 4th largest among all observation types. • The MISR total impact is 0.7% of the impact from all obs.

3.11e-6

0.7%

Observation count by class Per obs impact by class Percent total obs impact by class

SatwindMISR

Dropsond

WindSat

MISR

Comparison to the MISR wind impact in NAVGEM

• Per observation impact of MISR winds on GEOS-5 is smaller than in NAVGEM• Percent total obs impact is the same as in NAVGEM.

Per obs impact on GEOS-5 Per obs impact on NAVGEM

3.1e-66.9e-6

0.7%

Percent impact on NAVGEMPercent total obs impact on GEOS-5

0.7%

Note: NAVGEM for Hurricane Sandy case Courtesy of Dr. N. Baker

Largest positive impact on the SH mid latitudes

• MISR have largest number of observations around 900 hPa.• Largest positive impact over the SH mid latitudes. • Negative impact ~45° in each hemisphere.

MISR wind impact vs. other wind impact

• The observation coverage over the SH mid latitudes is sparse in control GEOS-5;• MISR wind observation coverage complements the current default wind observation

coverage in GEOS-5.

All wind impact (not including MISR) All wind # (not including MISR)

MISR wind impact MISR #SH NH

12

The Multi-Angle Imaging SpectroRadiometer (MISR) retrieves along-track and cross-track

winds

• Mission Lifetime: 1999 -> ~2018+• Swath Width ~ 380 km• Observations covers from the north pole to the south pole. • MISR NRT winds became publicly available September 1, 2014• Two Camera Retrieval

– Cross-track cloud motion, ∆t = 50 seconds between two cameras. – Cloud Top Height (along-track error 90 m/ms-1)

• Three Camera Retrieval:– Full cloud motion vector, ∆t = 200 seconds

Along-tra

ckCross-track

Up

∆t ~ 7 minutes

Cross-track

Along

-trac

k

MISR wind coverage within 6 hours

MISR u-component MISR v-component

U-component has much larger impact on improving the 24-hour forecast

Per o

bs in

nova

tion

-90 90

Obs

#To

tal i

mpa

ct

• U-component has largest positive impact in lower levels; v-component has largest positive impact in higher levels.

Detecting the bad orbits with observation impact calculation and O-F statistics

• Observation impact calculation shows large negative impact from certain orbits; may have georegistration errors.

• Removing the orbits have improved the results.

MISR wind impact Dec 16, 2009 – Jan 15, 2010

Mean O-F for v-component

Summary• MISR is expected to operate till 2020+. MISR retrieves height

along with winds. Cross-track wind is more accurate than along-track winds

• MISR wind observation coverage complements the current winds assimilated in GEOS-5.

• Assimilating MISR winds has improved the 24-hour forecast accuracy. The per obs impact of MISR is the 4th largest among all observation types.

• The per obs MISR wind impact in GEOS-5 is smaller than the per-obs MISR wind impact in NAVGEM. However, the percent total MISR wind impact is 0.7%, the same as in NAVGEM.

• The u-component has much larger positive impact on the forecast error reduction than the v-component.

Ongoing work and future plans

• Developing observation operator that can assimilates cross-track and along-track winds

• Optimize MISR wind quality control method• Assess MISR wind impact on hurricane Sandy

forecast• Test the MISR wind impact with different

baseline observations (e.g., obs assimilated in NAVGEM) in GEOS-5.

• Transition from O2R to R2O