Risk perception of local stakeholders on natural hazards: implications for theory and practice

In Romania, local stakeholders’ knowledge plays a decisional role in emergencies, supporting rescue officers in natural hazard events, coordinating and assisting, both physically and psychologically, the affected populations. However, despite in Iași Metropolitan area (NE of Romania), the occurrence and severity of natural hazards are increasing there is a lack 10 of knowledge of local stakeholders to address the population toward safety actions. For this reason, 118 local stakeholders were interviewed to determine their risk awareness and preparedness capacities over a set of natural hazards to understand where the lack of knowledge, action, and trust are exacerbated the most. Results reveal substantial distinctions among stakeholders and the different threats based on their cognitive and behavioral roles in the communities. The role of responsibility and trust has been seen as important driving factors shaping their perception and preparedness. Preparedness 15 levels were low, and, not for all, learning and preparatory actions are needed to withstand the negative occurrences of natural hazards. As their role is to refer with direct interventions in affected areas managing communication initiatives with the entire population of the community, there is the need to create stakeholders’ networks, empowering local actors that could serve as a bridge between authorities’ decisions and local people in order to make effective risk management plans and secure more lives and economies. 20

communes that in only 11 years (between 2007 and 2017) had exceptional increased the number of inhabitants (Valea Lupului,102.8%;Miroslava,93,4%) or new constructions (Miroslava,164,8%;Valea Lupului,141.4%). The same trend is highlighted by the number of building permits issued in 2017, which in some cases (Miroslava and Valea 135 Lupului) exceed that of the main urban center. A new peri-urban area is developing spontaneously around Iaşi City, which is https://doi.org/10.5194/nhess-2021-37 Preprint. Discussion started: 4 February 2021 c Author(s) 2021. CC BY 4.0 License. demolished dwellings, 34,582 homeless families, 763 industrial units affected, and many other damages in all sectors of the economy (Georgescu and Pomonis, 2008). Although located relatively far from the epicentral zone, Iaşi county was the most 205 affected in Romania in terms of percentage of dwelling stock affected: 47% was affected, from which 11% destroyed, 13% of dwellings requiring strengthening, and 23% dwellings requiring repair (Georgescu and Pomonis, 2008). In the last decades, earthquakes of over 6Mw were those from 1986, 1990, and 2004 and minor damages were reported.

Climate trends in NE Romania
In the 1900-2005 period, mean annual temperature in Northeastern Romania has increased by around 0.2 and 0.3° C (Haylock 210 et al., 2008;Kurnik et al., 2017), while for the 1961-2007 period, the trend of the increase is between 1 and 1. 2° C (Busuioc et al., 2010). The current and future climate changes trends, and effects for Romania are not very well studied, and the existing results based on observational and modeling data are very often contradictory (Busuioc et al., 1997;Cuculeanu et al., 2002;Busuioc et al., 2010Busuioc et al., , 2013Croitoru and Minea, 2015;Croitoru et al., 2016), this being the motive that the European level downscaled scenarios need to be taken into account. Currently, CORDEX (COordinated Regional climate Downscaling 215 EXperiment) framework is used for European regional forecasting at a 12.5 km resolution and for the RCP4.5 and RCP8.5 emission scenarios (Moss et al., 2010;van Vuuren et al., 2011), through the EURO-CORDEX initiative (Jacob et al., 2014).
The interpretation of modeling on a continental scale shows for the study area a change of 2071-2100 period temperature compared to 1971-2000 period, of 2 to 5° C for mean annual, summer, and winter values (Jacob et al., 2014;Kurnik et al., 2017). The historical climate data show for Northeastern Romania an increase of annual mean values with up to 20% and a 220 decrease of summer precipitations with up to 5% (Haylock et al., 2008;Kurnik et al. 2017), although strong spatial variability is shown (Croitoru and Minea, 2015;Croitoru et al., 2016). The forecasts show a further continuation of these trends (Jacob et al., 2014;Kurnik et al., 2017). Climate change driven by anthropogenic emissions is expected to increase precipitation extremes in both wet and dry regions as it happened in the historical period, although the intensity cannot be predicted (Donat et al., 2016;Donat et al., 2017;Ingram, 2016). Anyway, the fact that the precipitation intensity will increase should be enough 225 to alarm the authorities and the citizens (Ingram, 2016). In NE Romania, the forecasts are that precipitation extremes will increase (Jacob et al., 2014;Kurnik et al., 2017), continuing the trend of the historical data (Croitoru et al., 2016). Heatwaves are expected to be more frequent and more intense considering the increase in temperature (Velea and Bojariu, 2018). The historical trends of droughts in NE Romania are of increasing frequency but decreasing magnitude (Minea andCroitoru, 2015, 2017;Minea et al., 2016;Spinoni et al., 2015), while the forecast is of slight increase (Stagge et al., 2015). North Atlantic 230 Circulation has a delayed effect on the spring flow (Bîrsan, 2017) in NE Romania, which will continue to remain a future trend, while the runoff should decrease (especially in summer), continuing the historical trends (Stahl et al., 2012;Croitoru and Minea, 2015). These assumptions are based on the upward trend of precipitation and evapotranspiration due to increasing temperatures (Cuculeanu and Bălteanu, 2004). The minimum discharge will decrease, and the water deficits will increase (Forzieri et al., 2014). The flood magnitude instead will increase in NE Romania (Alfieri et al., 2015;Reker et al., 2017), so 235 probably the number of deaths in Romania will continue to be one of the biggest in Europe (Vanneuville et al., 2017).

Data collection and methods
Local stakeholders have been selected representing different characteristics in terms of power, legitimacy, and urgency (Mitchell et al., 1997;Mainardes et al., 2012). Further, the dominant stakeholders (mayors, police officers), discretionary stakeholders (farmers), and dormant stakeholders (professors and priests) have been selected. Semi-structured in-depth 240 interviews have been run from March 2017 until October 2018 involving 118 people: 23 mayors, 27 farmers, 25 priests, 21 police chiefs, and 22 school heads. (Fig. 1). As in many other countries, in Romania, public institutions are organized at administrative levels, village/town halls, schools, police headquarters. The leaders of these institutions (mayors, police chiefs and school heads, and in few cases, their deputies) were recruited directly to participate in the present study. Priests and local entrepreneurs (farmers) were randomly selected and interviewed on-site. 245 The questionnaire was organized into two parts: the first with pre-defined questions regarding the assessment of risk perception induced by natural hazards: level of threat, personal experience, level of knowledge, level of preparedness, risk management, communication, and trust (Table A1, Appendix) and a second part in which discussions have been focussed on environmental and hazardous phenomena that threaten the places where they live and work. Interviews were run from 30 to 50 minutes according to the desire of the participant to expand the open questions with his/her personal experience. In most of the cases, 250 there were constructive discussions, some stakeholders inviting other members of the community (especially the mayors) into the dialogues considering it an enriching approach for the community.
There is a clear gender imbalance in the sample of stakeholders considered for the interviews (Fig. 2). This is due to the specificity of certain professions in Romania (priests are exclusively men, while police officers predominantly) or the perpetuation of older mentalities regarding the occupation of positions at the top of public administration (the case of mayors 255 100 % men). Only for school heads, we found a balanced situation: 63% were women. The majority of the stakeholders have a university degree, being a mandatory requirement for school heads, priests, and police officers. A large proportion of stakeholders (88%) live in the area where they work (same community or neighborhood communities), and this could suggest an amplification of perception of high-probability risks and reducing low-probability ones (Bernardo, 2013). The age distribution is skewed toward older persons, especially in the case of mayors (mean age 53.6 years) or school heads (49.2 260 years) in contrast with a younger age of policemen (39.4 years).
To test some assumptions, we formulated the following questions: Q1: Is there a dependency relationship between the threats of different natural hazards?
Although the selected stakeholders have different roles within the communities and a different timing in the evolution and management of the events related to natural hazards, all of them bear extra responsibility (legislative, educational, 265 communicational, and moral) compared to the lay public. In this sense, we stated the second hypothesis.
Q2: Do different stakeholders have different perceptions and preparedness level according to a set of natural hazards? Since Iaşi Metropolitan Area is situated in two main geomorphological settings (hilly areas and floodplains) and during the last decades, there have been registered localized hazards (such as landslides in hilly areas and floods in floodplains), this factor could influence the risk perception. As a consequence, another question was formulated. 270 Q3: Do geographical and topographical characteristics of locations affect stakeholder's risk perception of different natural hazards?

Statistical analysis 275
Data coding was performed using a tabular data application (Open Office Calc) by assigning codes from 1 to 5 for the Likert scale data and from 0 to1 for dichotomous responses. The continuous variables were coded using numbers. After the coding, the raw data was exported to R stat (R Core Team, 2018), where the data was manipulated to obtain the format required by the specific functions used to analyze them. The statistical analysis was performed in three main steps (Openheim, 2000): (i) first, the univariate analysis was performed by plotting on the Likert or arithmetic scales the sample distributions in order 280 to have a first overview (descriptive statistics) of the data. Also, we have chosen to comply with the standard statistical assumptions (especially regarding the failure of parametric statistics in the case of extreme values of ordinal data and unequal interval scales, Baker et al., 1966, Armstrong, 1981 and use both univariate and bivariate analysis with graphical analysis to provide a more in-depth analysis (Knapp, 1990;Mircioiu and Atkinson, 2017). Also, we avoided considering Likert data as nominal categories since the ordering will be lost (Agresti, 2010;Mangiafico, 2016). 285 (ii) secondly, the bivariate analysis consisted of computing the cross-tabulation and various independence and association measures between the variables. First of all, it has been tested the independence of the responses toward the risks involved in the study for the stakeholders' categories and, after that, association tests for the assessment of the significance of stakeholder type and other categorical variables. Kruskal-Wallis rank-sum test was used to assess if there are differences in the responses (Magnifiaco, 2016) for every category of risks and natural risks and decedent type, village, commune, flooded or non-flooded, 290 age, gender, education. When the dependence exists (the null hypothesis is rejected), the statistic Freeman's epsilon-squared was used to assess the association's strength between one ordinal variable and one nominal variable (Mangiafico, 2016). This statistic ranges from 0 to 1, with 0 indicating no association and 1 indicating perfect association. Values bigger than 0.5 were regarded in our case as a measure of powerful association in the presence of dependence. This association measure was computed using the epsilonSquared() function from the rcompanion R package (Mangiafico, 2016). 295 (iii) finally, we applied a multivariate method, correspondence analysis for those questions and risks that were found conclusive in the bivariate analysis step.
Besides, CA (Correspondence Analysis) graphical methods have been applied for exploring the relationships between variables in contingency tables (Greenacre, 2007). The method's theory is straightforward, based on the singular value decomposition of the contingency table's matrix data structure. We have chosen this method because it describes our data graphically in terms 300 of showing the differences between stakeholder types or other categorical variables, especially for those with big Freeman's epsilon-squared values. The Likert scale with the answer to the question is considered the dependent variable, and the variants of the response or the categories of stakeholders or other associated categorical data (flooded or non-flooded communes) are the independent data. The column variables (e.g., stakeholder type) are displayed as oriented vectors, while the Likert scale counts are displayed as dots. The orientation of the stakeholder type vector toward one of the axes shows its contribution to 305 that axis's variance. If the angle between the vector and the lines is 45°, then the contributions to the two axes are the same, while if the angle is smaller toward a certain axis, the greater the contribution to the variance of that axis is. The length of the arrow vectors is proportional to their contribution to the two-dimensional solution. Since we have an ordered variable, and the distances between the categories are not the same, there is no logic to take into account the distances along the axes of the CA plot and to make comparisons (although this type of plot allow this, in the sense that the axes are scaled to a common scale).

The level of threat
The first question addressed to the interviewees was designed to assess the main socio-economic and environmental factors which could affect the communities' quality of life. The majority of stakeholders (61%) consider that the level of development is the main factor that can threaten the quality of life in their territory (Fig. 3). It follows the risks induced by natural hazards 315 (57% of responses), climate change (40%), criminality (37%), environmental pollution (27%), and technological risks (8%). Generally, the stakeholders that participated to the present survey consider droughts as the most threatening natural hazard both for their communities but also for them personally (Fig. 4). 320 Water scarcity is a direct consequence of the continental climate of the region that affected the agricultural economy of North-Eastern Romania for centuries (Mărgărint and Niculiţă, 2016;Niculiţă et al., 2020). Many stakeholders reported a drastic reduction in the number of cattle, which, in the driest years, can reach 80% of the total animals of the households in the villages.
"There are ten years since I had serious problems every year. I achieved a special car-tanker to get water for livestock. And very little remains for vegetable crops. I get water from the reservoir (5 kilometers away), and I don't know what will happen 325 when it disappears." (farmer, 35 years, managing 300 hectares of agricultural land and 35 cows. They also consider that this hazard will affect their communities for many years from now. Alongside the dramatic reduction of agricultural production, the most dangerous problems occur regarding livestock).
Earthquakes represent the second threatening hazard. The memory of the 1977 earthquake, when Iaşi County registered the highest number of buildings affected in Romania (Georgescu and Pomonis, 2008) is still vivid in the memory of many 330 stakeholders. Although the norms in constructions were strongly upgraded after this event, after 1989, the discipline in buildings decreased suddenly due to the lack of legislation. How many dwellings have been built up in the last years is not far from the knowledge of the interviewees and, from this point of view, many raised serious questions regarding the resistance of the new constructions. "Many who bought new homes think they are new and strong, but at the next big earthquake, they will find that they were built just to be sold." (mayor, 58 years, personally affected by the 1977 earthquake). The population's 335 level of dissatisfaction is constantly increasing concerning public works, transportation, and the environment. Considering that any significant event did not trigger these permanent stressors, the real situation of risks associated with natural hazards can be much more profound, almost unknown to many of the inhabitants and their leaders. A middle position is occupied by the hazards which registered a higher frequency: rainstorms and snowstorms had a growing trend in the last decade in the study area and, as a consequence, their impact on communities is quite essential. During the year, the strongest storms occur in late spring and summer. In some cases accompanied by hail, the most significant damages 345 are recorded in agriculture and in newly built areas that do not have an adequate drainage infrastructure. When they have a large area of development, they can affect the transports, trigger soil erosion, and lead to the increase of the lower order hydrographic network flows, leading to the destruction of the bridges, the siltation of the canals, etc. These issues were invoked as the most pressing by farmers and mayors, and police chiefs. "I am here for few years. In the center of the locality, there are no problems, there is asphalt on the street, but towards the valley, those who have moved to the house in the last four years 350 live a nightmare every time it rains. The road is muddy and becomes impassable." (a police officer in a settlement with many new dwellings, 34 years).
Climate-related hazards that have a relatively low temporal frequency, like floods, landslides, and soil erosion, are perceived as imposing a low threat, in general. The landslide risk is high in hilly regions of NE Romania (Micu et al., 2017, Mărgărint andNiculiţă, 2017). In the last century, one of the most significant events inside the settlements took place 50 years ago in a 355 succession of years with high precipitations (Pujină, 2008). With few exceptions, the memory of those events seems to erase.
But the risk is still high, and people will face again with landslide reactivations in the years with the same increased pattern of precipitations (Niculiţă, 2020). There is a lack of prevention behavior in terms of recent expansions of built areas due to several factors: investors' desire to build and sell, lack of knowledge and awareness of the danger of those who buy, and those who should take decisions regarding the expansion of built-up areas. "In our commune, the landslide risk has been solved: we have 360 the study regarding landslide hazard and risk in an updated form, so we are in line with the legislation." (mayor of a commune affected by landslides in 1969-1972, 66 years).
The outputs of The Kruskal-Wallis rank-sum test and Freeman's epsilon-squared statistics show correlations among every category of natural risks and a set of socio-economic and geographic variables (for further results, see Table A3 of the Appendix). The most significant differences are in stakeholder type (answering the second research question), gender, age, 365 and spatial localization, and geomorphological context. At the same time, education does not influence the response. The results indicate that the risk perception is dependent on stakeholder types, which partially confirms the third research hypothesis. In addition, it has been found that the age of the respondents is an essential factor regarding certain risks (Table   A2, Appendix) because some of them might be born after certain important hazard events such as the 1977 earthquake, 43 years ago, or the landslides events such as those between the '70 and the '80 (Niculiță et al., 2017(Niculiță et al., , 2018. For floods, climatic 370 hazards, and soil erosion, it seems that younger respondents are more aware. The CA contribution biplot for Question 1 from Fig.5, is showing the correspondence between the perceived role of natural hazards as threats to the local community by different stakeholder type, considering the first two dimensions, that sum 96,8% of the variance. Police chiefs and priests who perceive natural hazards as low and medium threats, mayors and farmers perceive them as high threats, and school heads that perceive them as high threats. The explanation of the low perception of hazards as threats to the community's quality of life in the case of priests and police chiefs is given by their relative low knowledge of natural hazards, given by their profession. School heads, mayors, and farmers have a high level of awareness associated with the threats for the quality of life of the following factors: level of development (91% of school heads), natural risks (82% of school heads and 81 % of farmers) and climatic change (78% of farmers, 55% of school heads). The exception is related to technological risks, given the predominant rural background of the communities. Priests and police chiefs, in general, 380 expressed a low level of perception regarding the threats to local communities, with some exceptions: e.g., police chiefs regarding criminality, which is their duty (the same threat is seen by school heads, in association with their high level of childcare). The highest values of the perceived threat associated with droughts have been registered in the case of school heads (95%) and farmers (93%) who expressed a great concern compared to the other stakeholders. Also, the earthquakes are seen as a significant threat by school heads (77%), farmers (56%), and priests (52%). By interpreting the enlarged discussions during the interview, this could be considered as a consequence of still lively memories of the 1977Vranceaearthquake (Armaş, 2006), 390 a social trauma of Romanian people, but also to present-day other factors: (i) a high vulnerability characterizes the majority of institutional buildings (especially schools and churches) to earthquakes (Mosoarca and Gioncu, 2013;Albulescu et al., 2020) and (ii) the frequent exercises for the improvement of the earthquake preparedness (in schools usually these exercises take place annually). The problem of the vulnerability of old buildings in Romania represents a constant public and scientific debate (Armaş, 2012;Banica et al., 2017) and, in this sense, we also raise on this occasion an alarm signal regarding the need for 395 essential investments in the modernization of public spaces in urban and rural areas in Romania.
From these general results, significant differences have been recorded among the two geomorphological types of the administrative units ( Fig. 1 and Fig. 6): floodplain administrative units (FAU) and hilly administrative units (HAU).
The results highlight that stakeholders have different levels of perception related to different hazards, according to the main past events that have been recorded in the last decades: in the floodplain administrative units (FAU in Fig. 6) there is a 400 significantly higher degree of awareness concerning flood risk and possible threats, while in the hilly administrative units (HAU) the level of threat associated to landslides and soil erosion is higher than in the FAU.

405
Again, droughts are the most life-changing natural hazards with the highest likelihood of occurrence. Rainstorms, snowstorms, and earthquakes follow them. A lower level of probability was assigned to soil erosion, landslides, and floods (Fig. 7). But here, there are important differences, depending on the geomorphological type of the locality. The stakeholders who come from floodplain settlements have indicated a higher probability for floods than the others (HAU stakeholders) and a lower probability for landslides and soil erosion. 410 The main geomorphological characteristics which can influence different hazardous processes and the distance to the potential risk areas constitute important factors of how different people perceive different risks (Bickerstaff and Walker, 2001;Heitz et al., 2009;Gao et al., 2020). Some natural hazards affect large areas (droughts, earthquakes, or snowstorms), while others (e.g., landslides, floods) are spatially concentrated in direct relation to topography characteristics at the local scale. From this point of view, the settlements from the study area, as part of the Moldavian Plateau, have been constantly affected by landslides and 415 floods (Văculişteanu et al., 2019), and their consequences are found in the answers given by the interviewees. Table A1 of the Appendix (Q2 and column 6) and Fig. 6 shows that the geomorphological context of the area where the stakeholder works is important in its perception regarding floods and landslide risk. These results are seen in the context of a social trauma of the inhabitants managed by the stakeholders during the evacuations of some settlements along Prut Valley in 2008 and 2010. Due to the risk of flooding of the inhabited areas, in July 2008, over 3000 inhabitants from Iaşi County, including Victoria, Ungheni, 420 and Ţuţora ATU3 (Fig. 1), were evacuated (Ziarul de Iaşi, 2008). Concerning the likelihood of occurrence of natural hazards (the fifth question), some types of natural hazards are perceived to increase in the near future, especially climatic induced hazards: droughts (86%), rainstorms (68%), and snowstorms (64%). 425 Landslides and soil erosion are perceived as not increasing, while for earthquakes, the results are balanced.

Personal experience and knowledge
Personal experience is one of the most critical factors influencing risk perception (Weber, 2006;Van der Linden, 2014;Knuth et al., 2015;Öhman, 2017). The study participants indicated that they were affected mainly by droughts, rainstorms, and snowstorms, with farmers bearing the major costs (Fig. 8). A large proportion of them was affected by droughts (93%), 430 rainstorms (78%), snowstorms, and soil erosion (48%). Stakeholders are affected by natural hazards according to their activities and responsibilities in their daily life, exposing them to different vulnerabilities. The other stakeholders were affected in a smaller measure by soil erosion. This process can generally pose problems only to 435 those who directly connect with the land, which affects lesser the build-up areas. It is shown that experience is higher with age, especially for the analysis with the earthquake occurrence. These are disasters that, for their high magnitude, can be impressed vividly in people's memory. Their role in disaster risk management and coordination allows them to remember the most significant events they served the community. In contrast, slow onset events (e.g., droughts or soil erosion) can disappear quickly. 440 The knowledge of participants about natural hazards has been asked through several sub-questions. Stakeholders get information differently about the probability of occurrence and the severity of these events. The majority get information from the TV/radio (82%), friends/family and community peers (60%), and social networks on the internet (53%). The more official channels are the least represented with national information initiatives (47%), school (44%), local administration (41%), and volunteer associations (40%). Looking at the triggering factors of those events, stakeholders mentioned all sub-sections from 445 the questionnaire (Table A1, Appendix) that they consider having an important influence on the negative impact of natural hazards. Some exceptions have been registered for 57% of mayors who responded that uncontrolled urbanization and unmanaged land use planning are not influencing the occurrence of any hazard. Local administration is controlling the land use planning, and, in any case, this might be the cause of negative consequences derived by climate extremes and geological movements. 450 The majority of priests and mayors do not consider that climate change can exacerbate the negative consequences of natural hazards (56% and 22% of them indicated "low" and "very low" respectively. Among the solutions to avoid the negative consequences of natural hazards, results indicated a uniform answer among all stakeholders, except the compensation scheme for the victims especially marked from mayors. Financial compensation schemes represent a particularly neuralgic issue in the post-communist society of Romania. Many interviewees highlighted that these compensations could be an encouragement of 455 non-compliance with the law, especially regarding unauthorized constructions on lands at risk of floods and landslides.

The level of preparedness
The level of preparedness was investigated individually, and regarding the community, they belong. Overall, the results indicate a low level of preparedness in the case of all the natural hazards discussed. The lowest ranks were given to soil erosion (64%), droughts (58%), earthquakes and landslides (55%), floods (52%), rainstorms (50%), and snowstorms (35%). It seems 460 that, despite a low level of readiness, stakeholders feel a bit more prepared to withstand the consequences of storms and floods.
Snowstorms affect the communities in winter (and exceptionally in spring, the case of April 2018), and agriculture do not suffer. Life in rural areas can be more comfortable compared with urban areas. In Romania, after the recent intense snowstorms such as those from January 2008 (Georgescu et al., 2009) or January-February 2012(Bălteanu et al., 2013, rural settlements have been endowed with specialized equipment in rapid intervention, especially in the case of roads, and these endowments 465 seem to improve the respondents' concerns. Similarly, the existing embankments along rivers (Prut, Jijia, and Bahlui) have often been invoked during discussions as ensuring a relatively good level of protection, especially of built-up areas. The lower level of preparedness is associated with soil erosion and landslides, for which many stakeholders declared their lack of knowledge concerning the processes themselves and related protective measures. The results of the survey made us accept the second hypothesis, which states that the level of 470 preparedness depends on the risk type.
The same pattern of the answers has been registered in the case of the assessment of the preparedness level of the communities.
However, preparedness was low, and stakeholders affirmed strongly that by good training and knowledge of natural hazards occurrence and mitigation practices, their and community preparedness could increase. Question 11 (How much do you think that your personal knowledge might increase the level of preparedness of your community? Table A1 of the Appendix) reveal 475 significant differences among stakeholders; while in the case of school heads, "high" and "very high" responses reached 95%, for police chiefs the percentage of the same responses dropped to 14%. Intermediate values have been recorded for the other stakeholders: "high" and "very high" answers were given by 67% of farmers, 56% of priests, and 39% of mayors. Police chiefs and mayors are responsible for risk management during an emergency, and for them, preparedness is at the base of the training.
For this reason, they might think that their role is the management of situations and, in any case, is the responsibility of 480 individuals. School heads who have the obligation of small infants feel that individual preparedness is the key to successful disaster management, evacuation, and recovery. In this regard, participation in simulation evacuations is a crucial step for a positive disaster outcome. Most of the stakeholders declared that they had participated, especially in the simulations concerning earthquakes, and few of them indicated other specific hazards (e.g., fires). Seventy-two of stakeholders (61%) declared that they participated in simulations in the last years, most of them to earthquake simulations (especially school heads and mayors). 485 Stakeholders from floodplains communes stated participation in flood simulations. In a particular case (Aroneanu settlement, located close to Iaşi International Airport), stakeholders participated in a technological disaster exercise (aircraft crash). The period elapsed since the last simulation varies from few months to over ten years, the most recent being mostly declared by the school heads.
Some of the most representative CA biplots represent the position of stakeholder types in the case of preparedness to cope 490 with different types of natural hazards ( Fig. 9 and 10).
The same differentiated pattern of the stakeholder responses was recorded in the case of the level of their communities' preparedness.

Risk management, trust, and communication
Several factors have been listed (Fig. 11) and discussed as representing long term solutions to improve current risk management 495 plans.
Most of the participants agreed with all the items proposed. On the other side, priests seemed to be the most pessimistic, especially in terms of predictability, people's preparedness, intervention, and recovery capacity. Again, the role of trust in depicting a negative situation in which stakeholders evidenced low trust on mitigation and management measures (Fig. 12).
As mayors followed the same trend, it is plausible to think that they delegate the responsibility during emergencies to other 500 institutions, imputing ineffective planning and organization. Question 16 ("In your judgment, how much are the opinions of the following actors taken into account in the decisions about 505 measures to adopt for preventing or reducing damage from natural hazards phenomena?") presents a grouping of "high" and "very high" responses around 70% for followings sub-sections: local communities, technicians/engineers, elective representatives at local and national levels. A lower percentage (34% of "high" and "very high" responses) has been registered for the sub-section "environmental organizations." Among stakeholder types, we should highlight the higher percentages of "low" and "very low" responses in the following cases: priests for "elective representatives at the local level" (16%) and 510 "technicians/engineers" (16%), school heads (50%) and mayors (43%) for "environmental organizations," farmers for "local communities" (16%), and "state elective representatives" (26%).

515
The stakeholders' role as leaders of their institution during the events generated by natural hazards is critical. They refer to direct intervention in the affected areas and the management and communication with the entire population of the community.
These issues were addressed in the following question. The gathered answers are generally in line with the level of social responsibility of the institutions that stakeholders represent according to the legislation but also to the moral leadership in the community. "high" and "very high" responses were acquired as follows: priests (88%), police chiefs (86%), mayors (74%), 520 school heads (64%), and farmers (52%). There are interesting absences of "low" and "very low" responses in the case of mayors, school heads, and priests, and the low proportion of these responses in the case of police chiefs (5%) and farmers (7%).

Discussions and conclusions
The current study's importance lies in the intrinsic characteristics of Iaşi area, being exposed and vulnerable to major natural 525 hazards and overlapped with recent and historical contradictory socio-economic dynamics of Romania (Ignat et al., 2014). In line with a competitive European economy with increasing educational level and income of the last 20 years, the Romanian society tried to follow the positive trends and numbers, with a rapid urban sprawl. The fast development was characterized by a lack of planning and infrastructural investments leading to an increased vulnerability to natural hazards. At the same time, the dissatisfaction and the feeling of the danger of people were felt even at the political level that, since 1989, has led to a 530 constant decrease of trust in national institutions and their leaders. In this fragile socio-economic and political environment, local stakeholders were involved in national programs to help communities (primarily rural areas) to prevent, manage and recover from emergencies, including weather extremes or natural hazards, because, very often, media, politicians or other public actors demonstrated to discredit these phenomena and their potential negative impact. However, history showed that disaster communication was poorly managed, and local stakeholders lacked in coordinating people in all phases of risk management. The lacking knowledge and preparedness understanding of stakeholders pushed the need to investigate their actual perception of natural hazards occurrence to set the scene for improved management at the local level. The results found with 118 interviews in Iaşi Metropolitan Area showed that, in general, there is a moderate level of threat 540 toward the negative influence of climate-related hazards and earthquakes with different levels. The three main themes that are resumed in the questions posed (Q1, Q2, and Q3) reveals differences in risk perception concerning various stakeholders' types and risks, and an obvious specific behavior related to the local geomorphological settings which favor local scale hazards (e.g., landslides and floods). Farmers are more concerned, especially to climate-related hazards, that can directly affect their livelihood and source of income. The literature has found that they might already receive incentives to protect the economic The perspectives of this study should be continued in the next years to assess the changes of the behavior of the stakeholders regarding the awareness of the threats posed by natural hazards induced risks in a dynamic perspective, taking into consideration the future events and their negative effects as well as the changes that the citizens will register at the level of 575 increasing (or not) the inter-community cooperation and the compliance with legislation.

Author contribution
MCM, MN, GR and PT designed the conceptualization, MCM and GR the questionnaire and MCM and MN carried it out.
MN carried the statistical analysis and the plotting. MCM prepared the manuscript with contributions from all co-authors.