AMS Annual Meeting, Seattle, WA, January 2017

Jeff Key*, Dave Santek+, Jaime Daniels#, Steve Wanzong+, Andrew Bailey@, Hongming Qi^, Wayne Bresky@, Chris Velden+, Walter Wolf#

*NOAA/National Environmental Satellite, Data, and Information Service, Madison, WI + Cooperative Institute for Meteorological Satellite Studies, University of Wisconsin-Madison #NOAA/National Environmental Satellite, Data, and Information Service, Camp Springs, MD ^NOAA/National Environmental Satellite, Data, and Information Service, Camp Springs, MD

@I.M. Systems Group (IMSG), Rockville, MD USA

15 Years of Satellite-Derived Polar Winds and Counting: Products and Impacts

Perspective

The polar winds project started with new generation satellites (Terra and Aqua) in 2000. It grew with the new generation Suomi NPP. It will continue with JPSS-1 and beyond. What has been the impact? Where do we go from here?

Background: Deriving Polar Winds

• While geostationary satellites repeatedly image the same area, the swaths of polar-orbiters only partially overlap.

• The method is to • Track cloud and water vapor

features in three consecutive orbits.

• Assign the wind vector height based on one or more methods

• Perform consistency and quality control tests

• Result: wind speed, direction,

and height with quality indicators

The white diamond is the area of overlap for three orbits.

Main Product: Single-Satellite Winds

Single-satellite winds are derived from the Moderate Resolution Imaging Spectroradiometer (MODIS), the Advanced Very High Resolution Radiometer (AVHRR), and the Visible Infrared Imaging Radiometer Suite (VIIRS)

Mixed Satellite Winds

• MODIS Aqua and Terra data streams combined. Could be A-T-A, T-A-T, T-T-A, etc.

• Pros: Complete polar coverage; lower latency (100 min rather than 200 for a triplet); somewhat lower latitude coverage (poleward of 65o or less).

• Cons: Smaller area of overlap so fewer vectors

each pass. Parallax correction and per-pixel times are necessary.

Terra only

Aqua, Terra, Aqua

This procedure has given rise to combined geostationary and polar-orbiting satellites (discussed later)

VIIRS Polar Winds

The Suomi National Polar-orbiting Partnership (S-NPP) satellite was launched 28 October 2011.

The new “nested tracking” algorithm developed for GOES-R is used for a higher-quality product.

VIIRS provides:

• High spatial resolution (375 & 750 m)

• Constrained pixel growth toward edge of swath

• A wider swath

These instrument and algorithm features result in a larger number of more accurate wind vectors.

7

Validation Validation datasets include: • Radiosonde wind observations serve as a

key validation data source • Aircraft wind observations • Model Analysis Wind Fields The various polar wind products have accuracies similar to geostationary winds, and system requirements are being met.

All Levels (100-1000 hPa)

VIIRS Polar Wind vs. Radiosonde Winds (m/s)

NHEM SHEM Accuracy 5.67 5.71 Precision 3.41 3.25 Speed bias 0.38 -0.04 Speed 17.61 14.22 Sample 9650 866

High Level (100-400 hPa)

NHEM SHEM

Accuracy 6.21 6.81 Precision 3.55 3.36 Speed bias -0.06 -0.23

d

Measurement Accuracy 7.5 m/s

Measurement Precision 3.8 m/s

Specifications:

Problem: Latency

The time it takes to generate winds after the time of the last image in an orbit triplet is, on average, 2.0-2.5 hrs for the operational winds.

Solution: Direct Broadcast (DB) Sites

Clockwise from upper left: McMurdo, Rothera (both Antarctica); Barrow, Alaska; Sodankylä, Finland; Svalbard, Fairbanks, Alaska

Latency: DB and Operational

The time it takes to generate DB winds is generally 0.5-0.8 hrs. Recall that for operational winds, the time delay is 2.0-2.5 hrs.

The Polar Winds Product Suite

MODIS • Aqua and Terra separately • Aqua and Terra combined • Direct broadcast (DB) at

– McMurdo, Antarctica (Terra, Aqua) – Tromsø, Norway (Terra) – Sodankylä, Finland (Terra) – Fairbanks, Alaska (MODIS discontinued)

AVHRR • Global Area Coverage (GAC) for NOAA-

15, -18, -19 • Metop-A, -B • HRPT (High Resolution Picture

Transmission = direct readout) at – Barrow, Alaska, NOAA-18, -19 – Rothera, Antarctica, NOAA-18, -19

• Historical GAC winds, 1982-2015. Two satellites throughout most of the time series.

Operational

Operational

Operational

VIIRS • S-NPP (GOES-R algorithm) • Direct broadcast at Fairbanks and

Sodankylä (heritage algorithm) LEO-GEO (experimental) • Combination of may geostationary

and polar-orbiting imagers • Fills the 60-70 degree latitude gap

Operational

Operational NWP Users of Polar Winds

European Centre for Medium-Range Weather Forecasts (ECMWF)

NASA Global Modeling and Assimilation Office (GMAO)

Deutscher Wetterdienst (DWD)

Japan Meteorological Agency (JMA)

Canadian Meteorological Centre (CMC)

US Navy, Naval Research Lab, Fleet Numerical Meteorology and Oceanography Center (FNMOC)

Met Office (UK)

The various polar winds products are used by 13 NWP centers in 9 countries:

National Centers for Environmental Prediction (NCEP)

Météo-France

National Center for Atmospheric Research (NCAR)

Australian Bureau of Meteorology

China Meteorological Administration (CMA)

HydroMeteorological Center of Russia (Hydrometcenter)

Wind Data Used at ECMWF

(Courtesy K. Lean)

Impact of Polar Winds Early on, the forecast impact due to the MODIS winds was significant and positive, with forecasts extended 2-6 hrs. Today the overall impact is generally less (more types of data), though the polar winds still reduce the forecast “dropouts”.

ECMWF March 2001: Anomaly correlation as a function of forecast range for the geopotential at 1000 hPa (left) and 500 hPa (right). Control (blue) and with MODIS winds (red).

CMC December 2006: Daily 500 hPa 72-hour anomaly correlation for Europe. Control (red) and with MODIS winds (blue).

Polar Winds Forecast Impact at ECMWF

The forecast sensitivity to observations (FSO) impact of polar winds (all) relative to other types of data. AMV: atmospheric motion vector

Polar winds have a significant impact on forecasts.

(Courtesy K. Salonen)

VIIRS Winds Forecast Impact at ECMWF

Change in vector wind error verified against own analysis

(Courtesy of K. Lean)

Looking ahead…

Courtesy M. Drinkwater

Combining Satellites: LEO-GEO Polar Winds

By combining data from many GEO and LEO satellites into hourly composites, the product fills a gap in the 60-70 degree latitude zone. The impact at NRL/FNMOC is positive and similar to that provided by the MODIS polar winds.

24-hr forecast error reduction in the FNMOC model for various observation types. The reduction in error resulting from the incorporation of the LEO-GEO wind product is circled. This wind product will be incorporated into FNMOC’s operational system after further testing.

New Spectral Channels: VIIRS and MODIS Winds from the Shortwave Infrared

MODIS

Vis (0.6 𝜇𝜇m)

(animations)

IR (11 𝜇𝜇m)

SWIR (2.1 𝜇𝜇m)

SWIR allows for better tracking of low, liquid clouds

SWIR, IR, and Water Vapor Retrieval Examples

MODIS VIIRS

SWIR WV & IR

The SWIR band provides more low-level vectors

New Spectral Bands: VIIRS Day-Night Band (DNB)

The VIIRS Day/Night Band (DNB) provides new possibilities for estimating wind properties at night.

VIIRS Day-Night Band (DNB) NCC EDR Albedo, 21-Aug-2013

02:29-02:35 Full Moonlight Night Scene

thin clouds

sea ice

sea ice

thin clouds

(Courtesy of CIRA)

3D Winds from Satellite Sounders Method: • Perform retrievals of vertical temperature

and humidity • Create images horizontal fields of humidity • Track humidity features over time Advantages: • 3D wind distribution • Implicit AMV height • Clear sky and above cloud Disadvantages: • Lower spatial resolution compared to

MODIS (13.5 vs. 1 km) • Narrower swath

Impact per observation (not overall impact), 1-24 July 2012

Retrieval winds from AIRS

Global Winds from JPSS-1 and S-NPP in Tandem

JPSS-1 and S-NPP will be in the same orbit separated by about 50 minutes, similar to Metop-A and –B. This provides an opportunity for global wind retrieval if the quality of the wind vectors is sufficient (two orbits are used rather than three as is currently done).

• First UV (λ = 355nm)

Doppler wind lidar (ALADIN)

• Laser transmitters qualified for flight

• Flight acceptance and launch - end 2017

• Observations of wind profiles for analysis of global wind field

• Understanding of atmosphere dynamics and climate processes

• Improved weather forecasts and climate models

ADM-Aeolus: ESA’s Doppler Wind Lidar Mission

• UV lidar (355 nm) with Mie and Rayleigh receivers • Doppler shift used to retrieve Horizontal Line of Sight component of wind velocity

25

Summary

• Automated, satellite-derived polar wind products were first developed and model-tested in 2001.

• Early NWP impact studies demonstrated a positive impact not just in the polar regions, but globally. The positive impact can still be seen, though lessened somewhat due to the availability of other data.

• In some sense, the polar winds created a significant “mutation” in the observing system. Will there be another?

Acknowledgements: From the NWP community: Mary Forsythe, Randy Pauley, Niels Bormann, Lars Peter Riishojgaard, John LeMarshall, Jim Jung, Real Sarrazin, Alexander Cress, Koji Yamashita, Masahiro Kazumori, Christophe Payan, Yan-Qiu Zhu, Claire Delsol, Lueder von Bremen, Iliana Genkova. From the remote sensing community: Matthew Lazzara and Paul Menzel.

Polar Winds Products: History

2001 2002 2004 2003 2006 2005 2007 2008 2009

10-day MODIS winds case study made available

Real-time Terra and Aqua MODIS winds

ECMWF and NASA DAO demonstrate positive impact

Daily AVHRR winds

Terra MODIS winds in ECMWF operational system

MODIS DB winds at McMurdo

MODIS DB winds at Tromsø

Real-time AVHRR winds

DB winds distributed via EUMETCast

Mixed-satellite MODIS winds

MODIS DB winds at Sodankylä

Metop AVHRR winds

AVHRR HRPT winds from Barrow

NESDIS MODIS winds on GTS

AVHRR HRPT winds from Rothera

MODIS DB winds at Fairbanks

2015 2014 2013 2010 2011 2012

VIIRS winds

LEO-GEO winds

AVHRR historical winds (v2)

VIIRS DB winds at Fairbanks

MODIS, AVHRR winds with nested tracking algorithm

Extra Slides

AMV Performance Metrics

where:

Ui and Vi ---> AMV Ur and Vr ---> “Truth”

AMVs (QI>60) are matched and compared against RAOBS or GFS model analysis winds. Metrics:

Comparison statistics of VPW product computed using the M15 band (10.76um),

29

All Levels (100-1000 hPa)

VIIRS Polar Wind vs. Radiosonde Winds (m/s)

GFS Forecast Winds vs. Radiosonde Winds (m/s)

NHEM SHEM NHEM SHEM Accuracy 5.67 5.71 4.54 4.77 Precision 3.41 3.25 3.06 2.99 Speed bias 0.38 -0.04 -0.30 -0.57 Speed 17.61 14.22 16.93 13.69 Sample 9650 866 9650 866

High Level (100-400 hPa)

NHEM SHEM NHEM SHEM

Accuracy 6.21 6.81 5.08 5.56 Precision 3.55 3.36 3.23 3.14 Speed bias -0.06 -0.23 -0.69 -0.55 Speed 23.62 18.05 22.99 17.73 Sample 3054 301 3054 301

Mid Level (400-700 hPa)

NHEM SHEM NHEM SHEM

Accuracy 5.65 5.24 4.48 4.48 Precision 3.40 3.12 3.04 2.87 Speed bias 0.56 0.07 -0.32 -0.75 Speed 16.69 12.51 15.81 11.69 Sample 4468 471 4468 471

Low Level (700-1000 hPa)

NHEM SHEM NHEM SHEM

Accuracy 4.95 4.55 3.90 3.70 Precision 3.08 2.39 2.69 2.40 Speed bias 0.64 0.04 0.32 0.28 Speed 10.91 10.52 10.58 10.76 Sample 2128 94 2128 94

Comparisons to Radiosondes September, 2013 – January, 2014

Measurement Accuracy 7.5 m/s

Measurement Precision 3.8 m/s

Specifications:

Comparison of the VPW product with aircraft data. There were insuffient data from the Southern Hemisphere for reliable statistics for different height bins.

30

All Levels (100-1000 hPa)

VIIRS Polar Wind vs. Aircraft

Winds (m/s)

VIIRS Polar Wind vs. Aircraft

Winds (m/s) NHEM SHEM

Accuracy 5.77 6.77 Precision 3.56 3.83 Speed bias 1.08 -1.67 Speed 21.62 29.97 Sample 34998 354

High Level (100-400 hPa)

NHEM SHEM

Accuracy 6.48 6.77 Precision 3.70 3.83 Speed bias 0.45 -1.67 Speed 27.27 29.97 Sample 14781 354

Mid Level (400-700 hPa)

NHEM SHEM

Accuracy 5.50 NA Precision 3.64 NA Speed bias 1.52 NA Speed 19.59 NA Sample 14775 NA

Low Level (700-1000 hPa)

NHEM SHEM

Accuracy 4.59 NA Precision 3.04 NA Speed bias 1.57 NA Speed 11.75 NA Sample 5442 NA

Comparisons to Aircraft June, 2014 – September, 2016

Measurement Accuracy 7.5 m/s

Measurement Precision 3.8 m/s

Specifications:

Comparisons of the algorithm’s derived winds against raob and aircraft winds at all levels (100-1000 hPa), high level (100- 400 hPa), mid level (400-700 hPa), and low level (700-100 hPa) in the northern hemisphere. In each case, the observed precision meets the requirement. The accuracy and precision of the VIIRS winds fall well within the accuracy and precision specifications.

VIIRS Coverage: Wider Swath

MODIS VIIRS

VIIRS has a wider swath (3000 km) than MODIS (2320 km), so the coverage will be better and will extend further south.

A wider swath means more winds with each orbit.

Improved Resolution at Edge of Swath • VIIRS method of aggregating detectors and deleting portions of the scans near the swath

edge results in smaller pixels at large scan angles. • For thermal bands, VIIRS is 0.56 km2 (0.75 x 0.75 km) at nadir and 2.25 km2 at the edge of

the swath (0.37 -> 0.8 km for imager bands; 0.74 -> 0.74 km for DNB) • In contrast, AVHRR and MODIS are 1 km2 at nadir and 9.7 km2 at edge of swath. • Additionally, VIIRS scan processing reduces the bow-tie effect.

Effects of Scan Angle on Winds

Many of the winds toward the swath edges are filtered out by QC.

Operational ECMWF system September to December 2008. Averaged over all model layers and entire global atmosphere. % contribution of different observations to reduction in forecast error.

Courtesy: Carla Cardinali and Sean Healy, ECMWF

Forecast error contribution (%)

0 2 4 6 8 10 12 14 16 18

O3: Ozone from satellites METEOSAT IR Rad (T,H)

MTSATIMG: Japanese geostationary sat vis and IR imagery GOES IR rad (T,H)

MODIS: Moderate Resolution Imaging Spectroradiometer (winds) GMS: Japanese geostationary satellite winds

SSMI: Special Sensor MW Imager (H and sfc winds) AMSRE: MW imager radiances (clouds and precip)

MHS: MW humidity sounder on NOAA POES and METOP (H) MSG: METEOSAT 2nd Generation IR rad (T,H)

HIRS: High-Resol IR Sounder on NOAA POES (T,H) PILOT: Pilot balloons and wind profilers (winds)

Ocean buoys (Sfc P, H and winds) METEOSAT winds

GOES winds AMSU-B: Adv MW Sounder B on NOAA POES

SYNOP: Sfc P over land and oceans,H, and winds over oceans QuikSCAT: sfc winds over oceans

TEMP: Radiosonde T, H, and winds GPSRO: RO bending angles from COSMIC, METOP

AIREP: Aircraft T, H, and winds AIRS: Atmos IR Sounder on Aqua (T,H)

IASI: IR Atmos Interferometer on METOP (T,H) AMSU-A: Adv MW Sounder A on Aqua and NOAA POES (T)

Note: 1) Sounders on Polar Satellites reduce forecast error most 2) Results are relevant for other NWP Centers, including NWS/NCEP

MODIS and AVHRR Winds Forecast Impact

at ECMWF

The per-obervation impact of MODIS water vapor and IR winds and AVHRR IR winds.

The impact of cloudy and clear sky polar winds is similar to AMVs from geostationary satellites.

Direct Broadcast Polar Winds

Arctic Antarctic

?