1 Introduction

More than 326,000 deaths have been reported from floods worldwide between 2000 and 2018, coupled with global economic losses of more than USD 1.7 trillion (Perera et al. 2019). Flooding and other weather-related hazards are expected to shift under climate change (Jongman et al. 2015), with flood-related damages rising (Hallegatte et al. 2013). Flooding causes devastating impacts on local communities, resulting in damage to homes, businesses, and infrastructure (Azevedo de Almeida and Mostafavi 2016). Despite this, evidence suggests that many countries do not prioritise water security (Brown et al. 2013). Australia is exposed to a wide range of natural hazards that include flooding, droughts, cyclones, and bushfires – impacts that have grown in recent years (Hein et al. 2019). These hazards inflict direct and indirect damage to communities (Bhattacharya-Mis et al. 2018), impairing national economic development (Brown et al. 2013). Flooding in Australia is among the top three most costly hazards (Ladds et al. 2017).

A series of flood disasters in 2021 and 2022 along Australia’s eastern coast brought to light ongoing challenges in managing flood risk. In the Greater Sydney region, home to more than 5.2 million people (Australian Bureau of Statistics 2021), at least 18,000 people were forced to evacuate after the Hawkesbury and Nepean Rivers flooded in March 2021 (Elsworthy 2021). Coastal New South Wales, including Sydney, saw the wettest week since records began in 1900 during the March 2021 floods (Australian Bureau of Meteorology 2021), only for this record to be broken less than a year later in March 2022 (Australian Bureau of Meteorology 2022). While there was much focus on inadequate planning and response mechanisms (AFAC 2021; NSW Government 2022), there has been comparatively less attention given to understanding how people’s perceptions of flood risk have shaped these and other flood disasters in the Greater Sydney region.

A vast literature on flood risk perceptions has developed robust theory of the links between risk perception and personal characteristics, such as age, education, and experience (Tierney 1999; Plapp 2004; Botzen et al. 2009; Lechowska 2018, 2022). Despite knowing that personal factors shape how individuals perceive flood risk (Wachinger et al. 2013; Mills et al. 2016; Wang et al. 2018), previous research has rarely focused on youth. Further understanding the factors that mould youth flood risk perceptions is critical as they have the most to gain from disaster risk reduction efforts. In the context of this research, we adopt the United Nations Office for Disaster Risk Reduction (UNDRR 2020) classification of youth ranging from ages 15 to 30, which is inclusive of young adults. This definition expands upon the more common youth definition used by the United Nations (ages 15 to 24) (United Nations 1981). Youth may benefit from a better understanding of their susceptibility to floods and how to prepare (Rufat et al. 2015; Zhong et al. 2021).

According to the World Bank (2020), it is estimated that there are more than 223 million university students globally. In Australia, there are more than 641,000 students enrolled in all Higher Education Institutions (Department of Education 2020). Despite the large number of university students in Australia, there been lacking research on students’ perceptions of flood risk and ability to understand commonly communicated flood terms. The objective of the research thus sought to assess the factors affecting university students’ perceptions and comprehension of flood risk in the Greater Sydney region, shown in Fig. 1.

Fig. 1
figure 1

Overview of research objectives

Full size image

We aimed to answer the following research questions:

  1. 1.

    How are social-demographic characteristics, preparedness, knowledge, and experience associated with university students’ flood risk perceptions?

  2. 2.

    How do university students understand different flood probability terms?

2 Background

In Australia, urbanisation and climate change are driving increased disaster risk (Masud et al. 2019). There are a growing number of studies exploring flood resilience and risk perception in light of these trends. As floods become more intense due to climate change (Kundzewicz and Schellnhuber 2004; Neumayer et al. 2014), there is a need to examine current and future risk perceptions to shape preventative actions (Kellens et al. 2013). Bubeck et al. (2012) and Hood and Jones (2003) define perceived risk as a product of the “likelihood” and “result” of occurrence. Risk perception comprises multiple dimensions (Holtgrave and Weber 1993). Wachinger et al. (2013) propose a separation of flood risk perception into four sets of influence factors: (1) risk factors (probability and frequency of hazardous events perceived or experienced), (2) informational factors (media attention, a source of accurate information, and risk management specialists’ participation), (3) personal factors (educational, occupation, age, gender, familiarity with disasters, trust in expert knowledge and past flood experiences), and (4) contextual factors (economic, vulnerability, house category, nation, region of residence, proximity to the water, and size of the community). These four factors provide a comprehensive view of how people perceive and respond to flood risk, interacting to shape an individual’s risk perception. Survey methods can be used to quantify risk perception (Rohrmann 2008) with this quantification being helpful in understanding how individuals and societies perceive and assess the risks associated with flooding.

There are differences between stakeholders’ perceptions of flooding risks, for instance, on the Gold Coast of Australia emergency experts, local government officers, and residents had differing perceptions of flood risk (Godber 2005). However, understanding students’ perceptions of flood risk have often been overlooked. Student populations are particularly vulnerable to flooding due to contributing socioeconomic factors, such as young age, not having flood insurance, and lack of experience. In the case of international student cohorts, there can also be cultural barriers (Ponstingel et al. 2019). Hung et al. (2016) note that students are particularly susceptible to flood impacts and will have higher socioeconomic losses because of their inadequate financial resources. Universities have an important role to play in supporting students (Simms et al. 2013).

2.1 Flood preparedness

Preparedness for disasters requires understanding the public’s perception of disaster risk. According to Peek and Mileti (2002), people who live in regions prone to disasters frequently fail to act or do very little to reduce their risk of death, injury, or property loss, which has been extensively documented. This inaction is because individuals often equate risk with probability. For example, in Switzerland, residents were found to miscalculate risk, creating gaps in disaster preparation plans to deal with flooding (Siegrist and Gutscher 2006). People often believe they are more prepared than they actually are, and can overestimate their ability to cope with a flood event (Paton and Johnston 2006). Flood mitigation infrastructure often masks dangers, such as levees and dikes (Terpstra and Gutteling 2008). In the case of university students, there is evidence that students overlook their own efficacy in preparing for disasters (Wu et al. 2017). This can lead to a false sense of security and complacency (Ludy and Kondolf 2012) which can result in people not taking the necessary precautionary measures. Environmental psychologists describe this phenomenon as “the levee effect” (Baan and Klijn 2004). In the Netherlands, this is referred to as “the myth of dry feet,” referring to people’s belief that the government can ensure adequate flood protection (Grant 2018). However, research has found that decision-makers often place a greater emphasis on measures that address immediate protection rather than those that affect long-term and indirect outcomes (Mehryar and Surminski 2022).

2.2 Flood knowledge

Knowledge plays a critical role in risk perception research (Johnson 1993; Lijklema 2001). Knowledge may enhance disaster preparation – both directly and indirectly. Risk education can help communities better prepare for floods by strengthening their resiliency (Dufty 2008). Hoffmann and Muttarak (2017) also note that risk education is critical in decreasing catastrophe risk and improving community disaster resilience. Based on a four-year study in the United States, university students expressed low flood risk perceptions, but as they aged, these perceptions increased (Ponstingel et al. 2019). The New South Wales State Emergency Service (SES) found in an evaluation that those communities who received education programs for more than one year were much more prepared and had a more significant desire to evacuate (Webber and Dufty 2008). Despite educational programs focused on disseminating how to understand flood communication, many people still struggle to understand messaging (Árvai 2014).

2.3 Past flood experience

People’s past experiences are a crucial factor influencing their perception of hazards (Weinstein 1989; Lindell and Hwang 2008; Miceli et al. 2008; Kellens et al. 2011; Terpstra 2011; Gotham et al. 2018; Wang et al. 2018). Fakhruddin et al. (2015) illustrate that person’s perception of future risks is sharpened by their experience. Generally, flood survivors have a higher perception of flood risk, particularly if they have suffered property damage or felt stressed during a flood (Wachinger et al. 2013). Their past experiences largely determine their understanding of the future. For example, although low elevations make flooding a common issue for the Netherlands, it is difficult for pupils to identify flood risks in their communities (Bosschaart et al. 2016). As a densely populated country with efficient drainage systems and a sophisticated network of dikes and pumps to prevent flooding (TeBrake 2002), students thus rarely encounter flooding firsthand (Parker et al. 2009). Students thus perceive little danger from floods and have tremendous confidence in water safety (Bosschaart et al. 2013). However, the link between experience and flood preparedness is contested with others noting it may only play a role in awareness, not action (Santoro et al. 2022).

2.4 Flood risk underestimation

Unlike engineers and planners, the general public is often biased in how they evaluate the likelihood of events (Weinstein 1999). Many individuals interpret flood return periods as deterministic rather than probabilistic (Kousky and Kunreuther 2010). While return periods denote the likelihood of an event occurring, people often perceive flood events as occurring on fixed intervals rather than a reoccurring possibility. Thus, people often underestimate the frequency and severity of floods (Zabini et al. 2021). Evidence shows that many students poorly comprehend fundamental statistical concepts (Garfield and Ahlgren 1988; Freeman et al. 2008). Even though students are exposed to probability concepts in their education, personal experiences often override mathematics (Kazima 2007). These misconceptions and misunderstandings frequently result in low-probability, high-consequence flood events being considered improbable (Salman and Li 2018). Numerous investigations have shown that people struggle to understand common flood phrases such as “1-in-100 year” or “1% Annual Exceedance Probability (AEP)” events (Godber 2005). The use of “1-in-100-year flood” or “1-in-50-year flood” as terms may thus conceal the risk magnitude associated with flood events (Gruntfest et al. 2002).

Heems and Kothuis (2012) note that while officials make significant efforts to connect with the public, there is still a gap between communities and authorities regarding how they see flood danger. Generally, public perceptions of risk may be simpler than expert perceptions – floods are possible, or they are not (Botterill and Mazur 2004). This phenomenon is because risk perception significantly divides professionals and the general population. Engineers and planners approach uncertainty in a structured manner (Salman and Li 2018), while the public often acts based on emotion and experience (Barnes 2002).

2.5 Raising flood risk perceptions

López-Marrero and Yarnal (2010) identify that flood hazards are often viewed as a less significant threat in the face of other household concerns, such as health, family well-being, and livelihoods. Providing individuals with relevant, valid, and updated flood information is thus vital to allow households to assess their risk and reinforce these perceptions with environmental cues (Bosschaart et al. 2016). University disaster risk reduction education (DRRE) can be improved by students’ knowledge and attitudes about disasters (Chen and Adefila 2020). The United Nations Educational, Scientific, and Cultural Organization reinforces the role of risk education in building community resilience (UNESCO 2016). Therefore, risk communication and education are essential in altering individual risk perceptions (Wachinger et al. 2013).

3 Methods

This study assessed how university students’ preparedness, knowledge, and experience predict flood risk perceptions in the Greater Sydney region. We also sought to understand how university students recognise different flood probability terms.

3.1 Data collection

We recruited students at the University of Sydney to participate in this research between May and August 2022 through in-person announcements at the start of lectures and by distributing flyers with QR codes at the entrance of each school or department to increase the diversity of our sample. A total of 500 invitations to participate in the research were distributed, out of which 272 (54.4%) online survey responses were returned. We omitted 10 incomplete responses from the analysis, resulting in 262 (52.4%) valid surveys. Survey responses were anonymous to the research team, with participants being told this before commencing the survey. To incentivise participation in the study, (50) AUD 10 gift cards for groceries were randomly raffled to participants.

There were 69,200 students enrolled in the University of Sydney in 2022, including 39,507 undergraduates, 29,693 postgraduates, and 3,669 doctoral students. An online survey questionnaire was designed and administered through Qualtrics. In this survey, our focus was on riverine and coastal flooding. The first component asked participants about demographic information, including their gender, age, place of residence (postcode), and current degree they were pursuing. In the second part, students were asked about the number of previous flood events they had experienced, their flood preparedness, their understanding of flood-related terms, and their perception of flood risk in Sydney. This research was approved by the Human Research Ethics Committee at the University of Sydney under project 2022/184.

3.1.1 Flood preparedness

Students were asked to respond to three questions about flood preparedness, drawn from elements highlighted by the NSW State Emergency Services in its public messaging. A five-point Likert scale (1 = strongly disagree, 2 = somewhat disagree, 3 = neither agree nor disagree, 4 = somewhat agree, 5 = strongly agree) was used for the following statements:

  • P1: I know where I will go in an evacuation and how I will get there.

  • P2: I know what I will take with me in an evacuation.

  • P3: I have talked with my household about what we will do if we need to evacuate.

We conducted a confirmatory factor analysis of the relationship between the measurement factor (preparedness) and the scale items. The average variance extracted (AVE) was 0.510, the composite reliability (CR) was 0.755, and with loading factors 0.755 (P1), 0.611 (P2), 0.765 (P2). Based on AVE and CR in this analysis, there was reasonable convergent validity in the data. According to the reliability analysis (Cronbach’s Alpha = 0.752), the scale reliability was also acceptable. Then we aggregated these three survey items to obtain students’ flood preparedness by calculating the total score per student. The range of student scores was between 3 and 15, with a mean of 7.51 and median of 7.00 with overall scores conforming to a normal distribution. Generally, a lower score indicated that a student’s preparation was less adequate.

3.1.2 Flood knowledge

Three questions were used to assess students’ knowledge of flood terms. Flood knowledge was hypothesised to impact how students formulated flood risk perceptions. We used three flood probability terms used by the NSW State Emergency Services in its communication of flood risk (NSW SES 2024):

  • K1: a 1-in-100 year flood.

  • K2: a flood with a 1% of happening in a given year.

  • K3: a flood with a 55% chance in an 80-year lifetime.

Each of these flood events has an equivalent probability of happening in any given year. Thus, we sought to test whether students could identify this commonality. Students were asked to evaluate the likelihood for each of the three flood events happening next year using a 7-point Likert scale (-3 = extremely unlikely, -2 = moderately unlikely, -1 = slightly unlikely, 0 = neither likely nor unlikely, 1 = slightly likely, 2 = moderately likely, 3 = extremely likely). The used CFA which resulted in an AVE of 0.516, CR of 0.759, and factor loadings for K1, K2, and K3 were 0.630, 0.841, and 0.667, respectively. The scale reliability was found to be acceptable (Cronbach’s Alpha = 0.752).

We hypothesised that a strong level of flood knowledge would result from a student answering the three questions consistently since they describe the same probabilities of occurrence. Thus, if students responded consistently, they were deemed to have a strong understanding of the three terms. Any deviation was categorised as diverging knowledge of flood probability terms. By comparing the deviations between the absolute values K1, K2 and K3 in pairs, we calculated the total deviation distance for each student’s understanding of the different flood terms, ranging from 0 to 8. To facilitate interpretation, we reversed coded our scale to provide better consistency when interpreting our model results. Lower scores thus indicate a lower level of flood knowledge. In addition to asking likelihoods for the above statements, we also asked students to provide an open-ended answer to describe what a “1-in-100 year flood” event meant to them given the frequent use of this term in the media following recent flood events in Sydney in 2022 which preceded the survey.

3.1.3 Flood experience

To assess flood experience, we asked students how many flood events they had encountered (0 = no experience with flood events, 1 = 1 flood event, 2 = 2 flood events, 3 = 3 flood events, 4 = 4 flood events, 5 = 5 flood events, 6 = 6 or more flood events). We treated this variable as continuous for assessing the influence of experience on flood risk perceptions.

3.1.4 Flood risk perceptions

To assess flood risk perceptions, we asked students about the probability of flood events in Sydney and the impact of flooding. Flood risk perception was determined by the product of perceived probability and estimated severity of flooding (Bubeck et al. 2012). To assess the likelihood of flooding in Sydney, students were asked whether they thought they would encounter floods in the future, using a five-point Likert scale (1 = definitely not, 2 = probably not, 3 = might or might not, 4 = probably yes, 5 = definitely yes). Students were also asked to assess the potential severity of these events by answering their level of agreement with the statement: ‘Flooding in Greater Sydney can cause significant damage.’ Severity was asked using a five-point Likert scale question (1 = strongly disagree, 2 = somewhat disagree, 3 = neither agree nor disagree, 4 = somewhat agree, 5 = strongly agree). Combining these together into a risk matrix (Markowski and Mannan 2008; Ni et al. 2010), we expressed flood risk perception as a product of the likelihood and severity of impact. A product of 1 to 8 represented low risk perception, 9 to 15 represented medium risk perception, and 16 to 25 represented high risk perception. This resulted in the matrix shown in Table 1. In general, those with a high flood risk perception are more likely to take proactive steps to protect themselves, such as creating an emergency plan. Those with a lower flood risk perception may be less likely to take these steps and can be more exposed in the event of a flood.

Table 1 Flood risk perception matrix
Full size table

3.2 Participants

We collected survey responses from 262 students. The proportion of men, women, and non-binary respondents was 59.2%, 39.3%, and 1.5%, respectively. In our sample, 71.4% of respondents were bachelor’s degree students, 20.2% were pursuing a master’s degree, and 8.4% of those surveyed were pursuing a doctoral degree. These proportions approximately corresponded to the student population of the university at the time of the research (see Table 2). Furthermore, we mapped the distribution of where sampled students reported living in the Greater Sydney area, shown in Fig. 2. Student places of residence were aggregated into five districts: Central (53.1%), North (17.9%), South (11.8%), West, West Central, South West (15.7%) and Outside the Greater Sydney (1.5%). In a small number of cases, a postcode may cross two districts. In this case, we assumed that a postcode only belonged to one administrative district based on which district comprised a higher percent of the postcode area. As shown in Fig. 2, the majority of students surveyed were residing in more central suburbs.

In our sample, 24.0% of students were younger than 20 years old, 54.2% of students were between 20 and 24 years old, 14.9% of students were between 25 and 29 years old and 6.9% of students were older than 29. The average flood preparedness level was 2.5 (out of 5) with a standard deviation of 1.04. The mean difference in students’ understanding of flood terminology was 4.6 (out of 16) with a standard deviation of 2.6. 3 in 5 students had no flood experience (59.2%), while 17.2% had experienced one flood event, 13.0% two flood events, 3.8% three flood events, 0.8% four flood events, and 6.1% five or more flood events. A summary of the sample and population characteristics is shown in Table 2.

Table 2 Descriptive statistics
Full size table
Fig. 2
figure 2

Distribution of surveyed university students in Greater Sydney

Full size image

3.3 Analysis

Considering that this study relied on self-reported data, it likely contains common method biases (CMB). We used a Harman’s single-factor test to assess CMB. To identify the relationships between flood preparedness, knowledge, and experience with risk perceptions that students form, we used ordinal logistic regression (OLR). Our regression included seven independent variables: (1) gender, (2) age, (3) degree level, (4) district (5) flood preparedness, (6) flood knowledge, and (7) flood experience. Gender considered women, men, or non-binary categories, while age, flood preparedness, knowledge, and experience were taken as continuous variables. The degree level was taken as bachelors, masters, or doctoral. We adopted the five districts used by the NSW Department of Planning for place of residence: Central, North, South, West (include West, West Central, South West) as well as a sixth group for those not included of these districts. Since the main campus of the University of Sydney is in the Central district, we used this as the reference residence group. Finally, we selected the most appropriate link function (logit, complementary log-log, negative log-log, probit and cauchit) by comparing candidate models through the parallel line test and goodness-of-fit. Based on these criteria, we selected a cauchit function.

Our general regression expression was as follows:

$$\begin{aligned} &Cauchit\left(Risk\right)={\beta }_{0}+{\beta }_{1}*Age+{\beta }_{2}*Gender+\\&{\beta }_{3}*Degree+{\beta }_{4}*District+{\beta }_{5}*Preparedeness+\\&{\beta }_{6}*Knowledge+{\beta }_{7}*Experience\end{aligned}$$

Where \({\beta }_{0}\) is the constant term; \({\beta }_{N},\) refers to the parameters to be estimated. We used Taylor-linearised variance estimation to adjust the confidence intervals for clustering effects. 95% confidence intervals were used for the results. There was a 6% margin of error in our study. Before any analysis, the data was cleaned for consistency. This involved reviewing the quality of comments and responses made by participants and categorising data quality. Analysis was completed using SPSS statistical software.

To investigate whether there were differences in students’ understanding of the three flood probability statements a Friedman test was used to assess differences among the three statements. Additionally, the Kruskal-Wallis H test was used to explore whether students with different levels of risk perception responded differently to probability statements. As part of this study, students were divided into three groups based on their risk perception: high risk, medium risk, and low risk. These three groups of students were compared with the three flood probability statements.

4 Results

In the following section, we perform an ordinal logistic regression to identify predictors of flood risk perceptions, before proceeding to discuss differences in how student’s understand common flood terminologies using Friedman and Kruskal-Wallis H tests.

4.1 Predictors of flood risk perceptions

In our regression model, student flood risk perceptions were the dependent variable. The assumption of proportional odds was met, using a likelihood ratio test comparing the fit of the proportional odds model to a model with varying location parameters, χ2 = 11.993, p = 0.446. Using a likelihood ratio test, the model significantly predicted student risk perceptions better than the intercept-only model (χ2 = 47.873, p < 0.001). Based on the Harman’s single-factor test, 35.53% (< 40%) of the variance was explained by the first factor, which indicates that CMB wasn’t significant.

Age, degree, postcode, and knowledge did not have statistically significant effects on students’ flood risk perceptions. We found gender, preparedness, and experience did however have statistically significant effects. Women were 1.95 times more likely than men to have higher risk perceptions (95% CI OR: 1.159–3.290, p = 0.012). We identified a negative relationship (OR = 0.908) between a student’s preparedness and flood risk perception (95% CI OR: 0.836–0.987, p = 0.024). Finally, for each additional flood event experienced by a student, the likelihood of having a high perception of flood risk was 1.889 times higher (95% CI OR: 1.426–2.499, p < 0.001). A summary of our regression results is shown in Table 3; Fig. 3.

Table 3 Ordinal logistic regression results
Full size table
Fig. 3
figure 3

Forest plot of flood risk perception predictors

Full size image

4.2 Differences in understanding flood probabilities

We also compared how students assessed the likelihood of ‘1-in-100 year’, ‘1% in any given year’, and ‘55% in 80-years’ flood statements. We used a Shapiro-Wilk test for check for normality, finding all the probability terms ‘1-in-100 year’ (p = 0.000), ‘1% in any given year’ (p = 0.000), and ‘55% in 80-years’ (p = 0.000) were not normally distributed. Thus, a Friedman test was used to explore students’ sensitivity between different flood probability terms. The results showed that students’ sensitivity of the flood term was statistically different across flood probability terminology (χ2 = 58.385, p < 0.001). Further pairwise comparison using the Bonferroni method showed that there were statistically significant differences between ‘1-in-100 year’ (Median = 0, Mean = 0.04) and ‘55% in 80-years’ (Median = 0, Mean = 0.720) (p < 0.001). There were also differences between the ‘1% in any given year’ (Median = 1, Mean = 0.031) and ‘55% in 80-years’ statements (p < 0.001), but not between ‘1-in-100 year’ and ‘1% in any given year’(p = 1.000). We can thus conclude that students generally identified the first two terms as the same but identified the third term as different.

A Kruskal-Wallis test revealed there were not statistically significant differences across different risk perception groups when assessing ‘1-in-100 year’ (χ2 = 0.539, p = 0.764) and ‘1% in any given year’ (χ2 = 1.892, p = 0.388) flood probability statements. However, we did identify statistically significant differences across risk perception groups for the ‘55% in 80-years’ (χ2 = 8.771, p = 0.012) statement. The distributions of understanding of flood probabilities scores were similar for all groups. Further pairwise comparison using the Bonferroni method identified statistically significant differences between the “high” (mean rank = 147.56) and “moderate” (mean rank = 119.19) (p = 0.019) risk perception groups, but not between the “low” (mean rank = 121.56) and “high” (p = 0.100) or “low and “moderate” (p = 1.000) risk perception groups. We show a comparison of the ratings of flood probabilities across risk perception groups in Fig. 4.

Fig. 4
figure 4

Distribution of student perceived likelihood of flood probability terms

Full size image

Qualitative analysis offered further insights into the gaps between interpretation and cognition of flood probabilities. We identified four primary conceptualisations put forward by students in how they understood the presented annual exceedance probability statement (‘1-in-100 year’). These interpretations focused on evaluating the probability statement solely through severity (23%), frequency (54%), a combination of severity and frequency (15%) and other dimensions (8%), shown in Fig. 5. For example, a student who focused on frequency described: “I think that this means we are experiencing flood events that would normally be very rare and not likely to occur frequently.”

Fig. 5
figure 5

Student understandings of ‘1-in-100 year flood’ statement

Full size image

5 Discussion

The individual factors affecting university students’ understanding of flood risk are diverse and complex. In our study, we found that gender, flood preparedness, and flood experience were the main factors associated with university students’ flood risk perceptions. Age, postcode, education level, and flood knowledge were inconclusive predictors of flood risk perceptions, but their function may either modify or amplify existing risk perceptions (Wachinger et al. 2013). In addition, we also studied students’ views on flood terms and identified how students understand probability statements presented in different formats. Our results show that students are less sensitive to the likelihood of flooding in the term “1-in-100 year flood”, which is often used by the media, as compared to flood terms that incorporate percent probabilities.

5.1 Factors associated with risk perceptions

Our regression analysis did not indicate that age was associated with differences in flood risk perception. However, our study differed from others (Kellens et al. 2011; Eryılmaz Türkkan and Hırca 2021) in that we narrowly focused on youth which may explain why we did not find differences in perceptions for this variable. We did find differences across gender in flood risk perception among university students, with women students being more likely to have higher flood risk perceptions than men. This may suggest that women students are more aware of the risks posed by floods as suggested by other studies (Miceli et al. 2008; Kellens et al. 2011; Gotham et al. 2018; Perić and Cvetković 2019). Hence, our results support previous research, such as Saleh Safi et al. (2012) who found gender is seen as a factor influencing risk perception, whilst age has a diminished role.

We did not find significant differences in risk perceptions across different education levels among students. In other words, a higher degree did not mean students were more aware of floods. This may however be explained by different disciplines which were not controlled for in the study. We did not find significant differences in flood risk perceptions across districts which we initially hypothesised might occur due to closer proximity to the Hawkesbury-Nepean catchment which is particularly susceptible to flooding (Gillespie et al. 2002; Masud et al. 2019). As claimed by Kellens et al. (2011), residential location may not have a strong correlation with flood risk perception. This might also be offset by exposure to coastal flooding for students residing closer to coastlines. Individual residential flood risk assessments were beyond the scope of this study but would offer more conclusive understanding of this factor on risk perceptions.

Furthermore, our results did not show a statistically significant difference between knowledge of flood terminology and risk perceptions. Although some studies (Botzen et al. 2009; Chen and Adefila 2020) have found that flood knowledge and flood risk perception are positively correlated, it is also possible for flood knowledge to have an opposite effect on perceptions of risk (Johnson 1993; Lijklema 2001). For instance, students may tend to overestimate potential risks when they have higher understanding of relative risk probabilities (Cox et al. 2010). Typically, flood knowledge is self-reported in studies concerned with flood-risk perception (Bosschaart et al. 2013) but knowledge may be better considered as a conceptual structure (Johnson 1993). People’s risk perception is thus influenced by their trust in the responsible risk manager when they lack knowledge of a hazard (Kellens et al. 2013). Our inconclusive link between flood knowledge and risk perception may stem the fact that we limited our analysis to flood terminology without considering other knowledge factors which merit further research.

A negative correlation was found between flood preparedness and perception among students. More specifically, participants who were better prepared tended to have lower flood risk perceptions. This could suggest that for students who were prepared better, they accepted their flood risks more openly. One possible explanation for this phenomenon is that, despite having strong flood risk perceptions, students may become complacent in their preparations due to a misplaced trust in government flood preparedness measures. Terpstra (2011) argues that trust can prevent flood risk anxiety through the affective process, which affects flood preparedness. The relationship between personal preparedness and risk perception may also be affected by fear and trust (Ejeta et al. 2018). In further analyses, the relationship between preparedness and cognition should be considered in terms of emotional factors, such as trust and worry, to identify potential connections.

Flood risk perceptions were positively correlated with flood experience. This is supported by Lee (2021) who found that individuals who personally experienced floods were more likely to be aware of flood terminology and to say they comprehended it. Wachinger et al. (2013) points out that exposure to flooding can heighten the potential for future risk and threat. For example, individuals are likely to be fearful of floods if their house was flooded before (Siegrist and Gutscher 2008). Moreover, this direct experience may enable future motivation for flood prevention (Miceli et al. 2008). However, Bradford et al. (2012) argue that individuals are at high risk of flooding whether or not they have experienced a flood.

5.2 Sensitivity of students to flood terms

We also identified differences in how students comprehend common flood probabilities. We found that students believe that the likelihood of flooding varied for different terms with overall stated likelihood decreasing in the following order: ‘55% in 80-years’ > ‘1% in any given year’ > ‘1-in-100 year’. Students stated that a flood with a “55% chance in an 80-year lifetime” was more likely to occur while a “1-in-100 year” flood was the least likely. This may suggest that students struggle to grasp the likelihood of the same flood happening in any given year, but better expect the same event occurring over their entire lifetime. Regardless of the mechanisms of understanding, our results show the differences in students’ conceptualisation of probability statements presented in different formats. Our research supports the need for further statistical and probability education among students in Australia, especially in application to complex real-world situations.

Moreover, we also found that students identified higher likelihood of floods when their probability is presented over a longer period. This may be because the larger percentage was more impressionable to students. We hypothesise that when a student saw the large percentage, they assumed that the probability was higher – supported by open ended responses from students in the survey. However, it is also possible that this is because students believe that flooding is more likely to occur during their life. In other words, if a flood probability is reported over a lifetime, students are more likely to think that this event will happen. They thus may think of the flood event as more likely and make decisions differently on this basis. This study also supports Keller et al. (2006), who found that probabilities for a longer time period increases the perceived threat leading to higher perception.

We did not find evidence of differences in how students evaluated the likelihood of flooding for the terms ‘1-in-100 year’ and ‘1% in any given year’ across high, moderate, or low risk perception groups. In other words, students with different levels of risk perception did not perceive differences in the expression of the flood probability terms. This may indicate that the two terms are indistinguishable in conveying risk information to students and do not increase or decrease the likelihood of perceived flooding. However, we found that there were significant differences in students’ perceptions of the term ‘55% in 80-years’ among students with different levels of risk perception. Specifically, students with high-risk perceptions responded with higher likelihoods for this flood term, as compared to those with moderate risk perceptions.

5.3 Rethinking the presentation of probabilities

Misunderstanding flood terms among students has significant implications for risk communication. A notable trend was the divergence in statements among students regarding the meaning of a ‘1-in-100 year flood’ between frequency, intensity, or some combination of both. Scambos et al. (2008) illustrate how misunderstandings of flood probabilities can lead to poor decisions and disastrous outcomes. This may lead to a false sense of security when it comes to rare events and can lead to students making dangerous decisions. For example, some students described in the survey the ways in which this problematically manifests in media, such as noted by individual: “I don’t think those news outlets truly understand the concept of the return period of floods, and they sometimes misinform the public with this terminology.” This depiction of media, as well as policymakers, may mislead audiences outside flood zones, to some degree, when reporting flood hazards. The technical term “100-year-flood” is frequently misused by the media, misrepresenting the risks that it is meant to convey (Baker 2018; Lee et al. 2021).

Many students also talked about how historical probabilities were changing under climate change. One student commented: “It should statistically not happen, but because of a rapid environmental progression we are seeing the extremes of annual environmental events in Australia.” This aligns with calls by some to re-examine the way in which we present flood probabilities (Holmes Jr and Dinicola 2010) – not only under climate change but also as a means to improve risk communication more fundamentally. Our findings suggest that describing the likelihood of flood events over longer durations may connect with a stronger realisation of potential impacts.

5.4 Limitations

There are a few limitations to this study. First, this study examines only students’ personal factors, while missing other potentially important contextual factors, such as the local flood exposure of student’s residences, socio-economic status, and cultural backgrounds. We expect that students who live near flood-prone areas will have a higher flood perception (Bosschaart et al. 2013). At the time of writing – there was not publicly available and standardised flood exposure data across Sydney. Previous studies have noted that students from higher income backgrounds may have a higher perception of flood risk (Berman et al. 2015; Ahmad and Afzal 2020). We also did not capture detailed contextual information about some of the variables due to necessary brevity in the survey – such as the nature of experience with previous floods – which merits further in-depth study.

The survey was also limited to students at the University of Sydney. Our approach was unable to employ formal random sampling, however as we describe above, we undertook measures when collecting data to ensure diversity of responses. Finally, our exploration of student knowledge was limited to the probability of flooding and did not include other aspects of flood-related knowledge, such as flood generation principles and local knowledge of exposure. Future research can continue to compare the effects of other factors on flood perceptions. Not all students have direct experience with flooding, thus it is also significant to find methods of risk communication that can create emotional links, such as virtual reality. Simulations based on virtual reality may be able to address this challenge, by providing users with a safe, immersive, interactive, and engaging means of learning about disasters (Calil et al. 2021).

6 Conclusion

We collected 262 responses from an online survey of university students in the Greater Sydney region, finding that higher flood experience and lower flood preparedness were associated with increased odds of higher flood risk perceptions. We also found that women had higher odds of higher flood risk perceptions. Age, location, educational, and flood knowledge were not found to be significant predictors. To further unpack how students perceive flood risk, we asked students to evaluate the likelihood of a 1-in-100 year flood, a flood with a 1% of happening in a given year, and a flood with a 55% chance in an 80-year lifetime – all of which express the same probability. We did not find differences in how students assessed the first two statements but found that students rated higher likelihoods for a ‘55% chance in an 80-year lifetime’ event. Our study is one of the first to examine university student flood risk perceptions in Sydney.

It is promising to see that students generally understood that a 1-in-100 year flood event is indeed an equivalent expression to an event with a 1% of happening in a given year. However, our findings point to lower recognition of near future flood events among students when compared over longer periods of time. This bodes well for longer term preventative actions that students may take but may signal a need to focus on short term flood awareness and risk communication. While policymakers and news media commonly use ‘1-in-100 year’ descriptions of flood events, our results point to the potential for expressing probabilities over longer durations to draw attention to heighten perceptions of flood risks.

There is thus a need to improve students’ understanding of flood terms so that they can better understand flood consequences. Media and higher education institutions have an important role and responsibility to play in sha** these perceptions among university students. It is also important to consider the direct or indirect experience of floods in risk education to improve students’ perception of flood risk. Incorporating flood direct or indirect experience into higher education is a effective way to improve students’ flood perception. In the context of climate change and achieving carbon neutrality in Australia, our research can help universities and governments design and improve flood risk education and communication programs.