NATIONAL WEATHER DIGEST

Forecasting

A TECHNIQUE FOR FORECASTING HEAVY PRECIPITATION IN THESOUTHEASTERN UNITED STATES

R.J. Soptei and W. K. Henry

Texas A & M universityCollege Station, Texas

ABSTRACT

Flooding has beco.e the nation's leadingweather proble•. There~ore, in an effort toprovide better precipitation input for floodevents, a .ethod of forecasting rainfallusing facsi.ile products to identif~ thesynoptic situations producing heav~ pre-cipitation has been developed. The 1000-500~b thickness is the ke~ para~eter forforecasting heavy precipitation in thesoutheastern United States. Twelve hourforecasting guides are presented. Theseforecasting guides were verified withsubsequent occurrences and found to iMprovethe HOS and OPF guidance.

1. INTRODUCTION

Every year floods cause millions of dollarsof property damage and hundreds of deaths inthe United States. Floods are common to allsections of the country and no area escapesthe devastation they may bring. The need formore accurate forecasts of heavy precipi-tation was stressed in a report by theAmerican Meteorological Society (1) on flashfloods. In addition, Fawcett (2) states thatprecipitation is one of the difficult of themeteorological parameters to forecast. Evenwith the increased number of numericalmodels the skill of the precipitationforecast has not improved over the last 10years. Further research into the presentforecast methods or development of new onesis warranted.

2. WINTER RAINFALL AND FRONTS

Since heavy rainfall often accompaniesfronts a count was made of the frequency offronts in the southeastern United States(which for the purpose of this study in-cludes eastern Texas, Mississippi, Alabama,Loulsiana, Georgia, and Florida) for thewinter seasons (November through March) of1973-1977. The fronts were counted from thesurface analyses of the National Meteor-ological Center (NMC). The results of thiscount are shown in Table 1. A total of ISSfronts entered the southeastern UnitedStates during these five winter seasons. The

8

average duration of a front in the area was3.1 days. The precipitation associated withthese fronts was identified and the fractionof the winter rainfall determined (Table 2).This study shows that for all states andyears at least 60 percent of the totalwinter season precipitation was associatedwith fronts.

Some investigations have been performed inan attempt to find the synoptic patternsaccompanying frontal passages associatedwith significant precipitation. Garcia, etal. (3) found heavy frontal-associatedprecipitation in the Dominican RepUblicoccurring with strong surface east winds inconjunction with strong upper level west-erlies and lower tropospheric ridging behindthe front. Maddox, et al. (4) found that oflSl flooding events occurring in the UnitedStates between 1973-1977 roughly 25 percentwere associated with fronts. Maddox and hisassociates went one step further and cal-culated mean atmospheric conditions accom-panying these frontal-type flooding events.Other research in this direction has beenperformed by Kreitzberg and Brown (5), Smithand Younkin (6), Bosart (7), Estogue (8),and Jensen(9).

3. ANALYSIS

3.1 Data

The precipitation data used in this studywere taken from the National Climatic Center(NCC) daily precipitation records for theindividual states. Data for the synopticanalyses are taken from the NMC facsimilemaps. In addition, Service A and C teletypedata were available. NMC analyses were usedso that the parameters utilized to develop aheavy precipitation forecasting techniquewould be available to all forecasters, andto eliminate bias and increase objectivityon our part. We believe that some frontswere removed from the analyses prematurely,which would have inc[eased the numbers inTables 1 and 2. However, in this study theNMC analysis was used.

3.2 Selection of Cases

The first step was to compute daily

VOLUMB 6. NUMBBR ,Table 1- Frontal occurrences in the southeastern United States during theperiod of study. TOp numbers represent the number of fronts per month,bottom numbers are the total days per month front(s) were analy2ed in thestudy area.

Year Nov D., J.n Fe' M&< Total

1973 , , , , 7 "15 20 28 13 3D 10'1974 , 7 , , , "" 24 26 19 17 1021975 , 7 , 7 • 32

12 " " 23 23 "1976 3 , 7 , , 27

11 15 15 10 " "1977 , • • • • 3•

21 24 20 " 21 104

1SS47S

Table 2. Percentage of wintertime precipitation associated with fronts forthe southeastern United States during the period of study.

State 1973 1974 1975 1976 1977

Louieiana • 3 ., 75 ,. .,Mississippi '0 '0

., .7 "Florida ., 91 74 '0 '4Georgia ., 93 " 73 93Alabama 93 ., ., 7' 91Eaat Texas 75 75 76 62 .,

9

NATIONAL WBATHBR DIGBSr

precipitation averages for each of the NCCsubsQctions within the six states involvedin this study. This was done for every dayof the winter months of 1973 through 1977.These daily precipitation averages were thenplotted on sectional maps and isoplethed at10 mm intervals. Next, using NMC surfaceanalyses, a record was made of each frontalepisode that occurred over the southeasternUnited States. An episode was comprised ofthe time before a front entered the studyarea to the time the front exited. This wasaccomplished by recording the 0000 GMTposition of the front for each day of thefrontal episode on a map. The 0000 GMTfrontal position was selected because itoccurs approximately in the middle of theobserve 24 hour total precipitation used tocalculate the daily precipitation averages.Therefore it was representative of the meanposition of the front during the 24 hourperiOd. In order to check for frontalassociated precipitation, the 0000 GMTfrontal position for a given 24 hour periodwas superimposed over the daily precipi-tation averages for the same period. Allprecipitation within 100 km of the front wasconsidered to be produced by the front. Theheavy precipitation events (the precipi-tation deemed most likely to cause flooding)were identified. A heavy precipitation eventwas identified to be an area of at least 1.5x 10· km! enclosed by the 25 rom isohyet.Each frontal episode was checked for thesecriteria. About 10 percent of the frontscaused heavy precipitation. No heavy pre-cipitation was found that was not associatedwith a front. Five cases were selected from15 frontal episodes which met these cri-teria. These cases included fronts witheast-west and north-south orientations. Inaddition, four cases of frontal passage withlittle or no precipitation were chosen as acontrol. Table 3 is a list and briefdescription of each case.

3.3 Synoptic Analysis

The synoptic situation was examined for eachof the nine cases. Upper air maps from 850to 200 mb inclusive were examined forheight, temperature, wind speed and direc-tion, and where appropriate dewpoint depres-sions were tabulated. Also included in theanalyses were stability indices, freezinglevels, mean surface-SaO mb relativehumidity, 1000-500 mb thickness, verticalvelocities, and vorticity patterns. Surfacemaps were examined for frontal positions inrelation to many of the abOve parameters.Other surface parameters such as wind,temperature, etc. were examined but did notrelate to the rainfall as well as the upperair parameters, thus the study of thesurface was discontinued. When the indi-vidual synoptic analyses were compiled theywere then compared to each other. Noting thesimilarities and differences in theseanalyses lead to the technique to forecast

10

heavy precipitation events associated withfronts.

4. RESULTS

From the synoptic comparisons described inthe previous section the most representativesynoptic parameters were selected and modelswere developed for heavy and light precipi-tation associated with north to south andeast to west oriented fronts. Since all mapsfor a given time (either 0000 or 1200 GMT)were readily available, these models showconditions for approximately 12 hours priorto the occurrence of precipitation (precipi-tation will begin after 1200 or 0000 GMTdepending on whether the initial mapset wasfrom 0000 or 1200 GMT), so that a 12 hourforecast may be made, and also for following12 hour periods, of the continuation of theheavy precipitation inclUding the locationand movement of the heavy precipitationarea. In actuality, though, since there is adelay in receiving the facsimile products,the 12 hour forecast periods will besomewhat displaced. In all likelihood, theinitial mapset used will be complete at 0400or 1600 GMT. Therefore, the initial forecastperiod will be approximately 8 hours, withthe assumption that the forecast will bevalid until the next mapset is available(either 1600 or 0400 GMT). The abovereasoning will follow throughout the con-tinuation of the precipitation. Therefore inreality, the forecast periods will be fromeither 0400 to 1600 or 1600 to 0400 GMT.

4.1 Heavy precipitation Associated with ColdFronts Oriented North to South

The 1000-500 mb thickness was the mostimportant parameter which indicated thatheavy precipitation would occur. When athickness ridge starts to increase as thefront approaches, the forecaster shouldconsider forecasting heavy precipitation tobegin within the next 12 hours. The areawhich will receive the heavy precipitationis located between the thickness ridge lineand the downstream inflection point. Inaddition, the following conditions shouldexist in the area for a forecast of heavyprecipitation. Some of these conditions areshown schematically in Figure 1.

1. Mean surface- 500 mb relativehumidity> 70 t.

2. Vertical motions> 1 ub s-~

3. K-indices > 15, withrepresentative values about 20.

4. A pressure ridge moving overthe area at the standard levels700-200 mb inclusive. The rainarea will be located between theridge line and upstream inflectionpoint.

VOLUMB 6, HUHSER 2

Table 3. List ot cases selecte~ tor Stu~y .

,

,

, H~..1973

.. ...~.. to _ eoU 'root WIlI.ldo __ ".-p ,~.....tT .....a.ol_.y

..1" .....11o.'fJ! roh .....4.....1_ f.__.L

h ....n 1.0\1101.... ,. ro.........'",•• , ..1ppl•• tId1.101>0...

,•·•j

, .

... .

!l

1~••f

, .

• •

••!•:··o

,

•

••"..,..•11 ·th•••~.: :;:1;'--..·-.

•

,

•

,•

•

•

•

·•~,•,"

_Gin'. U~~cll. UIO

1200 GIft'. ~.;1._....,. I'"

-. -. _ fo..e ... -.nell CIfI'. " " _""" ~r.b.......,.. un , , • • • , , , .1$, _~......h.....mIInl"d Ito....

Ofl' 0 .... A. fo..·e..toLeNIoI ...... A. fo ..o...e. Loulol ......0000 Gll,.......n .. II""... ."" ....U.n"1"10,, ... , Mnn> h.Il,I.l_.

No at ISO .nd ~oo ., y.... ,ood. ~IIO."OO ••~u boo.....n 'S·'O,. , ........ U_I... oeC\lrrod h\ LeNIoI..... 4u~I....IIL. po,,,,•.

y.... ISO OlD, ...... sao OlD. h ••' no .... '00 OlD....... Soo ••

11

NATIONAL ~BATHBR DIGBST

Pig. 1. Synoptic situation accorapanyingfronts oriented north to south 8 h priorto the occurrence of heavy precipitation.In this and. all following figures thelongitude identifiers have been changedto avoid the concept that the rains arein a geographic area. For the8e schemat-ic composite8 the rain area was locatedand the parameters were placed in relationto the rain area. TO locate the heavyrain area the opposite procedure should befollowed. Legend:_Rain arear----X:-indexr-- --Mean surface-SOO IlIb rela-tive hUJllidity (\); ••• Thickness and850 IIlb ther'\llllll ridge line: •••Verticalmotion (lib 8-1 ).

++,:+

-,II b s .

values of

2. Vertical motions > 1

1. Mean surface-SaO mb relativehumidity> 60 t.

5. Wind directions SSW-WSW.

3. K-indices > 15, with30 being common .

4. The pressure ridge aloft willhave moved eastward so that theprecipitation area is under theinfluence of a trough-ridge system(with the inflection point locatedover the heavy precipitationarea) .

6. Strong diffluence at 300 mb.

7. A thermal ridge located overthe precipitation area at the850-500 mb levels.

cribed above if the heavy precipitation isto persist. Figure 2 is a schematic of thesynoptic patterns .

+."

•1~. +

•+....,-, • 1.5

•

•

6. Slight diffluence at 300 mb.

5. Wind directions wsw-SSW at alllevels with wind speeds 40-50 ktat 850 mb.

8. A thermal ridge at all levelslocated over the potential rainarea. The 850 mb thermal ridgeline will usually coincide withthe 1000-500 mb thickness ridgeline.

+.m

Fig. 2. Synoptic situation accompanyingfronts oriented north to south during theoccurrence of heavy precipitation:Legend:_Rain areal X-index;-- --Mean surface-500 mb relativehumidity (\): ••• Thickness ridge liner••• Vertical motion (lJb 8-1 ) .

..+

,,+ •+•"'+-m

VOLUME 6. NUMBER 2

,,+- r' -;% .. +""+ + +•"+'. •"'+- + -+ .+ + +"'--------,,+- • '+---+ + +". "~ O-r- ..."' , .. .. .~

roi.,. 3. Synoptic situation accompanyingfronts oriented elllst to west 8 h prior tothe occurrence of helllvy precipitation.Leqeoo: Rllli.n area; lC:-index:

Mellin surface-SOO mb relativehtunidity (\cl; • •• Thickness ridge line;••• Vertical lIIOtion (\1b S-l).

4.2 Light precipitation Associated withFronts Oriented North to South

Prior to and during the occurrence of lightprecipitation the synoptic situation wasvery different from the conditions occurringwith heavy precipitation. If the followingconditions exist as the front moves towardand through the study area heavy precipi-tation should not be forecast.

1. A 1000-500 mb thickness troughintensifying and moving throughthe area.

Z. Mean surface-500 mb relativehumidities < 60 t.

3. K-indices < 15.

4. A pressure trough at all levelsmoving into and across the area.

5. Winds WSW-NW at all levels (anexception to this was at 850 mbprior to the occurrence of lightrain, the winds were NE-NW at 5-25kt) .

6. oewpoint depressions , 12 C at850 and 500 mb.

7. No diffluence at JOO mb.

a. Vertical motions 1 ,b -1< s

4.3 Heavy Precipitation Associated withFronts Oriented East to West

The synoptic situation for east to westoriented fronts was not defined as well as

13

1. Mean surface-SaO mb relativehumidity> 701.

3. K-indices > IS, withrepresentative values about 30.

will be located over the western portion ofthe rain area and the area will be boundedon the north and south by the 5610 and 5690mb thickness lines respectively. In addi-tion, the following conditions should existover the area defined above if the rain isto presist and will serve further to definethe area receiving heavy precipitation.Figure 4 shows the schematic relationship.

NATIONAL WBATHBR DIG6ST

it was for north to south oriented fronts.Again, the 1000-500 mb thickness was thebest parameter for identifying the heavyprecipitation forecast (although not asdefinitive as it was for north to southoriented fronts). If a 1000-500 mb thicknessridge should begin developing when the frontis along the Gulf coast a forecaster shouldconsider the occurrence of heavy rain in thenext 12 hours. The area of heavy rain willoccur east of the thickness ridge line. Theeastern border of the rain area can be foundwith the aid of the other parameters thatshould exist if heavy rain is forecast.Figure 3 is a schematic representation ofthe synoptic conditions.

2. Vertical motions > 2 -1,b s

1. Mean surface-SaO mb relativehumidity" Sat.

2. Vertical motions" 1 ~b s-l.

3. K-indices " 15 withrepresentative values being 20 to]0.

4. A trough-ridge system over thestudy area at 850, 300, and 200 mbwith the system at 850 mb having ahigh amplitude and the rain arealocated between the trough lineand inflection point under northto south oriented contours.

5. A thermal ridge located overthe precipitation area at 850 andSao mb.

4. A trough-ridge system locatedover the study area at 850, 300,and 200 mb with the system having

6. Wind directions SWat 850, 300,and 200 mb.

pig. 4. Synoptic situation accOlIIpanyingfronts oriented east to west during' theoccurrence of heavy precipitation..Leqend:_JUlin area: J(-indexJ-- --Mean surface-SOO lab relativehulllidity ("I: ••• Thickness ridge line:••• Verticlll motion (JJb s-l) .

+"___IS

+'"

+

+.,••..,..,

• +2 •

*- .-+-

••..,...,

"'+

a high amplitude at 850 mb and therain occurring under the south-westerly winds.

5. A thermal ridge located overthe area at 850 and 500 mbs.

6. Oewpoint depressions < 6 C at500 mb.

Once the precipitation starts, as with northto south oriented fronts, the synoptic con-ditions for the continuation of the rainfallare somewhat different. The 1000-500 mbthickness was once again the key parameterin this situation. The thickness ridge line

4.4 Light Precipitation Associated withFronts Oriented East to We~t

Prior to and during the occurrence of lightprecipitation the synoptic situation wascompletely different than that occurrin~with heavy precipitation. If the followlngconditions exist as the front moves towardand through the area heavy precipitationshould not be forecast.

1. A 1000-500 mb thickness troughlocated over the study area.

2. Mean surface-SaO mb relativehumidities < 50%.

14

VOLUME 6, NUHBER 1

Table 5. verification of forecast model for fronts oriented north to southduring the occurrence of rain. Forecasts are made for 12-h periods usinginitial data. Legend as in Table 4.

.' .

1200 CM1'. nooסס- em. 11,,-.clt.l"O...·:0;) C;~

I1J" r.Y.T. 11..... cl,. , ••0

000•;••i;.1 ult-Hn tllI.lulFP.'POI':> 65\.

OFf" 0 11-•••t.rn "1 •• I.~lppAh~......""-Geor.:;I.,:> foP 55\.

"on, t ... "Pt,PC, o~•• loutll-,"lIu, Unite"Stlt.. " 25\ •

"onl ta'" cpr,pop Oy........ tll-Clucrn Unlu

NATIONAL WEATHER DIGEST

Table 6. Verification of forecast model for fronts oriented east to westprior to the occurrence of rain. using initial maps, any heavy rain shouldbe forecast to begin in approximately 8h. Legend as in Table 4.

I". 0, , , ! ,

~I~ " • , ~ ,~~ ~ H :1"'i ; " ! h••~~ - ~t~ g • " "j r i fl ; 1:, i~ ; , ., Il , ".... .- ., •

110:0 ""_lUO

3. R-indices < 15.-1

4. Vertical motions < 1 lib s

5. A thermal trough located overthe study area at 850 and 500 mb.

6. Dewpoint depressions ~12 C at500 mb as the front becomesstationary.

7. Wind directionS of W-NW at 850,300, and 200 mb.

5. VERIFICATION

Frontal passage situations that occurredafter March 1977 were used to verify theeffectiveness of the proposed forecastingmodels. These will serve as an example ofapplication. These verifications are shownin Tables 4-7. Each forecast parameter forthe test cases was examined to see if it metthe model criteria and entered into a tableby the use of yes (Y), no (N), or marginal(M). After this was done, the heavy rainarea (if any) as forecast by these modelswas compared to the Quantitative Precipi-tation Forecast and Probability of Pre-cipitation rainfall areas forecast by theNational Weather service, and to the actualprecipitation areas. As the tables show, fornorth to south oriented fronts, the de-veloped forecast models compared favorablywith the National Weather Service guidance.On a positive note, the proposed modelsverified one heavy rainfall event that theNational Weather Service missed. For e~~t towest oriented fronts the developed modelsalso performed reasonably well. In mostcases when the models and National WeatherService forecast no widespread heavyprecipitation none occurred (although therea few areas of localized heavy precipita-tion, this probably would not lead toextensive severe flooding). Again, in onecase when the Quantitative PrecipitationForecast forecast a widespread area of heavyprecipitation, and the new model d4d not,the new model was a slight improvement (somesmall sections of the acea in question didhave 25mm of precipitation).

Although not formally presented, thesynoptic conditions accompanying all thefronts for the winter season of 1979-1980were observed. No cases were observed thatmet the conditions of the model that heavyrainfall did not occur and no heavy rainoccurred when not forecast. It is realizedthat one year is not a large sample and thatthe model could forecast heavy rain and therain not occur.

6. SUMMARY AND CONCLUSION

The meteorological parameters causing heavyor light rains associated with wintertimefronts in the southeastern United States

VOLUME 6, NUHBER 2

have been identified. It was found thatcertain synoptic characteristics distinguishthe heavy from light precipitation cases.The key parameter for both nort:l to southand east to west oriented fronts was the1000-500 mb thickness. A thickness ridgewill be present in the heavy rain cases,while the light rain cases will have athickness trough. In addition several otherparameters are noted as being unique to theheavy rain cases. To test the use of theabove information as a forecasting tool averification was performed. This verifi-cation showed the technique worked well atforecasting and pinpointing areas of heavyprecipitation.

ACKNOWLEDGEMENTS

This research was supported in part, by theAtmospheric Research Section, NationalScience Foundation, Grant number ATH76-22282.

RBPBR8NCBS

(1) AMS, 1978: Statement ot concern: t1ashE1oods-~ n~tiona1 problem. Bull. Amer.Meteor. Soc., 59, 585-586.

(2)Fawcett, E.B., 1977: Current capabilitiesin prediction at the N~tiona1 WeatherServices' National Meteorological Center.Bull. Amer. Heteor. Soc., 58, 143-149.

(3) Garcia, 0., L. Gosart, and G. DiMego,1978: On the nature ot the winter seasonrainfall in the Dominican Republic. Hon.Wea. Rev., 106, 961-982.

(4) Haddox, R.A., c.r. Chapell and L.R.Hoxit, 1979: Synoptic and mesoscale ~spectsof flash tlood events. Bull. Amer. Meteor.Soc., 60, 115-123.

(5) Jl:reitzberg, C.W., and H.A. Brown, 1970:Mesoscale weather s9stems within anocclusion. J. App1. Meteor., 9. 417-432.

(6) Smith, W., and R.J. Younkin, 1972: Anoperationally useful relationship betweenthe polar jet stream and he~vy precipi-tation. Mon. Wea. Rev., 100, 434-440.

(7) Bosart, L.P., 1973: Detailed analyses ofprecipitation patterns ~ssociated withmesoscale features accompanying UnitedStates east coast c9c1ogensis. Hon. We~.Rev., 101, 1-12.

(8) Estoque, 8., 1976: 8ehavior ot trontsover the Gulf of Mexico. Proceedinqs ofConference on Meteorology Over the Gult ofMexico, College Station, Texas A & HUniversit9. 1-8.

(9) Jensen, R., 1977: Record storllls andflood over the Red River Va11e9, June-July1975. 8ull. Amer. Meteor. Soc., 58, 502-504.

17