Volume 1, Issue 1 p. 87-102
Open Access

Effects of climatically shifting species distributions on biocultural relationships

Matthew O. Bond

Matthew O. Bond

Department of Botany, University of Hawai'i at Mānoa, Honolulu, Hawai'i

Manaaki Whenua Landcare Research, Dunedin, New Zealand

Search for more papers by this author
Barbara J. Anderson

Barbara J. Anderson

Manaaki Whenua Landcare Research, Dunedin, New Zealand

Search for more papers by this author
Te Hemo Ata Henare

Te Hemo Ata Henare

NorthTec, Whangarei, New Zealand

Search for more papers by this author
Priscilla M. Wehi

Corresponding Author

Priscilla M. Wehi

Manaaki Whenua Landcare Research, Dunedin, New Zealand

Te Pūnaha Matatini, Auckland, New Zealand


Priscilla M. Wehi, Te Pūnaha Matatini, Manaaki Whenua Landcare Research, Dunedin, New Zealand.

Email: [email protected]

Search for more papers by this author
First published: 27 March 2019
Citations: 15



  1. Local and Indigenous Peoples play critical roles in safeguarding global biological and cultural diversity. However, species distribution modelling has yet to incorporate perspectives that assess threats to the linked biological and cultural systems of local and Indigenous Peoples.
  2. Here, we provide the first example of integrating species distribution modelling with benefit-relevant indicators. This novel approach assesses how human access to culturally important species may change over time.
  3. Focusing on two culturally significant species used by the Indigenous Māori people of New Zealand, we first identified predictor variables relevant to the habitat of each species. We used species distribution models (SDMs) to estimate the recent (1961–1990) potential distribution for each species based on occurrence records and predictor variables, then generated future climate suitability maps.
  4. Our models show that future suitability for one species shifts to the south, in line with changes in temperature and precipitation, while the second species range expands into higher latitudes, driven primarily by increased temperature. When we combined these models with knowledge of tribal boundaries and cultural practices, results indicated that these distributions might decrease access to culturally important plants. Future suitability for one species shifted substantially from where it is most valued for weaving, while the second species range expanded to include more of its primary medicinal users.
  5. Climate change-mediated shifts in the ranges of these species are likely to affect intergenerational human–environment relationships, sense of place, cultural identity and knowledge on a regional scale, as well as cultural identity and social cohesion on a national scale.
  6. By interpreting SDMs within a socioecological framework, this research illustrates a new approach to assessment of vulnerabilities to climate change and identifies strategies for adaptation.

A plain language summary is available for this article.



He wāhi nui nō ngā Iwi Kāinga, Iwi Taketake hoki ki te tiaki i te kanorau ā-ao o te koiora, o te ahurea. Heoi anō kāore anō kia tau te whakaahua i te horahanga o te momo koiora ki te whakauru i ngā tirohanga ā-Iwi Kāinga, ā-Iwi Taketake ki ngā haukotinga i ngā hononga o ngā tikanga ā-Iwi Kāinga, a-Iwi Taketake. He hāngai tonu tā mātou titiro ki ētehi momo hirahira ki te Ao Māori, ka matua tirohia ngā taurangi matapae o ngā whaitua oranga o ia momo. Whakamahia ai e mātou ngā tauira horahanga ā-momo (SOMS) he whakatau tata i te horahanga rāpea o te wā (1961–1990) o ia momo mai i te kōpae takahanga me ngā taurangi matapae, kātahi ka hangaia he mahere mō te arotau ā-ahurangi o te wā e heke mai nei. Kei ā mātou whakatauiratanga te kitenga ko te arotau ā-ahurangi o tētahi momo he nuku whakatetonga mai i ngā rerekētanga o te pāmahana me te whakatīwharatanga; ā ko tā te momo tuarua he haere whakateraki, tuatahi mai i te kaha kē ake o te pāmahana. Ki te tūhonotia tahitia ēnei whakatauiratanga ki te mōhiohio o ngā rohe me ngā tikanga ā-iwi, he kitenga tērā pea ka haukotia te wātea o ētahi koiora e hirahira ana ki te ahurea ā-iwi. Te arotau o tētahi momo i nukuhia nuitia mai i tōna wāhi hirahira mo te raranga, ā ko tō te tuarua i whakarahia ake kia uru mai tētahi tokomaha anō o ōna kaiwhakamahi i tōna rongoā. Mai i te rereke haere o te ahurangi i rerekē haere ai ngā tūāhua o ēnei momo he pānga pea ki te huāngatanga ā-whakapuruanga o te taiao ā-ira tangata; ki te tūrangawaewae; ki te tuakiri ā-ahurea; me te koi o te mārama ki te rohe me ona whanaungatanga, taea noatia te motu me ōna nā ano hoki. Ko tā tēnei rangahau he whakaahua hōu i te arotakenga o ngā whakaraerae ki te rerekētanga o te ahurangi, he tautuhi i ngā rautakinga o te takatūranga mā ngā tauira o te horahanga ā-momo ki roto o te anga oha-kaiao.


International policy recognizes Indigenous Peoples’ knowledge and customary use of biological resources as essential to biodiversity conservation (Ens, Scott, Rangers, Moritz, & Pirzl, 2016; International Union for Conservation of Nature [IUCN], 2016; United Nations, 2015a, 2015b). Since Indigenous Peoples manage 80% of the world's biodiversity on at least 28% of the planet's land surface (Garnett et al., 2018; Sobrevila, 2008), the stability and sustainability of their biocultural systems (linked biological and cultural systems) is of global concern. The Aichi Targets in the Strategic Plan for Biodiversity 2011–2020 signal the preferred pathway to meet the vision of the Convention on Biodiversity (United Nations, 2010) with the goal that ‘by 2050, biodiversity is valued, conserved, restored and wisely used, maintaining ecosystem services, sustaining a healthy planet and delivering benefits essential for all people’. The Hawai'i Commitments, developed by the 2016 IUCN World Conservation Congress, likewise emphasize the need for policy, laws and best practice based on equitable and sustainable use of natural resources, and recognize Indigenous Peoples as key stewards of the world's biodiversity (IUCN, 2016).

In recognition of the ‘inextricable link’ between culture and nature (Posey, 1988), the Hawai'i Commitments, along with prior global initiatives such as the United Nations Sustainable Development Goals and the Paris Agreement, stress that climate change is a particular threat for Indigenous Peoples (IUCN, 2016; United Nations, 2015a, 2015b), in part because Indigenous needs and rights are often marginalized (Nilsson, 2008; United Nations, 2009). In addition, Indigenous Peoples often inhabit areas predicted to have severe climate change impacts (IPCC, 2014; McLean, 2010). Therefore, Indigenous Peoples’ resource use and livelihoods are being quickly and disproportionately disadvantaged by changing rainfall patterns, increasing temperatures, sea level rise and other climate change impacts (IPCC, 2014; McLean, 2010; United Nations, 2009). Because Indigenous ecological knowledge or use of local ecosystems is linked to the diversity and conservation of these ecosystems (e.g., Wilder, O'Meara, Monti, & Nabhan, 2016), loss of biological diversity is strongly correlated with loss of Indigenous culture and language (e.g., Turvey et al., 2010). Thus, climate change-mediated local extirpations and ecological shifts (e.g., species range, abundance and phenology) are impacting Indigenous social systems (Durkalec, Furgal, Skinner, & Sheldon, 2015; Ford et al., 2013; Gómez-Baggethun, Corbera, & Reyes-García, 2013; McLean, 2010).

Disregarding biocultural practices in conservation actions can contribute to community opposition, project failure and even ecological damage (Ban et al., 2013; Tipa & Panelli, 2009; West & Brockington, 2006). Therefore, the knowledge and participation of Indigenous Peoples may be critical to prioritize ecosystems or species at risk under changing climatic conditions (Guerrero, McAllister, Corcoran, & Wilson, 2013; Sterling et al., 2017; Wehi & Lord, 2017). Indigenous People's relationships with the environment profoundly shape landscapes, so the biological and cultural qualities of landscapes cannot be separated (Berkes, 2008; Chan et al., 2011; Stephenson, 2008). These relationships are embedded in practices such as the gifting or exchange of natural resources, maintenance of resources as an obligation to past and future generations, and inclusion of places and resources in definitions of identity or ancestry; practices which require protection and use of accessible biological systems (Carey & Lydon, 2014; Pascual et al., 2017). Despite the critical ramifications for Indigenous identity, climate change studies that integrate cultural impacts on Indigenous Peoples with effective conservation actions are rare (Ban, Picard, & Vincent, 2009; Brugnach, Craps, & Dewulf, 2017).

Although conservation planning has begun to include spatial components to account for ecological and evolutionary processes (Rouget, Cowling, Pressey, & Richardson, 2003; Wilson, Cabeza, & Klein, 2009), it has largely failed to do the same for human–ecological interactions (Chan et al., 2011) or to consider how climate change might impact these interactions (Tschakert et al., 2017). Because Indigenous spatial concepts and relationships tend be dynamic, multilayered and context specific, they frequently differ from local government, or ecosystem, boundaries (Woodward & Lewis, 1998). Thus, integrating Indigenous boundaries in conservation prioritization and future planning is a critical step towards the inclusion of biocultural perspectives and practices.

This investigation is the first to map species range shifts using benefit-relevant indicators. By doing so, we examine how climate change may affect human access to resources, cultural practices and intergenerational relationships between people and landscape. Thus, we demonstrate how climate change modelling can embrace a socioecological systems approach to anticipate future risks and challenges in ecosystem management and biocultural conservation. We illustrate these principles using two plant species highly valued by indigenous Māori people in New Zealand. First, we calculate the recent (1961–1990) potential distribution of exemplar species using the relationship between climate, geophysical variables and occurrence records. Next, we generate future climate suitability maps, analyse these models in conjunction with ethnobiological knowledge and Māori tribal boundaries, and discuss cultural issues arising from predicted species distribution shifts. Finally, we highlight how these methods can be used by Indigenous Peoples, conservationists, scientists and policymakers to support biocultural conservation.


2.1 Exemplar plant species and biocultural system

Kūmarahou (Pomaderris kumeraho A.Cunn. ex Fenzl) and Kuta (Eleocharis sphacelata R.Br.) are culturally important plants used for medicinal and weaving purposes, respectively, in New Zealand. Both species contribute to all four ecosystem service categories of the Millennium Ecosystem Assessment (2005). Kūmarahou is an endemic shrub that flourishes on well-drained clay soils in northern North Island (Clarkson, Smale, Williams, Wiser, & Buxton, 2011; Enright, 1989; Figure 1b). Kūmarahou flowers and leaves are used by many Māori across New Zealand's North Island as the premier treatment for respiratory ailments both historically and currently (Brooker, Cooper, & Cambie, 1998; Māori Plant Use Database 2018a; Wehi & Wehi, 2010; Williams, 1996).

Details are in the caption following the image
Suitability maps for Kūmarahou (Pomaderris kumeraho A.Cunn. ex Fenzl), an endemic medicinal shrub, show that suitable climate for Kūmarahou is expanding south along the coasts of North and South Island. Suitability is represented by the median value of each 1 km2 grid cell across 70 species distribution model (SDM) projections (10 replicates of seven SDMs) at baseline (1961–1990) climate and across 910 species distribution model projections (10 replicates of seven SDMs using 13 General Circulation Models) for each of two future projections (2021–2050, 2051–2080) at low-emission and high-emission scenarios. Suitability is plotted on a scale from 0 (white, lowest probability of suitable conditions) to 1 (dark green, highest probability of suitable conditions). Uncertainty among models is represented by the standard deviation plotted on a scale from 0 (white, no difference between models) to 0.5 (black, maximum difference between models) in the inset map. Observed occurrences were compiled from fieldwork, online databases and herbarium records

Kuta, on the other hand, is a native wetland sedge distributed across New Zealand, Australia and New Guinea (Figure 2b and Figure S1a). Kuta is used for weaving by Indigenous Peoples throughout its global range (Akerman, 1998; Beck, Haworth, & Appleton, 2015; Feachem, 1973; Hole & Dutoit, 2005; Kapa, 2010; Wehi, 2006). In New Zealand, Kuta is especially valued in the northern regions of the North Island, where expert weavers transform its stems into highly prized mats and other items that are strongly associated with their tribal areas (Kapa, 2010; Wehi, 2006). Kuta also plays pivotal roles in food webs, community structure and primary productivity of New Zealand wetlands (Sorrell & Tanner, 1999), which are linked to Māori sustenance, culture, identity, and kinship (Tipa & Teirney, 2006). However, 90% of New Zealand wetlands have been destroyed since European colonization in the early 19th century (McGlone, 2009), and less than half of New Zealand's wild vegetation is now native (Cieraad, Walker, Price, & Barringer, 2015). Because of these changes, it is critical to identify and protect remaining habitat for species such as Kuta and Kūmarahou, and consider how habitat suitability might change in the future of continued climate change (King, Penny, & Severne, 2010).

Details are in the caption following the image
Suitability maps for Kuta (Eleocharis sphacelata R.Br.), a native sedge used for weaving. Kuta range has little change, but suitability is decreasing across North Island and increasing in southern South Island. Suitability is represented by the median value of each 1 km2 grid cell across 70 species distribution model (SDM) projections (10 replicates of seven SDMs) at baseline (1961–1990) climate and across 910 species distribution model projections (10 replicates of seven SDMs using 13 General Circulation Models) for each of two future projections (2021–2050, 2051–2080) at low-emission and high-emission scenarios. Suitability is plotted on a scale from 0 (white, lowest probability of suitable conditions) to 1 (dark green, highest probability of suitable conditions). Uncertainty among models is represented by the standard deviation plotted on a scale from 0 (white, no difference between models) to 0.5 (black, maximum difference between models) in the inset map. Observed occurrences were compiled from fieldwork, online databases, and herbarium records

In New Zealand, Māori families traditionally harvest resources within tribal boundaries and geographic regions as part of their cultural responsibility to maintain these resources (Harmsworth & Awatere, 2013; Stevens, 2006). The quality and accessibility of natural resources within tribal boundaries is linked to Māori cultural practices and spiritual wellbeing (Gibbs, 2003; Harmsworth & Awatere, 2013; Stevens, 2006; Wehi & Lord, 2017; Wehi & Wehi, 2010). High quality and abundant resources are a source of tribal prestige (e.g., Kāi Tahu as providers of Tītī, Puffinus griseus; Stevens, 2006), and may be ceremonially gifted, sold or exchanged with people from other tribal regions. Thus, access to regionally restricted resources (such as Kūmarahou) or populations with esteemed phenotypes (such as Kuta from Northland, Kapa, 2010; Wehi, 2006) affects tribal prestige and social connections on a national level (Gibbs, 2003; Harmsworth & Awatere, 2013).

2.2 Predictor variables

We selected predictor variables based on previous ecological studies and cultivation knowledge of Kūmarahou and Kuta (Table S1). For both analyses, Variance Inflation Factor (VIF) values (Table S1) were calculated using the usdm vif command (Naimi, Hamm, Groen, Skidmore, & Toxopeus, 2014). Continuous variables with VIF values greater than three were removed to avoid multicollinearity (Guisan, Thuiller, & Zimmermann, 2017; Zuur, Ieno, & Smith, 2007). For all remaining variables, we calculated the variable correlation for all thinned presence records (Figure S5) using the pairs command of the raster package (Hijmans, 2017), showing all variable correlations to be below the multicollinearity threshold of 0.7 (Guisan et al., 2017). To improve SDM accuracy by limiting model complexity, variables that did not increase the area under the receiver operating characteristic curve (AUC) or True Skill Statistic (TSS) were not included in SDMs (Merow et al., 2014).

Kūmarahou variables included in the model were mean temperature of coldest quarter, precipitation seasonality (coefficient of variation; Hijmans, Cameron, Parra, Jones, & Jarvis, 2005), sand content at depth 0.3 m, soil organic carbon content at depth 0.3 m, and soil pH at depth 0.3 m (Hengl et al., 2014). We included the mean temperature of the coldest quarter as an indicator of frost tenderness (Clayton-Greene, 1978; Grace, 1987; Kauriparknurseries.com, 2017), while all other variables are associated with the well-drained, nutrient-poor, acidic soils in which Kūmarahou often grows (Clarkson et al., 2011; Enright, 1989; Kauriparknurseries.com, 2017). Highly collinear variables, such as alternate precipitation variables, were removed from the analysis.

Kuta predictor variables were mean temperature of the coldest quarter, annual mean precipitation (Hijmans et al., 2005), slope (De Ferranti, 2009), soil bulk density at soil surface and silt content at soil surface (Hengl et al., 2014). The mean temperature of coldest quarter was chosen to model growing season length and frost tenderness (Grace, 1987), while all other variables were selected to approximate wetlands (Asaeda, Manatunge, Rajapakse, & Fujino, 2006; Bantilan-Smith, Bruland, MacKenzie, Henry, & Ryder, 2009; Campbell, Cole, & Brooks, 2002; Sorrell, Tanner, & Sukias, 2002; U.S. Environmental Protection Agency, 2008). Future projections of each climate variable were obtained from 13 standard General Circulation Models (GCMs) of world climate (Emori et al., 2016; Table S2) for both a high (RCP 8.5) and low (RCP 2.6) greenhouse gas emission scenarios at two future timeframes (2021–2051 and 2051–2080). In RCP 2.6, the increase in global mean temperature is less than 2°C, but observed climate trends are already tracking above the ‘worst case scenario’, RCP 8.5 (Peters et al., 2012). Therefore, we selected RCP 2.6 and RCP 8.5 as optimistic and realistic climate change scenarios respectively.

2.3 Species distribution modelling

All analyses presented in this study were performed in R software (version 3.5.1; R core team, 2018). The distribution data for Kūmarahou (n = 406) and Kuta (n = 3559) were derived from databases, herbarium specimens and field observations (Table S3). To improve species distribution model (SDM) performance, occurrence records were cleaned; points were removed if they were obviously inaccurate (e.g., points in the ocean), missing coordinates, located exactly at the centre of the country (which suggests incorrect georeferencing) or had fewer than three decimal digits in the latitude or longitude (which suggests insufficient spatial accuracy; Gueta & Carmel, 2016). True absences were not available for these species; to account for climatic bias of presence data, pseudo-absences were randomly distributed within a buffer region 200 km from presence points (Figures S2 and S3) (Barbet-Massin, Jiguet, Albert, & Thuiller, 2012). Buffer region size was determined by varying the size while monitoring changes in model performance and predictions; optimum distance in this study and in another study (VanDerWal, Shoo, Graham, & Williams, 2009) was 200 km. To reduce spatial sampling bias, counter residual spatial autocorrelation and improve SDM performance (Boria, Olson, Goodman, & Anderson, 2014; de Oliveira, Rangel, Lima-Ribeiro, Terribile, & Diniz-Filho, 2014; Hijmans, 2012), occurrence and pseudo-absence points were spatially thinned by randomly removing a single point from within a 10 km radius using the spThin package (Aiello-Lammens, Boria, Radosavljevic, Vilela, & Anderson, 2015). A 10 km radius was chosen because of the environmental heterogeneity of this system, which includes many islands (Liu, Vellend, Wang, & Yu, 2018; Radosavljevic & Anderson, 2014). The resulting thinned data samples retained 105 and 941 points for Kūmarahou and Kuta respectively (Figures S2 and S3). Thinned occurrence records were projected onto a 30 arc-second2 (~1 km2) gridded map matching the predictor variable data for both Kūmarahou (105 grid cells) and Kuta (941 grid cells). To optimize SDMs, Kūmarahou and Kuta pseudo-absences (210 and 1,882, respectively; Figures S2 and S3) were selected for a ratio of pseudo-absence to presence of 2:1 (Barbet-Massin et al., 2012; Liu, Newell, & White, 2018).

The recent potential distribution for each species was determined from the relationship between the gridded presence/pseudo-absence data and environmental variables, including climate under recent (1961–1990) conditions (Anderson et al., 2009). We used seven standard SDMs available in the BIOMOD2 package (Thuiller, Georges, & Engler, 2013) to increase accuracy and ensure that our results were robust to differences in model type (Elith et al., 2006; Liu, Newell, & White, 2018): Artificial Neural Network (ANN), Flexible Discriminant Analysis (FDA), Generalized Boosting Models (GBM), Generalized Additive Model (GAM), Generalized Linear Model (GLM), Multiple Adaptive Regression Splines (MARS), and Random Forest (RF). For each species, SDMs were run in 10 replicates; each replicate was built using 70% of the presence and pseudo-absence data (randomly sampled) and evaluated with remaining 30% of presence and pseudo-absence data (Guisan et al., 2017).

To assess performance of SDMs (Figure S4), the AUC and TSS were calculated with the BIOMOD2 get_evaluations function, using the default cut-off to maximize sensitivity and specificity (Barbet-Massin et al., 2012; Thuiller et al., 2013). Relative importance of variables was similarly obtained with the BIOMOD2 get_variables_importance function (Figure 3). Ecological realism of SDMs was assessed by averaging the response curves from each SDM for each variable (Figure 4; Guisan et al., 2017; Merow et al., 2014).

Details are in the caption following the image
Contribution (in %) of each variable for Kūmarahou and Kuta species distribution models. For Kūmarahou, mean temperature of the coldest quarter was the main predictor variable. For Kuta, soil bulk density was the main predictor variable. Key: Cold. Temp = mean temperature of coldest quarter, Precip. Seas. = precipitation seasonality (coefficient of variation, Soil Sand = sand content (50–2,000 μm) mass fraction in % at depth 0.3 m, Soil pH = Soil pH at depth 0.3 m, Soil Carbon = soil organic carbon content at 0.3 m, Ann. Precip. = annual mean precipitation, Slope = % slope, Soil Carbon = soil organic carbon content (fine earth fraction) in g/kg at depth 0.3 m, Bulk Density = bulk density (fine earth fraction) in kg/cubic metres at soil surface, Soil Silt = silt content (2–50 μm) mass fraction in % at soil surface. Boxplots show median, upper and lower quartiles, and whiskers that extend to the most extreme data point that is no more than 1.5 times the interquartile range from the box, and outliers
Details are in the caption following the image
Response curves for the variables used in Species Distribution Models (SDMs) for Kūmarahou (a) and Kuta (b). Curves depict average probability of occurrence from 70 species distribution models (SDMs; 10 replicates of seven SDMs) along the environmental gradients. Key: Cold Temperature = Mean temperature of the coldest quarter, Precipitation Seasonality = precipitation seasonality (coefficient of variation), Soil pH = soil pH in H2O at depth 0.3 m, Sand = sand content (50–2,000 μm) mass fraction in % at depth 0.3 m, Soil Organic Carbon = soil organic carbon content (fine earth fraction) in g/kg at depth 0.3 m, Annual Precipitation = annual mean precipitation, Slope = % slope, Bulk Density = bulk density (fine earth fraction) in kg/cubic metres at soil surface, Silt = silt content (2–50 μm) mass fraction in % at soil surface

Recent potential suitability maps (Figures 1a, 2a; Figure S1b) were generated using the median and standard deviation of suitability values over 70 SDM projections for each grid square. Future suitability was projected using each combination of SDM and GCM at two future timeframes and two emission scenarios, giving a total of 3,640 model projections for each species (7 SDMs × 10 replicates × 13 GCMs × 2 emission scenarios × 2 future timeframes). Future suitability maps (Figures 1c–f and 2c–f) were generated by calculating the median and standard deviation of suitability values at each grid square from all 910 projections for each of the four combinations of emission scenario and future timeframe. Recent potential and future suitability maps show the probability of suitable environmental conditions on a scale from 0 (least) to 1 (most), and regions of uncertainty (the standard deviation of models, which can range from 0 to 0.5; Brito et al., 2011; Johnson, Kotz, & Balakrishnan, 1996; Martínez-Freiría, Argaz, Fahd, & Brito, 2013).

2.4 Cultural impacts of climate change

To link climate change-mediated species range shifts with impacts on use of these species by Indigenous Peoples, we used benefit-relevant indicators (also known as ecosystem service indicators), which address how accessible or valuable a natural resource is for a specific human use (Boyd et al., 2015; Olander et al., 2018; Ringold, Boyd, Landers, & Weber, 2013). To evaluate access to Kuta and Kūmarahou, we identified cultural boundaries used for historic and current harvest. Although a number of such regions are documented, we highlight four important harvesting regions from geographic extremes of their recent potential distributions (Figures 5 and 6). Kūmarahou is widely used across northern New Zealand. It is harvested opportunistically from a variety of sites, but preferentially within the harvester's tribal territory (Wehi & Wehi, 2010). Therefore, we interpreted Kūmarahou access in relation to tribal boundaries (Te Puni Kōkiri, 2016). As examples, we selected areas from four tribes with documented use of Kūmarahou: Ngāpuhi (Brooker et al., 1998; Williams, 1996), Ngāti Maniapoto (Māori Plant Use Database, 2018a; Wehi & Wehi, 2010), Waikato (Māori Plant Use Database, 2018a; Wehi & Wehi, 2010), and Whakatōhea (Māori Plant Use Database, 2018a).

Details are in the caption following the image
Change in the projected future suitability for Kūmarahou differs across tribal regions. Relative suitability is increasing in four selected iwi (Māori tribal groups; boundaries represented as black lines) where Kūmarahou is used medicinally. The suitability values of every 1 km2 grid cell within each region are shown in a boxplot for the recent potential distribution (white; 1961–1990); two boxplots for a low-emission scenario (light grey; RCP 2.6 2021–2050 and 2051–2070); and two boxplots for a high-emission scenario (dark grey; RCP 8.5 2021–2050 and 2051–2070). Suitability values of each 1 km2 grid cell are the median values of all recent (70) and future (910) model projections. Boxplots show median, upper and lower quartiles, and whiskers that extend to the most extreme data point that is no more than 1.5 times the interquartile range from the box, and outliers. n = number of 1 km2 map units per iwi. Iwi (Māori tribal group) boundaries are represented as grey lines on the central panel
Details are in the caption following the image
Change in the projected future suitability for Kuta differs across harvesting sites. Relative suitability for selected harvest sites (1 km2 map units where Kuta is currently harvested for weaving) is decreasing in northern North Island and stable or increasing further south. Iwi (Māori tribal groups) boundaries are shown in grey on the central map. For each 1 km2 grid cell, the suitability values of every recent (70) and future (910) model projection are shown in a boxplot for the recent potential distribution (white; 1961–1990); two boxplots for a low-emission scenario (light grey; RCP 2.6 2021–2050 and 2051–2070); and two boxplots for a high-emission scenario (dark grey; RCP 8.5 2021–2050, and 2051–2070). Boxplots show median, upper and lower quartiles, and whiskers that extend to the most extreme data point that is no more than 1.5 times the interquartile range from the box, and outliers

Although Kuta is used by weavers throughout New Zealand, it is most commonly used by weavers in the northern and upper central regions the North Island (Hole & Dutoit, 2005; Kapa, 2010; Wehi, 2006). Weavers typically return to the same harvesting site each year, often the same site that has been used for generations. These sites are highly prized and carefully used to support future harvests (Kapa, 2010; Wehi, 2006). Therefore, we interpreted Kuta access by analysing 1 km2 map units to approximate individual harvesting sites. As examples, we selected four notable harvesting sites for Kuta: Lake Mahinapua, used by Kāi Tahu weavers (Hole & Dutoit, 2005); Lake Ngātu, perhaps the most famous Kuta harvesting site in the country, favoured by weavers from many tribal areas including Te Aupōuri, Ngāi Takoto and Te Rarawa (Kapa, 2010; Wehi, 2006); Lake Kaiwai, a traditional Ngāpuhi site (Te Hemo Ata Henare, personal communication, October 22, 2018); Lake Rotomā, a traditional site for Te Arawa weavers (Kapa, 2010; MTEC Consultants Ltd, Wildland Consultants Ltd, & Toi Ora Associates, 2012).

2.5 Statistical analyses

For each example region, the difference in median suitability across the recent potential and all four future scenarios was tested using Kruskal–Wallis one-way ANOVA, which is resistant to skew and can compare differently shaped distributions.


Within New Zealand, the recent potential distribution of both Kūmarahou (Figure 1a) and Kuta (Figure 2a) is centred on northern North Island. For Kūmarahou and Kuta, median AUC was 0.87 and 0.86 respectively (AUC scale: 0–1, perfect score = 1, no skill = 0.5), while median TSS was 0.68 and 0.59 respectively (TSS scale: −1 to 1, perfect score = 1, no skill = 0; Figure S4). For Kūmarahou, the mean temperature of the coldest quarter is the most important predictor variable in our SDMs (Figure 3a), with almost no probability of occurrence below 6°C (Figure 4a). The recent potential distribution for Kuta is largely driven by soil bulk density (Figure 4b), with very little probability of occurrence above 1,400 kg/m3 (Figure 4b).

Future projected suitability for both Kūmarahou and Kuta increases at the southern edge of the recent potential distributions (Figures 1c–f and 2c–f). The median projected future suitability for Kūmarahou increases significantly across all four of the selected tribal regions (Figure 5; Ngāpuhi: H = 865.64, 4 df, p < 2.2 × 10−16; Ngāti Maniapoto: H = 24,597, df = 4, p < 2.2 × 10−16; Whakatōhea: H = 6,632.1, 4 df, p < 2.2 × 10−16; and Waikato: H = 5,583.6, 4 df, p < 2.2 × 10−16). Recent potential suitability for Kūmarahou is highest in the Ngāpuhi region and remains so across all future emission scenarios and time projections. In contrast, recent potential suitability in Waikato is generally moderate, but increases considerably in future scenarios at both low and high emission scenarios. In the tribal regions of Whakatōhea and Ngāti Maniapoto, recent potential suitability is generally very low, but in a future high-emission scenario, suitability of Whakatōhea and Waikato regions become similar and suitability of Ngāti Maniapoto region approaches the recent potential suitability of Waikato region. In a future low-carbon world, Whakatōhea and Ngāti Maniapoto regions reach similar levels of suitability for Kūmarahou. In the high-carbon world, there is a large increase in suitability, but uncertainty increases at the northern and southern edges of the projection.

For Kuta, future projected suitability decreases in northern New Zealand and increases in southern New Zealand (see Figures 2a and 2f). A comparison of relative change in the projected future suitability at four selected traditional harvesting sites of Kuta (Figure 6) shows a large suitability decrease in northern North Island, but only a slight suitability decrease further south. At Lakes Ngātu (H = 836.3, 4 df, p < 2.2 × 10−16) and Kaiwai (H = 279.19, 4 df, p < 2.2 × 10−16), the median suitability decreases significantly with time and increasing emissions. At Lakes Rotomā (H = 120.88, 4 df, p < 2.2 × 10−16) and Mahinapua (H = 130.47, 4 df, p < 2.2 × 10−16), the median suitability also decreases significantly under some time and emission scenarios, but less dramatically than in the lakes further north.


This paper demonstrates how to interpret SDMs within a sociocultural context to identify changes in human access to resources. Including socioecological context is the first step towards advancing climate change spatial conservation prioritization beyond mapping ecological and evolutionary processes to mapping important human–ecological processes. For our exemplar species, future suitability will be reduced in some harvesting regions and increased in others (Figures 5, 6). The results show that climate change will affect access to these culturally valued plant species, with the potential to place biocultural connections at risk where human access to resources does not shift with species shifts. Mapping future suitability for currently accessible and culturally significant populations of useful species exposes risks to biocultural connections and reveals where and how adaptations should be targeted to support biocultural resilience.

4.1 Species distribution models

Model fit across the 70 SDMs was good, as assessed by AUC and TSS according to the thresholds of Hodd, Bourke, and Skeffington (2014). The importance of temperature in Kūmarahou distribution (Figure 4a) confirms that this species is frost sensitive (Clayton-Greene, 1978; Grace, 1987; Kauriparknurseries.com 2017). Projected southern expansion of Kūmarahou distribution (Figures 1, 5) is consistent with predicted temperature increases in New Zealand (Royal Society of New Zealand, 2016). The importance of soil bulk density in Kuta distribution (Figure 4b) aligns with the observed soil bulk density in mineral and organic wetlands (U.S. Environmental Protection Agency, 2008), which are both habitat for Kuta. Because SDMs for one species can identify habitat for species with similar fundamental niches (Araújo & Peterson, 2012), some habitat outside the observed range of Kuta (Figure S1a), such as in Indonesia and Fiji, is identified as suitable on the recent potential suitability map (Figure S1b). This habitat is likely occupied by Eleocharis dulcis, a closely related sympatric species (Roalson, Hinchliff, Trevisan, & da Silva, 2010), whose range also includes Sulawesi, Java, Malaku Islands, Lesser Sunda Islands, New Caledonia, Fiji (Kew World Checklist of Selected Plant Families 2018). Projected changes in Kuta distribution (Figures 2, 6) are consistent with predicted rainfall change in New Zealand (Royal Society of New Zealand, 2016) and observed climate change impacts, which occur at faster rates in wetlands than in most other ecosystems (Loarie et al., 2009).

4.2 Biocultural implications

For many species, such as Kūmarahou (Figures 1, 5), range is predicted to expand under climate change, rather than shift or decrease as described for many other species (Chen, Hill, Ohlemüller, Roy, & Thomas, 2011; Parmesan, 2006). In these results, as suitability for Kūmarahou increases, it may be added to regional pharmacopoeia, or replace currently used species. For example, Kawakawa (Macropiper excelsum Miq.) is currently used in many parts of New Zealand to treat illnesses that those further north might treat with Kūmarahou (Māori Plant Use Database, 2018b). Harvesting of species with expanding ranges, such as Kūmarahou, instead of previously harvested species, such as Kawakawa, may alter dynamics of ecological communities (Dao & Hölscher, 2018; Ticktin, 2004; Trauernicht & Ticktin, 2005) and cultural knowledge (Clavero, 2014; Hall, 2009; Shackleton et al., 2007). Species expansions can affect sociocultural practices, especially for regionally restricted species such as Kūmarahou. For example, Kūmarahou is presently so highly valued in the Tainui tribal confederation (which includes Waikato and Ngāti Maniapoto, Figure 5), that plant material is requested from Northland by some medicinal users. The dramatic increase in projected suitability for Ngāti Maniapoto and other southern iwi may facilitate local harvesting of Kūmarahou in coming decades. This kind of species expansion could affect inter-tribal reciprocity and gifting practices (Henrich & Gil-White, 2001; Kawharu, 2000; Roberts, Norman, Minhinnick, Wihongi, & Kirkwood, 1995; Thomas, 1991; Walter, Jacomb, & Bowron-Muth, 2010), because similar increases in access to prestige goods have destabilized other Indigenous sociocultural systems (Bayart, 1978; Peregrine, 1999; Wesson, 2008).

Sociocultural factors also add complexity to assessing the ecological effects of another common response to climate change: range contractions (Parmesan, 2006). For example, projected changes in Kuta distribution (Figures 2, 6) suggest that this species will remain abundant. However, with the addition of tribal boundary data, we show that decreasing suitability for Kuta in Northland is likely to inhibit local Kuta availability and continued use of harvest sites that have been treasured for generations, particularly Lake Ngātu. Reduced local access to species is typically accompanied by a loss of biocultural knowledge (Aswani, Lemahieu, & Sauer, 2018; Hanazaki, Herbst, Marques, & Vandebroek, 2013; Soga & Gaston, 2018). Using traditional Kuta harvest sites remains important for cultural identity and direct ancestral connections to the environment (Harmsworth & Awatere, 2013; Kapa, 2010; Kawharu, 2000; Roberts et al., 1995), especially in the northern and central regions of New Zealand's North Island (Kapa, 2010; Wehi, 2006).

Ecological changes like those predicted in this study with climate change are a critical issue for Indigenous Peoples around the world, because they transform resource availability and landscape in ways that affect cultural identity, sense of place, knowledge, and social cohesion (Foale, 2008; Norgaard & Reed, 2017; Tschakert et al., 2017). Nonetheless, cultural practices adapt to changing situations over time (Berkes, 2008; Graham, Dayton, & Erlandson, 2003; Masterson et al., 2017). Indigenous Peoples have successfully translocated many useful species, especially in the Pacific (e.g., Hall, 2009; Mcdowall, 1996; Whistler, 2009).However, it is also common, including among Māori weavers, to use social networks to obtain material, or harvest, in distant regions (Carey & Lydon, 2014; Robinson & Williams, 2001; Thomas, 1991; Walter et al., 2010). Yet others go to extraordinary lengths to access natural resources important to cultural practices themselves, despite difficulties such as crossing international boundaries and facing jail time (Addo, 2013; Stevens, 2006; Wehi & Lord, 2017). However, humans also experiment and substitute novel species in response to environmental changes, although these substitutes may not have the same qualities as the original resource (Athayde & Silva-Lugo, 2018; Berkes & Jolly, 2002; De Medeiros et al., 2012; Kujawska, Hilgert, Keller, & Gil, 2017; van Andel et al., 2014; Vick, 2011). Weavers can, and do, use alternative weaving plants to replace Kuta, but the colour, texture, and quality of Kuta are unique (Evans & Ngarimu, 2005; Kapa, 2010).

4.3 Conservation implications

Where biocultural connections are at risk, collaborative management strategies may be required to adapt valued practices for continuing climate change. However, biocultural research and conservation applications must allow Indigenous groups to maintain agency of their knowledge and skills and share them in culturally germane pathways (Sobrevila, 2008). Efforts to conserve species such as Kuta should focus on assisting local communities to identify accessible populations that have been used historically and are in areas least susceptible to potential negative effects of climate change. It may also be necessary to transfer biological material to habitats outside the recent potential ranges where modelling identifies high future suitability (Figure 2F). To ensure culturally suitable selection of plant material and locations, translocations should be guided by Indigenous communities. Establishing new populations collaboratively with local communities may increase species resilience to support continuing or even expanding cultural use. Because harvesting is targeted towards trait qualities within species that are important for cultural use, identifying and evaluating climate change effects on these traits is a further step towards long-term biocultural resilience.

Holistic strategies are critical for maintaining resource access because climate change is not the only factor affecting habitat suitability. In our exemplar species, geophysical variables explain a large part of recent potential species distribution (Figures 3 and 4). While pollution (human and agricultural), vegetation clearance, drainage, and altered land use (e.g., agriculture, recreation) are not included in our SDMs, they directly affect geophysical variables (i.e. soil structure and composition). Reducing these threats at a regional watershed level could dramatically increase local suitability, effectively mitigating the negative effects of climate change. Such community-led and holistic approaches to natural resource management align with Indigenous conservation ethics and facilitate transmission of cultural knowledge (Harmsworth & Awatere, 2013; Stevens, 2006).


The reciprocal relationships between humans and the natural world underpinned discussions at the 2016 IUCN World Conservation Congress. Aloha ‘āina, kaitiakitanga and related concepts of other Indigenous Peoples (embodying the reciprocal dependence of humans and the natural world) exemplify a ‘Culture of Conservation’, as highlighted by the Hawai'i Commitments. To be consistent with these perspectives, best practice climate change research must incorporate cultural resource use parameters and acknowledge the importance of traditional ecological knowledge and cultural values. Incorporating relevant cultural information to develop socioecological climate change models provides clear pathways that enhance adaptation and resilience planning. Through their alignment to United Nations principles and Sustainable Development Goals, such models will promote both human wellbeing and the protection of natural resources in the face of continued climate change. This will lift the standard for meeting global targets and shift potential solutions to those that focus on the retention of biocultural diversity and support community well-being.


We thank herbarium staff Leon Perrie (WELT), Elise Arnst (CHR), Reijel Gardiner (CANU), and Toni Cornes (WAIK), as well as Janice Lord and Mieke Kapa, who provided distribution records. Weavers and elders from Te Tai Tokerau provided invaluable support, especially Toi Te Rito Maihi, Ngapuhi Brown, Elaine Te Pania and Phoebe Rawhiti, as well as Christina Wirihana and Te Roopu Raranga Whatu o Aotearoa. We appreciate constructive feedback from Heather McMillen, Tom Etherington, Bill Lee and three anonymous reviewers. Rutherford Discovery Fellowships 12-LCR-001 and 14-LCR-001 supported BJA and PMW respectively. MOB was funded by EAPSI-1613891, the Royal Society of New Zealand and RDF 14-LCR-001, and a National Science Foundation Graduate Research Fellowship. The study sponsors played no role in the study design, in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the paper for publication.


    The authors declare no conflict of interest.


    P.M.W. conceived the research, from which M.O.B., B.J.A. and P.M.W. designed the project. T.H.A.H. led cultural aspects of the project, and P.M.W. and T.H.A.H. identified field sites for Kuta and Kūmarahou. M.O.B., P.M.W. and T.H.A.H. collected the cultural data. M.O.B. and B.J.A. performed the SDM analysis. All authors contributed to writing the manuscript.


    Our online Supporting Information contains public sources for variables used in SDMs (Table S1) and for observed records for Kūmarahou and Kuta used in species distribution modelling (Table S3).