A large earthquake and tsunami took place in February 2010,
affecting a significant part of the Chilean coast (Maule earthquake,
The vulnerability variables that best explained the extent of the disaster were housing conditions, low household incomes and limited knowledge about tsunami events, which conditioned inadequate reactions to the emergency. These variables still constitute the same risks as a result of the reconstruction process, establishing that the occurrence of a similar event would result in a similar degree of devastation. For post-earthquake conditions, it was determined that all neighborhoods have the potential to be restorative environments soon after a tsunami. However, some neighborhoods are still located in areas devastated by the 2010 tsunami and again present high vulnerability to future tsunamis.
A tsunami is a phenomenon known for its great destructive power in a short period of time; however, the process of post-disaster reconstruction usually lasts a long time and generates significant socio-territorial transformations. A total of seven destructive tsunamis affected the coasts of Indonesia, Samoa, Chile and Japan in the last decade alone: 2006, 2007, 2009, 2010 (27 February and 24 October) and 2011. These tsunamis took the lives of 237 981 people and generated an estimated USD 456 million in economic losses (Løvholt et al., 2012, 2014). These degrees of destruction have been explained by a number of factors, such as ineffective early warning systems, inadequate information management by the population, lack of emergency mechanism coordination and high levels of social vulnerability (Rofi et al., 2006; Løvholt et al., 2014; Reese et al., 2007). Although scientific research has led to significant advances in the understanding of the generation and propagation mechanisms of these phenomena (Aránguiz et al., 2013; Løvholt et al., 2014), other aspects linked to social components (vulnerability and resilience) are less understood, primarily for post-disaster conditions, given social system dynamics and complexity. The latest events have shown that increased mortality may be associated with intrinsic aspects of vulnerability, which in the natural disaster context is defined as the inability of society to respond to an event, in this case a dangerous natural phenomenon (Anderson and Woodrow, 1989 in Cardona, 2001; Wilches-Chaux, 1993). Intrinsic aspects include population characteristics such as age and gender (Rofi et al., 2006), income levels and occupations (Birkmann et al., 2007; Prerna et al., 2015), ideological and cultural factors, levels of knowledge and inadequate reactions to the emergency (Ruan and Hogben, 2007). Other studies, through a line of still incipient work, have established that elements associated with social capital and territorial identity foster social resilience, which would be an enabling framework for overcoming the negative effects of a disturbance (Pelling, 2003).
The 2010 Chile tsunami showed the high fragility of social and institutional systems in coastal areas, as significant destruction along 600 km of coastline was observed (Quezada et al., 2010; Fritz et al., 2011; Yamazaki and Cheung, 2011; Jaramillo et al., 2012; Sobarzo et al., 2012; Bahlburg and Spiske, 2012; Martínez et al., 2012). Historical records show that these phenomena are not sporadic in the country but rather highly recurrent, causing significant devastation (Lomnitz, 1970; Monge, 1993, Lagos, 2000; Ruegg et al., 2011; Palacios, 2012).
Territorial planning in Chile, as in much of the rest of the world, has been focused primarily on interventions for mitigation (Herrmann, 2015), with policies and instruments for reconstruction (e.g., sustainable reconstruction plans and master plans) focused on housing production rather than social reconstruction of territories (Rasse and Letelier, 2012; Martinez, 2014). Thus, interdisciplinary approaches necessary for the reconstruction of human settlements in an integrated manner – i.e., studies which identify, assess and integrate physical, economic, social, environmental and perceptual factors – have been neglected. This complex approach has already been addressed in an international context, with the application of different study models of urban resilience to disaster (e.g., Cutter et al., 2008; Norris et al., 2008). Resilience refers to the ability of a community to adapt and recover after a disturbance without losing its character (Cutter et al., 2014; Walker and Salt, 2006) and is expressed multi-dimensionally (Cutter et al., 2014); in Chile, however, its psychological and social dimensions are the least considered in post-disaster planning. This is the case despite the fact that the integration of these dimensions in planning can promote community recovery after a disaster, with the potential to rebuild “the place where the restoration occurs” (Allan and Bryant, 2010). A restorative experience is described as “the process of recovering psychological and social resources that have become diminished in the efforts to meet the demands of everyday life” (Hartig, 2007, p. 164). After a large tsunami, the city “takes on a new meaning [and] its spaces and components are re-evaluated (by the people)” for their capacity to provide restorative experiences (Allan and Bryant, 2010). Thus, post-disaster reconstruction processes are an opportunity to effectively reduce risk and generate mechanisms of physical as well as social resilience.
In this context, we analyze pre- and post-disaster tsunami inundation risks in one of the coastal towns most affected by the earthquake and tsunami on 27 February 2010, which presented an intense transformation as a result of post-disaster reconstruction. It is unknown whether this reconstruction process has reduced vulnerability and provided a restorative urban system, enhancing urban resilience, or if it has generated new risk areas. Questions were asked regarding the neighborhoods being rebuilt in Dichato: e.g., do they have the potential to be restorative environments? Which specific sites provide restoration? Are restorative environments pre-existing areas that persist after the disaster, or are they new sites built during reconstruction? These questions seek to determine whether the reconstruction process has promoted the population's ability to adapt after a tsunami and whether reconstruction decisions have decreased the potential for damage in the case of future events.
Dichato is a town located on Coliumo Bay (36
Dichato's urban layout is characterized by a coastal plain of approximately
2 km
Geographical context of the study area. Dichato is located on Coliumo
Bay (36
In order to give risk a value in pre- and post-disaster conditions, the
equation
The tsunami hazard was estimated by means of a numerical simulation
considering the tsunami on 27 February 2010. The Non-hydrostatic Evolution
of Ocean WAVEs (NEOWAVE) numerical model (Yamazaki et al., 2009, 2011) was
used. This model solves linear and nonlinear shallow water equations using
nested grids with different spatial resolutions. In this case, four nested
grids were used, with resolutions of 120
Nested computational grids. The inset in grid 2 shows the tsunami initial condition (Hayes, 2010). TG in grid 4 indicates the location of the virtual tide gauge.
A Manning roughness coefficient of 0.025 was used and the total simulation
time was 6 h, with output results (water surface elevation and current
velocity) of 1 min. The tide level was set to the sea level at the time
of the maximum inundation. To do this, preliminary numerical simulations
were conducted to find the maximum tsunami wave. The tide level was
estimated to be
Hazard level according to flow depth.
Reference: modified after Walsh et al. (2005).
Hazard level according to flow current velocity.
Reference: modified after Jelínek et al. (2012) and González-Riancho et al. (2013).
The overall vulnerability analysis was performed for two scenarios: pre- and post-disaster. This was done in order to establish the variables related to the level of destruction and the effects generated by the process of post-disaster reconstruction in the configuration of new risk areas. Pre-disaster vulnerability was determined in 2010, while post-disaster conditions were determined using data collected in 2014.
In the case of pre-disaster conditions, the analysis unit corresponded to census blocks. Due to the destruction of most of the census blocks after the tsunami, for post-disaster conditions the analysis unit was the neighborhood. Neighborhood units were first defined with respect to their pre- and post-tsunami existence. In this line, newly reconstructed areas and old built zones were identified. Within each of these larger areas, neighborhoods were subdivided based on urban morphology characteristics such as type of grid, housing and land use. Finally, main streets such as avenues were used to define the boundaries of neighborhoods. Figure 1 illustrates the location of the nine neighborhoods in Dichato.
Variables associated with each dimension of vulnerability pre- and post-2010 tsunami.
For the overall vulnerability analysis, variables selected for both scenarios were representative of physical, socioeconomic and educational dimensions; however, some variables were modified according to the pre- and post-disaster availability of data (Table 3). For pre-disaster conditions, the variables related to the physical and socioeconomic dimensions were obtained from the 2002 census (INE, 2002). However, due to the destruction of the post-tsunami census blocks, the use of the census was invalidated and data were collected through surveys. To this end, variables related to those of the pre-disaster data collection were selected (for example, in the physical dimension the housing type variable is related to the number of floors variable). The survey was also used also to collect data related to the educational dimension in pre- and post-disaster scenarios, including the following variables: knowledge of evacuation routes, knowledge of safe zones and participation in educational programs or talks.
Three levels of vulnerability for each dimension were established (high, medium, low). The variables associated with each vulnerability dimension were given equal weight in the final matrix. This was done because previous studies that included the dimension of vulnerability in risk (Martínez et al., 2012; Rojas et al., 2014) used similar criteria, which have been found to be representative of local conditions. Variables were incorporated into the GIS using ArcGis 10.1 to generate thematic maps and summary charts through map algebra.
The capacity of the neighborhoods of Dichato to provide restorative post-disaster experiences was assessed through a perceived restoration study (Hartig et al., 1997). This type of study explores the capacity of environments to be restorative, which is particularly crucial in cities prone to natural disasters such as tsunamis. For Dichato in particular, this type of study can reveal the specific landscape attributes that influence perceptions, indicating whether restorative post-tsunami experiences are based on newly reconstructed neighborhoods and features or pre-existing neighborhood character. In other ways, it provides information regarding whether or not post-tsunami reconstruction contributes to a restorative experience and hence to resilience.
To undertake this study, the nine neighborhoods of Dichato were defined as analysis units (Fig. 1). People living in each of these neighborhoods assessed their environment by means of the perceived restorativeness scale (PRS), an instrument constructed based on the Attention Restoration Theory (Kaplan and Kaplan, 1989). The PRS has been used to identify landscape attributes that can be restorative to people subjected to high levels of stress and mental fatigue (Hartig et al., 1997; Korpela and Hartig, 1996; Ulrich et al., 1991). Three factors of the PRS were used to evaluate the interaction of people with the neighborhood they inhabit: being away (BE-AW), which reflects the need to escape from everyday life or daily mental activities that require major concentration; fascination (FAS), which is found in environments that attract and hold our attention without any effort; and compatibility (COMP), which refers to a sense of oneness with environments that provides the capability to meet our desires and needs. A survey was developed based on previously validated PRS questionnaires, including 15 sentences that refer to each of these factors (five for each). This survey was used to interview people in their homes during the same time the vulnerability questionnaire (described in the previous section) was administered. Each interviewee was informed about the survey takers' interest in understanding their experience in their current neighborhood. Accordingly, they were asked to assess each sentence from 1 to 7 (Likert scale), with 1 indicating that the sentence did not apply to their neighborhood at all and 7 indicating that it applied to the highest degree. As an example, a sentence referring to the BE-AW factor was “Being in this neighborhood is like a break from my daily routine.” Another sentence, referring to the FAS factor, was “There are many interesting things that catch my attention in this neighborhood.” A sentence referring to the COMP factor was “The activities that can be done in this neighborhood are activities that I enjoy.” After evaluating each of the 15 sentences, each person was asked to describe the neighborhood areas they recalled during the evaluation.
Numerical data obtained using the Likert scale were organized in Excel sheets for each interviewee and neighborhood and were later used to perform an analysis of variance (ANOVA) and calculating mean and standard deviation (SD) values for each factor and neighborhood. Textual data were organized in Excel sheets as well, including the frequency of mention of each site that interviewees recalled during the evaluations (expressed as a percentage). The results allowed the following aspects to be identified: significant differences among neighborhoods with respect to restoration capacity, neighborhoods with the highest and lowest restoration values and the sites that were more meaningful for providing a restorative experience to the inhabitants.
For the application of pre-disaster surveys oriented at determining
vulnerability to and perception of the phenomenon, stratified sampling was
conducted, with groups (strata) corresponding to 95 census blocks (Fig. 1)
(INE, 2002). Population was defined as the number of inhabitants older than
15 years of age (
The determination of post-disaster vulnerability and restoration was also
addressed by stratified sampling, where groups (strata) corresponded to nine neighborhoods (Fig. 3). Population was defined as heads of households
(male or female) who live in the town of Dichato permanently (
Data collected in these surveys were subjected to multivariate descriptive
analysis, cluster analysis and principal component analysis. Cluster
analysis allowed the similarities and dissimilarities among neighborhoods to
be easily observed, whereas principal component analysis allowed the
association of the different dimensions of vulnerability with each analyzed
neighborhood to be observed. The chi-squared test was used to compare
proportions and a one-way ANOVA was conducted for the
numerical variables. The Tukey test was applied for comparison, using a
significance level of
Inundation by tsunami risk matrix for the town of Dichato.
Risk ranges: low (1–2), medium (3–4) and high (6–9).
Risk factors were integrated into a matrix (Eckert et al., 2012; Jelínek et al., 2012; Martínez et al., 2012) and three risk levels were obtained from the multiplication: high, medium and low, with scores from 1 to 9 (Table 4). Risk level is applied to analysis units, according to pre- and post-event conditions, in the GIS vulnerability section.
Figure 3a shows the inundation area obtained from the numerical simulation.
Dots indicate inundation height measurements from field surveys (Mikami et al., 2011; Fritz et al., 2011),
while the asterisk indicates the synthetic tide
gauge location. Figure 3b shows a comparison of recorded and simulated
data, where the error obtained from Eq. (1) was
Tsunami hazard maps:
Pre-event tsunami vulnerability of the town of Dichato.
In the case of physical vulnerability, 51 % of census blocks reported high vulnerability levels, which involved 47 % of the total inundated area and 57 % of the total population (Fig. 5). Vulnerability levels relate to a deficiency in housing materials (e.g., wood, fiber cement, clay and waste). In relation to economic vulnerability, 73 % of households reported a medium level, which involves 61 % of the inundated area and 67 % of the total population. These vulnerability levels can be explained mainly by the location of the residential areas, in which more than 75 % of the inhabitants reside, where there is no overcrowding but income levels are low, with approximately 44 % of the population receiving monthly incomes of less than CLP 118 000 (about USD 170 at time of writing). In addition, 54 % of the population has high educational vulnerability. Low levels of schooling influence overall vulnerability as 42 % of the population has only a primary school education or has not completed this level and only 55 % has finished secondary school. It is important to note that 58 % of the population indicated that the tsunami was the result of the earthquake, while 42 % attributed the tsunami to other causes, sometimes related to religion, including global warming and the apocalypse.
Post-event tsunami vulnerability of the town of Dichato.
For post-disaster conditions, the reconstruction plan applied to Dichato, known as PRBC-18, modified 29 % of the total town area, with 15 % established as a tsunami-resistant building area, not including expropriation (Figs. 6 and 7). Elevated houses (stilt houses) and community buildings were designed and built in these areas (coastline). Twelve percent of the total area was reserved for mitigation parks, which were built along the coastline and river banks, where the tsunami surged and the greatest destruction was generated. Construction of the mitigation park began in 2015, with a line of trees that covered a surface area 95.7 ha, with a width of more than 20 m. The fishing area accounts for 1.6 % of the total town area, including the construction of a fishing pier and a market in Villarrica Cove.
Master plan for reconstruction for Dichato.
The post-event vulnerability measures presented variations in relation to
pre-disaster levels. The physical dimension presented an average level in
53 % of the total area, explained by the number of floors (
The ANOVA showed significant differences in the physical and educational
vulnerability dimensions (
Regarding feelings assessed regarding the possibility of a future tsunami
(Table 4), five feelings showed no significant differences by neighborhood
(
The overall vulnerability analysis for pre- and post-tsunami conditions indicates that vulnerability was reduced from high to medium levels and that the spatial distribution of vulnerable areas was maintained for both conditions (Fig. 9). For pre-tsunami conditions, 90 % of the area evaluated presented high vulnerability, 5.4 % medium vulnerability and 4.6 % low vulnerability. For post-tsunami conditions, 55 % of the area presents high vulnerability levels and 45 % exhibits medium vulnerability, while low vulnerability was not found. These findings lead to the conclusion that the entire area has high or medium levels of vulnerability.
Pre- and post-tsunami vulnerability in Dichato.
Safety perception in current residential areas was evaluated as safe or very
safe by 65 % of the local population (
The 1–7 Likert scale was applied to assess five topics, namely reconstruction,
process quality, equipment and the role of the National Emergency Office
(ONEMI). The results showed no differences among neighborhoods. Positive
evaluations were obtained for the reconstruction process (mean
Results with significant analyzed variable differences by
neighborhood. Indications are as follows: T_S (total sample),
Stat (statistic), df (degrees of freedom),
In relation to the perceived restoration study, the ANOVA showed significant
differences among neighborhoods (
Little difference was presented between the surface area (0.11 km
Tsunami risk areas for pre-event conditions.
Pre-disaster
Neighborhood B presented the greatest post-earthquake transformation, replacing a fish market with beachfront buildings, a mitigation park and a boulevard with a striking design in order to attract tourism. However, behind these buildings, a mixture of stilt houses and other types of one-story housing, made of wood or masonry, were built, giving rise to new neighborhoods and post-earthquake risk areas. Other areas, such as neighborhood A, went from being provisional neighborhoods to consolidated settlements (e.g., El Molino) and in turn absorbed part of the relocated population. In general, new post-disaster risk areas affect neighborhoods C, D, E, F, G and H up to a height of about 20 m; however, Dichato stream extends the propagation area into more inland neighborhoods.
The main results of this research were that in this urban area, with its strong reliance on natural resources (fishing) and associated tourism, high risk levels are presented for both pre- and post-disaster conditions. These conditions are not new and have already been reported in other areas in the country under the reconstruction process (Rojas et al., 2014). However, few studies exist worldwide on how coastal towns evolve in response to post-disaster reconstruction processes, generating transformations that do not contribute to risk reduction or urban resilience.
The main factors explaining high risk levels are building quality and materials, which are highly related to the degree of destruction caused by the 2010 tsunami. Many of these houses were one-story buildings made of wood or lightweight materials and built in a do-it-yourself manner. Lack of infill and reinforced concrete masonry (failure of brick masonry infill walls and lightly reinforced concrete columns) was a damage factor, coinciding with studies by Palermo et al. (2013) in the area. According to these authors, residential housing consisting of light timber frame and concrete frame construction with brick masonry infill walls suffered widespread damage throughout the surveyed coastal region of Chile.
According to numerical modeling, tsunami wave heights reached between 5 and
8 m, with current velocities greater than 2 m s
Other important factors were the socioeconomic status and education level of the population, which were relevant mainly in pre-disaster conditions. In this case, 44 % of the population has low incomes and a widespread lack of knowledge concerning emergency plans or evacuation routes, resulting in inadequate reactions. One year after the event, people still had symptoms of post-traumatic stress, indicating feelings of panic, fear and sadness (Venegas, 2011). Most of this population, which consisted of owners affected by the tsunami, moved to provisional neighborhoods where they remained for 2 years without basic services. Some of these provisional houses became final settlements. The study conducted in Dichato by Shahinoor and Kausel (2013) stated that risk of tsunamis is not well addressed in planning and community-oriented programs and that the pre-established mechanisms for post-disaster recovery are not appropriate, which is why risk is not reduced. The latter is not a specific problem of the location but derives from the lack of coordination between planning instruments and risk management in Chile, which is essentially reactive and not preventative (Martínez, 2014). However, Chile lacks a public policy oriented at establishing criteria or a reconstruction model to implement in case of a disaster and usually gives priority to physical reconstruction rather than social reconstruction. Yet physical reconstruction continues to take place in inappropriate locations and therefore, in considering only spatial location, risk areas fail to be managed in a way that effectively reduces risk.
A limitation in this research was that not all of the variables used for post-disaster conditions were similar to those used for pre-disaster conditions, for which there were census data at census block level. Even so, the results were concordant with the observed reality and the behavior of the variables coincides with other studies carried out in affected locations (Martínez, 2014; Rojas et al., 2014). The lack of standardized open-access geospatial databases prior to the disaster makes it difficult to manage risk, overcome the emergency and begin the reconstruction process. In the case of Chile, this has been critical, limiting the monitoring of post-disaster socio-territorial changes as since 2002 there has been no update to the national census, which only recently has been planned for 2017. In addition, the lack of coordination among various ministries in the country causes replication of information, which in turn is collected according to various criteria or standards, and to which access is not easy. These aspects limit continuity in risk studies. Given the high social cost of these and other threats, debate on this matter has recently begun in the country.
Regarding restoration results, it is interesting to find that the restoration capacity of neighborhoods varies with respect to the presence and absence of natural and built elements. Natural elements such as the presence of hills and views of bodies of water contribute to perceived restoration. These results are in line with previous restoration studies indicating that the presence of natural elements such as water and vegetation are related to restorative environments (Korpela and Hartig, 1996; Hartig et al., 1997). However, in this case, it is not only the mere presence of these elements that is relevant but most probably the sense of security they give to the community as well. Hills are useful for refuge in case of tsunami as well as for observation points, which are much needed to keep people informed about what is happening in the event of a disaster. Consequently, it is important for future planning processes to consider the potential of natural elements to remain in place after a disaster in order to restore communities post-disaster. For instance, access to these natural sites from different neighborhoods should be enhanced during the reconstruction process by, for example, including evacuation routes that lead to these areas for recreation purposes in everyday life. The latter would contribute to post-disaster adaptation and social resilience (Pelling, 2003).
In addition, the restoration results show that new elements introduced during the reconstruction process such as the coastal infrastructure for mitigation and anti-tsunami housing are also characteristic of neighborhoods that provide restorative experiences (Khew et al., 2015). It is possible that these elements, although they are built features, give a certain sense of security to respondents, which could explain these results. This study did not focus on establishing relationships between perceived safety and post-disaster restoration factors; however, it is highly recommended that this possible relationship be expanded in future studies. It may be that post-disaster restorative experiences are found in new built and mitigation sites that give security to the community. This would also be important to consider in the reconstruction process, as built features of the kind described here play a role not only in mitigation but also in post-disaster community functioning, contributing to social resilience (Pelling, 2003).
In this sense, vulnerability and resilience are distinct elements but superimposed in their role in natural disasters and brought together in the cornerstone of sustainability (Turner, 2010). In the case of Dichato, vulnerability exhibited a close relationship to lack of resilience because few lessons were learned from the 2010 event and the same mistakes are still being made, with a nearly completed rebuilding process that presents vulnerability conditions very similar to those that existed prior to the 2010 earthquake. This situation is explained by the emphasis on physical rather than social reconstruction, the lack of public policies to face a rebuilding process of this magnitude despite recurring events in the country and, especially, by the poor consideration and assessment of risk areas in planning at a local scale, since other affected areas were repopulated in the same manner and relocated to the same risk areas (Martínez, 2014). In some neighborhoods, increased social and environmental problems such as pollution, crime and poverty occurred as a result of reconstruction processes (Rojas et al., 2014). The main disadvantage of these programs is that they were implemented as similar projects in 18 affected coastal towns, regardless of geographic reality and territorial identity. In addition, the programs did not distinguish between rural or urban areas. Small fishermen's coves located in coastal wetlands and small bays that underwent semi-urbanization processes had to absorb relocated populations from affected areas, resulting in increased population densities in new risk areas and loss of cultural and territorial identity. The latter was reflected in that between 57 and 75 % of the population most affected by the tsunami 6 years ago identified with Dichato prior to the tsunami. In this respect, most current approaches establish that resilience is characterized by socio-ecological system responses to natural disturbances, capacity for self-organization, learning and adaptation to change (Folke, 2006; Turner, 2010). These elements present a challenge from an institutional point of view in Chile, as risk management and its link to organized society must be strengthened to ensure that investments in reconstruction processes effectively reduce risks associated with phenomena that will undoubtedly continue to occur in the country. In addition, reconstruction involves addressing physical, social and environmental territory components to facilitate the development of post-disaster resilience, meaning that the country must change its approach to natural disaster management, moving towards sustainability of its cities and coastal towns.
The vulnerability factors that best explained the extent of the 2010 tsunami disaster were housing materials, low incomes and poor knowledge about such phenomena, which conditioned inadequate reactions at the time of emergency. The current town configuration, resulting from the reconstruction process in the 6 years after the event, has generated new risk areas that occupy the same locations as in pre-event conditions, meaning that risk is not reduced. For post-earthquake conditions, it was determined that all neighborhoods have the potential to be restorative environments after a tsunami but with different intensities, depending on the type of natural and built features they have kept and incorporated during the reconstruction process. However, risk analysis indicates that neighborhoods with greater post-disaster restoration ability remain in the same areas devastated by the 2010 tsunami and will likely be destroyed again in a future event, a situation that forces us to reflect on how to plan coastal zone land use and manage risk in the country. Finally, the methods used in this investigation, in which cluster analysis, principal component analysis and a GIS platform are integrated, prove highly beneficial in an interdisciplinary approach such as this, where the selection criterion of the variables is fitted to the knowledge of the place and the variables are validated as a function of the local reality, which facilitates their adaptation to other geographical areas of interest.
No data sets were used in this article.
The authors declare that they have no conflict of interest.
Grant funding came from FONDECYT nos. 1151367, 11140424, 1150137 and FONDAP no. 15110017 from the National Scientific and Technological Research Commission (CONICYT) of Chile and from DID I-2014-11 from the Universidad Austral de Chile. The authors thank the referees (A. Armigliato) for the comments on the manuscript. The authors thank Terry Hartig for his valuable advice on the application of the perceived restorativeness scale. Edited by: M. Gonzalez Reviewed by: A. Armigliato and one anonymous referee