Assessment of island beach erosion due to sea level rise: the case of the Aegean archipelago

18 Pages·2017·7.98 MB·English

Nat.HazardsEarthSyst.Sci.,17,449–466,2017
www.nat-hazards-earth-syst-sci.net/17/449/2017/
doi:10.5194/nhess-17-449-2017
©Author(s)2017.CCAttribution3.0License.
Assessment of island beach erosion due to sea level rise: the case of
the Aegean archipelago (Eastern Mediterranean)
IsavelaN.Monioudi1,AdonisF.Velegrakis1,AntonisE.Chatzipavlis1,AnastasiosRigos1,2,TheophanisKarambas3,
MichalisI.Vousdoukas4,1,ThomasHasiotis1,NikolettaKoukourouvli5,PascalPeduzzi6,EvaManoutsoglou1,
SerafimE.Poulos7,andMichaelB.Collins8
1DepartmentofMarineSciences,UniversityoftheAegean,UniversityHill,81100Mytilene,Greece
2DepartmentofCulturalTechnologyandCommunication,UniversityoftheAegean,UniversityHill,
Mytilene81100,Greece
3SchoolofCivilEngineering,AristotleUniversityofThessaloniki,UniversityCampus,
54124Thessaloniki,Greece
4EuropeanCommission,JointResearchCentre(JRC),DirectorateforSpace,SecurityandMigration
DisasterRiskManagementUnit,ViaE.Fermi2749,21027Ispra(VA),Italy
5DepartmentofGeography,UniversityoftheAegean,UniversityHill,81100Mytilene,Greece
6UNEP/DEWA/GRID-Geneva,InternationalEnvironmentHouse,11ChemindesAnémones,
1219Châtelaine,Switzerland
7FacultyofGeologyandGeoenvironment,NationalandKapodistrianUniversityofAthens,
PanepistimioupoliZografou,15784Athens,Greece
8PlentziakoItsasEstazioa,UniversityoftheBasqueCountry,Areatzaz/g,48620Plentzia-Bizkaia,Spain
Correspondenceto:MichalisI.Vousdoukas([email protected])
Received:17October2016–Discussionstarted:7November2016
Revised:17January2017–Accepted:4February2017–Published:21March2017
Abstract.Thepresentcontributionconstitutesthefirstcom- vulnerabletomeansealevelrise(MSLR)andepisodicSLRs
prehensive attempt to (a) record the spatial characteristics due to (i) their narrow widths (about 59% of the beaches
of the beaches of the Aegean archipelago (Greece), a crit- have maximum widths <20m), (ii) their limited terrestrial
ical resource for both the local and national economy, and sediment supply, (iii) the substantial coastal development
(b) provide a rapid assessment of the impacts of the long- and(iv)thelimitedexistingcoastalprotection.Modelingre-
termandepisodicsealevelrise(SLR)underdifferentscenar- sultsindeedprojectsevereimpactsundermeanandepisodic
ios.Spatialinformationandotherattributes(e.g.,presenceof SLRs, which by 2100 could be devastating. For example,
coastalprotectionworksandbackshoredevelopment)ofthe under MSLR of 0.5m – representative concentration path-
beachesofthe58largestislandsofthearchipelagowereob- way (RCP) 4.5 of the Fifth Assessment Report (AR5) of
tainedonthebasisofremote-sensedimagesavailableonthe the Intergovernmental Panel on Climate change (IPCC) – a
web. Ranges of SLR-induced beach retreats under different storm-induced sea level rise of 0.6m is projected to result
morphological,sedimentologicalandhydrodynamicforcing, inacompleteerosionofbetween31and88%ofallbeaches
andSLRscenarioswereestimatedusingsuitableensembles (29–87%ofbeachesarecurrentlyfrontingcoastalinfrastruc-
ofcross-shore(1-D)morphodynamicmodels.Theseranges, ture and assets), at least temporarily. Our results suggest a
combinedwithempiricallyderivedestimationsofwaverun- very considerable risk which will require significant effort,
upinducedflooding,werethencomparedwiththerecorded financial resources and policies/regulation in order to pro-
maximumbeachwidthstoproviderangesofretreat/erosion tect/maintain the critical economic resource of the Aegean
andfloodingatthearchipelagoscale.Thespatialinformation archipelago.
showsthattheAegean“pocket”beachesmaybeparticularly
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
450 I.N.Monioudietal.:Assessmentofislandbeacherosionduetosealevelrise
1 Introduction coastal infrastructure and assets and/or as significant envi-
ronments of leisure (e.g., Paula et al., 2013). Therefore, as-
Beaches are critical components of the coastal zone; not sessmentsofthebeachmorphologicalevolutionatdifferent
only are they significant habitats in their own right (e.g., spatiotemporal scales are required based on advanced nu-
Defeo and McLachlan, 2013), but they also provide protec- merical,analyticaland/orempiricalmodelsconstructedand
tion from marine flooding to other transitional ecosystems appliedbyexperiencedoperatorsandsetup/validatedusing
andthecoastalassets,infrastructureandactivitiestheyfront appropriate field data and backed by expert analysis (e.g.,
(e.g., Neumann et al., 2015). At the same time, tourism has Roelvink et al., 2009; Bosom and Jiménez, 2010; Ding et
been increasingly associated with beach recreational activ- al., 2013). However, such efforts are usually hampered by
ities according to the dominant “Sun, Sea and Sand - 3S the(a)scarcityofrelevantinformationinmanycoastalareas
tourism” model (Phillips and Jones, 2006). Consequently, and(b)dearthinthenecessaryhumanandfinancialresources
beaches have become very important economic resources (e.g., Parker et al., 2013); this is particularly true when as-
(Ghermandi and Nunes, 2013) formingone of the pillars of sessmentsofbeacherosionarecarriedoutoverlargerspatial
tourism, an economic sector that contributes an estimated scales.Nevertheless,itisnecessarytoassessfuturebeachre-
5% of Gross Domestic Product (GDP), and about 6–7% treat/erosionandfloodriskatlargespatialscalesinorderto
of global employment (directly and indirectly) (Hall et al., identify“hotspots”andplanforeffectiveadaptationpolicies
2013). andefficientallocationofresources.
Beaches are also very dynamic environments, controlled Againstthisbackground,theobjectiveofthepresentstudy
by complex forcing-response processes that operate at var- istoassesstheerosionandtemporaryinundation/floodrisks
ious spatiotemporal scales (Short and Jackson, 2013). They of the beaches of the Aegean archipelago islands (Greece)
are generally under erosion (Eurosion, 2004; IPCC SREX, under different scenarios of sea level rise (SLR). Towards
2012;IPCC,2013),whichcanbedifferentiatedinto(a)long- thisobjective,spatialcharacteristicssuchasthearea,length,
termerosion,i.e.,irreversibleretreatoftheshoreline,dueto maximumwidth,orientation,sedimentsandthepresenceof
meansealevelrise(MSLR)and/ornegativecoastalsedimen- coastal works and backshore development of the Aegean
tary budgets that force either beach landward migration or beaches were recorded. This information was then used in
drowning(NichollsandCazenave,2010),and(b)short-term conjunctionwithprojectionsfromensemblesofcross-shore
erosion caused by storm surges and waves which, may or morphodynamicmodelstoobtainestimatesoftherangesof
may not, result in permanent shoreline retreats but can be potential beach retreat/erosion and flooding under different
nevertheless devastating (e.g., Smith and Katz, 2012; UN- MSLRsandstormconditions.
ECE,2013).TheacceleratingMSLRcoupledwithepisodic
stormeventswillaggravatethealreadysignificantbeachero-
sion with severe impacts on coastal activities, infrastructure 2 Aegeanarchipelagobeaches:significance,
and assets (e.g., Jiménez et al., 2012), and the beach carry- environmentalsettingandsealevelrise
ingcapacityforrecreation/tourism(ValdemoroandJiménez,
2006;McArthur,2015). 2.1 SignificanceoftheAegeanarchipelagobeaches
Beach erosion appears to be particularly alarming in is-
lands. Island beaches are increasingly vulnerable to erosion TheAegeanarchipelago(Fig.1)consistsofseveralthousand
due to their (generally) limited dimensions and diminish- islands and rock islets with a combined area of 17550km2
ing sediment supply (e.g., Velegrakis et al., 2008; Peduzzi and a total coastline length of about 5880km (Eurosion,
et al., 2013). At the same time, island beaches are amongst 2004). Few of these islands and islets are populated; less
themostsignificant3Stourismdestinations.Forexample,3S than 70 islands have more than 100 permanent inhabitants
tourismaccountsformorethan23%oftheGrossDomestic and 45 have more than 1000 (http://www.statistics.gr/en/
Product in many Caribbean Small Island Developing States interactive-map). Yet Aegean islands form very significant
andinsomecases,e.g.,AntiguaandBarbuda,formorethan touristdestinations.About50%ofallGreekhotelbeds(and
75%(ECLAC,2011).Mediterraneanislandsarealsomajor >60% of all 5 star hotel beds) are located in the Aegean
tourism destinations; in Greece, most of the hotel capacity archipelago, with 43% of the foreign arrivals to Greece in
andforeigntouristarrivalsandearningsareassociatedwith 2015(7.4outofatotalof17.1million)arrivingatits11in-
theGreekislands(SETE,2016). ternationalairports(SETE,2016).
Underavariableandchangingclimate,projectionsonthe In recent years, tourism has become a most signif-
futureevolutionofbeachmorphologyarenoteasyduetoun- icant economic activity in Greece. In 2013, foreign
certainties regarding both forcing and beach response (e.g., earnings of the Greek tourist industry were about
Short and Jackson, 2013). Nevertheless, beach erosion is EUR15.5billion (http://www.bankofgreece.gr/Pages/el/
amongst the first issues to consider when planning for the Statistics/externalsector/balance/travelling.aspx). As recent
sustainable development of the coastal zone, particularly in studies suggest that for each EUR1 generated by tourism
areaswherebeachesfunctionasnatural“armor”tovaluable in Greece an additional EUR1.2–1.65 is created by related
Nat.HazardsEarthSyst.Sci.,17,449–466,2017 www.nat-hazards-earth-syst-sci.net/17/449/2017/
I.N.Monioudietal.:Assessmentofislandbeacherosionduetosealevelrise 451
economic activity (a multiplier of, at least, 2.2; see IOBE, tan Arc straits (Fig. 1) as the most energetic areas of the
2012), it follows that direct and indirect earnings from Aegeanarchipelago;forexample,maximumwaveheightsof
tourism may account for up to about 20% of the country’s about11m(T of13.3sanddirectionof345◦N)havebeen
p
GDP(and30%oftheprivatesectoremployment).Tourism reported for the Mykonos–Ikaria strait (22 January 2004).
isevenmoreimportantfortheisland(local)economies.For AnAnalysisofERA-Interimwaveinformation(1979–2013)
example, in 2012 tourism accounted for about 48% of the fromdifferentrepresentativeareasoftheAegeanarchipelago
GDP of Crete and 60% of the GDP of the Cyclades and carried out as part of the present study shows (a) mean sig-
Dodecaneseislandcomplexes(SETE,2016). nificant wave heights (H ) of about 1m in all areas, apart
s
In the Aegean archipelago, 3S tourism is the dominant from an area to the northeast of eastern Cretan Arc straits
model. A most critical component of 3S tourism is the (meanH ofabout0.8m);(b)meanmaximumwaveheights
s
availability of beaches that are environmentally and aes- ofabout2.4m;and(c)significantinterannualvariability.
theticallysoundandretain adequatecarryingcapacity(e.g., Recent studies on the future wave climate of the Aegean
McArthur, 2015; Cisneros et al., 2016). Therefore, the as- archipelagoprojectsmallchangesinsignificantwaveheights
sessment/management of beach erosion which could con- for the 21st century. For the period 2001–2049, significant
stitute a major risk for the sustainable development of the wave heights in the N. Aegean are projected to slightly in-
Aegean islands beaches should be prioritized; a decade-old crease for the SW waves relative to the 1950–2000 refer-
approximation had suggested that about 25% of the total enceperiod;whereasfortheendofthecentury(2050–2099),
coastline of the Aegean islands was already under erosion waveoccurrenceandheightsareprojectedtoshowhighspa-
(Eurosion,2004). tiotemporal variability (e.g., Prinos, 2014; Tsoukala et al.,
2016).
2.2 Environmentalsetting
2.3 Meanandextremesealevels
The Aegean archipelago is located in the Aegean Sea, a
peripheral sea of the Eastern Mediterranean that covers an
area of some 160×103km2, drains high relief basins with Mediterranean MSLR rates were 1.1–1.3mmyr−1 for most
a total area of 200×103km2 and is connected to Black of the 20th century. Since the late 1990s, however, much
Sea through the Dardanelles straits and to Eastern Mediter- higher rates (2.4–3.8mmyr−1) have been recorded, an in-
raneanthroughtheCretanArcstraits.TheAegeanSeahasa crease attributed mainly to additional water mass inputs
highreliefduetocomplexregionaltectonicsandcomprises rather than to steric contributions; these changes have been
different geomorphological units (Poulos, 2009), including also related to North Atlantic Oscillation (NAO) modula-
an extensive shelf (N. Aegean shelf), a tectonic trough (N. tions (Tsimplis et al., 2013). For the Aegean archipelago in
AegeanTrough),acentralplatform(Cycladesplateau)with particular, satellite altimetry suggests recent MSLR rates of
alargeconcentrationofislandsanddeepbasins(Fig.1).The 4.3–4.6mmyr−1 (Mamoutos et al., 2014), with some peri-
Aegean Sea shows complex hydrographic patterns and cir- odscharacterizedbyevenhigherrates(e.g.,Tsimplisetal.,
culation(e.g.,Theocharisetal.,1993)whicharemainlycon- 2009). In terms of future projections, recent studies project
trolled by the cold and low salinity water inputs from the decreasesintheMSLRratesinthe21stcentury,althoughthis
Black Sea through the Dardanelles strait and the warm and mightbeanunderestimationduetotheuncertaintiesregard-
salinewaterinputsfromtheLevantineSeathroughtheeast- ing the mass exchanges particularly between the Black and
ern Cretan Arc straits (Skliris et al., 2011). Under certain Aegean Seas (Mamoutos et al., 2014). Hinkel et al. (2014),
conditions(theEasternMediterraneantransient),theAegean using an approach accounting for changes in ice mass and
basinscanbeimportantlocationsofdeepwaterformationin oceancirculation,projectedthefollowingMSLRsforthere-
the Eastern Mediterranean (e.g., Zervakis et al., 2000; An- gionoftheAegeanarchipelago(33.5–40◦N,18.5–28.5◦E).
droulidakisetal.,2012). In2050,MSLwasprojectedtobe0.13–0.15and0.14–16m
The complex physiography of the Aegean archipelago higher than that of the 1985–2005 reference period for a
controls its wind and wave climate, which is generally rel- medium land ice scenario and RCPs of 4.5 and 8.5, respec-
ativelymildduetotheshortfetchesanddurations.Northerly tively; in 2100, under the same scenarios, MSL was pro-
winds (44% frequency of occurrence, Androulidakis et al., jected to be 0.46–0.48 and 0.66–0.72m higher than that of
2015) and waves (Soukissian et al., 2007, 2008) appear to the1985–2005referenceperiod,respectively.
prevail.Althoughwavesaregenerallymoreenergeticinwin- In addition to MSLR, changes in the intensity, frequency
ter,therearealsoenergeticeventsinsummer,forcedbyN– and/or patterns of extreme storm surges and waves can, at
NE winds (“the etesians”). Highly energetic wave events of least temporarily, induce beach erosion and flooding, par-
relatively short duration may also occur, particularly along ticularly when combined with increasing mean sea lev-
islandstraits.Soukissianetal.(2008)suggested(i)thearea els (MSLs)(e.g., Xu and Huang, 2013). Extreme sea levels
to the N–NE of the Cyclades platform (particularly the (ESLs) in the Mediterranean have a seasonal footprint with
Mykonos–Ikariastrait)and(ii)thewesternandeasternCre- extremepositivesoccurringmostlyinwinterandundercer-
www.nat-hazards-earth-syst-sci.net/17/449/2017/ Nat.HazardsEarthSyst.Sci.,17,449–466,2017
452 I.N.Monioudietal.:Assessmentofislandbeacherosionduetosealevelrise
Figure1.TheAegeanarchipelago.
tain North Atlantic Oscillation modulations (Tsimplis and 3 MaterialsandMethods
Shaw,2010).
Future ESLs are projected to show high spatial variabil- 3.1 GeospatialcharacteristicsoftheAegean
ity, being sensitive to the evolution of the thermohaline cir- archipelagobeaches
culation and the Black Sea buoyant inputs (e.g., Mamoutos
ThegeospatialcharacteristicsofAegeanarchipelago(“dry”)
et al., 2014). Generally, extreme levels due to storm surges
beaches have been recorded on the basis of the imagery
and waves are projected to decrease in terms of magnitude
andotherrelatedopticalinformationavailableintheGoogle
overtheMediterraneanbasintowardstheendofthe21stcen-
EarthProapplication.Inthisstudy,drybeachesweredefined
tury(e.g.,ConteandLionello,2013),althoughmodelchoice
asthelow-lyingcoastalsedimentarybodiesboundedontheir
andresolutionandthequality/resolutionofthecoastalobser-
landwardsidebyeithernaturalboundaries(vegetateddunes
vations available for model validation may have influenced
and/or cliffs) or permanent artificial structures (e.g., coastal
these projections (e.g., Calafat et al., 2014). For example, a
embankments, sea walls, roads, and buildings) and on their
recentstudy hassuggestedsubstantial changesin thereturn
seaward side by the shoreline, i.e., the median line of the
periodsofESLsfortheE.Mediterraneanfortheendofthe
foaming swash zone shown on the imagery. Regarding the
21st century, with the current 1000-year event projected to
lateralextentofindividualbeaches,theseweredelimitedby
occureveryfiveyears(Vousdoukasetal.,2017).Regarding
naturalbarrierssuchasrockpromontories.Tinybeaches(ar-
the storm surges in the Aegean archipelago, these are rel-
easlessthanabout20m2)wereignoredandnotincludedin
atively small (heights of up to about 0.5m, Krestenitis et
the data set. Digitization of the remote-sensed imagery was
al.,2011)andprojectedtoshow(generally)smallheightin-
carriedoutbyfew(four)analystswhoconsistentlyfollowed
creasesuntil2050aswellaschangesintheirtemporaldistri-
the above beach delimitation rules. To assess inconsisten-
bution(e.g.,Marcosetal.,2011;Androulidakisetal.,2015;
cies, 400 beaches from different islands (about 12% of the
Vousdoukasetal.,2016a).
recordedbeaches)wereprocessedbyallfouranalystsandthe
standard deviation of the extracted shoreline positions was
estimatedatlessthan0.3m;thiswasconsideredacceptable
forthescopeofthestudy.
Nat.HazardsEarthSyst.Sci.,17,449–466,2017 www.nat-hazards-earth-syst-sci.net/17/449/2017/
I.N.Monioudietal.:Assessmentofislandbeacherosionduetosealevelrise 453
However,itshouldbenotedthattheprimaryinformation (Leont’yev, 1996), XBeach (Roelvink et al., 2010) and
used,i.e.,satelliteimagerysnapshots,cannotprovidesynop- the Boussinesq model, whose hydrodynamic component
ticinformationattheAegeanarchipelagoscale,astheavail- involves high-order Boussinesq equations (Karambas and
ableimageshavebeencollectedindifferentperiodsandun- Koutitas, 2002). The Bruun model is a widely-used (e.g.,
der variable hydrodynamic conditions. Regarding the beach Hinkeletal.,2010;Ranasingheetal.,2013)analyticalmor-
width estimations, although astronomical tidal effects are phodynamic model that estimates long-term coastal retreat
smallduetothemicrotidalregimeoftheAegeanarchipelago (S) under a SLR (α) on the basis of the equilibrium pro-
(tidal ranges in most areas less than about 0.15m) and the file concept; its results are controlled by the height of the
generally increased beach slopes, there are still uncertain- beach face and the cross-shore distance between the beach
ties as the estimations were based on “snapshots” of the closure depth and the shoreline. The Edelman model esti-
shorelinepositionswhichareverydynamiccoastalfeatures mates beach erosion/retreat using the initial height of the
controlled, amongst others, by the spatiotemporal variabil- beach face, the water depth at wave breaking and the surf
ity of the nearshore waves and currents (e.g., Velegrakis et zone width, whereas the Dean model estimates retreats on
al.,2016).Satelliteimagery,aswellasother“one-off”land- thebasisofthewaveheight,thewaterdepthatwavebreak-
based, airborne or lidar surveys, provides only snapshots of ing and the surf zone width. It should be noted that due to
beach widths which may not represent “mean” conditions; lackofaccuratebathymetrydata,thenecessaryinputvalues
nevertheless, this cannot be avoided when working at large of the analytical models, i.e., the distance of the shoreline
spatialscales(e.g.,Allenbachetal.,2015;Vousdoukasetal., fromtheclosuredepthandthewidthofthesurfzone,were
2016a). estimatedonthebasisofthebeachslope.
Beaches were digitized as polygons and exported to Ar- The SBEACH model (Larson and Kraus, 1989) is a nu-
cGIS for further analysis. There has been no georectifica- mericalmorphodynamicmodelconsistingofthreemodules:
tion, as the aim of the exercise has not been to provide a hydrodynamic, a sediment transport and a morphologi-
definitive locations and elevations of beach features, but to cal evolution module. It can describe wave transformation
extract/record(horizontal)geospatialcharacteristics.Tothis in shoaling waters, with the coastal sediment transport con-
end, a custom made AML (ARC Macro Language, propri- trolledbythecoastalwaveenergyfluxes;thesedimentconti-
etary language for ArcInfo applications in Esri software) nuityequationinafinite-differenceschemeanda“stair-step”
script was used to estimate beach areas, lengths, maximum beachprofilediscretizationisusedinitsmorphologicalmod-
widthsandorientations(Allenbachetal.,2015). ule. The numerical model based on Leont’yev (1996) uses
In addition to beach dimensions, other relevant infor- the energetic approach, with the cross-shore wave energy
mation was recorded and codified, including the presence balance controlled by wave propagation angle and dissipa-
of (a) natural features, such as river mouths, back-barrier tion;sedimenttransportratesareestimatedseparatelyforthe
lagoons and cliffs, and beachrock outcrops; and (b) artifi- surf and swash zones. The XBeach model (Roelvink et al.,
cial features such as coastal protection schemes and back- 2010)isanopen-source,widelyusednumericalmodelofthe
shore infrastructure/assets. Classification of the beach sed- nearshoreprocessesintendedtoestimatetheeffectsoftime-
iments of the Aegean island beaches in terms of sediment varyingstormconditions;itcontainsatime-dependentwave
textureclasses(e.g.,sandorgravel)wasalsocarriedouton action balance solver and allows for variations in the wave
the basis of all available information from the scientific lit- action over time and over the directional space. Finally, the
erature/reports (e.g., Monioudi et al., 2016) as well as web- Boussinesq model used computes nonlinear wave transfor-
basedphotographicmaterial. mation in the surf and swash zone based on a wave prop-
agation module involving high-order Boussinesq equations
3.2 Beachretreatpredictionsduetosealevelrise (KarambasandKoutitas,2002);itssedimenttransportmod-
ule can estimate sheet flow as well as bed and suspended
Sea level rise represents a most significant hazard for load over uneven sea beds (e.g., Karambas, 2006). Detailed
beaches, forcing their retreat/erosion; a sea level rise α will descriptionsofthemodelsusedcanbefoundelsewhere(e.g.,
result in a shoreline retreat S due to erosion of the beach Vousdoukasetal.,2009a;Monioudietal.,2016).
face, the sediments of which are transported/deposited off- In the present contribution, all models were used in a
shore, with the extent/rates of the cross-shore retreat con- stationary mode. Validation of model results was provided
trolled(amongstothers)bybedslope,thetextureandsupply throughcomparisonswiththeresultsofphysicalexperiments
of beach sediments, and the hydrodynamic conditions (e.g., that took placein the large wave flume(GWK) in Hanover,
Dean,2002). Germany (Sect. 4.2.1). The models were used in an ensem-
In the present study, seven cross-shore (1-D) morpho- blemodeinordertoassesstherangeoflong-andshort-term
dynamic models were used to project beach response to beach retreats/erosion for different beach slopes, sediment
SLR: the Bruun (Bruun, 1988), Edelman (Edelman, 1972) grain sizes and wave conditions, and under different sce-
and Dean (Dean, 1991) analytical models and the numeri- narios of MSL changes and/or storm sea levels. Two model
cal models SBEACH (Larson and Kraus, 1989), Leont’yev ensembleswerecreated:a“long-term”ensembleconsisting
www.nat-hazards-earth-syst-sci.net/17/449/2017/ Nat.HazardsEarthSyst.Sci.,17,449–466,2017
454 I.N.Monioudietal.:Assessmentofislandbeacherosionduetosealevelrise
of the Bruun, Dean and Edelman analytical models; and a R = (1)
2%
“short-term”ensemblecomprisingthenumericalSBEACH, (cid:2)H L (cid:0)0.563β2+0.004(cid:1)(cid:3)1/2!
Leont’yev, XBeach and Boussinesq models. The former is 1.1 0.35β(H /L )1/2+ o o ,
o o
used to assess beach retreat/erosion under MSLR and the 2
latter to assess retreat due to temporary SLR (i.e., due to foralldata,and
storm surges and waves). The adopted approach was based
onthepropositionthatasmodelshavedifferentialsensitivity
to the controlling environmental factors, ensemble applica- R2%=0.043(Ho/Lo)1/2R2%=0.043(Ho/Lo)1/2,
tions may provide “tighter” prediction ranges than the indi- fordissipativebeaches(ξ<0.3), (2)
vidualmodels(Sect.4.2.1).
where R , the 2% exceedance of the peak run-up height,
Experimentswerecarriedoutusingvariousplausibleener- 2%
H andL arethedeepwaterwaveheightandlength,β the
geticwaveconditionsintheAegeanarchipelago(Sect.2.2), o o
beachslopeandξ theIribarrennumber(ξ =β/(H /L )1/2).
i.e.,waveswithoffshoreheights(H )of1,1.5,2,3and4m o o
s
Wave run-up excursions were then calculated from the
and periods (T) of 4, 5, 6, 7 and 8s. Likewise, in order to
wave run-up heights (R ) for all tested bed slopes and
addressthesedimenttexturevariabilityoverthearchipelago 2%
waveconditionsandaddedtothebeacherosion/retreatpro-
beaches,experimentswerecarriedoutforsevendifferentme-
jectionsoftheseven1-Dcross-shoremorphodynamicmod-
dian (d ) grain sizes (d of 0.2, 0.33, 0.50, 0.80, 1, 2 and
50 50
elstoprojectfinalfloodingexcursions(S(i)).Thebestfitsof
5mm);notethattheresultsoftheanalyticalmodelsarenot
thelowerandupperlimitsofthefinalprojectionsofflooding
controlled by beach sediment size. Five (5) different linear
byallmodelswerethenestimated.
profile slopes (bed slopes of 1/10, 1/15, 1/20, 1/25 and
1/30) and 11 SLR scenarios (0.05, 0.15, 0.22, 0.30, 0.40,
0.50, 0.75, 1, 1.25, 1.50 and 2m) were examined. Experi-
4 Results
mentswerecarriedoutforallcombinations(about5500ex-
periments),andthemeans(bestfits)ofthelowestandhigh- 4.1 CharacteristicsofthebeachesoftheAegean
estprojectionsbyallmodelsofthetwoensembleswerees- archipelago
timated. With regard to combined SLRs, i.e., coastal ESLs
due to storm surges and wave set-ups superimposed on the Atotalof3234beacheswererecordedalongthecoastsofthe
MSLRs (e.g., Vousdoukas et al., 2016b), the long-term and 58 larger islands of the Aegean archipelago. These beaches
short-termensembleswereusedconsecutively. were found to be “pocket” beaches, having a total area of
Recent MSLR projections for the area that take into ac- about 21.35km2; this indicates that the total carrying ca-
count contributions of ice mass were used (Hinkel et al., pacity of the Aegean archipelago beaches (i.e., the num-
2014):(i)0.15m,averageoftheRCP4.5andRCP8.5sce- ber of visitors that can be simultaneously hosted) is about
narios in 2040; and (ii) 0.5 and 0.7m under RCP 4.5 and 2.135 million according to the Rajan et al. (2013) criterion
RCP8.5(2100).Withregardtoshort-termSLR,recenttrends (10m2foreachbeachuser).Aroughestimationonthebasis
forextremesealevelsintheAegeanarchipelago(upto0.5– oftheaveragenumberoftouristdaysspentintheAegeanis-
0.6mhigh)(TsimplisandShaw,2010)areused;ESLsinthe landssuggeststhat,withtheexceptionofthebusiestmonths
region are projected to only slightly increase in the future (JulyandAugust)duringwhichabout7milliontouristdays
according to recent projections regarding their storm surge permontharerecorded(SETE,2016)andmanybeachesop-
component (Androulidakis et al., 2015; Vousdoukas et al., erate close to their full carrying capacity, the archipelago
2016a). beaches in their present condition still have (as a total) po-
The above approach is designed to project beach re- tentialfordevelopmentasenvironmentsofleisure.
treat/erosion, but not temporary inundation/flooding due to MostoftheAegeanbeachesarenarrow,withabout59%
wave run-up. Although the wave run-up is dealt within the of them having recorded maximum widths of less than
numericalmodelsoftheensemble,itseffectsaremanifested 20m (Fig. 2). Regarding beach sediment type, three dif-
intheresultsonlyifitinducessedimenttransportthatforces ferent classes were recorded: (i) sandy beaches (36.4%),
morphological changes (e.g., Leont’yev, 1996). However, (ii) gravel/pebbled beaches (44.4%), and (iii) cobble/rocky
waverun-upinducedtemporaryfloodingthatdoesnotresult beaches (about 2%). For the remainder of the beaches, no
in beach retreats might also be a significant beach manage- sedimenttypecouldbeassignedonthebasisoftheavailable
ment issue (e.g., Jiménez et al., 2012; Hoeke et al., 2013). information(Fig.2).
Therefore,estimationsofwaverun-upexcursion/inundation Beachorientationwasexaminedby(i)takingintoconsid-
were also undertaken on the basis of run-up heights; these eration the eight main compass directions and (ii) consider-
were estimated for all tested conditions, using the expres- ing the eight main compass directions for each of the eight
sions of Stockdon et al. (2006) which have been validated island complexes of the Aegean Sea – N. Aegean, Dode-
for the beaches of the Aegean archipelago (Vousdoukas et canese,Cyclades,Crete,SWAegean,Argo-Saronic,Euboea
al.,2009b): (Evia), and Sporades, Fig. 1. In the first case, occurrence
Nat.HazardsEarthSyst.Sci.,17,449–466,2017 www.nat-hazards-earth-syst-sci.net/17/449/2017/
I.N.Monioudietal.:Assessmentofislandbeacherosionduetosealevelrise 455
Figure2.StatisticsoftheAegeanarchipelagobeaches.(a)Piediagramsofmajorcharacteristics;(b)relationshipbetweenthebeachmaxi-
mumwidthandthepresenceofrivermouths;and (c)boxplotsshowingrelationshipsbetweenbeachwidthsandsedimenttypes.
of beaches with S, SE and SW orientations appears more asignificantcontrolforbothbeachwidthandsedimenttype;
prevalent (16.1, 13.9 and 12.5% respectively) than that of beaches with river mouths are more likely to be wider and
beachesfacingtowardstheotherfivecompassdirections;this builtonfinersediments(sands)(Fig.2andTable1).
wasfoundtobestatisticallysignificant(Table1).Prevailing Another interesting finding is that perched beaches (Gal-
beach orientations were also found to differ over the differ- lop et al., 2012) form a significant fraction of the Aegean
ent island complexes. For example, the prevailing (statisti- archipelago beaches; beachrock outcrops (e.g., Vousdoukas
callysignificant)beachorientationswerefoundtobetowards etal.,2009a)wererecordedonthebeachfaceof23%ofall
the SE (18%) in the Dodecanese islands and towards the S beaches.Intermsoforientation,althoughthereappearsthat
(22%)inCrete.Thereisalsosignificantcorrelationbetween beacheswithbeachrocksare moreprevalentatthesouthern
beach orientation and recorded maximum width. Statistical coasts of islands (192 out of 744 total occurrences), statis-
tests (Table 1) showed that W and SW facing beaches are tical testing did not show any significant trends (Table 1).
associated with greater maximum widths (24.5 and 24.0m, Sincetherecouldbeasignificantnumberofbeacheswhich
respectively). may contain buried beachrocks (i.e., not outcropping at the
Theaboveresultsindicatesomehydrodynamiccontrolin time of the analyzed imagery), it seems that beachrocks are
the development and maintenance of the Aegean beaches: quitewidespreadalongtheAegeanislandbeaches.
therearemoreandwiderbeachesalongtheislandcoaststhat Coastal protection schemes have been recorded at about
are(atleast)partiallyprotectedfromtheprevailingnortherly 15% of the Aegean island beaches (Fig. 2a). In compari-
wind/waves (Soukissian et al., 2007, 2008). However, the son,manybeachesareassociatedwithcoastaldevelopment:
correlation is weak (Table 1) due to other important factors 80.8%ofallbeachesfrontpublicandprivateassetssuchas
controlling island beach development/maintenance, such as coastalroads,housingandtouristinfrastructure.Thedensity
theantecedenttopographyandgeologicalhistory,theterres- of these assets is variable, ranging from a coastal road and
trialsedimentsupplyandthehumandevelopmentateachin- afewhousesfoundinremoteislandbeachestotheplethora
dividualbeach. ofvaluablepublicandprivateassetsfoundbehindtheurban
In terms of the terrestrial sediment supply, few beaches beaches of, e.g., Heraklion (Crete) and Rhodes. Generally,
(about 18%) were found to be associated with intermittent about32.7%oftheAegeanbeachesfrontcoastalinfrastruc-
(very rarely permanent) flow river mouths; most of those tureandassetswithmoderatetohighdensity.Theconsider-
were found in the large islands of the North Aegean (314 ablecoastaldevelopment,coupledwiththenarrowwidthsof
outofatotalof728beaches).Riverinesupplyappearstobe
www.nat-hazards-earth-syst-sci.net/17/449/2017/ Nat.HazardsEarthSyst.Sci.,17,449–466,2017
456 I.N.Monioudietal.:Assessmentofislandbeacherosionduetosealevelrise
Table1.StatisticalanalysisofthecharacteristicsoftheAegeanarchipelagobeaches.
Comparison NullHypothesis(H0) Statisticaltestresults Observations
Beach orientation and H0:thedataareuniformlydis- Rayleigh’s test: H0 rejected Prevailingbeachorientationto-
occurrence tributed around the compass (p<0.001). Beach orientation not wardsthesouth(S)(occurrence
rose uniformlydistributedoverthecompass 16.1%)
rose.
Beachorientationand H0:theeightcompassorienta- Kruskal–Wallis (Analysis of variance WandSWfacingbeacheshave
maximumwidth tions have equal mean values byranks):H0rejected(p<0.001). greatermaximumwidths(mean
regardingmaximumwidth Mean beach maximum widths differ maximum widths of 24.5 and
(µ1=µ2=...µ8) withorientation. 24m)
H0: ρM=0, no correla- Mardia’s correlation coefficient (Zar,
tioninthedatapopulation 2010):H0rejected(p<0.01).
Significantcorrelationbetweenorienta-
tionandmaximumwidth.
Beach orientation and H0:thetwovariablesareinde- χ2 testsfor(i)fourand(ii)maincom- Nosignificanttrend
beachrockoccurrence pendent passorientations:
H0rejected(p<0.001).Smallassocia-
tion power, Cramer’s (i) V =0.13 and
(ii)V =0.17.
Sediment type and H0: beach sediment type inde- χ2test:H0rejected(p<0.05).Signifi- Sand-sized beach sediments
presenceofrivermouth pendentoftherivermouthpres- cantcorrelation,butsmallpowerofas- are more associated with river
ence sociation(Cramer’sV =0.064). mouths
Beachmaximumwidth H0: the sets are independent Student’sttest(two-tailedtest): Beaches with large maximum
and presence of river (meanvalues H0rejected(p<0.001). widthsareassociatedwithriver
mouth µ1=µ2) mouths(Fig.2b)
Beach sediment type H0: mean values of the three ANOVA,aftersplittingthedatasetinto Beacheswithcoarsesediments
andmaximumwidth categoriesareequal thethreesedimenttypes: are associated with smaller
µ1=µ2=µ3 H0rejected(p<0.001). maximumwidths(Fig.2c)
theAegeanislandbeachesincreaseexposureunderavariable pendent of the beach sediment texture. Generally, models
andchangingclimate. showeddifferentialsensitivitytoinitialconditionsandforc-
ing,whichjustifiestheircollectiveuseinensembles.
4.2 Predictionsoferosion/retreatandfloodingforthe Modelresultshavebeencomparedwiththosebyphysical
Aegeanislandbeaches experimentsconductedatthelargewaveflumeinearly2013
(details in Vousdoukas et al. 2014). In these experiments,
the initial slope of the beach was set to about 1/15, tested
4.2.1 Modelsensitivityandvalidation
waveshadanoffshoreheight(H)of1mandaperiod(T)of
5s, and the seabed consisted of well-sorted sand of median
Model sensitivity tests undertaken within the present study
grainsize(d )equalto0.3mm.ThreeSLRscenarioswere
haveshownthatbeachretreatsarecontrolledbybeachtypol- 50
tested(risesof0.2,0.4and0.6m),withtheinitialprofilesof
ogy. All models show higher retreats with decreasing Irib-
theseexperimentscontrolledbythefinalprofile(waveforc-
arren numbers – i.e., for beaches with milder beach slopes
ing only) of the first experiment (i.e., the test without level
(β)and/oroffshorewavesofincreasedsteepness(H /L )–
o o
rise).Simulationtimesweresetto3000s.
withtheexceptionoftheBruunmodelwhoseresultsarein-
In Fig. 3, profiles by the numerical models Leont’yev,
dependentofξ forlinearprofiles.Themostsensitivemodels
SBEACH, XBeach and Boussinesq under the baseline level
to beach slope are the Edelman and SBEACH models, and
aswellasthreeincreasedsealevels(+0.2,+0.4and+0.6m)
the least sensitive is the XBeach model; with regard to the
are compared against those that resulted from the physical
offshorewaveclimate,XBeachappearstobethemostsensi-
experiments. It appears that there are some discrepancies,
tive, while Leont’yev and Boussinesq appear to be the least
particularly with regard to the dynamics of the offshore bar
sensitivemodels.Theeffectofsedimenttextureisnotalways
andtrough.TheLeonty’evandXBeachmodelsseemtocon-
clear,althoughaweaknegativecorrelationbetweenbeachre-
siderably“smooth”thesefeatures,whereastheSBEACHand
treatandthemediansedimentsize(d )mightbediscerned
50
Boussinesqresultstendtobetterrepresenttheresultsofthe
in the numerical models; analytical model results are inde-
Nat.HazardsEarthSyst.Sci.,17,449–466,2017 www.nat-hazards-earth-syst-sci.net/17/449/2017/
I.N.Monioudietal.:Assessmentofislandbeacherosionduetosealevelrise 457
physical experiments; the best performance was that of the the lowest projections from the four numerical models) is
Boussinesqmodel(Fig.3).Nonetheless,thereappearstobea given by S=0.1α2+9.7α+0.4 and the “high” mean by
goodagreementbetweenthebeachretreatsestimatedbythe S=0.7α2+28.5α+4.8(Fig.5a),whereSisthebeachretreat
models and those recorded during the physical experiments andαistheSLR.Also,thelowprojectionmeanofthelong-
(Fig. 3 and Table 2). The Leont’yev, SBEACH and Boussi- termensembleisgivenbyS=0.1α2+10α+0.3andthehigh
nesqmodelsappeartoslightlyunderestimatebeachretreats, projectionmeanbyS=1.6α2+29.8α+2.3(Fig.5b).Ranges
and XBeach appears to overestimate them. The Bruun and in beach temporary inundation/flooding due to wave run-
Edelmanmodelsappeartooverestimatebeachretreatsunder up combined with (α) episodic (short-term) SLRs (Fig. 5c)
higherSLRs,whereastheDeanmodelappeartooverestimate were estimated as S(i)=0.1α2+9.7α+4 (minimum) and
themunderallSLRstested. S(i)=−0.7α2+31.2α+27.2 (maximum) and with long-
Whenthemodelsareusedinanensemblemode,thecom- term SLRs (Fig. 5d) as S(i)=0.4α2+10.1α+3.7 (mini-
parisonbetweentheretreatsprojectedbythenumericalmod- mum)andS(i)=0.4α2+30α+28.1(maximum).
els(short-termretreat)andthoserecordedinthephysicalex- Ranges of decreases in dry beach widths were projected
perimentsimproves(Table2).Conversely,analyticalmodels through the comparison between the ranges of beach re-
tend to overestimate beach retreats by up to 20%. Interest- treat/erosion (S) and the recorded maximum widths of the
ingly, when the results of all (7) analytical and numerical 3234 pocket beaches. Ranges in beach temporary inunda-
modelsarecombinedinaunifiedensemble,thenretreatover- tion/floodingwereestimatedbythecomparisonbetweenthe
estimationsdecreasetolessthan10%forthetestedSLRs. ranges of combined beach retreat and wave run-up excur-
The above results refer to natural profiles. In order to re- sions(S(i))andthebeachmaximumwidths.InTable4,es-
latethevalidationtothelinearprofilesusedhere,furthertests timationsarepresentedfornineSLRscenarios:(i)0.15,0.5
wereundertaken.Inthesetests,themodelswererunusingan and0.7mMSLRsaccordingtorecentprojectionsforthearea
“equivalent”linearprofilehavingaslopeof1/15tocompare (Hinkeletal.,2014)usingthelong-termensemble;(ii)short-
withtheresultsofthephysicalexperiments.Theequivalent term SLRs due to storm surges and waves of 0.2, 0.4 and
profile was estimated using the best linear fit to the natu- 0.6m using the short-term ensemble; and (iii) combined
ral profiles used in the physical experiment (between water MSLRs and short-term SLRs of 0.55, 1.1 and 1.3m using
depthsof3.5andelevationsof1.5m). thelong-andshort-termensemblesconsecutively.
The results of this exercise are shown in Table 3. Gener- According to the projections of the long-term ensem-
ally, the comparison showed that, at least for the conditions ble, MSLRs of 0.15, 0.5 and 0.7m will result to beach re-
tested,theresultsofthemodelssetwiththeequivalentlinear treats/erosion of about 1.8–6.8, 5.3–17.6 and 7.3–24m, re-
profilewerereasonablyclosetothoseofthephysicalexper- spectively,whereasshort-termensembleestimatesshowthat
iments,withthecomparisonworseningwithincreasingwa- storm-induced levels of +0.2, +0.4, +0.6m could result in
terlevels.Whenthenumericalmodelsareusedasan(short- rangesofbeachretreat/erosionofabout2.3–10.6,4.3–16.3,
term)ensemble,beachretreatforcedbywavesisonlyoveres- 5.3–19.3and6.2–22.2m,respectively(Table4).Forthecon-
timated;underSLR,however,ensembleprojectionsimprove sidered SLR scenarios, there will be significant impacts on
(overestimations of up to 19%, Table 3). Regarding the an- theAegeanarchipelagobeachesasshownbythepercentages
alyticalmodels,thelong-termensemblegaveimprovedpro- ofbeachesthatareprojectedtobeeroded/shiftedlandwardto
jections, with differences between models and experiments adistanceequaltobetween50and100%oftheirmaximum
rangingfromabout3to11%. width(Table4).
Even under an MSLR of 0.15m (RCPs 4.5 and 8.5 in
4.2.2 Beacherosionandinundation/flooding 2040),therecouldbesubstantialimpactsonthebasisofthe
projections mean minimum and maximum projections of the long-term
ensemble(Table4).Temporaryinundation(S(i))couldover-
Modelingresultsshowthatsealevelrisewillcauseshoreline whelmbetween4.9and81%ofthebeaches,flooding(occa-
retreatsaswellassignificantmorphologicalchanges(Fig.4). sionally) 4.6–80% of the beaches fronting currently exist-
Numerical models showed differential profile changes in ing coastal infrastructure/assets. Under an MSLR of 0.5m
both breaker and surf zones, supporting their use in an en- (RCP 4.5, 2100), projected impacts will be severe. From 5
semblemode. to 54% of all Aegean beaches will be completely eroded
Models displayed differential behavior for almost all in the absence of appropriate coastal defences, whereas be-
tested conditions, showing, as expected, significant ranges tween about 18 and 90% of all beaches are projected to be
of results (Fig. 5) due to the varying initial conditions and occasionallyoverwhelmedbyflooding.For2100,underthe
forcingused,i.e.,differentbedslopes,sedimentsizes,wave highemissionscenario(RCP8.5,MSLRof0.7m),impacts
conditions and SLRs. The means (best fits) of the lowest couldbecatastrophic(Table4andFig.6):12.4–68and28–
and highest projections of all models were calculated. It 93% of Aegean island beaches will be completely eroded
was found that the “low” mean of the beach retreat pro- andoccasionallyoverwhelmedbyflooding,respectively.As-
jections by the short-term ensemble (i.e., the best fit of sociatedinfrastructure/assetsarealsoprojectedtobegreatly
www.nat-hazards-earth-syst-sci.net/17/449/2017/ Nat.HazardsEarthSyst.Sci.,17,449–466,2017
458 I.N.Monioudietal.:Assessmentofislandbeacherosionduetosealevelrise
Figure3.Profilesbynumericalmodelsplottedagainstresultsfromphysicalexperimentsatthelargewaveflume(GWK,Hanover):(a)ini-
tial/present water level; (b) water level rise of 0.2m; (c) water level rise of 0.4m; and (d) water level rise of 0.6m. Both numerical and
physicalexperimentsweresetupforthesameinitial(nonlinear)profile.Simulationtimesforthenumericalexperimentssetto3000sto
matchthoseofthephysicalexperiment.
Table 2. Comparison of results from models and physical experiments. Negative values represent erosion. Key: SLR, sea level rise; Ex-
per,physicalexperiment;Leo,Leont’yevmodel;SB,SBEACHmodel;Xb,XBeachmodel;Bous,Boussinesqmodel;STEns,Short-term
ensemble;Br,Bruunmodel;Edel,Edelmanmodel;Dean,Deanmodel;andLTEns,Long-termensemble.
SLR(m) Shorelineretreat/erosionoradvance/accretion(m)
Exper Leo SB Xb Bous STEns Br Edel Dean LTEns
0 −0.35 −0.01 0.02 −2.32 −0.06 −0.59 – – – –
0.2 −3.57 −3.2 −3.04 −5.63 −3.0 −3.72 −3.01 −3.09 −6.7 −4.27
0.4 −6.23 −5.94 −5.83 −8.35 −5.6 −6.43 −6.13 −6.43 −8.82 7.13
0.6 −8.66 −8.22 −8.02 −10.27 −7.9 −8.60 −9.25 −9.97 −11.5 −10.24
impacted, with 12–66 and 26–92% of all beaches fronting backofthebeach.About27and51%ofallAegeanbeaches
currentlyexistingassetsprojectedtobelosttobeacherosion will retreat by more than their maximum width and 82 and
andoccasionallyoverwhelmedbyflooding,respectively. 87%willbecompletelyoverwhelmedbytemporaryflooding
According to the mean low projections of the short-term under short-term SLRs of 0.2 and 0.4m, respectively (Ta-
ensemble,stormcoastalsealevelsof+0.2and+0.4mwill ble4).
result in moderate (temporary) beach retreats (0.1–2.3% of Inthecaseofa0.4mshort-termSLR,upto50and86%
allbeacheswillretreatmorethantheirmaximumwidth)and of all beaches fronting assets will be affected by beach re-
temporary flooding (5–12.7% of the beaches will be occa- treat/erosion and temporary flooding, respectively (Fig. 6),
sionallycompletelyflooded).Onthebasisofthemeanhigh whereasimpactswillbemoresevereunderhighercoastalsea
projections(forcedbythehighwaveconditionsexpectedin levels(+0.6m,Table4).Thepicturedoesnotchangemuch,
storms, see also Tsoukala et al., 2016), beach retreats and evenwhenprojectionsareadjustedforthemaximumoveres-
flooding will be substantial with severe reductions in dry timationobservedintheresultsofthenumericalmodels(by
beach widths and potential damage to assets located at the 19%)whenlinearprofileswereused(Sect.4.2.1);forexam-
Nat.HazardsEarthSyst.Sci.,17,449–466,2017 www.nat-hazards-earth-syst-sci.net/17/449/2017/