Walt Petersen and Kevin Knupp UAH/ESSC November 7, 2007 [email protected] UAH THOR Center Radar Infrastructure: Exploring QPE Algorithm Development.

Walt Petersen and Kevin Knupp UAH/ESSC

November 7, 2007

[email protected]

UAH THOR Center Radar Infrastructure: Exploring QPE Algorithm Development for Operational Support of TVA River Management

KBMX

RSA

68 km

KGWX

UAH/NSSTC THOR Center and Hazardous Weather Testbed

MIPS/NSSTC

ARMOR

KHTX

75

DD lobe

1 km Res.

1.5 km Res.

LMA 100-500 m

LMA Antenna

NEXRAD

ARMOR

MIPS Profiler

MAX ?

MAX

Outline• Objectives for TVA-UAH interaction• Radar QPE Problem, Dual-pol solution

• Brief overview of dual-pol• How do we improve QPE

• UAH/NSSTC infrastructure• Simple example• data processing/flow• Rainfall algorithm

• Where we are: Example products• The future• Appendix: BREAM

24 Hour Rain Totals July 6, 2007

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Rain Gauge (inches)

Rad

ar (

inch

es)

• 88D Dual-pol upgrade imminent- improved QPE a primary driver.

• Can dual-pol QPE replace significant % of TVA gauge network?

• Demonstration project with ARMOR in advance of WSR-88D dual-pol upgrade

• Dual-pol rain rate estimator, NO gauge input

• 6-24 hour QPE over basin scales

• Real time web-products

• Facilitate/reintroduce radar QPE tailored to TVA needs for use in river management

• Future customer specific extensions (e.g., NOAA/NWS QPE/F products, site specific terrain corrections etc.

• HWT/COMET-NWS Synergies: pass products direct to WFO HUN- test utility, development

E.g., Warm season precipitation event Favorable comparison to gauges-

BUT much of the heaviest precip missed gauges (typical)!!!

Heterogeneity of rain field presents problems for gauge-adjusted Z-R totals but not for dual-pol

Dual-Pol QPE Applied to Operational Hydrology

Walter A. Petersen, University of Alabama Huntsville

QPE: Problem with conventional radar-rainfall approaches: Reflectivity Factor (Z) - Rainfall Rate (R) Relations

Sample of current operational relationships:Z = 300 R1.4 - convective rain Z = 250 R1.2 - tropical rain

Z = 200 R1.6 - summer stratiform rainZ = 130 R2.0 - winter stratiform (eastern US)Z = 75 R2.0 - winter stratiform (western US)

• NEXRAD measures rainfall using one variable- Z – at single polarization (H)

• Problem dates from 1940’s: Numerous rainfall-reflectivity relationships, which one is correct?

• Errors of 100-200% are common. Why?

• Measurement is sensitive to rain drop size distribution, presence of hail/ice/snow, and radar calibration.

• Unacceptable errors for high resolution hydrological application (e.g., flash flood nowcasting, runoff modeling)

• Even gauge corrections are still beholden to gauge calibration/error/sample mismatch- a problem at times.

Z (dBZ)

R (

mm

/hr)

Z (dBZ) vs R(mm/hr)

50 vs 100 mm/hr at 50 dBZ over a valid range of observed DSDs!!

0

100

50

Proprietary information, Walter A. Petersen, University of Alabama Huntsville

Polarimetric Radar Variables1. Reflectivity factor Z at horizontal polarization

- Measure of drop size and concentration;

• most sensitive to SIZE (D6)

2. Differential reflectivity ZDR

- Measure of median drop diameter→ SIZE/SHAPE

- Useful for rain / hail / snow discrimination→ SIZE/SHAPE

3. Differential phase ΦDP (Specific Differential Phase- KDP)

- Measure of content and size→ NUMBER/SHAPE

- Immune to radar miscalibration, attenuation, and partial beam blockage

4. Copolar-correlation coefficient ρhv

- Indicator of mixed precipitation → SHAPE/PHASE/CANTING (Depolarization)

- Useful for identifying non-meteorological scatterers

U.S. Research

NCAR

NSSL

CSU

NASA

UND

NOAA ESRL

UMASS

UAH ARMOR

Operational:

NEXRAD, TV

Advantages: Obtain a better description of particle types and shapes in a given volume of space

• More accurate rain rates (improved QPE)

• Hydrometeor ID and non-meteorological scatterers

• Consistent calibration

U.S. BroadcastHuntsvilleNew YorkHoustonChicagoTampa

Dual-Pol: Improved Quantitative Precipitation Estimation (QPE) and Hydrometeor Identification (HID)

Radar “sees”

Tumbling and lower dielectric strength makes hail look like a spheres

Unless they start to melt…

Hail/Graupel

Melting Hail/Graupel(Toroid or ice core;

looks like a huge drop)

a

b

1 mm

6 mm

Axis ratio ~ 1

Axis ratio < 1

RainSmall Drops (1 mm)

Large Drops (> 4 mm)

Axis ratio decreases with size- more oblate

Particle-Size Controlled

Smaller ZDR Larger ZDR

Smaller KDP Larger KDP

vs

vs

InsectsRain

vs

Hail/Graupel Rain

Small Drops Large Drops

Sm

aller Z

Larg

er Z

Small Drops

vs

Large Drops

Smaller KDP Larger KDP

Larger ZDR Smaller ZDR

Smaller # Larger #

Number Concentration Controlled

Microphysics

Dual-Pol Interpretation

Proprietary information, Walter A. Petersen, University of Alabama Huntsville

How do we get to “improved QPE”: UAH QPE Infrastructure

Walter A. Petersen, University of Alabama Huntsville

MAX: Mobile X-band Dual-Polarimetric Radar

MAX OPS:

• Adaptive rapid sector/full volumes, RHI’s, vertically pointing

• Mobile targeted QPE studies; severe wx

Raingauges and Disdrometers

• 2 Parsivel optical disdrometers, 2D Video Disdrometer (CSU)

• Geonor rain gauge, three tipping bucket, 1 WXT-510 impact gauge/disdrometer, CHARM rain gauge network, TVA gauge network

ARMOR: C-band Dual-Polarimetric Radar• Variables: Z, V, W, ZDR, DP,KDP, hv,

LDR, HID

ARMOR Standard Ops (24/7):

• 3-Tilt dual-pol scan every 5 min.

• 1-Tilt surveillance every 2.5 min.

Significant Weather/Research Ops:

• Adaptive rapid sector/full volumes, RHI’s, vertically pointing

MIPS: Mobile Integrated Profiling System

• 24/7 Ops

• Vertically pointing, adaptive vertical resolution, DSD, wind, temperature, pressure, humidity profiles

HID Z hv

ZDR KDP DP

Drizzle

Lt. Rain

Mod. Rain

Heavy Rain

Hail

Hail/Rain

Small Hail

Rain/Sm. Hail

Dry Snow

Wet Snow

Cloud

Ice Crys.

How do we get to “improved QPE”? ARMOR PPI at 19:38 UTC

Use of raw variables takes more work for less experienced..........

Combined polarimetric variables offer a powerful means to discriminate liquid from frozen precipitation: Improved QPE, land surface hydrology, warning decision support.

Proprietary information, Walter A. Petersen, University of Alabama Huntsville

How do we get to “improved QPE”

UAH ARMOR Polarimetric Data Processing

Walter A. Petersen, University of Alabama Huntsville

NSSTC/UAH Raw IRIS files (w/Vaisala HCLASS)

IRIS Images

Clean using Z, hv, var (DP)

Correct Attenuation (Z) Z and ZDR = KDPDifferential Attenuation (ZDR)

Filter DP, recompute KDP (adaptive FIR filtering)

Compute Hybrid rain rates (R[KDP,ZDR,Z])

Write UF

Compute HID Accumulated Rainfall

Imag

es/T

able

s

Optional DD

Level II (sweeps)WFO HUN

(AWIPS, GR)UF

TVA

Net

CD

F

ARMOR

T1

ICE PRESENT?

NO

YES

KDP 0.3 and ZH 35? R = R(KDP)YES

NO

ZH BAD? YE

S

R = R(ZHRAIN)

R=BADNO

KDP 0.3, ZH 35.0 dBZZDR 0.5 dB?

YES

R > 50 mm/hr, dBZ > 50 ,or Z, ZDR corr. too large ?

ZH > 30 dBZ, ZDR 0.5 dB?

R = R(ZH,ZDR)

R = R(ZH)

ARMORRAIN RATE

ALGORITHM

(1) R(KDP,ZDR)(2) R(KDP)(3) R(ZH,ZDR)

R = R(ZH)GOOD DATA? YES

NO

R=BAD

KDP ≥ 0.5?

KDP< 0.5?

YES

R = R(KDP)

YES

R =R(KDP,ZDR)

YES

R =R(ZH,ZDR)

no

no

NO

YES

NO

UAH Rainfall algorithm

Proprietary information, Walter A. Petersen, University of Alabama Huntsville

1-hrAccumulation

6-hr (N-hr)Accumulation

Product Access: http://www.nsstc.uah.edu/ARMOR/webimage

Products

• All dual-pol variables for first 3 sweep elevations

• Hydrometeor types (fuzzy and table based)

• Rainfall 1-hour (image)

• Rainfall 6-hour (image)

• 6-hour TVA basin rain statistics (text)

Operational

• Current image

• Last 10 image and 3-hour loops

• Scan comparisons between variables

• Automatic updating

UAH ARMOR RADAR 10/16/2007 (PM): 6-Hour Rainfall Accumulation Products

6-Hour Rainfall Accumulation Image Product

• Centered on ARMOR radar in Huntsville

• TVA Basins and 25 km range rings indicated with white contours.

• TVA gauge locations indicated as points

• Numeric table summarizing basin mean rainfall statistics (area mean, maximum, minimum and standard deviation of 1 km pixels in each basin).

• % RPxl = Percentage of basin area covered by > 0.005 inches of accumulated rainfall in the past 6-hours

http://www.nsstc.uah.edu/ARMOR/webimage/

Tennessee River Basins

MS

AL

GA

NC

KY

TN

VA

SC

(Unregulated area between Kentucky dam and mouth of Tennessee river, 710 Square Miles.)

System refinements (more products, delivery methods, verification statistics etc.)

Correction scheme for 88D (dual-pol tuning)

Test terrain following rain-map with BREAM* database (*next talk)

Topography: A problem area for radar

i.Leave gauges in steep terrain for now

ii.Target work (MAX/MIPS, other leveraged opportunities) site specific 88D corrections over terrain?

100 km

Future TVA Work?

ROCKCASTLE AT BILLOWS 0.713 in. 0.027 in. 0.059 in. 0.367 in. 0.042 in. 0.000 in.

WOLF CREEK LOCAL 0.504 in. 0.003 in. 0.069 in. 0.141 in. 0.029 in. 0.003 in.

MOUTH OF OH TO BARKLEY DAM 0.000 in. 0.000 in. 0.000 in. 0.000 in. 0.000 in. 0.000 in.

LAUREL R @ MUNICIPAL DAM @ CORBIN 0.255 in. 0.011 in. 0.035 in. 0.304 in. 0.068 in. 0.002 in.

DOVER TO BARKLEY 0.000 in. 0.000 in. 0.003 in. 0.002 in. 0.000 in. 0.000 in.

MOUTH OF TN TO KY DAM 0.000 in. 0.000 in. 0.000 in. 0.007 in. 0.000 in. 0.000 in.

CUMB AB WILLIAMSBURG 0.074 in. 0.071 in. 0.090 in. 0.306 in. 0.111 in. 0.089 in.

NORTH FORK HOLSTON RIVER NEAR GATE CITY 0.003 in. 0.873 in. 0.000 in. 0.774 in. 0.557 in. 0.603 in.

CADIZ TO MOUTH 0.000 in. 0.000 in. 0.000 in. 0.000 in. 0.000 in. 0.000 in.

CLINCH RIVER ABOVE TAZEWELL 0.021 in. 0.801 in. 0.026 in. 0.422 in. 0.331 in. 0.527 in.

. . . . . . .

. . . . . . .

Future: Test Utility of High-Resolution SPC WRF BASIN SPECIFIC QPF?