ERAD 2014 - THE EIGHTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY

ERAD 2014 Abstract ID 7.5 1 schuur@ou.edu

1Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, Oklahoma, USA 2NOAA/OAR/National Severe Storms Laboratory, Norman, Oklahoma, USA

1 Introduction

Precipitation types common to transitional winter events, such as freezing rain and sleet, typically result from thermodynamic and microphysical processes that occur in very shallow layers above the surface. Because of this, it is not uncommon for the lowest elevation scan from the radar to overshoot transitional winter weather precipitation types at ranges as close as even a few tens of km from the radar. The problem, therefore, is not always one of correctly identifying the precipitation type in the radar beam, but rather that of being able to diagnose what is reaching the surface below it, a problem that is compounded at even more distant ranges where the precipitation type sampled by the radar is even less representative of what is reaching the surface. We have therefore adopted a classification approach where thermodynamic output from numerical models is used to create a “background classification” that is then later modified, when warranted, by radar observations.

In this paper, we describe recent modifications to a new, surface-based polarimetric Hydrometeor Classification Algorithm (HCA) that uses thermodynamic output from the High Resolution Rapid Refresh (HRRR) model. An earlier version of this HCA has been discussed by Schuur et al. (2012) and Schuur et al. (2013). Recent modifications include efforts to take what was initially designed to strictly be a surface-based winter weather HCA and modifying it to be an “all-season” precipitation classification algorithm.

2 Background Classification

As noted, because the dominant precipitation type observed at the surface in winter storms often results from thermodynamic and precipitation processes that occur well below the radar’s beam, an accurate background classification of precipitation type is a fundamental component in the development of our algorithm. Schuur et al. (2012) reported on initial work to develop a background classification for this work. Using Tw profiles from the 13 km Rapid Refresh model, they devised a scheme whereby the number of Tw layers found in the vertical profile above each surface location were used to decide whether the “background” precipitation type in transitional winter weather events was rain, snow, freezing rain, ice pellets, or a freezing rain/ice pellets mix. The scheme did not consider the depth of each layer. Reeves et al. (2014) have since conducted a much more thorough investigation in which the model-based background classification scheme of Schuur et al. (2012) was compared to the results of other model-based classification techniques, including those of Ramer (1993), Baldwin et al. (1994), and Bourgouin (2000). More recently, work at NSSL has also explored the potential of creating a background classification using a simple one-dimensional model with spectral (bin) microphysics that explicitly treats the processes of melting / refreezing by taking the initial size distribution and density of dry snowflakes into account. It describes the evolution of mass water fraction and density of hydrometeors as they fall to the ground separately in 80 size bins. Ultimately, testing and a comparison of all background classification scheme results to surface observations from the Meteorological Phenomena Identification Near the Ground (mPING, Elmore et al. 2014) project will be used to identify the optimal background classification for our new surface-based HCA.

3 Algorithm Description

Figure 1 presents a schematic that summarizes the classification procedures of the new HCA. The first step of the process is to run the fuzzy-logic-based classification that is currently deployed on the WSR-88D network (Park et al. 2009). The algorithm then allows fuzzy-logic-based classifications from the lowest elevation sweep to be projected to the surface as snow or ice crystals for cold season events where the entire atmospheric column above a location has T < -5ºC and as rain, big drops, or hail for warm season events where the surface temperature at a location has a T > 5ºC. For intermediate conditions typical of transitional winter weather events, the algorithm uses vertical profiles of model wet bulb temperature profiles to provide a background precipitation classification type. Polarimetric radar observations are then used to either confirm or reject the background classification. For example, if a radar bright band (suggesting an elevated warm layer) is observed immediately above a background classification of dry snow (suggesting the absence of an elevated warm layer in the model output), the background classification is found to be inconsistent with observations and is modified according to a set of empirical rules. The polarimetric radar data are also used to provide further refinement of precipitation type categories when the observations are found to be consistent with the background classification.

Recent modifications to a new surface-based polarimetric Hydrometeor Classification Algorithm for the WSR-88D network

Terry J. Schuur1,2, Alexander V. Ryzhkov1,2, Heather D. Reeves1,2, John Krause1,2,

Kimberly L. Elmore1,2, and Kiel L. Ortega1,2

ER

prsedean(am“hclfu

FiTw

elTw

thevpo

4

cocletusco5

mevsuanbydoco

suwparebalo

ERAD

RAD 2014 Abstract

Since the radresence/abseneparate paper)epend on rangnalyses are usaccurate ML d

more distant rahybrid” ML dlassification puzzy-logic HC

igure 1: Schew through thelevation scan w at the surfac

he ground as events, the fulolarimetric ra

Example

We now presomprehensive lassification ot al. (2014) issing the proceolumn above adB.

The first examore detail by ventually exteubfreezing witn excellent daty the new suromain, and 2onditions (cold

Figure 2 preurface-based a

with the new suart of the radaegion of elevaased algorithmogic-based ML

2014 - THE EI

ID 7.5

dar-based modce of an elev is also introd

ge from the rased to create adetection, butanges, but oftdetection algoprocess is to rCA identified h

ematic showinge entire atmoare projectedce is > 5ºC, teither rain, bigll model-base

adar observati

sent examplestesting to d

f Schuur et als similarly in edures describany surface lo

ample we presSchuur et al. (nding northwth a small regtaset to illustr

rface-based alg2) an exampled-season, war

esents the KPalgorithm at 08urface-based Har domain wheated ML. Them performed bL detection alg

IGHTH EURO

dification of tvated warm laduced. With thadar, horizontaa “blended” m limited to sh

ften contaminaorithm is descrun the hail-sihail.

g the classificospheric columd to the groundthe fuzzy-logicg drops, or haed backgrounions (condition

s of the algoridetermine the l. (2012) is usa stage of deed by Schuur

ocation contain

sent is an even(2013). On tard to a point ion on the farrate two featurgorithm) for ce of an eventrm-season, and

BZ results of802 UTC on 2HCA providinere both the H

e reason the fubetter) is that gorithm (whic

PEAN CONFE

the underlyingayer, a new hhis new meltinal gradients in

map of ML dehort ranges froated by effeccribed in detaizing algorithm

cation processmn is < -5ºCd as either icec-based precipail (condition 2nd classificatn 3).

ithm output froptimal bac

ed for the exaevelopment an

et al. (2012).ns a radar gate

nt from the Pithis day, and e~10 km to th

r southern edgres: 1) the limconditions wht where the sd transitional w

f the fuzzy-lo21 January 20ng a much moHRRR model uzzy-logic-basthe elevated

ch relies on M

ERENCE ON R

2

g backgroundybrid techniqng layer (MLn model wet-betections by coom the radar)ts of beam b

ail by Schuur m of Ortega

s of the new sC, the fuzzy-loe crystals or spitation type c2). For intermtion is deter

rom several dikground clas

amples providnd testing, the That is, the e with a thres

ttsburgh, PA extensive regie south of the

ge of the radarmitations of thehere an elevatsurface-based winter weathe

ogic-based cla012. The differe realistic traand KPBZ ZD

sed scheme pML always r

ML detections a

RADAR IN ME

classificationque to detect t) detection tecbulb temperatuombining con), low elevatio

broadening), aet al. (2014,

et al. (2013)

surface-basedogic-based prsnow (conditiocategories fromediate condirmined and l

ifferent eventsssification. Fded. Since thee algorithm reradar-based mholds of Z > 2

(KPBZ) on 2on of warm ai

e radar. Surfacr domain withe current fuzzted layer of w

algorithm ther) illustrated

assification foerences betweeansitional wint

DR and HV fieerformed so pemained to that elevation an

TEOROLOGY

n is strongly dthe melting lachnique, Gausures, and time

ntributions froon radar scan

and model su, this conferenand applying

d HCA. For crecipitation tyon 1). For w

om the 0.5º eleitions typical olater modified

s. As noted eFor that reasoe “hybrid” MLesults shown hmodification is25 dBZ, ZDR >

1 January 201ir advanced to

ce air underneah a surface Tw zy-logic basedwarm air coverhat is proposein Figure 1.

or this event aen the 2 classiter weather clelds indicatedpoorly for thishe south of Kngles between

Y AND HYDRO

dependent on ayer (discussessian weightine from the moom high elevans (ML detectggested ML nce). The fin

g it to all loca

cold-season coype categoriearm-season cevation scan aof transitionald, when war

arlier, we are on, the “NSSL detection alghere rely upos conducted w> 0.8, HV < 0

12. This evenowards KPBZath the air waof >0C. This

d HCA (and brs only a ported in this pap

and the resultification schemlassification od the presences event (and t

KPBZ. As a ren 4 and 10º) ne

OLOGY

detection of ted in detail inng functions thost recent modation radar scation available locations. Thnal stage of tations where t

onditions whees from the 0onditions wheare projectedl winter weathrranted, by t

still conductiSL” backgrougorithm (Schu

on ML detectiwhen the vertic0.97, and SNR

nt is discussedZ from the sou

s predominans event providenefits providtion of the radper meets all

ts from the nemes is dramatver the southe

e of an extensithe new surfacesult, the fuzzever identified

the n a hat del ans

to his the the

ere 0.5º ere

d to her the

ing und uur ion cal

R >

d in uth, ntly des ded dar l 3

ew tic, ern ive ce-zy-d a

ERAD

RAD 2014 Abstract

ML layer in theeep column ofwith small poco the south of this classificati

igure 2: Preclevation, and (lutter (GC), beezing rain/icnd giant hail (

Figure 3 shoonditions depiolumn, warm-f the radar dom

igure 3: Clasanuary 2012. eason” conditocations wheronditions wheonditions typic

2014 - THE EI

ID 7.5

e near vicinityf cold air. Asckets of graupthe radar that ion is far mor

cipitation type(right panel) Wbiological scace pellets (FR/(GH).

ows the same icted in Figur-season with Tmain, with the

ssification con Condition 1

tions when thre all ice caten the surface cal of transitio

IGHTH EURO

y to the radar s a result, precpel in region o

consisted of ee consistent w

e classificatioWsHCA at 08 atterers (BS), /IP), freezing

grid as in Fire 1 were metTw at the surfae transition be

ndition follow1 (green) showhe entire atmotegories are p

temperature onal winter w

PEAN CONFE

and thereby ecipitation clas

of high Z). Onextensive regi

with both the H

n centered onUTC on 21 Jacrystals (CR

rain (FR), big

igure 2, but ilt. As can beace > 5ºC, andtween the con

wed in the decws locations wospheric coluprojected to tat a location eather are me

ERENCE ON R

3

erroneously asssified over thn the other haions rain, freeHRRR model

n the KPBZ raJanuary 2012. R), dry snow g drops (BD),

llustrating thee seen, all 3 cd transitional nditions provid

cision tree in where snow a

umn above ththe surface ais > 5ºC, and

et and the full

(

RADAR IN ME

ssumed that thhe entire domaand, the surfaczing rain, ice output and rad

adar from the Precipitation(DS), wet sn rain (RA), he

e regions overconditions (cowinter weatheding a seamle

Fig. 2 centerand ice crystaat location h

as either raind condition 3 WsHCA is ru

TEOROLOGY

he entire radarain was classice-based algorpellets, and a dar observatio

(left panel) fun type categornow (WS), graeavy rain (HR

r the radar doold-season witer) were met ass transition.

red on the KPals are projec

has T < -5ºC,n, big drops,

(red) shows n to determin

Y AND HYDRO

r domain was ified as snow rithm providefreezing rain

ons to the sout

fuzzy-logic-basries are no echaupel (GR), i

R), hail (HA),

omain at whicth Tw < -5ºCat this time ov

PBZ radar atcted to the su, condition 2 or hail for locations whee surface prec

OLOGY

located withinand ice crystd classificatio/ ice pellet m

th of the radar

sed HCA at 0ho (NE), grouice pellets (IPlarge hail (LH

ch each of theC over the entver some porti

t 08 UTC on urface for “co

(yellow) sho“warm seasoere intermediacipitation type

n a als

ons mix.

r.

0.5º und P), H),

e 3 tire ion

21 old ws n” ate e.

ER

ettaprsu

10NanraH

thtraob

Fi(Uas

ThUfogea bth

ERAD

RAD 2014 Abstract

Shortly after t al. 2014) thaagged observarecipitation tyurface observa

On 8 Februa00 cm in dept

National Weathnd HV fields, adar-indicated

HV and surfacehe north. Anansition regiobservations ov

igure 4: RadaUpton, NY) rads that of the su

he fuzzy-logicUnlike for the Kor this event. Heometrically pbroad “bull’s

hat the surface

2014 - THE EI

ID 7.5

r the above eveat provides thations of preciype observatioations overlaid

ary 2013, the Nth. Figure 4her Service Oit is clear thatML extendin

e-based HCA n analysis of on slowly prover much of L

ar reflectivity (dar at 1504 Uurface HCA.

c-based HCA KPBZ event wHowever, sinc

projected outweye” of rain th

e HCA provide

IGHTH EURO

ent, our groupe general pubipitation type.ons that have d.

Northeastern U4, at 1504 UTffice, located t a widespread

ng over Long Ifields at this tthis same eve

ogressed to thLong Island.

(Z), DifferentiUTC on 8 Febr

results at thiswhere no ML wce the algorith

ward along eachat was centees much bette

PEAN CONFE

p developed anblic an opport. In the 2 wigreatly aided

U.S. experienTC on 8 Febru

on Long Island region layerIsland Sound)time show goent at a later he south over

ial reflectivityruary 2013. O

s time are showwas observedhm then assumch azimuth, thred on the radr agreement w

ERENCE ON R

4

nd released antunity to partiinter seasons d our research

nced a historicuary 2013, shond at Upton, Nr of warm air) at this time. od overall agrtime by Schuthe next sev

y (ZDR), CorreOverlaid mPIN

wn, and comp above the rad

mes that the Mhe resultant fuzdar site and exwith both the r

RADAR IN ME

n iPhone and Acipate in our since the app

h. We now p

c winter stormows Z, ZDR, aNY) WSR-88to the north sh Precipitationreement, with uur et al. (201veral hours, ev

elation coefficNG surface ob

pared with the dar site, the fu

ML detected inzzy-logic HCA

xtended well inradar and mPI

TEOROLOGY

Adroid app (rresearch by ps release, we

present results

m that resulted and HV from8D radar. Frohore of Long

n type reports rain over mu

14, this confeventually resu

cient, and surfbservations fo

surface baseduzzy-logic-basn regions closeA (mapped tonto ConnecticING observati

Y AND HYDRO

referred to as mproviding timhave collecte

s from 2 even

in snow accum the KOKX (om an examin

Island (with aoverlaid on toch of long isla

erence) demonulting in snow

face HCA for llow the same

d HCA resultssed algorithm e to the radar co the conical sucut to the northions.

OLOGY

mPING, Elmoe-stamped, geed over 600,0nts with mPIN

umulations up(New York Cation of the Za few pockets op of the Z, ZD

and and snownstrated that tw and ice pel

the KOKX e color scheme

s, in Figure 5. identified a Mcan then be urface) showeh. It is clear

ore eo-000 NG

to City ZDR

of DR,

w to the llet

e

ML

ed

ER

FiO

Oa raapobwclcowte

ERAD

RAD 2014 Abstract

igure 5: FuzzyOverlaid mPIN

Finally, we pffice, located transitional w

ain / ice pellet ppears to overbservations as

wet snow catelassification, iomplete our co

was presented esting.

2014 - THE EI

ID 7.5

y-logic-based NG surface obs

present Figurein Sterling, V

wither weathermix and wet

r predict the fs very little fregory with nindicating weomparison anwith these re

IGHTH EURO

HCA and surfservations foll

e 6, which shoVA) WSR-88Dr event over a snow over thefreezing rain reezing rain onumerous mPet snow. As nd testing of vesults will be

PEAN CONFE

rface HCA forlow the same

ows Z, ZDR, aD radar at 120

major metrope DC area. A/ ice pellet ca

or ice pellets wPING reportsnoted earlier,arious scheme

e replaced wit

ERENCE ON R

5

r the KOKX (Ucolor scheme

and HV from t05 UTC on 13politan area, w

According to Rategory. Thiswere reporteds agreeing w, we expect ies. Ultimatelyth the “optim

RADAR IN ME

Upton, NY) radas that of the

the KLWX (W February 201

with the surfacReeves et al. (2s appears to bd. There was,with the radarimprovement y, the “NSSL

mal” backgrou

TEOROLOGY

dar at 1504 Usurface HCA

Washington, D14. As with thce HCA show2014), the NS

be apparent wh, however, verr-based modiin the backg” classificatio

und classificat

Y AND HYDRO

UTC on 8 Febr.

DC National Whe above exam

wing broad regSSL backgrounhen comparedry good consiification to tround classifi

on of Schuur etion as determ

OLOGY

ruary 2013.

Weather Servimple, this showgions of freezind classificatid to the mPINistency with tthe backgrou

fication after wet al. (2012) thmined from o

ice ws ing ion NG the

und we hat our

ER

Fi(Ssc

Su

ouloattealtybaaninthcrdera

WsuclScthpeen

ERAD

RAD 2014 Abstract

igure 6: RadaSterling, VA) rcheme as that

ummary and

In thisutput from thowest elevatiotmospheric coemperature at lgorithm uses ype. Polarimeased modifican elevated wantroduced. Wihe radar, horizreate a “blendetection, but langes, but ofte

While new elemurface for truelassification anchuur et al. (2

hese “new anderiod of extenntire continent

2014 - THE EI

ID 7.5

ar reflectivity (radar at 1205 of the surface

Future Wor

s paper, we dee High Resol

on sweep to bolumn above a

a location havertical profitric radar obse

ation of the unarm layer, a neith this new mzontal gradiended” map of limited to shoen contaminate

ments have be cold- and wand threshold b

2014), the bacd improved” consive testing thtal U.S.

IGHTH EURO

(Z), DifferentiUTC on 13 F

e HCA.

rk

escribe recent lution Rapid Rbe projected a location has as a T > 5ºCiles of model ervations are t

nderlying backew hybrid tec

melting layer (nts in model wf ML detectioort ranges froed by effects o

been introducearm-season cobased ML det

ckground clasomponents ofhat will fully

PEAN CONFE

ial reflectivityFebruary 2014

modificationsRefresh modeto the surfacT < -5ºC and

. For intermwet bulb temthen used to e

kground classichnique to det(ML) detectionwet-bulb tempons by combiom the radar),of beam broad

ed to the algoonditions, the tection detailesification and

f the algorithmutilize both p

ERENCE ON R

6

y (ZDR), Corre4. Overlaid m

s to a new, suel. The algorce as snow ord as rain, big dmediate condi

mperature profieither confirm ification is strotect the meltinn technique, Gperatures, andining contribu, low elevatiodening), and m

orithm, such aresults presened by Schuur

d ML detectionm are introducolarimetric W

RADAR IN ME

elation coefficPING surface

urface-based prithm allows r ice crystalsdrops, or hail itions typical files to provid

or reject the bongly dependeng layer (discGaussian weigd time from thutions from h

on radar scansmodel suggest

as the projectinted in this pap

et al. (2012).n algorithm aed into the alg

WSR-88D rada

TEOROLOGY

cient, and surfe observations

polarimetric Hfuzzy-logic-b for cold seafor warm seaof transitiona

de a backgrounbackground clent on detecti

cussed in detaghting functiohe most recenhigh elevations (ML detectited ML locatio

ion of fuzzy-lper still rely u. As outlined

are still beinggorithm in thear data and mP

Y AND HYDRO

face HCA for s follow the sa

HCA that uses based classificason events wason events wal winter weand precipitatiolassification. on of the pres

ail in a separatons that depennt model analyn radar scansion available ons.

logic algorithupon the “NSSd by Reeves edeveloped an

e coming yearPING observa

OLOGY

the KLWX ame color

thermodynamcations from twhere the ent

where the surfaather events, ton classificatiSince the rad

sence/absencete paper) is ald on range froyses are used s (accurate Mto more dista

m results to tSL” backgrouet al. (2014) and tested. Whr, we will enteations across t

mic the tire ace the ion ar- of lso om

to ML ant

the und and hen er a the

ERAD 2014 - THE EIGHTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY

ERAD 2014 Abstract ID 7.5 7

Acknowledgement

Funding was provided by NOAA/Office of Oceanic and Atmospheric Research under NOAA-University of Oklahoma Cooperative Agreement #NA11OAR4320072, U.S. Department of Commerce, and by the U.S. National Weather Service, Federal Aviation Administration, and Department of Defense program for modernization of NEXRAD radars.

References

Baldwin, M., R. Treadon, and S. Contorno, 1994: Precipitation type prediction using a decision tree approach with NMCs mesoscale eta model. Preprints, 10th Conf. on Numerical Weather Prediction, Portland, OR, Amer. Meteor. Soc., 30-31

Bourgouin, P, 2000: A method to determine precipitation type. Wea. Forecasting, 15, 583-592.

Elmore, K. L., Z. L. Flamig, V. Lakshmanan, B. T. Kaney, H. D. Reeves, V. Farmer, and L. P. Rothfusz, 2014: mPING: Crowd-sourcing weather reports for research. Bull. Amer. Meteor. Soc., in press.

Ortega, K. L., A. V. Ryzhkov, J. Krause, P. Zhang, and M. R. Kumjian, 2013: Evaluating a Hail Size Discrimination Algorithm for dual-polarized WSR-88Ds using high-resolution reports and forecaster feedback. 36th Conference on Radar Meteorology, Breckenridge, CO, American Meteorological Society, Boston, 7B.4.

Park, H.-S., A. V. Ryzhkov, D. S. Zrnic, and K-.E. Kim, 2009: The hydrometeor classification algorithm for the polarimetric WSR-88D: Description and application to an MCS. Wea. Forecasting. 24, 730-748.

Ramer, J., 1993: An empirical technique for diagnosing precipitation type from model output. Preprints, 5th Int. Conf. on Aviation Weather Systems, Vienna, VA, Amer. Meteor. Soc., 227-230.

Reeves, H., K. Elmore, A. Ryzhkov, T. Schuur, and J. Krause, 2014: Sources of uncertainty in precipitation type forecasting. Wea. Forecasting, in press.

Schuur, T. J., Park, H.- S., A. V. Ryzhkov, and H. D. Reeves, 2012: Classification of precipitation types during transitional winter weather using the RUC model and polarimetric radar retrievals. J. Appl. Meteor. Climate, 51, 763-779. DOI: 10.1175/JAMC-D-11-091.1.

Schuur, T. J., A. V. Ryzhkov, H. D. Reeves, M. R. Kumjian, K. L. Elmore, and K. L. Ortega, 2013: A new surface-based polarimetric Hydrometeor Classification Algorithm for transitional winter weather to be used on the WSR-88D network. 36th Conference on Radar Meteorology, Breckenridge, CO, American Meteorological Society, Boston, 7B.3.

Schuur, T. J., A. V. Ryzhkov, and J. Krause, 2014: A new melting layer detection algorithm that combines polarimetric radar-based detection with thermodynamic output from numerical models, 8th European Conference on Radar in Meteorology and Hydrology, Garmisch-Partenkirchen, Germany, 7NOW.P05.