The role of different factors related to social impact of heavy rain events: considerations about the intensity thresholds in densely populated areas

In the assessment of social impact caused by meteorological events, factors of different natures need to be considered. Not only does hazard itself determine the impact that a severe weather event has on society, but also other features related to vulnerability and exposure. The requests of data related to insurance claims received in meteorological services proved to be a good indicator of the social impact that a weather event causes, according to studies carried out by the Social Impact Research Group, created within the framework of the MEDEX project. Taking these requests as proxy data, diverse aspects connected to the impact of heavy rain events have been studied. The rainfall intensity, in conjunction with the population density, has established itself as one of the key factors in social impact studies. One of the conclusions we obtained is that various thresholds of rainfall should be applied for areas of varying populations. In this study, the role of rainfall intensity has been analysed for a highly populated urban area like Barcelona. A period without significant population changes has been selected for the study to minimise the effects linked to vulnerability and exposure modifications. First, correlations between rainfall recorded in different time intervals and requests were carried out. Afterwards, a method to include the intensity factor in the social impact index was suggested based on return periods given by intensity–duration– frequency (IDF) curves.


The role of different factors related to social impact of heavy rain events: considerations about the intensity thresholds in densely populated areas 1 Introduction
The study of the impact that severe weather events have on a territory and its population is of maximum interest for society, involving a broad range of sectors in addition to meteorology and climatology specialists, such as emergency services, policy makers and insurance companies.For this purpose, the Social Impact Research (hereafter SIR) Group was created under the frame of the MEDEX project (MEDiterranean EXperiment on cyclones that produce high-impact weather in the Mediterranean; http://medex.aemet.uib.es).Since 2008, the SIR Group has carried out several studies on this subject, centred in two regions of Spain, the Balearic Islands and Catalonia (Fig. 1).Requests related to insurance claims which are received in Meteorological Services were used as proxy data, as they proved to be a good indicator of social impact (Amaro et al., 2010).They reveal if an event has caused damages in accordance to the number of claims received because of adverse meteorological conditions.These claims can be translated into economic losses that citizens have to face.One of the advantages of this indicator is that it is reasonably objective because it depends on facts: the weather event causes a material damage.On the contrary, one of the drawbacks is that the requests of reports can be altered by some factors, such as the insurance company proceedings and the variations of freely available online data.On the other hand, the most severe weather events demand especial treatment and do not require usually individual reports, but are dealt globally by regions.
This proxy data can be contrasted with others, like the media coverage.Press news can be used to obtain an indirect estimation of the risk (Llasat et al., 2009b), so it is a good indicator to verify the social impact of the analysed events.
A methodology for treating the requests was defined by the SIR Group, in order to use them in social impact studies.The SIR Group analysed the events occurred in the Balearic Islands and Catalonia during the period 2000-2002 (Amaro et al., 2010).Afterwards, a cumulative index (CI) to indicate the social impact expected for rain events was established.An index is a valuable tool to foresee approximately the effects that an event can have, in which variables of different nature interact.
The CI was built as a non-dimensional index, including the following factors: the maximum precipitation in 24 h of the event, population affected by rainfall exceeding 60 mm, length of the event and finally the coincidence with a strong wind event.Each factor was quantified in levels from 1 to 10 according to different criteria for each of Introduction

Conclusions References
Tables Figures

Back Close
Full them (Amaro et al., 2010).As a result, the CI equation is as it follows (Eq.1): where L R -level associated to the maximum rainfall in 24 h; L P -level associated to the population affected by rainfall exceeding 60 mm; L TR -level associated to the length of the rain event; L TW -level associated to the coincidence with a strong wind event.
The events with a higher number of requests, and therefore, more social impact, should have a greater CI (being the maximum possible value 40).Even though the CI adjustment was suitable for most of the events, there were some exceptions which indicated that other factors ought to be considered.
In this study, the analysis of the exceptions in which the CI had a worst adjustment in Catalonia during the period 2000-2002 has been carried out in order to define which other factors should be also taken into account (Fig. 2a).The revision's results show that two main features stand out: population density, which is connected with vulnerability, and rainfall intensity, related to hazard.As it has been observed, rainfall intensity is particularly relevant in high-densely populated areas, so studies for these areas require priority treatment.Therefore, a study of the intensity has been carried out for the city of Barcelona.According to the availability of intensity data, this study has been completed for the period 2008-2011.
The main goal of this paper is to determine the role of rainfall intensity in the social impact that rain events have in a densely populated area.After determining the influence of the rainfall intensity, the next aim is to find a suitable way to include it in the CI.
This paper is structured in three sections: first, a revision of the events that had a worst CI adjustment will be made.Then, an assessment of the rainfall intensity factor will be carried out for the city of Barcelona, analysing its correlation with the requests related to insurance claims received after heavy rain events, and proposing a method to include this factor in the CI.Finally, some conclusions will be drawn and future work will be suggested.Introduction

Conclusions References
Tables Figures

Back Close
Full

Events' revision
The heavy rain events of the period 2000-2002 occurred in Catalonia that presented a worst CI adjustment have been analysed in order to determine which modifications should be applied to the index (Fig. 2a).
In the Western Mediterranean region, recent studies show an increase in vulnerability mainly related with the population displacements towards the coastal regions, where flash floods are frequent (Llasat et al., 2009a(Llasat et al., , 2010)).The more populated a region is, the more social impact a weather event causes in it.Because of this, the population factor had been included to build the CI when a maximum of rainfall over 60 mm in 24 h was recorded, according to the threshold established by MEDEX.Nevertheless, one single threshold for all Catalonia has proved to be insufficient.One of the best examples to illustrate this fact is the 14th and 15th of July 2001 event.The number of requests received (133) was higher than expected according to the maximum rainfall in 24 h (98 mm).In fact, the CI for this event was pretty low (Fig. 2a).Moreover, the analysis of requests by municipality showed that approximately half of them came from municipalities of Barcelona's Metropolitan Area, where the maximum rainfall was around 40 mm (Fig. 3a).It is Catalonia's most populated region (Fig. 3d), with more than 10 000 inhabitants km −2 , and the threshold of 60 mm in 24 h demonstrates to be too high for it, as there was a remarkable social impact caused by a lower precipitation.According to press, there were 40 departures of firefighters during this event.
On the contrary, other events had less social impact than it was expected according to the requests received.One of the most outstanding examples is the 21-25 October 2000 event.Only 36 requests were received, with a high maximum accumulated rainfall recorded in 24 h of 270 mm (Fig. 3b).The main reason was that it affected mostly rural areas, where the population density is low.
In the same way, Petrucci and Pasqua (2008) obtained high levels of damage in areas characterised by high population density while in scarcely populated areas hy-Introduction

Conclusions References
Tables Figures

Back Close
Full drogeological phenomena did not induce important damage.Consequently, the events' revision suggests that different rainfall thresholds for differently populated areas ought to be applied (Barbería et al., 2011).
The studied cases also indicate that rainfall intensity cannot be ignored.The 10 August 2002 event had a higher number of requests than it was expected according to the CI.It affected an extensive part of Catalonia, including areas of the Metropolitan Area of Barcelona (Fig. 3c).There, the rainfall intensity exceeded 20 mm in 30 min, which corresponds to the first level of warnings (Vilaclara et al., 2010) according to the thresholds applied in the Servei Meteorològic de Catalunya (hereafter SMC) for Civil Protection.Another similar example is the 14-15 July 2001 event, when rainfall intensity also exceeded 20 mm in 30 min in the Barcelona area.Consequently, the overcome of this threshold is likely to be related to a high social impact.
On the other hand, we must bear in mind that during the same event of 14-15 July 2001, the maximum precipitation intensity (36.4 mm in 30 min) was recorded in the north part of Catalonia, coinciding with the maximum rainfall in 24 h (Fig. 3a), but it caused a minor impact because this is low populated area (Fig. 3d).Therefore, the rainfall intensity should be considered to assess social impact of heavy rain events, but also in a different way for urban and rural regions (Amaro et al., 2011).
Other factors that stand out after the revision are the length of the event and the coincidence with strong wind, which seem to be less relevant than the other CI factors.
In some cases they are partly the cause for a high CI which is not verified by a high number of requests.The 4 January 2001 and 28 January 2001 events exemplify the overestimation of the CI caused by the level associated to the coincidence with a strong wind event.These events have a quite low number of requests, but their CI stands out compared to the other events with a similar number of requests.
On the other hand, the 20-24 September 2002 event illustrates the overestimation caused by the level associated to the length of the event.According to the number of requests, it should have had a lower CI, but the number of days increased the CI value.Introduction

Conclusions References
Tables Figures

Back Close
Full Finally, some phenomena that are more unusual can be decisive, like hail.In agricultural regions, during one single hail event, a high percentage of crops of an area can be damaged, causing a huge impact, regardless the other factors.This was observed during the 14 and 15 July 2001 event, when hail affected between the 80 % and 100 % of fruit tree crops in a specific rural area in west Catalonia, so a considerable number of requests were received from this region, despite the low population density.Hail also must be considered when urban areas are affected, as in the 10 August 2002 event, when hail of a diameter up to 4 cm was reported in cities near the Barcelona city area.This likely contributed to the significant number of requests received.

Conclusions of the revision
The events' revision confirms that, in social impact assessment, differently populated areas should be approached in a different way.To carry out a regionalization, the most relevant fact is not the total population of a municipality, but the concentration of population in relation with the land, which is also linked to the type of urbanization and the land use.The more elements at risk and the more valuable and susceptible that those elements are, the higher vulnerability to flooding (Scheuer et al., 2011).Thus, urbanisation enhances the risk (Nirupama and Simonovic, 2007) and cities densely populated become more vulnerable.
Regarding the CI, some improvements are introduced: the intensity and the population density should be included in it, and, on the contrary, the levels associated to the length of the event and to the coincidence with a strong wind event need to be reduced, as they have proved to be less relevant (Barbería et al., 2011).The lack of systematic hail information for the period 2000-2002 prevents from including this factor in the CI.
As a first approach, to build the reassessed cumulative index (RCI), a new threshold of 40 mm in 24 h for municipalities overcoming 10 000 inhabitants km −2 is suggested.
The levels associated to this new factor (L P40 ) follow the same criteria established by Amaro et al. (2010) for the levels associated to the population affected by rainfall exceeding 60 mm.Consequently, different levels have been obtained from Eq. ( 2), Introduction

Conclusions References
Tables Figures

Back Close
Full where Po40 is the number of people affected by rainfall between 40 mm and 60 mm in municipalities overcoming 10 000 inhabitants km −2 , and PoT is the total population of Catalonia.
Only the rainfall between 40 mm and 60 mm is taken into account as the population affected by precipitation overcoming 60 mm is already considered in the L P factor.Thus, the L P and L P40 factors together have a maximum value of 10.
On the other hand, the levels associated to the length of the rain event and to the coincidence with a strong wind event have to be reduced.In order to maintain 1 as the minimum value associated, root operation was selected to reduce the levels.The best results were achieved using the square root.
Accordingly, the resulting equation for the RCI is as it follows (Eq.3): where L R -level associated to the maximum rainfall in 24 h; L P -level associated to the population affected by rainfall exceeding 60 mm; L P40 -level associated to the population affected by rainfall between 40 mm and 60 mm in areas of more than 10 000 inh km −2 ; L TR -level associated to the length of the rain event; L TW -level associated to the coincidence with a strong wind event.
The Fig. 2b shows the RCI with these corrections.All the events in the first and second quartile (lower number of requests) have values below 8, whereas in the CI some events in the first quartile were above it.That was too high considering the low number of requests and in comparison with the events with a higher impact, so a better adjustment has been achieved.On the other hand, in this approach the events with a higher impact have a RCI above 11.Nevertheless, there are some exceptions that could be explained by the intensity and hail factors, which are not included in the equation yet.
Consequently, the intensity factor should be added.In the same way as total rainfall, this parameter must be incorporated in a specific way for differently populated regions.In order to achieve that, a first assessment on the intensity role has been carried out for the city of Barcelona.

Focusing on hazard: the intensity factor
Damages can be caused by intense precipitation that does not amount a high total quantity.According to Diakakis (2011), there is a strong connection between peak storm intensity and flood occurrence.Therefore, this variable is essential for social impact studies.For Catalonia, a typology of floods was set up according to their hazard and potential impact (Llasat, 2009).This classification includes short-lived events of very intense precipitation caused by isolated cells or multicells with a limited horizontal extension and episodes of heavy rain sustained for several hours caused by multicells or mesoscale convective systems.The occurrence of these flood events supports the fact that high intensity rainfall has a very remarkable impact on the Catalan territory regardless the total quantity recorded in 24 h.Risk has been described as the combination of hazard, exposure and vulnerability (UNISDR, 2012).As rainfall intensity is part of the hazard factor, in order to focus on this variable, a number of events where the other two factors are minimised will be analysed.To achieve that, a fixed area and a period without significant changes in population have been selected: Barcelona, which is the most densely populated city in Catalonia, and the period 2008-2011.

Considerations about the selected period
The period 2008-2011 has been chosen taking into consideration, firstly, the rainfall data availability.Since 2008, the SMC has three automatic rain gauges in the city of Barcelona, which are located in different city areas (Fig. 4), and cover the pluviometric regimes that have been suggested for the Barcelona urban area (Rodríguez Introduction

Conclusions References
Tables Figures

Back Close
Full  , 2012).Therefore, we can consider that most of the rain events that affected the city are included in the study.
For the city of Barcelona, an increase in the number of extraordinary floods was detected particularly since the 19th century due to rising vulnerability (Llasat et al., 2008), but on the other hand the number of catastrophic floods has decreased thanks to the renewal of the drainage system of the city (Barrera et al., 2006).Consequently, it is important to select a period with no significant changes in vulnerability in order to reduce it to the point of being negligible.During the period 2008-2011 there were no significant changes in the city population (Table 1) so similar vulnerability can be assumed.
Concerning the total number of requests received in the SMC during the analysed period, a considerable decreasing is detected in 2011 (Fig. 5a), even though the number of heavy rain events was higher that year.In Catalonia, there were ten events that exceeded the highest level of warnings established by the SMC (more than 40 mm in 30 min).Factors related to online data are likely to have affected, at least partly, the number of requests, according to the website statistics.Between 2009 and 2011, an increase in the number of page views at the SMC website (www.meteo.cat) is detected (Fig. 5b).Statistics of the year 2008 are not shown as they are not available.Data of the automatic weather stations are freely accessible in the SMC website, and citizens are likely getting acquainted with online data search.The highest increasing in page views took place during the year 2010, when the number of requests decreased, and this trend continued in 2011.Consequently, different thresholds of requests ought to be considered when selecting the events that had a significant social impact.

Heavy rain events in Barcelona during 2008-2011 and their social impact
The rain events that took place during the four-year period 2008-2011 in the city of Barcelona and their social impact have been analysed in this study.Rainfall data from the three automatic rain gauges that the SMC has in the city of Barcelona and requests received related to insurance claims have been collected.In the SMC a systematic Introduction

Conclusions References
Tables Figures

Back Close
Full quality control is applied to the data recorded by the automatic weather stations.It consists of plausible value check (if the values are within the acceptable range limits) and time, internal and spatial consistency checks.These checks are performed daily, monthly and yearly, hence this procedure guarantees data reliability.Moreover, data recorded by the rain gauges of the sewer system in the city of Barcelona (hereafter CLABSA) has been used to verify the maximum rainfall of the events.The selection of events was carried out as it follows: first, a list of rain events was made according to the requests related to insurance claims received at the SMC.On the other hand, another list of rain events was made based on the records of the three automatic rain gauges mentioned before.The maximum rainfall in 24 h of each event was obtained from the daily data, recorded from 00:00 to 24:00 UTC.Also, the rainfall accumulated from 08:00 UTC on one day to 08:00 UTC on the following day was built, in order to take into account the pluviometric day.Afterwards, the lists obtained from the requests and from the maximum rainfall were crossed.
The results showed that for the 2008-2010 period, the events that had five or more requests were the ones in which the maximum rainfall in 24 h exceeded 40 mm, with some exceptions.For the year 2011, the events that had three or more requests matched with that threshold, except for two of them.For all the exceptions, even though no SMC rain gauge did exceed an accumulated precipitation of 40 mm in 24 h, the exceeding of this value was recorded by the CLABSA rain gauges (Table 2).Therefore, the overcome of this threshold shows a clear relation with a high social impact in the area of Barcelona.It verifies that 40 mm in 24 h is a suitable threshold for densely populated areas, as it was proposed for the RCI equation (see Sect. 2.1).Consequently, the events with a maximum rainfall over 40 mm in 24 h were selected, as they showed relation with social impact.Twenty rain events met this criterion and the RCI was calculated for them (Table 2).The population factor was not included, as this study is focused in the city of Barcelona, and therefore, it does not change.Introduction

Conclusions References
Tables Figures

Back Close
Full

Correlations between rainfall in different time intervals and number of requests
In order to determine the influence of rainfall intensity in social impact, correlations between precipitation recorded in different time intervals and number of requests received have been carried out.Therefore, we will be able to determine not only if is the intensity correlated to social impact, but also which measurement periods are more significant.The periods of 24 h, 60 min, 30 min and 10 min have been considered (Table 3).For six out of the twenty events, only accumulated rainfall values in 24 h were available, so fourteen events have been included in the correlations.
Least squares regression has been applied and Pearson's correlation coefficient has been calculated (Fig. 6).Significance of the regressions has been tested and therefore, it has been proved whether the different variables considered are linearly related or not to the number of requests.The freely available statistical software package R has been used (http://www.r-project.org).
The results show that the maximum rainfall in 60 min, 30 min and 10 min have statistically significant linear correlation with a 95 % confidence level with the number of requests.For the 30 min and 10 min periods (Fig. 6c and d) the correlation is very similar, and the p values near zero indicate very strong evidence against the null hypothesis of no linear association.The maximum rainfall in 60 min (Fig. 6b) is also correlated to the number of requests, but not as strongly as the maximum rainfall in 30 min and 10 min.
On the contrary, the maximum rainfall recorded in 24 h (Fig. 6a) does not show statistically significant linear correlation (95 % confidence level).
Therefore, the maximum rainfall in 30 min or the maximum rainfall in 10 min should be included in the RCI in order to introduce the rainfall intensity factor, as these variables have the best ability to explain the number of requests.Introduction

Conclusions References
Tables Figures

Back Close
Full

Categories of intensity
In order to include the intensity factor in the RCI, different categories of intensity have been built.The first step is to determine which measurement period is selected, taking into account that both the 30 min and the 10 min periods presented a similar correlation with the number of requests.The 30 min period has been chosen as the range of values for the period of 10 min is quite limited, while the period of 30 min covers a wider range (Fig. 7), and therefore, would facilitate the creation of categories.Moreover, the 30 min period is used in the SMC for Civil Protection warnings, and the availability of historic data is higher for the 30 min measurement period than for the 10 min one.
For building categories, different criteria can be applied.In this study, the categories are based on return periods, so they are connected to the exceptionality of the events.The return periods are used for warning thresholds like Meteoalarm, where, for instance, return periods over 5 yr correspond to red warnings (Stepek et al., 2012).
To determine the values of precipitation in 30 min associated to different return periods, Intensity-Duration-Frequency (hereafter IDF) curves have been used.For the city of Barcelona (Observatori Fabra) IDF curves were built (Pérez, 2012), being the generalized equation the following (Eq.4): I = 15.2 + 4.26 ln T/(6.5 + t) 0.780  (4) where I -intensity in mm min −1 ; T -return period in years; t -duration in minutes.
The rainfall in 30 min was calculated for nine return periods (0.5, 1, 2, 5, 10, 15, 25, 50 and 100 yr) (Table 4).Based on the rounded values, ranges of rainfall in 30 min were ascribed values from 1 to 10 (Table 5) to build the levels of the intensity factor (L I ).The value 1 corresponds to the lowest intensity range and the 10 to the highest.Therefore, the minimum value (1) will be for a maximum rainfall in 30 min associated to a return period below 0.5 yr (between 0 and 20 mm), and the maximum (10) for a maximum rainfall in 30 min associated to a return period beyond 50 yr (more than 60 mm).

NHESSD Introduction Conclusions References
Tables Figures

Back Close
Full

Reassessed cumulative index adjusted for densely populated areas
The categorization of rainfall intensity allows the introduction of this variable in the RCI equation.Therefore, the RCI for highly populated areas is as it follows (Eq.5): where L R -level associated to the maximum rainfall in 24 h; L I -level associated to the maximum intensity in 30 min; L P -level associated to the population affected by rainfall exceeding 60 mm; L P40 -level associated to the population affected by rainfall between 40 mm and 60 mm in areas of more than 10 000 inh km −2 ; L TR -level associated to the length of the rain event; L TW -level associated to the coincidence with a strong wind event.
The L I factor will be different for each region.In this study, the L I intensity factor was introduced for the city of Barcelona.The RCI for the 2008-2011 events (Table 6) shows a better adjustment after introducing the intensity factor.The population factor has not been taken into account as it does not vary.
Considering the better adjustment achieved after the introduction of the intensity factor, the next stage should be to introduce it in the revision of the events recorded during the period 2000-2002 (Sect.2).Particularly, the 14-15 July 2001 was one of the events with the highest number of requests but with a quite low RCI (below 11).As it has been commented in Sect.2, rainfall intensity was remarkable in the area of Barcelona, where the maximum rainfall in 30 min was 20.8 mm.This intensity corresponds to the second level of the L I (Table 5).If we introduce the L I in the RCI of this event, we obtain a value of 11, which is comparable with other events that had a high social impact.However, this improvement cannot be applied to the other events yet because the L I is not still calculated for all Catalonia.Thus, a specific study for each region is needed.Moreover, the importance of intensity should be evaluated to determine if its correlation with social impact is high in other areas that are not so densely populated.We must keep in mind that the lack of IDF Figures

Back Close
Full ered.However, this factor has not been analysed due to the lack of systematic hail information in the SMC during the studied period.The importance of rainfall intensity has been demonstrated for a high-densely populated urban area like the city of Barcelona.The heavy rain events that took place during the period 2008-2011 have been analysed.The results show that there is a statistically significant linear correlation (95 % confidence level) between the rainfall recorded in 60 min, 30 min and 10 min and the requests related to insurance claims.The best correlation has been observed for the 30 min and 10 min periods, being both very similar.In accordance with the SMC data availability, the 30 min period has been used to asses the social impact of the rain events.
A method for including the intensity factor in the social impact index has been suggested, based on return periods obtained from IDF curves.The results show a better adjustment of the RCI to assess the social impact of weather events in highly populated areas when the intensity factor is considered.Future work will be aimed at carrying out a regionalization based on the vulnerability of the territory, suggesting different thresholds of rainfall for each area.Furthermore, in order to introduce the intensity factor in the equation, IDF curves need to be calculated for more regions.Therefore, the RCI would become a useful tool to foresee events with high social impact and it will allow a better management of contingency plans in vulnerable areas.Figures

Back Close
Full Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | et al.
Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Table 4 .
Rainfall in 30 min calculated for different return periods for the city of Barcelona, based on the generalised IDF curves equation.