Cultural valuation and biodiversity conservation in the Upper Guinea forest , West Africa

The cultural valuation of biodiversity has taken on renewed importance over the last two decades as the ecosystem services framework has become widely adopted. Conservation initiatives increasingly use ecosystem service frameworks to render tropical forest landscapes and their peoples legible to market-oriented initiatives such as REDD+ and biodiversity offsetting schemes. Ecosystem service approaches have been widely criticized by scholars in the social sciences and humanities for their narrow focus on a small number of easily quantifiable and marketable services and a reductionist and sometimes simplistic approach to culture. We address the need to combine methods from each of the “three cultures” of natural science, quantitative social science, and qualitative social science/humanities in conceptualizing the relationship between cultural valuation and biodiversity conservation. We combine qualitative data with forest inventories and a quantitative index of cultural value to evaluate the relationship between cultural valuation and biodiversity conservation in Upper Guinea forest in Liberia, West Africa. Our study focuses on “sacred agroforests,” spaces that are associated with Mande macro-language speaking groups such as the Loma. We demonstrate that sacred agroforests are associated with different cultural values compared with secondary forests. Although biodiversity and biomass are similar, sacred agroforests exhibit a different species composition, especially of culturally salient species, increasing overall landscape agro-biodiversity. Sacred agroforests are also shaped and conserved by local cultural institutions revolving around ancestor worship, ritual, and the metaphysical conceptual category “salɛ.” We conclude that to understand the relationship between cultural valuation and biodiversity conservation, interpretivist approaches such as phenomenology should be employed alongside positivist ecosystem service frameworks.


INTRODUCTION
The cultural valuation of biodiversity has taken on renewed importance over the last two decades as ecosystem services frameworks have become widely adopted, with the most influential and well-known being presented in the Millennium Ecosystem Assessment (MEA 2005).Conservation initiatives increasingly use ecosystem service frameworks to render tropical forest landscapes and their peoples legible to market-oriented schemes such as REDD+, PES, and biodiversity offsetting in terms of the provisioning, e.g., food and water; regulating, e.g., climate and disease; supporting, e.g., nutrient cycling and pollination; and cultural, e.g., spiritual and recreational, benefits they confer to society (Costanza et al. 1997, MEA 2005, Armsworth et al. 2007, Naidoo et al. 2008, Corbera 2012).Ecosystem service approaches have been widely criticized by scholars in the social sciences and humanities, however, for their narrow focus on a small number of easily quantifiable and marketable services and a reductionist and sometimes simplistic approach to culture (Robertson 2011, Dempsey and Robertson 2012, Kirchhoff 2012, McAfee 2012, Pröpper and Haupts 2014, Schnegg et al. 2014, Winthrop 2014, Plieninger et al. 2015).A major problem is the assumption that cultural services can be quantified, and then be correlated to ecological structures and functions (Daniel et al. 2012, Russell et al. 2013).Although this permits integration with ecological data, it misses the vital point that important dimensions of cultural valuation, e.g., the religious or the sacred, cannot be reduced to ecological objects, empirically observable through positivist methods.More profoundly, the conception of culture as a category of value that is completely separate from the material values of other ecosystem services is a Western, post-Enlightenment, Cartesian phenomenon (Kirchhoff 2012, Pröpper andHaupts 2014).Non-Western peoples often do not separate nature and culture as does the Western lay and scientific thinking behind ecosystem services frameworks (Ingold 2000, Latour 2009, Viveiros de Castro 2012, Descola 2013).Forms of cultural valuation can be fundamentally different to economic valuation, more of a "processual activity of meaning-making" rather than something commensurable to a market assigned monetary value (Pröpper and Haupts 2014, Winthrop 2014, Schnegg et al. 2014).It is now recognized in the conservation literature that understanding the cultural valuation of biodiversity requires interpretivist as well as positivist social science theory and methods (Adams 2007, Sandbrook et al. 2013, Moon and Blackman 2014).Understanding the biocultural relationships through which cultural values shape tropical forest diversity is now increasingly recognized as important for the conservation of both biodiversity and (tangible and intangible) cultural heritage (Gavin et al. 2015).This means investigating the full range of cultural values, from, for example, more utilitarian (Ickowitz et al. 2014) to more symbolic (Dold andCocks 2012, Cocks andWiersum 2014).In trying to understand this diversity of cultural valuation, we attempt to overcome problems raised by Kirchhoff (2012) and Turnhout et al. (2013) associated with conflating economic and cultural valuation by combining methods from each of the "three cultures" of natural science, quantitative social science, and qualitative social science/ The research presented in this paper focuses on a regionally specific example of this phenomenon, which we term "sacred agroforests" (SA), in the Upper Guinea forest of West Africa.SA are located at the sites of ruined towns belonging to Mande macro-language speaking groups who have occupied the region for more than 500 years (Brooks 1989).Owing to considerable cultural continuity in the region, despite historic and recent wars, current populations are normally descendants of the historical occupants of these towns, and are conferred tenurial rights by ritualized relationships with their ancestors (Fraser et al. 2015).We focus on one group in particular, the Loma.For the Loma, like other Mande speakers and many other societies of sub-Saharan Africa (Fortes 1965, Calhoun 1980), ancestors are present and have social roles in the world of the living.This reflects the wider tendency in sub-Saharan Africa for the religious and the social to be fundamentally intertwined (Bloch 2008).It is through relationships with ancestors that Loma descendants can access land through status and tenurial rights.These sites are both ritually and symbolically associated with ancestors through graves and certain trees.Ownership normally lies with the male head of a single family within the population of descendants.SA are sometimes used by gender-specific initiation societies (Poro for men, Sande for women), and for ritual and religious activities including divinations and sacrifice to protect individuals and groups (Little 1965, 1966, Leopold 1991, Hojberg 2007, McGovern 2012).These factors mean that clear cutting and burning these sites is strictly forbidden in customary law, thereby conserving plants and trees growing on them.By contrast, the surrounding natural old fallows are not preserved by any cultural laws and may be cut for shifting cultivation or logging.
Although the sacredness of SA is related to the symbolic presence of ancestors, their use as agroforests is related to material practices of ancestors.The activities of ancestors transformed the soil through deposition of charred and fresh organic materials at loci of domestic and food processing.These actions created, over time, anthropogenic soils called "African dark earth" (AfDE) that have double to triple the amount of organic carbon, significantly higher pH, cation exchange capacity, and more plant-available nutrients when compared with background Oxisols (Solomon et al. 2016).SA feature a canopy of culturally significant mature rainforest tree species, and a midstory of cocoa (Theobroma cacao), kola (Cola nitida), and various native and exotic fruit trees.AfDE is attractive for planting cocoa in Liberia because upland soils are extremely infertile; this tree crop normally only grows well in this country on fertile but limited lowland soils.Conversely, Côte d'Ivoire and Ghana have a much more diverse mix of upland soils (Jones et al. 2013), making cocoa cultivation much more viable on upland soils in these countries.
The historical-ecological phenomenon of SA in NW Liberia therefore provides an interesting and hitherto unexplored arena within which to examine the role of cultural valuation in shaping and conserving biodiversity.Studies of sacred forests in West Africa have tended to focus on cultural dimensions with less attention to ecological processes (Fairhead andLeach 1994, Chouin 2009).Studies of agroforests in West Africa tend to be purely ecological and lack recognition of the legacy effects of historical human agency on current species abundance (Oke and Odebiyi 2007, Sonwa et al. 2007, Bisseleua and Vidal 2008, Correia et al. 2010), despite the fact that the distribution of several species found in the agroforests reported in these studies is heavily influenced by historical cultural practices.Forests on anthropogenic soils in Africa, meanwhile, have not been http://www.ecologyandsociety.org/vol21/iss3/art36/Fig. 1.Location of Wenwuta in Liberia (inset top left), and the sacred agroforests and old fallows sampled in the vicinity of Wenwuta.The areas of sacred agroforests and settlements were mapped with GPS in collaboration with local people.described, and only a handful of studies describe such phenomena in South America (Paz-Rivera and Putz 2009, Junqueira et al. 2010, Fraser et al. 2011a,b, Quintero-Vallejo et al. 2015).This article and the research it draws on aim to understand the relationship of cultural valuation to biodiversity conservation in the context of SA.First, we carried out a regional survey to determine the extent and typology of SA in NW Liberia.Second, we conducted a case study drawing on participant observation and open interviews to investigate how the cultural values as processes shape natural resource management practice.Third, we compared the ecological characteristics of the vegetation (density of individuals, basal area, species richness, floristic composition) of SA with unmanaged 45-65-year-old fallows (OF).We use OF as a baseline for our comparisons because although they are anthropogenic forests to some degree, they are more similar to natural old-growth forests because they are largely unmanaged.Fourth, we employ an index of cultural valuation to quantify utilitarian and symbolic cultural valuation of different species.Based on a prior literature review and initial observations, we formulated the following research questions: 1. How widespread are SA in NW Liberia?
2. What role do social or religious institutions play in conserving SA?
3. How does the species composition and cultural valuation of SA compare with that of OF?

Regional survey
To assess the frequency and distribution of SA we conducted an initial regional survey in four counties of NW Liberia: Gbarpolu, Bong, Lofa, and Nimba (Fig. 1, Appendix 1).We selected these counties and the areas visited within them based on descriptions of regions that were heavily populated during the 19th century (Fairhead et al. 2003) that we assumed would have the most prevalent anthropic influence and therefore the greatest area coverage of SA.During our regional survey we asked community leaders how many SA there were in the vicinity of their town or village, and visited the most easily accessible one at each community.In total we visited 83 different localities and at 51 of them we found at least one SA, 94 in total (Appendix 1).It emerged that the local term for SA is "old town spot," confirming that locals are well aware of the anthropogenic origins of these spaces.

Case study selection and participatory mapping of sacred agroforests
Following the initial survey, Wenwuta, a Loma settlement in the southern Zorzor district of Lofa County, NW Liberia (Fig. 1), and its surrounding landscape, were selected for an in-depth case http://www.ecologyandsociety.org/vol21/iss3/art36/study because this relatively remote area is characterized by a high degree of cultural continuity when compared with other regions of NW Liberia.We sought such an area because we wanted to investigate the relationship of cultural valuation to biodiversity conservation away from areas of rapid land use change closer to urban centers.The region has an average rainfall of 2900 mm (majority of rainfall occurs between July and November) and is capable of supporting wet evergreen forest (Poorter et al. 2004), although today there is no true old-growth forest in a landscape long dominated by rice fields, fallows, and anthropogenic forests (M.Diabaté, personal observation).The geomorphology of the landscape is characterized by low rolling hills that form a hill to valley continuum, within which three major physio-hydrographic positions are distinguished by their soil and the source of water for cultivation.These are pluvial (hilltop, cultivation reliant on precipitation), phreatic (hillside, groundwater from high water table ), and fluvial (valley bottom water from surface flow, i.e., run-on and flooding by streams).Natural soils at the top of the toposequence are typically infertile and highly leached Oxisols or Ultisols, whereas those toward the bottom are more fertile Inceptisols and Entisols (Andriesse and Fresco 1991).Wenwuta is a settlement of ~250 people and around 2.5 ha in size.C 14 dating shows the town was founded between 1670 and 1682 (Solomon et al. 2016).To estimate the extent of SA around the settlement, in collaboration with locals we GPS-mapped all SA within a 3km radius of Wenwuta (Fig. 1), which was the area that they knew best and we had easiest logistical access to.This encompasses the entire Wenwuta territory and extends into that of neighboring settlements.

Forest inventories and soil characterization of sacred agroforests and old fallows
Based on the participatory mapping in collaboration with locals, we chose nine patches of SA and four patches of OF to establish transects for forest inventories (Appendix 2).In each SA, we established between one and five "variable area" transects following Sheil et al. (2003), depending on the size of the patch, resulting in a total of 28 transects.In OF, we established between three and seven transects per patch, in total 21 transects.(Fig. 1, Appendix 2).We did more transects in SA than OF for two reasons: first, owing to the difficulty of finding OF within the area of study, and second, because SA were the focus of study, and their biodiversity is less well understood than that of OF in the region.Transects were based on a 40-m transect line, along which four cells with a length of 10 m and a variable width were delimited on both sides and perpendicular to the line (eight cells thus making up a sample-plot).Three size-classes of trees were sampled: • large trees (DBH > 20cm; L max 20 meters, L min 15 meters), • medium trees (DBH > 5, < 20 cm; L max 15 meters, L min 10 meters), • tree seedlings (DBH < 5 cm; L max 5 meters, L min 2.5 meters) In each cell, for each size class, the five trees closest to the transect line were recorded.If no tree was encountered up to a fixed minimal distance (L min ), the cell was tallied as empty and structural parameters (tree density and basal area) for this cell were counted as zero; if less than five trees were encountered up to the maximal distance (L max ), the sampling stopped and density was estimated from the number of trees recorded and the maximal distance; if five trees were encountered before the maximal distance, the distance to the line of the fifth tree was recorded and used to estimate tree density (Sheil et al. 2003).For each size-class, a maximum of 40 sampled trees per plot could thus be obtained.
Because the number of sample trees per plot varied, to make data on species richness comparable between transects we calculated a species richness index provided by Z=log(species counts)/log (stem counts), following Sheil et al. (2003).
To characterize the soil under SA, we took one soil sample at 0 -15cm depth from each of the eight cells in each transect.These were then mixed to create a single composite sample for each transect in the SA.Samples were taken to Cornell Nutrient Analysis Laboratory, and were analyzed for chemical properties (aluminium, phosphorous, copper, zinc, manganese, calcium, lead, potassium, sulphur, iron), organic matter content and pH (Appendix 3).A separate study showed that the upland soils in the region were uniformly Oxisols (Solomon et al. 2016).
Although we did not collect soil samples under OF, we examined in situ a sample taken in each transect to confirm that the OF were located on yellowish/red Oxisols.

Cultural dimensions of sacred agroforests
To understand the processes by which the Loma value, interact with, and conceive of their environment we used unstructured and semistructured interviews and oral histories (n = 64), focus group discussions, participant observation, and transect walks (Cotton 1996) during nine months of fieldwork in Liberia.These methods were contextualized within in-depth ethnography during five months' residence in Wenwuta.In all types of interviews we sought to balance numbers of male and female, youth and elder informants.
To quantify the cultural values associated with species in SA and OF, we used a freelisting technique (a type of cultural domain analysis, see Bernard 2011).We interviewed 116 randomly selected individuals at Wenwuta, surrounding villages, and at the town of Borkeza (Fig. 1).These other locations were selected because of the presence of SA and communities with long histories of interaction with them.We asked each individual the following questions: (1) Which trees can you see when you are in an old town spot (SA)?; (2) Which trees can you see when you are in an OF?We then used the Cognitive Salience Index, proposed by Sutrop (2001) to measure species' cultural salience, which we used as an index of cultural valuation.This index uses free-listing data to elicit and rank the members of a certain cultural domain (in our case: SA and OF).The general idea is that terms (i.e., ethnospecies, that is, different kinds of plants and trees as perceived by locals) that are mentioned more frequently (i.e., in more interviews) and earlier (i.e., among the first terms to be mentioned in each interview) are more "culturally salient" (Quinlan 2005) and therefore have a higher value for the index.Our cultural salience index includes both utilitarian (since most useful plants tend to be mentioned more frequently,) and spiritual/ symbolic (since some plants that are culturally salient might not have a high utilitarian value) values (Reyes-García et al. 2006).It provides a means to quantify cultural valuation that provides reliable data whilst being relatively quick and easy to conduct interviews (Quinlan 2005).http://www.ecologyandsociety.org/vol21/iss3/art36/Local names for the trees were linked with their scientific equivalents by regional expert botanists Diabaté, Beavogui, and Guilavogui.We then calculated a cultural salience index (adopted from Sutrop 2001) of each tree species using this formula: CS = F / (N*mP) where: CS = cultural salience index, F = frequency (# of people who cited the species), N = number of people interviewed, mP = mean order in which the species was mentioned in the lists (or mean "rank" of the species; Sutrop 2001).Using the data obtained in the floristic inventories, we then calculated the average cultural value per tree in each transect and per transect, which were compared between SA and OF.
To compare SA and OF in terms of vegetation structure (density of individuals, basal area), species richness, and cultural valuation index, we used analyses of variance (ANOVA).To visualize differences between SA and OF with regard to their floristic composition, we used an ordination technique (nonmetric multidimensional scaling -NMDS) based on the Bray-Curtis dissimilarity.To test the statistical significance of the difference in species composition between SA and OF, we used a nonparametric multivariate analysis of variance (MANOVA) with 1000 permutations.All statistical analyses were performed with the R statistical software using the vegan package (Jari Oksanen et al. 2016).

Regional survey and coverage estimates
Open interviews and oral histories conducted during the initial survey indicate that SA exhibit the same fundamental characteristics (that we describe at length below) among Kpelle, Mano, and other speakers of languages belonging to the same Mande macro family as the Loma, all of whom share similar cultural and ritual practices.Clearing and burning these spaces is forbidden under customary law because these sites feature ancestors' graves and certain trees tied to living individuals, neither of which may be disturbed.At the site of old towns, this sacredness allows the continuing existence of mature trees within these spaces while allowing the planting of shade-tolerant economic tree crops such as kola, native to West Africa (Ford 1992), cocoa, native to South America and cultivated in West Africa from the 40s and 50s (Westphal 1987), and coffee (Coffea canephora), native to West Africa.Our participatory GPS mapping within a 3-km radius of Wenwuta (~2827 ha; Fig. 1) shows that SA cover 18.6 ha of the landscape surrounding the settlement.Given the wide historical distribution of Mande speaking peoples, it seems reasonable to conclude that SA are widespread throughout the Upper Guinea forest of NW Liberia and southern Guinea (Fairhead andLeach 1994, Fairhead et al. 2003), and Sierra Leone as well (Leach 1994, Frausin et al. 2014).

Cultural processes shaping sacred agroforests
Our interviews and oral histories at Wenwuta and during the wider regional survey revealed that SA are cultural spaces located on the sites of old towns (Appendices 1 and 2), characterized by the presence of ancestors and graves.Ancestors, supernatural beings, and ritual artefacts such as shrines, masks, divinatory media, fetish objects, and herbal medicine are expressed in the concept of salɛ (expressed in Liberian English as "medicine"; Hojberg 2007, McGovern 2012).We found that that in SA, ancestors and some trees are seen as being endowed with this metaphysical power.The head of the Wenwuta Poro society informed us that certain trees are salɛ, although one must be familiar with particular salɛ properties or powers associated with different trees to exploit them for ritual purposes.The Loma explained that the presence of ancestors is manifest in gravestones and also in individual trees that are linked to the living and the dead.Kola trees at SA are often related to living individuals, i.e., they were planted to commemorate the birth of an individual, or other significant events (Ford 1992).One local man explained, "our forefathers moved and planted tree crops for memory.Nobody can make farm on old spots [SA], they still bury people there.When a child is born a kola tree is planted, which is then protected.These are still planted in old town spots or adjacent to them."Cotton (Ceiba pentandra) trees at these sites are also protected.They are said by elders to have once been planted in lieu of gravestones.This in turn imbues mature Ceiba with an ancestral significance.
Owing to this cultural valuation of SA, the cutting and burning of trees is forbidden, but these areas are still used for the cultivation of tree crop species, which can be done without disturbing large, mature emergent rainforest species, and without the need for burning.The Loma recognize that the AfDE that is associated with SA permits cocoa and kola cultivation on upland soils, which is otherwise difficult on highly infertile yellowish red Oxisols that dominate in Liberia [1] .A local chief explained, "I brought cacao here in 1964... when we arrived we tried planting cacao everywhere but it did not grow.I planted cacao in the red soil, after 20 years it did not bear fruit.It only grows in the lowland, on old town spots, and around the town... Kola also needs good soils to produce well, so it is good to plant it in the old town spot [SA]." We found that the settled life of Loma and by extension other Mande groups over time results in a mosaic of SA.A female elder explained, "we plant trees around villages and towns, cocoa; coffee; kola and cotton trees on graves, but after time they get big and the place gets cold and dark.We are prevented by our customary law from cutting the kola and cotton trees, so people decide to move.The town space is then planted with cocoa, kola and other trees, and it becomes an old spot, where people still come for meetings, to gather crops and to bury our dead."Therefore, the physical production or "domestication" of space by the Loma, follows an arc through time from the initial clearance of forest and establishment of the town-space, decades or even centuries of daily domestic activities, and the spatial and temporal dynamics of the planting and management of tree crops and other trees, forest regrowth, and the eventual abandonment of spaces as a living area (Fraser et al. 2014).In this final stage, the surrounding forest island makes the town "too cold," referring to a phenomenon where forest growth over time makes the town "colder," which is associated with a growing power of ancestors and also malevolent spirits, leading to the need to establish a new town in a safer space (McGovern 2012).The abandoned townspace then becomes protected precisely because of the traces of the former settlement, the presence of graves, cotton trees and kola linked to the living.This system, however, is under threat from a generational shift in cultural valuation.Youth across NW Liberia participated in the recent (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005) war, which accelerated already existing processes of change wherein youth begin to challenge various aspects of the Loma worldview, including the sacredness of SA, i.e., that they should not be cleared for annual crop production (Fraser et al. 2015).This is demonstrated in our survey that found that when SA were slashed, it was often by youth in violation of their elders' wishes.This was particularly acute in Nimba County, the cradle of the war, where youth had greatest participation in armed conflict (Appendix 1).

Comparing sacred agroforests and old fallows
Sacred agroforests and old fallows differed in several of the vegetation parameters that we analyzed (Fig. 2, Table 1).The species richness index tended to be lower in SA than in OF for all strata, but this difference was statistically significant only for the middle strata (Table 1, Fig. 3).The greater disturbance and management intensity to which SA have historically been subject may reduce the number of species in the community, favoring a limited number of species of particular cultural value.This is clearly seen in cacao, for example, a species highly valued and intensively managed that dominates the understory in many patches of SA, particularly at the middle strata.Despite this more intensive management, however, our results show that the upper strata and the seedling strata of SA can maintain a comparable number of species with OF.The density of seedlings was significantly higher on OF than on SA, while the density of trees in middle strata was higher in SA than in OF (Table 1, Fig. 2).Similar to the pattern observed for species richness, the lower density of seedlings in SA can be due to the removal of seedlings from nonuseful species, but once these species are favored, they can occur in high densities at the middle and upper strata.
Our index of cultural valuation showed that SA are of higher overall cultural (utilitarian and symbolic) value to the Loma than OF.This is supported by our finding that the average cultural valuation indices per individual and per hectare were significantly higher in SA for trees in the middle and upper strata, when compared to OF.For seedlings, however, it was higher only when calculated per individual (and not per hectare; Table 1, Fig. 3).This shows that the historical and current management of SA favors the occurrence of species that are of high cultural importance to local people.The lack of difference for seedlings http://www.ecologyandsociety.org/vol21/iss3/art36/ is probably the result of the regeneration of these useful species together with species that regenerate naturally during secondary succession.We found that basal area was more heterogeneous and significantly higher in SA than in OF in the middle strata, but in the upper strata it was slightly higher in OF than in SA (although not statistically significant; Table 1, Fig. 3).This is likely the result of the occurrence of useful species in high densities in SA, particularly in the middle strata.The large variation in basal area in SA is likely due to a more heterogeneous land-use history of these areas when compared with OF.These results indicate that SA can have biomass values comparable to those of OF.

Fig. 3.
Boxplots showing comparisons between sacred agroforests (SA, white boxes) and old fallows (OF, grey boxes) regarding (A) species richness index (Sheil et al. 1999), cultural salience index [per individual (B) and per hectare (C)], density of individuals (D), and basal area (E) for three vegetation strata: lower ("low"), middle ("mid") and upper ("up").For our definition of the species richness index, the cultural salience index, and the vegetation strata see section "Case study and methods." The floristic composition was clearly different between SA and OF for seedlings (NPMANOVA F = 16.08;p = 0.001), for trees in the middle strata (F = 48.09;p = 0.001), and for trees in the upper strata (F = 14.13; p = 0.001; Fig. 3).The NMDS ordination also shows that the difference in floristic composition between SA and OF tends to be larger for trees than for seedlings (Fig. 2).
This indicates that species that occur in OF regenerate in the understory of SA but do not reach maturity, either because they are not adapted to that specific environment or because they are removed through management.Figure 2 also shows that in OF the difference between the composition of seedlings and trees is greater than in SA.This shows that, although species composition will likely change with time in both environments, future composition on SA will tend to stay closer to the current composition than will OF.

DISCUSSION AND CONCLUSIONS
The relationship of cultural valuation to biodiversity conservation is often examined through single disciplinary lenses, leading to partial views and incommensurability between them.
We have attempted to overcome these shortcomings by combining methods from each of the three cultures of natural science, quantitative social science, and qualitative social science/ humanities to show the distribution and cultural valuation of SA, their relationship to OF, and the ways in which social institutions mediate the management of these spaces.In this concluding section, we reflect on how successful our combination of different methods and data sets was in addressing our research questions, and their broader utility in exploring the relationship between cultural valuation and biodiversity conservation.
Our first research question sought to establish the extent of the hitherto undescribed phenomenon of SA in NW Liberia.This was the most straightforward in terms of the question and methods, i.e., oral histories and GPS points, yet the most time consuming owing to difficulty of travel both by car and on foot.Our regional survey, interpreted in the light of 19th and 20th century historical descriptions, found that SA are common features of the landscape in northwestern Liberia.Because it relied on local knowledge of old town spots (the local term for SA) we consider the findings to be reliable and the sample size to be big enough to support the finding that SA are widespread in the Upper Guinea forest and therefore potentially providing an important role in biodiversity conservation, along with tangible and intangible cultural heritage values.In Madagascar, for example, culturally protected small islands of forest have been shown to be essential for maintaining ecosystem services (Bodin et al. 2006).We found locals to be well aware that the fertility of AfDE permits and/or improves the production of tree crops, particularly cocoa, which, during our regional survey, we did not observe growing outside patches of AfDE or naturally fertile lowland soils.
Our second research question sought to examine and explain the role of social or religious institutions in conserving SA.We drew on a range of ethnographic methods from qualitative social science and humanities.This revealed the cultural processes through which both SA as places and individual species within them are valued and how this valuation shapes their conservation.We found that SA were valued as places for initiation society activities and as sites of fertile soil suitable for farming cocoa.In addition, the presence of ancestors in SA, and their manifestation in graves, trees, and belief in salɛ, is the cultural value of these places that contributes to their conservation.This cultural valuation of place in terms of the presence of ancestors, is significant in legitimating current inhabitants' claims to land tenure.This important form of cultural valuation cannot be http://www.ecologyandsociety.org/vol21/iss3/art36/captured by quantitative approaches (natural or social), and underlines the importance of qualitative investigation.Moreover, some individual plants and trees in these places were valued both for their utilitarian affordances, while others were symbolically valued because they were seen to be imbued with the metaphysical power salɛ, while yet others were valued because they were understood to be linked to the living people and the dead ancestors.
Our third research question compared species composition and cultural valuation of SA to that of OF.This question was addressed using quantitative methods from natural science and quantitative social science, strictly oriented toward species and ethnospecies respectively, but not toward these spaces as places, as question two did.We found that although SA have biodiversity and biomass comparable to OF, they have a different species composition, not only adding biodiversity to the landscape, but also conserving forest species that are common in SA but rare (or lacking) in OF.Our index of cultural valuation showed that species within SA are of significantly higher cultural value for the Loma when compared to those of OF.Our index did not differentiate utilitarian and symbolic dimensions of cultural valuation; salience may reflect either or both of symbolic and utilitarian values, and the distinction between these kinds of cultural valuation is better explored through qualitative methods, which we attempted to address in question two.
Our approach was successful to the extent that it used methods from each of the three cultures of science in addressing research questions.However, as with most mixed methods research, it did not integrate the methods as such, rather, they remained separate.
The first two research questions were addressed through an interpretivist lens using qualitative methods (GPS points aside) with the underlying assumption of a processual or relational world, whereas the final question was addressed by way of a positivist approach using quantitative methods which, although belonging to different disciplinary groupings (natural science and quantitative social science), are similar in terms of underlying assumptions of a world that can be split up and understood in terms of constituent parts.
This brings us to the question of whether a deeper integration of methods from the three cultures of science is a feasible way to advance ecosystem services frameworks.Our conclusion is that, although the two cultures of natural science and quantitative social science are readily compatible, given their shared underlying positivism, it is a much greater challenge to incorporate methods from qualitative social science and the humanities, given their underlying interpretivism.We are hesitant to suggest that, for example, forms of cultural valuation such as salɛ, or belief in the presence of ancestors, can be incorporated into ecosystem services approaches without losing their fundamental nature.This is because they are part of indigenous ontologies that are radically different to the Western scientific framing that underlies the ecosystem services framework.The issue is that these two ontologies are incommensurate; the incorporation of salɛ and/or ancestors on their own terms would undermine the positivist assumptions of the ecosystem services framework.Conversely, were these concepts to be incorporated into the ecosystem services framework on its terms they would lose their situated meaning vis-à-vis the lived experience or "lifeworld(s)" of local people(s).Nevertheless, mixed methods approaches do hold potential for cross-fertilization; the use of each distinct approach can potentially improve interpretations of the other, which can foster a more holistic understanding overall.
We started from the idea that studies combining methods from each of the three cultures of science are likely to provide a more multidimensional understanding of tropical forest landscapes.This approach generates more holistic knowledge that takes into account other values in addition to ecological and economic ones.
We advocate the use of methods from the three cultures of science to advance discussion on how to better align and design conservation policies with a diversity of cultural values.However, we suggest that it is unlikely that this would be possible within a singular ecosystem services framework.A more realistic alternative would be to continue to situate methods from the two positivist cultures of natural science and quantitative social science within the ecosystem services framework, with the third interpretivist culture of qualitative social science and humanities located within a suitable alternative theoretical framing such as, for example, phenomenology (e.g., Ingold 2000, Willerslev 2004, Harris 2005, Roth 2009, Jackson 2012, Fraser et al. 2014).Therefore, we recommend that the three cultures of science should be used to consider ecosystem services from different positions (see Williams 2014), rather than attempting to integrate these positions and associated methods into a unified framework, because there are limits to the extent that different approaches can be integrated without one or another losing its essence.
In conclusion, we recommend that given the widespread occurrence of SA in the Upper Guinea forest, these cultural forests should be subject to dedicated conservation policies in West African countries, and policies should be sensitive to the cultural values that we have shown to be related to both the formation and conservation of these sites.With regard to cultural valuation, we note that the utilitarian and symbolic cultural values of SA that we revealed can be expressed through ideas of tangible and intangible cultural heritage, which can be used to articulate cultural valuation and biodiversity conservation to a wider audience.Conservation policies should incorporate intangible cultural heritage or symbolic cultural values into their conceptualization of local cultural valuation, along with better known tangible cultural heritage or utilitarian cultural values.
__________ [1] Rainfall may also be a factor in the infertility of Liberian soils.
The country has one of the heaviest annual rainfalls in all of Africa.In the Upper Guinea forest, rainfall decreases from east to west (see Fig. 2 in Bongers et al. 1999), meaning that over millennia the soils have become significantly much more heavily leached than comparative soils in Ghana, where cocoa cultivation is widely practiced on upland soils.It is claimed that Bacurtazi was established when a woman from the Delema quarter married a man from Zolowo.Delema people had been farming in the Bacurtazi area before the town was established, and the man from Zolowo decided to found the village in this area.Bacurtazi was abandoned within living memory, and its inhabitants then moved to a new spot nearby which, after being a village for decades, is now reduced to four farm kitchens after being destroyed during the war.

Baema Oxisols
This is an area of secondary forest 45-55 years in age.The reason for such a distance to Wenwuta is that this was identified by elders as the 'oldest' forest within a day's walk from Wenwuta.However, upon inspection Diabate concluded it was younger than the Wotizi forests close to Wenwuta.Even in this sparsely populated area we found mature Kola trees in the secondary forest.

Dadazu AfDE
This the oldspot for the current village of Dadazu 2 Delema AfDE Delema is actually now under the authority of the chief of a nearby village, Karwalawuta.Delema had been haunted by an evil spirit and therefore Wenwuta people avoided the areas.The chief's father, who had been living in Monrovia, gained access to the old spot by marrying a woman from Wenwuta, and then sought permission to plant cacao, kola, coffee, banana and orange trees there 5 Doyahou Oxisols This is an area of secondary forest 50-60 years in age 5

Famene AfDE
Famene belonging to a different ethno-linguistic group, the Barlain-Kpelle that inhabited the area prior to the arrival of the Loma and therefore so ancient as to have been long abandoned even in oral histories.Famene, meaning "hear-hear" in Kpelle (Memene in Loma) (Figure 3) was named as a place where people from around the region met and discussed issues.

Gbokolomie AfDE
This is the oldspot for the current village of Gbokolomie

Kpazaquelle AfDE
This is an oldspot belonging to the town of Tinsue.It has a single male owner resident in this town.

Wenwuta Old Spot AfDE
This is the oldspot for Wenwuta.Wenwuta is dated to 1670-1682, implying this site is long abandoned.Oral histories also contradict.There was no consensus as to whether this was Wenwuta oldspot, or the oldspot of Duala, the founding quarter of Wenwuta.Also, it is not clear whether this site was abandoned and re-occupied or simply abandoned

Fig. 2 .
Fig. 2. Nonmetric multidimensional scaling (NDMS) showing differences in floristic composition between sacred agroforests (SA) and old fallows (OF) in different strata: seedlings, trees with 5 < DBH < 20 cm (middle strata) and trees with DBH > 20 cm (upper strata).Each point in the graph represents a transect, and distances between points are proportional to their biological dissimilarity, calculated with the Bray-Curtis index.

Table 1 .
Analysis of variance (ANOVA) for the parameters measured in transects established in sacred agroforests (n = 29) and old fallows (n = 21) in Northwestern Liberia.Values shown are means ± their standard deviation.
Over the long term therefore, inhabitation of the Upper Guinea forest by Mande speaking peoples creates culturally valuable areas of anthropogenic forest.http://www.ecologyandsociety.org/vol21/iss3/art36/ article can be read online at: http://www.ecologyandsociety.org/issues/responses.php/8738 Figure A1.1:Localities surveyed where Sacred Agroforests were found in North Western Liberia