Carbon dioxide removal to combat climate change? An expert survey on perception and support

The most recent IPCC report considers Carbon Dioxide Removal (CDR) as an unavoidable climate change mitigation measure, although controversial discussions about CDR have taken place in the past. This study contributes to the ongoing debate by providing insights into academic expert opinions about Bioenergy with Carbon Capture and Storage (BECCS) and Direct Air Carbon Capture and Storage (DACCS). An online survey was conducted to examine how academic experts (N = 172) perceive and to what extent they support BECCS and DACCS. Overall, the results show rather high expert support for research on and the deployment of BECCS and DACCS. Perceived feasibility of the technologies consistently predicted support for BECCS and DACCS, with means in the upper medium range. Further significant predictors were the extent to which experts perceived the technologies to be necessary, the experts’ amount of academic experience, the associated moral hazard, and the perceived tampering with nature.


Introduction
According to the IPCC Sixth Assessment Report (IPCC 2022), Carbon Dioxide Removal (CDR) is no longer considered as a potential climate mitigation measure but rather as a necessary component of climate protection. About 533 GtCO 2 have to be removed from the atmosphere between 2020 and 2100 by using CDR to (likely) stay below two degrees of global warming (IPCC 2022). In particular, the use of four CDR approaches has been considered, namely, (1) Afforestation and Reforestation (A/R), (2) Enhanced Weathering (E/W), (3) Bioenergy with Carbon Capture and Storage (BECCS), and (4) Direct Air Carbon Capture and Storage (DACCS), whereby the latter two are also being labeled as CDR-CCS (Mintz-Woo and Lane 2021). BECCS is expected to remove 291 GtCO2 and DACCS 19 GtCO 2 between 2020 and 2100 (IPCC 2022). Because a goal has been set to achieve net zero emissions, CDR-CCS is considered to be a Net Emission Technology (NET) (Minx et al 2017, Rickels et al 2019, Otto et al 2021. CDR-CCS is based on the idea that carbon dioxide can be captured and then stored in geological storage sites. In general, the IPCC assumes that the geological storage of CO 2 is 'a mature technology' and that 'the CO 2 can be permanently isolated from the atmosphere' (IPCC 2022, p 38). In this way, CDR-CCS is closely related to conventional CCS, which is based on an idea to capture CO 2 emitted by fossil-fueled power plants on site (L'Orange Seigo et al 2014, Mintz-Woo andLane 2021). In the case of CDR-CCS, CO 2 is either captured in biomass power plants (BECCS) or directly captured from the atmosphere using filters with chemical absorbents (DACCS) (Tavoni and Socolow 2013).
What makes CDR-CCS controversial is the fact that the current deployment rates are insufficient to limit the global temperature increase to 2°C (IPCC 2022); thus, the feasibility of CDR-CCS has been questioned (Anderson and Peters 2016, Low and Schäfer 2020). The real economic costs of CDR-CCS are yet to be

General factors General Mean (SD)
Technology optimism (α = 0.70) 4.23 (0.92) All items adopted (Lutzke and Árvai 2021). Technological advancements can be used to solve future problems. Risks connected to new technologies should be seen as temporary problems that will be solved later. Science and technology will make our lives easier. Careful scientific research will help to prevent accidents at facilities that rely on new technologies. Support for bioenergy (α = 0.91) 3.82 (1.18) Personal items Support for solid biofuel Support for liquid biofuel Support for gaseous biofuel Perceived risk of geological storage of carbon dioxide (α = 0.82) 2.72 (1.11) Personal items How would you rate the technological maturity of geological storage of carbon dioxide? How would you rate the risk of carbon dioxide leaks from geological storage sites? How would you rate the danger for humans and ecosystems posed by the geological storage of carbon dioxide?

Mean (SD)
Perceived cost-efficiency (α BECCS = 0.76; α DACCS = 0.77) 3.05 (1.25) 2.05 (1.10) The deployment of BECCS/DACCSK K is a cost-efficient measure for reducing the Earth's temperature. (Item adopted by Jobin and Siegrist (2020) (Jobin and Siegrist 2020). Support for further research regarding BECCS/DACCS Support for the deployment of BECCS/DACCS to mitigate climate change in the 21st century Note. SD = standard deviation; all factors were measured on six-point scales, and high values represent high levels of the respective factor. determined (Lehtveer and Emanuelsson 2021), and geologically storing carbon dioxide has been described as presenting a risk to human health (Rochon et al 2008). Moreover, questions remain open regarding the moral permissibility of CDR (Gardiner 2013, Preston 2013, Anderson and Peters 2016, Shue 2017, Lenzi et al 2018, Haikola et al 2019, e.g., whether humans may interfere or tamper with nature (Jamieson 1996, Preston 2013. BECCS might increase competition for land and, at the same time, decrease biodiversity due to the demand for biofuels (IPCC 2022). Furthermore, CDR-CCS has been accused of representing a moral hazard, because it legitimates countries to emit CO 2 based on the belief that emissions could be removed ex post facto (Anderson and Peters 2016). However, there is no absolute certainty that it will be feasible to remove substantial amounts of CO 2 in the future, and uncertainties related to the deployment of CDR demand further climate mitigation measures (Grant et al 2021).
Recently, academic discussions about CDR-CCS have rapidly evolved due to the ongoing progress of research and development (Haikola et al 2019). The role allocated by the IPCC (IPCC 2022) suggests that CDR-CCS is currently approved by a broad range of scientists. While some studies have investigated how the public perceives CDR-CCS (Wright et al 2014, Wolske et al 2019, Cox et al 2020, Jobin and Siegrist 2020, Wenger et al 2021, research on the perceptions of experts is scarce. Existing studies suggest that experts are ambivalent towards NETs (Delina 2021) and that experts have concerns about the technological feasibility of BECCS and the impact of BECCS on biodiversity (Vaughan and Gough 2016). At the same time, it has been shown that the acceptance for BECCS is increasing (Romanak et al 2021) and that climate system modeling experts perceive NETs as necessary (Rickels et al 2019). Comprehensive surveys of expert opinion stemming from various scientific disciplines, however, are lacking.
To address this gap, we conducted a quantitative survey among academic experts. In particular, we investigated expert support for CDR-CCS, characteristics commonly associated with CDR-CCS, and analyzed factors that influence expert support for CDR-CCS. Differentiating between BECCS and DACCS, we analyzed whether the two technologies are perceived and supported differently. As previous research has shown that support for novel technologies can vary based on differences in scientific discipline (Hochschild and Sen 2015), we also analyzed whether the experts' academic backgrounds influenced their perceptions of CDR-CCS. This study, therefore, contributes to the emerging discussion related to these technologies by quantifying expert support and identifying underlying factors.

Methods
To investigate how experts perceive CDR-CCS, a survey with general and technology-specific questions was sent to experts on CDR. The survey was online from 14 June until 4 July 2022 and was implemented by using the tool LimeSurvey.

Questionnaire design
Before starting, all participants gave written informed consent to participate in the study. After asking a first general question related to support for diverse energy sources, participants were informed that they would have to answer questions about BECCS and DACCS and were asked to self-report based on their knowledge of BECCS and DACCS. Participants also had to provide information regarding thematically relevant publications and their participation in the academic discourse on BECCS and DACCS. Only participants who indicated that they had knowledge of BECCS and/or DACCS (i.e., publication of at least one paper or actively following the academic discourse) were asked the respective technology-specific questions and were considered in the final sample.
Next, all experts were asked to describe their opinion about the geological storage of carbon dioxide. Subsequently, some specific questions regarding BECCS and DACCS were asked regarding the perceived risk of geologically storing carbon dioxide, the cost-efficiency, biodiversity-friendliness, moral hazards associated with, feasibility, necessity of using, and respective support for the technology, as well as their feelings about whether using this technology tampered with nature (see table 1 for an overview of the factors as well as used sources, if applicable). If the participants were experts on both BECCS and DACCS, the question sets related to the two technologies were randomized to minimize order effects. We then asked a question to assess experts' level of optimism about technology, which appeared jointly for all experts. Lastly, experts were invited to answer some questions about their scientific background and demographic characteristics. In particular, participants were asked to describe their current academic position and academic background, as well as their age, gender, and the location of their main workplace. The age was not included in any of the analyses and excluded from the provided data set, as some survey participants remarked that this might reduce anonymity.
During the survey, many items appeared in a randomized order, and some were formulated in a reverse order to reduce response biases. Participants were able to navigate back and forth through the survey, since the goal of this study was not to analyze affective responses but to analyze sophisticated estimates by experts. Apart from some additional open-ended comment fields, experts were forced to answer all questions to complete the survey.

Measured factors
For each factor, two to four items were identified. To ensure that the factors were comparable, all related items were measured on a six-point Likert scale (only the end points were labeled). However, the exact wording of the scale varied between the factors. Cronbach's α was calculated to check for the internal consistency of the items. Regarding the technology specific factors, Cronbach's α was calculated separately for BECCS and DACCS. Whenever Cronbach's α was above 0.6, the internal consistency of the items was considered as acceptable (Streiner 2003), and scales were built by calculating the mean values of the connected items. Except for one factor (perceived biodiversity-friendliness), Cronbach's α was acceptable for all factors. In the case of perceived biodiversity-friendliness, only one item was considered in the analysis. An overview of the measured factors and resulting mean values and standard deviations is shown in table 1.

Sample description
The sample was created in June 2022 based on authorships in relevant papers. By defining search terms such as 'carbon dioxide removal climate change', 'bioenergy carbon capture storage', and 'direct air carbon capture storage', we could query the Scopus database to identify such studies about CDR, irrespective of the methodological approach used. Subsequently, studies were excluded if they were older than 2005 to limit the sample to a set of experts with up-to-date expertise, resulting in a total number of 1,951 identified papers and respective corresponding authors. Whenever an e-mail address was available, an e-mail with an invitation was sent out to these corresponding authors (in total: 1,350). Furthermore, invitations were sent to all first and coauthors of empirical studies on CDR-CCS cited in the present study. After a week elapsed, an e-mail was sent out to all experts who had participated by then and who indicated that they could be contacted with the request to forward the survey link to their research colleagues. Moreover, several reminders were sent to all authors in the final week of June 2022. No financial incentives were provided.
In total, 199 people completed the survey. 179 participants indicated that they knew something about BECCS and/or DACCS. The remaining 20 participants were excluded from further analysis. Additionally, seven participants were excluded based on their short response time (< 5 min). This resulted in a final sample of 172 experts whose answers were analyzed. Of these, 145 participants were identified as experts regarding DACCS and 167 participants as experts regarding BECCS. Of these, 140 participants indicated that they had knowledge regarding both technologies.
The mean participation time of experts was 12.87 min (SD = 7.68), and the average age of experts was 45.03 (SD = 13.31). Of all experts, 132 experts (76.74%) identified themselves as male, 39 experts (22.67%) as female, and one expert (0.06%) as non-binary. Over 90% of experts have their current main workplace in Europe and Central Asia, North America, or East Asia and the Pacific region. Regarding academic experienrce, the biggest group of experts was composed of full professors (27.91%), followed by senior researchers (19.77%), and associate professors (13.37%). By far, most experts indicated that they had an academic background in either Engineering and Technology (36.05%) or the Natural Sciences (31.98%). After that, the two next largest academic background groups were Social Sciences (12.21%) and Economics (7.56%).

Data analysis
The data were analyzed using R (4.2.0). The analysis consisted of three main steps: (i) an analysis of predictors for support for BECCS and DACCS (multiple linear regressions), (ii) a general comparison of the expert's perception of BECCS and DACCS (Wilcoxon signed-rank tests), and (iii) a comparison of the perception of academic background groups (ANOVAs).
First, two multiple linear regression analyses were conducted to investigate which factors predict support for BECCS and DACCS (using a mean value of 'support for research' and 'support for deployment'). As independent factors, three joint factors (no differentiation between BECCS and DACCS) and six technology specific factors (differentiation between BECCS and DACCS) were used in addition to the experts' level of academic experience. The models were checked for multicollinearity and heteroscedasticity (see Supplement A for further information), revealing no signs of multicollinearity between the used factors. Since the data for the DACCS model were heteroscedastic, robust standard errors were calculated and reported instead of the default standard errors in this case.
Second, we tested whether experts who indicated that they had knowledge of both technologies (N =140) perceive BECCS and DACCS differently with regard to the technology-specific factors. Multiple Wilcoxon signed-rank tests were calculated because the data were not normally distributed. Finally, to test whether significant differences existed among the three biggest academic groups (Engineering and Technology, Natural Sciences, and a joint group of Social Sciences and Economics), several one-way ANOVAs without repeated measurement were conducted. Experts who did not belong to one of these three groups were excluded from the ANOVAs. Whenever the ANOVAs indicated a significant difference between the academic groups, the respective dependent factors were tested visually (using QQ plots) for normal distribution, and Levene's tests were conducted to test for homoscedasticity. If these requirements were fulfilled, Bonferroni post hoc tests were performed to analyze which groups differed significantly from each other.

Results
The final sample consisted of a total of 172 experts (145 DACCS experts, 167 BECCS experts, and 140 with expertise on both). Table 2 depicts the results of the regression analysis. The DACCS model explained 65.1% of the variance in support for DACCS and the BECCS model explained 68.9% of the variance in support for BECCS. The perceived feasibility of BECCS and DACCS significantly predicted support for both technologies. In contrast, the perceived necessity was a significant positive predictor of support for BECCS, but not a significant predictor of support for DACCS. Furthermore, perceived tampering with nature was a significant negative predictor of support for DACCS, whereas perceived moral hazards was a significant negative predictor of support for BECCS. Finally, with regard to academic experience, junior scientists support BECCS significantly more than groups with higher seniority (i.e., full professors, research group leaders). Table 3 shows the means and standard deviations of all technology specific factors as well as the results of the conducted Wilcoxon tests for differences between BECCS and DACCS (N = 140, sub-sample of experts on both technologies). We found significant differences (p < 0.001) between the technologies regarding their perceived biodiversity-friendliness, cost-efficiency and moral hazard, with effect sizes ranging between r = 0.36 and r = 0.70. While BECCS was associated with higher cost-efficiency and lower moral hazard, DACCS was perceived to be more biodiversity-friendly.

Differences in evaluation of general and technology-specific factors based on scientific disciplines
One-way ANOVAs show significant differences between the academic background groups with regard to two general factors (see figure 1 for  With regard to BECCS, significant differences were detected between the academic groups related to some technology-specific factors. A one-way ANOVA shows that BECCS is perceived significantly differently concerning its biodiversity-friendliness, depending on the academic groups (F(2.143) = 3.94; p < 0.05). The respective Bonferroni post hoc test results show that social scientists and economists (M = 2.33, SD = 1.19) consider BECCS to be significantly less biodiversity-friendly than engineers and technologists (M = 3.12, SD = 1.39). Moreover, how cost-efficient experts perceive BECCS to be differs significantly between the academic groups (F(2.143) = 7.95; p < 0.001). In particular, social scientists and economists (M = 2.45, SD = 1.16) perceive the cost-efficiency of BECCS as significantly lower than natural scientists (M = 3.13, SD = 1.14), as well as engineers and technologists (M = 3.50, SD = 1.30). Lastly, significant differences were detected between the groups related to the perceived feasibility of BECCS (F(2.143) = 8.54; p < 0.001). Engineers and technologists (M = 4.42, SD = 1.13) perceive BECCS as significantly more feasible than natural scientists (M = 3.73, SD = 1.30), as well as social scientists and economists (M = 3.39, SD = 1.29). In contrast, regarding DACCS, the one-way ANOVAs did not indicate any significant differences between the groups with different academic backgrounds.

Discussion and conclusions
Unlike previous studies, we analyzed perceptions of CDR-CSS held by experts rather than by lay people. While the public opinion is undoubtedly important to consider to improve the acceptance of technology, it has been shown that public opinions can be influenced substantially by misinformation and conspiracy-based criticism (Bolsen et al 2022). Furthermore, the knowledge about CDR-CCS technologies in the general public seems so far to be rather limited ( The results should be considered in light of a few technical limitations related to the survey-based research approach. The sample, albeit quite large for an expert sample, consisted of academics who publish in academic journals or follow the topic from a mainly academic perspective, but did not include (local) stakeholders or policy makers. The grouping into engineers, social scientists, and natural scientists is also rather crude, and the survey was quantitative in nature, meaning that reasons behind the answers and specific contextual factors could not be explored. The study involved a limited set of relevant factors rather than addressing all kinds of conceivable risks related to CDR-CCS (e.g., water resources or energy demand). Nevertheless, the regression models could be applied to predict the variance in support for BECCS and DACCS more broadly and reliably than in previous studies (Jobin and Siegrist 2020), indicating that the survey included the most crucial factors. The present study results suggest that experts show moderately strong levels of support for BECCS and DACCS research and deployment. In contrast to a previous study (Rickels et al 2019), our results show no strong expert consensus and only a medium level of agreement regarding the necessity of the technologies for climate mitigation. Still, it must be noted that the survey took place only a few weeks after the contribution of WG III on the mitigation of climate change to the Sixth IPCC report (IPCC 2022) was published. The relatively new weight placed on CDR-CCS in the IPCC mitigation pathways might also have influenced expert views on the topic.
With regard to support for CDR-CCS in general, the perceived feasibility could be identified as a consistent predictor in both regression models. Experts who see CDR-CCS as technologically feasible tend to support the technologies. The two technologies, albeit having similar levels of support, are perceived differently by the experts with regard to certain aspects. The academic background of the experts influenced the results regarding whether BECCS was perceived as feasible or not; interestingly, no such connection was found for DACCS. Perceptions of technological feasibility might also vary between different world regions (Fridahl and Lehtveer 2018).
Furthermore, the analysis results suggest that junior scientists show higher levels of support for CDR-CCS than researchers with higher levels of seniority (i.e., professors or group leaders). This finding can be interpreted in two ways: Junior experts might be more familiar with recent insights than more senior experts, given the rapid progress in research on CDR-CCS in recent years, but more experienced participants might also have a more holistic view on challenges and problems for the technologies and might, therefore, be more cautious. The available data do not permit us to come to a definite conclusion, and whether age and experience predict how optimistic (or pessimistic) experts are about technology has, to the best of knowledge of the authors, not been sufficiently studied.
The perceived necessity was the second most important predictor of support for BECCS but was not a significant predictor of support for DACCS. The perceived necessity and feasibility have rarely been considered in previous studies, and it remains unclear why the perceived necessity is only a significant predictor for BECCS. Another difference between the two technologies was observed regarding perceived tampering with nature: Support for DACCS was significantly (and negatively) influenced by this factor, but support for BECCS was not. Previous studies found that tampering with nature is a key predictor of lay people's support for NETs (Sweet et al 2021). The present results confirm that tampering with nature can also be a relevant predictor for experts' support for NETs. This indicates that not only lay people but also experts have value propositions that may influence how technologies are perceived (Silva et al 2007).
The lack of observed significant differences in the overall support of BECCS and DACCS contradicts previous findings (Delina 2021). The current study also did not find significant differences between BECCS and DACCS with regard to the perceived feasibility, perceived necessity, or perceived tampering with nature, indicating that the two technologies are viewed similarly in several regards-an observation that also was made with a sample of lay people (Wright et al 2014). The lack of observed differences with respect to perceived feasibility is surprising at first, as the IPCC (IPCC 2022) attributes a much higher mitigation potential to BECCS; however, this might be explained by the fact that the feasibility of large-scale deployment was not quantified or specified in the used questionnaire.
Significant differences in the perception of the two technologies were recorded with regard to biodiversityfriendliness and cost-efficiency. These results are in line with the fact that BECCS is usually considered as a greater threat to biodiversity, but as cheaper in terms of costs per removed ton of CO 2 (IPCC 2022). In previous studies, which aggregated biodiversity-friendliness and cost-efficiency into one factor labeled perceived benefits, this difference could not be detected (Jobin andSiegrist 2020, Wenger et al 2021).
Experts also associated a significantly higher risk of a moral hazard associated with DACCS as compared to BECCS. However, the concern that the future availability of technologies would serve as an excuse to avoid reducing emissions in the present only explained the reduced support for BECCS but not for DACCS. This finding is interesting, because previous results have associated the moral hazard issue with all kinds of NETs (Anderson and Peters 2016). Another study also suggested that a moral hazard effect could concern BECCS and DACCS similarly (Campbell-Arvai et al 2017) and that measures with lower mitigation potentials would possibly have no or lower moral hazards.
The experts' scientific background seemed to influence their perception of BECCS, while they had more homogeneous opinions about DACCS. These differences were particularly significant between the two groups 'Engineering and Technology' and 'Social Sciences and Economics', but they are also small and limited to a few factors. Additional significant differences were only noted between other pairs of groups in two cases: engineers and technologists differ significantly from natural scientists regarding the perceived feasibility, and social scientists and economists differ significantly from both engineers and technologists and from natural scientists regarding the perceived cost-efficiency. Regarding all significant differences, engineers and technologists had more positive opinions towards BECCS than social scientists and economists. One reasonable explanation could be that engineers and technologists are generally more optimistic about technological solutions than social scientists and economists, which has been shown in a previous study in the context of genomics (Hochschild and Sen 2015). In the end, engineers and technologists are trained to find technical solutions, while social scientists and economists are trained to investigate broader social and economic implications. It would be worthwhile to explore these potential reasons for differences in further studies.
Compared to the findings from previous studies with samples of lay people, the present study results In both cases, the perceived feasibility is by far the strongest predictor for support. Nonetheless, policy decisions for or against concrete CDR-CCS projects will require more than only considering experts' opinions and public acceptance: CDR-CCS projects need to be considered holistically. This implies including technical considerations but also regional as well as political contexts. According to the results of this study, experts tend to support the deployment of CDR-CCS as a climate solution, but this does not imply unconditional support. Expert support and public acceptance could still vary regarding the deployment of specific projects. Another open question is if and how far expert opinions change over time. With more experience from larger-scale deployments current assessments might be reevaluated. It would be worthwhile to repeat the study reported here in a few years. Overall, however, the necessity of CDR is likely to continue to outweigh concerns regarding the potential risks, moral hazards, and moral legitimacy.