1 of 21

Probability of Detection of SPLASH* Using Polarimetric Radar

Aaron Ward¹, Matthew Kumjian², Karly J. Reimel², Alex P. Ferguson¹, Stephen Bieda III¹, Matthew Bunkers³ and B.J. Simpson¹

1. NOAA/National Weather Service, Amarillo, Texas

2. Department of Meteorology and Atmospheric Science, The Pennsylvania State University, University Park, Pennsylvania

3. NOAA/National Weather Service, Rapid City, South Dakota

*SPLASH = Storms Producing Large Amounts of Small Hail

2 of 21

SPLASH Background

Recent spike in interest concerning storms that produce large accumulations of small hail:

Kalina et al. 2016
Ward et al. 2018
Kumjian et al. 2019
Wallace et al. 2019
Friedrich et al. 2019

Case data have been limited when considering SPLASH radar signatures.
More data are needed for statistical analysis.

Photo/Video Credits: Aaron Ward

3 of 21

SPLASH Signature Review

Large Reflectivity (Z)
Large Specific Differential Attenuation (ADP)
Anomalously large Specific Differential Phase (KDP)

Z_H (dBz)

Z_DR (dB)

Phi_DP (°)

K_DP (°/km)

>= 7.5 °/km

Reduced Z_DR values down radial

>= 50 dBZ

Kumjian et al. 2019

Large Diff Phase

4 of 21

Specific Differential Phase (KDP)

Small Melting Hail

Slower

Faster

Large Hail

5 of 21

Melting Hail Microphysics

S-band Scattering calculations From Bringi & Seliga (1977a,b).
Regardless of the particle size distribution selected, smaller melting stones dominate contributions to K_DP and A_DP.

Surface

0C Lvl

-10C Lvl

-30C Lvl

Two-layer spheroids w/max allowable liquid water mass based on Rasmussen & Heymsfield (1987), Ryzhkov et al. (2013a,b), and Kumjian et al. 2018.

From Kumjian et al. 2019

6 of 21

Hypothesis

SPLASH POD using polarimetric signatures will be high near the radar, but may not be detectable at long range (> 40km).
SPLASH POD using polarimetric signatures will be high when the wet bulb zero height (WBZ) is high, but may not be detectable when the WBZ is low (< 2500 m).

7 of 21

Case Selection and Examples

April 11, 2012 – Amarillo, TX

Credit: NWS Amarillo

March 8, 2015 – Jonesboro, TX

March 31, 2016 – Gluckstadt, MS

June 8th, 2017 – Canyon, TX

Credit: NWS Amarillo

May 18, 2015 – Pecos, TX

July 14, 2015 – Rapid City, SD

8 of 21

Methods: Calculating KDP & ADP

Specific Differential Phase (KDP)

Specific Differential Attenuation (ADP)

Used WSR-88D archived radar data.
Only considered the 0.5° slice for simplicity.
Calculated based on Ryzhkov et al. 2005 (method used operationally).
With help of Py-ART, gate filters were used to isolate true max KDP value.

Used WSR-88D archived radar data.
Only considered the 0.5° slice for simplicity.
Calculation based on linear dependence with PhiDP (Ryzhkov et al. 2013).
Attenuation was computed up to WBZ from most valid RAOB.
With help of Py-ART, gate filters were used to isolate true max KDP value.

9 of 21

Methods: Adding Gate Filters

Specific Differential Attenuation(dB/km)

Specific Differential Phase (°/km)

10 of 21

Methods: Moment Based Gate Filter (Z & CC)

Specific Differential Attenuation(dB/km)

Specific Differential Phase (°/km)

pyart.filters.moment_based_gate_filter

11 of 21

Methods: Remove Insignificant Features

Specific Differential Attenuation(dB/km)

Specific Differential Phase (°/km)

pyart.filters.moment_based_gate_filter + echo.correct.noise.significant_detection

12 of 21

Methods: Haversine Formula

Specific Differential Attenuation(dB/km)

Specific Differential Phase (°/km)

pyart.filters.moment_based_gate_filter + echo.correct.noise.significant_detection + Haversine

SPLASH Location =

Box = 5x5 km

5km

SPLASH Location =

Box = 5x5 km

13 of 21

Methods: Radar Range and Wet Bulb Zero Height

In cases where one WSR-88D was within 50km of the SPLASH point, and a second was within 150km, data from both radars were used (true for six cases).

Short Range KDP (KAMA)

Long Range KDP (KLBB)

Wet bulb zero heights from the Plymouth State Weather Center Text Listings were used for the nearest RAOB in space and time.

14 of 21

Results: Max KDP/ADP vs. Radar Range

NWS KDP (°/km) vs. Range (km)

Phi Linear ADP(dB/km) vs. Range (km)

15 of 21

Results: Max KDP/ADP vs. WBZ

NWS KDP (°/km) vs. WBZ (m) (cmap = Gate Height (m))

Phi Linear ADP(dB/km) vs. WBZ (m) (cmap = Gate Height (m))

16 of 21

Final Numbers: SPLASH POD

Criteria:

ADP > 1.0 dB/km

KDP > 7.5 °/km

Results:

KDP all (50): 0.54

ADP all (50): 0.50

KDP r <= 40km (20): 0.80

ADP r <= 40km (20): 0.90

KDP r > 40km (30): 0.37

ADP r > 40km (30): 0.23

KDP WBZ => 2500m (43): 0.58

ADP WBZ => 2500m (43): 0.54

KDP WBZ < 2500m (7): 0.29

ADP WBZ < 2500m (7): 0.18

KDP r < 40km & WBZ => 2500m (17): 0.94

ADP r < 40km & WBZ => 2500m (17): 0.94

KDP r < 40km & WBZ < 2500m (3): 0.00

ADP r < 40km & WBZ < 2500m (3): 0.67

Step through each parameter. For the “all”, we might mention how we had a good distribution of cases representing different environments and difference radar ranges. We the r< 40km, we should note that the signatures are proven to be very reliable to detect SPLASH cases, but the reliability falls off quickly beyond 40km. WBZ also shows some value with high POD for cases where WBZ > 2500 m. It is more difficult to say with WBZ < 2500 m, because there is not a good sample size.

For the 6^th column, 16/17 cases verified for both KDP & ADP (94%). For the 7^th column, KDP was 0/3 and ADP was 2/3.

kdp all: 0.54

50

kdp short: 0.8

20

kdp long: 0.36666666666666664

30

kdp low wb0: 0.2857142857142857

7

kdp high wb0: 0.5813953488372093

43

kdp high rwbz: 0.9411764705882353

17

kdp low rwbz: 0.0

3

adp all: 0.5

50

adp short: 0.9

20

adp long: 0.23333333333333334

30

adp low wb0: 0.2857142857142857

7

adp high wb0: 0.5348837209302325

43

adp high rwbz: 0.9411764705882353

17

adp low rwbz: 0.6666666666666666

3

17 of 21

Potential Use in Decision Support Services Regime

The high POD values using KDP and ADP suggest SPLASH radar signatures provide unique information that can provide operational meteorologists “just in time” information about potentially hazardous accumulations of hail.
This information can be passed on to public safety personel (e.g. Dept. of Transportation for snow plow deployment) to help protect life and property.
A few offices (e.g. KAMA & KUNR) are issuing warnings and advisories to warn the public.

This Photo by Unknown Author is licensed under CC BY-SA-NC

18 of 21

Future Work

SPLASH Environments
SPLASH algorithm
SPLASH in WoFs?

19 of 21

Conclusion

Scatter plots revealed notable trends with range, but the trends with WBZ were not as easy to determine because of differences in gate height. More data is needed to get better R²values.
SPLASH POD was very high (94%) for KDP & ADP when both r < 40 km and WBZ => 2500 m, but POD values drop off at ranges > 40 km (KDP = 37%, ADP = 23%) and WBZ < 2500 m (KDP = 29%, ADP = %18).
POD values suggest polarimetric radar SPLASH signatures can be used for life saving decision support services.
False alarm and lead time statistics are unreliable due to limitations in reporting and verification procedures (or lack thereof).
Verification procedures for SPLASH cases should be employed across the NWS.
More work is needed to better identify the environments that are favorable for SPLASH cases, with utilization of model proximity soundings and additional SPLASH cases.

20 of 21

SPLASH False Alarm & Lead Time

The lack of a dedicated reporting method for SPLASH cases results in unreliable statistics for FAR and lead time.
One goal of this research is to prove that SPLASH cases are worthy of National Weather Service warnings and verification, similar to that of a severe thunderstorm.
Once verification occurs across NWS offices on a regular basis, FAR and lead time statistics will become more reliable.

21 of 21

Haversine Equation for Small Distances