Automated Quality Control and Analysis of NOAA WP-3D Airborne Doppler Data and its Use during the

2005 Hurricane Season

John Gamache

NOAA/AOML/Hurricane Research Division

WP-3D Aircraft and their Radars

Vertically scanning X-band Doppler radar aboard the NOAA P-3 Aircraft

NOAA Antenna

Radar Characteristics

• 3.22 cm wavelength--X band• Beam width normal to scanning is 1.35 deg--1 km

resolution at 40 km--long range horizontal resolution

• Beam width along scan is 1.9 degrees--translates to 1 km at 30 km range or 2 km at 60 km range--long-range vertical resolution

• Radial (along beam) resolution is 150 m. This is the resolution directly above and below the aircraft.

JHT Project: Real-Time Dissemination of Hurricane Wind Fields Determined from Airborne Doppler Radar Data

Project officially ended June 30 2005

Goals:• To demonstrate capability to automatically quality control

airborne Doppler data well enough that they may be assimilated in “real-time”, and also used to produce analyses of the hurricane core winds--demonstrated in 2004

• To demonstrate capability to produce three-dimensional wind analyses in real time on the aircraft--demonstrated in 2004

• To demonstrate capability to transmit the analyses to the NHC/TPC--demonstrated in 2005

• To demonstrate transmission of the quality-controlled Doppler radial-velocity data to EMC for assimilation into HWRF--real time quality control has been done but transmission presently severely limited by bandwidth

HRD Software Prior to JHT

• Three-dimensional variational analysis of winds from airborne Doppler radar data

• Higher resolution radius height cross-section of wind along the inbound and outbound legs

Development Required for JHT Project

• Automated Quality Control of Raw Doppler Data (completed)

• Software to generate batch job files (completed, although present implementation may still require too much operator input--improved in last 2 months)

• Software to generate output files that can be transmitted off the aircraft by SATCOM to TPC (completed) and EMC (completed but not yet transmitted)

• Path to TPC (completed, but will be revised to make more operational) and EMC (not yet completed)

Variational Analysis--Minimize the following simultaneously:

• The difference between the projection of wind analysis on the Doppler radials and the original Doppler radial velocities.

• The three-dimensional mass divergence (analestic approximation).• Second derivative in all three directions, as well as cross

derivatives. This is a smoother and should be given a light weight.• Difference between vertical wind at vertical boundaries, and wind-

analysis vertical wind.

€

FN+1 = (∂ρ (z)uijk

∂x+

∂ρ (z)vijk

∂y+

i=1

I∑

j=1

J∑

k=1

K∑

∂ρ (z)wijk

∂z)2

€

Fn = δijkm[VRmn −α mnuijk − βmnvijk −γ mn (wijk −VTijk )i=1

I∑

j=1

J∑

k=1

K∑

m=1

Mn

∑ ]2

Doppler projection equations

Anelastic mass-continuity equation

€

FN+2 =

∂ 2u∂x2

⎛

⎝ ⎜

⎞

⎠ ⎟

2

+∂ 2u∂y2

⎛

⎝ ⎜

⎞

⎠ ⎟

2

+∂ 2u∂z2

⎛

⎝ ⎜

⎞

⎠ ⎟

2

+

2∂ 2u∂x∂y

⎛

⎝ ⎜

⎞

⎠ ⎟2

+ 2∂ 2u∂x∂z

⎛

⎝ ⎜

⎞

⎠ ⎟2

+ 2∂ 2u∂y∂z

⎛

⎝ ⎜

⎞

⎠ ⎟2

+

∂ 2v∂x2

⎛

⎝ ⎜

⎞

⎠ ⎟

2

+∂ 2v∂y2

⎛

⎝ ⎜

⎞

⎠ ⎟

2

+∂ 2v∂z2

⎛

⎝ ⎜

⎞

⎠ ⎟

2

+

2∂ 2v

∂x∂y

⎛

⎝ ⎜

⎞

⎠ ⎟2

+ 2∂ 2v

∂x∂z

⎛

⎝ ⎜

⎞

⎠ ⎟2

+ 2∂ 2v∂y∂z

⎛

⎝ ⎜

⎞

⎠ ⎟2

+

∂ 2w∂x2

⎛

⎝ ⎜

⎞

⎠ ⎟

2

+∂ 2w∂y2

⎛

⎝ ⎜

⎞

⎠ ⎟

2

+∂ 2w∂z2

⎛

⎝ ⎜

⎞

⎠ ⎟

2

+

2∂ 2w∂x∂y

⎛

⎝ ⎜

⎞

⎠ ⎟2

+ 2∂ 2w∂x∂z

⎛

⎝ ⎜

⎞

⎠ ⎟2

+ 2∂ 2w∂y∂z

⎛

⎝ ⎜

⎞

⎠ ⎟2

⎡

⎣

⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢

⎤

⎦

⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥

i=1

I∑

j=1

J∑

k=1

K∑

Filtering:

€

FN+3 = wij12

i=1

I∑

j=1

J∑

€

FN+4 = δTijki=1

I∑

j=1

J∑

k=1

K∑ wijk

2 ,

where ijk = 0 if ijk does not represent a top boundary, ijk = 1 if ijk does represent a top boundary.

Top boundary condition:

Bottom boundary condition:

€

F = λ 1F1 +L + λ N FN

+λ N+1FN+1 + λ N+2FN+2 + λ N+3FN+3 + λ N+4 FN+4

€

∂F∂uijk

= 0,

€

∂F∂vijk

= 0,

€

∂F∂wijk

= 0.

and

For every point ijk there are three equations:

If there are Np points then the system to be solved will have 3Np equations. After solution points notconstrained by data are flagged.

The total cost function F is then given by:

Bousquet and Chong (1998)

Doppler de-aliasing/unfolding

•Doppler velocity comes from phase shift (maximum unambiguous phase shift is - to + ).

usually represents 20 m/s or Nyquist velocity (VN). 21 m/s away from the Doppler radar will be interpreted as 19 m/s (-19 m/s) toward the radar.

•Thus VRobs = VDop + 2n * (VN). De-aliasing or unfolding means determining n at each radar bin.

•Bargen-Brown de-aliasing consists of using the average of several radar bins inward from the present bin to determine the most likely value of n at the present bin. First bin determined from wind measured at radar.

•This will even operate with a gap in data, although the process then becomes more unreliable.

Automatic quality control

• Remove data with high velocity spectral width• Remove reflection of main- and side-lobe by sea surface• Remove isolated speckles which can fool the automatic Doppler

unfolding process• Use single ray “Bargen-Brown” automatic unfolding• Use a full sweep de-aliasing method developed at HRD• Produce a wavenumber 0 and 1 analysis of wind velocity• Use low-wavenumber analysis to improve “Bargen-Brown” and

sweep dealiasing and further quality control• Interpolate quality-controlled sweeps into three-dimensional

cartesian analysis and higher-resolution radial-vertical cross-sections

Doppler Radial Velocity Hurricane Humberto232855 UTC 23 Sep 2001

First Pass--Wave No. 0 and 1 total wind speedHurricane Katrina 28 August 2005

96 km-96 km

96 km

96 km-96 km

96 km

No

rth

-so

uth

dis

tan

ce f

rom

cen

ter

East-west distance from center 0 45 90 m/s

0.5 km altitude 1.0 km altitude

Second Pass--Fully 3D cartesian analysisHurricane Katrina 28 August 2005

East-west distance from center

96 km-96 km

96 km

No

rth

-so

uth

dis

tan

ce f

rom

cen

ter

96 km-96 km

96 km

0 45 90 m/s

0.5 km altitude 1.0 km altitude

Radius-height cross-section of total wind speedHurricane Katrina 1725-1818 UTC 28 August 2005

96 km0 km

90 m/s60300 m/s175 kts58 117

radius

hei

gh

t

18 km

Radius-height cross-section of radial wind speedHurricane Katrina 1725-1818 UTC 28 August 2005

-40 -20 0 20 40 m/s

96 km0 km radius

hei

gh

t

18 km

Dramatic 12-h change in Katrina Wind Profile: CAT5-CAT3Dramatic 12-h change in Katrina Wind Profile: CAT5-CAT3

Doppler Wind Profile - 29 Aug 1000-1040 UTCDoppler Wind Profile - 29 Aug 1000-1040 UTC

W NE

Doppler Wind Profile - 28 Aug 1725-1820 UTCDoppler Wind Profile - 28 Aug 1725-1820 UTC

1 km

SW NE

Flight Level

1 km

Flight Level

12 km

12 km

Products produced on aircraft

• Three-dimensional wind field centered on storm centerdomain: (44 x 44 x 37), resolution: (3-5 km x 3-5 km x .5 km)variables: horizontal and vertical wind velocity, radar reflectivity

• Radial-vertical cross-sections of wind along the average azimuth flown outward from the storm center

domain: (59 x 121), resolution (1.5 km radial, 150 m vertical)variables: tangential, radial, vertical, and total wind speed

• Trimmed set of quality-controlled Doppler radial-velocity measurements

Products transmitted from aircraft via SATCOM in 2005

• u- and v-components of wind at 0.5- and 1.0-km levels (1640 and 3280 ft) sent in knots as 4 gzipped ASCII text files (4-5 km horizontal resolution

• Inbound and outbound vertical-radial profiles of total wind speed (1.5 km radial resolution, .15 km vertical resolution, out to 88 km)

Wind speed in m/s at 1-km altitudeHurricane Ivan 12 September 1115-1145 UTC

(digital camera picture of N43RF airborne workstation)

40 Real-time analyses transmitted in 2005

• Hurricane Katrina--14 analyses25 August (Florida Landfall)--5, 27 August--4, 28 August--5

• Hurricane Ophelia--13 analyses8 September--2, 9 September--311 September--6, 12 September--2 (Extratropical Transition)16 September--1, 17 September--1

• Hurricane Rita--6 analyses20 September--1, 21 September--4, 22 September--1

• Hurricane Wilma--7 analyses22 October--5, 23 October--2

• Higher resolution post-season analysis of all 2005 cases except Gert can be found in the data section at: http://www.aoml.noaa.gov/hrd

Real-time 1600-ft analysis of Hurricane Katrina Wind1725 - 1818 UTC, 28 August 2005

wind in knots

-89 -88

27

26

-89 -88

27

26

Real-time Radius-heightProfile of total

wind speed

Azimuth 045from center

Azimuth 225from center

HurricaneKatrina

1725-1818 UTC28 Aug 2005

50 nm0 nm

10,000 ft

169 kt

168 kt

he

igh

t

Radial distance (nm)

dBZ dBZ dBZ

dBZ dBZ dBZm/s m/sm/s

m/s m/sm/s

Sonde - Doppler total-wind-speed intercomparisonHurricane Katrina 28 August 2005

Sonde

DopplerWind

Radar reflecitivity

• Radial wind mean difference (sonde – Doppler): -6.7 m/s -13.0 kts• Tangential wind mean difference (sonde – Doppler): -1.5 m/s -2.9 kts• Wind speed mean difference (sonde – Doppler): -0.4 m/s -0.8 kts• Radial wind RMS difference: 13.6 m/s 26.4 kts• Tangential wind RMS difference: 7.5 m/s 14.6 kts• Wind speed RMS difference: 6.4 m/s 12.4 kts

Comparison for all drops on 28 August 2005 by NOAA 43with real time airborne Doppler analysis (0.5 - 1.0 km layer)

*note--nearly all drops are eyewall (wind max) dropswhere wind variability is high

Radial wind mean difference (sonde – Doppler): -1.4 m/s -2.7 ktsTangential wind mean difference (sonde – Doppler): 0.2 m/s 0.4 ktsWind speed mean difference (sonde – Doppler): 0.8 m/s 1.6 ktsRadial wind RMS difference: 7.3 m/s 14.2 ktsTangential wind RMS difference: 6.0 m/s 11.7 ktsWind speed RMS difference: 6.2 m/s 12.1 kts

Comparison of all NOAA drops in 2004with automatic quick-look airborne Doppler analyses

(from 0.5 km to flight level)*note--nearly all drops are eyewall (wind max) drops

where wind variability is high

Future work

•Automatically determine navigational coordinates for analysis (with some help from operator)

•Extend analysis beyond ~100 km from storm center--make it work more effectively with no well defined center (disturbances and depressions)--mainly involves allowing larger radial gaps in data in weaker systems with much weaker velocity gradients

•With faster link in future--run analysis and quality software on the ground

•Assimilation testing in HWRF

•Determine errors in radial velocity better.

34

Aircraft & Aerosonde Plots/Ophelia 16 Aug, 2005

96

35

Aerosonde/Doppler radar Ophelia 20050916

Doppler radar analysis at time of closest approach of Aerosonde to wind center and just after WP-3D SFMR penetration across the eye.

96

36

96

Real-time Doppler winds and vertical profiles

NE

SW

96

96

Aerosonde/Doppler 20050916

Nominal Aerosonde ht: 2500 ft (~.75 km)

Hurricane Wilma 1900-1945 UTC 23 October 2005Total Wind Speed

0.5 km altitude 1.0 km altitude

0 45 90 m/s0 45 90 m/s

No

rth

-so

uth

dis

tan

ce f

rom

cen

ter

(km

)

East-west distance from center (km)96

No

rth

-so

uth

dis

tan

ce f

rom

cen

ter

(km

)

East-west distance from center (km)

96-96

96

96-96

Hurricane Wilma 2025-2110 UTC 23 October 2005Total Wind Speed

0.5 km altitude 1.0 km altitude

0 45 90 m/s0 45 90 m/s

96

No

rth

-so

uth

dis

tan

ce f

rom

cen

ter

(km

)

East-west distance from center (km)

96-96

96

No

rth

-so

uth

dis

tan

ce f

rom

cen

ter

(km

)

East-west distance from center (km)

96-96

Hurricane Wilma 23 October 2005Total Wind Speed at 3.0 km altitude

1900-1945 UTC 2025-2110 UTC

0 45 90 m/s 0 45 90 m/s

96

No

rth

-so

uth

dis

tan

ce f

rom

cen

ter

(km

)

East-west distance from center (km)

96-96

96

No

rth

-so

uth

dis

tan

ce f

rom

cen

ter

(km

)

East-west distance from center (km)

96-96

Hurricane Wilma 1920-1945 UTCRadius-height cross-section along Azimuth 045

23 October 2005 Total Wind Speed

0 45 90 m/s

Radial distance from center (km)0 96

18

Hei

gh

t (k

m)

Hurricane Wilma 2047-2110 UTC Radial-vertical cross-section along azimuth 135

23 October 2005 Total Wind Speed

0 45 90 m/s

Radial distance from center (km)0 96

18

Hei

gh

t (k

m)