1. Introduction In environmental conditions of convective or latent instability any mechanism able to produce initial ascent from low levels or the ascent in block of the whole air-column can lead to the release of convection. Orographic upslope forcing or horizontal wind con- vergence can be good mechanisms to provide initial ascent from low levels (Doswell, 1982; Bluestein, 1993). To sustain convective precipitation long enough to accumulate large quantities of rainfall, a feeding current of relatively warm and wet air is necessary to replace the water removed by precipitation. On the other hand, a relatively warm and wet low-level inflow favours ver- tical instability and can even lead to the instability if the environment is only just stable, though not unstable, at the beginning. Therefore a very good place for collect- ing a large amount of precipitation in a few hours is the limiting boundary of an organised warm and wet low- level flow, where the convergence is guaranteed (Doswell, 1982). In general, this kind of bounded flow can readily be found in the eastward and poleward sec- tor of surface cyclones (Bluestein, 1993). The very high frequency of surface lee and/or thermal depressions in the Mediterranean often produces con- vergence bands, localised moisture supply areas or the combination of both, i.e., the bounded warm and wet flow just mentioned. In Part 1 of this work (Jansa et al., 2000) we identified this kind of mechanism as a con- tributory mechanism in one selected case of Mediterranean heavy rain and we have also mentioned its probable contribution to others. Here in Part 2 we try to determine whether the presence of a cyclone cen- tre close to the area affected by heavy rain is a system- atic factor or only an occasional one. That is, we need to consider as many heavy rain events as possible from a statistical approach. We can do this by using our own databases of cyclonic centres, heavy rain events and mesoscale convective systems (MCSs). Sections 2, 3 and 4 below are devoted to describing these databases. Sections 5 and 6 analyse the relationship or simultane- ity between heavy rain and cyclone centres, through cross-referencing information from the heavy rain and cyclone databases. In section 7 the relationship between MCSs and heavy rain and between MCSs and cyclones is studied. Finally, section 8 summarises the results of this work. 2 Cyclone databases The word ‘cyclone’ is usually used in mid-latitudes in its common synoptic sense, referring to the low pres- sure centres formed by baroclinic instability processes, Meteorol. Appl. 8, 43–56 (2001) Western Mediterranean cyclones and heavy rain. Part 2: Statistical approach Agustin Jansa 1 , Ana Genoves 1 , M Angeles Picornell 1 , Joan Campins 1 , Ricardo Riosalido 2 and Olinda Carretero 2 1 Centro Meteorológico en Illes Balears, Instituto Nacional de Meteorología, E-07071 Palma de Mallorca, Spain 2 Servicio de Técnicas de Análisis y Predicción, Instituto Nacional de Meteorología, E-28071, Madrid, Spain This is the second part (statistical approach) of a work concerning the relationship between heavy rain and cyclonic centres in the western Mediterranean. Using a statistical approach we seek to verify the indirect role of the cyclone centres in locating, triggering or focusing heavy rain: a cyclonic centre – even if neither strong nor deep – may contribute to the low-level flow organisation and so to the creation or intensification of a low-level warm and wet current that can feed and sustain convective rain in favourable environmental conditions. The foundation for the statistical approach is several Several databases maintained and updated by the Spanish Institute of Meteorology (INM). These databases of cyclonic centres, heavy rain and strong wind events, and mesoscale convective systems (MCSs), all covering the western Mediterranean or part thereof, can be cross-referenced by looking for simultaneity between heavy rain events (or MCSs) and cyclone centres in the vicinity. We have found that in most of the heavy rain events (around 90%) there is a cyclonic centre in the vicinity, usually located so that its presence favours the creation or intensification of a feeding flow of Mediterranean air towards the area affected by heavy rain. The same occurs with the MCSs. 43
Transcript
1. Introduction
In environmental conditions of convective or latentinstability any mechanism able to produce initial ascentfrom low levels or the ascent in block of the wholeair-column can lead to the release of convection.Orographic upslope forcing or horizontal wind con-vergence can be good mechanisms to provide initialascent from low levels (Doswell, 1982; Bluestein, 1993).
To sustain convective precipitation long enough toaccumulate large quantities of rainfall, a feeding currentof relatively warm and wet air is necessary to replacethe water removed by precipitation. On the other hand,a relatively warm and wet low-level inflow favours ver-tical instability and can even lead to the instability if theenvironment is only just stable, though not unstable, atthe beginning. Therefore a very good place for collect-ing a large amount of precipitation in a few hours is thelimiting boundary of an organised warm and wet low-level flow, where the convergence is guaranteed(Doswell, 1982). In general, this kind of bounded flowcan readily be found in the eastward and poleward sec-tor of surface cyclones (Bluestein, 1993).
The very high frequency of surface lee and/or thermaldepressions in the Mediterranean often produces con-vergence bands, localised moisture supply areas or the
combination of both, i.e., the bounded warm and wetflow just mentioned. In Part 1 of this work (Jansa et al.,2000) we identified this kind of mechanism as a con-tributory mechanism in one selected case ofMediterranean heavy rain and we have also mentionedits probable contribution to others. Here in Part 2 wetry to determine whether the presence of a cyclone cen-tre close to the area affected by heavy rain is a system-atic factor or only an occasional one. That is, we needto consider as many heavy rain events as possible froma statistical approach. We can do this by using our owndatabases of cyclonic centres, heavy rain events andmesoscale convective systems (MCSs). Sections 2, 3 and4 below are devoted to describing these databases.Sections 5 and 6 analyse the relationship or simultane-ity between heavy rain and cyclone centres, throughcross-referencing information from the heavy rain andcyclone databases. In section 7 the relationship betweenMCSs and heavy rain and between MCSs and cyclonesis studied. Finally, section 8 summarises the results ofthis work.
2 Cyclone databases
The word ‘cyclone’ is usually used in mid-latitudes inits common synoptic sense, referring to the low pres-sure centres formed by baroclinic instability processes,
Meteorol. Appl. 8, 43–56 (2001)
Western Mediterranean cyclones and heavy rain.Part 2: Statistical approachAgustin Jansa1, Ana Genoves1, M Angeles Picornell1, Joan Campins1, Ricardo Riosalido2 andOlinda Carretero2
1 Centro Meteorológico en Illes Balears, Instituto Nacional de Meteorología, E-07071 Palmade Mallorca, Spain2 Servicio de Técnicas de Análisis y Predicción, Instituto Nacional de Meteorología, E-28071,Madrid, Spain
This is the second part (statistical approach) of a work concerning the relationship between heavy rainand cyclonic centres in the western Mediterranean. Using a statistical approach we seek to verify theindirect role of the cyclone centres in locating, triggering or focusing heavy rain: a cyclonic centre – evenif neither strong nor deep – may contribute to the low-level flow organisation and so to the creation orintensification of a low-level warm and wet current that can feed and sustain convective rain infavourable environmental conditions. The foundation for the statistical approach is several Severaldatabases maintained and updated by the Spanish Institute of Meteorology (INM). These databases ofcyclonic centres, heavy rain and strong wind events, and mesoscale convective systems (MCSs), allcovering the western Mediterranean or part thereof, can be cross-referenced by looking for simultaneitybetween heavy rain events (or MCSs) and cyclone centres in the vicinity. We have found that in most ofthe heavy rain events (around 90%) there is a cyclonic centre in the vicinity, usually located so that itspresence favours the creation or intensification of a feeding flow of Mediterranean air towards the areaaffected by heavy rain. The same occurs with the MCSs.
with a typical scale of the order of 1000 km. However,in this work we will also include as cyclones orcyclonic centres any kind of surface depressions, evensmall, weak and shallow low centres of orographic orthermal origin.
In this sense, many low-level Mediterranean cycloniccentres are indeed small and shallow and easily misrep-resented in large-scale analyses and forecasts. Onlymanual analyses or high-resolution objective analysesare able to detect and describe them.
Two databases of cyclones in the westernMediterranean are used in this work. The first is basedon subjective manual analyses, where the detection anddescription of the cyclonic centres are subjective. Thesecond is based on objective high-resolution analyseswhere the detection and description of the cyclones areobjective.
2.1. Subjective cyclone database
The subjective database covers the period fromDecember 1991 to November 1995, during which onlya large-scale operational weather forecasting suite wasused in the Spanish National Meteorological Institute(INM). The suite, based on the LAM–INM model, wasan adaptation of the ECMWF grid point model run-ning at a horizontal resolution of 0.91 degrees in lati-tude and longitude, both in analysis and forecasting.Due to this large grid size, LAM–INM was unable todetect many of the surface Mediterranean cyclones wewere interested in catching. Therefore we chose the useof subjective hand-made analyses to create a cyclonecatalogue.
To perform a manual analysis significantly better thana large-scale objective analysis is not so straightforwardin the western Mediterranean area, because it is a regionwith a poor density of surface observations, particu-larly in the marine and North African zones. Analystshave to keep in mind that analysis is not the same asinterpolation (Doswell, 1982) and they have to evaluatethe data, look for help from non-conventional informa-tion (satellite images, for instance) and use their ownknowledge about conceptual models of the mesoscalestructures (orographic, kynematic or thermal) thatoccur frequently in the region (Jansa, 1990). As a con-sequence, manual analysis is subjectively non-homoge-neous. Usually the zone closest to the analyst is betterunderstood and therefore better analysed than moredistant areas.
Our subjective database of cyclones was extracted fromfour years of daily manual surface analyses at 00 and 12UTC. These were operational analyses, re-analysed bysome of the authors. From a visual inspection of everychart, all the low-pressure centres were considered, notonly those that appeared with closed isobars, but also
some relative minima supposed to correspond to a sig-nificant maximum of geostrophic vorticity (open struc-tures or troughs). The location of the centre, the centralpressure value and the pressure values at eight sur-rounding points were stored. A filter consisting of athreshold for vorticity was used to reject non-signifi-cant structures. The threshold was low (0.8×10−4 s−1 ofcentral average geostrophic vorticity obtained with agrid length of 200 km) because we wished to retain evenquite weak cyclones within the database. A completestatistical−climatological analysis of westernMediterranean surface cyclones from our subjectivecyclone database can be found in Campins et al. (2000).
The geographical area explored for cyclone centres isshown in Figure 1. Within this area, a large number ofcyclones are encountered − more than 6000 in fouryears, with 1609 being the yearly average. To comparecapabilities, a database of cyclones objectively identi-fied from LAM–INM analyses was compiled parallel tothe subjective one. LAM–INM detected slightly morethan the half as many cyclones as the subjective method(946 per year) in an area slightly larger than the first(see Figure 1).
2.2. Objective cyclone database
In 1995 the operational model at the INM was replacedby the HIRLAM model (see Gustafsson, 1991), run-ning with a horizontal resolution of 0.5o in latitude andlongitude. The boundary conditions are given by theglobal ECMWF model. Analysis in the operationalchain is based on an optimum interpolation scheme, thefirst guess being the 6-hour forecast provided by theprevious run of the HIRLAM–INM. It is a multivari-ate three-dimensional analysis for the mass and windfield and univariate for humidity (Diaz-Pabón, 1996).
The method used here to identify a cyclone in theobjective analyses consists of looking for relative min-ima in the whole mean sea-level pressure grid, i. e., anygrid point at which the pressure value is lower than theeight pressure values at the grid points surrounding it.In the following step, the pressure values are interpo-lated to a geometrically regular 50 km grid and thepressure gradients in the eight main directions aroundthe provisional cyclone centre are explored in the newgrid until a distance of D km from the centre. If theaverage pressure gradient reaches G hPa/100 km fromsomewhere to the centre in N or more of the eight maindirections the provisional cyclone is included. In othercase the cyclone is rejected. The values of D, G and Ncan be selected as desired.
Although depending on the selection of D, G and N,the above criterion may be considered roughly equiva-lent to the visual detection of any closed low or opentrough in a subjective analysis (an exact translation isnot possible).
A half year period (summer and autumn 1995) of over-lapping manual analyses, LAM–INM analyses andHIRLAM–INM analyses was used to check the abilityof the new model to detect cyclones (Picornell et al.,1997). To do so, the following values were taken: D =1000 km, G = 0.1 hPa/100 km, N = 4. Some low centresso detected (objectively or subjectively) can be veryweak. Therefore, to make possible the comparisonbetween the results from both methods, the same filteras in the subjective method was applied after detection:the geostrophic vorticity computed by finite differ-ences with a grid length of 200 km has to be bigger than0.8×10−4 s−1. Figures in Table 1 give an indication of theresults from the comparison. Note that only the com-mon area of exploration (that is, the area used in thesubjective method, Figure 1) has been considered inthis comparison.
It can be inferred from Table 1 that HIRLAM–INM0.5° has overcome the inability of LAM–INM 0.91° todetect small and weak cyclones. In fact,HIRLAM–INM 0.5° is able to detect more cyclonesthan the subjective method. When looking at details inthe comparison (not given here, see Picornell et al.,1997) it is found that the differences are not randomlydistributed and not imputable to a model tendency to
produce spurious lows, but they are mainly centred inthe regions usually least explored by the human ana-lysts. When focusing on the ‘central areas’ of the analy-ses, direct comparisons between methods show thatmost of the lows are detected with equal success (seeexamples in Picornell et al., 1997).
When using a more restrictive criterion for detection(D = 850 km, G = 0.5 hPa/100 km and N = 6) and thesame threshold for vorticity, the number of cyclonesdetected objectively (943 in summer 1995 and 419 inautumn 1995) decreases, closely approaching the sub-jective results (we had hoped to detect at least the samecyclones and others besides). When avoiding the filterfor vorticity, the figures increase by around 20%(model depressions with average vorticity within 200km lower than 0.8 s−1 are around 20% of the total).Taking into account the fact that any objective analysistends to be smoother than a subjective one, we considerthat these are the best results.
Therefore in the present work we have used the morerestrictive parameters for the detection (D = 850 km, G= 0.5 hPa/100 km and N = 6), without filtering forvorticity. The objective database based onHIRLAM–INM has been used for the periodDecember 1995 to November 1996, the window ofexploration of cyclones being defined by 12oW to 18oEand 29oN to 49oN.
The database of cyclones used in this work is thereforemixed, with subjective analyses (December 1991–November 1995) and objective HIRLAM/INM analy-ses (December 1995–November 1996) being made overa total of five years. Two daily analyses, at 00 and 12UTC, are considered in both sections.
Western Mediterranean cyclones and heavy rain – statistical approach
45
Figure 1. Area of study. The polygonal frame (dashed line) indicates the area of exploration of cyclones by the subjective method(see text). The rectangular frame (full line) marks the area of exploration by the objective method with the old coarse gridLAM–INM model. Shaded areas are mountainous terrain over 500, 1000, 1500, 2000 and 2500 m high.
Table 1. Number of cyclones in the westernMediterranean using different analyses.
Analysis Number of cyclones
Subjective LAM–INM HIRLAM–INManalysis 0.91° 0.5°
Summer 1995 523 170 1234Autumn 1995 423 123 617
3. Database of heavy rain events
Our database of heavy rain is principally built on theco-operation of several national (and regional) weatherservices and is consequently quite heterogeneous.
Figure 2 shows the area covered by the heavy rain data-base. A substantial part of the western Mediterraneancoastal regions is included. The whole area is dividedinto territorial units of different sizes: in southernFrance the territorial units are the administrativedepartments; in Spain they are the provinces, except forthe Balearic Islands where they are islands; in Algeriathe territorial units are climatological areas used by theOffice Nationale de la Météorologie; and in northernItaly they are regions. Ticino (in southern Switzerland)and the islands of Corsica and Sardinia are alsoincluded in the database.
Except in northern Italy, where only GTS (WMOGlobal Telecommunication System) data have beenused, non-GTS data have been used to compile thisdatabase. Non-GTS rainfall data have been providedby the respective national (or regional) meteorologicalservices. In some cases non-GTS rainfall data are avail-able only as 24-hour accumulated precipitation during
the ‘pluviometric’ day. The pluviometric day runs from07 (or 06 or 08) UTC on one day to the same time onthe next. Although partial data were available for someregions, only data on 24-hour accumulated precipita-tion (over the pluviometric day) have been consideredin this work.
The general threshold considered in our database fordefining a heavy rain event is 60 mm/24 h. The excep-tion is Algeria, where the threshold has been lowered to30 mm/24 h. We define a heavy precipitation event in aterritorial unit as a day in which the correspondingthreshold has been exceeded at any weather stationtherein. The number of events during a period in everyterritorial unit depends not only on the degree of inci-dence of heavy rain in the territory but also on theextension of the territorial unit and the density of avail-able weather stations in it. Although we have receivedor collected data for the whole period December 1991to November 1996 in some territorial units, the datafrom other units are partial and refer to only one or afew years within the period. Therefore the numbers ofevents in different territorial units are not comparable.A true climatology of heavy rain cannot be inferredfrom our database. Note also that several events cancorrespond to the same date if several territorial unitsare affected by heavy rain at that date. Table 2 givessome details about our database. To simplify, theperiod December 1991 to November 1992 is indicatedas 1992 and so on.
4. Mesoscale Convective System database
During autumn, from 1989 to 1994 and since 1997, theINM’s Forecasting and Analysis Techniques Servcie(STAP) has stored objective information from satelliteinfra-red (IR) images about Mesoscale ConvectiveSystems (MCSs). These have been defined by Houze(1993) as ‘a cloud system that occurs in connectionwith an ensemble of thunderstorms and produces acommon contiguous precipitation area on the order of100 km or more in at least one direction’. MCSs aresupposed to be potentially related to heavy rain andflash flooding, as has actually occurred in some severecases (e.g. Rivera & Riosalido, 1986; Riosalido et al.,1998; Ramis et al., 1994; Senesi et al., 1996).
The database was constructed from half-hourlyMeteosat IR images during the months of September,October and November. Using the INM McIDAS sys-tem facilities, a quantitative analysis was performedalong the complete life cycle of every system. The areaexplored was a window of the IR image defined by10°W to 15°E and 33°N to 51°N (the Iberian and west-ern Mediterranean area).
The initial criterion used to define a convective systemin the database was that its cloud-shield (as defined bythe –32°C cloud-top temperature) must extend at least
A Jansa et al.
466
Figure 2. Territorial units as used in the heavy rain database(see text). BA = Barcelona (Spain). TA = Tarragona (Spain).LL = Lleida (Spain). GI = Girona (Spain). CS = Castellón(Spain). VC Valencia (Spain), AL = Alicante (Spain). MU =Murcia (Spain). AM = Almería (Spain). GR = Granada(Spain). MA = Málaga (Spain). PM = Mallorca (Spain). MH =Menorca (Spain). IB = Ibiza (Spain). R1 = Oran (Algeria). R2= Algiers (Algeria). R3 = Annaba (Algeria). R4 = Salada(Algeria). R5 = Bousaada (Algeria). R6 = Constantine(Algeria). 20 = Corsica (France). 11 = Carcassone (France). 13= Marseille (France). 83 = Toulon (France). 06 = Nice(France). 66 = Perpignan (France). 30 = Nîmes (France). 34 =Montpellier (France). 84 = Avignon (France). 04 = Digne(France). 05 = Embrun (France). 48 = Mende (France). SA =Sardinia (Italy). LI = Liguria (Italy). LO = Lombardy (Italy).PI = Piedmont (Italy). TI = Ticino (Italy).
100 km in one direction. After quality control, a selec-tion criterion (combined size and duration) was definedto keep only those systems ranging within a predefinedscale. The final criterion combines aspects of theFritsch et al. (1986) and Augustine et al. (1988) criteria.The MCSs that meet the following two conditions dur-ing its life cycle are kept as selected MCSs and form thedataset: (a) the size of the –52°C area must be greaterthan 10000 km2 and (b) this size must be maintained forat least three hours.
A summary of the climatalogy of MCS in the Iberianand western Mediterranean can be found in Riosalido(1997) or Riosalido et al. (1998).
Only a subset of the information about MCS has beenused for the present work. The subset consists of thelocation (latitude and longitude), plotted every threehours, of the centre of any MCS included in the maindataset, independent of the state of evolution of thatMCS. The same MCS can therefore appear several
Western Mediterranean cyclones and heavy rain – statistical approach
47
Table 2. Heavy rain data
Country Territorial unit Precipitation data Number
Code Name Extension Density Type Provider Periodof events
France 04 Digne Small High Non-GTS Météo-France (DIRSE1) 1993–96 1405 Embrun Small High Non-GTS Météo-France (DIRSE1) 1993–96 906 Nice Small High Non-GTS Météo-France (DIRSE1) 1993–96 3213 Marseille Small High Non-GTS Météo-France (DIRSE1) 1993–96 1683 Toulon Small High Non-GTS Météo-France (DIRSE1) 1993–96 2984 Avignon Small High Non-GTS Météo-France (DIRSE1) 1993–96 2011 Carcassone Small High Non-GTS Météo-France (DIRSE1) 1993–96 3130 Nimes Small High Non-GTS Météo-France (DIRSE1) 1993–96 7034 Montpellier Small High Non-GTS Météo-France (DIRSE1) 1993–96 3848 Mende Small High Non-GTS Météo-France (DIRSE1) 1993–96 3566 Perpignan Small High Non-GTS Météo-France (DIRSE1) 1993–96 2120 Corse Moderate Moderate Non-GTS Météo-France (DIRSE1) 1993–96 49
Spain BA Barcelona Moderate High Non-GTS INM2 (CMT3 Catalunya) 1992–96 34GI Girona Moderate High Non-GTS INM2 (CMT3 Catalunya) 1992–96 36LL Lleida Moderate High Non-GTS INM2 (CMT3 Catalunya) 1992–96 18TA Tarragona Moderate High Non-GTS INM2 (CMT3 Catalunya) 1992–96 22AL Alicante Moderate High Non-GTS INM2 (CMT3 Valencia) 1992–96 44CS Castellon Moderate High Non-GTS INM2 (CMT3 Valencia) 1992–96 25VC Valencia Moderate High Non-GTS INM2 (CMT3 Valencia) 1992–96 57MU Murcia Moderate High Non-GTS INM2 (CMT3 Murcia) 1992–96 19AM Almeria Moderate Moderate Non-GTS INM2 (CMT3 Andalucia Or.) 1992–95 6GR Granada Moderate Moderate Non-GTS INM2 (CMT3 Andalucia Or.) 1992–95 7MA Malaga Moderate Moderate Non-GTS INM2 (CMT3 Andalucia Or.) 1992–95 16IB Ibiza Very small Very high Non-GTS INM2 (CMT3 Illes Balears) 1992-96 15
MH Menorca Very small Very high Non-GTS INM2 (CMT3 Illes Balears) 1992–96 9PM Mallorca Small Very high Non-GTS INM2 (CMT3 Illes Balears) 1992–96 82
Algeria R1 Oran Large Low Non-GTS Office Nationale de la Météorologie(Algeria) 1992–96 8
R2 Alger Large Low Non-GTS Office Nationale de la Météorologie(Algeria) 1992–96 30
R3 Annaba Large Low Non-GTS Office Nationale de la Météorologie(Algeria) 1992–96 40
R4 Salda Large Low Non-GTS Office Nationale de la Météorologie(Algeria) 1992–96 5
R5 Bousaada Large Low Non-GTS Office Nationale de la Météorologie(Algeria) 1992–96 0
R6 Constantine Large Low Non-GTS Office Nationale de la Météorologie(Algeria) 1992–96 16
Italy SA Sardegna Large Moderate Non-GTS SAR4 19967 18LI Liguria Large Low GTS MAP Data Centre5 1992–968 9LO Lombardia Large Low GTS MAP Data Centre5 1992–968 24PI Piamonte Large Low GTS MAP Data Centre5 1992–968 5
1DIRSE=Direction Interregionale du Sudest2INM=Instituto Nacional de Meteorología (Spanish National Institute of Meteorology)3CMT=Centro Meteorológico Territorial4SAR=Servizio Agrometeorologico Regionale della Sardegna5Mesoscale Alpine Programme Data Centre, ETH, Zürich, Switzerland6SMA=Schweizerische Meteorologische Anstalt (Swiss Meteorological Institute)7Data for 1996 were provided by SAR. A few data for other years were taken from MAP Data Centre and they are only GTS data8As collected by the MAP Data Centre, that is, only MAP-Seasons, from June to November
times in our MCS subset in nearly the same or in dif-ferent locations.
5. Heavy rain versus cyclone centres
5.1. The method
The databases of cyclones and heavy rain events havebeen cross-referenced so that we can search for simul-taneities between heavy rain and the close presence of acyclone centre. To do so, every heavy rain event storedin the heavy rain database is considered. The cyclonedatabase is explored looking for any cyclone centreregistered with the same date as the heavy rain event at12 UTC and with the following date at 00 UTC. Thatis, if the heavy rain event corresponds to day D, theanalyses that are explored correspond to day D at 12UTC and to day D+1 at 00 UTC, both times beingwithin the pluviometric day D and no other analysiscorresponding to the same pluviometric day. If severalcyclone centres meeting this condition are found, onlythe closest to the geographical centre of the territorialunit affected by heavy rain is stored.
Unfortunately we do not know the time of the heaviestrain – this information is not available from our data-base, as explained above – and therefore we can onlyspeak of rough simultaneity between cyclone presenceand heavy rain. Moreover, if a cyclone centre was actu-ally influencing or partially determining the heavy rainit may have been a cyclone other than the closest. Inany case, there is no guarantee of the effective influenceof the closest cyclone on the heavy rain. What we willhave been obtaining is only a statistical result aboutsimultaneity between heavy rain and cyclones, that is,answers to the following question: How frequently isthere a cyclone centre roughly close to an area withheavy rain and where is this cyclone usually located?
We have explored this kind of simultaneity for everyheavy rain event in every territorial unit. A summary ofour results is shown in Table 3. For each territorial unit(see Figure 2) with 15 or more cases of heavy rain, wegive the frequency of cyclone presence relative to thecases of heavy rain within various distances (radii) fromthe site (geographical centre of the territorial unit)where the heavy rain occurs.
5.2. Frequency of simultaneity of cyclone andheavy rain, and relative location
Note that the frequency of cyclone presence within aradius of 1000 km from the site of heavy rain is almost100% for all the territorial units, but due to the veryhigh frequency of cyclones in the westernMediterranean this result cannot be considered as verysignificant. That a cyclone can influence a heavy rainoccurrence at a distance of more than 600 km is quitehard to accept; it becomes more credible within a radius
of 600 km. Over this distance, the frequencies rangebetween 85 and 100%, except for southern Spain(Almeria, Granada, Malaga: AM, GR, MA), Algeriaand Switzerland. In any case, we know that somecyclones, even those very close to the heavy rain area,are not actually affecting the heavy rain. A typical caseis the southern Pyrenean orographic lows, very fre-quent in synoptic northern situations (Bessemoulin etal., 1993), which would automatically be selected as thelow closest to the heavy rain site for some heavy rainevents in Catalonia (BA, GI, LL, TA) and even in someother places. They are probably not directly related tothe heavy rain formation. But in most cases there doesappear to be a relationship between the occurrence ofheavy rain and the presence of a cyclone. In order toclarify this point, we need to know where the cyclonecentre is placed in relation to the heavy rain site. Inaddition, in order to make comparisons, we have useda random sample of dates (seasonally consistent withthe Mediterranean heavy rain climatology).
(a) Establishing the location of a cyclone centre
To establish where the cyclone centre is placed in rela-tion to the heavy rain site we have constructed a vectorthat links the heavy rain site and the closest cyclonecentre for every case of heavy rain. Management of thisvectorial magnitude is non-trivial. If the distribution ofcyclone centre locations around a site of heavy rain wasGaussian (or normal) the complete distribution of loca-tions would be given by the true vectorial average loca-tion (not included) and the true vectorial dispersion(bi-dimensional standard deviation) (also notincluded). This is not the case and so we have lookedfor other ways of presenting results numerically andgraphically. As is known, a direct vectorial average ofvectorial magnitudes smoothes the result, especially innon-normal distributions. To avoid this problem wehave preferred to obtain a ‘dominant vector’ instead ofthe mean vector, as is usual in wind climatology. Themean length of the dominant vector is obtained sepa-rately, as a scalar average of all the observed lengths,and the dominant direction is obtained as a vectorialaverage of the unit vectors associated with each indi-vidual vector. In this way the average distance and thedirection of the cyclone centre from the heavy rain siteare obtained. The dominant directions are indicated asusual in meteorology (cyclone centre to the north ofthe heavy rain site is indicated by 0 or 360°, to thesouth, 180°, and so on). Both results are included inTable 3, which also gives results for groups of sites.
(b) Comparisons based on a sample of dates
In order to have a reference for comparison of the resultsof Table 3, we have constructed a random sample ofdates that is roughly consistent with the climatologicalseasonal distribution of heavy rain in the Mediterranean.The sample consists of 135 dates: 27 per year, 12 inautumn (September to November), 6 in winter
A Jansa et al.
48
(December, January and February), 6 in spring (Marchto May) and 3 in summer (June to August). For autumnwe have used 5 September at 00 UTC, 10 September at12 UTC, 15 September at 00 UTC, 20 September at 12UTC, 5 October at 00 UTC, and so on; for winter, 10December at 00 UTC, 20 December at 12 UTC, 10January at 00 UTC, and so on; for spring, as for winter;for summer, 15 June at 00 UTC, 15 July at 12 UTC, etc.Considering every one of the aforementioned dates as aheavy rain event, we have explored the cyclone databaselooking for the closest cyclone to every site. The resultsfor a selection of sites are include in Table 4. Randomly,
a cyclone is found within a radius of 600 km from ‘06’ in56% of dates (Table 4). In cases of heavy rain, this fre-quency increases to 94% (Table 3). On average, the ran-dom frequency of appearance of a cyclone within 600km from any site is 53%, but the corresponding fre-quency in cases of heavy rain is 88% for the same set ofsites. They seem to be quite significant differences.
Some significant differences also concern the dominantlocations of the closest cyclone to some of the sites. Forinstance, the dominant location for the closest cycloneto Mallorca (PM) for the random set of dates is around
Western Mediterranean cyclones and heavy rain – statistical approach
49
Table 3. Simultaneity between heavy rain and cyclones. Groups of sites are: France-E (04, 05, 06, 13, 83, 84),France-W (11, 30, 34, 48, 66), Catalonia (BA, GI, LL, TA), Valencia-Murcia (AL, CS, MU, VC), Andalusia (AM,GR, MA), the Balearic Islands (IB, MH, PM), Algeria (R1 to R6), northern Italy (LI, LO, PI) and northern Italyplus Ticino. The locations representing the groups of sites, as well as the frequencies of simultaneity and theaverage distance and direction of the cyclone centre, are obtained as weighted averages, the weight of eachterritorial unit in the group being the number of cases of heavy rain in it.
Country Territorial Number of Relative frequency (%) of cyclone Average location ofunit code cases of presence within a given radius from cyclone centre from
heavy rain the heavy rain site heavy rain site
≤400 km ≤600 km ≤1000 km Distance Direction(km) (degrees)
Country Territorial Number of Relative frequency (%) of cyclone Average location ofunit code random presence within a given radius from site cyclone centre from site
LO 135 26 40 64 554 207Switzerland TI 135 24 36 63 578 198
400 km away to the north-west (Table 4) − i.e. south ofthe Pyrenees, where there is a very high frequency oflee depressions − but in the case of heavy rain the dom-inant location of the closest cyclone to the same site is229 km to the south-south-west (Table 3), in theAlgerian Sea: these are the cyclones that could actuallyinfluence the occurrence of heavy rain in Mallorca.
5.3. Graphic presentation
To visualise the main results of Table 3 and some oth-ers which are not presented, Figure 3 has been con-structed. For a selection of individual sites or groupsthe ‘dominant’ location of the closest cyclone centre tothe heavy rain area appears as the centre of an ellipse,whose axes give an idea of the dispersion of locations.The radial axis is the standard deviation of the distancesbetween heavy rain site and cyclone centre. The trans-verse axis, Rt, is given by:
where:
R—
= average distance between heavy rain site andcyclone centreσux
= standard deviation of the x-component of u→
σuy= standard deviation of the y-component of u→
u→ = unitary vector of the direction between heavy rainsite and cyclone centre
The smaller the ellipse in Figure 3, the more represen-tative or confident are the results concerning the mostfrequent location of the cyclone centre simultaneouswith heavy rain. Thus the most reliable results are thoseobtained for northern Italy and Ticino, France-E,Valencia-Murcia and the Balearic Islands. A little lesssignificant is the cyclone location simultaneous withheavy rain in France-W and Catalonia (we have alreadymentioned the spurious effect of some Pyrenean lows)and in Andalusia. Still more spread are the cyclone
locations for the cases in Corsica and Sardinia. Themost spread cyclone distribution corresponds to theheavy rain events in Algeria.
The dispersion of the location of the closest cyclone ismuch larger for the set of random dates than for the setsof heavy rain events. So for Mallorca (PM) the axes ofthe ellipse are 186 km (radial) and 373 km (transverse)for the random set and 140 km and 226 km for theheavy rain set. For Corsica (‘20’), the correspondingfigures are 272 km and 391 km, instead of 232 km and248 km, etc.
Figure 4 shows the true location distribution of thecyclone centres simultaneous with the four sites that canbe considered the most interesting, due to the high num-ber of events in them: France-30 (Nîmes), France-20(Corsica), Spain-VC (Valencia) and Spain-PM (Mallorca).
5.4. Intensity, size and zero vorticity band ofcyclones occurring simultaneously with heavy rain
Concerning the intensity or depth of the cyclones, wehave computed an average intensity of all the closestcyclones occurring simultaneously with heavy rain forevery site and group of sites. We define the intensity ofthe cyclone as the difference in sea-level pressurebetween the surroundings and the centre of the cycloneusing two grid sizes (or distances from the surroundingsto the centre): 200 and 400 km (Campins et al., 2000). Ingeneral, the cyclones are weak or moderate, with aver-age intensity values of 2.5 hPa (grid size 200 km) and 3.9hPa (grid size 400 km), which correspond to averagepressure gradients of 1.3 hPa/100 km within a range of200 km and 1.0 hPa/100 km for 400 km from thecyclone centre. The corresponding average vorticitiesare ζ200 = 2.0×10−4 s−1 (from 200 km around to the cen-tre) and ζ400 = 0.8×10−4 s−1 (from 400 km around to thecentre). The important difference between both averagevorticities suggests a small size for the average cyclone.
A Jansa et al.
50
Rt R= +σ σu2
u2
x y
(In the hypothesis of sinusoidal pressure distribution,the cyclone radium would be around 250–300 km, if theradius of the cyclone is defined by the circle of zero vor-ticity or maximum pressure gradient.) Noting that themost common distance between heavy rain sites and theassociated cyclone centres is around 250–300 km, if thecyclone was sinusoidal we could imagine heavy rainoccurring in the cyclonic peripheral band, where thepressure gradient is maximal.
Differences are observed at some sites. For instance, forTicino ζ200 = 2.9×10−4 s−1 and ζ400 = 1.2×10−4 s−1, i.e. thecyclones associated with heavy rain are a little moreintense and slightly larger. Note that in these cases theaverage distance between the heavy rain site and thecyclone centre is 390 km.
The average values for the whole subjective cyclonedatabase are very similar (Campins et al., 2000), andtherefore the cyclones that have been observed in asso-ciation or simultaneity with heavy rain are not espe-cially deep or intense in relation to all the cyclones ordepressions observed in the region.
5.5. Discussion of the main results
The existence of a cyclone centre near a heavy rain siteis much more frequent than when heavy rain does notexist. On the other hand, the location of cyclone cen-tres around heavy rain sites is not randomly distributedalthough there is important dispersion. Some locationsare preferred over others. In general, there is very goodagreement with the conceptual model described above:the cyclone centres usually associated (simultaneous, infact) with heavy rain are located so that they favour theinflow of Mediterranean air towards the area of heavyrain. This feeding flow is southerly for northern Italyand Ticino, south-easterly for France, and easterly forCatalonia, the Balearics and Valencia-Murcia.
Note that, owing to the orographic structure of thewestern Mediterranean (see Figure 1, Part 1, Jansa et al.,2000), in many of the places considered here the coastalor inner relief intersects the feeding flow and forcesupward motion and orographic rain. Also note that theorographic mechanism may be very effective in cases ofconvective instability, providing the initial ascent to
Western Mediterranean cyclones and heavy rain – statistical approach
51
Figure 3. Location of the closest cyclone centre in the case of heavy rain atdifferent sites. From left to right and from top to bottom the sites areFrance-W, France-E, northern Italy and Ticino, Catalonia, the BalearicIslands, Corsica, Valencia-Murcia, Algeria, Sardinia and Andalusia. Blackdots indicate the heavy rain site and the most frequent location of thecyclone centre. The size of the axis of the ellipse indicates the dispersion inthe location of cyclone centres.
release the convection. In cases of instability and veryhumid inflow, a quite moderate ascent (a moderaterelief) should be enough.
In many of the Mediterranean heavy rain events, theorographic component is relevant with the cycloneonly having an indirect effect. It is the mechanism thatforces the warm and humid flow against the mountains.But in other cases the convergence within the low –usually concentrated in some frontal or non-frontalbands – is enough to provide the initial ascent in thefeeding current. The cyclone is then providing the feed-ing flow and also some convergence zones.
The results obtained here agree with the extreme casesdiscussed in Part 1 of this work (Jansa et al., 2000). Notethe location of the low centres in the cases of Piedmont-Liguria, Ticino, Gandia (Valencia) or Tarragona(Catalonia) in Figures 5, 7, 8 and 9 of Jansa et al. (2000)and compare them with Figure 3 of this study.
Romero et al. (1999a) derived eight characteristic rain-fall spatial patterns for 449 torrential rain events in theSpanish Mediterranean area (defined as when at least2% of stations registered more than 50 mm) during theperiod 1964–93. Romero et al. (1999b) investigated therelationship between the previously extracted heavyrain patterns and some synoptical patterns obtainedfrom ECMWF analyses, the most significant fields inthe definition of the synoptic patterns being the 500and 925 hPa geopotential fields. Some of their resultsare quite consistent with our present findings. In
particular, in their synoptic pattern 12, the presenceof a cyclone in the Algerian marine region at lowlevel is very clear and this synoptic pattern is closelyconnected with their heavy rain pattern 5 (i.e. heavyrain in Valencia and the Balearics). Their heavy rainpattern 7 (heavy rain in Catalonia) is in good correla-tion with their synoptic pattern 8, in which a cyclonecentre at low level in the Catalonian−Valencian areais also clear. Some indication of the presence of alow-level cyclone centre between the Algerian andSpanish coasts also appears in some of the synopticpatterns corresponding to heavy rain in the areas ofValencia and Murcia, although this is not a clear result.Poor consistency between our results and those ofRomero et al. (1999b) could stem from the inabilityof the analyses they used (resolution 0.75° latitude–longitude) to detect small cyclone centres. In any case,the two works can be considered as complementaryand globally consistent.
6. Heaviest rain events
In order to detect possible particularities in the eventsof heaviest rain, we repeated the analysis explained inthe preceding section for the cases of rainfall over 100mm/24 h. Table 5 summarises the main results. Becausecases of heaviest rainfall are much less frequent thancases of heavy rain (60 mm/24 h), not all the individualsites or groups are included in Table 5. Only the resultsfor sites or groups with ten or more events of precipi-tation over 100 mm/24 h are shown.
A Jansa et al.
52
Figure 4. Location of the closest cyclone centre simultaneous to heavy rain in selected sites (black dot), in squares of 200×200km around the site. In % of the total of cases of heavy rain, expressed in grey scale: less than 1% (white), 1–2%, 3–5%, 6–10%,11–20%, 21% or more (darkest).
Almost all the results included in Table 3 and Table 5are quite similar. There are only a few differencesworth mentioning. For Catalonia (BA, GI, LL, TA) thedistance between heavy rain and a cyclone increases a little in the cases of heaviest rain, and the cyclonelocation turns nearly 30° to the south-south-east. Thisis probably due to a partial cancellation of the spuriouseffect of the Pyrenean lee lows: in the cases of heaviestrainfall in Catalonia the associated low is located moreconsistently than in the cases of less heavy rain.
For the cases in northern Italy (mainly in Lombardy),the results for the heaviest cases are worse than in gen-eral: the frequency of simultaneity between heavy rainand cyclone within 600 km decreases from 89% to 73%and the average distance increases from 315 to 430 km.Probably there are a few cases of very heavy rain innorthern Italy (among only 11) that are not influencedat all by any Mediterranean low.
The results for the intensity of the cyclones associatedwith events of heaviest rainfall are also very similar tothose obtained in general. The average values of vortic-ity are ζ200 = 2.1×10−4 s−1 and ζ400 = 0.8×10−4 s−1.
7. MCS and heavy rain: simultaneity withcyclones
7.1. Relationship between MCSs and heavy rain
The first thing we have done is to check the relationshipbetween MCSs and heavy rain, although this aspect isnot very relevant for the scope of this paper.Furthermore, it is not possible to obtain completeresults pertaining to this question, mainly because most
of the MCSs are over water, at least during part of theirlife cycle (Riosalido et al., 1998). In any case, we fol-lowed two methods to achieve partial results.
First of all, we selected the cases of heavy rain thatoccurred in the same period for which MCS data areavailable (autumn 1992, 1993 and 1994). For each of theselected events (remember that an event refers to onedate and one territorial unit) we explored the MCSdatabase looking for any simultaneous MCS centredwithin a radius of no more than 150 km from the heavyrain site. Although an MCS may have a significant spa-tial extension and the maximum rain intensity does notneed to coincide with the MCS centre, we have consid-ered it quite unlikely to have heavy rain associated withan MCS at distances greater than 150 km from the MCScentre, allowing for exceptions. Note that in theMediterranean MCS, the mean radius of the –52 °Carea is around 120 km and the mean length of the majoraxis of the –32 °C area is 200 km (Riosalido et al., 1998).To determine the simultaneity, we have analysed theMCS locations at 09, 12, 15, 18 and 21 UTC of theheavy rain date (D) and at 00, 03 and 06 UTC of D+1,in accordance with the definition of the pluviometricday. If an MCS is found at less than 150 km from theheavy rain site at any of the times indicated, we con-sider that there is simultaneity between heavy rain andMCS, although this introduces another uncertainty,because heavy rain could have occurred at a differenttime of the pass or approximation of the MCS, foranother cause.
Only 19 of the 352 heavy rain events of the selected sub-set (autumn of 1992, 1993 and 1994) of our heavy raindatabase have been simultaneous with the presenceof a MCS close enough to the heavy rain site to be
Western Mediterranean cyclones and heavy rain – statistical approach
53
Table 5. Simultaneity between very heavy rain (≥ 100 mm/24 h) and cyclones.
Country Territorial Number of Relative frequency (%) of cyclone Average location ofunit code cases of rain presence within a given radius from cyclone centre from
100mm/24h heavy rain site heavy rain site
≤400 km ≤600 km ≤1000 km Distance Direction(km) (degrees)
reasonably sure of a true relationship. Only 5.4%of the events of heavy rain in the Mediterraneancan be related to the presence of an MCS. If only theheaviest rainfall events (more than 100 mm/24 h) areconsidered, the ratio increases to 13:101; i.e. 12.9% ofthe cases of heaviest rainfall can be associated withMCS presence.
From the inverse point of view, simultaneous heavyrain in the vicinity of 56 individual MCSs was sought.Without giving more details, we concluded that in atleast half the cases of MCSs passing across or near a ter-ritory there is heavy rain (in some places).Furthermore, in over half the positive cases the rainfallat some points surpasses 100 mm/24 h. In fact, some ofthe most important heavy rain events in our databasewithin the sub-period 1992–94 were associated withMCSs. For example, the 240 mm/24 h registered in thedepartment of Avignon (France) on 21 September 1992during the ‘Vaison-la-Romaine’ flash flood event(Senesi et al., 1996), the 282 mm/24 h of precipitationon 27 September 1992 in the region of Liguria, the 301mm/24 h on 22 September 1993 in the department ofNîmes, and the 390 mm/24 h on 10 October 1994 in theprovince of Tarragona (Ramis et al., 1998; Romero,1998; see also Figure 9 of Jansa et al., 2000).
In spite of the inaccuracy of the above approaches, itcan be asserted that most MCSs do produce heavy rainand some of them produce very heavy rain, althoughonly a few of the quite frequent cases of heavy rain inthe Mediterranean are related to MCS.
7.2. Relationship between cyclones and MCSs
The results of cross-referencing the contents of ourMCS and cyclone databases are presented below. Whatwe want to know is how often a cyclone centre can befound in the vicinity of a MCS in a relative locationcompatible with the idea that the cyclone centre is amechanism that contributes to forming or sustainingthe MCS. The methodology is almost the same as insections 5 and 6, with two main differences: the simul-taneity between MCS and cyclone can be exact and themarine or terrestrial locations of MCS are equally con-sidered.
Here we have considered the locations of MCS centresonly at 00 and 12 UTC. For each MCS centre at a cer-tain time on a certain date we have explored the cyclonedatabase looking for the closest cyclone centre atexactly the same time and date. To do so, we haveseparated the MCS centre locations in four spatialwindows defined by 35°N to 40°N/3°E to 5°W (south-west sector), 35°N to 40°N/10°E to 3°E (south-eastsector), 40°N to 45°N/3°E to 5°W (north-west sector)and 40°N to 45°N/10°E to 3°E (north-east sector). Thedata we have obtained are summarised in Table 6 andFigure 5.
Although the relative frequency of cyclone centressimultaneous with MCSs within a radius of 600 kmfrom the MCS centre is lower (71%) than for the heavyrain study, the present results can still be consideredsignificant if they are compared with the cyclone fre-quencies for a ‘random sample of dates’ (see Table 4)and are quite consistent with the previous ones andwith our conceptual model.
When studying the environmental conditions in whichMCSs develop in the Mediterranean, Riosalido et al.(1998) found that MCSs frequently appear in low-levelair-mass boundaries, where the warm and wet inflowtends to increase or sustain instability and where con-vergence exists. In the poleward and eastward sector ofa cyclone centre these conditions can easily be reached.This is another argument for the consistency of ourpresent results. Note that an important part of theMCSs appear on the sea and the orographic factor can-not be argued.
8. Conclusions
Although the direct and immediate cause of heavy rainis usually convection, a front, the orographic ascent ora combination of some of these and other synoptic andmesoscale mechanisms, a possible concurrent factor forheavy rain in the Mediterranean is the presence of alow-level cyclone centre that contributes to floworganisation and in particular to the establishment of awarm-wet destabilising and feeding inflow current.
This hypothesis was checked in Part 1 (Jansa et al.,2000) through consideration of particular cases and inparticular through the detailed analysis of one of thesecases, the ‘Piedmont flood event’.
In Part 2 we have tried to check the hypothesis from astatistical approach. Databases of surface cyclones,heavy rain and MCSs have been prepared and cross-referenced looking for simultaneities between heavyrain and cyclones and between MCSs and cyclones.
In around 90% of all cases of heavy rain in theMediterranean (more than 60 mm/24 h of precipitation,948 cases in five years) and of very heavy rain (morethan 100 mm/24 h of precipitation, 259 cases in fiveyears) and in 71% of all cases of MCS (44 cases in threeautumns) there is a cyclone centre located within 600km of the heavy rain site or the MCS centre. In a fewgeographical areas – Andalusia (southern Spain),Algeria and Ticino (Switzerland) – the frequency ofsimultaneity between heavy rain and a cyclone centrein the vicinity decreases to 80%. For some places(Valencia and the Balearic Islands, in Spain), this fre-quency increases to 99%. When only the heaviest rainevents are taken into account, the percentage of simul-taneity between heavy rain and a cyclone centre in thevicinity reaches 100% for Valencia and the Balearics
A Jansa et al.
54
and 97% for some French areas and Catalonia (Spain).The frequency decreases for northern Italy, although itremains over 70%.
The cyclones simultaneous with heavy rain are usuallyweak or moderate (0.8×10−4 s−1 of average geostrophicvorticity in an inner disc of 400 km of radius) and small.Their location is not random, the cyclone centre usu-ally being close – around 300 km – to the heavy rainsite, and in a location where it can produce significantinflows of Mediterranean air (warm and wet) into theheavy rain site.
Our statistical approach therefore reinforces the resultsof Part 1 of this work and gives consistency to our ini-tial hypothesis.
Consequently, the correct forecasting of heavy rain in thewestern Mediterranean entails very accurate prediction ofsurface cyclones, their intensity, shape and size and loca-tion, as well as the correct forecasting of direct factors.
Acknowledgements
Non-GTS precipitation data used in this work havebeen provided by J.C. Rivrain (DIRSE, Météo-France),Alejandro Martínez Albaladejo and Luis Vázquez(CMT Catalunya, INM, Spain), Victor Alcover andMercedes Alemán (CMT Valencia, INM, Spain),Fermín Gallego and Ramón Garrido (CMT Murcia,INM, Spain), José Luis Ramírez Olalla (CMTAndalucia Oriental, INM, Spain), Bachir Hamadache(Office Nationale de la Météorologie, Algeria),Alessandro Delitala and Piero Chessa (ServizioAgrometeorologico Regionale della Sardegna) andPaolo Ambrossetti (Osservatorio Ticinese, SMA,Switzerland). GTS precipitation data for northern Italyhave been obtained from the MAP Data Centre, ETH,Zürich. This work has been carried out in the frame-work of the Project CLI95-1780, partially sponsoredby the Spanish Plan Nacional de I+D, CICYT.
Western Mediterranean cyclones and heavy rain – statistical approach
55
Table 6. Simultaneity between MCS and cyclones.
Window Window Number of Central location Relative frequency Average location oflimits cases of MCS of MCS (%) of cyclone cyclone centre from
(lat.–long.) presence MCS centrewithin 600 kmof MCS centre Distance Direction
Figure 5. Location of the closest cyclone centre in case of MCS in different sectors. From left to right and from top to bottomthe sectors are north-west, north-east, south-west and south-east. Black dots indicate the average location of the MCS in eachsector and the most frequent location of the simultaneous cyclone centre. The size of the axis of the ellipse indicates the disper-sion in the location of cyclone centres.
References
Augustine, J. A. & Howard, K. W. (1988). Mesoscale convec-tive complexes over the United States during 1985. Mon.Wea. Rev., 116: 685−701.
Bessemoulin, P., Bougeault, P., Genoves, A., Jansa, A. &Puech, D. (1993). Mountain pressure drag during PYREX.Beitr. Phys. Atmos., 66: 305−325.
Bluestein, H. B. (1993). Synoptic-Dynamic Meteorology inMidlatitudes, Volume II: Observations and Theory ofWeather Systems. Oxford University Press, 594 pp.
Campins, J., Genovés, A., Jansà, A., Guijarro, J. A. & Ramis,C. (2000). A catalogue and a classification of surfacecyclones for the Western Mediterranean. Int. J. Climatol.20: 969–984.
Díaz-Pabón, R. (1996). General characteristics of the modelof analysis and prediction HIRLAM/INM (in Spanish). IVSimposio Nacional de Predicción [available from InstitutoNacional de Meteorología, Madrid].
Doswell, C. A. (1982). The operational meteorology of con-vective weather, Vol. I, Operational mesoanalysis. NOAATech. Mem. NWS NSSFC-5, 157 pp.
Fritsch, J. M., Kane, R. J. & Chelius, C. R. (1986). The con-tribution of mesoscale convective weather systems to thewarm season precipitation in the United States. J. ClimateAppl. Meteorol., 25: 1333−1345.
Gustafsson, N. (1991). The HIRLAM model. In SeminarProc. on Numerical Methods in Atmospheric Models, Vol.II, ECMWF, 115−146.
Houze, R. (1993). Cloud Dynamics. InternationalGeophysics Series, Vol. 53, Academic Press, 573 pp.
Jansa, A. (1990). Notes on meteorological mesoscale analysis inlow atmospheric levels (in Spanish), Instituto Nacional deMeteorología, Madrid, 70 pp.
Jansa, A., Genoves, A. & Garcia-Moya, J. A. (2000). WesternMediterranean cyclones and heavy rain. Part 1: Numericalexperiment concerning the Piedmont flood case. Meteorol.Appl., 7: 323–334
Picornell, M. A., Genoves, A., Jansa, A. & Navascues, B.(1997). Mesocyclones detectability by using operationalobjective analyses. In Proc. of INM/WMO Symposium onCyclones and Hazardous Weather in the Mediterranean,
MMA-INM/UIB, Palma de Mallorca, 55−62 [availablefrom Instituto Nacional de Meteorología, Palma deMallorca].
Ramis, C., Llasat, M. C., Genoves, A. & Jansa, A. (1994). TheOctober 1987 floods in Catalonia: synoptic and mesoscalemechanisms. Meteorol. Appl., 1: 337−350.
Ramis, C., Romero, R., Homar, V., Alonso, S. & Alarcon, M.(1998). Diagnosis and numerical simulation of a torrentialprecipitation event in Catalonia (Spain). Meteorol. Atmos.Phys., 69: 1−21.
Riosalido, R. (1997). Mesoscale convective systems in thewestern Mediterranean, a satellite view. In Proc. ofINM/WMO International Symposium on Cyclones andHazardous Weather in the Mediterranean, MMA-INM/UIB, Palma de Mallorca, 353−359 [availablefrom Instituto Nacional de Meteorología, Palma deMallorca].
Riosalido, R., Elizaga, F., Carretero, O. & Martin, F. (1998).Satellite climatology of mesoscale convective systems in thevicinity of the Iberian Peninsula: application to forecastingtorrential rain (in Spanish). Nota Técnica STAP Nº 29,Instituto Nacional de Meteorología [available fromInstituto Nacional de Meteorología, Madrid].
Rivera, A. & Riosalido, R. (1986). Mediterranean convectivesystems as viewed by Meteosat. A case study. 6th MeteosatScientific Users’ Meeting, Amsterdam 25−27 Nov. 1986,EUM P. 01.
Romero, R. (1998). Numerical simulation of mesoscaleprocesses in the Western Mediterranean: EnvironmentalImpact and Natural Hazards. PhD. Thesis, Universitat deles Illes Balears (UIB).
Romero, R., Ramis, C. & Guijarro, J. A. (1999a). Daily rain-fall patterns in the Spanish Mediterranean area: an objectiveclassification. Int. J. Climatol., 19: 95−112.
Romero, R., Sumner, G., Ramis, C. & Genovés, A. (1999b).A classification of the atmospheric circulation patternsproducing significant daily rainfall in the SpanishMediterranean area. Int. J. Climatol., 19: 765−785.
