journal of hydrology - unipd.it...periods and areas for the island of crete, can be used as basic...

Journal of Hydrology xxx (2010) xxx–xxx

ARTICLE IN PRESS

Contents lists available at ScienceDirect

Journal of Hydrology

journal homepage: www.elsevier .com/ locate / jhydrol

Seasonality of floods and their hydrometeorologic characteristics in the islandof Crete

Aristeidis G. Koutroulis, Ioannis K. Tsanis *, Ioannis N. DaliakopoulosDepartment of Environmental Engineering, Technical University of Crete, Chania, Greece

a r t i c l e i n f o

Article history:Available online xxxx

Keywords:SeasonalityFloodsAtmospheric circulationCrete

0022-1694/$ - see front matter � 2010 Elsevier B.V. Adoi:10.1016/j.jhydrol.2010.04.025

* Corresponding author. Fax: +30 28210 37849.E-mail address: [email protected] (I.K. Tsanis).

Please cite this article in press as: Koutroulis, A.(2010), doi:10.1016/j.jhydrol.2010.04.025

s u m m a r y

The seasonality of the hydrometeorologic characteristics of floods that occurred in Crete during the per-iod 1990–2007 is presented. Hydrological characteristics were analyzed using seasonality indices basedon a dataset of 53 daily precipitation stations as well as 15 daily and 7 monthly recording flow stations.The atmospheric circulation conditions during the flood events were examined based on a joint subjec-tive classification and meteorological analysis. The flood event-based seasonality was found to coincidewith the seasonality of the daily precipitation maxima of November and December. The seasonality of thethree largest long term daily precipitation maxima indicates that 50% of the maximum precipitationevents occur from November to January (NDJ period). Analysis showed that the maximum annual streamflows in Crete are lagging by approximately 1 month from annual maximum daily precipitation in theregion. The circulation type classification of the flood events showed that most of the weather systemsoccurring in the Mediterranean and passing over Crete have SW, NW and W direction. For the majorityof the events, a common mean sea level pressure gradient field was observed over Europe. This compar-ison of the seasonality of selected hydrometeorologic characteristics reveals valuable information withinthe context of flood occurrence.

� 2010 Elsevier B.V. All rights reserved.

1. Introduction

Flood timing and magnitude data are used in local, seasonal andregional flood frequency analyses for estimating design flood val-ues required in engineering design and planning (Cunderlik et al.,2004a; Ouarda et al., 2006). On a larger spatial scale, the hydrolog-ical characteristics such as precipitation and stream flow depict theregional climate mechanisms and therefore the seasonality of floodoccurrence is strongly connected to the climate forcing mecha-nisms of each region. The knowledge of weather patterns associ-ated with extreme rainfall and runoff events can serve as areliable early flood warning system and a non-structural approachfor flood mitigation. Especially for the case of events such as flashfloods, that are seldom predictable and have severe consequences,understanding the ‘‘how” and the ‘‘when” can be indispensible forcivil protection.

Storm events are often associated with the development of low-pressure systems. Over the Mediterranean these systems areapproaching from three main directions which can be roughly dis-tinguished as West (W), Southwest (SW) and Northwest (NW).According to Barry and Chorley (2003), depressions that enter theMediterranean from the Atlantic Ocean (W source) and baroclinic

ll rights reserved.

G., et al. Seasonality of floods a

waves from the Atlas mountain range (SW source) account for 9%and 17% of the low-pressure systems respectively. The remaining74% form at the lee of the Alps and Pyrenees (NW source). This isa weather classification mentioned in literature (e.g. Barry andChorley, 2003) with the effects of each class being well documented.In general, a weather classification method can be an algorithm orconcept used for type definition and assignment of objects to thosetype. Any method (algorithm) may be applied in different varia-tions, e.g. using different distance metrics, numbers of types, spatialdomains, temporal configurations or input parameters. Weathertypes may be regarded as the driving forces of the hydrologic sub-systems which together form the complex hydrologic system of acatchment (Van de Griend and Seyhan, 1984), and thus the identifi-cation of dominant types can be used in a hydrologic context.

Flood seasonality can be effectively described in terms of direc-tional or otherwise orientation statistics (Mardia, 1972 and Fisher,1993). Bayliss and Jones (1993) also described previous flood sea-sonality in Great Britain by means of seasonality indices derivedfrom directional statistics. Castellarin et al. (2001) combined floodand rainfall seasonality descriptors and basin relative permeabilityin a regional model for ungauged basins in Northern Italy. Cunder-lik and Ouarda (2009) studied the trends in the timing and magni-tude of seasonal maximum flood events across Canada takinginto account the directional character and multi-modality offlood occurrences. Magilligan and Graner (1996) used directional

nd their hydrometeorologic characteristics in the island of Crete. J. Hydrol.

http://dx.doi.org/10.1016/j.jhydrol.2010.04.025

mailto:[email protected]

http://www.sciencedirect.com/science/journal/00221694

http://www.elsevier.com/locate/jhydrol


2 A.G. Koutroulis et al. / Journal of Hydrology xxx (2010) xxx–xxx

ARTICLE IN PRESS

statistics to express some of the hydroclimatological and geomor-phic controls on flooding in New England and to reveal the regionalgroupings of similarly responding basins. The authors concludethat directional statistics are an appropriate method for depictingregional hydrologic regimes and in describing some of the hydro-climatological controls on flood occurrence.

The objective of this paper is to narrow down specific seasonaland climatic characteristics that can be associated with extremerainfall and runoff in the island of Crete. For this reason, largeand smaller scale meteorological characteristics that precede ex-treme events in the area have been identified and analyzed. The di-rect relation of large and small scale weather structures withregards to flash flood generation is not clear (e.g. even though a sig-nificant number of flash flood generating storms have occurred un-der the presence of a given large scale structure, the presence ofthis structure is not alone indicative of a flash flood occurrence).Possible correlations between extreme daily and monthly rainfalland runoff with larger scale phenomena will allow for a betterunderstanding of the flood occurrence process. Determining floodoccurrence seasonality, and spatial extent and therefore flood riskperiods and areas for the island of Crete, can be used as basic buteffective flood hazard mitigation and risk management planning.At the same time, this kind of information may be used to supporta simplified flash flood alert system, based on the computationof critical rainfall threshold values, as shown by Martina et al.(2006) and Norbiato et al. (2008).

This paper is original in that it combines different approaches ofseasonality in order to identify flood occurrence characteristics forthe island of Crete. For one of the approaches a small modificationis proposed so that results have a better uniformity and are easierto evaluate. The advantages of a multi-scale approach are shown.The analysis presented in this paper is aimed towards a betterunderstanding of extreme events over Crete.

2. Methodology

2.1. Circulation types

A widely used atmospheric circulation classification system forEuropean conditions is that developed by Baur et al. (1944). The socalled Grosswetterlagen and later HB-GWL classification (afterHess and Brezowscy) is based on large scale weather observationssuch as the location of semi-permanent pressure centers (e.g. thepoles of the North Atlantic oscillation) the location and extensionof frontal zones and the presence of cyclonic or anti-cyclonic con-ditions. Since its introduction, this classification method has been

Table 1Order of weather type classification according to the Hess and Brezowsky Grosswetterlag

HBGWT GWL Definition HB

01 West Wa Anti-cyclonic westerly 06Wz Cyclonic westerly

WS South-shifted westerly 07WW Maritime westerly (Block E.Europe)

02 SWest SWa Anti-cyclonic south-westerly

SWz Cyclonic south-westerly 08

03 NWest NWa Anti-cyclonic north-westerlyNWz Cyclonic north-westerly

04 HighCE HM High over central europe 09BM Zonal ridge across Central Europe

05 LowCE TM Low over central europe 1006 North Na Anti-cyclonic northerly

Nz Cyclonic northerlyHNa Icelandic High, ridge Central EuropeHNz Icelandic High, trough Central Europe 11

Please cite this article in press as: Koutroulis, A.G., et al. Seasonality of floods a(2010), doi:10.1016/j.jhydrol.2010.04.025

revised (Hess and Brezowsky, 1952; Hess and Brezowsky, 1969and Hess and Brezowsky, 1977) and updated (Gerstengabe et al.,1999) and its variations are often used in literature (e.g. Caspary,1995 and Caspary, 1996). Table 1 shows the current HB-GWL clas-sification which reduces the initially 30 classes to 11, including atransitional class. Bearing in mind the inherent uncertainty ofany classification approach of the complex and dynamic large scalemeteorology (Yamal and White, 1987; Bardossy and Caspary,1990; Yarnal, 1993), the HB-GWL is considered as the best avail-able scheme (James, 2007) and is therefore used in this study.

2.2. Seasonality

Seasonality of hydrological time series can be effectively de-scribed in terms of directional statistics (Mardia, 1972; Fisher,1993; Magilligan and Graner, 1996). Following Bayliss and Jones(1993), Burn (1997) and Cunderlik et al. (2004b) the Julian dateof occurrence of an annual hydrologic event i can be transformedto a directional statistic:

hi ¼ ðJulian dateÞi2p365

� �ð1Þ

where hi is the angular value (in radians). The Julian date is Day 1 forJanuary 1st and Day 365 for December 31st. The estimated vectorhas a magnitude ri = 1 and a direction of hi radians. For a sampleof n events, individual directional statistics components can beaggregated in order to estimate the x- and y-coordinates of themean date of event occurrence (Burn, 1997):

�x ¼ 1n

Xn

i¼1

cosðhiÞ ð2Þ

�y ¼ 1n

Xn

i¼1

sinðhiÞ ð3Þ

The mean direction of the event dates is then obtained from(Burn, 1997):

�h ¼ tan�1 �y�x

� �ð4Þ

While the directional statistic of a single event has a unity mag-nitude, for the mean vector of n events, �r can be defined as themeasure of variability of the date of occurrence about the meandate (Burn, 1997).

�r ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�x2 þ �y2

pð5Þ

en (HB-GWL) Source: James, 2007.

GWT GWL Definition

North HB High over the British IslesTrM Trough over Central Europe

NEast NEa Anti-cyclonic north-easterlyNEz Cyclonic north-easterlyHFa Scandinavian High, Ridge Central Europe

East HFz Scandinavian High, trough Central Europe

HNFa High Scandinavia-Iceland, ridge Central EuropeHNFz High Scandinavia-Iceland, trough Central Europe

SEast SEa Anti-cyclonic south-easterlySEz Cyclonic south-easterly

South SA Anti-cyclonic southerlySZ Cyclonic southerlyTB Low over the British IslesTrW Trough over western Europe

U U Transition (no distinct type)



A.G. Koutroulis et al. / Journal of Hydrology xxx (2010) xxx–xxx 3

ARTICLE IN PRESS

Therefore, �r ranges from zero, denoting high variability or auniform distribution around the year, to unity, showing high regu-larity or that all events occur on the same day.

Another approach based on the Smax index introduced by Para-jka et al. (2009) describes the seasonality of the long-term meanmonthly characteristics of precipitation and runoff.

This method estimates the frequency with which the meanmonthly maximum of a given year occurred on the same monthas the long-term mean monthly maximum in that particular timeseries. For N years of mean monthly values, Smax is calculated inyears as:

Smax ¼XN

i¼1

PðM ¼ miÞ ð6Þ

where M is the month of the long-term maximum and mi is themonth of the maximum mean monthly value in year i. RespectivelyS2max is the mean monthly secondary maximum of N years. Theseasonality can be expressed as a vector much like the Burn index,with the direction representing the month M (January = East,April = North, July = West and October = South) and the magnitudeshowing the frequency Smax or S2max.

Herein, a modification of that approach is proposed, producingthe normalized index �Smax, that does not depend on the numberof years N for a sufficiently large dataset:

�Smax ¼PN

i¼1PðM ¼ miÞN

ð7Þ

�Smax takes values from 1N (tends to 0 for sufficiently large datasets), if

no annual maximum occurs on the same month as the long-termmaximum, to unity, if all annual maxima occur on the same month.Therefore, �Smax can be used as a measure event occurrence variabil-ity much like �r.

2.3. Case study

The island of Crete occupies the southern part of the country ofGreece (Fig. 1) and is divided into four prefectures from East toWest: Lasithi, Iraklion, Rethymno, and Chania. With an area of8265 km2, Crete covers almost 6.3% of the area of Greece. The meanelevation is 482 m and the average slope 228 m/km with thetopography fracturing into small catchments with ephemeralstreams and karst geology. Crete has a typical Mediterranean is-land environment with about 40% of the annual precipitationoccurring in the winter months while there is negligible rainfall

Fig. 1. Location of Crete Island and location of precipitation and flow station


during the summer. The average precipitation ranges from440 mm/year on the Plain of Ierapetra (Lasithi) to more than2000 mm/year on the Askifou upland (Chania), where orographiceffects tend to increase both in frequency and intensity of winterprecipitation (Naoum and Tsanis, 2004; Koutroulis and Tsanis, 2010).

Annual air temperature in Crete increases 1.5 �C from West toEast and 1 �C from North to South. Potential evapotranspirationvaries from 1370 to 1570 mm/year, representing 75–85% of themean annual precipitation in low elevation areas (less than300 m a.s.l.) and 50–70% in high elevation areas. Overall 70% ofprecipitation evaporates, 19% infiltrates and the remaining 11% isrunoff, almost irrespective of annual rainfall (Naoum and Tsanis,2004). The precipitation and ensuing runoff are greater in the norththan in the south because the prevailing wind direction is north-westerly. Generally, the western part of the island receives moreprecipitation than the eastern part (Naoum and Tsanis, 2004).

Daily precipitation data (which was then aggregated to monthlyvalues) was compiled by the WRDPC service (Water ResourcesDepartment of the Prefecture of Crete) for 67 precipitation stations.The stations mainly cover the eastern part of the island which has ahigher level of agricultural and tourism activity than the westernpart. Out of the entire dataset, 14 gauges recorded only for 4–10 years and were excluded from the analysis, while the rest hadsufficient measurements for 28–32 years. The gauges are locatedat elevations that range from sea level, in the prefecture of Iraklion(central Crete), to 905 m a.s.l, in the prefecture of Lasithi (easternCrete).

Mean monthly runoff values were available at 22 gauging sta-tions, while maximum annual flood records were available at 14gauging stations in Crete. Catchment area ranges from 22 to500 km2 with a median of 123 km2. Catchment average elevationranges from 300 to 1100 m a.s.l. with a median of 640 m a.s.l.The record length of all stations is at least 16 years in the period1970–2002 with an average of 28 years.

The analysis is based on a dataset observed in the period 1973–2005. It consists of mean monthly precipitation and annual maxi-mum daily precipitation at 53 precipitation stations, meanmonthly runoff at 22 stations and maximum annual floods at 15gauging stations. The precipitation and stream flow gauges are pre-sented in Fig. 1. The seasonality is discussed within the context ofits spatial and temporal variability and flood occurrence andmagnitude.

Evidence for 49 flood events were retrieved in collaborationwith the local Civil Protection Service Office of Crete Region forthe period 1990–2007 and used to produce regional maps of flood

s used for the seasonality analysis of their hydrological characteristics.




ARTICLE IN PRESS

occurrence at municipality level. Civil Protection Service is respon-sible for recording the impact cost, the location, the date and otherinformation of the flood events over Crete. A subset of these 49flood events was included in a European flash flood atlas duringa recent flash flood data compilation effort in the frame of HY-DRATE, EC funded research project (CN: 037024), (Gaume et al.,2009).

3. Results and discussion

3.1. Event-based seasonality analysis

The analysis of occurrence of the 49 flood events of the last dec-ade for the island of Crete is presented in Fig. 2. About 30% of allflood events occur in December. Regarding the seasonal distribu-tion more than 50% of the events occur during winter period whileabout 40% occur during autumn. Autumn events usually have thestronger impact due to higher rainfall rates and their relative tim-ing to dry-summer periods. The early timing and the unexpectednature of the autumn events finds sometimes the local authorities,responsible for the removal of riparian vegetation, unpreparedresulting to flooding due to stream blockages.

Regarding the spatial distribution of the events over Crete, 66%of them were reported for western Crete region, while the rest 34%were reported for eastern Crete (Fig. 3). The East to West precipi-tation gradient can be attributed to the regional NW to SE domi-nant meteorological atmospheric patterns and the higherelevation and steepest slope morphology of the western Crete incomparison to the eastern region of the island.

2

7

10

15

8

2

0

23

0 0 00

4

8

12

16

Sept

embe

r

Oct

ober

Nov

embe

r

Dec

embe

r

Janu

ary

Febr

uary

Mar

ch

Apr

il

May

June

July

Aug

ust

Num

ber

of F

lood

Eve

nts

Fig. 2. Monthly distribution of flood events in Crete (1997–2007).

Fig. 3. Spatial distribution and number of flood events (1990–2007) for the island of Crvalues were calculated using Log-Pearson Type III Distribution and the interpolation me


Aiming to a more thorough analysis at municipality level, thenumber of events was correlated with the 100-year return perioddaily precipitation, the mean elevation, the mean slope and themean aspect of each municipality. No significant trends were ob-tained through this analysis. A basin level analysis could result tomore valuable results but it was not feasible due to lack of infor-mation. Fig. 3 illustrates the number of flood events per municipal-ity and the 100-year return period maximum daily precipitation,showing that flood events are not clearly correlated with maxi-mum daily precipitation.

3.2. Atmospheric circulation types related to floods in Crete

The meteorological analysis is focused on the circulation pat-terns and the synoptic prevailing background conditions for heavy,localized rainfall over Crete. The atmospheric circulation condi-tions of the flood events since 2002 was determined according tothe HB-GWT subjective classification, based on data provided byCOST 733 Classification Catalogues (Huth et al., 2008) focusing tothe Eastern Mediterranean Region (defined as domain 10 withinCOST 733 including Balkans, SE Europe and Italy between 34N–49N and 7E–30E). The atmospheric circulation conditions of the2002 posterior events, was subjectively classified according tothe synoptic analysis based on NCEP/NCAR Reanalysis Project data-set (Kistler et al., 2001). Moreover, synoptic analysis based onNCEP/NCAR dataset was performed for all the events for classifica-tion homogeneity. METEOSAT images were used to support a moreconcrete explanation of the storm generation and propagation.High resolution visible channel METEOSAT images were used dur-ing daytime and infrared channel images were used during night-time for cloudy features (storm) tracking (Ottenbacher et al.,1997). The satellite images of 30 min temporal resolution wereprovided from the European Organization for the Exploitation ofMeteorological Satellites (EUMETSAT). The circulation patterns of22 flood events (Table 2) that affected 49 regions over Crete, duringthe 1990–2007 period, were diagnosed through the joint analysisof the analysis of large-scale atmospheric circulation fields, theMETEOSAT derived storm tracking and the intensity and spatialdistribution of the surface precipitation. Table 2 contains the infor-mation about the circulation type derived from the above analysis,the maximum recorded rainfall, and the Mean Sea Level Pressureover North, Central and South Europe during the event.

The classification shows that most of the weather systemsoccurring in the Mediterranean and passing over Crete have SW,NW and W directions. Fig. 4 presents the spatial distribution of

ete, and 100 years RP max daily precipitation. (The 100 years return period rainfallthod used is IDW.).



Table 2Circulation type, timing, precipitation and Mean Sea Level Pressure characteristics of flood events.

A/A Analysisperiod

Circulationtype

Circulation Max measuredprecipitation(mm)

NorthEuropecirculation

Low overNorth Europe(hPa)

High overCentralEurope (hPa)

Low overSouthEurope (hPa)

1 1-10/11/1991 W Cyclonic 40 Cyclonic 985 1021 10152 10-15/1/1994 W Cyclonic 180 Cyclonic 982 1024 10093 19-21/10/1997 N Cyclonic 140 Cyclonic 1000 1021 10094 1-7/1/1999 SW Cyclonic 200 Cyclonic 991 1024 10095 5-12/12/1999 W Cyclonic 40 Cyclonic 982 1030 10156 12-23/01/2000 SW Cyclonic 280 Cyclonic 1018 1030 10187 17-23/04/2000 SW Cyclonic 78 N/A 1003 1015 10068 3-9/12/2000 SW Cyclonic 270 Cyclonic 997 1027 10189 13-17/01/2001 HighCE Cyclonic 240 N/A 1018 1033 1009

10 1-5/11/2001 NW Cyclonic 173 Cyclonic 1003 1030 100911 1-9/12/2001 W Cyclonic 350 Cyclonic 1015 1033 100312 1-13/01/2002 NW Cyclonic 80 Cyclonic 997 1039 101213 7-13/09/2002 NW Cyclonic 35 Cyclonic 1021 N/A 101214 23-29/01/2003 NW Cyclonic 200 Anticyclone 1003 1030 100315 27-31/05/2003 SW Cyclonic 160 N/A N/A N/A 100016 3-7/10/2004 E Cyclonic 115 Cyclonic 991 1033 101217 3-16/11/2004 SW Cyclonic 90 Cyclonic 1003 1027 101218 16-21/12/2004 NW Cyclonic 120 Cyclonic 988 1007 99719 26-31/5/2005 NW Cyclonic 100 Cyclonic 1009 1024 100920 13-17/09/2005 NW Cyclonic 53 Cyclonic 994 1018 100621 1-4/11/2005 NW Cyclonic 15 Cyclonic 994 1027 101822 9-20/10/2006 SW Cyclonic 200 Cyclonic 1009 1024 1006

Fig. 4. Spatial distribution and Atmospheric Circulation type of flood events for the 1990–2007 period over Crete.

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

Autumn Winter Spring

Perc

ent o

f ev

ents

NW

SW

W

Fig. 5. Seasonality of Atmospheric Circulation type of flood events for the1990–2007 period over Crete.


ARTICLE IN PRESS

the circulation types related to flooding during the 1990–2007 per-iod for the island of Crete.

There is a pronounced difference among the origins of thesesystems when comparing western Crete with the rest of the island.Due to the shielding provided by the intense orography of westernCrete (Lefka Ori Mountains), the western part of the island receivesmostly the SW and NW circulation systems, while W circulatingsystems cross and reach the middle part of the Island, meetingthe second orographic barrier of Idi Mountain. Regarding the sea-sonality of the circulation patterns, Fig. 5 shows that most of Wtype events occur during winter while the dominant types duringautumn are NW and SW.

For the majority of the events, a common pressure gradientfield is observed over Europe. This common state can be de-scribed as a stream of low pressure over Southern Europe (Med-iterranean) combined with high pressure over Central Europeand low pressure over North Europe. The low barometric sys-tems are moving cyclonically eastwards crossing Mediterraneanand North Europe (Fig. 6). An average mean sea level (MSL) pres-sure drop of 16 ± 6.5 hPa from Central to South Europe and of27 ± 11 hPa from Central to North Europe is observed for mostof the cases. These depressions cause high intensity short dura-tion rainfall crossing the island of Crete, and are responsible forflooding.


3.3. Seasonality of mean monthly precipitation

The seasonality of mean monthly precipitation is presented inFig. 7. The maxima of mean monthly precipitation (Fig. 7a) gener-ally occur during winter with the exception of Central South part ofCrete (South of Messara Valley) where the maximum mean




ARTICLE IN PRESS

monthly precipitation is in November and September. For two sta-tions in NE Crete, the maximum mean monthly precipitation is inOctober, whereas in the central part, the maximum mean monthlyprecipitation occurs in January and December. A pattern of Decem-ber maximum is observed over the central north part of the islandwhich has lower elevation. A ‘‘strong” (higher occurrence) Januarymaximum pattern is observed at the central highlands. The analy-sis of �Smax indicates that on average (over 53 precipitation stations)the mean monthly maxima is 0.26 and in 40% of the precipitation

Fig. 6. Common pressure field observed for the majority of the flood events.

Fig. 7. Seasonality of long-term mean monthly precipitation in Crete. The top (a) and bomaximum (S2max) LMP in the period 1973–2005, respectively.


stations the mean monthly maxima coincide in 10 out of 32 years(1973–2005).

The secondary maximum (Fig. 7b) of the long-term meanmonthly precipitation is also generally observed during winter,seasonally close to the long-term mean monthly precipitationmaximum. An exception is for six stations located in eastern Cretefor which the secondary maximum of the long-term mean monthlyprecipitation occurs during September but with low frequency. Theevaluation of the S2max frequency shows slightly weaker seasonal-ity than assessed for the precipitation maxima. On average the�S2 max is equal to 0.25 while 34% of the stations have an �S2 max over0.3.

The number of precipitation stations with similar monthlymean precipitation seasonality is presented in Fig. 8. The left andright panels show the seasonality of the maximum and secondarymaximum of the long-term mean monthly precipitation, respec-tively. The frequencies are estimated separately for two groups ofprecipitation stations, according to their elevation. Fig. 8 showsthat the stations located at or below 200 m a.s.l. present their max-ima mostly in January (12 stations) and their secondary maximamostly in December (eight precipitation stations).

The rain gauges located above 200 m a.s.l. have their maximaand secondary maxima generally in January. Overall, the seasonal-ity of long-term mean monthly precipitation over Crete exhibits aclear pattern characterized by the occurrence of maxima and sec-ondary maxima in winter and a typical summer minimum. Onlysix out of 53 rain gauges are located at elevation above 700 ma.s.l. and none above 1000 m a.s.l. while the elevation of Crete var-ies from 0 to 2500 m a.s.l. It is obvious that there is a lack of pre-cipitation measurements at high altitudes as well as in thewestern part of Crete.

ttom (b) panels show the season and frequency of the maximum (Smax), secondary



0

5

10

15

20

25

Sep

Oct

Nov

Dec Jan

Feb

Mar

Apr

May Jun

Jul

Aug

Num

ber

of s

tati

ons

Maximum

below 200 m a.s.l. above 200 m a.s.l.

0

5

10

15

20

Sep

Oct

Nov

Dec Jan

Feb

Mar

Apr

May Jun

Jul

Aug

Num

ber

of s

tati

ons

Secondary Maximum


(a) (b)

Fig. 8. Distribution of rain gauges with similar (a) long-term mean monthly maximum and (b) secondary maximum precipitation seasonality. The frequencies are evaluatedseparately for rain gauges located at or below and above 200 m a.s.l., respectively.


ARTICLE IN PRESS

3.4. Seasonality of mean monthly runoff

The seasonality of mean monthly runoff is presented in Fig. 9.Fig. 9a shows the seasonality of the maximum long-term meanmonthly runoff. The spatial variability in runoff seasonality showsa different pattern when compared with precipitation, which is aconsequence of the complex runoff dynamics. This is due to theseveral aquifers and aquicludes of complex distribution and prop-erties contributing to the flow process of the gauged basins. Forthe entire island of Crete, maximum monthly runoff tends to oc-cur during January–February, while maximum monthly precipita-tion occurs during December–January. This shift can be explainedtaking into account the presence of the karstic properties of theaquifers contributing with high spring discharge flows, mainlyover the western part of the island. Similar patterns are observed

Fig. 9. Seasonality and frequency of (a) Smax and (b) S2max in the period


for the secondary maxima of long-term mean monthly runoff(Fig. 9b).

The number of catchments with monthly runoff seasonalityestimations is presented in Fig. 10. Catchments with mean eleva-tion above 550 m a.s.l. show a delayed runoff seasonality comparedto lower mean elevation basins. As indicated by the difference in�Smax values, this behavior can be attributed to the dominant karsticformations of the higher altitude basins.

3.5. Seasonality of annual maximum daily precipitation

In order to investigate the change of the seasonality with themagnitude of extreme events, the seasonality index (�r, �h) was sep-arately estimated from a different number of the largest annualmaximum values. In this study, the Burn seasonality index of the

1970–2002Areas in grey represent karstic geological formations.



0

2

4

6

8

10

Sep

Oct

Nov

Dec Jan

Feb

Mar

Apr

May Jun

Jul

Aug

Num

ber

of s

tati

ons

Maximum


0

2

4

6

8

10

Sep

Oct

Nov

Dec Jan

Feb

Mar

Apr

May Jun

Jul

Aug

Num

ber

of s

tati

ons

Secondary Maximum


(a) (b)

Fig. 10. Distribution of catchments with: (a) long-term mean monthly maximum and (b) secondary maximum runoff seasonality. The frequencies are evaluated separatelyfor catchments with mean elevation at or below and above 550 m a.s.l., respectively.

Fig. 11. Seasonality of annual maximum daily precipitation for the period 1973–2005 in terms of: (a) the seasonality estimated from the complete dataset, (b) the threelargest events and (c) the largest event. The direction of the vectors indicates the average occurrence of extreme precipitation in a hydrological year, and the magnitude is thestrength of the seasonality.


ARTICLE IN PRESS

Please cite this article in press as: Koutroulis, A.G., et al. Seasonality of floods and their hydrometeorologic characteristics in the island of Crete. J. Hydrol.(2010), doi:10.1016/j.jhydrol.2010.04.025



ARTICLE IN PRESS

largest extreme event (�r1, �h1) was compared to the seasonality ofthe events calculated from the two (�r2, �h2), three (�r3, �h3), five (�r5,�h5), ten (�r10, �h10), and complete dataset, largest precipitation andflood annual maxima records. The change in seasonality was thenassessed by comparing the variability (�r) and the difference in themean date of occurrences.

The spatial patterns of the maximum daily precipitation season-ality over Crete are presented in Fig. 11. The Burn index of annualmaxima estimated from the complete dataset (Fig 11a), from thethree largest precipitation maxima (Fig. 11b) and that from thelargest precipitation maxima (Fig. 11c), respectively. The Burn in-dex, �r30, calculated from all events in the period 1973–2005 exhib-its uniform patterns of a late December–early January occurrence,for the total area of Crete. This strong seasonality pattern has a lowvariability in the mean date of occurrence. The mean index (�r30,�h10) for all stations is (0.68, December 29th). The seasonality of

Fig. 12. Seasonality maximum annual flood peaks in the period 1970–2000 in terms of: (aand (c) the largest event. The direction of the vectors indicates the average occurrence of eseasonality.


the three largest precipitation maxima exhibits an increased spa-tial variability (Fig. 11b). The variability in the mean date of occur-rence (�r3) is a little stronger with an average �r3 ¼ 0:74 around �h3 of3rd of December. Finally, the seasonality of the largest precipita-tion maxima shows the larger spatial variability (Fig. 11c) withan average �h1 of 17th of November.

3.6. Seasonality of annual maximum flows

The spatial patterns of the maximum annual floods seasonalityare presented in Fig. 12. It is clear that the maps show a morehomogeneous spatial pattern than those of annual maximum dailyprecipitation which is due to similarities in catchment processeslike snowmelt during high precipitation events and spring (karstic)discharges. The strong seasonality is reflected by a value of 0.81 for�r15, around �h15 of 29th of January. The variability in the mean date

) the seasonality estimated for the fifteen largest events, (b) the three largest eventsxtreme precipitation in a hydrological year, and the magnitude is the strength of the




ARTICLE IN PRESS

of occurrence (�r3) is weaker with an average �r3 ¼ 0:75 around �h3 of8th of January, while the largest flow maxima (�r1) with a �h1 at the10th of December. Grey arrows of Fig. 12a–c shows the corre-sponding precipitation maximum and when compared with floodpeaks, a seasonality shift is observed.

4. Conclusions

Comparison of the results of event-based seasonality analysisand hydrological characteristics seasonality analysis for the period1990–2007 for the island of Crete revealed valuable informationwithin the context of flood occurrence. The results of the presentstudy concurs to the seasonality analysis of 550 documentedevents presented by Gaume et al. (2009), comprising the inventory(European flash flood atlas) of HYDRATE EC funded project. Thespecificity of the late autumn–early winter flood occurrence inCrete can be paralleled to the robust signal of autumn flash flood-ing of the Mediterranean area. The slight late time shift of floods inCrete when compared to the autumn Catalonia, France and North-ern Italy, is due to the southern geographic location of the island.

The seasonality of largest daily precipitation maxima (�h1 = mid-dle November) and the seasonality of the three largest precipita-tion maxima (�h3 = early December) for Crete, concurs with theresults of the event-based seasonality analysis for Crete (20% offlood events in November and 30% in December). The seasonalityof the three largest long term daily precipitation maxima (�h3) indi-cates that 50% of the precipitation maxima events occur during theNovember–December–January (NDJ) period. The seasonal charac-teristics of flood regimes across the Alpine–Carpathian range hasbeen examined by Parajka et al. (this issue). The authors classifiedthe seasonality of annual maximum daily precipitation and runoffin eight categories from West to East. It is interesting to observethat the seasonality of flood events in Crete is similar to the one ob-served for the Cevennes and surrounding areas (south France).Catchments in these two areas are, typically, impacted by stormsduring late autumn due to warm and moist air advection fromthe Mediterranean Sea. A time delay (winter rather than autumn)in flood flows is observed in both regions due to wetter soils andkarstic spring runoff contribution. The comparison also illustratesthat the dominant flood producing storms for Cevennes regionare triggered from southern air flows, while for Crete from wes-terly air flows. This is due to the low barometric systems that aremoving cyclonically eastwards, crossing the Mediterranean, andcreating the specific direction of above mentioned air flows.

Moreover, the evaluation of �Smax indicates that for morethan 65% of the stations the mean monthly maxima occur duringNDJ period of the long-term period 1973–2005. The comparisonamong the seasonality of the annual maximum daily precipitationand the maximum annual flows for the area of Crete shows a shiftof 1 month. This is attributed to the snow melting process, thelow soil percolation rates of winter period and the high baseflowrates of the local karstic aquifers, contributing to the maximumflows.

As far as the spatial distribution is concerned, 66% of the re-ported events took place in western Crete whereas the rest 34%took place in the eastern part of the island. This is rather expectedtaking into consideration the fact that western Crete receives high-er amounts and higher rates of precipitation than eastern Crete.This is partly due to the regional atmospheric patterns that aredominantly approaching from NW and are heading towards SE. An-other important factor is the morphological variability which pre-sents higher elevation and steepest slopes in the west part of theisland. For the majority of the events, a common pressure gradientwas observed that can be described as a stream of low pressureover Southern Europe (Mediterranean) combined with high pres-


sure over Central Europe and low pressure over North Europe.The classification of the atmospheric circulation types during theflood events concluded that most of the weather systems gener-ated within the Mediterranean area and passed over Crete withSW, NW and W directions.

As discussed, flash floods can seldom be predicted and thereforeany new evidence regarding the occurrence of events can be usefulfor their mitigation. Statistical analysis of large and medium scaleinformation reveals the spatio-temporal patterns that can lead to aflash flood. This analysis provides better awareness which is a keycomponent of hazard mitigation, planning and management. Thepresent integrated hydrometeorologic seasonality analysis can bea useful decision aid for civil protection authorities in the islandof Crete. European Flood Directive presented in 2007 asks fromEC member states to prepare flood risk maps by 2013. The researchpresented in this paper could provide valuable information for theIsland of Crete and could be a part of the work done to meet theDirective.

Acknowledgments

The research was supported by the European Communityfunded project, HYDRATE, Sixth Framework Programme, ContractNo. 037024. The authors would like to acknowledge the Water Re-sources Department of the Prefecture of Crete (WRDPC) for provid-ing most of the hydrological data for the island of Crete, the CivilProtection Service of the prefecture of Crete (Ms. Koutentaki) andthe Ministry of Environment, Physical Planning and Public Worksfor providing information regarding flood events. Finally, theauthors would like to thank the COST Action 733 project for theprovided atmospheric circulation results and the National Centerfor Atmospheric Research for the Global Reanalysis data that sup-ported the meteorological analysis of this study.

References

Bardossy, A., Caspary, H.J., 1990. Detection of climate change in Europe by analyzingEuropean atmospheric circulation patterns from 1881 to 1989. Theor. Appl.Climatol. 42, 155–167.

Barry, R.G., Chorley, R.J., 2003. Atmosphere, weather, and climate, eighth ed.Routledge, London, ISBN 0–415–27170–3.

Baur, F., Hess, P., Nagel, H., 1944. Kalender der Gmfl wefferlagen Eumpas 1881–1939, Bad Homburg, Homburg.

Bayliss, A.C., Jones, R.C., 1993. Peaks-over-threshold flood database: summarystatistics and previous seasonality. Report No. 121, Institute of PreviousHydrology, Wallingford, 61 p.

Burn, D.H., 1997. Catchment similarity for regional flood frequency analysis usingseasonality measures. J. Hydrol. 202, 212–230.

Caspary, H.J., 1995. Erhfhung des Hochwasserrisikos in SUdwestdeutschlanddureh lnstationaritat der winterliehen zyklonalen Westwetterlagen, In:Schultz, G. A. (Hrsg.) Verf’dgbarkeit yon Wasser, Schriflenreihe: ‘‘Hydrologic/Wasserwirtschaft”. Heft 14, Ruhr-Universit�it Bochum, 127–139.

Caspary, H.J., 1996. Die Winterhochwasser 1990, 1993 und 1995 inS0dwestdeutschland - Signale einer bereits eingetretenen Klimaanderung? In:Beehteler, Gfinthert, Kleeberg (Hrsg.): Klima�nderung und Wasserwirtsehaft,Tagungsband Int. Syrup. v. 27–28.1 I. 1995 Mfinehen, Mitteilungen Inst. farWasserwesen, Bd. 56a/b, Universit/lt der Bundes wehr MQnchen, 169–184.

Castellarin, A., Burn, D.H., Brath, A., 2001. Assessing the effectiveness of hydrologicalsimilarity measures for previous flood frequency analysis. J. Hydrol. 241, 270–285.

Cunderlik, J.M., Ouarda, T.B.M.J., Bobée, B., 2004a. On the objective identification offlood seasons. Water Resour. Res. 40.

Cunderlik, J.M., Ouarda, T.B.M.J., Bobée, B., 2004b. Determination of flood previousseasonality from hydrologic records. Hydrol. Sci. J. 49 (3), 511–526.

Cunderlik, J.M., Ouarda, T.B.M.J., 2009. Trends in the timing and magnitude of floodsin Canada. J. Hydrol. 375, 471–480. doi:10.1016/j.jhydrol.2009.06.050.

Fisher, N.I., 1993. Statistical Analysis of Circular Data. Cambridge University Press,Cambridge. p 277.

Gaume, E., Bain, V., Bernardara, P., Newinger, O., Barbuc, M., Bateman, A.,BlaškovicÀova, L., Bloschl, G., Borga, M., Dumitrescu, A., Daliakopoulos, I.,Garcia, J., Irimescu, A., Kohnova, S., Koutroulis, A., Marchi, L., Matreata, S.,Medina, V., Preciso, E., Sempere-Torres, D., Stancalie, G., Szolgay, J., Tsanis, I.,Velasco, D., Viglione, A., 2009. A compilation of data on European flash floods. J.Hydrol. doi:10.1016/j.jhydrol.2008.12.028.






ARTICLE IN PRESS

Gerstengabe, F.-W., Werner P.C., Rüge U., 1999. Katalog der GrosswetterlagenEuropas 1881–1998 nach P. Hess und H. Brezowsky. 5. Aufl. – Potsdam-Inst. F.Klimafolgenforschung, Potsdam, Germany, 138 pp.

Hess, P., Brezowsky, H., 1952. Katalog der Großwetterlagen Europas. Ber. Dt.Wetterd. in der US-Zone 33. Bad Kissingen, Germany.

Hess, P., Brezowsky, H., 1969. Katalog der Grosswetterlagen Europas, 2. neubearbeitete und ergänzte Aufl. Berichte des Deutschen Wetterdienstes 113.Offenbach am Main.

Hess, P., Brezowsky, H., 1977. Katalog der Grosswetterlagen Europas 1881–1976, 3.verbesserte und ergänzte Aufl. Berichte des Deutschen Wetterdienstes 113.Offenbach am Main.

Huth, R., Beck, C., Philipp, A., Demuzere, M., Ustrnul, Z., Cahynová, M., Kysely, J.,Tveito, O.E., 2008. Classifications of atmospheric circulation patterns – recentadvances and applications. Trends and Directions in Climate Research. Ann. NYAcad. Sci. 1146, 105–152.

James, P.M., 2007. An objective classification method for Hess and BrezowskyGrosswetterlagen over Europe. Theor. Appl. Climatol. 88, 17–42.

Kistler, R., Kalnay, E., Collins, W., Saha, S., White, G., Woollen, J., Chelliah, M.,Ebisuzaki, W., Kanamitsu, M., Kousky, V., Van den Dool, H., Jenne, R., Fiorino, M.,2001. The NCEP-NCAR 50-year reanalysis: monthly means CD-ROM anddocumentation. Bull. Amer. Meteor Soc. 82, 247–267.

Koutroulis, A.G., Tsanis, I.K., 2010. A method for estimating flash flood peakdischarge in a poorly gauged basin: case study for the 13–14 January 1994flood, Giofiros basin. Crete. Greece. J. Hydrol. doi:10.1016/j.jhydrol.2010.02.012.

Magilligan, F.J., Graner, B.E., 1996. Hydroclimatological and geomorphic controls onthe timing and spatial variability of floods in New England. USA. J. Hydrol. 178,159–180. doi:10.1016/0022-1694(95)02807-2.

Mardia, K.V., 1972. Statistics of Directional Data. Academic Press, New York.


Martina, M.L.V., Todini, E., Libralon, A., 2006. A Bayesian decision approach torainfall thresholds based flood warning. Hydrol. Earth Syst. Sci. 10, 413–426.

Naoum, S., Tsanis, I.K., 2004. Orographic precipitation modeling with multiple linearregression. J. Hydrol. Eng. 9 (2), 79–102.

Norbiato, D., Borga, M., Degli Esposti, S., Gaume, E., Anquetin, S., 2008. Flash floodwarning based on rainfall depth-duration thresholds and soil moistureconditions: an assessment for gauged and ungauged basins. J. Hydrol. 362 (3),274–290. doi:10.1016/j.jhydrol.2008.08.023.

Ottenbacher, A., Tomassini, M., Schmetz, J., Holmlund, K., 1997. Low-level cloudmotion winds from meteosat high-resolution visible imagery. WeatherForecast. 12, 175–184.

Ouarda, T.B.M.J., Cunderlik, J.M., St-Hilaire, A., Barbet, M., Bruneau, P., Bobeé, B.,2006. Data-based comparison of seasonality-based regional flood frequencymethods. J. Hydrol. 330, 329–339.

Parajka, J., Kohnová, S., Merz, R., Szolgay, J., Hlavcová, K., Blöschl, G., 2009.Comparative analysis of the seasonality of hydrological characteristics inSlovakia and Austria. Hydrol. Sci. 54 (3).

Parajka, J., Kohnova, S., Balint, G., Barbuc, M., Borga, M., Claps, P., Cheval, S., Gaume,E., Hlavcova, K., Merz, R., Pfaundler, M., Stancalie, G., Szolgay, J., Bloschl, G., thisissue. Seasonal characteristics of flood regimes across the Alpine-Carpathianrange. J. Hydrol.

Van De Griend, A.A., Seyhan, E., 1984. Statistical analysis of weather-type runoffphenomena in an alpine environment. J. Hydrol. 70, 51–69. 51.

Yarnal, B., 1993. Synoptic climatology in environmental analysis. In: Srudies inClimatology Series. Belhaven Press, London.

Yamal, B., White, D., 1987. Subjectivity in a computer-assisted synoptic climatologyI: classification results. J. Climatol. 7, 119–128.



http://dx.doi.org/10.1016/0022-1694(95)02807-2



journal of hydrology - unipd.it...periods and areas for the island of crete, can be used as basic...

Documents