Rapid assessment of flood susceptibility in urbanized rivers using digital terrain data: Application to the Arno river case study (Firenze, northern Italy)

lable at ScienceDirect

Applied Geography 54 (2014) 35e53

Contents lists avai

Applied Geography

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

Rapid assessment of flood susceptibility in urbanized rivers usingdigital terrain data: Application to the Arno river case study(Firenze, northern Italy)

Stefano Morelli*, Alessandro Battistini 1, Filippo Catani 2

Department of Earth Sciences, University of Firenze, Via La Pira n�4, 50121 Firenze, Italy

Keywords:Urbanized riverDikeOverflow susceptibilityFlood path modelingFlow directionArno river

* Corresponding author. Tel.: þ39 055 2757782; faxE-mail addresses: [email protected] (S. Mo

gmail.com (A. Battistini), [email protected] (F. Cat1 Tel.: þ39 055 2757548; fax: þ39 055 2756323.2 Tel.: þ39 055 2757559; fax: þ39 055 2756323.

http://dx.doi.org/10.1016/j.apgeog.2014.06.0320143-6228/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

As a result of the recent acquisitions of detailed geographic data in fluvial environments this studyproposes an analytical procedure in a GIS environment for an easy, rapid and cost-effective evaluation offlood susceptibility within urban and sub-urban areas. The method, based on data gathering in variouspublic offices and possibly integrating missing geomorphologic information with new records, has beentested in a real case study in central Italy from which operational instructions could be extrapolated forother similar areas in the world. A long reach of the Arno river, which is currently the most importantnational hydraulic emergency area in terms of civil protection, was selected for this purpose. Themeasured dike heights, extrapolated from a geographic database previously created by the authors onbehalf of the Provincia di Firenze administration, were compared with the results of hydraulic modelingconducted by the Arno River Basin Authority in which water flood levels were simulated for variousreturn periods. This data comparison has allowed us to identify critical points for overflow hazard. Inaddition, the flood discharges provided by the same model were related to the freeboard typically used inriver hydraulics for the major Italian watercourses (1.00 m for a 200-year flood). A deeper analysis of theflood susceptibility was subsequently carried over to the urban area of Firenze. A model of the surfacehydrological flows concentrated on the historic center has provided a comprehensive response of thisarea to the sudden appearance of abundant surface-water flows. The behavior of the urban water cir-culation in terms of path and areas involved were promptly identified with great precision during normalflow conditions of the Arno in the event of river obstruction at bridges and in case of undifferentiatedrunoff out from the riverbed.

© 2014 Elsevier Ltd. All rights reserved.

Introduction and problem description

Since ancient times, floods and flash floods (Barredo, 2007;Doswell, Brooks, & Maddox, 1996) have marked milestones in thedramatic histories of many countries, not only for the grievingtribute to human victims (Guzzetti, Stark,& Salvati, 2005; Penning-Rowsell, Floyd, Ramsbottom, & Surendran, 2005; Salvati, Bianchi,Rossi, & Guzzetti, 2010; Taubenb€ock et al., 2011) and irreparabledamage to cultural heritages (Drd�acký, 2010; Herle, Herbstov�a,Kupka, & Kolymbas, 2010; Holický & Sýkora, 2010; Lanza, 2003)but also for the extraordinary nature of the rainfall beating down on

: þ39 055 2757788.relli), alessandro.battistini@ani).

the basins (Brauer et al., 2011; Kale, 2003; Winston & Criss, 2002).Disasters associated with inundation in urbanized areas are causedby exceptionally intense rainfall events as well as by destructiveaccidents to manmade structures (such as dike erosion or slump-ing). They may cover large areas in the event of flooding from ahigh-flow river that disperses its excess water in a densely popu-lated conurbation (Krellenberg, Müller, Schwarz, H€ofer, & Welz,2013; Yen, 1995). However, a combined study of the physical andsocietal processes is essential to determine the degree of impact ofa flood event. For this reason, fluvial risk management should beproperly considered in urban planning (Tran, Shaw, Chantry, &Norton, 2009).

Moreover, because any intervention may require several de-cades before new measures are carried out (Merz, Hall, Disse, &Schumann, 2010), every decision concerning new territorialtransformations should also take into account potential futurechanges, such as economic development or climate variations.

S. Morelli et al. / Applied Geography 54 (2014) 35e5336

From a purely technical perspective, it is necessary to finalize aneffective system for improving flood risk preparedness and gath-ering information about the temporal distribution of the watersupply, the geographical distribution of the fluvial morphologiesand manmade structures, as well as the physiographic confor-mation of the surrounding area. For several rather extended riversof the world that cross urban areas, these data are qualitativelyand quantitatively insufficient (e.g., along the Klamath andMahanadi rivers in Oregon (United States) and in India, respec-tively; Patro, Chatterjee, Singh, & Raghuwanshi, 2009; Risley, Hess,& Fisher, 2006). In many cases, however, a rather large amount ofdata exists (e.g., along the Mississippi, Rhine and Meuse rivers inthe United States, Germany and Netherlands, respectively; Santato,Bender, & Schaller, 2013), but these data have not been exploitedto their full potential (Brown et al., 2013; Zerger & Wealands,2004).

Although hydrologically innovative and robust models arecurrently available for processing such data, they are poorly suitedto expeditious applications, are often not well integrated withinspatial modeling environments (GIS) and are not capable of non-expert management (Al-Sabhan, Mulligan, & Blackburn, 2003;Kulkarni, Mohanty, Eldho, Rao, & Mohan, 2014). These modelsrequire considerable expertise in hydrological data and the modelsand are unsuitable for rapid applications in extended fluvial areasbecause of the types and the interactive nature of the data required.On the basis of these assumptions and considering the currentmethods for obtaining urban flood prediction (Boonya-Aroonnet,Maksimovic, Prodanovic, & Djordjevic, 2007; Clement, 2013;Hammond, Chen, Djordjevi�c, Butler, & Mark, 2013; Henonin,Russo, Mark, & Gourbesville, 2013), we have developed an easyand cost-effective sequence of operating procedures to rapidlyevaluate flood susceptibility along very extended urban and sub-urban fluvial reaches. The proposed approach is mainly based onprocessing operations in a GIS environment using data collected bydifferent public agencies and possibly integrating missing geo-morphologic information with new data (Fig. 1). The latter usuallyare quickly obtainable and at reasonable costs, as already demon-strated by Casas, Benito, Thorndycraft, and Rico (2006) and Morelli,Segoni, Manzo, Ermini, and Catani (2012). This approach allows themapping of overflow susceptibility along anthropized rivers. It alsoallows us to define some detailed hydraulic risks within historicareas deriving from the analyses of the current urban morphology,especially with respect to our understanding of the paleogeographyand major historical flood events.

Fig. 1. Operative procedures proposed for rapid evaluations of the flood susceptibility inrepresents information from the local authorities while the green typeface represents new infigure legend, the reader is referred to the web version of this article.)

This paper thus aims to: i) readily use available data from thepublic domain; ii) suggest to public administration agencies how toprofitably use the territorial data in assessing overflow suscepti-bility; iii) develop a cost-effective and practical method of evalu-ating flood susceptibility in historic urban areas (in terms of thepath and impacted area); iv) apply this method to a real case (fromwhich operational instructions can be extrapolated for othersimilar areas in the world) showing that much important infor-mation for landmanagement can be obtained with just a few steps;and v) indicate to expert users where it is really most necessary tomake a specific hydraulic model or undertake a particular stabilityanalysis of the anthropic works for optimal safety management.

We chose the Arno river for this purpose because it is an Italianwatercourse with a greater hydraulic hazard. Indeed, it has sufferedfrom floods since the foundation of the ancient Roman city. Anddespite numerous structural interventions undertaken over time tosolve this problem, the Arno river still represents the second mostserious national hydraulic emergency in terms of civil protection(Morelli et al., 2012). Hence, we consider the Arno river animportant case study that deserves further investigation; wetherefore examined 75 km of its length in the present work. Such astretch corresponds to the extension of a state-of-the-art, high-precision GPS measurement campaign we recently performed in ashort time and published in a user-friendlyWebGIS of the Provinciadi Firenze Administration. This campaign provided for the first timea large amount of new and useful morphometric data about thepattern and shape of the dikes (made of earth, masonry or com-bined solutions), as well as the horizontal and vertical localizationof their most significant structural elements and critical featureswith centimeter accuracy (Morelli et al., 2012; Segoni et al., 2008).Additionally, with regard to the urban flooding analysis, themetropolitan area of Firenze was chosen because many of itsproblems are typical of historic cities built on large floodplains(Cembrano et al., 2004; Coulthard & Frostick, 2010; Daungthima &Kazunori, 2013; Marco& Cayuela, 1994; Schmitt, Thomas,& Ettrich,2004). In fact, this city, which was declared aWorld Heritage Site byUNESCO in 1982, is mainly affected by: i) floods of the Arno river(e.g., the population still remembers the disastrous flood of 1966and several flood marks on the main buildings of the city recall theflood of 1844); ii) outflows of water from subterranean canals (e.g.,in August 1984, shortly after its completion, the Mensola Creekbroke the tubing in which it was contained and spilled over); iii)outflows directly coming from the sewage system (e.g., on June 5th,2011, the sewage system could not collect all of the 55 mm of rain

long-anthropized fluvial stretches based mainly on existing data; the black typefaceformation obtained by the authors. (For interpretation of the references to color in this

S. Morelli et al. / Applied Geography 54 (2014) 35e53 37

that fell in about 1 h); iv) leakage from broken pipes (e.g., onFebruary 10th and 18th, 2008, two pipelines dating back more than50 years broke under the strategic ring road boulevards and alongtwo avenues close to the Arno river, creating security and hygieneproblems for the citizens, while on July 25th, 2011 an old cast ironpipe emplaced in the 1920s broke and flooded an ancient streetapproximately 300 m from the famous cathedral dome).

Study area

The Arno river

The study was undertaken in the Arno River Basin (central Italy).It included an interrupted long stretch of the Arno river withinFirenze Province for a total length of 75 km, which is slightly lessthan one third of the entire watercourse. In particular, the analyseswere concentrated along the riverbanks from the administrativeboundary with Arezzo Province at the confluence with Vaccher-eccia Creek in the Upper Valdarno to the boundary with PratoProvince at the mouth of the Bisenzio tributary (right bank) and tothe morphological narrowing downstream of the village of Bru-cianesi (left bank), both in the Middle Valdarno (Fig. 2). Withinthese limits the river flows through 15 different municipalities.

The entire course of the Arno river is strongly influenced by themorphology of the basin and especially by the occurrence of a se-ries of alluvial plains of different areal extension separated bynarrow thresholds that represent mainly calcareous and arena-ceous rock outcroppings (Canuti, Cencetti, Rinaldi, & Tacconi, 1994;Morelli et al., 2012; Tacconi,1994) (Fig. 2). Except for these stretches

Fig. 2. The stretch of the Arno river in which the analytical methodology has been implemValdarno sub-basin and the BeC distance represents the riverbed, which crosses the Middbetween narrow thresholds and larger alluvial plains) have been highlighted through the sustudy area in which the whole basin and the entire Arno river is shown. (For interpretation ofthis article.)

the entire course tends to have a quite mobile riverbed, which,however, has been greatly controlled in urbanized sites sinceancient times (Becchi & Paris, 1989; Billi, Rinaldi, & Simon, 1997;Rinaldi & Simon, 1998; Rinaldi, Simon, & Billi, 1997; Surian &Rinaldi, 2003). For example, numerous adjustments of theriverbedwere carried out in a large part of the Florentine floodplainsince the 12th century (Natoni, 1944), transforming it from ananastomosing river into a rectilinear channel with an extremelyreduced width (150 m at most instead of the 1000 m based onhistorical information) (Canuti et al., 1994).

From a hydrological point of view, the Arno river basin covers anarea of 8228 km2 (Morelli et al., 2012) on soils largely characterizedby a medium-low permeability (Autorit�a di Bacino del fiume Arno,2008). For this reason the runoff is mostly influenced by the trendof the precipitations, which fall in the drainage areawith an unevendistribution due to the great variation of the basin orography(Autorit�a di Bacino del fiume Arno, 2004). The flows are charac-terized by two maxima during the rainy period (December andMarch) and an absolute minimum (August) (Fig. 3), with an actuallag time between rainfall and peak runoff, which is generally due tothe seasonal conditions of the soil and depletion flows. This averagelag time is called the time of concentration and is approximatelythe time needed for water to flow from the most remote point in awatershed to the watershed outlet (Haan, Barfield, & Hayes, 1994).Several systems are available for calculating the time of concen-tration. In a complex system such as the Arno basin, the time ofconcentration can be roughly estimated through the contributionsof the sub-basins of which it is composed. The main sub-basin datawere obtained from Arno Basin Authority (sub-basin used for

ented: the AeB distance represents the riverbed in the terminal portion of the Upperle Valdarno sub-basin. The different morphological areas of the riverbed (alternationbdivision of the watercourse using red marks. In the middle image an overview of thethe references to color in this figure legend, the reader is referred to the web version of

Fig. 3. Average daily runoff and rainfall from 2002 to 2012 (both included) immediately downstream of Firenze (data from the Servizio Idrologico Regionale e Regional HydrologicalSurvey: http://www.sir.toscana.it/, last visited May 14, 2014).

S. Morelli et al. / Applied Geography 54 (2014) 35e5338

Water Balance purpose, http://www.adbarno.it/, last visited April12, 2014). The sub-basins upstream of Firenze (Fig. 4) include 16sub-basins (the complex system of eastern sub-basins is mainly dueto channeling in the Chiana river valley). Their characteristics andtheir corresponding times of concentration, calculated for this workusing the method proposed by Giandotti (1934), are shown in

Table 1. The total average annual runoff of the entire basin isapproximately 3 billion cubic meters, with an average annual flowof 50 m3/s just upstream of Firenze and 100 m3/s at the closure ofthe basin (Campolo, Soldati, & Andreussi, 2003; Cortecci et al.,2009; Dapporto, Rinaldi, Casagli, & Vannocci, 2003). The compe-tent authorities estimated amaximum flow rate of 4500m3/s in the

Fig. 4. Sub-basins upstream of Firenze used for calculating the time of concentration.

S. Morelli et al. / Applied Geography 54 (2014) 35e53 39

center of Firenze during the last devastating flood (November 4,1966).

The urban area of Firenze

In the urban stretch of Firenze the investigation was expandedto include the floodplain covered by the Old Town center and itssurrounding neighborhoods, which together can be considered aspart of a larger “metropolitan area” according to the definition ofBerry, Goheen, and Goldstein (1968) (Fig. 5). The current geomor-phological feature of this area is the result of the evolution of afluvial-lacustrine basin, which originated in the Plio-Pleistoceneepoch as a consequence of tectonic movements during the exten-sional phase of the northern Apennines orogenesis (Bartolini &Pranzini, 1981). This area exhibits a NWeSE-oriented basin(Capecchi, Guazzone, & Pranzini, 1975), which developed anendorheic drainage system characterized by short watercourses(many of them antecedents to the structure) with a sporadic regimeand a significant sediment transport (Coli & Rubellini, 2007). Thealluvial deposits of the paleo-Arno and the other minor water-courses gradually covered the bedrock, which is mainly formed ofLigurian units, for a maximum thickness of 160 m (90 m in thecenter of the city) (Boccaletti et al., 2001).

This considerable sediment accumulation gave rise to a very flatalluvial area where the processes of fluvial dynamics slowly su-perseded the tectonic movements and became the foremost factor

responsible for the physiographic modification (Coli, Agili, Pini, &Coli, 2004). In fact, the floodplain has been subjected to thelateral erosion of watercourses and meandering, flooding andswapping (Bartolini & Pranzini, 1988). In the particular case of theFirenze urban area, the most relevant watercourses besides theArno river have been the Mugnone, Terzolle and Affrico creeks.Their natural dynamics operated undisturbed until the appearanceof the first human communities, whose activities have interactedwith the local environmental processes and from time to timeinfluenced the geomorphological evolution of the area (Rinaldi,2003) (Fig. 6). In particular, human dynamism in the study areahad its origins in the founding of the Roman settlement dating backto 56 BC (Hardie, 1965), although the first traces of artificialdeforestation can be dated back to the Neolithic age (Caporali,Rinaldi, & Casagli, 2005). Initially, the fluvial interventionsuniquely involved the urban area. They then gradually extended tothe rest of the plain through a sort of natural trade-off coexistencebetween the necessity of anthropogenic development and theinevitability of natural processes. On balance, this cohabitation hascontinued uninterruptedly until the time of reconstructionfollowing the Second World War.

The anthropogenic measures that have slowly shaped the wholearea over the centuries include the canalization (through fixedriverbanks), straightening and width reduction of the Arnoriverbed, the diversion of smaller watercourses through the exca-vation of subterranean canals (e.g., San Gervasio) or new riverbeds(e.g., Mugnone) and the reclamation of swampy areas in addition tothe completion of new urban plans (Fig. 6) (Billi & Rinaldi, 1997;Canuti et al., 1994; Cencetti & Tacconi, 2005; Gumiero et al.,2009; Rinaldi & Simon, 1998). On the other hand, during the last60 years large urban growth has become the only predominantmorphologic agent for the evolution of the area, which in a fewyears has lost all natural appearances because of the unrelentingwork on the watercourses during the previous ages. Thus, theprimary anthropogenic structures that currently affect the presenturban district of Firenze are the dense housing scheme, the intri-cate road network and the resulting flow-regulation works thatrigidly guide the urban watercourses throughout the area (Coliet al., 2004). Among these structures, the handiworks having agreater importance for the control of the overflow movement areboth dikes and fills and even the replacement of the surfacedrainage with artificial subterranean canals.

The latest watercourses to suffer this intervention are the AffricoandMensola tributaries, which were forced to flow underground in1956 and in 1984, respectively. Other important anthropogenicfactors are the raised railways, road overpass ramps and archaeo-logical fillings related to urban stratification. These undertakingsbrought the original topographic plan down to several metersbelow the current road plan, especially with respect to the oldRoman town in the present midtown (Coli & Rubellini, 2007). All ofthese features can force the direction of hydraulic surface runoff inthe case of rife water coming from abundant rains or broken pipes,not properly drained by the sewerage system or even in the event ofunstoppable overflows that may arise both from the superficialhydraulic network and from the 12 subterranean canals (in partialstretches or entirely) just within the urban area.

Methodology

Evaluation of dikes subject to overflowing

The procedure by which we attempt to determine the suscep-tibility to the dike overflowing in large fluvial areas has beendevised to be performed in a GIS environment with straightforwardoperational steps. All of the topographical data of the perifluvial

Table 1Main physical characteristics of the sub-basins upstream of Firenze and their cor-responding time of concentration calculated by means of the Giandotti method(1934).

N Basin name Basinarea(km2)

Basinminimumheight(m)

Basinaverageheight(m)

Mainriverlength(km)

Time ofconcentration(h)

1 ValdarnoMedio

169 35 240 23 7.6

2 ValdarnoSuperiore A

141 87 440 11 4.2

3 ValdarnoSuperiore B

471 96 403 23 8.7

4 ValdarnoSuperiore C

168 148 331 20 7.5

5 Casentino A 126 201 433 12 5.26 Casentino B 308 255 678 17 5.87 Casentino C 448 320 797 29 7.48 Sieve A 260 85 559 18 5.29 Sieve B 542 150 443 30 10.110 Sieve C 36 243 442 6 2.911 Ambra 206 145 376 24 7.712 Chiana A 97 207 354 3 4.513 Chiana B 326 222 347 24 12.114 Chiana C 447 240 340 17 13.815 Chiana D 398 242 294 31 21.816 Tresa 106 243 343 19 8.6

S. Morelli et al. / Applied Geography 54 (2014) 35e5340

zones and the dike mapping included in the available geodatabasewere inserted in a dedicated GIS and then overlaid with the track ofthe channel cross-sections used by the Arno River Basin Authorityto calculate the gauge heights of different floods. Such gaugeheights were estimated using the SIMI model (Sistema Informativodel Modello Idraulico e Hydraulic Model Information System)

Fig. 5. Land use map of the study area in relation to the Arno river (data from Geoscope,www502.regione.toscana.it/geoscopio/usocoperturasuolo.html, last visited May 20, 2014).

(Autorit�a di Bacino del fiume Arno, 2004) in the most criticaltransverse sections (from a hydraulic perspective) and in relation tofloods with a recurrence interval of 30,100, 200 and 500 years (T30,T100, T200 and T500, respectively). We should remember that theSIMI uses independent models for the hydraulic processes in theriverbed and flooded areas; however, these are connected by ananalytical scheme, which describes the transfer of the overflowingvolumes. In particular, it is composed of a one-dimensional un-steady flow model with respect to the riverbed and the semi 2-Dmodel, with a schematization of the storage areas with respect tothe overflowing water volumes. This type of modeling also con-siders the water contributions from the main tributaries. After that,by a linear interpolation of the available GPS points belonging to thedike crown, their maximum elevations were precisely identified oneach Basin Authority hydraulic cross-section. These accurate valueswere compared with the gauge heights for each of the four pre-dicted events in the SIMI model, which verified the real floodretention capacity of the defensive structures that are assumed toremain completely undamaged during such flood episodes.

Obviously, this method produces point-wise results. Thus,through morphological analysis of the dikes, the examination oftheir state of maintenance (Morelli et al., 2012; Segoni et al., 2008)and the on-site control of their position with respect to the an-thropic infrastructures, extensive information relating to the realcriticality of the floodplain was obtained for the whole study area.Moreover, the flood discharges provided by the Arno River BasinAuthority were also related to the freeboard of 1.00 m for a 200-year flood because it is considered quite precautionary for thecorrect functioning of the dikes along the major Italian water-courses (Da Deppo, 2006), just as in other countries exposed to highhydraulic hazards, such as the United States (California Departmentof Water Resources, 2012). The considered value is a safety factor,

the Geographic and Environmental Information System of the Tuscany region: http://

Fig. 6. Geomorphological evolution of the floodplain around the city of Firenze and its urban development from the ancient Roman settlement (A) to the 19th century (C).

S. Morelli et al. / Applied Geography 54 (2014) 35e53 41

S. Morelli et al. / Applied Geography 54 (2014) 35e5342

which adequately takes into account all of the approximationsrelated to the hydraulic modeling and the various local phenomenanot usually considered by the model, but which nevertheless mayoccur during a flood event (Huffman & Eiker, 1991).

Hydraulic modeling in an urban area

With regard to the analysis in the metropolitan district, all of theobtained information about the overflow risk were subsequentlyconnected with different surface flow models which we namedflow direction representations (FDR). These cover a large sector of theadjacent built-up area like awater flow network. Suchmodels focusonly on superficial water paths. What happens with flow pathswhen a layer of water occurs in the area and flooding through timeis not taken into the account. It can be assumed that most of theflow passes by streets and intersections (Mignot, Paquier, & Haider,2006 and references therein). The main areas of water accumula-tion are inside buildings below ground level (basements, cellars,etc. or on the ground floor with rising water). These areas, onceflooded, do not constitute a significant obstacle to water flowbecause each area is isolated from the others by walls and watertransition through these areas is estimated to be much lower thanthe flow along the roads.

These flow models have been created using an analytic proce-dure employing a high-resolution topographic model and simu-lating different hydrodynamic conditions of the Arno river becauseit represents the main hydraulic manifold of the urban area. Ac-cording to documented flow rate variations and flood occurrencesalong the entire Arno river, three possible situations are assumed asboundary conditions for the model: i) the water of the Arno riverflows regularly downstream and the riverbed can receive and drainfluids from the urban sewage network, as well as from the surfacerunoff without any particular problems (high flow rates, but undercontrol); ii) the bridges crossing the Arno river are obstructed andthe flows near them have difficulty moving entirely within theriverbed (considerable flow rates with local problems); iii) theamount of water is so great that it has great difficulty in flowingregularly throughout the Arno riverbed far from the bridges(exceptionally high flow rates with large problems). The firstoccurrence could be associated with an indirect flood (overflowsonly in the most upstream neighborhoods), a flash flood from apoint-wise dike collapse, a pluvial flood, an inundation from theminor urban drainage network (surface or underground structuralbreaks) or a combination of these. Instead the second scenariocould be relatedmainly to clogging of the bridge spans, followed bya rapid rise and diffuse dike overflow (direct flooding). However, inconjunction with this event, some of the situations hypothesizedfor the first case can also occur simultaneously. Finally, the thirdcondition could involve the occurrence of floods directly comingsimultaneously from the near riverbed and the most upstreamsectors with an indirect impact. This possibility involves greatamounts of water and a very diffuse overflow in the broad river-bank sectors. Consequently, the three proposed scenarios, whichfollow an order of increasing criticality, have produced threedifferent FDR models.

The Hydraulic Digital Model (HDM) of 1 m/pixel resolution (inthe Roma40 Gauss-Boaga West reference system) is the surfacemodel used for flow-path analysis under the hydraulic conditionsdescribed above (Fig. 7A). It arises from a preliminary mixingoperation between the Digital Terrain Model (DTM) (Fig. 7C) andthe Digital Surface Model (DSM) (Fig. 7B). Both are derived from anaerial laser scanning survey with LIDAR (Light Detection AndRanging) techniques (Webster, Forbes, Dickie, & Shreenan, 2004;Wehr & Lohr, 1999), which was performed in November 2007 bythe CGRCompany on behalf of themunicipality of Firenze bymeans

of a TopoSys Falcon II laser scanner. The operating features of thisdevice, which are suitable for highly detailed surveys from lowaltitudes, are listed in Table 2. Few graphic processing steps arerequired to obtain it. Useless information inside the buildings wereremoved from the DTM.

To do this, we used Regional Topographical Maps at 1:2000 scale(obtained from the cartographic offices of the Tuscan region). Thesemaps are characterized by 40 cm of planimetric and 30 cm ofelevation tolerance. Accurate building perimeters can be extractedfrom these maps. These perimeters were used to create a polygonalmask. This mask was used to remove the DTM data and to workonly on data outside buildings. Afterward, data gaps within thebuildings were filled in by elevation data derived from the DSM. Nofalse or soft raised obstacles, such as electric lines or trees, arepresent in the resulting model (HDM). During a flood simulationonly real hard obstacles exist, especially the blocks constituted bybuildings (Mignot & Paquier, 2003). In some cases even the bridgescan be considered significant obstacles and for this reason threedifferent flood scenarios connected to abundant rainfall eventswere hypothesized. The first one is the situation in which the Arnoriver can regularly catch and disperse downstream the possibleoverflow water and all of the runoff. In this instance all of the ob-stacles inside the riverbed, notably the urban bridges, have beenexcluded from the HDM model and were replaced with the hypo-thetical underlying ground elevation. The second circumstance iswhen the Arno river can still receive water from the external area,but the maximum level of the river is so high that bridges partiallyobstruct the flow and the river locally overflows. This has beensimulated through an HDM where the bridges have been consid-ered as buildings that join one riverbank to another. The thirdscenario is when the Arno river is not able to accommodate morewater, which accumulates extensively outside its banks. This hasbeen simulated through a simulated HDM in which the riverbedarea is filled by pixel values of very high elevation.

At this point the superficial water movements have been tracedin relation to each assumed river condition. As a first step, theelevation values are opportunely modified in the input raster,whereas small imperfections in the data can produce erroneouslydiscontinuous flows. This problem usually occurs when water istrapped in a cell that is surrounded by cells with higher elevations.Abnormal cells could be already present in the models or could becreated by interpolation that mask and merge operations (Peddle,Rabe, Soenen, & Johnson, 2005; Wu, Li, & Huang, 2008). Thus, the“sink” cells have been eliminated and replaced by adequate valuesobtained from the neighboring cells (average values). This is aniterative process becausewhen a sink is filled, the boundaries of thefilled area may create new sinks. Then, a raster of flow directionswas created from each terrain cell to its steepest downslopeneighbor (Doan, 2000; Tarboton, 1997; Tarboton, Bras, &Rodriguez-Iturbe, 1991). All of the information about the possiblewater flow direction was stored according to the following D8encoding: 1 ¼ East, 2 ¼ South-East, 4 ¼ South, 8 ¼ South-West,16 ¼West, 32 ¼ North-West, 64 ¼ North, 128 ¼ North-East. On thebasis of the obtained data, a raster of accumulated flow to each cellwas created.

In a Flow Accumulation Model, the value of each cell is thenumber of upstream cells that pour water into it according to thecalculations proposed by Jenson and Domingue (1988). The phys-ical meaning of such a model is simple: the channels are repre-sented by cells with the highest values, as they collect more waterfrom upstream cells, and the ridges are represented by cells withlowest values. In the specific study, the possible rain runoff andfloods have been represented by the higher value points. To betterhighlight the most important water lines network, evident linearrepresentations have been calculated that impose a significant

Fig. 7. The Hydraulic Digital Model (HDM) (A) used for the GIS analysis derived from a mixing operation between the Digital Surface Model (DSM) (B) and the Digital Terrain Model(DTM) (C).

S. Morelli et al. / Applied Geography 54 (2014) 35e53 43

threshold value. Boolean maps (true or false value, 1 or 0) werecomputed, considering whether a cell receives more or less waterthan an established value. These maps are represented by contin-uous lines of cells with the highest values.

Some methods of determining the threshold value and ofobtaining a drainage network consistent with geomorphology inmountainous regions are described in the literature (i.e., Tarboton& Ames, 2001; Tarboton et al., 1991), but no one has dealt withour case of study. Different threshold values would result in streamnetworks with different stream lengths and densities. However, thedetermination of the threshold value is subjective and could bebased on the calibration of the stream network from the spatialmodel matching with the one extracted from surveyed topo-graphical maps (Rumman, Lin, & Li, 2005).

For our specific situation, a threshold value of 3000 pixels (thatis 3000 m2 of water outflow) was deliberately chosen to study therunoff in the flat urbanized area of Firenze. Not having maps of thestream network in the case of flooding through the buildings, sucha threshold valuewas chosen after some testing, starting from a lowvalue (300) and increasing it until the water runoff paths reportedin 1966 flood chronicles (Messeri & Pintus, 2006) were clearlyidentifiable. The 1966 flood was selected as a reference because itwas as catastrophic as many others in ancient times and the generalurban plan of most of the city has remained almost unchanged untiltoday. Then, the lines of runoff were converted into a vector format(polylines) to associate each segment to a hierarchical order

according to the definition of the number of tributaries proposed byStrahler (1952). Subsequently, the watersheds of every segmentwere calculated from the information on flow direction and theabove-mentioned hierarchical subdivision. In the end, the mainwater flow directions, which were similar to a real river network,were overlapped onto the urban fabric to interpret the response ofthe urban area to different flood events. In this case, the reliabilityof the outcomes was evaluated by comparing them with all of theavailable data related to the morphology of the Arno paleo-coursewithin the drained areas presently covered by urbanization, theprogressive urban expansion during different historical periods andthe maximum areal spread of the most exceptional and devastatinghistoric floods (1333, 1589, 1740, 1844, and 1966).

Results

Assessment of the actual retention capacity of the dikes

We compared the heights reached by the water during the floodevents simulated by the Authority Basin with the up-to-date andaccurate dike measurements. We verified whether the levels ofeach given flood event, as calculated by the SIMI model, represent areal condition of hydraulic safety or on the contrary, if the defensivesystem has critical dike sectors inwhich thewater can overflowandspread in the outside floodplain. In such an evaluation all of thedikes less than 6 cm higher with respect to the flood level were

Table 2Technical features of the LIDAR survey performed by the CGR company.

Measurement rate 83,000 pulses per secondSwath width 390 m (at the max flight altitude 1500 m)Average measurement density 3 measurements/m2 (at the max

flight altitude)Average density per hectare From 30,000 to more than 60,000 pointsField of view 14�

Elevation precision (2 sigma) 15 cmPlanimetric precision (2 sigma) 50 cm

S. Morelli et al. / Applied Geography 54 (2014) 35e5344

considered overflowable because this value corresponds to themean error in the height positioning obtained with the GPS map-ping and processing (Morelli et al., 2012). Moreover, such a pro-cedure also takes into account the possibility that small turbulentwaves can be created during a flood event.

This evaluation proves that the dikes prone to overtopping arequite extensive, evenmuchmore than could have been expected byexamining the hydraulic hazard maps created by the Arno RiverBasin Authority (Table 3). However, only certain stretches have veryhigh criticalities and they increase in extent as the estimated peakflow levels grow. Their locations are on the whole shown in Fig. 8for each recurrence interval.

Regarding a 30-year flood, two areas are potentially subject tooverflowing. They correspond primarily to some point-wise dikeportions in front of the Incisa-Figline plain andmore extensively onthe dikes built to protect both sides of the river in the Firenze plain(starting from approximately 6 km downstream the historic centerof the city). For a higher recurrence interval (100 years), the analysisin these areas shows a gradual lateral expansion of the criticalitiescreating very hazardous scenarios, especially for those plots of landsubjected to recent urbanization. At the same time three newsectors at risk of overflowing appear in other parts of the water-course. One of these is located just in front of the historic center ofFirenze and it extends for approximately 250 m from the bridge“Ponte alle Grazie” to the National Library (Fig. 8A). Therefore, evenif the unfavorable dike protection is shortly extended, our currentanalysis highlights potential problems of inundation in the mostancient district of the city beginning with not-too-excessive flows.

With respect to a 200-year flood, a further area at high risk ofoverflowing is added to those previously mentioned. This includesthe entire dike protecting the village of Brucianesi, which is locatedin the narrow rocky gorge downstream from the Firenze plain. Inthis case it is expected that all of the dikes that protect the village(built both with earth and masonry) will be unable to contain thepredicted flood. Fortunately, in the other sites considered hazard-ous for lower recurrence intervals, new dike sectors at risk increaseslightly in length in front of the urbanized areas, while they enlargemore in areas facing agricultural or fallow lands (Fig. 8B).

A deeper discussion is worth having for the city of Firenze,whose historic center was severely damaged in 1966 by an unex-pected flood of just this magnitude (Campolo et al., 2003).Assuming such event will occur again, the portions of the dikessubjected to overflowing perilously increase their overall lengthrelative to those estimated for a 100-year flood. They expand for

Table 3Estimate of the dike lengths subjected to overflowing for different recurrenceintervals.

Dikes km %

Total extension 39.2 100Overflowable for T30 6.9 17.6Overflowable for T100 23.7 60.5Overflowable for T200 31.5 80.4Overflowable for T500 37.5 95.7

about 187 m in front of the historic center almost up to the UffiziGallery (Fig. 8A). Furthermore, a new critical sector appears on theopposite riverbank precisely where the Arno river exhibitsconsiderable narrowing due to the right-of-way structures of theSan Niccol�o weir. Now even the outskirts of the city have two sitesin which water flows may leak out dangerously. One is located atthe Varlungo bridge on the right riverbank (less than 4 km up-stream from the old town); the other is in the hydrographic leftriverbank at the residential district of Isolotto (less than 3 kmdownstream from the old town). In both cases, the stretches ofriverbank with dikes are discontinuous, are fairly close and arespaced out by extensive sectors without any flood defense systems.Some of these areas may become critical during more recurrentevents (for example during a 100-year flood). However, consideringtheir extremely limited distribution, the maximum heights reachedby the flood and themorphology of the riverbanks, water spreadingshould remain contained.

Finally, for events with a recurrence interval of 500 years, thetwo critical reaches identified downstream in the study area(Brucianesi village and the terminal portion of the Firenze plain)have no substantial increase in overflowable dikes because they arealready fully submerged by the 200-year flood. Conversely, somecriticalities are identified in the floodplains upstream of the city ofFirenze. In fact, for the first time small stretches at risk appear in thetown of Incisa despite the fact that the riverbed is quite deeply dugand the buildings are erected at a considerable height from themean water level of the river. The same issues also arise in thehighly urbanized area in front of the bridge of Figline (Fig. 8A). Thestretch between the two towns remains rather critical, as is sug-gested for events of more frequent occurrence. Regarding the urbanarea of Firenze, the three areas previously identified as over-flowable are significantly expanded, with serious threats for largesectors of the urban and suburban areas. In conclusion, the historiccenter sustains widespread flooding along the dikes in relation tothe flowprovided. Practically, there is an upstream expansion of thepreviously discussed critical areas on the right bank and the for-mation of new and extended areas at risk on the opposite riverbank(Fig. 8B).

Freeboard assessment

In this subchapter, the comparison between the heights reachedby the simulated floods and the GPS-measured dike heights hasfocused on verifying the existence of an appropriate safety marginin the dike structure above the hypothesized floodwater. Also inthis analysis, all of the dikes whose top is less than 6 cm higher thanthe flood level were considered overflowable because of all of theuncertainties in the adopted measuring instruments. Currentlysome portions of the Arno dikes, which have been found effectivefor centennial-period events, do not have sufficient freeboard toembank higher flows and they are easily overflowed by 200-yearfloods. In general, the dike geometry in many cases provides asafety freeboard of only a few centimeters, which cannot guaranteethe full functionality of the dikes that consequently are prone tooverflow during events with a slightly longer return period. All ofthese latter criticalities are generally noticeable along many dikestretches immediately adjacent to structures that have beenconsidered prone to overflow for 100-year floods. The greatest riskis certainly close to the village of Brucianesi (downstream from theGonfolina Gorge) in which the entire dike, despite keeping all thedwellings safe during centennial events, is completely subject toabundant overflow in the case of a 200-year event.

This highlights the existence of significantly undersized defen-sive structures in relation to the narrow morphology of the valley,which markedly influences the water levels.

Fig. 8. Localization of the dike subjected to overflow in relation to flood events with different return periods and their areal extension in the cities of Firenze (A) and FiglineValdarno (B).

S. Morelli et al. / Applied Geography 54 (2014) 35e53 45

Moreover, where the dikes have a hydraulic freeboard of lessthan 1 m for a 200-year flood, floods with a 500-year return periodare expected to submerge all of them everywhere except for thestretch within the historic center of Firenze (Fig. 9). Table 4 showsthe estimated volumetric flow rates and floodwater levels for eachconsidered section that can lead to such flooding. Instead, withregard to the safe zone, the wall dikes on the right bank have apositive and increasing difference between their real maximumaltitude and the simulated flood level moving from the westernoutskirts to the old town center at the medieval bridge namedPonte Vecchio (the freeboard varies from 0.52 m to 0.95 m)(Fig. 9A). The same trend persists even on the opposite riverbank,albeit with major differences in height (the freeboard varies from0.44 m to 0.93 m) (Fig. 9A). In both cases the freeboard increase isnot constant, but small local variations emerge from the analysis ofthe urban sections.

Description of the surface water flows in the urban area of Firenze

For each of the three hydrodynamic scenarios of the Arno river,explained in subsection 3.2, rather significant information has beenobtained about the response capacity of the urban environment tothe sudden circulation of great amounts of water described by eachof the three FDR models.

Basically, the distribution of flow lines within the urban fabricduring the full functionality of Arno dynamics (corresponding towell-regulated high flow rates and including the predisposition toserve as a rainwater collector in case of excessive flow for thesewerage system and as a rapid way of escape for indirect flood

waters), reveals the existence of several differently distributed sub-basins in the urban area. In particular, there are two fairly largebasins in the most upstream area of the city; given their widespatial extent, these are very important for the amount of waterpotentially drained and discharged in the Arno river (Fig. 10). Onecovers the right floodplain and the convergence point of all itsrunoff is located at a little square facing the Arno river approxi-mately 360 m upstream of the Ponte Vecchio. Therefore, its finalsection is located close to the historic center and in particular, infront of the most ancient bridge of Firenze. This is unequivocallydefined as the most dangerous, both because of its construction (asegmental arch bridge with small openings) and its position rela-tive to the urban spatial structure. The second basin extends alongthe opposite bank and includes a narrow strip of floodplain and alarge section of hills. Its closing section is located next to the Gio-vanni da Verrazzano bridge approximately 2 km upstream of itscounterpart on the right bank.

In the first case the drainage area is predominantly flat and isentirely urban. Furthermore, this sub-basin is physically separatedfrom the current riverbed by some micro-basins that represent anatural barrier against low-to-medium floods (Fig. 10A). Because oftheir geomorphological arrangement it was not even necessary tobuild local or extensive hydraulic defenses such as wall-dikes. Theanalyses concerning the spatial evolution of the Arno riverbed inthe past 2000 years shows that the set of the micro-basins wasproduced by anthropic interventions on the river hydrodynamics(especially the width reduction), which forced the riverbed to actwithin increasingly restricted and confined areas. In particular, thisplot of land is between two alignments: one is more external and

Fig. 9. Results of the freeboard analysis, where existing, referred to the hydraulic sections used by the Arno River Basin Authority for its hydraulic model: the situation for the dikesof Firenze (A) and Figline Valdarno (B). In addition to the different levels of criticality the constituent material of the dike is also highlighted.

S. Morelli et al. / Applied Geography 54 (2014) 35e5346

corresponds to the maximum extension of the riverbed (reached inthe period from 50 BC up to the 12th century); the other is closer tothe current riverbed and corresponds to the maximum lateralextension of the Arno during the Renaissance period (a few metersaway from the present course).

Table 4Comparison between the GPS-measured dike height ant the main values estimatedby the SIMI model (volumetric flow rates and floodwater levels) for each consideredsection in which floods with a return period of 500 years are expected to submergeall of the dikes that have a hydraulic freeboard of less than 1 m for a 200-year flood.

Sectionn.

Riverbanks GPS measure SIMI results

200-years flood 500-years flood

h [m a.s.l.] h [m a.s.l.] Q [m3/s] h [m a.s.l.] Q [m3/s]

829 Right 129.62 129.30 2572 129.79 2839827 Right/left 129.09/

129.01128.69 2550 129.20 2792

826 Right/left 128.60/128.78

128.54 2549 129.06 2794

825 Left 128.24 128.18 2549 128.74 2796818 Right 126.92 126.59 2481 127.37 2650814 Left 126.10 125.97 2519 126.91 2709813 Right/left 125.88/

126.68125.86 2519 126.83 2710

767 Left 119.77 119.60 2591 120.49 3266616 Left 53.86 52.97 3873 53.63 4608615 Left 53.83 53.19 3876 53.93 4569597 Left 50.98 50.27 3836 50.85 4158595 Right/left 50.96/50.54 50.52 3836 51.09 4204594 Right 50.78 50.58 3836 51.17 4252593 Right/left 50.65/50.67 50.59 3836 51.18 4210584 Right 49.40 49.19 3519 49.89 3568554 Left 45.07 44.82 3514 45.17 3575546 Left 43.90 43.70 3440 44.21 3450543 Left 43.16 42.99 3318 43.51 3610

Thus, the evolutionary history of this area is completelydifferent from that of the surrounding land and begins in a sub-aerial environment only when it has been recently reclaimed fromthe river. On the contrary, the current geomorphological configu-ration of the adjacent urban basin, from which the flow networkoriginates, is the result of a long interaction between the naturalevolution of the Arno, the construction of hydraulic works andurban sprawl. However, despite centuries of interventions inthroughout area, the urban design (and thus the flow patterns) insome districts still follows the planimetric evolution of the Arnoriverbed that has occurred since the first human settlements. This isespecially true in sections nearest to the existing riverbed (e.g., thearea once covered by the course Arno in the 12th century)(Fig. 10A). In both of the main basins, the more intricate the roadnetwork is (usually in the older quarters), the more the hydraulicpattern deriving from the proposed model is expressed. At any rate,the hierarchical distribution of the flow lines (Fig. 10B) convergecontemporaneously toward a main line that is subject to receivingmost of the water hypothetically circulating. In each area theybelong to the fifth order in the hierarchical classification proposedby Strahler (1952). This flow line distribution clearly proves that themorphology can be effective (excluding any anthropogenic inter-ference or rapid spreading of immeasurable volumes) in disposingof all of the water on this portion of the floodplain and preservingmost of the urban area. Furthermore, the shape of the right banksub-basin seems to be able to remove surface waters before theyreach the old town center, thus protecting an important part of thecultural heritage (Fig. 10).

According to the second model, in which the Arno river can stillreceive water from the external catchment area but in which thebridges openings are partially occluded, more critical evidenceemerges about the surface flow distribution (Fig. 11). In thissimulation the pattern of runoff lines over the left basin, identified

Fig. 10. Surface water flows in the urban area of Firenze during the full functionality of Arno dynamics (A). The main basin in the right side (painted in yellow) is physicallyseparated from the current riverbed by differently colored micro-basins that represent a natural barrier against mediumelow floods. B, the distribution of the flow lines in ahierarchical order according to the classification proposed by Strahler (1952). Both images are in relation to the Arno riverbed from 50 BC to the 12th century. (For interpretation ofthe references to color in this figure legend, the reader is referred to the web version of this article.)

S. Morelli et al. / Applied Geography 54 (2014) 35e53 47

and described above, tend naturally to direct its fluids toward theriverbed. Then, they begin to flow into the Arno river and whenthey encounter obstacles at the bridges are forced to transfer someof their liquid load into the right adjacent basin that is topo-graphically lower. So the excessive flows spread in the oppositefloodplain, creating a single, large inundated area. In particular, thewater tends to run directly from the closing section of the left basintoward the opposite bank because it is located in front of a partiallyoccluded bridge. A flow inversion occurs at the closing section of

the former basin in the right bank (upstream of the Ponte Vecchio);consequently all of the fluids move directly toward the old towncenter (Fig. 11A).

According to several historical accounts (Nencini, 1966), thehypothesis of simultaneous cloggings at all of the bridges isunlikely to be realized, while it is reasonable to presuppose a firstoccurrence at the Ponte Vecchio with a possible involvement ofthe other bridges occurring only later. However, the occurrenceof this single event before the others does not change the

Fig. 11. Surface water flows in the urban area of Firenze in the event that the Arno river can still receive water from the external catchment area but when the bridges openings arepartially occluded (A). B, the distribution of the flow lines in a hierarchical order according to the classification proposed by Strahler (1952). Both images are in relation to the Arnoriverbed from 50 BC to the 12th century.

S. Morelli et al. / Applied Geography 54 (2014) 35e5348

geometry of the final FDR model for which the resulting flowlines on the right bank remain the preferential paths of waterremoval in the urban area. So in this simulation spreading beginsin the surroundings of the overflow points, and only if the flowlevel increases due to high intensity or the long duration of anevent, can the most external runoff lines become progressivelyinvolved. This scenario is also confirmed by the hierarchicaldistribution of the flow lines (for example the higher order linemoves through sensitive areas: ring road boulevards, the oldtown center, railway/bus stations and underground parkingspaces) (Fig. 11B).

Finally, the third modeled scenario, under which the riverbed ischarged by great amounts of water so that it must laterally transfermost of its liquid load, simulates uncontrolled flooding within theurban area followed by a temporary floodwater stagnation (Fig. 12).In particular, the two banks remain separate and independent ba-sins over their respective sides. The left basin has a surface circu-lation system, which gathers water down from the hills andconveys it to the Arno. However, the river itself pushes its waters tojoin with them in a small flat district representing an ancient riverbend with the lowest elevations. The outflow of these combinedwaters thus stops immediately and a rapid flooding occurs with

S. Morelli et al. / Applied Geography 54 (2014) 35e53 49

water depths proportional to the duration of the event. The scale ofthe assumed event physically separates the left basin from the rightone, which reacts differently to the flooding. Because the latterdevelops on a more extended flat area, it can spread the floodwaters farther. The flow lines of this basin are approximately equalto those described in the previous simulation and any large stag-nation areas occur as it happens on the opposite basin (Fig. 12A).Assigning hierarchical classes to the flow lines practically coincideswith the sole difference in the terminal direction of the flow linewith the highest order (Fig. 12B). In this case it flows parallel to theArno, while in the case of clogged bridges it moves away from theriver because of the high flows diverted by the occlusion of the firstbridge downstream of the old town (Ponte alla Vittoria). In the left

Fig. 12. Surface water flows in the urban area of Firenze in the event that the Arno river is chcausing uncontrolled flooding within the urban area (A). The extension of the right basin isflow lines in a hierarchical order according to the classification proposed by Strahler (1952)interpretation of the references to color in this figure legend, the reader is referred to the

basin the flow lines are instead exclusively influenced by the above-mentioned water stagnation and tend to move along these floodedareas.

Discussion

The methodology proposed in this paper could provide a valu-able tool to help deal quickly with flood risk management, espe-cially for those public institutions with increasingly limitedeconomic resources, and may be applied in different ways to urbanplanning. For example, in a possible risk managing cycle the in-formation obtained could provide the basic tools to plan an opti-mized and wide sensor monitoring network (as also required for

arged by great amounts of water so that it must laterally transfer most of its liquid load,roughly equal to that of Fig. 9 except for the orange sub-basin. B, the distribution of the. Both images are in relation to the Arno riverbed from 50 BC to the 12th century. (Forweb version of this article.)

Fig. 13. Maximum extension of the inundated areas in Firenze during some historical floods compared with the modeled water spreading in the case of normal outflow in the Arnoriverbed (A) and in the event of occlusion of the bridge openings (B). In the latter case, some of the heights reached by the 1966 flood are shown, while in both cases the position ofthe urban sewer system is noted.

S. Morelli et al. / Applied Geography 54 (2014) 35e5350

the planning of contemporary Smart Cities; Su, Li, & Fu, 2011) toadequately distribute emergency facilities and train a resilientcommunity. Because a prevention program that includes thesetopics would be actuated in Firenze before the fiftieth anniversaryof the 1966 flood, the present work is fully relevant to the imme-diate needs of the city.

Moreover, some useful applications that were tested in differentcontexts of the study area are here suggested for all institutionsinterested in replicating this approach in similar situations world-wide. Startingwith a 30-year flood return time, the adjacent strip of

floodplain close to the Arno river has been classified into 3 types ofrisk: high, medium and low. This expeditious distinction is ach-ieved using a heuristic criterion based on the assignment of acertain level of danger centered around an expert judgment, whichextends the current approach to the management of hydraulicworks. Evaluating and assigning a score tomany different elements,such as the results related to overflow, the areal distribution of thesimulated flood and the spatial distribution of anthropogenic ob-jects that may be affected by the spreading water, makes possiblesuch a classification. In particular, the presence of sensitive areas in

S. Morelli et al. / Applied Geography 54 (2014) 35e53 51

terms of civil protection (e.g., farmlands, commercial or serviceactivities, residential complexes or working settlements), roadnetworks with different typologies of traffic density and the con-fluences of rivers with various hydrological regimes has been takeninto account and evaluated according to their different influenceson safety. However, the more the spatial arrangement (density anddistribution) of the above elements is known in detail the morerefined the resulting mappings can become.

Another important application, which was tested on the walldikes of Firenze, concerns the use of simulated flow heights toevaluate the design characteristics of a possible remediation mea-sure regarding the rise of the dike summit that can prevent over-flow under specific flow rates and to hypothesize, in case of aninsufficient freeboard, a definite size of a possible reinforcementwork (permanent or temporary). For example, in a small portion ofthe dike wall in front of the historic center of Firenze, the flow levelof a 100-year flood is a few centimeters above the maximum heightof the dike (up to a local maximum of 30 cm). Modest adjustmentworkswould be sufficient to prevent a damaging overflow, whereasslightly more extensive works would be required to strengthen thedike against a 200-year flood. All of these observations clearlysuggest that the interventions carried out along the urban reachafter the 1966 event (e.g., decimetric raising of the dike walls andlowering of the foundation slabs of the Ponte Vecchio and at thebridge further downstream) are not yet sufficient to prevent cata-strophic overflows with a similar flow magnitude (a recurrenceinterval of 200 years). According to our new research this is alsotrue for more recurrent events. Currently, several further inter-vention techniques could be used here for this purpose. Consid-ering the prestigious location in which the dikes are located,hydraulic protections that usually remain hidden and enter intooperation only when necessary could be successfully employed(e.g., the inflatable dike systems such as the Hydro-Air-Bag™,Benedetti, Ferri, & Baruffi, 2008).

All of the previously suggested analyses highlight how some ofthe most critical issues can be easily determined along the peri-fluvial strips and have practical implications for themanagement ofthe hydraulic policy when high-detail integrated data are available.In addition because the assessment of flood susceptibility isessential for urban areas built in large floodplains, this paper alsodemonstrates that once the main hydrodynamic conditions forwhich a river can create criticalities are established, a further GISanalysis concerning the evaluation of the flood dispersion overdensely populated flat areas can be successfully performed startingjust from the previously identified overflow-prone critical points.

The opportunity to significantly extend our knowledge aboutflood distribution has been tested by applying some specific pro-cedures to a densely inhabited city such as Firenze in which theareal spreads of the main historical floods are well known anddocumented over many centuries. In fact, the hypothesis that theurban fabric is able to react to flooding according to our modeling isconfirmed by much historical evidence. Whenever floodingoccurred in Firenze, the water flows followed the same path anddirection described by the smaller basins of themodel if the bridgesdid not obstruct the current. According to the chronicles, thisoccurred in 1589, 1740 and 1844 (Fig. 13A). During floods when thebridges blocked the current of the Arno (e.g., in 1333 with theclogging of Ponte Vecchio), the flood distribution, especially on theright-side plain, was comparable to the large basin spread identi-fied in the corresponding model (Fig. 13B).

Therefore, an almost perfect correspondence (estimated up to90% for the mapped 1333 episode) was found between real oc-currences in the past and the hypothesized flood movementsproposed by our model for any catastrophic situations in the nearfuture. This confirms that the proposed model is trustworthy. The

only slight difference with the oldest events (in terms of maximumareal expansion) can be mainly attributed to the recent populationgrowth and urban expansion in the more distant areas from thehistorical center (Fig. 13). Finally, we note that in our model, theflow direction is considered in the absence of movable objects, suchas cars, stands and street furniture. Because these objects are easilyremovable and are provided with a certain porosity if accumulatedby the flow strength, we believe they are not able to significantlyinfluence the flow path within the urban area. Rather they maycause unpredictable effects just in terms of time (the larger theaccumulation, the greater the time of discharge).

In conclusion, this study reveals that all of the analytic pro-cedures, given basic topographic data, can lead in a few steps to theaccumulation of a vast amount of useful information about floodsusceptibility in large anthropized areas (with particular effec-tiveness concerning ancient settlements). For this reason this lowcost working methodology could be successfully adopted by publicadministrations, which govern such types of districts. Moreover, acomplete automation could make these routine operations withwhich any users could directly identify the foremost problems,further saving time and money. Such cost savings could also bereinvested in targeted interventions in the territory to improve ourknowledge about increasingly specific issues.

Conclusion

Because of the importance of rapidly assessing the flood sus-ceptibility in extended fluvial urbanized areas (river stretches oftens of kilometers and more), we have developed an expeditiousand cost-effective sequence of procedures to preliminarily evaluatethe most critical overflowing dikes, and consequently, to obtain thedegree of dispersion of the floodwater in metropolitan areas withrespect to different hydrodynamic conditions of the main river.Such assessments, based exclusively on GIS analyses, can producein a short time a huge amount of territorial information that canhelp public administrations concentrate their economic effortsaccording to more specific hydraulic models or particular analysesin specific areas. This awareness has been gained by using themethodology along an Italian test site. In particular, the Arno riverwas chosen because it is characterized along the entire reachstudied by a highly variable fluvial regime in which long periods oflow water levels alternate with ruinous peak flows. Moreover, thecity of Firenze located approximately in its central portion wasselected for the areal analysis because it has existed on its river-banks starting from the ancient Roman period. Consequently, all ofthe most important floods have been properly documented.

The implementation of this methodology for this real case hasrevealed significant situations from which a general list of usefulapplications can be synthesized. In particular, the proposed analysisallows us to i) definitely position the overflow points on the dikesduring events with different return periods; ii) achieve an expe-ditious mapping which classifies the level of criticality of thefloodplain behind the overflowable dike; iii) estimate raising thedike up to the minimum height as a useful measure to prevent anew flooding episode during specific flow rates (especially in casesof reduced height differences); iv) assess the exact sizing of rein-forcement works on the dikes (permanent or temporary) in case ofinsufficient freeboard and possibly suggest how to minimizelandscape impacts; and v) promptly evaluate the areal dispersion offloods over densely populated urban areas (possibly starting fromthe overflowable points) according to the different hydrodynamicconditions of the main watercourse and the expected floodtypology.

All of these outcomes have already created interest among thetechnicians of the Firenze municipality because they wish to

S. Morelli et al. / Applied Geography 54 (2014) 35e5352

optimize their economic resources to preserve as much as possibletheir immense cultural heritage. Effectively, in our opinion, therapid delimitation of the most vulnerable areas (overflowable dikesand inundated streets), obtained by means of data that are almostall available, is a step in just that direction. For this reason our re-sults have been made available for the implementation of a newscientific-technical program (in collaboration with the responsibleauthorities) dedicated to the risk analysis to which the existingmuseum collections and the monumental structures of the Fran-ciscan complex of Santa Croce are subjected. In addition, as a resultof a certain similarity of the management problems of the casestudy with those of other watercourses in the world, we believethat the suggested procedures and their consequent applicationscould also be an inspiration for many other riverine communities.

This is especially true for those social contexts that, to achieve asmart and possibly sustainable growth with a very resilient com-munity, need to experiment with new policies and targeted prac-tices of land management that can successfully go beyond the localadministrative boundaries, as in the case study. This stems from thecurrent and increasingly widespread awareness that the life andthe well-being of a city depend on more or less intense relation-ships with other cities, with “detached” settlement, and in general,with all of the other parts of the same basin. These theoreticalconstructs fall within the concept of “metropolitan areas” towardwhich a new and significant international convergence has recentlyoccurred. Even the most current Italian policies seem to be orientedin this direction. As it is well known, attempts made in the past torelaunch large-scale planning (essentially based on topedown hi-erarchical mechanisms) produced unsatisfactory results in someEuropean countries and were gradually abandoned. For example,the metropolitan government of Rotterdam was established in1964 and abolished in 1985, the English metropolitan countieswere created in 1972 and suppressed in 1986, and the Barcelonametropolitan agency has been operating from 1974 to 1987.Nevertheless, the institutional and procedural innovations and themodern techniques for large-scale planning, mainly based on sus-tainable development that have spread in recent years allow us tohypothesize a new type of enlarged governance to which thepromising findings of this study can certainly contribute.

References

Al-Sabhan, W., Mulligan, M., & Blackburn, G. A. (2003). A real-time hydrologicalmodel for flood prediction using GIS and the WWW. Computers, Environmentand Urban Systems, 27, 9e32. http://dx.doi.org/10.1016/S0198-9715(01)00010-2.

Autorit�a di Bacino del fiume Arno. (2004). Piano Stralcio per l’Assetto Idrogeologico:relazione integrativa (in Italian). Published online on http://www.adbarno.it.Firenze. Accessed 27.08.13.

Autorit�a di Bacino del fiume Arno. (2008). Progetto Piano Stralcio: Bilancio Idrico (inItalian). Published online on http://www.adbarno.it. Firenze. Accessed 27.08.13.

Barredo, J. I. (2007). Major flood disasters in Europe: 1950e2005. Natural Hazards,42, 125e148.

Bartolini, C., & Pranzini, G. (1981). Plio-quaternary evolution of the Arno basindrainage. Zeitschrift fur Geomorphologie N.F., 40, 77e91. Suppl., BD.

Bartolini, C., & Pranzini, G. (1988). Evoluzione dell'idrografia nella Toscana centro-settentrionale. Bollettino del Museo di Storia Naturale della Lunigiana, 6e7,79e83.

Becchi, I., & Paris, E. (1989). Il corso dell'Arno e la sua evoluzione storica. Acqua Aria,6, 645e652.

Benedetti, C., Ferri, M., & Baruffi, F. (2008). Urban restoration of Valstagna riversiderelating to its present hydraulic risk problem. In B. Gumiero, M. Rinaldi, &B. Fokkens (Eds.), Proceeding of the 4th, ECRR international conference on riverrestoration, Venice, 16e21 June 2008 (pp. 339e348).

Berry, B. J. L., Goheen, P. G., & Goldstein, M. (1968). Metropolitan area definition: Areevaluation of concepts and statistical practice. Working Paper No. 28. Wash-ington D.C.: U.S., Bureau of the Census, U.S. Government Printing Office.

Billi, P., & Rinaldi, M. (1997). Human impact on sediment yield and channel dy-namics in the Arno River basin (central Italy). In D. E. Walling, & J. L. Probst(Eds.), Human impact on erosion and sedimentation, proceedings of rabat sym-posium (pp. 301e311). Wallingford: IAHS Publications n. 245.

Billi, P., Rinaldi, M., & Simon, A. (1997). Disturbance and adjustment of the ArnoRiver, Central Italy. I: historical perspective, the last 2000 years. In S. S. Y. Wang,

E. J. Langendoen, & F. D. Shields (Eds.), Management of landscapes disturbed bychannel incision, stabilization, rehabilitation, restoration (pp. 595e600). Oxford,MS: Center for the Computational Hydroscience and Engineering, University ofMississippi.

Boccaletti, M., Corti, G., Gasperini, P., Piccardi, L., Vannucci, G., & Clemente, S. (2001).Active tectonics and seismic zonation of the urban area of Florence, Italy. Pureand Applied Geophysics, 158, 2313e2332. http://dx.doi.org/10.1007/PL00001172.

Boonya-Aroonnet, S., Maksimovic, C., Prodanovic, D., & Djordjevic, S. (2007). Urbanpluvial flooding: development of GIS based pathway model for surface floodingand interface with surcharged sewer model. In Proceedings of the 6th NOVATECHinternational conference: Sustainable techniques and strategies in urban watermanagement (pp. 481e488), ISBN 2-9509337-7-7. Lyon, France.

Brauer, C. C., Teuling, A. J., Overeem, A., Van der Velde, Y., Hazenberg, P.,Warmerdam, P. M. M., et al. (2011). Soil buffer limits flash flood response toextraordinary rainfall in a Dutch lowland catchment. Hydrology and Earth Sys-tem Sciences Discussions, 8, 111e150. http://dx.doi.org/10.5194/hessd-8-111-2011.

Brown, M., Sharples, S., Harding, J., Parker, C. J., Bearman, N., Maguire, M., et al.(2013). Usability of geographic information: current challenges and future di-rections. Applied Ergonomics, 44(6), 855e865. http://dx.doi.org/10.1016/j.apergo.2012.10.013.

California Department of Water Resources. (2012). Urban levee design criteria.Published online on http://www.water.ca.gov/floodsafe/leveedesign Accessed30.03.13.

Campolo, M., Soldati, A., & Andreussi, P. (2003). Artificial neural network approachto flood forecasting in the river Arno. Hydrological Sciences Journal, 48(3),381e398. http://dx.doi.org/10.1623/hysj.48.3.381.45286.

Canuti, P., Cencetti, C., Rinaldi, M., & Tacconi, P. (1994). The fluvial dynamics of theArno River: 2. Historical evolution of the Arno Riverbed. Memorie della Societ�aGeologica Italiana, 48(3), 851e864.

Capecchi, F., Guazzone, G., & Pranzini, G. (1975). Il bacino lacustre di Firenze-Prato-Pistoia, geologia del sottosuolo e ricostruzione evolutiva. Bollettino della Societ�aGeologica Italiana, 94(4), 637e660.

Caporali, E., Rinaldi, M., & Casagli, N. (2005). The Arno River floods. Giornale diGeologia Applicata, 1, 177e192.

Casas, A., Benito, G., Thorndycraft, V. R., & Rico, M. (2006). The topographic datasource of digital terrain models as a key element in the accuracy of hydraulicflood modeling. Earth Surface Processes and Landforms, 31(4), 444e456. http://dx.doi.org/10.1002/esp.1278.

Cembrano, G., Quevedo, J., Salamero, M., Puig, V., Figueras, J., & Martí, J. (2004).Optimal control of urban drainage systems: a case study. Control EngineeringPractice, 12(1), 1e9. http://dx.doi.org/10.1016/S0967-0661(02)00280-0.

Cencetti, C., & Tacconi, P. (2005). The fluvial dynamics of the Arno River. Giornale diGeologia Applicata, 1, 193e202.

Clement, A. R. (2013). An application of geographic information system in mappingflood risk zones in a north central city in Nigeria. African Journal of Environ-mental Science and Technology, 7(6), 365e371. http://dx.doi.org/10.5897/AJEST12.182.

Coli, M., Agili, F., Pini, G., & Coli, N. (2004). Firenze: il suo impatto sull'evoluzionegeomorfica dell'area. Il Quaternario, 17, 195e211.

Coli, M., & Rubellini, P. (2007). Note di geologia Fiorentina. Firenze: S.E.L.C.A.Cortecci, G., Boschetti, T., Dinelli, E., Cidu, R., Podda, F., & Doveri, M. (2009).

Geochemistry of trace elements in surface waters of the Arno River Basin,northern Tuscany, Italy. Applied Geochemistry, 24, 1005e1022. http://dx.doi.org/10.1016/j.apgeochem.2009.03.002.

Coulthard, T. J., & Frostick, L. E. (2010). The Hull Floods of 2007: implications for thegovernance and management of urban drainage systems. Journal of Flood RiskManagement, 3, 223e231. http://dx.doi.org/10.1111/j.1753-318X.2010.01072.x.

Da Deppo, L. (2006). Difese arginali dei territori di pianura. In G. Frega (Ed.), Tec-niche per la Difesa dall'Inquinamento, Proceedings of 26� refresher course, 22e25June 2006 (pp. 1e46).

Dapporto, S., Rinaldi, M., Casagli, N., & Vannocci, P. (2003). Mechanisms of riverbankfailure along the Arno River, Central Italy. Earth Surface Processes and Landforms,28, 1303e1323. http://dx.doi.org/10.1002/esp.550.

Daungthima, W., & Kazunori, H. (2013). Assessing the flood impacts and the culturalproperties vulnerabilities in Ayutthaya, Thailand. Procedia Environmental Sci-ences, 17, 739e748. http://dx.doi.org/10.1016/j.proenv.2013.02.091.

Doan, J. H. (2000). Hydrologic model of the Buffalo Bayou using GIS. In D. Mariment,& D. Djokic (Eds.), Hydrologic and hydraulic modeling support (pp. 127e143).Redlands, CA, USA: ESRI Press.

Doswell, C. A., III, Brooks, H. E., & Maddox, R. A. (1996). Flash flood forecasting: aningredients-based methodology. Weather and Forecasting, 11, 560e581. http://dx.doi.org/10.1175/1520-0434(1996)011<0560:FFFAIB>2.0.CO;2.

Drd�acký, M. (2010). Impact of floods on heritage structures. Journal of Performanceof Constructed Facilities, 24, 430e431. http://dx.doi.org/10.1061/(ASCE)CF.1943-5509.0000152. Special Issue: Flood Impact to Heritage Structures.

Giandotti, M. (1934). Previsione delle piene e delle magre dei corsi d’acqua. Min-istero LL.PP. Memorie e studi idrografici, 8(2). Servizio Idrografico Italiano, Rome(in Italian).

Gumiero, B., Maiolini, B., Rinaldi, M., Surian, N., Boz, B., & Moroni, F. (2009). Italianrivers. In K. Tockner, C. T. Robinson, & U. Uehlinger (Eds.), Rivers of Europe (pp.467e495). Amsterdam: Academic Press.

Guzzetti, F., Stark, C. P., & Salvati, P. (2005). Evaluation of flood and landslide risk tothe population of Italy. Environmental Management, 36(1), 15e36. http://dx.doi.org/10.1007/s00267-003-0257-1.

S. Morelli et al. / Applied Geography 54 (2014) 35e53 53

Haan, C. T., Barfield, B. J., & Hayes, J. C. (1994). Design hydrology and sedimentology forsmall catchments. San Diego, CA, USA: Academic Press.

Hammond, M. J., Chen, A. S., Djordjevi�c, S., Butler, D., & Mark, O. (2013). Urban floodimpact assessment: a state-of-the-art review. Urban Water Journal, 1e16. http://dx.doi.org/10.1080/1573062X.2013.857421.

Hardie, C. (1965). The origin and plain of Roman florence. Journal of Roman Studies,55, 122e140.

Henonin, J., Russo, B., Mark, O., & Gourbesville, P. (2013). Real-time urban floodforecasting and modelling e a state of the art. Journal of Hydroinformatics, 15(3),717e736. http://dx.doi.org/10.2166/hydro.2013.132.

Herle, I., Herbstov�a, V., Kupka, M., & Kolymbas, D. (2010). Geotechnical problems ofcultural heritage due to floods. Journal of Performance of Constructed Facilities,24, 446e451. http://dx.doi.org/10.1061/(ASCE)CF.1943-5509.0000058. SpecialIssue: Flood Impact to Heritage Structures.

Holický, M., & Sýkora, M. (2010). Assessment of flooding risk to cultural heritage inhistoric sites. Journal of Performance of Constructed Facilities, 24, 432e438.http://dx.doi.org/10.1061/(ASCE)CF.1943-5509.0000053. Special Issue: FloodImpact to Heritage Structures.

Huffman, R. G., & Eiker, E. (1991). Freeboard design for urban levees and floodwalls.In U.S. Army Corps of Engineering Hydrologic Engineering Center (Ed.), Pro-ceedings of a hydrology and hydraulics workshop on Riverine Levee Freeboard,Monticello, Minnesota, USA (pp. 5e11).

Jenson, S. K., & Domingue, J. O. (1988). Extracting topographic structure from digitalelevation data for geographic information system analysis. PhotogrammetricEngineering and Remote Sensing, 54(11), 1593e1600.

Kale, V. S. (2003). Geomorphic effects of monsoon floods on Indian rivers. NaturalHazards, 28, 65e84. http://dx.doi.org/10.1023/A:1021121815395.

Krellenberg, K., Müller, A., Schwarz, A., H€ofer, R., & Welz, J. (2013). Flood and heathazards in the Metropolitan Region of Santiago de Chile and the socio-economics of exposure. Applied Geography, 38, 86e95. http://dx.doi.org/10.1016/j.apgeog.2012.11.017.

Kulkarni, A. T., Mohanty, J., Eldho, T. I., Rao, E. P., & Mohan, B. K. (2014). A web GISbased integrated flood assessment modeling tool for coastal urban watersheds.Computers & Geosciences, 64, 7e14. http://dx.doi.org/10.1016/j.cageo.2013.11.002.

Lanza, S. G. (2003). Flood hazard threat on cultural heritage in the town of Genoa(Italy). Journal of Cultural Heritage, 4(3), 159e167. http://dx.doi.org/10.1016/S1296-2074(03)00042-6.

Marco, J. B., & Cayuela, A. (1994). Urban flooding: the flood-planned city concept. InG. Rossi, N. Harmancioglu, & V. Yevjecich (Eds.), Coping with floods (pp.705e721). Dordrecht: Kluwer. http://dx.doi.org/10.1007/978-94-011-1098-3_43.

Merz, B., Hall, J. W., Disse, M., & Schumann, A. (2010). Fluvial flood risk managementin a changing world. Natural Hazards and Earth System Sciences, 10(3), 509e527.http://dx.doi.org/10.5194/nhess-10-509-2010.

Messeri, S., & Pintus, S. (2006). 4 Novembre. 1966: l’alluvione a Firenze [4th November1966: The flood in Florence]. Pisa: Ibiskos editrice.

Mignot, E., & Paquier, A. (2003). Impact-flood propagation case study e the modelcity flooding experiment. In Proceedings of the third IMPACT workshop (EU-funded research project on investigation of extreme flood processes and uncer-tainty), Louvain La Neuve, Belgium, 5e7 November 2003, HR Wallingford.

Mignot, E., Paquier, A., & Haider, S. (2006). Modeling floods in a dense urban areausing 2D shallow water equations. Journal of Hydrology, 327(1e2), 186e199.http://dx.doi.org/10.1016/j.jhydrol.2005.11.026.

Morelli, S., Segoni, S., Manzo, G., Ermini, L., & Catani, F. (2012). Urban planning, floodrisk and public policy: the case of the Arno River, Firenze, Italy. Applied Geog-raphy, 34, 205e218. http://dx.doi.org/10.1016/j.apgeog.2011.10.020.

Natoni, E. (1944). Le piene dell’Arno e i provvedimenti di difesa (in Italian). Firenze:Felice Le Monnier editore.

Nencini, F. (1966). Firenze i giorni del diluvio (in Italian). Firenze: Sansoni editore.Patro, S., Chatterjee, C., Singh, R., & Raghuwanshi, N. S. (2009). Hydrodynamic

modelling of a large flood-prone river system in India with limited data. Hy-drological Processes, 23, 2774e2791. http://dx.doi.org/10.1002/hyp.7375.

Peddle, D. R., Rabe, N. J., Soenen, S. A., & Johnson, D. L. (2005). Derivation of top-oclimatic indices for alpine snowpack analysis in Alberta rocky mountain wa-tersheds. In ERDC-CRREL (Ed.), International proceedings Canada-US easternsnow conference, Waterloo ON, Canada, June 7e11, 2005 (pp. 173e190).

Penning-Rowsell, E., Floyd, P., Ramsbottom, D., & Surendran, S. (2005). Estimatinginjury and loss of life in floods: a deterministic framework. Natural Hazards,36(1e2), 43e64. http://dx.doi.org/10.1007/s11069-004-4538-7.

Rinaldi, M., Simon, A., & Billi, P. (1997). Disturbance and adjustment of the ArnoRiver, Central Italy: II. Quantitative analysis of the last 150 years. In S. S. Y. Wang,E. J. Langendoen, & F. D. Shields (Eds.), Management of landscapes disturbed bychannel incision, stabilization, rehabilitation, restoration (pp. 601e606). Oxford,

MS: Center for the Computational Hydroscience and Engineering, University ofMississippi.

Rinaldi, M., & Simon, A. (1998). Bed level adjustment in the Arno River, central Italy.Geomorphology, 22(1), 57e71. http://dx.doi.org/10.1016/S0169-555X(97)00054-8.

Rinaldi, M. (2003). Recent channel adjustments in alluvial rivers of Tuscany, CentralItaly. Earth Surface Processes and Landforms, 28, 587e608. http://dx.doi.org/10.1002/esp.464.

Risley, J. C., Hess, G. W., & Fisher, B. J. (2006). An assessment of flow data from Kla-math river sites between Link river Dam and Keno Dam, south-central Oregon: U.S.Geological Survey Scientific Investigations report 2006-5212. Reston, Virginia: U.S.Geological Survey.

Rumman, N., Lin, G., & Li, J. (2005). Investigation of gis-based surface hydrologicalmodelling for identifying infiltration zones in an urban watershed. Environ-mental Informatics Archives, 3, 315e322.

Salvati, P., Bianchi, C., Rossi, M., & Guzzetti, F. (2010). Societal landslide and floodrisk in Italy. Natural Hazards and Earth System Sciences, 10, 465e483. http://dx.doi.org/10.5194/nhess-10-465-2010.

Santato, S., Bender, S., & Schaller, M. (2013). The European floods directive and op-portunities offered by land use planning. CSC Report 12. Germany: Climate Ser-vice Center.

Segoni, S., Morelli, S., Kukavicic, M., Camorani, A., Catani, F., & Ermini, L. (2008).Applicazione di tecniche di mappatura ad alta risoluzione nel fiume Arno(Firenze). Giornale di Geologia Applicata, 9(2), 215e225.

Schmitt, T. G., Thomas, M., & Ettrich, N. (2004). Analysis and modelling of floodingin urban drainage systems. Journal of Hydrology, 299(3e4), 300e311. http://dx.doi.org/10.1016/j.jhydrol.2004.08.012.

Strahler, A. N. (1952). Hypsometric (area-altitude) analysis of erosional topography.Geological Society of America Bulletin, 63, 1117e1142.

Su, K., Li, J., & Fu, H. (2011). Smart city and the applications. In Proceeding of in-ternational conference on electronics, communications and control (ICECC),Wuhan, China, Sept. 2011 (pp. 1028e1031). http://dx.doi.org/10.1109/ICECC.2011.6066743.

Surian, N., & Rinaldi, M. (2003). Morphological response to river engineering andmanagement in alluvial channels in Italy. Geomorphology, 50, 307e326. http://dx.doi.org/10.1016/S0169-555X(02)00219-2.

Taubenb€ock, H., Wurm, M., Netzband, M., Zwenzner, H., Roth, A., Rahman, A., et al.(2011). Flood risks in urbanized areas e multi-sensoral approaches usingremotely sensed data for risk assessment. Natural Hazards and Earth SystemSciences, 11, 431e444. http://dx.doi.org/10.5194/nhess-11-431-2011.

Tacconi, P. (1994). The fluvial dynamics of the Arno River: 1. Methodology and studyproject. Memorie della Societ�a Geologica Italiana, 48(3), 755e764.

Tarboton, D. G. (1997). A new method for the determination of flow directions andupslope areas in grid digital elevation models. Water Resources Research, 33(2),309e319. http://dx.doi.org/10.1029/96WR03137.

Tarboton, D. G., & Ames, D. P. (2001). Advances in the mapping of flow networksfrom digital elevation data. In Proceedings of the ASCE, world water and envi-ronmental resources congress, Orlando, Florida, May 20e24.

Tarboton, D. G., Bras, R. L., & Rodriguez-Iturbe, I. (1991). On the extraction of channelnetworks from digital elevation data. Hydrologic Processes, 5(1), 81e100. http://dx.doi.org/10.1002/hyp.3360050107.

Tran, P., Shaw, R., Chantry, G., & Norton, J. (2009). GIS and local knowledge indisaster management: a case study of flood risk mapping in Viet Nam. Disasters,33, 152e169. http://dx.doi.org/10.1111/j.0361-3666.2008.01067.x.

Webster, T. L., Forbes, D. L., Dickie, S., & Shreenan, R. (2004). Using topographic lidarto map flood risk from storm-surge events for Charlottetown, Prince EdwardIsland, Canada. Canadian Journal of Remote Sensing, 30(1), 64e76. http://dx.doi.org/10.5589/m03-053.

Wehr, A., & Lohr, U. (1999). Airborne laser scanning e an introduction and overview.Journal of Photogrammetry and Remote Sensing, 54, 68e82. PII: S0924-27169900011-8.

Winston, W. E., & Criss, R. E. (2002). Geochemical variations during flash flooding,Meramec River basin, May 2000. Journal of Hydrology, 265(1), 149e163. http://dx.doi.org/10.1016/S0022-1694(02)00105-1.

Wu, S., Li, J., & Huang, G. H. (2008). A study on DEM-derived primary topographicattributes for hydrologic applications: sensitivity to elevation data resolution.Applied Geography, 28, 210e223. http://dx.doi.org/10.1016/j.apgeog.2008.02.006.

Yen, B. C. (1995). Urban flood hazard and its mitigation. In F. Y. Cheng, & M. S. Sheu(Eds.), Urban disaster mitigation, the role of engineering and technology (pp.245e260). Oxford, UK: Elsevier Science.

Zerger, A., & Wealands, S. (2004). Beyond modelling: linking models with GIS forflood risk management. Natural Hazards, 33, 191e208. http://dx.doi.org/10.1023/B:NHAZ.0000037040.72866.92.