A flash-flood event hit the northeastern part of
Mallorca on 9 October 2018, causing 13 casualties. Mallorca is
prone to catastrophic flash floods acting on a scenario of deep landscape
transformation caused by Mediterranean tourist resorts. As global change may
exacerbate devastating flash floods, analyses of catastrophic events are
crucial to support effective prevention and mitigation measures.
Field-based remote-sensing and modelling techniques were used in this study
to evaluate rainfall–runoff processes at the catchment scale linked to
hydrological modelling. Continuous streamflow monitoring data revealed a
peak discharge of 442 m
Flash floods are related to high-intensity precipitation, mainly convective
in origin and with restricted spatiotemporal occurrence. For this reason,
they usually strike basins
Characterising the response of a catchment during an extreme flash-flood
event is important because it clarifies flood severity and the activating
hydrological processes and their dependency on natural and anthropogenic
catchment properties (Borga et al., 2007). Numerous
studies have tried to determine these driving factors
(Braud et al., 2014) for
which geological heterogeneity associated with the presence of karst
features is crucial in Mediterranean catchments
(Vannier et al., 2016;
Wainwright and Thornes, 2004). Likewise, flash floods are closely related to
land use: the devastation of plant cover in the Mediterranean may increase
the risk of flooding because bare soil leads to larger runoff coefficients
(Wainwright and Thornes, 2004). However, the limited
spatial and temporal scales of flash floods make these events particularly
difficult to monitor and document. In the case of rainfall monitoring, the
spatial scales of the events are in general much smaller than the sampling
potential offered by apparently dense rain networks
(Borga et al., 2008; Amponsah et al., 2016). In the case of streamflow
monitoring, there is a lack of flash-flood discharge (
Earlier flash-flood forecasting systems were based on the flash flood
guidance (Georgakakos, 1986), which calculated the a priori
amount of rainfall needed to trigger specific
In the Mediterranean region, the planning and management of flood hazards are hydrosociologically crucial (Gaume et al., 2016), especially in the current global change context. It is clear that the understanding of flash floods requires an integrated scientific approach (Marchi et al., 2010), in which technological advances create opportunities to investigate simultaneously in the areas of Earth and social sciences (Wohl et al., 2019). Firstly, geomorphometric techniques applied to topographic surveys can be valuable planning and decision-making tools for producing flood hazard maps that represent flood-prone areas (Kalantari et al., 2017). Geomorphometry and digital terrain analysis by means of a topography-based connectivity index could also be used to simulate preferential flow paths for high-magnitude events to determine the main erosion and deposition areas, as a tool for better and faster responses to catastrophic flooding events. In addition, post-storm assessments to capture the landform signature of the event are needed. The use of high-resolution field measurements here is critical to understanding the effects of storms on fluvial dynamics (Westoby et al., 2012) and providing input data sets for numerical modelling. These data can be difficult to obtain, as traditional post-storm survey techniques are expensive and time consuming (Duo et al., 2018). In recent years, unmanned aerial vehicles (UAVs) have been used to improve traditional, expensive, or time-consuming mapping approaches in catchment science research. UAVs can be deployed rapidly and achieve accurate high-resolution topographic data for monitoring geomorphic changes (Estrany et al., 2019; Langhammer and Vacková, 2018).
Secondly, the concept of vulnerability relates to the predisposition of certain stakes to damage or malfunction, implying that a multitude of direct and indirect factors, often interacting in a dynamic and complex way, should be integrated in its assessment. Thus, vulnerability particularly relates to the damage to exposed stakes (Defossez and Leone, 2017). Estimations of the elements at risk in flash floods and the damage-driving factors are insufficiently understood in the Mediterranean region, even though assessment of damage processes caused by flash floods have been assessed in several European and national projects and publications (see Llasat et al., 2013; Gaume et al., 2016). Observed spatial distributions of costs and fatalities are the result of complex interplay between different explanatory factors. Useful information may reveal the economic and social impact of floods on our societies, but its interpretation is questionable (Gaume et al., 2016). In addition, the largest public world databases on disaster events that contain flood events do not include all the catastrophic events in the Mediterranean region (Llasat et al., 2013), which underlines the need for further research. Since 2012, the Copernicus Emergency Management Service (Copernicus EMS) has evaluated the intensity and scope of the damage caused by natural disasters, human-made emergency situations and humanitarian crises throughout the world (Copernicus Emergency Management Service, 2019). To understand the damage processes of flash floods better, comprehensive damage assessment that links hydrological process dynamics and intensities to damage and loss is needed (Laudan et al., 2017). In addition, a comparison of “ground-based” assessment and “remote-based” Copernicus EMS may shed light on the accuracy of this rapid and helpful tool for assessing most catastrophic flash floods.
This paper aims at improving understanding of the hydrological and
socioeconomic processes of the devastating flash floods that frequently
affect Mallorca (Estrany
and Grimalt, 2014; Llasat et al., 2013), a paradigmatic Mediterranean
flood-prone region under intense human occupation and geologically shaped by
karstic features. The study focuses on the catastrophic flash-flood event
that hit the northeastern part of the island on 9 October 2018,
causing 13 casualties. This was the worst local natural disaster in decades
and caused a marked interest among the scientific community, producing early
studies of the event. Lorenzo-Lacruz et al. (2019) reconstructed this same flash-flood event through the application of
hydrological and hydraulic simulations, with a focus on the meteorological
input. The specific objectives of this study are (1) to explain the runoff
response and so clarify the dependency of flood severity on catchment
properties and human influence, by using the flash-flood
To obtain better understanding of the flash-flood and damage processes, a comprehensive and combined analysis of the meteorological synoptic situation, precipitation and discharge was developed through the analysis of the rainfall–runoff processes at a small spatial scale during this extreme event. Likewise, this analysis was linked to hydrological modelling to check the internal consistency of the information gathered by instruments and to flood damage and losses at the village of Sant Llorenç des Cardassar. Finally, HR-DTMs were generated by lidar 2014 data from the Spanish National Geographic Institute and by imagery captured through a low-cost UAV just 6 d after the catastrophe, to calculate a sediment connectivity index (IC) and measure geomorphic changes (Fig. 1).
Workflow of the experimental design.
Mallorca is a Mediterranean flood-prone region, historically affected by flash floods. Since the late Middle Ages, devastating flash floods have been systematically documented, particularly in Palma, the capital of the island. In this town, a catastrophic event caused approximately 5000 deaths in 1403 (20 % of its population), showing that floods are the main natural hazard in this type of environment (Petrus et al., 2018). In the rest of the island, the historical distribution patterns of human settlements were related to fluvial systems but avoided the occupation of floodplains until the increase of urban areas in the 19th century during the Industrial Revolution. However, in the second part of the 20th century, this urban expansion became exponential, with many more urban and tourist settlements (Pons Esteva, 2003), often in flood hazard areas.
In such a high-energy environment, Mallorca's eastern (“Llevant” in Catalan)
county contains a dramatic combination of physical and human factors that
have created a flood-prone environment with a very strong coupling of
climate and geomorphology to constant urban expansion since 1850, which has
led to more than 10 catastrophic flash-flood events during this period
(Estrany and Grimalt, 2014).
This county consists of two main relief units (Álvaro et
al., 1991): (1) the Llevant ranges occupy the headwaters with altitude
ranging from 300 to 500 m a.s.l. They consist of a series of alpine
mountains and hills mainly of Jurassic limestones and dolomites and
Cretaceous marls; and (2) the Marinas, a reefal Upper Miocene tabular
platform composed of calcarenites, calcisiltites and terra-rossa post-reef
sediments affected by significant karstic processes on the eastern slopes of
the Llevant ranges. A morphometric analysis of the catchments
by Estrany and Grimalt (2014)
showed differences that depended on both the width of the platform (i.e.
the distance between the coast and the base of the Llevant ranges through
the Miocene platform) and the hypsometry and geological settings at the
headwaters. The torrentiality and clinometric variables were the ones most
closely related to the geological settings. Therefore, the catchments with
the highest values of torrentiality (i.e.
This study focuses on the Ca n'Amer River catchment (78 km
Main characteristics of affected basins during the 9 October
2018 flash flood.
The 9 October storm affected the two northernmost catchments of Llevant county, i.e. the Ca n'Amer and Canyamel rivers (Fig. 2) with nine and four casualties, respectively, and significant damage. The synoptic situation was characteristic of flash-flood events in the Western Mediterranean (Fig. 3a). A cut-off low at midlevel was located in the eastern part of the Iberian Peninsula and shallow low-level pressure affected the same region, driving warm and wet air from the Mediterranean Sea to the Balearic Islands and the eastern part of the Iberian Peninsula. This occurred in early October, when the sea surface temperature is close to its annual maximum in the Western Mediterranean, providing high quantities of moisture. Moreover, the cut-off low showed a typical divergence at midlevel on its eastern flank, affecting the Balearic Islands and favouring the development of deep convection. Convection started on the sea between the Balearics and the Iberian Peninsula (Fig. 3b and c) and, due to SW winds at mid-tropospheric levels, the convective cells started to move towards the Balearics, where heavy rainfall triggered the flash-flood event (Fig. 3d and e).
In order to assess the rainfall–runoff processes during the flash-flood
event in the Begura de Salma River, the continuous 10 min precipitation
record of the event was obtained from radar (see location in Fig. 2b)
images that were initially calibrated through rainfall data downloaded at
The radar, located in the west of Mallorca, obtains reflectivity data (in
dBZ) at
Two pairs of coefficients were tested: (i) the pair
Calculation with the two pairs of coefficients results in high amounts of
precipitation (100 mm in the first case and 150 mm in the second case) but
an amount much lower than the maximum rainfall observation, which is higher
than 250 mm: this shows the complexity of calculating the precipitation of
this event accurately by radar data. In the scientific literature, many
methodologies to enhance precipitation calculation using radar data can be
found. While some methods focus on the correction of the reflectivity data;
others are based on the calibration of the estimation of precipitation using
reflectivity data, as these are the two main sources of errors
(Harrison et al., 2000). Mountains may partially or totally
block the electromagnetic radar signal and affect radar reflectivity and
precipitation estimations (Germann and Joss, 2004). The
study area is mountainous but with low maximum altitudes (
With the set of coefficients
A hydrometric gauging station is located where the Begura de Salma River
enters the village of Sant Llorenç des Cardassar, 50 m upstream from the
Ma-3323 road bridge (Fig. 4a). The station was built by the water authority
in the 1970s. After years of abandonment, in 2015, the MEDhyCON research
group installed within the gauge house a Hobo water level U20L-04, which
measures the water stage by 1 min readings, accumulating 15 min
average values. The station is located at the very beginning of the concrete
channel that takes the river through the village. Closing a drainage basin
of 23 km
The transformation of water stage (hereinafter WS; m) to
For the generation of the 2-D hydraulic model, a HR-DTM derived from a 1 m
lidar-based DTM dating from 2014 was used;
As boundary conditions, an input hydrograph was arranged at the most
upstream part of the section studied, whilst the WS was subsequently
arranged in accordance with the slope of the channel bed at the most
downstream part. Hydraulic simulations were always started under no-flow
conditions. Taking into account the hydraulic function characteristics for
obtaining the SDRC, the calculation method used in these simulations was the
diffusion wave. To design an accurate SDRC for the hydrometric station, the
hydraulic effects of the concrete channel and bridges were assessed, running
iteratively the model by making the representative
Accordingly, a first hydrograph was designed step by step by using
The SDRC was divided into two sections. The first was related to a scenario
where bridges do not affect the hydraulic function at the gauging section,
despite flow being bankfull or even overbanking the channel. The second
section was a scenario under the influence of bridges, i.e. the flow below
them and also over their decks. It is then clear that the design, dimensions
and proximity to the hydrometric station of these bridges affected directly
flow behaviour at the hydrometric station due to obstruction (see pictures
in Fig. 4). The obstruction of bridge 1 occurred when the upstream section
of the concrete channel reached the bankfull level, such that dragging and
floating elements began to collide against the back of the bridge,
triggering the obstruction process. Consequently, the hydraulic model had to
integrate the obstructions observed at the downstream bridges, at least the
two nearest ones (Fig. 4a). These post-event pictures and maximum water
stages observed in situ 10 h after the event were useful for calculating the
obstruction percentages of these two bridges: 85 % at bridge 1, 40 % at
bridge 2 and with no obstruction at the other ones. The overbank flow
coefficients were 2.2 for bridges 1, 4 and 5, and 2.1 for bridges 2 and 3.
The SDRC obtained allowed calculation of this phenomenon, which is activated
with
The SDRC was finally designed (see Fig. 5), showing clearly the two
differentiated sections with a gap between them in WS of approximately 1.5 m because
1.4 m is the edge of road bridge 1 located just 50 m downstream from the
hydrometric station. The SDRC is fitted to a power equation for both
sections, obtaining high values of significance (
Stage–discharge rating curve performed by means of two-dimensional hydraulic modelling with two differentiated sections in accord with the influence of bridge 1 (see Fig. 3a) and its potential obstruction.
A semi-distributed hydrological model was used to reproduce the hydrological response of the catchment during the flash-flood event: the routing system (RS) model (García-Hernández et al., 2007; Jordan, 2007). The SOil CONTribution (SOCONT) was the rainfall–runoff model used. This comprised an appraisal of hydrological processes such as snowmelt as well as surface and subsurface infiltration-induced flow and groundwater flow due to percolation (Jordan, 2007; Schaefli et al., 2005). The catchments were divided into elevation bands to incorporate the influence of temperature evolution with altitude and orographic effects within mountainous catchments. In this model, the subcatchments were divided into 100 m elevation bands. The SOCONT model was applied to each elevation band, which was added at the outlet of each subcatchment. The input data of the SOCONT were the temperature obtained from meteorological stations and the precipitation derived from radar measurements (see previous section).
The SOCONT model is shown in Fig. S2. First, the temperature was
interpolated for each elevation band, based on inverse distance weighting
using the Shepard method (Shepard, 1968). Resampled 1 km
resolution radar data (see Sect. 2.3) were used in the model to obtain
precipitation for each elevation band by including all 1 km resolution
points falling within each elevation band. The soil-infiltration model was
based on modified GR3 equations (Schaefli et al., 2005).
Infiltration and evaporation were determined by the soil saturation; i.e.
infiltration is higher for lower soil saturation, whereas evapotranspiration
is higher for high soil saturation. Surface runoff was computed with the
Storm Water Management Model (SWMM). Soil infiltration was modified to simulate karstic hydrological
dynamics, as shown in Fig. S3. Precipitation infiltrated the soil, as in
Schaefli et al. (2005), as a function of soil saturation (Fig. S3a).
The resulting outflow from the reservoir (
Rapid mapping is a mature Earth observation (EO) service with many years of
user-oriented development since the international charter “Space and Major
Disaster” was established in 1999. EO satellite-data-derived disaster
mapping during emergencies is provided to civil protection and humanitarian
user communities at national, continental and worldwide scales. Given the
great helpfulness of the Copernicus EMS reports attained by rapid mapping
techniques, the damage assessment of the event focused on the comparison
between two information sources. The first one, a “ground-based” report, was
the damage analysis carried out by the Directorate General of Emergencies of
the Balearic Islands Government (Pol, 2019a). This “ground-based”
report provided a detailed description of the resources mobilised in the
emergency phase and also a damage inventory. The second information source
was the “remote-based” damage assessment by the Copernicus EMS
(
The damage assessment in the two sources was compared by means of a
cartographic overlay with GIS tools. To provide greater accuracy, detailed
territorial information and flow direction were also incorporated. First,
the type of buildings and land use at the urban plot scale from the General
Directorate for the Cadastre (
As well as the hydrogeomorphological monitoring tasks, the 15 October
2018 MEDhyCON research group collaborated in the emergency operation to
search for a missing person during the flash flood. Firstly, and taking into
account the emergency situation, the index of (water and sediment)
connectivity at the catchment scale was applied to find the areas with the
greatest sediment deposition potential, which were where victims could have
been buried by the flash flood. The sediment IC
proposed by Borselli et al. (2008) and modified by Cavalli et
al. (2013) determined the preferential flow paths by
exploring the water and sediment transference patterns in different
landscape compartments of the entire Ca n'Amer River. Thus, the IC is a
dynamic property of the catchment that indicates the probability of a
particle at a certain location reaching a defined target area, which in this
study was established at the catchment outlet
(Trevisani and Cavalli, 2016). This
morphometric index was mainly derived from a HR-DTM, in this case, a 1 m
lidar-based DTM dating from 2014;
HR-DTMs facilitate the improvement of sediment connectivity as a powerful tool to determine preferential flow paths and those areas with the greatest potential sediment deposition. The evaluation of the flash-flood landform signature by UAVs is the second part of creating a tool for a rapid response of post-catastrophe search-and-rescue tasks by applying hydrogeomorphological precision techniques. The estimation of overbank sedimentation allowed the calibration of predicted large sedimentation by IC mapping and its reliability in detecting sites where victims might be buried by flood sediment.
The latest technological advances in remote data acquisition (i.e. UAVs) and
topographic modelling (i.e. structure from motion – SfM; multi-view
stereo – MVS) have led to a huge advance in Earth and environmental
sciences. Following the incorporation of MEDhyCON to the emergency
operations, several UAVs flew all along the Ca n'Amer River, from the
headwaters (Begura de Salma River) to its outlet into the Mediterranean Sea
at the village of S'Illot (Fig. 2c). This fieldwork involved the
establishment and survey of more than 250 ground control points (GCPs),
needed for an appropriate georeferencing of the aerial photographs taken by
the drone. Therefore, on 15 October 2018, just 6 d after the
flash flood, evidence of erosion was recorded by aerial photographs taken
with a small unmanned aerial vehicle (UAV DJI Phantom 4 Pro,
Imagery acquired during the aerial campaign enabled (1) the creation of
mosaics of aerial georeferenced images and (2) the generation of
high-resolution digital terrain models. These were produced by Agisoft Metashape Pro®
v1.5.3 using automated digital photogrammetry techniques. This software obtains
high-quality results easily from algorithms known as SfM. Further details
on the implemented algorithms can be found in Lowe (2004) and Westoby et al. (2012). For
the proper acquisition of the imagery, flight altitude was set at 70 m,
ensuring ground resolution close to 0.02 m px
Once all the drone images were georeferenced and properly mosaicked,
topographic modelling (i.e. SfM) generated the
post-flash-flood very-high-resolution DTM (i.e. 5 cm pixel size). The
comparison of that DTM to that of the catchment prior to the catastrophic
event (lidar-based DTM dating from 2014) allowed the quantification and
assessment of the actual magnitude (competence) of the event in terms of the
volume of sediments eroded and/or deposited and the alteration of the
fluvial morphology. It is worth noting that no geomorphic changes were
observed between 2014 and October 2018, by photointerpretation of aerial
imagery (PNOA, 2015) and the continuous measurement of water stages since
January 2015, with no overbanked flood events. Consequently, geomorphic
changes were estimated in a floodplain downstream from Sant Llorenç des
Cardassar to evaluate the amount of overbank sedimentation in the area of
the rescue where IC suggested the search. Measurements were taken with a
procedure, similar to DEM of difference (DoD),
that compared the elevation of the ground class
points extracted from the lidar topography collected in 2014 (
The hydrogeological and geomorphological characteristics of Mallorca's river catchments control their surface water–groundwater interactions and thus generate different streamflow regimes (see Estrany et al., 2009). The headwaters of all subcatchments and the tributaries that drain the Llevant ranges and Marinas are ephemeral due to the high degree of fracturing, fissuring and karstification, which favour infiltration and percolation through perched karstic aquifers unconnected to the main stream channels.
The hydrological monitoring period assessed in this paper by using data from
the hydrometric station was from 10 January 2015 to 30 September 2018
(Fig. 6). The month of October 2018 was reserved for a
singular and deeper study that could describe the catastrophic flash-flood
event better (see results, Sect. 3.2). This gives a series of almost 4
hydrological years under hydrometeorological conditions illustrating
ephemeral behaviour of the Begura de Salma River that was average in terms
of precipitation (see the inset table, Fig. 6). In terms of
Discharge at 15 min intervals measured at the MEDhyCON hydrometric station located at the beginning of the concrete channel of the Begura de Salma River in Sant Llorenç des Cardassar. Likewise, the daily rainfall measured at the AEMET-B630 Ses Pastores during the monitored period (10 January 2015–30 September 2018), prior to the catastrophic flash flood of 9 October 2018, is shown. The bottom table displays the rainfall, runoff and peak discharge for hydrological years during study period. Rainfall data are from AEMET-B630 Ses Pastores, located 10.5 km from the Begura de Salma River catchment outlet and representative of the rainfall dynamics of the Llevant range's headwaters.
The hydrological response of the flash flood was analysed through variables
derived from the rainfall (Table 1a, seven variables) and runoff (Table 1b, nine
variables) of the catchment: event rainfall duration: duration from the
beginning of rainfall until it stopped; time of maximum rainfall: time of
the highest rainfall intensity; centroid storm: central time of the rainfall
event; average radar rainfall: mean rainfall obtained by radar; IP
Map of isohyets of the rain storm of 9 October 2018 in the
two headwater catchments of the Ca n'Amer River, i.e. the Sa Blanquera River and
Begura de Salma River. Source: 10 min radar images obtained from the web
Rain started to fall at 15:00 LT (official time; UTC
The hydrological model described above helped to understand the process
better during the event. The results of the hydrological model simulation
can be seen in Fig. 7. The input data used in the model were the
continuous radar data set described above and the temperature measured at the
surrounding meteorological stations, i.e. three stations within a radius of
12 km. The model was calibrated to reproduce the event and the final
parameters were set to
Initial conditions of the Génie Rural (GR) reservoir:
The relative volume error between the simulation and the measurement was
6 %. The simulated peak ratio was 437.7 m
The same parameters were used for the entire headwater catchment.
Accordingly, the
The flash flood had enormous social and economic repercussions in Llevant county and the whole of Mallorca, as well as extensive national and international media coverage. The flash-flood event was a catastrophe with 13 deaths and economic damage that had great impact on the population and infrastructure. The number of casualties in one of the most important international tourist resorts, considered traditionally safe, shook national and international opinion. For further assessment of the media impact, see Table S1.
The damage assessment report by the emergency services of the Balearic Islands Government showed an unprecedented mobilisation of resources in the region during the first week after the catastrophe in line with the high number of victims and amount of damage (Table 2; Pol, 2019a). The initial costs of the emergency works exceeded EUR 1.5 million including the following actions: cleaning and restoring river channels, demolition of walls and structures affected, removal of potential polluting sources (Pol, 2019b). The declaration of a disaster area is regulated by Spanish Law 2/2018, other complementary laws (BOE, 2018, 2019) and Decree 33/2018 (GOIB, 2018). These laws established public support for alleviating the basic needs of families, deaths, housing assistance, aid for loss of vehicles and support for the economic sectors affected. The laws provided assistance for the repair of public infrastructure and environmental damage, specifying the amount of aid and the administrative procedures to receive it. These regulations referred to all the affected areas, including the municipalities of Sant Llorenç des Cardassar but also Artà, Capdepera, Son Servera and Manacor. Recovery was financed jointly by the different public administrations: the Spanish Government, the Balearic Islands Autonomous Government, the Mallorca Government and the Sant Llorenç des Cardassar City Council. In April 2019, the expenditure of the Autonomous Government had reached EUR 30.4 million in recovery and mitigation actions (GOIB, 2019). This expenditure included aid to the affected towns and villages of EUR 11.27 million (EUR 2.7 million for the Sant Llorenç des Cardassar City Council), aid to companies of 3.3 million, for rehabilitation of homes (1.6 million), vehicle recovery (1.5 million), social aid (1.2 million) and EUR 0.264 million for deaths. The Sant Llorenç des Cardassar City Council, deployed various funds from the Spanish Government and Autonomous and Mallorca Governments for an investment plan in Sant Llorenç des Cardassar of EUR 3.51 million (Ajuntament de Sant Llorenç des Cardassar, 2018). In parallel, the Insurance Compensation Consortium (CCS, 2018), the Spanish public agency that handles payments to affected people in cases of damage caused by catastrophic events, processed claims for the flash flood as well as all the payments following damage assessment after the disaster in Sant Llorenç des Cardassar. A total of 774 claims were processed, with EUR 6 842 468 paid out (see Table S2 in the Supplement).
Damage summary and emergency actions after the 9 October 2018 violent flash flood in Llevant county, Mallorca. Source: Pol (2019b).
A territorial and hydrological analysis of the damage assessment is
developed here. The location of the affected buildings and the WS reached in
the streets and buildings provided by the emergencies department of the
Balearic Islands Government and by the Copernicus EMS enabled three affected
zones within the urban area of Sant Llorenç des Cardassar to be mapped
(Fig. 8a). Zone 1 was due to the overbank flow of the Begura de Salma River and corresponded mostly to the affected areas defined by the
Copernicus EMS. In this zone (1), the highest WS in the streets was reached,
exceeding 3.3 m. Zones 2 and 3 were those urban areas affected by the
overbank flow of the Sa Muntanyeta creek, located in the northernmost area
of the village. The streets of Sant Llorenç des Cardassar rerouted the
overbank flow from both the Begura de Salma River and the Sa Muntanyeta
creek (Fig. 8b), causing significant damages to vehicles and movable public
property. In addition, as most of the buildings in Sant Llorenç des
Cardassar use the ground floor as a home or business, the event caused major
flooding by water and mud that made their use impossible and required
cleaning and restoration. According to the Balearic Islands Government, 392
damaged buildings and plots were inventoried in the urban area of Sant Llorenç
des Cardassar, most of them in zone 1 (Fig. 8c and 8d). The
flow direction illustrated how the north-to-south direction, parallel to
the Begura de Salma River, caused most damage in zone 1, with 349 affected
buildings and a WS average of approximately 1.03 m. Zones 2 and 3 had lower-intensity
damage, with 37 and 6 affected buildings and a maximum WS of
1.80 m and 1.60 m, respectively. In these zones (2 and 3), the flow direction had no clear
pattern because the Sa Muntanyeta creek has a small catchment (2.2 km
The maps included in the Balearic Flood Risk Management Plan (GOIB, 2016) indicate the urban area of Sant Llorenç des Cardassar as a maximum risk area. Accordingly, the plan developed an analysis of the potentially affected areas for recurrence periods of 10, 100 and 500 years (Fig. 8e). In addition, Table 3 analyses the damaged buildings to see if they are included in these official flood risk areas in accord with the recurrence periods. None of the flood risk maps for different return periods encompassed the areas affected as a result of the event. The 10-year recurrence map only included 25 % of the affected areas; the 100-year map covered 48 % of the area damaged, while the 500-year map only reached 60 % (Table 3).
Damaged buildings in the village of Sant Llorenç des Cardassar caused by the violent flash flood on 9 October 2018 and those encompassed in the official flood risk maps for 10-, 100- and 500-year recurrence periods.
On comparing the affected zones where damaged buildings are also depicted in the Copernicus EMS, some differences between the initial flash-flood definition carried out by the Copernicus EMS-EU and the distribution of damaged buildings were found (Fig. 7f) in zones 2 and 3. It is worth noting that the post-event definition of the Copernicus EMS (Copernicus Emergency Management Service, 2018) covered approximately 90 % of the real damage.
The search for the only person missing during the flash flood who had not yet been found 6 d after the storm caused considerable social consternation in the Balearic Islands and beyond. Subsequently, hydrogeomorphological precision techniques were crucial. A very intense topographical survey constructed very high-resolution (i.e. 5 cm pixel size) digital elevation models and orthophotomosaics.
First, given the emergency situation, the index of (water and sediment) connectivity at the basin scale was used to identify those areas with the greatest sediment deposition potential (Fig. 9a). The IC shows well the sediment transfer processes within drainage catchments. The most connected areas of a basin are those in which their different compartments are more powerfully linked. That favours the largest water surface flow generation and thus erosion and, potentially, larger soil losses. On the contrary, the zones with low connectivity are those whose topographical characteristics disconnect water and sediment flows, acting as storage or deposition areas. The IC was applied to the whole Ca n'Amer River basin but was only analysed from the point where the missing person was last seen (Fig. 9b, point 1). That was the exact point where the car in which he was driving was swept away by the flood wave. Therefore, the preferential water and sediment paths most likely to be followed by the flood flows were identified, as well as the most important deposition areas downstream from the last point the person was seen. The most likely deposition zone was identified and immediately communicated to the emergency authorities, upstream from the bridge of the Ma-15 road which crosses the Ca n'Amer River about 1 km below Sant Llorenç des Cardassar. The search activities concentrated on that area, which is where the last victim was found (Fig. 9b, point 2) when the emergency authorities had decided to move their search activities to the mouth of the Ca n'Amer River and beyond into the Mediterranean.
Spatial patterns of hydrological and sediment connectivity
(deposition zones in blue colours)
In addition to the ability of HR-DTMs to improve sediment connectivity as a
powerful tool to determine preferential flow paths and deposition areas, the
present study evaluated the landform signatures of the event by using UAVs
as a rapid-response tool for post-catastrophe search and rescue tasks along
the whole downstream section of the Ca n'Amer River from the village of
Sant Llorenç des Cardassar, in order to measure effectively the sediment
deposits generated by the flash flood and to locate and quantify the most
important deposition areas downstream from where the person was last seen.
As the last missing person was found by using the connectivity index, in the
end the sediment deposition measurement was not needed during the emergency
operation. However, this study checked its validity by assessing the
floodplain area where the last person was found. Accordingly, the elevations
of each of the 7103 lidar points on the right bank of the Ca n'Amer River
were compared. From the differences interpolated (TIN), an elevation raster
for a total volume of 844.28 m
The flash-flood event described in this study fits with the monthly
distribution of flash floods in Spain carried out by Gaume et al. (2009):
October is the month with the highest number
of this type of flood event. In addition, the hydrological characteristics
of the event were comparable with the flash-flood requirements established
by Amponsah et al. (2018) for inclusion
in the EuroMedeFF database, which are a
The hydrological model was calibrated specifically for the flooding event.
The parameters of the modified GR reservoir as well as the initial
conditions were adjusted to best represent the flooding event. A very
sensitive parameter is the
The predictability of flash-flood events is unresolved, especially because forecasting of intense thunderstorms has also not been solved by operational meteorology. Even using one of the best state-of-the-art weather forecasting models, HARMONIE/AROME, the Spanish National Weather Service (AEMET) only activated a yellow warning for 1 h accumulated precipitation of 20 mm beforehand. In contrast, the synoptic situation was forecast well by global forecasting models some days before the event. An experienced forecaster could anticipate the occurrence of an intense thunderstorm by using these models but would lack any quantitative or geographical precision, which are two key factors in flash-flood forecasting. However, nowcasting products, based on radar, satellite and ground conditions may anticipate severe weather situations better. These products are updated often (several minutes to 1 h) and compensate the weather forecasting models which are updated less often. The main challenge in using the hydrological model as a flash-flood early warning system is to include correctly initial soil saturation conditions as well as accurate rainfall forecasts. For the latter, the scientific community is working on nowcasting products that typically deliver short-term (few hours lead time) rainfall forecasts that are updated very often, from a base of 10 min to an hour. These forecasts are based on real-time measurements that combine data from radar, satellites and meteorological stations. However, it is hard to calculate initial conditions automatically, as the river is dry most of the time and there are no soil moisture measurements in the catchment. Data assimilation and automatic adjustment of initial conditions, which are usually applied in operational forecasts, are therefore not relevant here. However, an early warning system can be built using the model proposed in this paper by assessing the uncertainty of the forecast. At present, Mallorca does not have any sort of early warning system to assist flood risk management and neither does Sant Llorenç des Cardassar. Similarly, no hydrometeorological early warning was issued by the competent authorities, as the Balearic Islands have no operational hydrological control network releasing real-time information on discharge. In October 2018, Sant Llorenç des Cardassar was one of the four municipalities in Mallorca with a flood risk emergency plan. However, it was not operational at the time the emergency was declared. As a result, the population was completely unaware of how to defend themselves, even during the emergency phase, although Sant Llorenç des Cardassar had significant social vulnerability to floods, as most of the casualties were tourists and the elderly.
The addition of the MEDhyCON research group on 15 October 2018 to the emergency operation brought in the application and testing of hydrogeomorphological precision techniques. The fundamentals are that flood risk plans and emergency activities are based on a thorough understanding of linkages between sediment and catchment compartments at all stages of flood events. Integrating topography-based connectivity assessment (Kalantari et al., 2017) and geomorphic change detection may be a crucial support to decision-making in flood risk planning and in emergency surveys, as this study shows. The combination of hydrological and sediment connectivity (IC in various forms) with other key natural characteristics (i.e. soil type and topography by using lidar-based HR-DTM), along with the integration of territorial information such as land cover/uses by using Cadastre databases (Piaggesi et al., 2011), results in a powerful tool. Accordingly, the easy-to-calculate IC can be an effective tool for rescue tasks after extreme flash-flood events with a huge erosion capacity.
In addition, the post-event delimitation and damage assessment released by the Copernicus EMS (Copernicus, 2018) identified approximately 90 % of the real damage in this traditional Mediterranean village, consisting of compact blocks of buildings and plots. The synthetic aperture radar (SAR) technology with very high spatial resolution (1–3 m; Plank, 2014) is fundamental to obtaining high efficiency and accuracy of this rapid mapping tool at low cost. Consequently, emergency resources can be directly concentrated on the most damaged areas without having to check the entire affected area on the ground.
The increase in the torrentiality of rainfall as a result of climate change in the Mediterranean region may exacerbate the level of exposure of urban areas and infrastructure to floods. Catastrophic events will increase in quantity and intensity. Local government bodies will need to adapt continuously prevention and management of flood risk tools to these new scenarios. The legal framework for flood risk planning and management (GOIB, 2016) showed that the level of risk exposure was extensively known. In addition, the analysis of current regulations shows that the appropriate preventive measures were being taken to minimise possible damage in a potential event in the Balearic Islands. However, the magnitude of this flash flood exceeded any type of forecast carried out by the risk and emergency plans. The consequences of the catastrophe revealed deficiencies in prevention by the local government, both at the level of urban planning and infrastructure and in risk management itself. In addition, the population was also unprepared due to a very low level of risk culture.
The hydrogeomorphological analysis and damage assessment developed in this
paper has provided a comprehensive understanding of the Sant Llorenç des
Cardassar flash-flood event of 9 October 2018 by means of an
integrated approach with a meteorological, hydrological, geomorphological,
damage and risk data analysis. The use of rainfall radar data – corrected
with measurements from rainfall stations in the surrounding region –
combined with
The flash-flood event was a catastrophe that caused 13 casualties, huge economic damage and an unprecedented mobilisation of human resources in the Balearic Islands. Rapid mapping from Copernicus EMS and detailed damage reported by regional authorities, linked to territorial information from the Cadastre and hydrogeomorphological processes, showed very accurately the damage-driving factors in the urban area of the village of Sant Llorenç des Cardassar. Although flood risk planning showed the high level of risk exposure, the disaster was generated by a very high exposure of buildings and infrastructure to floods, the absence of early warning systems with efficient action protocols and the lack of municipal regulations to instruct the population on how to act when struck by an event of this magnitude. The incorporation of hydrogeomorphological precision tools during emergency post-catastrophe operations was highly effective. Then, the simple application of a geomorphometric index from easy-access lidar-based topographic data resulted in a rapid identification of deposition zones in the different compartments of a catchment, which helped in the search and rescue of missing persons. In addition, the evaluation of landform signatures by using UAVs measured effectively the sediment deposits generated by the flash flood and/or mobilised by the emergency operations during rescue searches.
This study represents a first step to further improvement of flash-flood risk management in Mediterranean flood-prone regions such as Mallorca, which are likely to recur due to global change. Mediterranean regions are subject to violent flash floods that may intensify – especially in terms of peak discharge – in the future due to forest fire, land use and/or climate changes. These future consequences of global change should lead to the modification of hydrological and flood risk models, enabling the development of a rule-based system on the catchment scale, with adaptive and resilient measures to be taken.
The hydrological data analyzed in this paper can be freely downloaded from the PANGEA data repository; see Estrany et al. (2020a,
The supplement related to this article is available online at:
JE, MR, RM, AC and FV developed the experimental design, whilst JE, JF and JG were responsible for data curation, fieldwork and figure elaboration, and MT carried out the meteorological analysis. BN and FV performed the hydraulic modelling. RM and XP developed the hydrological model code and performed the simulations. MR completed the damage assessment. AC performed the sediment connectivity and geomorphic change detection. Resources and funding acquisition were supervised by JE and MR. JE prepared the manuscript with contributions from all co-authors.
The authors declare that they have no conflict of interest.
Josep Fortesa has a contract funded by the Vice-presidency and Ministry of Innovation, Research and Tourism of the Autonomous Government of the Balearic Islands (FPI/2048/2017). The contribution of Miquel Tomàs-Burguera was supported by the project CGL2017-83866-C3-3-R, also funded by the AEI. Julián García-Comendador is in receipt of a pre-doctoral contract (FPU15/05239) funded by the Spanish Ministry of Education and Culture. Compensation payments were facilitated by the Spanish Insurance Compensation Consortium, whilst the type of buildings and land use at the urban plot scale were provided by the Spanish Directorate General for the Cadastre, and the damage report was by the Directorate General of Emergencies of the Balearic Islands Government. Meteorological data were facilitated by the Spanish Meteorological Agency (AEMET). We are grateful to BalearsMeteo for providing subhourly rainfall data of the event. The authors want to thank Xurxo Gago, Carlos J. Oliveros, José A. López-Tarazón and Hassan Ouakhir for their assistance during fieldwork. Finally, we want to pay tribute to all the professionals and volunteers who worked with such determination in the rescue tasks.
This research was supported by the Spanish Ministry of Science, Innovation and Universities, the Spanish Agency of Research (AEI) and the European Regional Development Fund (ERDF) through the project CGL2017-88200-R “Functional hydrological and sediment connectivity at Mediterranean catchments: global change scenarios -MEDhyCON2”.
This paper was edited by Kai Schröter and reviewed by two anonymous referees.