Combining high-resolution remotely sensed data with local and Indigenous Knowledge to model the landscape suitability of culturally modified trees: biocultural stewardship in Kitasoo/Xai’xais Territory

Biological diversity is strongly linked to social, cultural, and economic resilience. Increasingly, there is a need to connect these dimensions in environmental management and monitoring (EMM) projects to reverse global declines in biological diversity and promote resilient cultures and communities (Lertzman 2009; De Groot et al. 2010; von der Porten et al. 2019; Zurba et al. 2019; Walker et al. 2020). Barriers to implementation can manifest when conservation oriented policies and decision-making processes are not locally informed or led (DeRoy et al. 2019). In cross-cultural biodiversity conservation or stewardship implementation, innovating new ways to support locally led stewardship initiatives is critical to meeting both local and global biodiversity conservation goals (Gavin et al. 2018). New strategies must directly include social and cultural values and support Indigenous rights and authority in EMM, which are integral to social–ecological systems (Sterling et al. 2017b; Artelle et al. 2018; Ban et al. 2018, 2020; Salomon et al. 2018; Pinkerton et al. 2019; Zurba et al. 2019; Caverley et al. 2020). “Biocultural approaches” can offer an efficacious method to guide EMM objectives towards achieving these outcomes, because they focus conservation goals and interventions on values that are important to local communities (Verschuuren 2012; Pungetti 2013; Gavin et al. 2015; Sterling et al. 2017a; DeRoy et al 2019).

Biocultural approaches seek to protect and steward linkages between humans and ecological systems by supporting locally led EMM objectives and by adopting a partnership-based approach to bridge social–ecological linkages across spatial and institutional scales (Caillon et al. 2017; Sterling et al. 2017a; Wali et al. 2017; Gavin et al. 2018; DeRoy et al 2019; von der Porten et al. 2019). Addressing biocultural EMM objectives requires monitoring processes that focus on locally important values, species, and the relationships among them (Gallagher and Josephs 2008; Verschuuren 2012; Artelle et al. 2018; DeRoy et al. 2019). Success in protecting and stewarding social–ecological linkages can be measured through diverse methodologies, ontologies, and knowledge systems (e.g., Indigenous methodologies, perceptions-based approaches, natural sciences, or empirical social sciences; Newing 2010; Bennett 2016; Caillon et al. 2017).

A range of these innovative cross-cultural approaches to EMM that explicitly adopt values that reflect local peoples and respect Indigenous rights and laws are emerging. For example, researchers in Australia have developed an environmental management framework in collaboration with two Indigenous groups in the Kimberly region—the Bardi Jawi and Nyul Nyul—that is guided by key Indigenous wetland stewardship principles (Pyke et al. 2018). This framework has helped transcend inter-cultural and inter-agency barriers and promote a guiding stewardship planning framework in which “wetlands need people” (Pyke et al. 2018). Similarly, the Omora Ethnobotanical Park in Tierra del Fuego, Chile, has incorporated 10 principles based on Yaghan (or Yámana) Indigenous Knowledge and stewardship practice to guide development of the Park’s mission and to align objectives among institutions involved, including the Yaghan People and Chilean government (Rozzi et al. 2006).

Beyond guiding principles and establishing frameworks, analytical approaches that support interdisciplinary, cross-cultural approaches to implementing biocultural EMM are required. Suitability modelling has been used in “functional ecological” approaches to EMM (i.e., those that do not include biocultural indicators) for decades (Rodríguez et al. 2007; Elith and Leathwick 2009; Franklin 2010; Guisan et al. 2013). Common suitability modelling methods that are used in functional ecological approaches are machine learning based, such as random forest models or maximum entropy (Guisan et al. 2017). However, suitability modelling in this context has had mixed success in achieving desired EMM outcomes. Failures can often be attributed to error propagation in the models (Heuvelink 1998; Store and Kangas 2001). This can be caused by use of ex situ occurrence data (i.e., observational data gathered outside the management area), or by inclusion of environmental predictor variables that do not strongly contribute to, or limit, the given habitat or distribution of focal species (Store and Kangas 2001; Franklin 2010). Therefore, suitability modelling is more likely to support outcomes when occurrence data used to evaluate the model and predictor influence are locally derived, and (or) the model predictor variables are locally informed (Polfus et al. 2014). Such approaches provide new opportunities to link local values with local data in EMM.

One method that is particularly well suited for the incorporation of local and expert knowledge in a suitability modelling framework is multi-criteria evaluation (MCE; Eastman 1999). MCE is a modelling process in which a suite of potential criteria or predictor variables that contribute to the suitability of a given outcome are empirically compared against one another to determine the relative influence of each predictor on the expected outcome (i.e., location of an important resource; Eastman 1999; Store and Jokimäki 2003). MCE methods have been used to incorporate local expert knowledge as part of suitability modelling frameworks in a diverse array of applications, including decision-making models and habitat suitability modelling (Store and Kangas 2001; Polfus et al. 2014, 2016). Local expertise can be especially important when prior empirical study is limited or nonexistent (Store and Kangas 2001). However, local guidance and place-based expertise are critical to any biocultural approach to suitability modelling or other EMM endeavours regardless of whether prior empirical study has occurred or not (Salomon et al. 2018).

The area now referred to as the Great Bear Rainforest (GBR) of British Columbia, Canada, is a globally significant biocultural diversity hotspot that is rich in place-based EMM expertise. The GBR is comprised of intertwined socio-cultural and marine-influenced terrestrial biogeoclimatic systems that have been heavily influenced by thousands of years of stewardship practice by First Nations Peoples (Lepofsky et al. 2017; Lepofsky and Armstrong 2018). Indeed, First Nations People in this region have been engaged in EMM directly with many species of cultural importance for over 14 000 years (Mclaren et al. 2015; Mackie et al. 2018). The Kitasoo/Xai’xais (KX) First Nation, among many others, are continuing this stewardship in a contemporary context by developing and applying their own approaches to stewardship informed by their Indigenous laws to manage both ecological and cultural values. Further, recent provincial legislation that applies to the GBR (Great Bear Rainforest Land Use Objectives Order (GBR LUO; British Columbia Ministry of Forests, Lands and Natural Resource Operations 2016)), was developed to recognize and support the implementation of Indigenous-led stewardship in a collaborative government-to-government structure. The GBR LUO includes specific protection targets (often percentage based) for ecological features such as ecotypes and critical habitat for endangered species with less-defined goals for culturally significant features in this region.

This legal process recognizes culturally modified trees (CMTs) as a priority cultural value included as an “Aboriginal Heritage Feature”. CMTs can be defined as trees, both living and dead, that bear evidence of traditional or cultural use by First Nations People (British Columbia Archaeology Branch 2001). Examples of CMTs in coastal British Columbia include bark removal, felling and removal of logs for wood products (such as canoes or building and carving materials), pitch collection, inner bark collection, and wood plank removal from live standing trees. Although empirical study on CMTs is limited in the GBR—especially for bark-stripped CMTs—this work builds off of recent work by others in the region that focused on another common type of CMT referred to as “aboriginally logged” features that were harvested for wood resources (Benner et al. 2019).

We focus here on CMTs that have been harvested for bark—a material central to current and historical Indigenous cultural practice—in a way that typically allows the tree to heal and continue growing (and in some cases be harvested multiple times over many years; British Columbia Archaeology Branch 2001; Turner et al. 2009; Earnshaw 2017). Generally, trees suitable for harvesting bark need to have a bole surface free of knots and branches. Smaller diameter trees can be more suitable for pulling tapered bark strips (often on the upslope face), whereas larger diameter trees are more suitable for removing rectangular bark strips (British Columbia Archaeology Branch 2001). However, the practise of intergenerational stewardship provides sustained harvesting opportunities on the younger healing lobes that form after bark is pulled, thereby allowing bark to be pulled for hundreds of years on even the oldest and largest diameter cedar trees (Earnshaw 2017, 2019; Stafford 2017). We focus particularly on CMTs that are visible to field surveyors (i.e., not healed-over completely), which were likely culturally modified within the last ∼250 years. Importantly, CMTs provide evidence of Indigenous use spanning centuries and serve as biocultural indicators that can help identify culturally significant areas that may also be biologically rich (Garibaldi and Turner 2004; Turner et al. 2009; Sutherland et al. 2016; Benner et al. 2019; DeRoy et al. 2019).

Despite their value to Indigenous peoples, CMTs were not protected until relatively recently and at small spatial scales. They received protection as “archaeological sites” under the provincial Heritage Conservation Act (British Columbia Archaeology Branch 1996), but only for CMTs dating prior to 1846—an arbitrary colonial construct that limits heritage management protection status (Turner et al. 2009; Earnshaw 2017). The cumulative impacts that widespread commercial timber harvest imposed on CMTs prior to their initial protection in the 1980s is likely immense (Oliver 2007; Turner et al. 2009; Earnshaw 2017). Further, in other regions empirically informed estimates suggest that roughly half of all bark-stripped CMTs are not currently visible to archaeologists and CMT surveyors because the tree has completely healed over the scar, which can lead to continued undocumented removal of CMTs through commercial forestry even after their legislated protection (Earnshaw 2017, 2019). This undocumented removal of CMTs may impact the outcomes of Indigenous rights and title cases brought before Canadian courts, due to the removal of physical evidence of long-term occupation and use (Earnshaw 2017). Documented removal of CMTs can also occur via site alteration permits issued by the provincial Archaeology Branch, which typically require expressed support or approval by First Nations. The impacts from commercial forestry continue as old growth cedar including both western redcedar (Thuja plicata) and yellow-cedar (Cupressus nootkatensis) remain targets for commercial harvest in British Columbia. Against this background of historical and ongoing impacts and their value as markers of Indigenous heritage, occupation, and stewardship, as well as their legal protection, documenting remaining CMTs comprises a stewardship priority for many Indigenous Nations in western Canada.

Historically, colonial governments have largely excluded Indigenous governments from decision-making regarding biodiversity conservation and resource extraction in Canada and abroad. However, the political landscape is changing in Canada and abroad with the commitment to implement the United Nations General Assembly (2007) Declaration on the Rights of Indigenous Peoples and to reconcile the relationships between Indigenous and colonial governments. Although there is increasing interest in co-governance, approaches still commonly follow dominant western science paradigms to which Indigenous governments, communities, and knowledge holders are expected to conform.

For example, the GBR LUO takes a spatially explicit approach to managing commercial forestry in the region by allocating percentage-based retention targets for each landscape unit and ecotype to be protected from commercial forestry (British Columbia Ministry of Forests, Lands and Natural Resource Operations 2016). The province of British Columbia has invested decades of research and development with teams of spatial analysts to create spatially explicit, ecological inventory data sets that combine remotely sensed data with field observations to model the distribution of ecological features, like ecological communities and rare ecosystems, and economic features such as stand volume (British Columbia Ministry of Forests, Lands and Natural Resource Operations 2018). In parallel, but dating back millennia, many Indigenous governments also have abundant knowledge and data in the form of local and Indigenous Knowledge, oral histories and, more recently, traditional use studies. These forms of information and data, however, are not easily translated to interact with percentage-based targets for commercial timber harvest limits and, unlike ecological inventory data, most cultural heritage features have not been surveyed or inventoried.

The resulting political and practical barriers are manifold. For example in the context of spatial scale, data from traditional use studies often refer to specific points or culturally significant places with amorphous boundaries that cannot be divided in the same way as forest stand types. At a political and governance scale, the barriers are more hindering. Laws, regulations, and other colonial government systems are slow to adapt to new co-governance agreements. For example, the Chief Forester with the Province of British Columbia sets the Allowable Annual Cut for this region and many others based purely on timber supply modelling, which does not typically take into account Indigenous lifeways, practices, histories, or links to cultural well-being beyond excluding recorded archaeological sites from model outputs. Amendment to such colonial regulations, laws, and governance processes often requires the burden of proof or evidence on behalf of Indigenous governments. In the context of these political and data-oriented barriers, some First Nations Governments in the GBR are developing their own spatially explicit analyses based on ecological and cultural values that are important to their communities to support culturally relevant governance.

Given this context and focus, our team, which includes members from academia and the KX Stewardship Authority, used suitability modelling approaches that link local stewardship values and data. Specific research goals as identified by the KX Stewardship Authority for this project included: (i) gathering data from areas with low previous survey effort, (ii) evaluating potential biophysical and cultural predictor variables that might predict site suitability and distribution of CMTs, and (iii) providing directly relevant information to KX Lands Managers about where culturally modified trees are likely to occur, particularly in the context of commercial forestry. Towards these goals we developed CMT suitability models using records from field surveys in addition to archived records from the British Columbia Archaeology Branch to inform KX forest management and to support greater equity in spatial data and analysis available during co-governance.

We developed a unique suitability modelling approach to not only support Kitasoo/Xai’xais governance of their Territory but also as a generalizable framework that could be used by other Nations in the context of co-governance or Indigenous-led governance. This approach draws from a diverse range inputs including, field survey, archived archaeological site reports, local and Indigenous Knowledge, remotely sensed satellite and LiDAR data to model landscape suitability for CMTs, a key biocultural indicator. We found that Elevation and the cultural variables are among the most powerful predictors for CMT distribution in KX Territory. In particular, the Cost-Distance Submodel is more influential than Slope alone, which indicates that cumulative impediments (costs) to harvesters to reach a site in the form of slope are more likely to affect the suitability of a given area for bark harvesting compared to specific site (cell) level conditions (Table 3). LiDAR-derived Elevation and CHM were also important contributors to predict CMT suitability. Slope, Aspect, Canopy Cover of both yellow-cedar and redcedar, however, were not highly influential in predicting CMT suitability. Our final model based on best-fit (principal) weights provides a continuous heatmap of CMT suitability throughout the study area (Fig. 3), and performed well to identify known CMT occurrence from both recent surveys and from RAAD (Fig. 4). Finally, by comparing our principal weights CMT model with the commercial timber harvesting landbase, we find that the proportion of high suitability areas is much greater within the THLB than outside.

The model provides a vehicle to combine local and Indigenous knowledge with biophysical data to collectively provide evidence-based predictions for the occurrence of culturally important features. Accordingly, the model illustrates how biocultural stewardship can draw on both ancient knowledge and contemporary computing approaches. We also acknowledge several limitations and trade-offs. The development of the KX field surveys employed a nonrandom sampling strategy, owing to logistical constraints and the difficulty in conducting grid- or census-based methods in this largely roadless landscape. Clearly, if financial and time resources are abundant, census-based surveys provide a better approach for many suitability modelling endeavours (Muir and Moon 2000; Store and Jokimäki 2003). Additionally, although the contributions of existing data from previous archaeological survey efforts greatly increased our sample size and provided valuable occurrence data for regions that were difficult to access during our field season, the accuracy of the georeferenced points was unknown. Confidence in the models could be improved by ground-truthing the spatial accuracy of a subset of the georeferenced CMT locations from RAAD as well as information derived from satellite imagery. Lastly, incorporating community-based interviews that more explicitly informed the models on harvester preferences and laws might improve model performance.

The predictive power of the cultural variables in our model demonstrates the importance of incorporating local perspectives in suitability modelling of biocultural indicators. For example, other CMT potential models often use a “distance from shoreline” variable, where our work benefitted from Indigenous Knowledge and local perspectives to create a finer-tuned Cost-Distance Submodel. The design of both the Cost-Distance Submodel and the Distance from Known Habitation Sites variables were led and informed by local and Indigenous Knowledge to ensure that our approach offered a more detailed and culturally resonate perspective than the simple “distance from shoreline” variable. The somewhat lower standard deviation for our Cost-Distance Submodel suggests its use also provided less uncertainty than Elevation. Importantly these variables reflect the long history of KX biocultural stewardship and offered key information to predict the suitability of CMT locations.

Although we expected that the biophysical variables would have strong predictive utility for CMTs, variables such as Canopy Cover of yellow-cedar and redcedar performed relatively poorly. The poor performance of the redcedar variable could be attributed to a possible correlation between CMTs and mixed species stand composition or inaccuracy in the satellite-derived species composition data (Fleming et al. 2004). The low sample size of yellow-cedar CMT observations (n = 33) used in the CMT–MCE, paired with the relatively small amount of area with dominant yellow-cedar canopy, could have contributed to the minimal predictive utility of yellow-cedar in modelling CMT suitability. The low sample size of yellow-cedar CMTs could be due to challenges in surveying higher elevation yellow-cedar stands. Moreover, most CMT observations from RAAD in our study area come from surveys conducted below 250 m. Although we used a helicopter during our field surveys to access higher elevation areas (conducting ∼15 km of survey above 300 m) to address this potential bias, we did not locate any CMTs above 400 m. Future studies may benefit from modelling yellow-cedar CMTs separately from redcedar CMTs due to their potential divergence in suitable site conditions (Stafford 2017).

Other than Elevation, biophysical variables had modest predictive utility. The CHM had the most predictive power of these variables, indicating that although taller trees are increasingly suitable (Table 3), a variety of tree heights can provide adequate conditions for CMTs. Further, a diversity of canopy layers (which might lower the median crown height per 25 m² cell) likely creates suitable conditions for trees with fewer branches below the crown, thus offering a bole surface relatively unobstructed by limbs from which to pull bark (Sutherland et al. 2016; Benner et al. 2019). Additionally, the mean suitability scores for Slope and Aspect fell below that of the equal weights mean suitability score, indicating their modest predictive power (Table 3). The standard deviations for Slope and Aspect were among the highest in the models, indicating that CMT suitability occurs across a wide range of slopes and aspects. This latter observation is often a reason for omitting CMT survey efforts in steep terrain in other regions, which is not supported by this model.

We quantified important spatial patterns between CMT suitability and the commercial THLB polygons, the latter comprising ∼27% of the study area. We found that moderate and high suitability cells account for more than 87% of the area within the THLB (Table 6). Further, the relative proportion of high suitability cells within the THLB (13%) is 51% greater than the amount outside the THLB (9%; Table 6). These patterns, evident at fine to large scales, illustrate the utility of this model (Fig. 3) in mitigating potential impacts to existing and future CMTs from commercial timber harvest. For example, forestry planning can identify “hotspots” of CMT suitability (at multiple scales) before onsite surveying and engineering is considered. We caution, too, that our (and any) model cannot replace the utility of ground-based surveys. Additionally, despite archaeological surveys within a cutblock prior to cutting, the undocumented removal of CMTs can still occur because old CMT scars can completely heal over making them difficult to identify prior to harvest (Earnshaw 2017, 2019). Given the role of CMTs as biocultural indicators the principal weights model could also be used as a tool for improving archaeological predictive survey for these healed-over CMTS and other archaeological site types (Gallagher and Josephs 2008; Verhagen and Whitley 2012; Hesse 2013).

Although suitability models cannot replace ground-based surveys, they offer insight into how impacts to culturally significant features and areas can be minimized as part of forest and other natural resource management. One potential method for applying our suitability model would be to overlay proposed cutblock polygons (smaller spatial units compared to the THLB polygons) to evaluate the mean, maximum, and range of suitability scores within the boundaries. In the context of Indigenous governments like the KX, this could provide Lands Managers and the communities they serve valuable information about the potential risks of developing an area before plans proceed, and investments are made into cutblock engineering, etc. For example, if development proceeds in areas of high suitability, proposed blocks can be surveyed intensively (at 100% survey coverage) prior to development and followed by a post harvest assessment to identify completely healed-over CMT scars (Earnshaw 2016, 2019). This model can also be applied during the Landscape Reserve Design process, which is a planning exercise to implement the goals and regulations set out in the GBR LUO for both biodiversity and culturally significant resources. Further, in the GBR and abroad, modelling biocultural indicators may also help support the design and implementation of Indigenous Protected Areas (Murray and King 2012).

Beyond the GBR, locally informed models such as ours could be used to help identify biocultural diversity hotspots or facilitate bringing cultural values into national or global strategies for biodiversity conservation and protection in the context of reconciliation (Diggon et al. 2020; Supernant 2020; Wong et al. 2020). Suitability models of other biocultural indicators and culturally significant species or values could also be incorporated as part of community harvest management plans to coordinate contemporary resource use, protected area designation, and management or comprehensive assessments for nonextractive industries such as tourism (e.g., Lemelin et al. 2015; Kitasoo/Xai’xais 2019). Scaling up, multiple biocultural suitability models could be combined in a biocultural ecosystem services modelling framework (Nelson et al. 2009; Klain and Chan 2012; Pert et al. 2015). In this way, stewardship managers can identify areas with a high diversity of cultural ecosystem services (e.g., heritage and marine/terrestrial resources). Further, this biocultural modelling approach could also be incorporated into species-at-risk management planning so that conservation efforts can have compounding benefits for both biodiversity and the Indigenous communities that have strong place-based ties to key habitat areas and species (Westwood et al. 2019).

Importantly this model is the first locally informed continuous spatial representation of historical and contemporary landscape use (cedar bark harvesting) to be applied in the context commercial forestry in Kitasoo/Xai’xais Territory. Further, we provide a unique example of how local and Indigenous Knowledge can be incorporated both in the form of input data as well as throughout the spatial modelling process. Our unique approach that blends Indigenous and place-based knowledge with archaeological inventory data to model aspects of the cultural landscape provides one path to weave Indigenous-led approaches with western science approaches to EMM. By linking people and place to a biocultural indicator, our model has facilitated the implementation of Indigenous-led stewardship in the context of co-governance. In doing so, our approach provides a potentially generalizable strategy to confront—and overcome—common political barriers to implementation of biodiversity conservation projects, such as the lack of recognition and meaningful incorporation of Indigenous cultural values during planning and monitoring.

Biocultural EMM offers an emerging domain that couples local values and scientific approaches to overcome common political barriers to biodiversity conservation by protecting enduring linkages between people and nature. Biocultural indicators, such as culturally modified trees, offer a focal point where multiple knowledge sources, institutions, and agencies can engage to design and implement EMM effectively. Our case study illustrates how a locally led suitability modelling approach guided and enabled the incorporation of locally informed input predictor variables. Accordingly, we demonstrate how local and Indigenous Knowledge and values can direct both the development and application of science as well as guide the implementation of biocultural EMM decisions in co-governance scenarios. However, we also highlight the need to re-evaluate the equity of co-governance agreements to better recognize, uphold, and apply Indigenous ways of knowing in ways we do here and beyond (Reid et al. 2020). Respecting and actively supporting Indigenous authority, laws, institutions, values, protocols, and stewardship practice are critical to achieving biodiversity conservation and sustainable development goals in Canada and abroad.

We thank the KX Resource Stewardship Authority and specifically Douglas Neasloss and Evan Loveless for their support and guidance of this research. We also thank our colleagues in the field including Stephen Neasloss, Sarein Basi-Primeau, Chantal Pronteau, John May, Mitch Robinson, Santana Edgar, Krista Duncan, Sam Harrison, and Rosie Child. We thank Andra Forney, Nick Reynolds, Blake Evans, and Richard Hebda for their guidance and training. We also thank Jacob Earnshaw, Robert Gustas, Jordan Benner, and Ken Lertzman for their guidance and input that greatly improved this work. Finally, we thank Diana Cooper with the British Columbia Archaeology Branch for her assistance with the RAAD database. B.C. DeRoy was supported by Raincoast Conservation Foundation Graduate Fellowship, Vernon Brown and C.N. Service by the KX Stewardship Authority and Spirit Bear Research Foundation, M. Leclerc by Raincoast Conservation Foundation and Fonds de Recherche Nature et Technologies (FRQNT #255026), I. McKechnie by Hakai Institute and NSERC-DG (#531246), and C.T Darimont by Raincoast Conservation and SkyeMikko Foundations as well as the National Science and Engineering Research Council (NSERC) Discovery Grant (#435683). We also thank the anonymous reviewers who considerably improved this manuscript.