Mapping Restoration Opportunity for Collaborating with Land Managers in a Carbon Credit-Funded Restoration Program in the Makana Municipality, Eastern Cape, South Africa


A. T. Knight, email


Spatial prioritization techniques are commonly used in conservation planning, but are relatively new for planning restoration programs. Typically, ecological data, and more recently data on economic costs and vulnerability of sites, are used. However, the effectiveness of restoration action ultimately relies on a combination of the appropriate ecological restoration techniques and the human and social dynamics of social-ecological systems. Surveys were conducted with 29 land managers within the Makana Municipality of the Eastern Cape, South Africa, to identify a range of human and social factors hypothesized to define the potential effectiveness of restoration action. Land managers with similar characteristics were grouped using a cluster analysis, and the groups ranked and mapped in geographical information system (GIS) to provide a spatial representation of restoration opportunity. The total number of questions were reduced by 35.6%, a step toward developing a rapid assessment approach for assessing land managers' potential participation in restoration initiatives. Identifying and incorporating human and social factors that directly influence restoration prioritization should promote efficient and effective implementation of restoration actions by the Working for Woodlands programme, who are looking to funding landscape-scale restoration through carbon trading.


Global warming is one of the most pressing issues facing humanity (IPCC 2007). An extensive and growing body of scientific evidence concludes that the vast quantities of green house gasses released from human activities have induced an unprecedented rise in global atmospheric and sea temperatures over the last century (Kerr 2007). Predicted climate change impacts will have far reaching consequences, leaving no country unaffected (Krajick 2004; Kerr 2007).

The Kyoto Protocol provides a framework for countries to collectively combat climate change (UNFCC 1998). Particular attention has been paid to reducing carbon dioxide (CO2) emissions (Kerr 2007; Mann 2008). Legally binding targets exist for reducing CO2 emissions for countries that have ratified the protocol (De Leo et al. 2001). The Clean Development Mechanism (CDM) is one mechanism that facilitates achievement of these targets (Tucker 2001). Carbon trading allows for emission offsets to be purchased and traded on the open international market, which avoids implementing harsh, politically unpopular, financial penalties via strict, legally enforceable emission limits (Tucker 2001).

Carbon trading markets provide opportunities for land managers within the subtropical thicket biome (hereafter “thicket”) of the Eastern Cape, South Africa. A regionally common plant, spekboom (Portulacaria afra), sequesters carbon at comparatively high rates (Mills & Cowling 2006). Spekboom is a succulent shrub endemic to thicket (Vlok et al. 2003), and comprises the majority of biomass in several thicket communities, notably spekboomveld (Vlok et al. 2003; Mills & Cowling 2006). Mills and Cowling (2006) estimated that carbon storage in intact thicket exceeds 20 kg/m2/annum, which is equivalent to mesic forest ecosystems, an impressive statistic given spekboom occupies semiarid landscapes with canopies rarely above 3–4 m. Commercial agriculture, especially intensive livestock grazing (Mills et al. 2005), has extensively reduced spekboom cover throughout most thicket communities (Lechmere-Oertel et al. 2005), with 45% of their extents being severely transformed (Lloyd et al. 2002).

The South African Department of Water Affairs and Forestry (DWAF) actively promotes landscape restoration through the Working for Water (WFW) program (van Wilgen et al. 1998). WFW's primary aims are removal of invasive alien vegetation and restoration of indigenous plant communities to increase water availability for ecosystems and people, job creation and rural upliftment, and nature conservation (DWAF 2004). The Working for Woodlands program, an extension of WFW, has partnered with the Rhodes Restoration Research Group (R3G) at Rhodes University, whose role is to assess the potential for restoring degraded spekboom-dominated thicket (Powell et al. Unpublished), to capitalize on the carbon sequestration potential. This partnership aims to expand WfW restoration activities to provide a range of social and ecological benefits including carbon sequestration and income from trading carbon credits, increased forage for game and livestock, reduced soil erosion, and job creation for disadvantaged communities (Powell et al. Unpublished; Mills & Cowling 2006).

R3G are evaluating the feasibility of restoring landscapes through hand-planting of spekboom cuttings in 50 × 50 m plots on rural properties across a range of thicket communities, and requires information on optimal sites for trial replicates. Effective restoration programs are fundamentally social, as well as ecological, processes (Munro et al. 2005), as land managers fundamentally affect, and are affected by, the land they manage (Sisk et al. 2006). Prioritizing any land management activities, especially restoration, therefore requires identification of the opportunities for, and constraints upon, effectively implementing these activities (Knight & Cowling 2007). This necessitates an intimate understanding of land managers' values, attitudes, and behaviors (Knight et al. 2010), including factors such as willingness to participate, capacity, leadership, social capital, and burnout (Sisk et al. 2006; Knight et al. 2010). Targeting land managers who exhibit characteristics facilitating the effective implementation of restoration programs promotes efficient distribution of R3G's limited resources in both space and time.

We mapped restoration opportunity, based upon a conservation opportunity approach (Knight et al. 2010). Spatial prioritization has been extensively employed in the identification of candidate protected areas (Margules & Pressey 2000), which has been extended to the field of restoration prioritization, although this field is still in its infancy (Crossman & Bryan 2006; Miller & Hobbs 2007). The factors assessed are typically ecological, to target areas with a high likelihood for successfully restoring ecosystem composition and function (Crossman & Bryan 2006).

Although ecological factors are important for restoration prioritization, ultimately, human and social factors determine the effectiveness of implementing restoration programs. Our study mapped the human and social factors defining opportunities for undertaking restoration effectively, which have yet to be incorporated into a spatial restoration prioritization. This study aims to (1) illustrate the importance and utility of assessing human and social factors when mapping restoration opportunity, and (2) provide R3G with spatially explicit direction on land managers representing opportunities for participation in the restoration of spekboom thicket.


Thicket occupies the south-western portion of the Maputaland–Pondoland–Albany hotspot (Steenkamp et al. 2004), primarily in the Eastern Cape province of South Africa (Low & Rebelo 1996). Vlok and Euston-Brown (2002) identified 34 solid and 78 mosaic thicket communities, 13 of which are spekboom dominated. With R3G staff (M. Powell 2008, Rhodes Restoration Research Group, Department of Environmental Science Rhodes University, Grahamstown, South Africa, personal communication) we identified two spekboom-dominated communities—Fish Spekboom Thicket and Albany Spekboom Thicket—as our study area. These were appropriate for restoration from, firstly, an ecological perspective, as spekboom is a dominant, keystone species (Vlok & Euston-Brown 2002) depleted through over-grazing (Mills et al. 2005), and, secondly, from a logistical perspective, that is travel distance for R3G extension staff. The landscape is sub-divided into cadastres (land parcels) which were adopted as planning units. Groups of cadastres are privately owned and managed as properties by individual land managers, primarily for pastoralism, eco-tourism, or hunting of wild game.

Given the comparatively homogeneous natural capital of the two thicket communities forming the study area (Vlok & Euston-Brown 2002), we assessed 12 human and social capital factors hypothesized to define opportunities for, and constraints upon, effective implementation of the R3G restoration program (Knight et al. 2010) (Table 1). Factors were identified on face value (Babbie 1989) for their utility in under-pinning effective restoration, and from scientific literature. We identified seven factors defining human capital, and four factors defining social capital (Grootaert & van Bastelaer 2001), plus a measure of the potential for collaboration between land managers and other stakeholders (Table 1).

Table 1.  Preliminary human and social factors hypothesized to define local-scale restoration opportunity, identified from an extensive literature review.
  1. These were formulated specifically for the context of the R3G restoration initiative, and adapted from Knight et al. (2010).

Human capital
Climate change knowledgeKnowledge of climate change and development of mitigating efforts, e.g. Kyoto Protocol• Knowledge comprises the cognitive component of the tripartite model for describing attitudes (Bohner & Wanke 2002)
  • Land managers with better knowledge of climate change and mitigation efforts may be more likely to adopt restoration initiatives
Restoration knowledgeKnowledge of restoration activities and organizations• Knowledge comprises the cognitive component of the tripartite model for describing attitudes (Bohner & Wanke 2002)
  • Land managers with a better knowledge of restoration activities and organizations may be more likely to participate in restoration initiatives with these organizations
Restoration behaviorParticipation in restoration-friendly activities, such as invasive alien plant removal• Behavior is a component of the tripartite model for describing attitudes (Bohner & Wanke 2002).
  • Behavior is a better reflection of values than attitudes, e.g. a strong stewardship ethic is not linked to increased adoption of best practice land management (Curtis & De Lacy 1998)
  • Land managers already practicing restoration-friendly activities may require fewer incentives to maintain or expand these practices
Willingness to participateIdentifies the restoration instruments and incentives a land manager will and will not engage, and the level of reduced production they will accept• Private land restoration initiatives are often voluntary and so rely on incentives and encouragement, rather than coercion or enforced involvement, as it is in private land conservation initiatives (Tisdell 2004)
  • These require a better understanding of social and economic drivers if needed to engender participation (Tisdell 2004), as well as identifying channels for encouraging participation
Local champion—personalCharacteristics of leadership and drive exhibited by a land manager• Champions are fundamental to leading private land conservation initiatives (Knight et al. 2003)
  • Acceptance by leaders in the community could possibly allow for a wider acceptance of restoration initiatives
Local champion—peersLand manager well-regarded by his/her peers• Allows for the identification of well-respected and potentially influential land managers who could be targeted for primary restoration activities
Willingness to not sellIdentifies land managers who will remain on their land• Allows for the identification of land managers who are not going to sell their land
  • This allows for potential long-term success, as collaboration with land managers could be for an extended period of time
Social capital
Local sense of belongingLand managers' level of trust and the strength of norms of reciprocity and sharing within their community• Land managers who trust and have confidence in each other will probably work more effectively together, and will likely require less input to foster collective action (Knight et al. 2010)
Local networksLand managers' level of involvement in his/her social networks as well as in community institutions and organizations• Effective restoration activity will be enhanced by involvement of the entire community (Knight et al. 2010)
  • Individuals with wide local networks will be better suited to engaging a variety of stakeholders
Broader networksLand managers' involvement with regional, provincial, or national institutions and networks• Land managers actively involved with government initiatives or restoration organizations may be more willing to participate
  • Shows possible reaction to potential “outside” involvement in local affairs
Confidence in governanceLand managers' level of trust in governance systems• Civil and political liberties, political stability and the absence of political violence, and measures of contract enforcement, expropriation risk, corruption, and the quality of government bureaucracy impact upon economies (Grootaert & van Bastelear 2001)
  • Restoration initiatives involve government institutions and as a result confidence in government is crucial to measure
  • Poor confidence in government could lead to failure in achieving Local Agenda 21 initiatives
Willingness to collaborateIdentifies the organizations a land manager will/will not engage and their preparedness to work with them• To be effective, restoration initiatives must be sure with whom land managers are prepared to collaborate
  • Collaboration is fundamental for effective land management initiatives (Wondolleck & Yaffee 2000)

Data were collected through semi-structured interviews with 29 land managers at their residences during August and September 2008. The interview protocol was developed through an iterative process of repeated review with experienced social researchers. Questions for each of the 12 factors were generally Lickert statements, but also included closed- and open-ended questions (Appendix S1). The first land manager interviewed was identified from the telephone directory, with subsequent land managers identified through “snowballing” during interviews (Goodman 1961). Locations of cadastres were monitored as the survey proceeded, and land managers located within selected thicket communities adjoining those already interviewed were prioritized for interviews. Interviews lasted between 45 min and 3 h.

An index or scale was developed for each of the 12 factors. Three coefficients of internal consistency were applied. Cronbach's inline image (Cronbach 1951) is the most common internal consistency test, and was complemented with ReVelle's β, as both coefficients, respectively, reflect scale quality and homogeneity (Cooksey & Soutar 2006a, 2006b). However, Cronbach's inline image over-estimates the proportion of variance within a scale when indicators of general factors are multi-dimensional (Zinbarg et al. 2005), which is common in many datasets. Accordingly, we also calculated McDonald's inline image, the most reliable coefficient of internal consistency (Zinbarg et al. 2005). Scales for internally consistent factors were developed by trialling different subsets of questions for individual factors to find the combination of questions which achieved the highest β and inline image values for the smallest number of questions (Knight et al. 2010). The reliability of each factor (i.e. the degree to which a subset of questions represented the informational content of the full original set of questions for each factor) was then calculated using the RV coefficient (Robert & Escoufier 1976). Reliable factors were then chosen based upon the inline image and RV coefficients. The reduction in the number of questions for each factor was sought to ensure the consistency and reliability of our factors, and to identify the smallest possible subset of questions for each factor, so as to develop a rapid assessment methodology for future application (Knight et al. 2010).

Thresholds of internal consistency depend on whether research is theoretical or applied (Nunnally 1978). Knight et al. (2010) suggest that values for inline image of 0.60 are low in an applied context. Coefficients above 0.80 for inline image and 0.70 for β are regarded as acceptable in an applied context (Rossiter 2002), or 0.70 for inline image in theoretical studies (Nunnally 1978). β values below 0.50 are low and may indicate the presence of sub-scales (Revelle 1979).

Knight et al. (2010) hypothesized that land managers exhibiting similar characteristics require similar investments to encourage effective implementation of conservation action. Accordingly, the data on land managers' human and social characteristics were subjected to cluster analysis, with factors displaying unsatisfactory internal consistency values excluded, applying Ward's minimum variance method of agglomeration (Legendre & Legendre 1998).

Analyses were conducted with R open-source environment for statistical computation and graphics (R-core 2007). Specific packages included (1) psych (Revelle 2007) for calculating Cronbach's inline image, Revelle's β, and McDonald's inline image, plus producing ICLUST plots, which also used Rgraphviz (Gentry et al. Undated) and Graph. Statistica was used to construct cluster figures (StatSoft Inc. 2008).


Identifying Human and Social Factors

Internal consistency of factors was generally satisfactory (Table 2). Of the human capital factors, Climate Change Knowledge ranked high (inline image, inline image, and RV = 0.98). Restoration behavior returned a low, but acceptable inline image coefficient (0.67) and a strong RV of 0.79. Willingness to participate exhibited moderate values (inline image, inline image) and a satisfactory RV (0.74). Local champions (peers) comprised an index summed from land managers' responses and so did not require an internal consistency test. Willingness to not sell included all questions, as they were identified a priori as being internally consistent and expectedly delivered high inline image and β coefficients (both 0.90; Knight et al. 2010). One factor was excluded, Restoration knowledge, as it presented a low inline image coefficient (0.52).

Table 2.  Results of analyses for internal consistency and reliability for factors of restoration opportunity.
 FactorsQuestion ReductionInternal ConsistencyReliability (RV)
inline imageβinline image
  1. *Denotes factors calculated as an index (not a scale) using the full set of questions, rather than the subset of questions to calculate a scale. Measures of internal consistently are: inline image, Cronbach's alpha; β, Revelle's beta; inline image, McDonald's omega. The reliability coefficient is Robert and Escoufier's RV coefficient and represents the degree to which the subset of questions captures the informational content of the full set of questions. Factors in italics denote those excluded from the subsequent cluster analysis to identify land managers representing a conservation opportunity because their coefficients were low.

Human capitalClimate change knowledge11–70.890.660.860.98
 Restoration knowledge4–
 Restoration behavior4–20.850.850.670.79
 Local champion: personal*3–3N/A
 Local champion: peers*1–1N/A
 Willingness to participate14–40.750.540.700.74
 Willingness to not sell*2–2N/A
Social capitalLocal sense of belonging10–40.810.750.810.76
 Local networks*23–23N/A
 Broader networks*13–13N/A
 Confidence in governance13–30.870.840.870.62
 Willingness to collaborate3–30.810.770.81N/A

Of the social capital factors, Confidence in Government ranked the highest among inline image coefficients (0.87). Local sense of belonging also ranked high (inline image and inline image) with a moderate RV value (0.76). Local and broader networks required no coefficients to be calculated, as this index was calculated as the sum of all networks land managers are involved in, and so it is, by definition, internally consistent. Willingness to collaborate exhibited high inline image and inline image coefficients (0.81).

Reduction of questions for factors was effective, with most factors showing a substantial decrease in question numbers (Table 2). Local and Broader networks and Local champion (peers) were not reduced as they are relatively weighted and comprise summed totals. The total reduction of questions was successful with 101 questions reduced to 65—a 35.6% reduction (Table 2).

Defining Restoration Opportunity Factors

Factor values for land managers varied significantly (Table 2). Knowledge of climate change, restoration, and the carbon market was moderate to strong, with 10 land managers scoring 0.7 or higher. Restoration behavior was the lowest scoring factor (mean = 0.22); there is, however, significant variability (SD = 0.34) with some land managers currently engaged in restoration activities. Willingness to not sell was particularly high, with a mean of 0.83 indicating land manager's general intention to remain on their farms. Willingness to participate was relatively high (mean = 0.67). Four land managers (11, 17, 18, and 23, in rank order) were identified by their peers as prominent local champions.

For social capital factors, most respondents felt a strong Sense of belonging in their community (mean = 0.74). Local and Broader networks values were low (mean = 0.43) and very low (mean = 0.23), respectively. Confidence in government was generally low (mean = 0.37) but especially for local government, with confidence in national government higher, but still moderate. Willingness to collaborate was high (mean = 0.72). Participation in local organizations, measured with land managers holding leadership positions or keeping informed about local events, was high.

Mapping Restoration Opportunity

Four distinct clusters were identified at the 1.65 linkage distance (Fig. 1). Factors influencing clustering were (in rank order) Willingness to not sell, Willingness to collaborate, Willingness to participate, and Local sense of belonging. Scores for individual land managers varied within clusters, and clusters were ranked by trading-off the relative importance of different factors for implementing an effective restoration program. This included using qualitative data from interviews. Willingness to collaborate and to Participate were valued highly as they identified candidates most suitable for collaboration with R3G. Clusters were ranked according to restoration opportunity values, with Cluster 1 (Very High Opportunity) including 5 land managers, Cluster 2 (High Opportunity) 4 land managers, and Clusters 3 (Moderate Opportunity) and 4 (Low opportunity) both including 10 land managers. Cluster 1 includes land managers who are very willing to be involved, and who offer wide scope to collaborate with the surrounding community.

Figure 1.

Cluster analysis of land managers based on internally consistent human and social factors of conservation opportunity. Note that the Willingness to not sell factor has been excluded as it is a negative factor of conservation opportunity.

The 29 participants owned approximately 113,000 ha and exhibited varying demographic characteristics. Cluster 1 represents 25% (28,000 ha) of the study area which is managed by only five land managers. Cluster 2 accounts for 13.2% (15,000 ha) of potential restoration opportunity. In contrast, Clusters 3 and 4, which contain 10 respondents each, account for 36% (40,800 ha) and 25.6% (29,000 ha). Average age did not vary widely across the study area or between clusters, with a mean age of 50 represented in both.

The identified clusters were then mapped using GIS. This involved allocating land manager's cadastres to corresponding clusters. Maps depicting the spatial prioritization results are not presented so as to preserve the anonymity of individual land managers.


This study assists R3G to establish a more effective thicket restoration program by providing an understanding of the opportunities for, and constraints upon, implementing effective restoration activities. Applying a spatial restoration prioritization which incorporates human and social data allows the feasibility of activities to be evaluated. The quantification of human and social factors provides R3G with an overview of the social-ecological system, and aids in identifying key land managers for supporting implementation and the incentives they require. Clustering land managers who exhibit similar characteristics allows R3G to adopt targeted approaches to involve groups of land managers in the program, which better ensures active participation by individual land managers. This will enhance the efficiency and effectiveness of the R3G program, ensuring a greater return on their investment, by avoiding wasteful attempts to engage land managers who are not willing, or sufficiently capacitated, to participate effectively (Newburn et al. 2005).

Four distinct clusters of land managers were identified, each displaying distinct characteristics. We note that qualitative data provide essential context for quantitative data, and both types of data should be utilized to rank land managers' potential involvement in private land restoration activities. Land managers within Cluster 1 (see Fig. 1) represent a very high opportunity for effective involvement, displaying strong knowledge of climate change, current involvement in restoration activities, and a strong willingness to collaborate. These land managers should be approached for the pilot project, as they represent the strongest restoration opportunities and manage a comparatively large area (28,000 ha). Effective implementation of Cluster 1 land managers could trigger involvement of land managers in Cluster 2. Clusters 1 and 2 are relatively small groups, which facilitates implementation in the early stages of the restoration initiative. Clusters 3 and 4 represent land managers with moderate to low opportunity for involvement, and so should be implemented last. Each cluster, coincidentally, contains a peer-identified local champion (in rank order, land managers 11, 17, 18, and 23), who could be encouraged to assist in driving restoration activities for each specific cluster. Targeting these champions may promote widespread acceptance of R3G and its policies. Significant investments, however, will be needed to involve all land managers and subsequently achieve restoration targets.

Our results demonstrate a moderate understanding by land managers of climate change, the carbon market, and the carbon sequestration properties of spekboom. This is attributed to the attention that these topics have received in popular literature aimed at commercial livestock farmers in South Africa, for example the Farmers Weekly magazine ( Such publications are a potential communication channel to inform and encourage restoration action. Although the clearing of alien vegetation by land managers is required by South African law, the restoration of degraded lands in the study area has been minimal, primarily due to the rising costs associated with restoration and decreased government support. However, the majority of land managers believed that the overall condition of their land was currently better than it was under the management of the previous generation. This may illustrate an increased understanding of appropriate land management practices, or may be an overly optimistic view driven by personal pride. A high willingness for involvement in restoration programs, however, is encouraging considering the low levels of current restoration behavior, as is the high tendency toward collaborative efforts. All land managers were willing to accommodate R3G with 50 × 50 m test plots, although for most land managers, withdrawing a larger piece of land from production to set aside for restoration would require guaranteed financial incentives. This again highlights the importance of engaging with land managers, as 29 new opportunities for pilot sites were identified. Opportunities for collaboration are further supported by numerous land managers noting the importance of holding leadership positions and attending community meetings, such as local Farmers Associations, which act as forums for collective decision making and interaction. These meetings should be recognized as a primary medium through which R3G can inform local land managers of their activities and encourage participation in their restoration initiatives.

Gathering and applying human and social data present different challenges to collecting ecological data (Knight et al. 2010). Gathering human and social data is expensive and time consuming (Sisk et al. 2006; Knight et al. 2010). Identifying reliable subsets of questions provides a potential solution to these problems. The reduction of questions allows for a more efficient interview process which could be completed within 20–30 min, thus increasing the return on investment of collecting data. Heterogeneous and idiosyncratic individual values and perceptions make reliably extrapolating or predicting factors affecting restoration challenging. A meta-analysis of similar datasets would be needed if these factors were to be reliably extrapolated over a wider geographic area or to different communities.

A significant limitation of this project is that implementation is guaranteed to be more complex than suggested by the simplicity of the four clusters presented from our prioritization. The majority of respondents represent medium to low opportunity, which suggests that significant investments will be required to ensure effective participation. Project funding is also not guaranteed in the long term, making urgent the need to link this initiative with carbon-credit funding soon. Land managers representing high opportunity are spatially dispersed, which limits contiguity of restoration and collaboration between landowners. It is hypothesized that implementation of the initial targeted investments in “champion” land managers will encourage neighboring land managers to join the program. An expectation has been created amongst the community regarding possible financial support which must be delivered upon to ensure that relations between R3G (and other similar initiatives) and land managers prosper.


This is the first spatial restoration prioritization incorporating human and social factors defining the effective implementation of restoration activities. The mapping of restoration opportunity provides a pro-active approach to scheduling land manager involvement in landscape-scale restoration activities, such as the R3G initiative. It demonstrates how human and social data can be gathered and analyzed to promote increasingly effective decision-making for restoration projects by (1) providing a rapid social assessment which delivers understanding of the often idiosyncratic attitudes and behaviors of individual land managers, and their social context, as they affect implementation; (2) actively building relationships between land managers and implementers which enables distribution of information on restoration activities and the potential opportunities these provide in the context of carbon-credit funding; (3) knowledge on the institutions and processes required to sustain the long-term collaborations between researchers and land managers essential for effective restoration initiatives; and (4) efficiently allocating limited financial and human resources. Prioritizations based solely on ecological data will inevitably face the challenges of not understanding the opportunities and constraints revealed through human and social data, meaning implementation is likely to be less effective, less positive, and less harmonious (Knight et al. 2010). These benefits will better allow rapid initial progress as uptake from willing landowners will lend credibility and commitment to restoration initiatives. This rapid initial progress should satisfy stakeholders, including funders, and promote further uptake from less-willing landowners.

Implications for Practice

  • Mapping restoration opportunity using human and social factors increases the feasibility and cost-effectiveness of restoration.
  • Social surveys to gather human and social data offers possibilities for empowering land managers through information provision and involvement.
  • Involving private land managers in regional-scale restoration can provide substantial societal benefits toward ameliorating global warming and through job creation.
  • Funding regional-scale restoration through carbon-credits better ensures long-term funding.


The Department of Environmental Science, Rhodes University, and the Rhodes Restoration Research Group (R3G), notably Charlie Shackleton, provided funding and support. Muhammad Jamaal and Michael Curran provided support with statistical analyses, whilst Gillian McGregor provided GIS advice. Discussions with Mike Powell provided invaluable insights into the theory and practice of carbon trading, and the opportunity to conduct action research. We sincerely thank all the land managers who agreed to participate in this study for being so hospitable and helpful: without them there would not have been a study. Insights from four anonymous reviewers substantially improved the manuscript.