1

WSR-88D Radar Rainfall EstimationPresent and Near Future

Daniel S. Berkowitz

Applications BranchNWS Radar Operations Center

Norman, Oklahoma

2

Paradigms of the Past

• One R(Z) relationship (a.k.a. Z-R or Z/R) could be used for the entire coverage (currently a 230 km range)– Conventional radar (WSR-57 and WSR-74 radars) used “stratiform”

(i.e., Marshall-Palmer Z=200R1.6 derived in the late 1940s) or “convective” R(Z) (Z=55R1.6) in the 1960s to the 1990s.

– WSR-88D has used a default R(Z) (the “Miami Z-R relation” derived from studies in Florida in the mid- to late-1950s) from mid-1980s legacy Precipitation Processing Subsystem (PPS) development to the present.

– WSR-88Ds were authorized additional relationships:• “Rosenfeld Tropical” (Z=250R1.2) in 1997 • Relative to the Continental Divide, “East-Cool Stratiform” (Z=130R2.0) and

“West-Cool Stratiform” (Z=75R2.0) in 1999

3

Paradigms of the Past (continued)

• A “bias correction” (a.k.a. “multiplicative bias”) can be applied for the entire radar coverage.– The multiplicative bias ratio is computed from all available gauge-radar

pairs as follows:Mean Field Bias = (∑Gi)/(∑Ri).

– Gauge data are ingested in an external system (e.g., Multisensor Precipitation Estimator, MPE), where this bias is calculated.

– The spatial distribution of the gauge-radar pairs is considered irrelevant. – The computed bias table is sent from AWIPS to the relevant RPG, where,

based on Hydromet Adjustment adaptable parameters (minimum number of pairs and whether or not to apply this “correction”), the appropriate correction factor is selected from the table.

4

Paradigms of the Past (continued)

• Hybrid scan construction can mitigate contamination from residual ground clutter (not already removed by the RDA) and can compensate for blockages.– Vertical resolution is dependent upon the VCP being used. Some

legacy VCPs (especially 31, 32, 21, and 11) had rather poor vertical resolution at low elevation angles.

– Blockages were estimated from surface elevation data based on (for sites <60° N. latitude) the February 2000 Shuttle Radar Topography Mission (SRTM) with 1 arc-second (about 30 m.) horizontal resolution & ≤16 m. vertical error. Tree growth and man-made construction since then has made the data for many sites out of date.

– The time required to build the Hybrid Scan array may leave gaps when convective cells move quickly.

Elevation Angles in a “Precipitation Mode”

Volume Coverage Pattern (VCP)

0.5°

1.5°

19.5° }several other elevation angles

Mountains cause beam blockages(as do trees, buildings, etc.)

6

Hybrid Scan Illustration

0.5°

0.9°

1.3°1.8°2.4°

Object being excluded from the hybrid scan

Terrain-based Hybrid Scan (NE cross-section at Eureka, CA)

Note the height of the radar beam samples relative to the surface elevation.

8

Accumulation Period vs. Volume Scan Duration

Volume scan

Hybrid scan

Accumulation period

Top of clock hour during antenna retrace timeAvg. time of

hybrid scan

18:01The accumulation period is measured between the average times of the hybrid scan (which may range from a single 0.5° elevation angle to >7° of elevation based on blockages).

9

Digital Terrain Elevation DataRadar Echo Classifier Anomalous Propagation Detection Algorithm (REC-APDA)

Enhanced Precipitation Preprocessor (EPRE)

Hybrid Scan (reflectivity)

Reflectivity

Radial Velocity

Spectrum WidthBeam Blockage Algorithm

Clutter Likelihood

Precipitation Rate Algorithm

Precipitation Accumulation AlgorithmRain Gauge

Reports

Gauge/RadarBias Table

Precipitation Adjustment

(Rainfall) Precipitation Products(OHP, THP, STP, DPA, USP, etc.)

Snow AccumulationAlgorithm

DHR for Flash Flood Monitoring & Prediction (FFMP)

Components of the Precipitation Processing Subsystem

Dissemination

10

Dual Polarization QPE Relationships

• Echo classification-based relationships (next slide)• “Tropical” R(Z,ZDR) from Bringi & Chandrasekar (2001)

• “Continental” R(Z,ZDR) from Ryzhkov

• R(KDP) from Ryzhkov• R(Z) from “default convective” legacy PPS Hydromet

Rate algorithm (the “Miami Z-R relation”) • Melting layer determined from ZDR at 4°-10° elevation

angles• Hybrid Scan construction from Hydrometeor

Classifications

Hydro Class, Melting Layer, &

Dual Pol variables

HydrometeorClassification

Algorithm Quantitative Precipitation

Estimation and other dual pol

products

Super-resolution Products

MeltingLayer

DetectionAlgorithm

Super-resolution

Data (0.5°x0.25 km)

Legacy Algorithms

Recombined (Legacy

Resolution) Data

(1.0°x1.00 km for reflectivity and 1.0°x0.25 km for

Doppler)

RDA (Radar Data Acquisition)

RPG (Radar Product

Generator)

Recombined (Dual

Polarization Resolution)

Data (1.0°x0.25 km)

Legacy Products

Environmental

data

Blockage data

GC, BI, BD, RA, HR

GC, BI, WS, GR, BD, RA, HR

GC, BI, DS, WS, GR, BD

GC, DS, GR, WS, IC, BD

Note: HA can occur at all ranges/heights.

DS, IC, GR

Beam bottomBeam top

BD = big dropsBI = biologicalDS = dry snowGC = ground clutterGR = graupelHA = hail/rain mixtureHR = heavy rainIC = ice crystalsRA = light/moderate rainWS = wet snow

(0° C.)

Beam centerline

Conditions/Classifications R method (mm/hr)

Ground Clutter (GC) or Unknown (UK) Not computed

No Echo (ND) or Biological (BI) 0

Light/Moderate Rain (RA) or Big Drops (BD) R(Z, Zdr)

Heavy Rain (HR) and blockage < 20% and ≤ 45 dBZ R(Z, Zdr)

Heavy Rain (HR) and blockage ≥ 20% or Z > 45 dBZ R(Kdp)

Rain/Hail (HA) and blockage ≥ 5% R(Kdp)

Rain/Hail (HA) and echo is at or below the top of the melting layer (ML) and blockage < 5%

R(Kdp)

Rain/Hail (HA) and echo is above the top of the ML and blockage < 5%

0.8*R(Z)

Graupel (GR) 0.8*R(Z)

Wet Snow (WS) 0.6*R(Z)

Dry Snow (DS) and echo is at or below the top of the ML R(Z)

Dry Snow (DS) and echo is above the top of the ML 2.8*R(Z) !Ice Crystals (IC) 2.8*R(Z)

14

Paradigms of the Past (continued)

• Single radar rainfall estimation is adequate for all hydrologic purposes. Ignore these uncertainties:– Beam propagation path (considered to be determined

from “standard atmosphere”) varies. – Beam typically overshoots the ground at long ranges. – Wind shear displaces radar volume with surface location.– Echoing volume is often not filled with uniform drop size

distribution (DSD) nor with same precip. type.– Evaporation may occur near the ground.– Rain gauge reports are “ground truth”!

15

Near Future• Use of MetSignal to mitigate wind turbine clutter, ground traffic

returns, and radar frequency interference• Simplification of adaptable parameter selection• Improved guidance for parameter selection from Multi-Radar Multi-

Sensor (MRMS) and from 5 years of WSR-88D dual pol. experience, e.g., study to reduce multiplier for rainfall estimate from dry snow, currently having default of R=2.8*R(Z) using MRMS – NSSL DEV system.

• New VCPs with improved vertical resolution • Improved ground clutter suppression of base data• Continuous improvements of MRMS QPE by NSSL, including

– blockage compensation based on recent history of data– mitigation of melting layer effects– use of vertical profile information and High Resolution Rapid Refresh model

data to determine precip. type

16

17

• General Surveillance• Same elevations as VCP 12/212 below 10°• Same vertical coverage as VCP 11/211

above 10°• Same data quality as VCP 21• Duration: ~6 mins• AVSET & 1 SAILS cut

VCP 215

18

19

• Clear-Air• More angles than 31/32• Based on low-level elevations of VCP 12/212• Duration: ~7 mins• May eventually replace VCP 32• 1 SAILS cut

VCP 35

20

“Distant” Future

• Better utilization of new VCPs and MESO-SAILS cuts• Virtual volume scans - constant “voxel” (volume

element) updating, preferably via MRMS mosaics• Use of vertical profiles of radar variables• Increased use of model data for improved

microphysics (better hydrometeor classification and physical environment)

• Probabilistic (rather than deterministic) QPE• More use of MRMS, where applicable, instead of RPG

algorithms

21

Fig. 16 from Miller, D. A., S. Wu, and D. Kitzmiller, 2013: Spatial and temporal resolution considerations in evaluating and utilizing radar quantitative precipitation estimates. J. Operational Meteor., 1(15), 168–184.

Dealing With Uncertainty

“As we know, there are known knowns. There are things we know we know.

We also know there are known unknowns. That is to say, we know there are some things we do not know.

But there are also unknown unknowns, the ones we don’t know we don’t know.”

-- Donald Rumsfeld (2004)

Questions?