<<

Q.f.R.Meteorol. Soc.(1990). 116' pp. 113!1151 551.515.13:551.553. I 1(680)

On the generationand propagationof the southernAfrican coastallow

C. J. C. RensoNrand M. R. JunY2 I Atmospheric ScienceProgramme and Instituteof Applied Mathematics,Uniuersity of Brilish Columbia, Vancouoer,Canada*;2 Departmentof Oceanography,Unioersity of , SouthAfrica

(Received30 January 1989.revisd 7 March 1990)

Sut'lveny The time-heightvariability of the lower marineatmosphere during the eastwardspropagation of coastal lowsaround the coastof southernAfrica is presented.These coastal lows are shallow,mesoscale disturbances that are trappedvertically by a srrongsubsidence inversion and horizontally,through Coriolis effects, against the steepescarpment that fringesthe subcontinent.Prior to the event,warm offshoreflow descendingoff this escarpmentat about the inversionlevel providesan input of buoyancywhich, togetherwith the cyclonic vorticity acqu.iredduring the descent.acts to generatea coastallow rather than a high. Analysisof the datashows that coastallows are characterized by a markedlowering of theinversion level, a switchin the directionof the windsbelow the inversionand a drop in the near-surfacetemperature, which are all consistentwith the hypothesisthat the coastallow is an internalKelvin wavepropagating in the marine layer.Good agreementbetween the observedpropagation speed of the coastallow and the theoreticalphase speedof sucha wave reinforcesthis hypothesis. A model of the vertical buoyancyadvection into the marine atmospherethat resultsfrom the warm offshoreflow is presented.lt is illustratedthat an internalKelvin wave,in the form of a coastallow. results from this buoyancyadvection. Comparisonsare made betweencoastal-low dynamics and those characteristicsof other mesoscale, coastally-trappeddisturbances which propagate in the marinelayers of Californiaand south-easternAustralia.

7. INrnooucrron The coastallow of southernAfrica is a shallow,mesoscale low-pressure disturbance that propagateseastwards around the subcontinentand whose distinct features arise from the unique interaction between the atmosphericstratification, the synopticweather patternsand the topographyof the region. Owing to the frequencyof its occurrence, which may be takenas once every six days(Preston-Whyte and Tyson 1973;Coastal Low Workshop-hereafterreferred to as CLW 1984;Kamstra 1987), and the fact that the intensityof the disturbanceis greatestnear the coastand decaysseawards over a Rossby radius(Anh and Gill 1981),the coastallow is a dominantfeature in the weatherof this area. In all observedcases, the propagationof the coastallow is precededby the eastward movementof the migratory subtropicalanticyclones and followed by the mid-latitude frontal systemsof the SouthAtlantic and Indian Oceans.Acting as a horizontalobstacle to this westerly flow are the coastalmountains of the region, which rise steeplyto a height of over I km (seeFig. 1). With the exceptionof a few river valleys.the coastal mountainsform a relativelysmooth, semicircular barrier so that with the interiorplateau, SouthAfrica as a whole resemblesa domewith a horizontalradius of curvatureof about 900km. This steeplysloping, curved surface is thereforeable to steerlow-level coastal weatherefficiently around its periphery(Taljaard 1972). Owing to the presenceof a strongsubsidence inversion at about the heightof the coastalmountains, flow constrainedin the horizontalby the topographyis trappedin the vertical.Below this inversion,the constrainedflow is alsotrapped horizontally as a result of Corioliseffects on the along-shoreflow. The lattermay be assumedto be in geostrophic balancewith the across-shorepressure gradient (Gill1977:' Anh and Gill 1981).No such balanceneed exist below the inversionbetween the across-shoreflow andthe along-shore

' Presentaffiliation: C.S.I.R.O.. Division of AtmosphericResearch, Victoria, . 1133 i,iiii,,.,',i;;ii,i^,ii::r,.;:;ir:;iiirriftiiiL

1134 C. J. C. REASON and M. R. JURY

flow that leads forcineof the in Gill (1977), offshoreflow, of a coastal In low is conside three case hypothesis. 1984:Hevde that the coasta February,Apri

Fieurel. Topographicand location map for southernAfrica. AB refersto AlexanderBay. L to Langebaanweg, Ciio Cup" foo.-n.G to George,PE to ,EL to EastLondon, and D to Durban.The firstcontour In Fig. 2 inlandfrom the coastlineis th-e500m contour, that enclosingthe stippledregion is the 1000mcontour and that in all three ca the shadedregion is the 1500mcontour. enclosing was preceded landmass.The pressuregradient. Instead, the across-shoreflow must vanishat the coastalmountains usingships of t""uur" of ttr. barrier effect, and is generallynegligible within a Rossbyradius (100- readilvavailab 200km) seawardsfrom thesemountains (Gill1977; Anh and Gill 1981;Bannon 1981; The ridsi ( CLW i9S4). As a result, the initial flow is constrainedto move anticyclonically(i'e. the south-west eastwards)around the coastalmountains within a narrowzone of the orderof the Rossby level Rossby radiusin width. from the ridging Gill (1977)first suggestedthat the structureand propagationfeatures of the coastal causeoffshore lows are similar to those of coastallytrapped waves in the ocean.Through use of the measuredat the Durban), nonlinearshallow-water equations on the /-plane, with a reducedgravity model of the i stratification.Gill showedhow interactionof incidentbarotropic westerly waves with the As the ri escarpmentgenerated coastal lows in the form of a Kelvin wave.The nonlinearadvection southon the we terms were shown to causesteepening of the wave front. thus simulatingthe rapid to the sout changesin wind speedsand direction and in inversion height which are sometimes ridginganl observedduring coastal-low passage. inducesoffshore propagate Followingon from Gill's study,Bannon (1981) and Anh andGill (1981)presented arou modelsthat elamined the interactionof theseincident synopticwaves (the baroclinic trailing casewas also considered) with the escarpmentin more detail and gavemore information systems on on the dispersionproperties of the therebygenerated Kelvin waves.To accomplishthis, the right of front a two-layermodef of the atmosphericstratification on the beta-planewas employed and, of the trail the to allow an analyticsolution, the nonlinearterms were neglected. Despite the seemingly cyclonecan because restrictiveassumptions made, the analyticalmodels were able to explainmost of the of the typical observedfeatures of coastallows and are consistentwith the resultsobtained from a of su numericalforecast model (de Wet 1979'1984). Figure2 of One of the featuresthat the modelshave not explainedentirely is the factthat one 4-5 degrees has tendsto observepredominantly coastal lows rather than highs.A linearKelvin wave reachedAlex Cape modeis unableto accountfor this featureso the preferentialpropagation of coastallows Town ( wasattributed to nonlineareffects by Gill (1977)and Anh and Gill (1981).While not southcoast. On t disputingthis mechanism,Bannon (1981) also considered important the effectsof the of the svstema forcing synoptic anticycloneand its associatedoffshore flow at the inversion' In this .When (summer paper, data and a model will be presentedthat isolatethe contributionof this offshore mode). SOUTH AFRICAN COASTAL LOW 1135 flow that leadsto the formationof a coastallow as opposedto a high. While the overall forcing of the low occursvia the interactionof the synopticflow with the escarpment,as (1981),it is arguedhere that it is the in GitI egTi), Anh and Gill (1981)and Bannon offshoreflow, resultingfrom this interaction,that providesa mechanismfor the formation of a coastallow rather than a high. In agreementwith Gill (1977),Anh and Gill (1931)and Bannon (1981), the coastal low is co-nsideredto be some form of Kelvin wave and the resultspresented here for three case studies of coastallow propagationare shown to be consistentwith this hypothesis.These casestudies were chosen,on the basisof previousresearch (CLW tgb+; geyAenrych1987), as being representativeof coastal-lowbehaviour. To illustrate that the coastal low is an all-year-roundphenomenon, events from the months of February,April and Septemberwere selected.

2. SvNosrtc coNDITIoNS In Fig. 2 are SouthAfrican WeatherBureau surface synoptic maps which showthat in all three cases,coastal-low formation near 25'S on the west coastof southernAfrica was precededby the ridging of the South Atlantic anticycloneto the south of the landmass.The offshorescale of the coastallow is determinedby the Weather Bureau using shipsof opportunity and offshoreoil-rig data. Unfortunately,these data are not readily availableand so are not discussedhere. The ridging of the SouthAtlantic ,together with the trailing cycloneto the south-wJrt(in the region4H5"S, 0-10'E), are the surfaceexpressions of an upper- level Rossbywave in the generalsynoptic cycle of the southernmid-latitudes. Resulting from the ridgingare easterly winds over the interior which,on descendingthe escarpment, causeoffshore flow alongthe westcoast initially. Figure3, whichconsists of wind profiles measuredat the four coistal stations(Alexander Bay, Cape Town, Port Elizabethand Durban), indicatesthat this offshoreflow occursat about the inversionlevel' As the ridging processcontinues further eastwards,offshore flow extendsfurther south on the west coastand hencecoastal-low propagation is induced.Meanwhile, far to the south-westthe trailing frontal systemis beginningto develop.Eventually, the ridging anticyclonepinches off to form a migratory high which, as it trackseastwards, ind-ucesoffshore flow alongthe south and eastcoasts and, in tandem,the coastallow to propagatearound the Capeand hencealong the southand, finally,east coasts. Since the iraiiing cyclonesare fastir than the coastallow (Fig. 2), the distancebetween the two ,y.t"*lt d""r"ur.r as theymove eastwards. In cases2 and 3 (middleand bottompanels on the right of Fig. 2), tire coastallow is graduallyovertaken and dissipatedwithin the front of the trailing cylonewhereas for caseI the low dissipateson the eastcoast before 2). latter situation arises the cyclone can up with it (top right panel, Fig. -The "itc'h becauseof the more southerlytrack of the mid-latitudecyclone in case1 and is more typicalof summerconditions when the subtropical high-pressure belt lies nearer the pole. Figure2 showsthat the coastallow movessouthwards along the westcoast at a rate of 4-5l"gt""r of latitudea day. so that abouta day afterformation near 25"S' the low hasreachid Alexander Bay (29"S, 16"E). By thefollowing day, the coastal low approaches CapeTown (34'S,18'E) after which the systemrounds the Capeto propagatealong the southcoast. On the sourh'andeast coasts, both the speedof propagationand the lifespan of the systemare influencedby the intensityand proximityof the trailingmid-latitude cyclone.When this cyclone is either relatively weak or liesfar to thesouth of thelandmass (iummer mode). the low is able to propagatemore quickly on the southcoast. Figure 2 1136 C. J. C. REASON and M. R. JURY

-E o O =^- u;bg r i.lS € I oo= v.89,-.

o-oc

Figure 3. Ti at 4ms-r inte of the datum). speedsrecorded hours after arrival

also showst (34'S,26'E)in Speedson henceless probably Agulhas Curre offshore.With reachDurban 1) or is overt mountarns steerthe low i lnverslon over easterncoast able to dissipa system In low around cycle of the continual I osrc z asEc g asEc the coastal SOUTH AFRICAN COASTAL LOW 1137

(a) (b) I

(contoured Figure3. Time-heightplots ((a) referstocase l, (b) to case2 and (c) to case3) ofwind speed ui"ar.:i inierval)un"o air"ction (urrorur,where the'point of the arrow refersto the time-heightco-ordinates wind oi tft" Out,t.l. Wind speedsgreiter than 8ms-rare stippled.The dashedline refersto the minimum "event. 'd'-refers to 12 ip".J. ,".or,i.d during each On the time axis, to the time of coastal-lowarrival"lT' millibars. hoursafter arrivaland;-12'to 12hours before arrival, etc. On thevertical axis, the pressuresare in alsoshows that the low moves7-9 degreeof longitudeto a positionnear Port Elizabeth (34"S,26'E)' in the day following its arrival at CapeTown. Speedson the east coast are generally slower becauseof a weaker inversion and henceless effective trapping of the low in the vertical.This weakeningof the inversion probably resultsfrom the greatersurface heat fluxesover the warm sea surfaceof the Agulhas Current which skirts the coastfrom Durban to Port Elizabethbefore moving oflshore. Within a day or lessof arriving at Port Elizabeth,the low is seen (Fig.2) to reachDurban (30'S,31'E) after which the systemmoves out to seaand dissipates(case "I) or is overtaken by the trailing frontal system(cases 2 and 3). Since the coastal mountainsturn significantlyinland north of Durban, the ability of the topographyto steerthe low in thi horizontalis quickly lost so that togetherwith the weakersubsidence inversionover the warm Agulhaswaters, there is little effectivetrapping on the north- easterncoast of southernAfrica. Surfacefriction and the synoptic-scaleflow are then ableto dissipatethe mesoscalecoastal low rapidly.For example,Fig.2shows that the system- disappears in a daYor less. In summary,it is clear from Fig. 2 that the anticyclonicpropagation of the coastal low around the coastof southernAfrica is very closelytied in with the synopticweather cycleof the southernsubtropics/mid-latitudes. It is this closeassociation which provides cbntinualforcing for the coastallow and,as the analysisof sections3-5 will show,enables the coastallow io propagatein the coherent,simple wave-like manner observed. 1138. C. J. C. REASON and M. R. JURY

3. AruospueRlc soUNDINGDATA

Figure 4(a-c) consistsof temperatureand dew-pointtemperature profiles through the lower atmospheretaken from soundingsprovided every lZhours by the SouthAfrican WeatherBureau for the four coastalstations (Alexander Bay, CapeTown, Port Elizabeth and Durban-see Fig. 1 for locations).In virtually all examples,a pronouncedtem- peratureinversion (indicated by the arrow in Fig. (a-c)) of 5-10 degCexists. Note that 6 in some cases,pronounced surface heating preventsthe formation of a temperature inversion. In these cases,the inversionlayer can still be detectedby the substantial decreasein dew-point temperature that always occurs over this layer (CLW 1984; Heydenrych1987) and which marksthe transitionfrom a moistermaritime airmass below to the dry continentalairmass above the inversion.As the coastallow approachesa given station, the inversionis observedto lower and strengthenand then increasein height and weakenafter coastal-lowpassage. The actualminimum in inversionheight occurs at the timeof passageof the coastallow; behaviourwhich is consistentwith a linearKelvin- wave model of the disturbance. Apart from the drop in inversionheight, coastal-lowpassage is also characterized by variationsin the temperatureand dewpoint,a shift in wind speedand directionand mostreliably. a minimumin the geopotentialheight of the variouspressure levels below the inversion. Comparisonof the wind directions(Fig. 3) with the coastlineorientation (Fig. 1) showsthat the coastallow is generallypreceded by offshoreflow nearthe inversionlevel. which switchesto an onshoreor along-shoredirection as the systemarrives. The marine- layer winds reversedirection from a trade-windflow prior to coastal-lowarrival to the cycloniccirculation (generally onshore) of the systemitself as the leadingedge passes through.Since the trade-windflow is southerlywith the coaston the right, there will be an Ekman divergenceof the marine-layerair due to the surfacefrictional contrast betweenland and sea. This divergencetogether with the regionalsubsidence beneath the SouthAtlantic anticycloneallows the dry continentalair movingoffshore to descend off the escarpmentand influencethe marine layer. Sincethe offshoreflow is warm and dry, beingof continentaltropical origin, near- surfacetempcratures rise prior to the coastal-lowarrival, reachinga maximumwhen the inversiondrops to its lowest point. At this point, the extent of mixed layer subjectto surfaceheating is shallowest,hence the higher temperatures.The onshoreflo* behind the trailing edgeof the coastallow is of maritime origin and hencerepresents a cooler andmoister airmass. As a result.near-surface temperatures tend to decreasesubstantially after passageof the coastallow and there is usuallya markedrise in the dew point. The latter leads to a-significant increase in cloud cover below the inversion,invariably as stratus. As the trailing edge passesthrough, surfacepressure and the inversion height increaseand the near-surfacewinds shift to along-shoreand strengthen(Figs. 3 and 4). Associatedwith the rising inversionis an increasein the near-surfacemoisture content. :33 Typically,dew pointsin the layer near 900mb rise by about 8 degCafter passageof the coastallow. Another commonfeature is the generallyhigher inversion level on the southand eastcoasts compared with the west coast.This featureis a result of the greatersurface heatingon the Indian Oceanside of southernAfrica with its warm AgulhasCurrent. Sea surfacetemperatures on the Indian Oceancoast are typically10 degC warmer than those on the cold Benguela-CurrentSouth Atlantic side (CLW 1984;Shannon 1985) where coastalupwelling often further reducesthe temperatureof the inshorewater. t

SOUTH AFRICAN COASTAL LOW 1139

^' ^E ad

(tt

N_O oi 26 v6 L-1 99 !.=

o2 ax

.b9 a.= ^o st oo

t Eo :F 94

i^ol

L

-go c>ga o-.9 9; ---l oo od>

L€ 3.c -ttt-' | -','- o- :-7<| -i | - YO \. l 'ic l9o

I ca do I f"P

rX

:S o F l! o C) o. 9.9 EO o> Fo

g'J .9p IL C. J. C. REASON and M. R. JURY

-o

:u 6 a R9 .9 e /

-t 6 E{ I --J c n

-t 6: iP :i 1E

I o 9a oE eg t I -t 9'l "tiE "E

+ o oo

--^ 6g l& 9E .: t ffi -t-'i

i;

G }4/ -9 9E a t I

EAaaEs- E38383" 888883 (r - (G) (r - (n) (r . (-) (r - (sl .l| fulor, r€l e@o') r€* Ul do,, rl* O mor) b G

' "'i;W,i!j.ii,i,;7.:I,,i: SOUTH AFRICAN COASTAL LOW tl4l

_).--: 1,

6 a oi

o 93 ___l U q o

o0

9v

I ol i

E "1 ;E

sIErYx- --=-e-l_c8"'' (r - (e) In* r,'- *.o,t r*l ,o* Nor) 3q F 1142 C. J. C. REASON and M. R. JURY

Case AB CT PE AD Case AB CT PE AD Case AB CT PE AD

Notes (i) (r o

E00

4. Cotutp

t000 The -21 -12 0 12 21h -21-12 0 12 zah -21-n 0 fl ah is now considere Figure 5. Time-height variation (15gpm contour interval) of the geopotentialanomalies (gpm) from the an internal Kelv meanrecorded during the event:(a) refersto case1, (b) to case2 and (c) to case3. labelson the horizontal (time) axis are as in Fig. 3. Anomaliesbelow -rj,i,?#rl:. stippled.On the verticalaxis, pressuresare in

whereil is the Figure 5 exhibitsthe departurefrom the meanof the geopotentialheight at various on the di heightsduring passageof the coastallow. As mentionedpreviously, the time of passage of the low at a given station is coincidentwith the surfaceminimum in geopotential height.Thus, for a casewhich haslittle or no interferencefrom the synoptic-scaleflow From the inve at levelsabove the inversion,time-height plots of geopotentialdifferences should consist as givenin T of a core of negativevalues around the time of passageof the low and confinedto levels speedsis seen below the inversion.This type of behaviouris seen(Fig. 5) to occur most often. There completelyrelia are two typesof exceptionto this behaviour,however. Where upper-levelcyclogenesis separatedfrom hasoccurred (case 3 at Durban), the extrememinima in the geopotentialheight are at probablyvi the levelsabove the inversion.More common(cases 2 and 3 at CapeTown and 2 at Port valid.In additi Elizabeth)are geopotentialminima which extendfrom the surfaceto levelswell above effectivetr the inversionat or just after the time of passageof the coastallow at thesestations. This the topography type of behaviouris indicativeof influencesresulting from the Rossbywave associated Two other with the forcing anticycloneand trailing cyclone.As expected,this structurein the coastal-low geopotentialis not observedat AlexanderBay or Durban becausethese stations are well Kelvin wave is north of the track of thesesystems. the southernhe It is consideredthat the most reliable signatureof coastal-lowpassage at a given secondreoui stationis the surfaceminimum in geopotentialheight. Using this indicator,the time of coast with an passagefor eachevent is determinedand hencethe speedof propagationfor eachcase can be calculated(see Table 1). In the next section,this value is comparedwith the theoreticalphase speed of a linearKelvin wave. where f is the SOUTH AFRICAN COASTAL LOW 1143

TABLE l. OssrRvrocoAsrAL Low AND LINEAR KELvIN-wevr sneeos (ms-')

Coastal-lowarrival Observedspeed Kelvin-wavespeed

Case1 AB 7 Feb. l98l (12) 7 7 CT 8 Feb. 1981(0) 9 8 PE 9 Feb. 1981(0) 9 9 AD 9 Feb. 1981(12) 5 3 Case2 AB 17Sep. 1e8s(0) 6 6 CT 17 SeP.1985(12) 8 7 PE 18Sep. 1e8s(0) e 9 AD 19 SeP.1e8s(0) 8 6 Case3 AB 19APr. 1980(12) 9 l0 CT 20 Apr. 1980(o) 10 8 PE 21 APr. 1e80(o) 8 8 AD 22 APr. 1980(12) 3 J

Notes: (i) (0) refersto 0000hours GMT and (12) to 1200hours GMT' Thus7 Feb' 1981 (12) refersto coastal-lowarrival nearestto the 1200GMT radiosondeascent on thatday. (ii) AB refersto AlexanderBay, CT to CapeTown, PE to Port Elizabethand D to Durban.

4. COVpRRISON oF COASTAL-LOw BEHAVIOUR wITH KeI-vlu-wnvE THEORY The consistencyof the observedcoastal-low behaviour with Kelvin-wavedynamics is now considered.Thus, the coastallow is assumed(e.g., Gill 1977)to take the form of an internal Kelvin wave propagatingin the inversionlayer with a phasespeed c: c : (g,lf)tl2 (4.1) wherel/ is the heightof the inversionlayer and g' is the reducedgravity, which depends on the differencein potential temperatulesof the upper (Or) and lower (@2)layer as g'=g(@r-@)l@2. (4.2)

From the inversiondata presentedin Fig. 4, the Kelvin-wavephase speed is calculated as given in Table 1. In general,good agreementbetween the observedand theoretical spe;ds is seento exist. At Durban however,the theoreticalvalues are not considered completelyreliable for severalreasons. Only in case1 (Fig. 2) is the coastallow still well sepaiatedfrom the trailing frontal system.In the other cases,the linear dynamicsare probably violatedby adveitive forcing from the nearbyfront and so (4.1) is no longer valid. In addition, the generallyhigher and weakerinversion at Durban makesfor less effectivetrapping of a Kelvin-wavefeature and this is enhancedby the turninginland of the topographyof this station. fwo ottrei crucial propertiesof Kelvin waves,which must alsobe satisfiedby the coastal-lowcases examin"d h"te if this model is to be valid, are as follows. Firstly' a Kelvin wave is requiredto propagatewith the coaston the left (lookingdownstream) in the southernhemisphere (Gill i9b2). This conditionis clearlymet here(see Fig. 2). The secondrequiremeni for a Kelvin wave is that this waveshould decay offshore from the coastwith an e-foldingdistance given by the Rossbyradius R: R = (r'H)tt2llfl (4.3) given 1, the where / is the Coriolis parameter.Using the data for (8'H)tl' in Table tt44 C. J. C. REASON and M. R. JURY

Rossbyradius is seento be of the order of 100km in eachcase. Such a value is clearly during which consistentwith the offshorescale of the coastallow evidentin Fig. 2. the coastt Additional support for the Kelvin-wave hypothesisarises from the coastal-low difference) is behaviouras displayedby the soundingdata of Fig. 3. The relativelycoherent variations Cape south in the inversionlayer height and wind speedand direction are characteristicof Kelvin- In addition wave dynamicsas opposed to the gravity-currentdynamics observed for other coastally- of cyclonicvorti trappeddisturbances that propagatealong the west coastof North America (Massand terms.the Albright 1987;Dorman 1987).It has been stated by the former authorsthat gravity- boundary-layer current dynamicsmay alsoapply to the coastallow. However, characteristicfeatures of potential vorti gravitycurrents such as unsteady, surging propagation (Simpson 1987), abrupt (occurring potentialvortici in minutes)changes in marine-layerwinds, temperature,pressure and inversionheight vorticity is (Dorman 1987;Mass and Albright 1987),and the formation of a reservoirof denseair cyclonicvorti to feed the coastalgravity current (Dorman 1987;Reason 1989) were not observedin the lower densi the coastal-lowcases nor in previousstudies (CLW 1984;Heydenrych 1987; Hunter low is then tr 1987).For thesereasons, and becausethe coastalJowspeed is not matched(Table 1) by on the the theoreticalgravity-current speed: mountalns. c:0.i9(6P/p)tlz (4.4) To conside which dependson the surfacepressure difference 6P betweenthe gravitycurrent and the marine-layer ambientair and the densityp of this air (Seitterand Muench 1985;Mass and Albright equationderi 1987),the coastallow cannotbe considereda gravity current. approxlmatlon The solitary Kelvin-wavehypothesis, suggested by Dorman (1985)as explanation inversion.A for a Californian trapped event of.4-7 May 1982,is also not consideredlikely here. by Gill (1e77) Firstly, the observedcoastal-low speeds were shown (Reason 1989) to be more accurately of the describedby a linear Kelvin-wavemodel. Also, the coastallow is observedin all cases model sincethe to propagatewithout lossof identityaround the pronouncedbend in the coastalmountains are much larger just east of CapeTown. Such behaviouris consistentwith that of linear Kelvin waves (Miles 1972;LeBlondand Mysak 1978)but not with solitaryKelvin waves(Miles 1977; Dorman 1985;Reason and Steyn 1988). To summarize,the linear Kelvin-wavemodel of the coastallow is felt to be the most appropriate.Nonlinear effects are considered to be smallin the formationstages (Reason 1989),but may becomesignificant once the low haspropagated some distance by leading to steepeningof the leadingedge and, hence, lags between the inversionheight minimum and the wind shift (e.g. Gill 1977).

5. THe coNrRtBUTroNoF oFFSHoREFLow ro coAsrAl-Low cENERATToN where suffixes L denotepartial de While the interactionof the synopticflow with the escarpmentis responsiblefor the ft is the inversi< overall forcing of a propagatingcoastally-trapped disturbance (Gill 1977;Anh and Gill (assumedzero), 1981;Bannon 1981),analysis of the data abovehas indicated that it is the offshoreflow respectively,/is (Fig. 3) resultingfrom this interactionthat providesa mechanismfor coastal-lowrather of transfer of air than coastal-highformation. Theoretical backing for this observationis providedin this is downwards) section. Attention is confined to the generationof the systemsince the propagation layer from a characteristicsof the coastallow as forced by the synopticflow havebeen dealt with in flux of density detailby Gill (1977),Anh and Gill (1981)and Bannon(1981). For the purposesof this layer) if the air discussion,the offshoreflow will be referred to as a berg wind (a local term). After A descendingthe escarpment,the bergwind reachesthe low-lyingcoastal areas as a warm, -.r represents offshore breeze. of the southern Sincethe temperaturedifference between the adiabaticallywarmed berg-wind air- significantly massand the maritimetropical air usuallypresent at the coastmay be 10degC or more, of the coastallo the berg flow representsa considerablebuoyancy input into the marine atmosphere mation.It is SOUTH AFRICAN COASTAL LOW 1145 duringwhich denseair is replacedby lessdense. As a result,pressures tend to fall along the cJastthroughout berg-windevents. The extent of this input (i'e'' the temperature difference)is ofien observedsome distance offshore. e.g., about30km seawardsofthe Capesouth coast(Hunter 1987). In addition to the effect of the buoyancyinput, the berg wind also acts as a source of cyclonicvorticity owing to vortex stretchingas it descendsthe escarpment'In general due to t.rrnr, the anticyclonicvoiticity present-lg7g). over mountainousregions tends to decay boundary-layerfriction (Smith However, this decayrepresents an increaseof potentiai voiticity so that when the fluid columnsleave the mountain, their increased potentialvorticity is convertedto cyclonicrelative vorticity so that, asa whole, potential vorticity is conserved.Thus, a descendingberg'wind airmasswill lead to an area of cyclonicvorticity at the coastwhich, togetherwith the falling pressuresresulting from the lower densityof this airmass,will give riseto a coastallow-pressure cell. This coastal low is then trappedto within a Rossbyradius of the escarpmentthrough Coriolis effects on the along-siroreflow and the vinishing of the across-shoreflow at the coastal mountains. To considerthe above physicalargument for the effect of the berg wind on the marine-layervorticity dynamici in mathematicalterms, consider the potential-vorticity equationierivable fiom the reducedgravity shallow-rvaterequations in the Boussinesq approximationwith the interfacebetween the layersrepresenting- the marinesubsidence innersion.Application of a reducedgravity model to coastal-lowdynamics was first made by Gill (1977)and subsequentlyby Anh and Gill (1981)and Bannon (1981). Examination gravity oi ttre jeopotential data (ploited in Fig. 5) confirmsthe validity of a reduced model sincethe anomaliei'causedby coastal-lowpassage at and beneaththe inversion are much larger than those in the upper layer. The linear equationsare ut,-ft)r = -81' -8fy u1,* fu1= l,+Hr(u**ur")=h,-w, uzr-fuz=-BI'-8'h' u2,*fu2 = -8f., - g'hr- h,*H2(u2,*u,,):tr', where suffixesI andZ denote the upper and marinelayers respectively, suffixesx'y, I denotepartial derivatives with respectto thosevariables, u, u arethehorizontal velocities, ft is the inversion displacement,I is the displacementat the top of the upper layer (assumedzero), Hr and-H2are the undisturbeddepths of the upper and marine layer gravity(4.2), w" is the net rate iespectively,/ is tfre Coriolii parameter,g' is the reduced net transfer of transferof air from the upper into the lower layer(counted negative if the into the lower is downwards)and modelsihe subsidencevelocity of warm berg-windair a vertical layer from above.As shown by Gill (1932)and Ikeda (1984),lrys represents (i'e' the marine flux of densitybetween the layersthat is negativeinto the activelayer layer) if the air abovethe interfaceis transferredto this layer. ' y-axis, A co-ordinatesystem has been chosen such that the coastlinelies alongthe -x representsdistante seawards and z is positiveupwards' Since the radiusof curvature Taliaard 1972) is -: 'i: of the southern African coastal mountains (900km accordingto \ iignincuntlylarger than the typical across-shorescale (a Rossbyradius-about 100km) oittre coaital lows, modellingthese mountains as a straightbarrier is a valid approxi- \s. mation. It is also assumedthit the steepnessof the mountains(Fig. 1 showsthat they 1.t46 C. J. C. REASON andM. R. JURY rise to a heightof over 1km within 5G-70kmof the coast)allows their representationas radius.The bou a verticalwall. Considerthe casein whichtt1.D14u2,u2ltrd Ht*Hz (i.e.a deeppassive upper atmosphereover a shallow active marine layer). Writing u:tt2-ttt=ttz a'r1du: Although it is u2-Dt = r'2 yields,after includingthe previouslyneglected nonlinear terms, obtain a general -g'h, it is moreinst tr, + Lut,* uu, - fu : (5.1a) characteristicof -g'h, D,I Ltu.,* uu, * fu: (5.ib) scaleis much and sincethe t h, + {(h * H)ul, + {(h + H2)u}r: w,. (5.1c) (5.2)can be Cross-differentiationof (5.la. b) and substitutionfor the horizontal divergencefrom becomes (5.1c)then leadsto an equationfor the potentialvorticity: - -(u, - - DlDt(u, uy 1f): uy * fl(r, DhlDt)l(h + H) which hasthe g by (u,-u,*f) can be rewrittenas whichon dividing w = A(y, t) Olntlnlu, - uy t.fl) = -(r. - DhlDt)l(h + H). Hence, it is seenfrom this equationthat the contributionof a subsidingberg wind (w, - exp(-x negative)to the marine-layervorticity budget is to yield an increasewith time of the absolutevalue of the absolutevorticity lu,-u"+fl which for an /-plane means the whereA(y.t) a productionof cyclonicrelative vorticity. Sincethe In addition, the berg wind itself will possesscyclonic relative vorticity after it dependenta descendsthe escarpment.To seethis, apply Eqs. (5.1a-c) for the casewhere u and u suchthat are berg-windhorizontal velocities and h is the depth of the berg-windlayer (w, is now degenerate)to get the conservationof potentialvorticity of this layer: DlD4@, - u, * fll(h + H)l = 0. Equation(5.6) Hence,any increasein the depth of this layerdue to descentmust lead to an increasein absolutevorticity (u,-ur*fl and henceproduction of cyclonicvorticity on an /-plane. Thus, it is clearthat the berg wind itself acquirescyclonic vorticity duringthe descentas which.from the a result of vortex stretchingand that it actsas a relativesource of this vorticity for the with phase marine layer into which it is subsiding. Thus. the solut Although the potentialimportance of the bergwinds on the dynamicsof the coastal the coastfor low has been previouslysuggested (Bannon 1981;CLW 1984),its contributionhas not this form of beenexplicitly modelled. With this in mind, a simplelinear model of the effectson the of Thomson(1 inversionheight of introducinga coastalbuoyancy advection into the marineatmosphere Now consi will be developed.The modelis adaptedfrom the work of Ikeda(1984), who studieda indicatesthat tl similarsituation in coastalocean dynamics, and is basedon the shallow-waterequations hence,a relati (5.1a--c). loweringof the Althoughthe assumptionof lineardynamics is not expectedto be validfor coastal- inversionhei low behaviouras a whole, it is believedto be sufficientlyaccurate in the initial stages, that the berg w consistentwith the analysisof Reason(1989), since the flow here is characterizedby a Sinceberg Rossbynumber significantly less than unity (typicalwinds being less than 5 m s-r and the 1984),it would offshorescale being 200-300kmat 30'S). Note too that the low wind speedsalso mean for the fact t that the Froude number of the flow is subcritical(i.e. particle or wind speedsare less argumentof Gi than the longwaveor Kelvin-wavephase speeds given in Table 1, for example).Lin- sincethe local r earizing(5.la-c) and eliminatingthe horizontalvelocitiesyields an equationfor w: h,: dissipateas a + a,,+ = (a,,+ (s.2) cannotexplain Gf2ntvz f')* fz)w, it can only ex whereV2 is the horizontalLaplacian operator, 6u:02lAP andR is the internalRossby than a high. 1,1,47 SOUTHAFRICAN COASTAL LOW radius.Theboundaryconditionofnonormalflowatthecoastthenimplies w,,*fw,-=0 atx:0' (s.3) Fourier and Laplacetransform techniques to Although it is possiblethrough the useof (5.2) and (5.3) (see,lor example,Thomson 19?0), obtain a generalunutytl" sot,ition ro underthe sub-inertial,longwave conditions it is more instructiveto considerthe equations characteristicof the coastallow' In otherwo I scaleis much greaterthan the across-shore and sincethe lypical timescaleis a day or t (5.2)can be ndected in comparisonwith becomes u...,.- R-lrn,: -R-2w, (5.4) which has the generalsolution: ( ft - - (2R)-t{exp(x/n) w: A(y,r) exp(x/R)+ B(y,r) erp(-x/R) J" "*n(-t7p)w,dx (s.s) - exp(-xlR) dx) J "*P1rln)'v. respectto 'r' whereA(y,t)unO a(y,r) are constantswith assumeB(y,t):0. Then the time- since the no* rnoll'o uanistrfar from the coast, from the boundarycondition (5'3) dependentalong-shore structure A(y, r) is determined suchthat ' 1o -/R(w)."}d'r = (5'6) + - explx/R){(',), lzn o' A, fRA, |

Equation(5.6) is of the form: u, * cu,- = f(Y, t) .g. Whitham 1974),represents a forcedwave ;:'ll:iild#:fu*r:}ht',ffir exactlythat found in the more generalmodel of Thomson(1970) mentioned above' w.. For w, negative,i.e. subsiding'(5'5) Now considerthe effectsof the signof is to causea relativedecrease in w and' indicatesthat the contiibution of the beig winds , inversionheight' In other words' a relative er berg-windconditions' This loweringof the on of iyclonic vorticity derivedabove shows her than coastal-highformation' sociatedwith coastal-lowpropagation (CLW ,roc€SSoffers a more satisfactoryexplanation high is formed than the heuristicnonlinear oi, *"t" viewed as being likely to intensify ;"Y.::r;l; ::iilT[',l':ltxi] il:f"11 rmonlygenerated and coastal highs very rarely: iiiturbance will tend to be a low rather it can only explainwhy a propagatlngcoastal than a high. 11,18 C. J. C. REASON and M. R. JURY

The influenceof periodicvariations in the bergwinds is now considered.Motivation for consideringperiodic forcing follows from the observationof a distinct6-day cycle in the surfacepressure spectra measured at coastalSouth African sites(Preston-Whyte and Tyson 1973;Kamstra 1987) and the identificationof this cyclewith coastal-lowdynamics (Gill1977; Bannon 1981).Thus, the berg-windsubsidence velocity is expandedinto a Fourier series,one componentof which is written as w,: W(x,y)exp(-iAt) (s.7) wherethe frequencyQ correspondsto a period of 6 days. The spatialdependence W(*,y) canbe determinedto someextent from the sounding data analysedin sections3 and 4. In all cases,the coastallow at a given station was precededby berg winds near the inversionlevel which usually occurredfor 24 hours prior to the coastal-lowarrival. Once the coastal low passedthrough, these winds generallyshifted within a day or so to the oppositedirection, depending on the closeness of the approachingfrontal system.With this in mind, the along-shoredependence of the berg-windsubsidence velocity can be written as w(r.y) = w(x) expii(ky+ e)) (s.8) Figure6. The free ( wherek is the wavenumberand e is the phaselag correspondingto the bergwind leading the coastallow by 1 day. As far as the offshoredependence is concerned,all that is and (5.15). availableis that the effectsof berg-windevents have been regularlyobserved 30km or -7.29x 10-ss-r, so offshore from the south coast (Hunter 1987)and, occasionally,as far as lCIkm plotted in Fig. 6, (Shannon1985). Thus, it is assumedthat displacement,as W(x): Wo,aconstant for -s (x < 0 (s.e) :0 for x<-a wherea is giventhe value50km. In this paper Hence, the inversiondisplacement has the form: examinedto indi disturbancethat - h: h(x)exp{i(ky a4} (s.10) mountainsof the so that (5.2) and (5.3) become The hypothes Kelvin waveis - - -R-2w* h,, R-2h exp{i(n/2 + e)} (s.11) the three events and interface,the sur the shift in the wi ch,-fh=0 atx=0 (s.r2) Kelvin wave where betweenthe obse speedof a linear c: elk: (g,H)t/z W* : Wo(l- Ar/f)lA. (s.13) Forcing of tl The solutionto equation(5.11) is scaleanticyclone the interior platea h : Hoexp(.r/R)- w* exp(-i5.rr/6){exp(-xlR) + exp(xlR)-21 -a < x 10 (5.14) or berg wind. cyclonicvorticity and high formation. - - h = IHo W* exp(-i5nl6)lexp(2alR) 2exp(alR)+l)lexp(;/R) , < -o (5.15) In eachcase, convexbend in t where H. is the initial inversiondisplacement at the coastand the value e= -nl3 with what is correspondingto a l-day phasedifference has been used.To seethe magnitudeof the 1977a;I977b) effectof the berg-windforcing, typical values of the parametersare substitutedin (5.14) phasespeed of the It49 SOUTH AFRICAN COASTAL LOW

0

- 10c e E -roo E o o 6 .E-soo I' c o 6 b -aoo

- 500

Distanceotfshore (km) andberg'wind forced linear Kelvin- Figure6. The free (unforced)linear Kelvin-wave.solution (r'.t,ffil:.*")

Wo= -0'05ms-r' = and (5.15).These are:Ho=l5pm, R= 137km' a=50km' ''f -i.Zi"lO ts-r, Q:1.2xi0-ss-r correspondingto a pelod of 6 days.The resultsare wind on the inversion ptotteOin Fig. 6, fro,,, which it is clear thit the effect of the berg bisplacement,as revealedby this rather crude model, is substantial.

6. DlscusstoN maps have been In this paper, atmosphericsounding data and synopticweather is a shallow mesoscale examinedto indicate thai the coastallow of southernAfrica around the coastal ' ,lg-1,'--,.Jti.:i::; disturbance that propagatescoherently in an eastwardsdirection mountainsof the subcontinent. The hypothesisof Gill (1977)that the coastallow is essentiallya coastally-trapped atmosphereduring Kelvin *un" i, reinforcedUy ttre iime-height variationsof the lower of the inversion the three eventsexamined here. It wasobserved that the displacement height at 1000mb and interface, the surface pressureas measuredby the geopote-ntial with an internal the shift in the winds fottowinga coastal-lowpissag" were all consistent agreementwas found Kelvin wave propagatingon the inversion. In addition, good the theoreticalphase betweenthe observel piopugutlonspeed of the coastallow and

the ridging of an eastward-movingSynoptic- ent. This ridgingcaused easterly winds over le escarpmentbecame a warm offshore flow rived from the berg wind together with the rt leadsto a coastal-lowrather than a coasthl- high formation. In each case,the coastallow was obsr convex bend in the coastalmountains nea with what is expectedtheoretically for bot 1977a;tg77b) Kelvin waves' These theori' phasespeed of the waveshould decreasean< 1150 C. J. C. REASONand M. R. JURY

the resolutionof the at a concavebend (bay) the reverseshould occur. Unfortunately, the theory data is not adequateto'determine whether this in fact occurs.Nevertheless. speedof the doesoffer a poisible explanationfor the observationthat the propagation Capesouth coast coastallow is often lessat Cape Town than that measuredalong the severalhundred kilometresto the east(cl-w 1984;Hunter 1987). boundary Other influenceson coastal-lowbehaviour include variations in the surface the east(warm conditions.Examples are the seasurface temperature variations between uprvellinginshore) egufhur current) and the west (cold Bengueiacurrent offshore and Thesevariations are coasts,local topographiceffects and land-seafrictional contrasts' coastal-lowpropa- likely io be resionsiUlefor the observeddifferences (Table l) in the gationspeed. " Comparisonof the southern African coastallow with similar coastally-trapped and western disturbaniesthat propagatein the marinelayers of south-ea-stern.Australia the interaction_of North America is of lnierest. Thesedisturbances are also forcedby Dorman 1987) the synopticflow with the coastalmountains (Holland and Leslie 1986; they arecoastal ridges rather but in sucha way that the inversionis alwaysraised. .i.e' the Australianand than lows. Consistentwith the hypothesisidvanced in this paper, by warm' offshore North American events are not iieceded in their formation stages onshore(Reason windslike the coastallow; insteadthe pre-disturbanceflow is generally 'i:: t ' , C' )4. 198e). about5 events itttrougfr coastalridging events appear quite commol t-ntl1T"t with the western per month in both southleaiternAuitralia (Holland and Leslie 1986)and has not LJnitedStates (Mass and Albright 1987),their existenceduring other seasons of the coastal beenreported. Thus, theseevenis cannot be regardedas dominant features low' sinceit climateof theseregions. On the other hand, the southernAfrican'coastal mustbe considered is observedat all timesof year on a fairly regularbasis (cl-w 1984)' low is intimately asa climatologicalphenorn"non of the r;gion. Note too that the coastal (section2) andthus occurs associatedwith the weeklycycle of the southernmid-latitudes whereasthe periodically(Preston-whyte and Tyson 1973;CLW 1984;Kamstra 1987) ioastal ridging eventsdo not appearto showthis regularity' African coastal Sincethe coastallow is tuitt u dominantfeature of the southern hasexplored weather,there is considerablemotivation for future research'This study that there may the most prominent aspectsof coastal-lowbehaviour and hasillustrated variationsin bottom be second-ordereffects on the dynamics.These effects are due to friction and surfaceheat fluxesand needto be investigatedfurther.

ACKNOWLEDGEMENTS of this We thank Dr. D. G. Steyn for helpful discussionsduring the preparation paper. {$r*&^ RereReNces by synoptic-scalesaves' Q' J' R' Anh, N. N. and Gill, A. F. l98l Generationof coastallows

Bannon,P. R.

Clarke,A. J. SOUTH AFRICAN COASTAL LOW 1151 resolutionof the 1984 South African CoastalLow Workshop,1984: Abstracts and t cLw the theory Summary. Institute of Maritime Technology,Simons- leless, town. SouthAfrica 'The tion speedof the de Wet, L. W. 1979 influenceof the SouthAfrican plateauon the generation Capesouth coast and developmentof atmosphericpressure features and circulations'.MSc Thesis(in ). Universityof Pretoria [urfaceboundary 1984 The dynamicforcing of coastallows. Pp 6.7 in SouthAfrican in the east(warm CoastalLow Workshop,1984: Abstracts and Summary. Institute of Maritime Technology.Simonstown. South rwellinginshore) Africa se variationsare Dorman,C. E. 1985 Evidenceof Kelvin wavesin California'smarine layer and rastal-lowpropa- relatededdv generation.Mon. WeatherReu.,ll3,827- 839 1987 Possiblerole of gravitl,currents in NorthernCalifornia's coastal Poastally-trapped summerwind reversals.J. Geophys.Res., 92, 1497-1506 alia andwestern cill, A. E. 1977 Coastallytrapped waves in the atmosphere.Q. t. R. Meteorol. he interactionof Soc.,103,431-410 1982 Atmosphere-OceanDvnamics. AcademicPress i; Dorman 1987) Griffiths,R. W. 1986 Gravity currentsin rotatingsystems. Ann. Reo.Fluid Mech., ;tal ridgesrather 18. 59-89 'A i Australianand Heydenrych,C. M. 1987 climatologyof thecoastal low in theSW Cape'. MSc Thesis, I warm, offshore Holland,G. J. and Leslie,L. M. 1986 Ducted coastalridging over South East Australia. Q. l. R. rnshore(Reason Meteorol.Soc., ll2,'l3l-7 48 'The Hunter,I. T. 1987 Weatherof theAgulhas Bank andthe Cape South Coast'. CSIR ResearchReport 634, Stellenbosch, th about5 events Ikeda,M. 1984 Coastalflows driven by a local density flux. J. Geophys.Res., and the western E9, 8008-8016 seasonshas not Kamstra,F. 1987 Interannualvariability in the spectraof the dailysurface press- ure at four stationsin the SouthernHemisphere. Tel/us, les of the coastal 39A. 509-514 stal low, sinceit LeBlond,P. H. and Mysak, L. A. 1978 Wauesin the Ocean.Elsevier. Amsterdam istbe considered Mass,C. F. and Albright, M. D. 1987 Coastal southerliesand alongshoresurges of the West Coast of North America:Evidence of mesoscaletopographically low is intimately trappedresponse to synopticforcing. Mon. WeatherReu., I and thusoccurs u5, 1707-1738 p7) whereasthe Miles,J. W. 1972 Kelvin waveson oceanicboundaries. J. Fluid Mech., 55, ll3- r37 1917 Diffraction of solitary waves.Zeit. ang. Math. Phys. ?3,889- African coastal 902 dy hasexplored Preston-Whyte,R. A. and Tyson, 1973 Note on pressureoscillations over SouthAfrica. Mon. Weather P. D. Reu.,l0l,65M59 'Coastally I that there may Reason,C. J. C. 1989 Trapped Disturbancesin the lower atmosphere'. rtionsin bottom PhD Thesis.University of BritishColumbia, Vancouver, Canada 'Diffraction Reason,C. J. C. and Steyn,D. G. 1988 Comment on of solitary Kelvin wavesat Cape Mendocino'.Mon. WeatherReu., 116,804-805 Seitter,K. L. and Muench, H. S. 1985 Observationof a cold front with rope cloud. ibid.,ll3,84O- 848 Shannon,L. V. 1985 South African OceanColour and Experiment.Sea paration of this FisheriesResearch lnstitute, Cape Town Simpson,J. E. 1987 Grauity Currens: In theEnuironntQnt and the Laboratory.Ellis Horwood, Chichester Smith,R. B. r979 The influence of mountains on the atmosphere.Adu. Geophys., 21, 81-230 Taljaard,J. J. 1972 Synopticmeteorologlr of the SouthernHemisphere. Meteorol. Monogr.,13, no. 35. AmericanMeteorological Society, waves.p. .I. R. Boston Thomson,R. E. 1970 On the generationof Kelvin-type waves by atmosphericdis- double Kelvin turbances.l. Fluid Mech., 42, 657470 313-32'7 Whitham,G. B. 19'74 Linear and Nonlinear Waues.Wiley (Interscience)New York coastal '.,7, 23t-247 alongan irregu-