“We will change whether we want it or not”: Soil erosion in Maasai land as a social dilemma and a challenge to community resilience

Soil erosion is a major environmental challenge that undermines economic development in many regions of the world. While much previous work explored physical processes behind this problem, less attention has been paid to social, cultural, and psychological parameters that make a significant impact on soil erosion through the land use practices that they support. The present paper addresses this gap by conducting a qualitative exploration of agro-pastoralist stakeholders’ experiences of soil erosion in northern Tanzania, using the community resilience framework and the social dilemmas approach as theoretical lenses. Interview data suggests that the factors that make communities vulnerable to soil erosion challenges include the centrality of cattle keeping practice to pastoralists’ cultural identity, lack of social cohesion, lack of alternative livelihood opportunities, and weak governance structures. We argue that the ways towards resolving the dilemma lie in addressing relevant cultural norms, building cohesive and open communities, and strengthening local governance. Abstract Soil erosion is a major environmental challenge that undermines economic development in many regions of the world. While much previous work explored physical processes behind this problem, less attention has been paid to social, cultural, and psychological parameters that make a significant impact on soil erosion through the land use practices that they support. The present paper addresses this gap by conducting a qualitative exploration of agro-pastoralist stakeholders’ experiences of soil erosion in northern Tanzania, using the community resilience framework and the social dilemmas approach as theoretical lenses. Interview data suggests that the factors that make communities vulnerable to soil erosion challenges include the centrality of cattle keeping practice to pastoralists’ cultural identity, lack of social cohesion, lack of alternative livelihood opportunities, and weak governance structures. We argue that the ways towards resolving the dilemma lie in addressing relevant cultural norms, building cohesive and open communities, and strengthening local governance.


Reviewers 1 and 2 commented that
is difficult to read. After considering it carefully, we decided to remove this figure from the revised version of the manuscript. We agree that this allows us to present the results in a more streamlined manner.
14. Reviewer 1 commented that Figure 4 uses the colours that are too dark and requires much interpretation. It seems likely that the reviewer meant Figure 5 here, not 4. On consideration, we have decided to retain Figure 5 (now labeled as Figure 4), while changing the colours to improve readability. It visualizes the final and key part of the analysis and aims to present the overall picture of the processes linked to soil erosion. The figure is accompanied by text on p. 29 that aids its interpretation, but at the same time, we feel, adds to the text by visualizing this information.
15. Reviewer 1 suggested that the paper would benefit from proofreading. It has now been proofread carefully, the wording has been adjusted in places, and the mistakes have been corrected.
We believe that these changes have improved the paper considerably, and we hope you agree. Due to paper length restrictions, and in order to make space for all the additions requested by the reviewers, we have had to remove the Limitations subsection. We would be happy to put it back if required. Of course, we are also happy to make any further changes that you or reviewers consider necessary. Thank you again for the helpful direction and for the constructive feedback on our work.

MS Number: JEVP-2019-184
Title: "We will change whether we want it or not": Soil erosion in Maasai land as a social dilemma and a challenge to community resilience" Dear Dr. Scannell, Please find attached a revised version of the above manuscript that we would like you to again consider for publication in the Journal of Environmental Psychology. We would like to thank you and the reviewers for taking the time to comment on our paper, for all the positive comments on our work, and for the critical suggestions. We are very grateful for the opportunity to incorporate your and the reviewers' feedback into a revised version of the manuscript. In the revised version of the paper we have sought to address the points made by yourself and the reviewers. In this letter we outline how we have dealt with the issues raised.
1. Reviewers 2 and 3 point out that the description of the Method is incomplete and does not provide enough detail to ascertain that the process described by Braun and Clarke was followed (Reviewer 2) or to clarify what type of TA was conducted (Reviewer 3). Reviewer 3 also points out that epistemological choices need to be stated upfront (rather than in the Discussion). We agree with the reviewers' criticism on this. In response to it, we have re-written the "data analysis" subsection (pp. 12-13). We have clearly stated the decisions we made as researchers with respect to epistemology and data analysis -specifically, to adopt a realist approach to the data, to conduct theoretical TA (based on the frameworks discussed in the Introduction) while retaining openness to themes not directly informed by the theory, and to look for semantic (rather than latent) themes. We have also provided a focused statement of the research questions we are addressing (p. 10), and we believe that the above methodological choices are justified by allowing us to address these specific research questions directly and in a focused way (this justification is provided on pp. 12-13). We have now also included a step-bystep description of the analytical process in the "data analysis" subsection (p. 13) -this matches the procedure proposed by Braun and Clarke. An effort is made throughout the Method and Results sections to acknowledge the active role of the researchers in the process of analysis.
2. Reviewer 2 listed several criteria from Braun and Calrke's work that they believe were not met by our analysis. Below we justify why we believe these criteria are now met.
Criterium 9 "Analysis tells a convincing and well-organised story about the data and topic". Our analysis is clearly organized, presents a rich and systematic account of the data, and brings the themes together into an overall model in the end ( Figure 4). Indeed, Reviewer 2 seems to agree that this final step of the analysis addresses criterium 9. As the other reviewers note, our account provides some interesting insights on the topic, which suggests that the analysis (at least for some readers and in some of its aspects) is convincing.
Criterium 12 "The assumptions about, and specific approach to, thematic analysis are clearly explicated". We agree that this criterium was not met particularly well in the previous version of the manuscript. We have now clarified the methodological assumptions and the approach taken (see point 1 above).

Criterium 13 "There is a good fit between what you claim you do, and what you show
you have done". Meeting this criterium was not obvious in the previous version due to the lack of detail about the process and assumptions. Now, when this detail is added, we believe that it is clear that the analysis and the stated method are consistent.
Criterium 14 "The language and concepts used in the report are consistent with the epistemological position of the analysis". We have now stated our epistemological position upfront, making clear that we are using a realist approach. We believe that there is nothing in our language use that is inconsistent with this approach.
Criterium 15 "The researcher is positioned as active in the research process; themes do not just "emerge". In the previous version of the manuscript in a couple of instances we referred to themes as "emerging". This language was used to make clear that the themes were not rigidly imposed based on a pre-existing set of theoretical concepts, and that we remained open to themes not directly informed by the theoretical frameworks. This approach is now clearly described in the Method section. We, of course, acknowledge the active role of the researchers in the analysis.
We have now carefully adjusted our language to reflect this, making clear that codes and themes are identified by the researchers through an analytical process guided by a set of epistemological assumptions and methodological choices, which are now clearly stated.
3. Reviewer 2 suggested that the literature on social dilemmas and factors contributing to cooperative solutions should be reviewed in more detail. In response to this, we have significantly extended the overview of previous research pertaining to the three cooperationenhancing factors that we focus on in this paper (group identification, norms, and communication, pp. 6-9). We also clearly justify the decision to focus on these three factors (pp. 6-7) by explaining that we selected socio-psychological factors (group processes) that have been frequently identified as drivers of cooperative behaviour in previous research and are potentially open to influence and change from within a group, providing opportunities for community-led change. In other words, we selected processes that have received a strong support from existing research and theory and offer opportunities for exploring practical implications.
We regret that within the scope of this paper it does not seem possible to provide a broader overview of the social dilemmas literature more generally. We certainly do not "brush it aside" (in fact, we point an interested reader towards some excellent reviews on the topic, p. 6).
However, this is a very large body of literature, and we are restricted by the word limit for journal submissions and the focus of our study. We do believe that reviewing the complete body of work on social dilemmas (rather than reviewing the literature on the three selected predictors of cooperation) would go far beyond the scope of this paper and would distract the reader from its focus. We also do not assert anywhere in the paper that the lab-based research on social dilemmas is "irrelevant" -in fact, we refer to some of it in our literature review. We do say, however, that the lack of field studies is a limitation of this area of research -it is important to test principles identified within the lab in field contexts. This point has been made before by researchers working in the field of social dilemmas, and we cite the relevant sources (p. 9).
Overall, in response to Reviewer's 2 point we considerably extended the literature review on the three factors central for the present paper. We continue to point readers towards the comprehensive general reviews of the work on social dilemmas, and maintain that the focus on laboratory environments is one limitation of previous research that the present paper addresses. 4. Reviewer 2 suggests that the results are not presented in a meaningful way. We have now clarified the presentation of the results in the following ways. A) All themes, subthemes, and brief theme descriptions are presented in Table 1. B) Figure 3 is removed to streamline the presentation. C) The subheadings within the Results section are brought in line with the terms used to describe community resilience framework. We have clarified that that these subheadings are used to organize the discussion of the themes, but they are not the themes themselves (which are presented in Table 1). D) We present the section that corresponds to former Figure 5 (now Figure 4) as a final step of the analysis. We agree with the reviewer that it corresponds to one of the final steps of the analytic process as specified by Braun and Clarke (identifying the overall picture that the data present) and that it meets criterium 9 (telling a well-organized story about the topic). We believe that the presentation of the results has been much improved as a result of these changes.

5.
Reviewer 3 suggests that the paper would be strengthened by drawing some broader theoretical insights from the findings and making explicit their relevance to other contexts, beyond the communities in focus. We agree with this point, and have made a number of additions to the text to address it, clarifying the links with other areas of research and spelling out some implications for broader theory and practice. Specifically, when describing the subtheme of transition to mixed livelihoods as a result of environmental pressures (p. 17), we discussed ways in which this subtheme connects to the broader literature on pastoralist adaptation. In the section describing the subtheme of "limits on migration" (pp. 18) we made clear how our observations contribute to the wider discussion of pastoralist land appropriation. We clarified the link between our conclusions about lock-ins associated with gender roles and research on gendered risk in other pastoralist communities (p. 21), and our conclusions on the role of community identification in cooperation (p. 22) are now explicitly linked to the existing theory and research on the same topic in laboratory contexts. On p. 23 we briefly consider the connection between our findings on perceived risk associated with cattle-raising and crop-growing and the existing work on ambiguity aversion and risk management in pastoralist societies, and on p. 26 we link the findings on the subtheme of "non-confrontational norms" to the theoretical model of the discussion effect in cooperation. We link our observations on adaptation to fast-and slow-onset shocks to wider literature on different pathways to resilience for these two types of environmental stressors (p. 24). Finally, as suggested by Reviewer 3, we specified some other contexts where the findings on the undermining role of pre-existing cohesion for open discussion (p. 31) and on identity content re-framing (p. 32) could be relevant to, and added a paragraph to outline theoretical implications of the community resilience analysis (p. 28).
6. Reviewer 3 suggests that there might be a contradiction between the aim to explore the issue from the stakeholders' perspective and the decision to use pre-specified theoretical frameworks for the analysis. We have now clarified that the aim of the present paper is to address three research questions (p. 10): What are participants' experiences of communal land degradation? How is this problem reflected within different community resilience domains? What processes may hinder or facilitate cooperation around this problem? These questions are addressed from the stakeholders' perspective since we are relying on qualitative self-report data.
At the same time, answering them requires using specific theoretical frameworks for the analysis. This choice has an advantage of producing detailed and focused analysis in relation to the specific research questions. However, we acknowledge that it also entails a possible weakness of omitting themes that do not directly fit within the chosen frameworks. This is one of the trade-offs that accompany analytical choices in qualitative analysis, and we make an effort to be explicit about these. 7. You asked us to comment on the sample size and the extent to which it allows to identify a sufficient diversity of themes. What represents a sufficient sample size in qualitative research is a contested topic, and there are no uniform guidelines. One concept that is frequently used to determine sample size is "saturation" (coined by Glaser and Strauss, 1967) -this is supposed to be reached when adding more observations does not result in new information or a better insight into the topic. Guest, Bunce, and Johnson (2006) conducted an empirical analysis on how quickly a saturation in thematic analysis can be reached, and concluded that saturation occurred within the first 12 interviews, while all basic elements of the themes were present in the first six interviews. Our experience with the data reported in this paper is consistent with Guest et al.'s analysis -we believe that saturation was reached by the end of the interviewing process.
How quickly saturation can be reached (and, therefore, how big a sample size should be) depends on a number of factors. For example, structured interviews and homogenous populations require smaller sample sizes, because saturation can be reached quicker. In our case, a relatively quick saturation is likely to be determined by using structured interview schedules, being guided by the specific research questions and theoretical approaches, as well as by the fact that Maasai pastoralist population in the Monduli district can be characterized as relatively homogenous.
8. Reviewer 1 made several comments about the paper's structure. In particular, they suggested that one of the sections ("transitions") dis not fit well within the structure of the Results section, and that not all subsections included analysis from the perspective of both theoretical frameworks in the same order. We have made changes to the manuscript's structure to respond to these concerns. In particular, we a) removed the "transitions" section, incorporating its content into other subsections, and b) standardized the presentation of the analysis across all subsections -in a summary paragraph for each subsection the results are analyzed first from the community resilience perspective, and then from the social dilemmas perspective. In line with the Reviewer's 1 suggestion we also adjusted the heading for the "governance" subsection to make clear that it discusses themes relevant to one of the resilience domains. 9. Reviewer 1 asked us to define soil erosion earlier. We have now moved the definition to the first sentence of the paper.
10. Reviewer 1 pointed out two sentences that may have been awkwardly phrased. We have now rephrased these sentences (pp. 3 and 30), and carefully edited the whole manuscript to improve the clarity of expression where possible.
11. Reviewer 1 noted that there is no definition for "Resilience Alliance". Resilience Alliance is a label for a collective of people who authored the work that we cited. It is a citation and not a reference to an organization or a concept -consequently, it does not seem to require a definition.
12. Reviewer 1 suggested that it would be useful to connect the social dilemmas approach back to the community resilience approach in the end of the subsection reviewing the literature for the former. We have now done this (p. 9). Figure 3 is difficult to read. After considering it carefully, we decided to remove this figure from the revised version of the manuscript. We agree that this allows us to present the results in a more streamlined manner.

Reviewers 1 and 2 commented that
14. Reviewer 1 commented that Figure 4 uses the colours that are too dark and requires much interpretation. It seems likely that the reviewer meant Figure 5 here, not 4. On consideration, we have decided to retain Figure 5 (now labeled as Figure 4), while changing the colours to improve readability. It visualizes the final and key part of the analysis and aims to present the overall picture of the processes linked to soil erosion. The figure is accompanied by text on p. 29 that aids its interpretation, but at the same time, we feel, adds to the text by visualizing this information.
15. Reviewer 1 suggested that the paper would benefit from proofreading. It has now been proofread carefully, the wording has been adjusted in places, and the mistakes have been corrected.
We believe that these changes have improved the paper considerably, and we hope you agree. Due to paper length restrictions, and in order to make space for all the additions requested by the reviewers, we have had to remove the Limitations subsection. We would be happy to put it back if required. Of course, we are also happy to make any further changes that you or reviewers consider necessary. Thank you again for the helpful direction and for the constructive feedback on our work.

Highlights:
 Soil erosion on Maasai land represents a typical example of the commons dilemma  Difficulty in tackling it is linked to lack of cohesion and cattle ownership norms  Community resilience in this context is weak to moderate across domains  Solutions lie in addressing relevant norms and building cohesive communities Soil erosion: A social dilemma and a resilience challenge 3 Soil erosion is defined as a displacement or wearing away of soil's upper layers, leading to a reduction of its productivity. It is a critical global problem that affects many areas across the world: A recent report by the UN Food and Agriculture Organization suggests that one third of the Earth's land is strongly degraded, estimating the economic cost of this degradation at 17% of the global GDP (UNCCD, 2017). Soil erosion undermines food security and successful economic development, especially in the regions that strongly depend on healthy soils for agricultural production. While much research effort has been invested into exploring the geophysical and geographical factors behind soil erosion (Blakie & Brookfield, 1987;Ionita et al., 2015;Valentin et al., 2005;Wilson & Juntti, 2005), less attention has been paid to social, psychological, and cultural processes related to this issue (cf. Blake et al., 2018;Wynants et al., 2019). In the present paper, we aim to start addressing this gap by conducting a multidisciplinary, qualitative exploration of stakeholder perceptions and experiences of soil erosion in the Monduli district of northern Tanzania and interpreting these in the context of associated socio-psychological and cultural factors. In doing this, we use two theoretical lenses adopted from different disciplines: The social dilemmas approach and the community resilience framework.

The community resilience framework
Community resilience relates to the local or community scale and encompasses the human and non-human resources and capacities within a community to take collective action to deal with problems and determine future development trajectories (Magis, 2010). Community resilience research considers system characteristics and processes through which human agency identifies and shapes collective futures (Wilson, 2012). In particular, this approach focuses on the importance of path dependencies (ways in which current choices are constrained by past Soil erosion: A social dilemma and a resilience challenge 4 circumstances) and lock-ins (sets of circumstances that make certain developmental pathways impossible to implement) to understand how the resilience of communities is affected by environmental, economic, or social challenges. Resilience is usually seen as a normative concept, i.e., a goal that stakeholders can strive to achieve (Wilson, 2017). With this in mind, in this paper we will refer to notions of 'strong', 'moderate' and 'weak' resilience, with strong resilience corresponding to the survival of a community as a cohesive unit able to withstand future shocks and disturbances.
To assess resilience at community level, we will use the conceptual framework developed by Emery and Flora (2006) and Kelly and colleagues (2015) which suggests that social, cultural, natural, economic and political/governance-related domains need to be taken into account to understand how resilient a community is ( Figure 1). Many authors have emphasised the importance of understanding the complex interplay between various domains for assessing community resilience (e.g. Emery & Flora, 2006;Buikstra et al., 2010;Wilson, 2012), suggesting that community resilience will be strongest where there is a balance between domains and they are equally well developed. As Figure 1 suggests, there are close interlinkages between the five domains and they broadly have equal 'weighting' -therefore, weakening one domain (e.g., weakening the social and cultural domains through outmigration) can also affect other domains (e.g., by reducing the availability of social capital for collective action). Building on Resilience Alliance (2007), Wilson (2012), and Kelly and colleagues (2015), in this article we will analyse evidence of community resilience and vulnerability to soil erosion across the five domains in Figure 1.
Soil erosion: A social dilemma and a resilience challenge 5 The community resilience model provides a useful framework for describing and evaluating parameters related to stakeholders' ability to overcome the impacts of soil erosion. At the same time, it represents a generic approach applicable to multiple impacts challenging community resilience and, as such, does not offer a tailored theoretical framework for the specific situation that the communities we study are facing (i.e., a need for collective cooperation around a shared natural resource). The social dilemmas approach was developed specifically to understand (and change) collective behaviour and individual choices associated with management of shared resources. It offers a complementary way of analysing human behaviour factors associated with soil erosion, as well as providing evidence for parameters that can contribute to problem resolution -and, as a result, increase community resilience.

The social dilemmas approach
The social dilemmas approach is a way of describing and theorizing situations where actors face trade-offs between individual and collective benefits. It is applicable to multiple real-Soil erosion: A social dilemma and a resilience challenge 6 life contexts and is widely used in psychology, economics, and other disciplines (for reviews, see Messick & Brewer, 1983;Van Lange et al,, 2014). Several types of dilemmas are described in the literature, but one that is most relevant to the problem of soil erosion in the context of agropastoralism is the "commons dilemma" (or "tragedy of the commons", Hardin, 1968) -in fact, an example most frequently used to illustrate unsuccessful resolution of this type of dilemma describes a (hypothetical) community where cattle grazing decisions lead to pasture and land deterioration. The "commons dilemma" is characterized by a situation where individuals have unrestricted access to a shared resource (e.g., communal land) and have a choice to contribute towards (protection of) this resource or not (e.g., reduce number of cattle grazed or not).
Contributions are always costly and do not affect immediate individual pay-offs from using the resource (e.g., it is costly to reduce cattle numbers) -therefore, an incentive not to contribute is very strong. However, if most community members choose not to contribute towards resource protection, this leads to resource depletion and collapse (Dawes, 1980). In other words, choices based on maximizing individual benefit are detrimental for the shared resource and the community as a whole.
Previous work on commons dilemmas has identified a number of factors that increase the likelihood of cooperative choices (i.e., choices that maximize collective, rather than individual, benefit) and ensure protection and stability of shared resources. While it is not possible to review all of these here (the following sources provide examples of comprehensive reviews: Messick & Brewer, 1983;Parks et al., 2013;Van Lange et al., 2014), several socio-psychological factors that are most relevant to the present context can be identified. Specifically, we chose to focus on group identification, group norms, and effective communication as the three socio-psychological parameters that have been frequently identified as drivers of cooperative solutions and are open Soil erosion: A social dilemma and a resilience challenge 7 to modification from within a group (thereby offering realistic opportunities for community-led change). We provide a brief review of the relevant evidence below.
Group identification refers to a psychological sense of connectedness with, and belonging to, a group (e.g., community, Levine & Hogg, 2009). According to social identity and selfcategorization theories (Tajfel & Turner, 1986;Turner et al., 1987), this connectedness develops as a result of incorporating group membership as an aspect of one's self-concept. Previous research demonstrates that contributions to shared resources increase when there is a strong sense of group belonging. For example, Wit and Kerr (2002) demonstrated that group identification (experimentally boosted within a lab environment) increased concern for collective interest and contributions to a shared resource. Similarly, Rabinovich and Morton (2011) showed that individuals who were given an opportunity to experience a sense of connection with a meaningful group were more willing to protect a natural resource the group shared (see also Heath et al., 2017). The role of group identification in motivating collective environmental action (that can be seen as a response to a large-scale commons dilemma) has received substantial support (e.g., Bamberg et al., 2015). According to the social identity model of collective action, group identification drives cooperative action by increasing collective efficacy and emotional dissatisfaction with the situation that the group finds itself in (van Zomeren et al., 2008).
Another factor that drives contributions to shared resources is cooperative group norms.
Self-categorization theory suggests that individuals who identify with their group internalize group norms and values and adopt these as guiding principles for individual choices (Turner et al., 1987). Biel and Thogersen (2007) provided a comprehensive overview suggesting that activation of relevant social norms contributes to cooperation in social dilemmas related to environmental resources, and Bicchieri (2002) demonstrated that group norm development is a Soil erosion: A social dilemma and a resilience challenge 8 crucial process linking collective discussion of a dilemma with cooperative choices. Earlier experimental work in laboratory contexts shows that developing cooperative consensus and internalizing corresponding group norms is fundamental for motivating decisions that prioritize collective (rather than individual) outcomes (Kerr et al., 1997;Orbell et al., 1988). Similar results were obtained in field settings, where relevant group norms were demonstrated to motivate engagement with sustainable agricultural practice among Australian farmers (Fielding et al., 2008) and decrease overconsumption of energy in the US (Goldstein et al., 2008). There is also an interplay between group norms and identification, such that the impact of norms is stronger when the level of group identification is high (Wildschut et al., 2002).
Finally, there is evidence that effective communication within a group promotes successful management of shared resources (see Sally, 1995, for a meta-analysis; Meleady et al., 2013, for a review). For example, Bouas and Komorita (1996) demonstrated that giving groups an opportunity to discuss a dilemma they are facing before making individual choices decreases the likelihood of shared resource depletion. Other research showed that such discussion effects are due to explicit promise-making (Orbell et al., 1988), and subsequent internalization of the commitment to cooperate (Kerr et al., 1997). Meleady and colleagues (2013) suggest that the discussion effect on cooperation takes place through a series of stages, including group identification strengthening, demonstration of cooperation benefits, and consensus development and internalization -in other words, discussion facilitates the two processes described above (identification and norms). There is also evidence that group discussion with like-minded individuals facilitates collective action for environmental sustainability (e.g., Thomas et al., 2009).
Soil erosion: A social dilemma and a resilience challenge 9 While psychological research on commons dilemmas is extensive and clearly identifies parameters responsible for cooperation around shared resources, much of it takes place in a laboratory context (but see Van Vugt & Samuelson, 1999;Van Vugt et al., 1996), and focuses on one predictor of cooperative choices at a time. It is crucial to extend this approach to field contexts, where both groups and dilemmas in question are established, and choices have vital consequences for individuals and communities -indeed, this is what reviews regularly call for (e.g., Kopelman et al., 2002;Van Lange et al., 2013). It is also important to explore the relevance of existing principles in non-Western contexts, where applying them could have dramatic effects on communities' welfare and resilience. The present paper proposes such an extension by applying the commons dilemma framework to analysing stakeholder experiences related to soil erosion in northern Tanzania. Identifying factors associated with willingness to cooperate around shared resources may open up opportunities for enhancing resilience of communities affected by soil erosion, thereby linking the two theoretical frameworks we use through their practical application.

Present research
The present study is an initial step in developing an understanding of some of the social, psychological, and cultural parameters related to soil erosion in northern Tanzania. Given the paucity of social science research on the topic, we started by conducting a qualitative investigation of stakeholders' accounts of the soil erosion problem, current land use practices, and perceived barriers and opportunities for adopting alternative practices, as well as social, cultural, and governance context framing these. The first part of the analysis used the community resilience framework to assess stakeholder resilience to soil erosion in their area. The second part used the commons dilemma framework to explore parameters associated with stakeholders' Soil erosion: A social dilemma and a resilience challenge 10 willingness to cooperate and take collective action in the wake of threats to the shared resource.
We aimed to addressed the following research questions: What are participants' experiences of soil erosion on their land? How is this problem reflected within different community resilience domains? and What processes may hinder or facilitate cooperation around this problem?
The study was conducted in the Monduli district of northern Tanzania, part of the East African Rift System (see Figure 2). The area suffers from significant soil erosion problems -a recent study demonstrated that over the last two decades, soil aggregate stability and infiltration capacity have been significantly reduced, which has led to sheet wash erosion and pronounced gully development (Blake et al., 2018). Maasai pastoralists constitute a large proportion of the area's population, and it is these communities that suffer most from soil erosion issues. While changing weather patterns may be a major factor contributing to this, land use practices also exacerbate the problem. The present study focuses on barriers and opportunities for changing these from the communities' perspective, while grounding the analysis in a wider set of parameters contributing to community resilience or vulnerability.
Soil erosion: A social dilemma and a resilience challenge 11

Participants and procedure
A qualitative approach was used, with semi-structured interviews as the method of data collection. Seventeen respondents from the Monduli district took part in the study (13 male, 4 female). Most participants were Maasai pastoralists and farmers (n = 14), the remaining interviewees were representatives of farmer organisations or local government officials advising Maasai communities on agricultural practices. Interviews were conducted in participants' home villages. For cultural reasons, responsibility for inviting community participants was devolved to local village leaders. They were asked to approach community members engaged in pastoralist or agri-pastoralist activities and invite them to participate on a voluntary basis, aiming for a gender Soil erosion: A social dilemma and a resilience challenge 12 balance wherever possible. Informed consent was sought from participants. After each interview was completed, participants were given the opportunity to ask questions and were provided with an information sheet.
Most interviews (twelve) were conducted by a member of the UK research team, with translation and cultural assistance from a member of the Tanzanian team, following guidance by Weiss (1994). The remaining five interviews were conducted by the Tanzanian team who were trained in interview techniques during the first stage of the data collection. Each interview lasted between 30 and 100 minutes. Interviews were audio-recorded and transcribed verbatim (in Swahili), using basic transcription (i.e., capturing all verbal, but not paraverbal information). The transcripts were subsequently translated into English. The interview schedule focussed on the following main topics: Community livelihoods, land use practices, soil erosion, community life, and cultural identity.

Data analysis
We employed a thematic analysis approach (Braun & Clarke, 2006) for data analysis. We chose to conduct theoretical thematic analysis, where the coding process was primarily driven by the two theoretical frameworks used (community resilience and social dilemmas frameworks).
This choice was made in order to give the analysis a specific focus and to provide a more detailed account of themes related to resilience and socio-psychological processes involved in the shared resource management, in the context of soil erosion. At the same time, we aimed to maintain a degree of flexibility, with some codes (and, subsequently, themes) being driven by our perception that they are central to participants' experience and essential for framing the focal issue. We used realist approach to the data, assuming that participants' use of language reflects their experiences and perceptions. This choice of epistemological approach was made with the Soil erosion: A social dilemma and a resilience challenge 13 aim of exploring processes framing the problem of soil erosion, based on the assumption that participants' accounts provide some access to such processes. In line with this approach, we coded for semantic themes, identified within the explicit meanings articulated by the participants.
The analytic process started with a careful reading of the complete dataset, followed by initial coding of data excerpts. We then sorted these codes into initial themes, and reviewed these using criteria of internal homogeneity and external heterogeneity (cf. Patton, 1990, see Table 1 for the final set of themes). Both processes of coding and theme identification were guided by the chosen theoretical frameworks. The dataset was re-read at this stage to make sure that the identified themes describe it well. Finally, we identified how the final themes fitted together to provide an overall account of the data. Below we present an overview of the main themes, structuring it according to the resilience domains (shown in Figure 1). This is followed by a reflection on the links between the themes and a construction of an overall account (Figure 4).
Excerpts from the interviews are referenced throughout by a letter showing participant gender and transcript number.

Natural and economic domains: Problem awareness and alternative livelihoods in the context of environmental and socio-economic change
As noted above, the extent of the soil erosion problem in the Monduli district is significant.
Participants demonstrated a high level of awareness of this issue. Many of them expressed strong concern about soil erosion and the implications that it has for their livelihoods and their children's future: "We are now worried that our children won't have a place to farm and graze their cattle" (F5). When asked about signs of erosion, participants mentioned gullies forming and increasing in size, but also soil loss on their farms, and accumulation of sediment in water Soil erosion: A social dilemma and a resilience challenge 14 reservoirs. Participants spoke about a range of solutions that they practice, directed both at adaptation to existing erosion (e.g., filling gullies with branches or manure) and mitigation of future damage (e.g., building barriers on farmland, using contour cultivation, hole planting, chemical weeding). At the same time, none of these solutions was perceived as sufficiently effective: "As a community we don't have any technique or knowledge … to solve the soil erosion problem" (F5). Overall, despite high awareness of the soil erosion issues, community resilience in the natural domain can be characterised as weak.
As is typical in Maasai communities, pastoralists in the Monduli district placed great emphasis on cattle as a valuable economic and socio-cultural resource (cf. Warren, 1995).
Participants talked about cattle as a 'liquid asset', which can be used as a buffer when needed during times of environmental or economic disturbances. Some interviewees stressed that other sources of sources of livelihood (i.e., crop farming) are unsuitable for this purpose and used this lack of immediate liquidity as an argument against alternative sources of livelihood: "I don't think that [shifting to farming] will happen because we use cattle as an alternative to farming for getting money whenever there is an emergency. In such situations you cannot wait until you sell crops because crops take a long time. So most people sell cattle and after they harvest they sell crops to replace the cattle..." (M8). Other barriers to economic diversification are lack of alternative skills, as stated by some participants: "Many of us know nothing about other business than cattle keeping…" (M7) and unavailability of dependable markets, as well as delays in payments from (commercial) buyers. For example, one participant (M17) commented on a failed initiative to tap into the global flower trade: "The price for dry flowers remained unchanged for over three years, which meant that some farmers have stopped growing those flowers. They also don't give you money on time, it can take up to three months to get all your money." From a community resilience perspective, the economic domain is characterised by pronounced lock-ins and path dependencies. While some participants expressed a willingness to embrace wider market possibilities (e.g., production of dry flowers mentioned above), a combination of geographical and socio-economic constraints has hampered the development of a more multifunctional economic base. Given that the main economic asset (cattle) is under significant threat from changing environmental conditions, the economic resilience of Maasai communities in the Monduli district can, therefore, also be described as weak.
The threats discussed above relating to the natural and economic domains are overlaid by higher spatial level demographic, social, and economic transitions that have been taking place in the recent decades. The first of these is the partial shift from cattle-keeping to mixed livelihoods (supplementing cattle-keeping with small-scale crop farming). The lack of pasture land (due to erosion and population growth) and changes in weather patterns were mentioned as the main reasons for this shift: "Our parents have been dealing mainly with livestock keeping, (but) due to …shortage of rain and shortage of pastures we decided to shift to agriculture" (M8). While this transition is a sign of vulnerability to changing environmental factors, it also demonstrates capacity for adaptation. This observation is consistent with previous research on pastoralist capacity for adaptation (e.g., Campbell, 1999;Huho et al., 2011;Opiyo et al., 2015;Wangui, 2018), and contributes to extensive empirical evidence on adaptability to changing environmental conditions noted in other pastoralist communities in different regions of the world. The shift to mixed livelihoods may, however, itself indirectly increase pressure on communal land and affect social cohesion within communities, as we will see below.
The second shift is related to migration patterns. Previously, Maasai families led a nomadic life-style, following the rains with their cattle. However, most families now remain in  Desmarais et al., 2015), weakening willingness to invest efforts in protecting shared pastures from erosion. (F1). At the same time, some interviewees suggested that this support was available in times of crisis (such as death in a household or severe drought), but outside of these situations they were mostly 'on their own'. Participants mentioned that if they asked neighbours for help, in most cases it would be given, but it would not necessarily be freely offered otherwise. There was also a gendered aspect to participants' reflections on connectedness and social support within their communities: Some interviewees mentioned strong support between women of the same community (see above), while others discussed the negative impact of the Maasai patriarchal system: "…the challenge is that [when a woman] harvests, the household head takes the whole produce to the market and leaves the spouse with nothing. Based on our culture, it's difficult for a woman to ask the household head where the produce was taken. Women toil but men benefit from women's hard work." (M1). It seems, therefore, that group identification and social support in many communities are split along gender lines: Connectedness within gender groups (especially among women) may be very high, but this does not necessarily lead to the whole community being bound together.
Overall, analysis from a community resilience perspective suggests that the social domain is partly characterised by negative lock-ins and path dependencies related to weak social support in some communities, and gendered power distribution across all of them. This latter finding is in line with the existing research on gendered risk distribution and vulnerability in pastoralist communities (e.g., Goldman et al., 2016;Talle, 1988) and calls on gender mainstreaming in international development more generally (e.g., Beneria et al., 2015;Parpart, 2014). At the same time, some evidence of cooperation within communities (and especially within gender groups) was also evident. Based on this evidence, the social domain can be characterised as moderately resilient.
Soil erosion: A social dilemma and a resilience challenge 22 Analysing the above evidence from a commons dilemma perspective, it can be concluded that community identification is often insufficiently strong for group members to recognize the degradation of the communal land as a priority that needs addressing. This conclusion is linked to an earlier observation that the transition to private land ownership shifted households' sense of responsibility away from communal land. Weak identification with fellow villagers and lack of open communication about the issue may represent a barrier to cooperation needed to address the soil erosion problem. These findings are consistent with the existing research on the role of group identification in cooperation in social dilemmas (e.g., Wit & Kerr, 2002;Rabinovich & Morton, 2011) and in collective environmental action (e.g., Bamberg et al., 2015), demonstrating that this parameter is likely to play an important role in the context of communal land management.

Cultural domain: Social norms, cultural identity content, and pathways to change
Some interviewees highlighted that Maasai cultural identity is closely associated with traditional extensive (transhumance) pastoralism and livestock rearing. Cattle were described as a central component of Maasai collective identity content (i.e., a sense of what it means to be a member of this cultural group). For example, one interviewee said: "…it is impossible to abandon cattle-keeping and still be called a Maasai! If you do that you are not a Maasai!" (M6), suggesting that it is impossible to retain one's cultural identity while shifting to a different type of livelihood. In addition, the size of herd appears to be a significant marker of a male Maasai person's status. One participant described the historical link between status and herd size in the following way: "If I hear that someone has 1000 cattle I must struggle to have 5000! We are competing to have more cattle than anybody else." (M6). These excerpts point towards a strong embeddedness of the practice of large herd keeping within the interviewees' collective identity, Soil erosion: A social dilemma and a resilience challenge 23 and the centrality of this practice for maintaining status within the cultural group, in line with existing accounts of Maasai culture (e.g., Fratkin & Mearns, 2003;McPeak et al., 2011).  (Bryan, 2013;Liebenehm & Waibel, 2014), and suggests that risk perceptions may lag behind the ongoing environmental change -for example, an increased risk of losing cattle during droughts may not be recognized.
Consequently, enhancing resilience may require adjusting risk perceptions associated with traditional and alternative livelihoods -a suggestion that may be of relevance to different types of societies undergoing environmental or social change.
There is some evidence that such an adjustment is happening for some participants. For example, some interviewees talked about their recent experience of drought, and the fact that this has changed how they think about cattle keeping: "If you advise us to reduce the number of livestock we will listen, because we normally lose many cattle during dry season" (M6). This Soil erosion: A social dilemma and a resilience challenge 24 experience seemed a stronger driver for change for some respondents than the soil erosion itself: "Interviewer: …might you think about reducing your livestock numbers because of soil erosion?
Interviewee: No, but I will do it because of drought." (M7). In resilience terms, this highlights that respondents may be more willing to adjust to fast-onset shocks such as droughts, but less willing (or able) to adjust to slow-onset disturbances such as soil erosion. This observation is consistent with the suggestion that different adaptation pathways may be activated for coping with fast and slow onset hazards and, consequently, societies may demonstrate different resilience levels with respect to each of these (Cutter et al., 2008;Handmer & Dovers, 1996).
Another pathway to change is offered by children's school education, and the influence they exert through this on their parents: "Children told us that keeping a big number of cattle isn't wealthy because at some point a big number will die due to drought and diseases" (F1).
Norms about sustainable ways of crop farming (e.g., hole planting, terracing) also seem to be changing through intra-community observation of successful

Political/ governance domain: Local governance and norms of decision-making
The data suggest that collective decision-making plays an important role in community functioning, although decision-making forums are not always used to discuss soil erosion issues.
At the same time, some interviewees pointed towards distinct difficulties in the domain of community governance. For example, a participant (M11) talked about "devolution of power" (i.e., the transfer of powers and responsibilities from higher-level government structures to the community level), and the consequences this has for over-exploitation of natural resources (e.g., forest grazing and clearance -a practice that has detrimental consequences for soil erosion): "…there is what they call a 'community approach to conservation'. within a community seems to be one of the key norms that is prioritized above natural resource protection.
The above analysis suggests that local governance of natural resources is associated with a number of problems and can be described as weak. While formal opportunities for managing resources locally exist, these seem to be undermined by Soil erosion: A social dilemma and a resilience challenge 27

Community resilience overview
Overall, the community resilience assessment has revealed a mixed picture (see Figure   3), where none of the domains could be characterized as strongly resilient. The social domain emerged as moderately resilient in line with cognate studies that discuss social capital in Maasai and other livestock herding communities (e.g. Fratkin & Mearns, 2003;McPeak et al., 2011;Silver, 2009) but with some variability in the strength of social connectedness and support noted across communities. The cultural domain also emerged as moderately resilient, largely due to strong lock-ins associated with male-focused cultural importance of cattle ownership as a status symbol. Echoing existing studies on Maasai culture, it seems culturally inconceivable for some participants to give up their cattle herds for alternative livelihood means (Fratkin & Mearns, 2003;Galvin, 2009). The most problematic domains, however, are the economic, governance, and environmental ones. The weakness of the economic domain is linked to negative lock-ins associated with a strong dependence on livestock grazing and lack of sustainable economic alternatives for more multifunctional rural livelihoods (cf. Wilson, 2012), while the local governance domain could be a potential stumbling block for successful implementation of policies and regulations alleviating soil erosion.
Soil erosion: A social dilemma and a resilience challenge 28  Figure 4. The factors above are exacerbated by the fact that cattle keeping constitutes a key element of Maasai cultural identity content and, consequently, the discourse of change represents a significant social identity threat (cf. Branscombe et al., 1999). At the same time, there is evidence of normative (and possibly identity) change taking place, through the pathways of peer influence, direct experience, and inter-generational transfer of education.
Soil erosion: A social dilemma and a resilience challenge 30 Overall, the above analysis is broadly consistent with previous work that discusses group identification (e.g., Wit & Kerr, 2002), social norms (e.g., Bamberg et al., 2015), and discussion opportunities (Meleady et al., 2013) as some of the central social psychological predictors of cooperation around shared resources. Our analysis suggests that these factors bear relevance to groups' ability and willingness to cooperate beyond experimental laboratory contexts, and may play a significant role in real world environments where ability to solve shared resource dilemmas has crucial implications for communities' livelihoods. The finding that avoidance of confrontation undermines protection of the shared resource may seem inconsistent with the claim that group identification and cohesion are essential for cooperation. In the present study, Soil erosion: A social dilemma and a resilience challenge 31 prioritising cohesion led community members to leave encroachment on the shared resource unchallenged, even when such behaviour was inconsistent with their personal norms. This represents an important nuance in the role of group identification in solving social dilemmascontrary to previous work, this role may not always be positive, especially when it prevents an open discussion of the issue. It is worth noting that existing models (often based on laboratory research) assume that group identification is developed in the process of discussing a dilemma (e.g., Meleady et al., 2013). The present study demonstrates how pre-existing community links and the value attached to them may represent a barrier for such discussions -a finding that may have implications for other contexts where groups facing social dilemmas are established and strongly interdependent (from fishing villages to deprived urban neighbourhoods).

Practical implications
Despite the explorative nature of the present research, it suggests a number of promising directions for addressing Maasai communities' willingness and ability to protect their land from further degradation. One crucial parameter is the centrality of livestock keeping to Maasai cultural identity. To make livelihood diversification conceivable, this possibility needs to be presented in such a way that it does not constitute a threat to cultural identity. One way of achieving this could be framing cattle keeping and nomadic lifestyle as practices that were making Maasai people resilient to harsh environmental conditions in the past, and presenting this adaptability and resilience (rather than the livestock keeping itself) as defining features of Maasai culture. Coupled with a clear understanding of changing environmental and social conditions (that undermine traditional ways of maintaining resilience), such re-framing could make adoption of alternative livelihoods more acceptable, since this would be consistent with maintaining adaptability as a key component of collective identity. Future research could explore Soil erosion: A social dilemma and a resilience challenge 32 this possibility further, focussing on context-appropriate approaches to re-framing identity content and producing corresponding normative change. The principle of identity content reframing could also prove relevant in a number of other contexts, including intergroup conflict (cf. Livingstone & Haslam, 2011).
Another dimension that needs to be addressed is enhancing community identification while opening up channels for discussion of soil erosion issues. Our analysis suggests that cooperation and agreement are highly valued in Maasai culture. These values can be harnessed to create opportunities for facilitated open discussion where differences in opinion are heard and accepted, rather than silenced. Such discussions may pave the way to building stronger group identification, that can then serve as a basis for developing cooperative solutions to the soil erosion challenge -for example, through enhancing a sense of collective efficacy and giving rise to the development of new norms and community self-monitoring systems.
The above steps, supporting the development of a socio-psychological basis for change (in the form of cultural identity content, community identification, and norms) need to be accompanied by work in three more directions: 1) support for education on processes and causes of soil erosion, 2) co-development of viable livelihood alternatives, and 3) support for strengthened local governance. The first of these steps should enable increased understanding of the link between certain land use practices and soil erosion, creating motivation for change; the second should make such change economically viable; while the third would co-design local institutions to support the transition. Achieving this is an ambitious task that requires a largescale multi-disciplinary intervention programme. Recognition of the need for change demonstrated by the study communities provides ample evidence that such work could enable a real and sustained transformation.