The Neglected Role of Risk Mitigation Perception in the Risk Governance of Underground Technologies—The Example of Induced Seismicity

In total, 808 respondents from the German- and French-speaking parts of Switzerland completed the survey, recruited through an online access panel. Because of unrealistic answers, we excluded one respondent from the analysis, leading to 807 respondents. We used a quota for gender, age, and the two language regions. Thus, gender and language ratio were balanced and representative for the Swiss population (Swiss Federal Statistical Office 2016). The age ranged from 18 to 85 years (M = 47.12 years, SD = 15.00), which is slightly older than the average Swiss population (M = 43.65 years) (Swiss Federal Statistical Office 2016). Compared to the average Swiss population, the sample was better educated. A minority of 43 respondents (5.3%) completed compulsory education (6 years of primary school and 3 years of secondary school, nine years in total), which was lower compared to the Swiss average of 12.6%. The number of respondents who completed secondary education (N = 429, 53.1%) was higher than the average in the Swiss population (46.2%). Respondents who had completed the tertiary education level (41.6%) were representative for the Swiss population average (41.3%) (Swiss Federal Statistical Office 2017). An additional 284 respondents did not complete the full survey. The highest dropout rate was found for CCS (N = 79), followed by CPG (N = 74), SG (N = 65), and DGE (N = 66).

We tested the perception of four underground technologies: deep geothermal energy (DGE), hydraulic fracturing for shale gas (SG), CO₂-plume geothermal (CPG), and carbon capture and storage (CCS). The respondents were randomly assigned to one of the four underground technologies (between subject design). We tested the perception of the four risk mitigation measures by a within subject design.

The survey consisted of several parts. It first provided respondents with necessary information to reach an informed opinion. The information text included an illustration, adopted from Knoblauch et al. (2017) for DGE, SG, and CPG, and L’Orange Seigo et al. (2013) for CCS. The information texts were structured similarly as follows:

Carbon Capture and Storage Carbon dioxide (CO₂) is one of the most relevant greenhouse gases, which is mostly released into the atmosphere by the combustion of fossil fuels. The combustion of fossil fuels (oil, natural gas, and coal) generates large quantities of electricity and heat worldwide. With carbon capture and storage (CCS) a large part of these CO₂ emissions are directly captured at the source of the power plants and stored in the underground. The captured CO₂ is pumped through a borehole to a depth of at least 800 meters. For this purpose, the CO₂ must be compressed so much that it is no longer gaseous. This ensures that carbon dioxide is not released into the atmosphere.

CO₂-Plume Geothermal Geothermal energy—also called underground heat—is energy stored in the subsurface. The amount of heat stored at the interior of the Earth is large. The deeper, the warmer it becomes. Temperature increases by 3 °C per 100 m on average from the Earth’s surface. There are different technologies to use this energy. A distinction is made between deep geothermal energy and near-surface geothermal energy. Deep geothermal energy uses heat in up to 5 km depth for heating and in many cases for electricity generation. Carbon dioxide (CO₂) is pumped down, which is heated in the underground. CO₂ is one of the most relevant greenhouse gases. A part of this CO₂, which is pumped into the depth, will be stored permanently in the underground. This ensures that carbon dioxide is not released into the atmosphere.

We used univariate analysis of variances (ANOVAs) to study the effects of the technologies on the respondents’ perceptions of risk mitigation measures. Repeated measures ANOVAs were conducted to analyze the results across different risk mitigation measures (within subject design). If Levene’s test of homogeneity of variances was significant for the dependent variables (p < 0.001) and the group sizes were unequal, we used Welch’s F-test and Games-Howell as a post hoc test for statistical significance (Field 2013). Otherwise, we used Bonferroni post hoc testing.

The four experimental conditions were balanced in terms of: language χ² = 0.85, p = 0.84 and gender χ² = 1.35, p = 0.72; whether respondents owned or rented a house or a flat χ² = 6.62, p = 0.68, and the size of the respondents’ residences F(3, 803) = 1.00, p = 0.39; and in terms of age F(3, 803) = 0.98, p = 0.40 and education F(3, 803) = 1.66, p = 0.18.

Respondents had to evaluate the effectiveness and the necessity of the risk mitigation measure, the subjective safety feeling, and if they would approve the technology project if such a measure was applied. The mean scores of these ratings for each mitigation measure are displayed in Table 2. The perception of effectiveness, necessity, safety, and technology approval differed significantly across the four mitigation measures (Fig. 1). The measures traffic light system and insurance are considered most needed, but relatively ineffective and less safe. Structural retrofitting and project relocation are perceived as less needed but more effective and safe.

With respect to effectiveness, safety, and technology approval, project relocation was perceived as the best and insurance the worst. Structural retrofitting was in-between, followed by traffic light system. Performing contrast analysis regarding safety revealed significant differences in all combinations. This is also true for effectiveness, except between the measures structural retrofitting (M = 4.48) and relocation (M = 4.63). Performing contrast analysis for the approval of the technology also revealed statistically significant differences between measures except for traffic light system (M = 3.64) and insurance (M = 3.56). However, with respect to necessity, traffic light system was perceived as the best, followed by insurance, structural retrofitting, and relocation. Performing contrast analysis only revealed statistically significant differences between relocation and all other measures.

To check which risk mitigation measure was preferred overall (see Fig. 1), we conducted repeated measures ANOVA. The results show that the perception significantly differed by the measure to mitigate the risk of induced seismicity, F(2.83, 2283.78) = 14.55. Post hoc tests using the Bonferroni correction revealed that the mitigation measure relocation (M = 4.49 ± 0.048) was preferred over insurance (M = 4.20 ± 0.050) and over the measure traffic light system (M = 4.32 ± 0.047), which was statistically significant in both cases (p < 0.001 and p = 0.003). Post hoc testing also revealed that the measure structural retrofitting (M = 4.41 ± 0.045) was preferred over insurance (M = 4.20 ± 0.050), which was also significant (p < 0.001). Therefore, we can conclude that the measures relocation and structural retrofitting are more desirable overall than the measures insurance and traffic light system.

A closer look at the evaluation of the mitigation measures showed that all four measures aggregated together are perceived as rather necessary (M = 4.90 ± 0.041). But effectiveness (M = 4.35 ± 0.043), safety (M = 3.95 ± 0.050), and approval (M = 3.73 ± 0.057) are seen as much more critical than necessity (Fig. 2).

Univariate ANOVA revealed that the perception of the four mitigation measures differed significantly across technologies (Table 3). Bonferroni post hoc testing showed a significant difference between DGE (M = 4.56) and SG (M = 4.03) across all four measures (p < 0.001). Thus, respondents to DGE preferred all risk mitigation measures compared with the respondents to SG (Table 3). The perception of the risk mitigation measures traffic light system and structural retrofitting differed significantly between all technologies, except that there was no difference between DGE and CPG (p = 1) and between CCS and SG (p = 1), indicating that these measures were preferred by respondents to DGE and CPG over respondents to CCS and SG. The risk mitigation measure insurance was also perceived more positively by respondents to DGE and CPG, showing a significant difference between DGE and CCS (p = 0.002), DGE and SG, as well as SG and CPG (p = 0.009).

We conducted repeated measures ANOVA to check which measures were perceived differently across technologies. There was a significant (but small) interaction, F(8.54, 2285.54) = 2.056, p = 0.03, indicating that the preference of the mitigation measures differed according to the technology. To break down this interaction, contrasts analysis was performed, comparing the perception of all risk mitigation measures across technologies. This revealed a significant interaction between the perception of the risk mitigation measures traffic light system and relocation across technologies, F(3,803) = 3.50, p = 0.015. This effect reflects that only respondents to CPG preferred the measure traffic light system over relocation. The other technologies preferred the measure relocation. For the rest, the preference ranking was the same (project relocation, structural retrofitting, traffic light system, earthquake insurance), but at different absolute levels (Table 3).

We conducted a univariate ANOVA to examine the effect of the technology perception (negative, neutral, or positive) on risk mitigation measures preference. Levene’s F test revealed that the homogeneity of variance assumption was not met (p < 0.05) for some of the risk mitigation measures. As such, we used the Welch’s F-test. The analysis showed that the effect of the technology perception on the preference of risk mitigation measures was significant for the measures traffic light system, insurance, and relocation (Table 4). A positive technology perception leads to a higher preference for the risk mitigation measures traffic light system and earthquake insurance. In the case of the measure relocation, a positive technology perception leads to a lower preference.

A closer look at the first choice of the risk mitigation measure revealed differences for the different technology perceptions. Respondents with a negative perception of the technology (regardless of the technology) preferred relocation, followed by structural retrofitting, traffic light system, and earthquake insurance. Respondents with a positive perception of the technology preferred different mitigation measures. They preferred the measure traffic light system, followed by relocation, structural retrofitting, and earthquake insurance (Fig. 3).

There are three substantive findings. First, mitigation measures are perceived differently. The mitigation measures project relocation and structural retrofitting are preferred over traffic light system and earthquake insurance. The least preferred mitigation measure, earthquake insurance, is also only reluctantly taken out for naturally occurring earthquakes (Heller et al. 2005; Brown 2013; Wachinger et al. 2013). Unlike most other mitigation measures, earthquake insurance needs action from the respondents themselves. As a result, insurance is probably less preferred. In general, direct mitigation measures, with the aim to reduce seismicity, are preferred over indirect measures. People prefer the hazard to be addressed directly and not for the potentially resulting consequences to be reduced.

Compensating the risk afterwards is viewed as less desirable than relocating the project and its potential hazards. The mitigation measure relocation is preferred over traffic light system and insurance. This could give us an indication that the public is especially sensitive with respect to their neighborhoods. This is in line with many studies around contested technology. With respect to CCS, for example, Krause et al. (2014) showed that 20% of the original proponents of a technology changed their minds after such a project was promoted in their neighborhood. Knoblauch and Trutnevyte (2018) also showed for deep geothermal energy that projects should be sited in remote areas, where the risk of induced seismicity is reduced, and the project is more accepted. The measure traffic light system is not perceived as particularly effective in comparison to the other mitigation measures. It might be that the public does not perceive this as a direct measure. In fact, real-time monitoring cannot always prevent seismic events, as there is a certain risk that some events occur even after the injection has stopped. Still, the traffic light system is considered to be best practice, especially for DGE (Kao et al. 2016). Thus, more dialogue with the public on the possibilities and challenges of traffic light systems would be reasonable. All risk mitigation measures are perceived more necessary than effective. This finding is in line with the longitudinal questionnaire study by Perlaviciute et al. (2017), which also found that most risk mitigation measures for earthquakes caused by gas production in the Netherlands were seen as more urgent than effective. The public wants mitigation measures, and finds them necessary, but people are not convinced that these measures will suffice. For risk governance, this means that polling risk mitigation perception is an essential input for choosing the appropriate set of mitigation measures and a respective risk governance scheme.

Second, mitigation measures are perceived differently across the four technologies. Our results show that all four mitigation measures are perceived more positively for a DGE project than for a SG project. The type of technology, and preexisting views and attitudes towards different technologies influence how the public perceives mitigation measures. This finding is comparable to the results of the study by Knoblauch et al. (2017) for risk communication with respect to induced seismicity. Attitude towards technology is a crucial factor in risk perception (Sjöberg 2000), and negative technology attitudes arguably lead to negative perceptions of risk mitigation measures as well. This can be supported with our results: DGE is perceived most positively, SG most negatively, similar to the risk mitigation perception. Still, the differences in risk mitigation perception are much smaller than for the technologies in general and not consistent across all four technologies. The differences are driven by other factors, arguably a combination of the differences in risk profiles, technological maturity, social discourse, and risk benefit perceptions (Dowd et al. 2011; Aposteanu et al. 2014; U.S. Department of Energy 2015; Knoblauch et al. 2017). Additional studies will be needed to better understand the variations in risk mitigation perception across technologies. For risk governance, this reaffirms the need to adapt measures according to the technology.

Third, we see that risk mitigation perception varies across and within different technologies. This has implications for risk governance, and concern assessment more specifically (IRGC 2012). We expect that the public will, dependent on the technology discussed, react differently to proposed risk mitigation measures. Thus, the public would conclude differently how well their concerns were taken into account and respond differently to the technology as well. This finding is comparable with a previous study, which analyzed three different hazards and showed that all hazard types required specific tailor-made mitigation approaches (de Vet et al. 2019). This is important for risk governance, as these frameworks generally assume that knowing about risk perception and designing risk mitigation measures accordingly (Trutnevyte and Wiemer 2017) would suffice. In fact, we can illustrate that neglecting such information will most probably lead to inadequate recommendations of risk mitigation measures within the scope of risk governance frameworks. Our study shows that we need to consider risk mitigation perception as well, as this can be important for the implementation of a technology.

There is no Swiss history of fossil-based fuels, unlike in other countries such as Germany or the United States. Underground technologies are not very well known in the broad Swiss population. This is especially true for CO₂-plume geothermal, which is still in the phase of basic technology research (Molnár et al. 2018; Wang et al. 2019). With respect to carbon capture and storage, there has not yet been a broad public discourse in Switzerland (unlike in other countries) and it is still unclear whether Switzerland is suitable for storing carbon dioxide. Thus, generalizations should be made with care and we strongly encourage researchers to replicate our study for other countries.

The sample in the study is slightly better educated and older than the Swiss population as a whole, comparable with other surveys (Mohorko et al. 2013). Also, a high dropout rate from the survey was recorded. But because the dropout rates were similarly high for the four technologies, the findings should not be biased when comparing between the technologies. However, it cannot be excluded, that this dropout rate had some impact on the findings.