Characterizing Tree Species Richness in Indigenous Peoples’ Lands: Addressing Often-Overlooked Data Limitations and Biases

Characterizing Tree Species Richness in Indigenous Peoples’ Lands: Addressing Often-Overlooked Data Limitations and Biases | Research Square Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Help Center Sign In Submit a Preprint Cite Share Download PDF Article Characterizing Tree Species Richness in Indigenous Peoples’ Lands: Addressing Often-Overlooked Data Limitations and Biases Sven Kock, Jorge C. Llopis, Miguel Fernández, Mónica Moraes R., and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8789639/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 10 You are reading this latest preprint version Abstract Several studies have characterized species richness distribution in Indigenous Peoples’ Lands (IPL) using global biodiversity datasets. However, these datasets contain substantial limitations and biases which may compromise the accuracy of spatial biodiversity assessments. In this study, we characterize tree species richness across IPL in Bolivia based on three different datasets to examine how spatial and taxonomic biases, as well as the data’s institutional provenance, shape characterization outcomes. Results closely reflect identified spatial and taxonomic biases, and as data generally come from various sources, the individual collections’ scope and our characterization do not align. We also notice that most data derive from international institutions without information about contributions from Indigenous Peoples, complicating sensitive interpretation and weakening visibility of Indigenous knowledge systems. Our findings underscore the limitations of relying solely on global biodiversity datasets to map species distributions in IPL and emphasize the need for locally grounded, ethically framed data collection practices. Biological sciences/Ecology Earth and environmental sciences/Ecology Earth and environmental sciences/Environmental sciences Spatial Analysis Bolivia biodiversity characterization IUCN Red List GBIF Indigenous Peoples’ Lands data biases data justice Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Biodiversity is rapidly declining at a global scale, driven primarily by land use change, overexploitation, and environmental pollution 1 . This decline represents a profound loss of nature’s intrinsic value 2 and results in a significant reduction of ecosystem services upon which human well-being depend 3 , making biodiversity loss a central concern in global sustainability efforts 4 . While mainstream conservation approaches have struggled to prevent this continuous degradation 5 , many Indigenous Peoples have been successful at maintaining sustainable ways of living for generations 6 . There is increasing recognition in both scientific and policy circles that Indigenous Peoples play globally important roles in biodiversity conservation 7 , 8 . However, Indigenous knowledge systems, which underpin many long-standing sustainable practices, are increasingly subject to considerable pressures, such as suppression, obliteration, assimilation and misappropriation 9 . In this context, numerous scientific analyses have investigated the spatial relation between Indigenous Peoples’ Lands (IPL) and different metrics of species richness and diversity. Several studies have focused on particular species groups such as mammals, bats or vertebrates, and have characterized the overlap between their distribution ranges and IPL 10 , 11 . Some of them have compared this overlap to that with Protected Areas (PA) and other areas (OA) 12 , 13 . Schuster et al. 14 used species ranges to identify the species richness in IPL, PA and OA 14 , and Seebens et al. 15 calculated the mean number of alien species in IPL and OA. These studies, which we call 'spatial characterizations’ in this article, aimed to characterize species composition or distribution in IPL (see Ref. 10 , 11 ), or compare them to those in other areas (see Ref. 12 – 15 ). The mentioned spatial characterizations draw on global biodiversity databases, such as the global International Union for Conservation of Nature (IUCN) Red List of Threatened Species 16 (hereafter Red List, see Ref. 10 – 14 ), or the database of the Global Biodiversity Information Facility 17 (hereafter GBIF, see Ref. 15 ). However, these databases are not designed to measure spatial species distribution, potentially undermining the accuracy and representativeness of these efforts. In fact, GBIF represents botanic collections accumulated over time 18 , so it cannot account for case-study-specific sample requirements, such as representative sites and species, or temporal regularity 19 . The Red List is a selective effort focused exclusively on assessing extinction risk, with no claim to spatial representativeness 20 . In general, it is well known that global inventories of biodiversity are incomplete 21 , 22 and suffer from considerable biases 23 . Numerous species are still undescribed 24 (i.e., Linnean shortfall, after Hortal et al. 25 ), and scientific knowledge about their distribution is meagre at best, particularly at the local scale required to distinguish between different area management types in spatial characterizations 26 (i.e., Wallacean shortfall, after Hortal et al. 25 ). In addition, biodiversity data are considerably shaped by the preferences and limitations of the individuals and institutions involved in their collections 23 . In this article, we assess how spatial and taxonomic biases in biodiversity datasets undermine the characterization of tree species richness in IPL, PA, and OA. Specifically, we focus on the case of Bolivia, a country with high biological and cultural diversity 27 – 30 , where, according to its national census, more than half of the population self-identifies as Indigenous 30 . As a basis for our assessment, we characterize tree species across these area types, applying methodologies similar to the spatial characterizations mentioned (Refs. 10 – 15 ). We conduct this characterization separately with datasets from two of the most used global databases, and from one national database. Then, we assess the spatial distribution of the datasets across Bolivia’s different ecoregions (described by Ibisch et al. 31 ), as well of the taxonomic distribution. We discuss the presence of biases and their influence on spatial characterizations by comparing the spatial and taxonomic distribution in the different datasets to each other, and to Ibisch et al.’s 31 descriptions. Moreover, to provide a deeper insight, we investigate which studies and institutions contribute to the datasets. We evaluate these dataset sources to determine whether their scopes align with our spatial characterizations and whether they include Indigenous Peoples’ knowledge and values. The three datasets we use are from the Red List, the GBIF, and the Bolivian Biodiversity Observation Network (hereafter BBON) 32 . The Red List provides estimated species ranges based on known occurrences, abiotic features, and expert scientific knowledge 33 . GBIF hosts the largest global occurrence database, offering a standardized framework that enables the integration of data from diverse sources and across extended timeframes 28 . BBON is the most comprehensive national species occurrence database for Bolivia, combining records from multiple sources, similar to GBIF. We focus on tree species, considering that, as of today, they have not been addressed in any spatial characterization related to IPL. Trees are particularly relevant because they are long-lived organisms that often embody enduring ethnobiological relationships, reflecting generations of place-based cultural practices and spiritual knowledge 34 . In Bolivia, numerous Indigenous communities maintain rich traditions of tree use and stewardship, including the cultivation of culturally-significant species in homegardens, the use of specific trees in rituals and healing practices, and the transmission of tree-related knowledge across multiple generations 35 – 37 . Results Spatial Characterization IPL cover 19% of the Bolivian Territory, with another 19% under PA status, and 4% with dual IPL and PA status (from here on IPL-PA). IPL thereby includes both Community Lands of Origin (TCO, according to their name in Spanish) and Indigenous Peasant Territories (TIOC, according to their name in Spanish), where third parties have legally acquired and respected rights. 38 IPL-PA areas are not independently designated, but represent zones where IPL and PA overlap spatially. Since IPL, PA, and IPL-PA areas overlap in these regions, the combined area type percentages exceed 100% by an amount equal to twice the percentage of IPL-PA. The remaining areas (Other Areas, OA) cover around 66% of Bolivia (Figure 1a). The relative estimated tree species richness over the different ecoregions is mapped in Figure 1b, based on the estimations by Ibisch et al. 31 . The highest richness is found in the northern part of the country, particularly the Southwestern Amazon, and the Yungas (ecoregions 1, and 6). In contrast, the Chaco Cerrado, the Flooded Savannas of the Pantanal, the Gran Chaco, the Mantane Chaco, and the Prepuna (ecoregions 2.4, 3.2, 5, 8, 9, and 10), located in the south and southeast, are estimated to have low tree species richness. Intermediate levels of species richness are estimated for all other Cerrado regions, the Flooded Savannas of the Llanos de Moxos, the Chiquitano Dry Forest, and the Tucumano-Bolivian Forest (ecoregions 2.1-2.3, 3.1, 4, and 7; see Table S2). The Puna ecoregions in the southwest, for which no tree species richness estimations are provided, are characterized as scarcely vegetated (ecoregion 11) or as too dry for any considerable vegetation (ecoregion 12). For Bolivia, the Red List provides ranges for 45 different tree species, the GBIF dataset 257,999 occurrences of 4,135 distinct tree species, and the BBON dataset 199,263 occurrences of 4,142 distinct tree species. Overlapping the species ranges from the Red List shows that most of them are concentrated in the central south (Figure 2a). The species richness hotspots identified through GBIF and BBON are more evenly distributed across the country. As they are based on observation points and not on ranges, they occur more punctually than those of the Red List, which cover all of Bolivia’s area (see Figure 2d and 2g). Many of the locations with the highest species richness in GBIF and BBON are in the northwest of the country. The middle column in Figure 2 (subfigures b, e, and h) presents the number of distinct species in each area type. OA has the highest values for all datasets. Fewer species are within PA and IPL, and the fewest in IPL-PA areas. Notably, this value relates to the area type sizes, as OA cover the highest and IPL-PA areas the lowest share of Bolivia. However, when considering species density (Figure 2f and 2i), which is the number of species per area size, IPL-PA has higher values and OA has lower values than IPL and PA. Based on the Red List data, higher relative species ranges are measured in PA and OA than in IPL and IPL-PA (Figure 2c), with differences between area types less pronounced than for species density based on GBIF and BBON. While the highest species density based on GBIF and BBON is more than six times higher than the lowest, the difference between the relative species ranges measured based on the Red List is no more than 30%. Overall, the three different calculations (present species, relative species range, and species density) yield notably different results with no shared apparent pattern. Moreover, the results for both the GBIF and the BBON datasets are very similar. Spatial distribution of the biodiversity data Across ecoregions (Figure 3), the highest number of species ranges in the Red List are found in the Inter-Andean Dry Forests (ecoregion 9), followed by the Yungas (ecoregion 6), and the Subhumid Puna (ecoregion 11). Slightly less ranges intersect with the Tucuman Bolivian Forests (ecoregion 7), the Eastern Cordillera (ecoregion 11.3), the montane Chaco (ecoregion 8), and the Gran Chaco (ecoregion 5). In the Amazon, Cerrado, savanna, and dry forest regions northeast of the Bolivian Yungas (ecoregions 1-4), as well as the Southern Dry Puna in the south-east (ecoregion 12), there are only few Red List species ranges. The distributions of GBIF and BBON are comparable to each other but differ considerably from those in the Red List. For both GBIF and BBON, most distinct species are found in the Pre-Andean Amazonian Forests (ecoregion 1.3) and Yungas (ecoregion 6), followed by the other Amazonian ecoregions (1.1, 1.2, 1.4, 1.5), and the Inter-Andean Dry Forests (ecoregion 9). The Pre-, Northern and Southern Puna (ecoregions 10 to 12), as well as most Cerrado ecoregions (2.1, 2.2, 2.4) stick out with particularly low species richness in the datasets. Zooming into the ecoregions, both the GBIF and BBON data are very clustered. In the GBIF dataset, 55 of the 11,189 grid cells (0.5% of the total area) concentrate more than 50% of the observations, while around 72% of the grid cells do not contain any observation. The BBON dataset shows a similar pattern, with 49 of 11,189 grid cells (0.4% of the total area) containing more than 50% of the observations, while 73% of the cells do not have any observation. Most of the grid cells with the highest number of observations are in the PA area type for both datasets (see Table S3). Relatedly, there are 2.5 and 1.9 times as many observations in PA than in IPL in GBIF and BBON, respectively. In general, the highest number of overall observations per area type in GBIF and BBON are in IPL-PA, followed by PA, IPL, and OA (See Table S4). Taxonomic distribution of the biodiversity data In the Red List, 21 out of 45 species, so almost half, are Cactaceae (Figure 4). The other 24 species are distributed across 17 families with a maximum of three species (<7% of the 45 species) per family. The families with most species in GBIF and BBON are Fabaceae and Rubiaceae. Together, they contribute almost 20% of more than 4,000 distinct species in each dataset, while they are not present in the Red List. The remaining species are distributed across 157 (GBIF) and 159 (BBON) families with no family with a share higher than 7%. Notably, only 22 of 4,135 species in GBIF and 20 of 4,233 in BBON (around 0.5%) are Cactaceae, the most abundant family in the Red List.  The average difference between the share of species in each family is 260% between Red List and GBIF, 312% between Red List and BBON, and 15% between GBIF and BBON, when only considering the families included in both respective datasets (see Table S5). Assessment of the dataset sources The Red List data was provided by five institutions (Figure 5a). 21 of the 45 species ranges are associated with a global cactus assessment; six are associated with the TROPICOS collection generated by the Missouri Botanical Garden (MBG) in cooperation with national institutions (according to the expert knowledge of M.M.R., although this information is not explicitly documented in the dataset metadata), three are associated with the Herbario Nacional de Bolivia (LPB), and three are associated with a joint effort between the MBG and the LPB. Three more entries are affiliated with Botanic Gardens Conservation International (BGCI), meaning that they were collected by Bolivian institutions with funding by BGCI. For the remaining five entries, no information about the origin is provided. Overall, at least 49% of the data are from international institutions. The GBIF dataset has occurrence entries from 322 different institutions, with almost 80% of the entries provided by six different herbaria. At least one-third of the herbaria is international (Figure 5b), particularly from Denmark, Spain, and the US. Around 78% of the entries are based on preserved specimens, 81% of which are provided by TROPICOS, which is associated with the data by all six herbaria, except the University of Aarhus Herbarium in Denmark (AAU). Another 22% of the entries in GBIF are observation data provided by the AAU (90%) and iNaturalist (9.4%). See Table S6 for more details. The BBON dataset does not include metadata about individual data sources. The authors report that 55.4% of the data included is hosted by institutions based outside Bolivia, and 44.6% by institutions based within the country. Moreover, 93.7% of their data is from voucher specimens and 6.3% from direct expert observations. However, these numbers are not comparable to those for the other datasets, because they are for all species in the entire dataset and not only the tree species we assess. Discussion The spatial characterization we conducted to assess tree species richness in IPL, PA, and OA in Bolivia generated strikingly different results based on the different datasets we used. As we will show, these differences are related to spatial and taxonomic biases in all datasets. A first indication of spatial biases is that the distribution of species richness across the country's ecoregions (Fig. 3 ) differs significantly between the Red List and GBIF and BBON. All three datasets indicate high species richness in the Yungas (ecoregion 6), which is generally consistent with analyses that found a mixture of Andean and Amazonian phytogeographical elements creates high levels of richness 39 , 40 . However, while the species ranges in the Red List further intersect with ecoregions to the southwest of the Yungas, most occurrences of distinct species in GBIF and BBON are recorded to the northeast, particularly in the Southeastern Amazon (ecoregion 1, Fig. 3 ). A second indication of spatial bias is that none of these distributions completely align with the expected distribution after Ibisch et al. 31 . According to their estimates, the highest tree species richness occurs in northern Bolivia, particularly within the Southwestern Amazon and the Yungas (ecoregions 1, and 6). However, few Red List species ranges intersect with the Southwestern Amazon (ecoregion 1). Instead, many species overlap with southern and southwestern ecoregions, such as the Inter-Andean Dry Forests, or the Chaco ecoregions (9, 8, and 5), which have low estimated species richness, and with the scarcely vegetated Northern Puna (ecoregion 11). GBIF and BBON have far less species in most of the Southeastern Amazon (ecoregion 1) than in the Yungas (ecoregion 6), though the Southeastern Amazon is Bolivia’s species-richest ecoregion in Ibisch et al.’s estimations 31 (see Table S2). Also, there are less different species in the Tucuman Forest (ecoregion 7) than in the Inter-Andean Dry Forests (ecoregion 9) and Gran Chaco (ecoregion 5). The identified spatial biases across ecoregions influence the spatial characterizations results because the different area types (IPL, IPL-PA, Pa, and OA) are unevenly distributed over the ecoregions (see Table S1 ). If an area type is more abundant in an ecoregion with a high data representation, it can likely be identified as more species rich. For example, many of the ecoregions that have a relatively higher species richness than expected in the Red List (particularly ecoregions 7, 8, 11) have almost no share of IPL-PA areas. This potentially contributes to the low relative species range measured based on the Red List (Fig. 2 c), in contrast to the high species densities measured based on GBIF and BBON (Figs. 2 f and i) in our spatial characterization. Data are also unevenly distributed within ecoregions. In the Red List, the mapped distributions do not reflect characteristic landscape features. For example, the Central Andean and Dry Puna (ecoregions 4 and 5), have valleys that are more vegetated than the surrounding dry environment 31 . However, as shown in Fig. 2 a, species distributions in southwestern Bolivia do not resemble the shape of valleys. Considering that the Red List ranges are estimations informed by data and expert knowledge 33 , the polygons were likely drawn too coarsely to capture such small-scale features. If this is the case, the distribution is also unlikely to reflect potential differences in species richness between the different area types. The uneven distribution of GBIF and BBON within ecoregions is demonstrated by their clustering. This issue has already been identified by Beck et al. 41 for GBIF. Data clustering can be expected to have a significant impact on spatial characterization because whatever area type is in the same region as the cluster cells is favored in the quantitative assessments. For example, both GBIF and BBON have an overproportionate share of cluster cells in IPL-PA and PA, in relation to the respective area sizes (see Table S1 and S3). This potentially contributes to the high species densities measured for these area types in our spatial characterization (see Figs. 2 f and i). In general, the total number of observations is distributed unevenly across the different area types (see Table S4), with two to three times more observations per area in IPL-PA and PA than in IPL and OA. This bias clearly relates to the patterns identified in our characterization (Figs. 2 f and 2 i). Other than in the case study by Beck et al. 41 in Switzerland, clustering cannot be resolved for Bolivia by sub-sampling because for much of the area, no data are available. Considering that biodiversity is generally better documented in the Global North than in the Global South 21 , 23 , this issue likely affects other regions in the Global South, too. In addition to spatial biases, we also identified taxonomic biases. A first indication for them is that the distribution of species across families differed substantially between the Red List, on the one hand, and the GBIF and BBON datasets, on the other. If all datasets accurately represented Bolivia’s biodiversity, the actual distribution of species across families would be reflected in them. In that case, all datasets would be similar. The deviations between the three datasets therefore indicate taxonomic biases, either in the Red List, in GBIF and BBON, or in all three datasets. A second indication for taxonomic biases is that the taxonomic distribution in the datasets, particularly the Red List, does not align with Bolivia’s characteristic and native tree species shares (see Ref. 31 , 42 ). The Red List overrepresents Cactaceae: almost half of the species in the dataset are from the family. However, it constitutes only around 1% of the country’s native tree species 42 and is characteristic of southwestern ecoregions, such as the Dry Chaco (ecoregion 2), but not of Bolivia’s tree-rich ecoregions 31 . The taxonomic distribution in GBIF and BBON are similar to that in Moraes et al.’s 42 list of Bolivia’s native tree species, indicating a better representativeness. Taxonomic biases affect spatial characterizations because area types located in the same regions as overrepresented families or species are more likely to be identified as species rich. For example, the Red List is highly dominated by Cactaceae species, which are typical found in the Dry Chaco, the Inter-Andean Dry Forests and the Puna ecoregions 31 (ecoregions 5, 9, and 10–12). The Puna ecoregions intersect with particularly many species in the Red List, but not in the GBIF and BBON. These ecoregions tend to have below-average shares of IPL and IPL-PA, and above-average shares of PA and OA, reflecting the divergence between the characterization results from the Red List, and the GBIF and BBON (Figs. 2 c, 2 f, and 2 i). To gain deeper insight into the observed spatial and taxonomic biases, we examined the dataset sources. All three datasets combine data from various sources, likely merging information with differing scopes 18 . This aggregation results in non-representative sampling of both locations and species, which are crucial factors for characterizing regional biodiversity 19 . For instance, almost half of the Red List species ranges originated in the Global Cactus Assessment by Goettsch et al. 43 , which evaluated global cactus threat status. As our spatial characterization targets all tree species rather than cacti alone, the sampled species are unrepresentative for our study (or any study with similar objectives) and help explain the taxonomic bias identified before. In GBIF, around one fifth of the entries (those by AAU) correspond to a study by Cámara-Leret et al. 44 , who inventoried palm species in 96 different sites across the Western Amazon to correlate them to the respective biophysical conditions. The GBIF dataset has up to 3,400 species records at these observation sites, which contributes to the clustering effect, and results in an unevenly high likelihood of detecting elevated species richness there. Most of Cámara-Leret et al.’s 44 sites were in the Southwestern Amazon (ecoregion 1, 42%) and the Yungas (ecoregion 6, 23%), so neither their spatial scope nor their taxonomic focus on palm species aligns with the scope of our spatial characterization. To account for potential biases resulting from measuring distribution on a national scale based on global datasets, we included an occurrence dataset from the BBON database. BBON is comparable to GBIF, as both combine available specimen data and direct expert observations. However, the scope of the BBON database is national, and GBIF international. Moreover, the dataset retrieved from GBIF has nearly 30% more observations than that from BBON. Possibly this is because GBIF is updated regularly, whereas BBON was last updated in 2015. Despite the difference in scope and size, the results based on both datasets are similar throughout all analyses, indicating that these factors play a minor role. This observation relates to our previous discussion: Regardless of the datasets’ overall spatial scope, they all integrate sources with varying scopes. As a result, most of the studies they rely on are likely to differ in spatial scope from the spatial characterization for which they are used. The observation that the distributions are not affected by the dataset size supports Chapman et al.’s 23 argument that collecting more data cannot be expected to mitigate biases if the individual and structural factors shaping data-generation processes remain unchanged. As preferences and limitations of the individuals and institutions involved in data generation shape the datasets 23 , information on these contextual factors is paramount for understanding the data 45 . Given our focus on IPL, we were particularly interested in whether Indigenous knowledge systems were considered in data collection efforts. Since the datasets do not include this information, we approximated it by identifying the host countries of contributing studies and institutions. A great share of the data in both the Red List and GBIF has been collected by non-Bolivian institutions. For the Red List, they stem from an international study, with no explicit acknowledgment of Bolivian participation. For GBIF, they were largely provided by herbaria in the Global North. This disparity between plant locations and data origin reflects a broader pattern of colonial legacies in global specimen collections 46 , 47 . Despite these insights, our analyses have important limitations in identifying the role of Bolivian Indigenous Peoples, and Bolivian individuals in general, in data collection. We cannot rule out the possibility that (Indigenous) Bolivian scientists and other individuals actively participated in data generation. In fact, it is known that many Indigenous Peoples have played a fundamental role in such scientific efforts 48 . Providing more information about the context in which the data was collected would help to interpret the results of the spatial characterizations more accurately. For example, potential limitations and biases related to the collectors’ access to certain areas 25 , 49 , 50 and their interests 23 , 51 , 52 could be considered. Das and Lowe 45 pointed out that transparency regarding Indigenous Peoples’ roles in data generation could also enhance their visibility and engage audiences from countries with colonial histories. After Bebbington 53 , empowering and involving Indigenous Peoples is essential to avoid deepening existing power imbalances when there is international interest in the local environmental resources. The biodiversity of the IPL is often considered to be of international interest, particularly due to multilateral agreements aimed at strengthening conservation efforts 54 , and the potential economic value of the biological data 55 . The numerous spatial characterizations of biodiversity in IPL in the literature also reflect this global relevance. However, the lack of information about Indigenous participation in the data collection obscures Indigenous roles 45 , potentially reinforcing existing inequalities, as Bebbington 53 described. One more factor that might obscure the role of Indigenous Peoples is that the framework to characterize species richness based on global datasets is grounded in Western science rather than Indigenous knowledge systems. As Devictor and Bensaude-Vincent 18 argue, the principle of compiling species data from diverse sources reflects a Western paradigm of managing and studying global biodiversity. They note that while global biodiversity data are instrumental for defining and monitoring policy goals, the simplifications required to merge data from different sources complicate the investigation of ecological systems. Fernández-Llamazares et al. 56 reinforce this idea by emphasizing that the environmental stewardship of Indigenous Peoples cannot be comprehensively described based solely on species numbers but rather must account for the complex interrelations between Indigenous Peoples and their environments. Other than with the Western idea to manage global biodiversity, data simplifications do not align with the holistic nature of Indigenous knowledge systems 57 . Overall, our findings suggest that the results of spatial characterizations are clearly shaped by the underlying datasets and their inherent spatial and taxonomic biases. These biases relate to likely mismatches between the scopes of the studies contributing to the datasets and that of the spatial characterization. As all datasets combine different sources, they all involve mixed geographic and taxonomic scopes, regardless of whether their overall scope matches that of the characterization. Additionally, the utilized datasets do not reflect the contributions of Indigenous Peoples, complicating accurate interpretation of the spatial characterization results, and weakening the visibility of Indigenous Peoples’ contributions. We recommend that future studies seeking to characterize spatial patterns of species richness use data that are well aligned with the specific objectives of the study, particularly in terms of geographic scope and taxonomic coverage. It is also paramount to account carefully for potential sources of bias in data collection and interpretation. In studies focused on biodiversity within IPL, we underscore the importance of Indigenous participation, not only as a matter of empowerment and ethical engagement, but also to enhance the accuracy, relevance and balance of biodiversity data. Methodology Datasets used For our analyses, we retrieved biodiversity data from three different databases: the IUCN Red List of Threatened Species, the database of the Global Biodiversity Information Facility (GBIF), and the database of the Group on Earth Observations Biodiversity Observation Network (BBON). We used the former two datasets because they have been used in spatial characterizations found in the literature (see Ref. 10–15 ), and the database by BBON to include one dataset that corresponds to the scale of our study area. For the Red List, we retrieved species distribution maps as vector data from Version 6.3 of the IUCN Red List 16 which characterize known and possible present and historical tree species ranges. For GBIF, we retrieved all the occurrence data for Bolivia from the website 58 as a cvs-file. For BBON, the complete occurrence dataset originating from Fernándes et al. 32 was provided as vector data by M.F.. In addition to these datasets, we used the ecosystem classification of Bolivia by Ibisch et al. 31 as a reference throughout the paper, to compare the observed spatial and taxonomic patterns with expected distributions. To map the different area types, we used vector data indicating the boundaries of Bolivia, PA, and IPL. The Bolivian boundaries were obtained from Natural Data ( naturaldata.com ), the PA boundaries from the World Database on Protected Areas (WDPA) 59 , and the IPL boundaries from the National Agrarian Reform Institute (INRA, after the name in Spanish) 38 . To map Bolivia’s ecoregions, we used the global vector dataset of Dinerstein et al. 60 . Data preparation All preparations and analyses were carried out in Python 61 . A scheme representing the structure of the codes can be found in Figure S1. For GBIF and BBON, we removed all entries lacking either species names or geographic coordinates. We used each latitude and longitude pair provided in the csv -file by GBIF to convert it into vector data. To analyze tree species only, we filtered the Red List, the GBIF, and the BBON datasets using the tree species list provided by Keppel et al. 21 . This list is based on the GlobalTreeSearch 62 and the GIFT 63 databases and is the most complete tree list available, including the scientific names of 58,044 globally known tree species. It is important to note that the list defines trees not based on taxonomy but based on the definition by the IUCN’s Global Tree Specialist Group. This definition includes all woody plants that grow at least two meters tall and have at least one stem with a diameter of five centimeters or more at breast height. Different from Seebens et al. 15 , we did not implement measures to account for spatial biases in the GBIF dataset. We adopt this approach to present the unmodified data distribution. Also, such measures can be contentious, particularly given that important sources of bias, such as systemic inequities or human preference, are challenging to quantify and adjust for 23 . For all five datasets, we extracted entries located within Bolivia using the national boundary of Bolivia. All datasets were projected to WGS 84 / UTM zone 20S to enable spatial analyses in meters. For the maps, we reprojected them to the WGS 84 dataset for more comprehensive axis descriptions. Spatial characterizations The IPL and PA area types were defined by the datasets adopted. Dual IPL-PA status areas are those where IPL and PA intersect. All other areas were defined as OA. The total areas of Bolivia and each area type were identified through the respective geometries. The relative areas were then calculated by dividing the area type sizes by that of Bolivia. To characterize the expectable tree species richness across Bolivia’s ecoregions, we mapped tree species richness estimations by Ibisch et al. 31 , using their vector data. Ibisch et al. 31 estimated the trees species richness for most of the ecoregions they characterized, either as ranges or as thresholds (e.g., <50, 250–450, or >1000). We used these estimations to group 18 of the 23 (sub-) ecoregions into three equally sized categories of relative species richness: ‘low’, ‘medium’, and ‘high’. Five subregions were grouped as ‘no info’ because Ibisch et al. 31 did not estimate their tree species (see Table S2). For the species distribution maps, we created one raster layer per dataset indicating the number of species in each pixel of a 10 km² grid. To do this, we created a separate raster layer for each species, where each grid cell overlapping with the species’ vector data was assigned a value of 1 and each other a value of 0. For the Red List, we directly used the species range entries for this, as each of them represents one species. For GBIF and BBON, we grouped the occurrences into separate sets (tuples), with each set containing all the records for a single species. Finally, we added up the raster layers of the individual species to generate a single layer per dataset. To determine how many species are present in the different area types, we counted the number of species, represented by species range polygons (for the Red List) or species occurrence tuples (for GBIF and BBON), that intersected with the area types’ polygons. As mask layers, we used the IPL and PA boundaries, prepared as previously described. For IPL-PA, we created the mask layer by intersecting IPL and PA. For OA we first combined IPL and PA and then defined the mask layer as all regions of Bolivia that are outside this combination. To determine the relative species ranges ( range rel , [ha/species/km 2 ]) based on the Red List in the different area types, we summed the species range areas ( range i , [m²]) of all species i within the defined mask layers. Then, we divided this total species range by the respective mask layer’s area ( A , [m²]) and the total number of species ( n ), and multiplied everything by 10 2 to convert the unit to hectare per species and square kilometer (see Equation 1). To calculate the species density ( ρ sp , [number / 1000 km 2 ]) based on GBIF and BBON, defined as the number of distinct species per thousand square kilometers, we extracted all occurrence points within each of the mask layers. Then, we counted the number of distinct species ( n sp [number]) present in each area. Finally, we divided n sp by A to obtain ρ sp and multiplied the result by 10 9 to convert the unit to number per 1000 km 2 , as shown in Equation 2. Spatial distribution To compare the spatial distribution of the species richness based on the different datasets, we mapped it across the country’s different ecoregions. Therefore, we measure how many species ranges of the Red List and how many occurrences of distinct species from GBIF and BBON are within each ecoregion of Bolivia. For the Red List, we counted the intersections between each ecoregion and all species range polygons. For GBIF and BBON, we extracted all occurrence points located within each ecoregion and then counted the number of distinct species. To show the clustering in the GBIF and the BBON datasets, we created a raster layer for each of them representing the spatial distribution of the observation numbers. As these raster layers are rectangular, empty cells could either be cells outside Bolivia or in areas within Bolivia with no observations. To account for this, we identified the number of grid cells in the raster layer corresponding to the shape of Bolivia by masking it to the country’s borders. The minimum number of cells having half of the observations was found iteratively based on a sorted list of cells derived from these raster layers. Taxonomic distribution To evaluate the taxonomic distribution of species in the datasets, we organized the species by their families. For GBIF and BBON, we first grouped all entries by species and then by family. For the Red List, we grouped entries directly by family, as each entry represented a distinct species. To measure the average difference in family distribution between (i) the Red List and GBIF, (ii) the Red List and BBON, and (iii) BBON and GBIF, we first identified the families shared by each pair of datasets. We then filtered each dataset to include those shared families only and sorted them alphabetically to ensure matching order. Finally, we calculated the ratio of the higher to the lower percentage for each family and identified the mean. Identifying the sources To gather information about the data sources, we used the metadata provided by each dataset. For the Red List, we used the source, citation, and compiler name, and for GBIF, we used the institution and collection. Red List entries were grouped by source, and similar descriptions were merged when they appeared to originate from the same institutions or studies. Based on this grouping, we researched the associated sources and used the expert knowledge of M.M.R. to determine whether they were affiliated with Bolivian institutions. For the GBIF dataset, we automated the process to count the number of entries per institution, per collection, and per collection-institution-pair. For the visualization in Figure 4, we applied a threshold of 10,000 entries and grouped institutions contributing fewer as “other”. As with the Red List, we used all available metadata to identify the origin of institutions or studies and determined whether they were Bolivian. Declarations Data Availability All datasets used in this study, except for those of the Group on Earth Observations Biodiversity Observation Network 32 and the spatial data by Ibisch et al. 31 , are publicly available and can be obtained from the referenced sources. We provide an explanation in the Supplementary Information on how to integrate these datasets into our code structure to reproduce the assessments. All results discussed in the article are shared either in the main manuscript or in the Supplementary Information, in some cases with thresholds applied to ensure clarity. They can be fully reproduced using our code. Code Availability All code used in this study is available on Zenodo: https://doi.org/10.5281/zenodo.18467099. Acknowledgements We thank Jocelyne Sze and Jordi Cristobal Rosello for inspiring discussions in earlier stages of this work. This study has received financial support from an ERC Starting Grant from the European Union (ERC, IEKCHANGES, 101117423) to Á.F.-L. Views and opinions expressed are those of the authors only and do not necessarily reflect those of the European Union and the European Research Council Executive Agency. Neither the European Union nor the grant authority can be held responsible for them. Á.F.-L. was further supported by a Ramón y Cajal research grant from the Spanish Ministry of Science and Innovation (RYC2021-034198-I). This work contributes to ICTA-UAB ‘María de Maeztu’ Programme for Units of Excellence of the Spanish Ministry of Science and Innovation (CEX2024-001506-M/funded by MICIU/AEI/ 10.13039/501100011033). Author Contributions S.K conceptualized the study, curated the data, developed the methodology and software, conducted the formal analysis and investigation, and wrote the original draft. J.C.L. contributed to reviewing and editing the manuscript.  M.F. provided data and contributed to reviewing and editing the manuscript. M.M.R. contributed to reviewing and editing the manuscript. Á.F.-L. conceptualized the study and contributed to reviewing and editing the manuscript. Competing Interests All authors declare no financial or non-financial competing interests. References Jaureguiberry, P. et al. The direct drivers of recent global anthropogenic biodiversity loss. Sci. Adv. 8 , eabm9982 (2022). Cardinale, B. J. et al. Biodiversity loss and its impact on humanity. Nature 486 , 59–67 (2012). Díaz, S. et al. Pervasive human-driven decline of life on Earth points to the need for transformative change. Science 366 , eaax3100 (2019). Arora, N. K. & Mishra, I. Life on land: progress, outcomes and future directions to achieve the targets of SDG 15. Environ. Sustain. 7 , 369–375 (2024). Secretariat of the Convention on Biological Diversity. Global Biodiversity Outlook 5 . https://www.cbd.int/gbo/gbo5/publication/gbo-5-en.pdf (2020). Gadgil, M., Berkes, F. & Folke, C. Indigenous Knowledge for Biodiversity Conservation. Ambio 22 , 151–156 (1993). Maffi, L. LINGUISTIC, CULTURAL, AND BIOLOGICAL DIVERSITY. Annu. Rev. Anthropol. 34 , 599–617 (2005). IPBES. Global Assessment Report on Biodiversity and Ecosystem Services of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services . https://zenodo.org/doi/10.5281/zenodo.3831673 (2019) doi:10.5281/ZENODO.3831673. Fernández-Llamazares, Á. et al. Scientists’ Warning to Humanity on Threats to Indigenous and Local Knowledge Systems. J. Ethnobiol. 41 , 144–169 (2021). O’Bryan, C. J. et al. The importance of Indigenous Peoples’ lands for the conservation of terrestrial mammals. Conserv. Biol. 35 , 1002–1008 (2021). Fernández-Llamazares, Á. et al. The importance of Indigenous Territories for conserving bat diversity across the Amazon biome. Perspect. Ecol. Conserv. 19 , 10–20 (2021). Sze, J. S. et al. Indigenous Peoples’ Lands are critical for safeguarding vertebrate diversity across the tropics. Glob. Change Biol. 30 , e16981 (2024). Estrada, A. et al. Global importance of Indigenous Peoples, their lands, and knowledge systems for saving the world’s primates from extinction. Sci. Adv. 8 , eabn2927 (2022). Schuster, R., Germain, R. R., Bennett, J. R., Reo, N. J. & Arcese, P. Vertebrate biodiversity on indigenous-managed lands in Australia, Brazil, and Canada equals that in protected areas. Environ. Sci. Policy 101 , 1–6 (2019). Seebens, H. et al. Biological invasions on Indigenous peoples’ lands. Nat. Sustain. 7 , 737–746 (2024). IUCN. The IUCN Red List of Threatened Species. (2022). GBIF Secretariat. What is GBIF? https://www.gbif.org/what-is-gbif. Devictor, V. & Bensaude-Vincent, B. From ecological records to big data: the invention of global biodiversity. Hist. Philos. Life Sci. 38 , (2016). Buckland, S. T. & Johnston, A. Monitoring the biodiversity of regions: Key principles and possible pitfalls. Biol. Conserv. 214 , 23–34 (2017). Guidelines for the Application of IUCN Red List of Ecosystems Categories and Criteria: Version 2.0 . (IUCN, International Union for Conservation of Nature and Natural Resources, Gland, Switzerland, 2024). Keppel, G. et al. Synthesizing tree biodiversity data to understand global patterns and processes of vegetation. J. Veg. Sci. 32 , e13021 (2021). Mugal, M. A. et al. Bridging global knowledge gaps in biodiversity databases: a comprehensive data synthesis on tree diversity of India. Biodivers. Conserv. 32 , 3089–3107 (2023). Chapman, M. et al. Biodiversity monitoring for a just planetary future. Science 383 , 34–36 (2024). Costello, M. J., May, R. M. & Stork, N. E. Can We Name Earth’s Species Before They Go Extinct? Science 339 , 413–416 (2013). Hortal, J. et al. Seven Shortfalls that Beset Large-Scale Knowledge of Biodiversity. Annu. Rev. Ecol. Evol. Syst. 46 , 523–549 (2015). Lomolino, M. V. & others. Conservation biogeography. Front. Biogeogr. New Dir. Geogr. Nat. 293 , (2004). Ibisch, P. L. Bolivia is a megadiversity country and a developing country. in Biodiversity: A Challenge for Development Research and Policy (eds Barthlott, W., Winiger, M. & Biedinger, N.) 213–241 (Springer, Berlin, Heidelberg, 2001). doi:10.1007/978-3-662-06071-1_15. Ibisch, P. L. Biodiversity Conservation in Bolivia: History, Trends and Challenges. in Environmental Issues in Latin America and the Caribbean (eds Romero, A. & West, S. E.) 55–71 (Springer Netherlands, Dordrecht, 2005). doi:10.1007/1-4020-3774-0_3. Moraes, R. M. & Sarmiento, J. Biodiversity in Bolivia. in Global Biodiversity (Apple Academic Press, 2018). Gigler, B.-S. Poverty, Inequality and Human Development of Indigenous Peoples in Bolivia . (2009). Ibisch, P., Beck, S., Gerkmann, B. & Carretero, A. Ecoregiones y ecosistemas. in Biodiversidad: La Riqueza de Bolivia 47–88 (2003). Fernández, M. et al. Challenges and opportunities for the Bolivian Biodiversity Observation Network. Biodiversity 16 , 86–98 (2015). IUCN SSC Red List Technical Working Group. Mapping Standards and Data Quality for the IUCN Red List Spatial Data . https://www.iucnredlist.org/resources/mappingstandards (2024). Chambers, J. Q., Higuchi, N. & Schimel, J. P. Ancient trees in Amazonia. Nature 391 , 135–136 (1998). Guèze, M. et al. Are Ecologically Important Tree Species the Most Useful? A Case Study from Indigenous People in the Bolivian Amazon. Econ. Bot. 68 , 1–15 (2014). Moraes R., M. Flora de palmeras de Bolivia . (Herbario Nacional de Bolivia, Instituto de Ecología, Universidad Mayor de San Andrés, La Paz, Bolivia, 2020). DeWalt, S. J., Bourdy, G., ChÁvez De Michel, L. R. & Quenevo, C. Ethnobotany of the Tacana: Quantitative inventories of two permanent plots of Northwestern Bolivia. Econ. Bot. 53 , 237–260 (1999). INRA. Tierras Comunitarias de Origen y Territorio Indígena Originario Campesino a nivel Nacional - Gestión 2018. Instituto Nacional de Reforma Agraria (2018). Swenson, J. J., Young, B. E., Beck, S., Comer, P. & Et., A. Plant and animal endemism in the eastern Andean slope: challenges to conservation. BMC Ecol. https://doi.org/10.1186/1472-6785-12-1 (2012) doi:10.1186/1472-6785-12-1. Moraes, M. R., Ríos-Uzeda, B., Moreno, L. R., Huanca-Huarachi, G. & Larrea-Alcázar, D. Using Potential Distribution Models for Patterns of Species Richness, Endemism, and Phytogeography of Palm Species in Bolivia. Trop. Conserv. Sci. 7 , 45–60 (2014). Beck, J., Böller, M., Erhardt, A. & Schwanghart, W. Spatial bias in the GBIF database and its effect on modeling species’ geographic distributions. Ecol. Inform. 19 , 10–15 (2014). Moraes R., M., Zenteno-Ruiz, F. & Fuentes, A. Árboles de Bolivia: Actualización e implicaciones del conocimiento. 13 , 1–90 (2017). Goettsch, B. et al. High proportion of cactus species threatened with extinction. Nat. Plants 1 , 1–7 (2015). Cámara-Leret, R., Tuomisto, H., Ruokolainen, K., Balslev, H. & Munch Kristiansen, S. Modelling responses of western Amazonian palms to soil nutrients. J. Ecol. 105 , 367–381 (2017). Das, S. & Lowe, M. Nature Read in Black and White: Decolonial Approaches to Interpreting Natural History Collections. J. Nat. Sci. Collect. 6 , 4–14 (2018). Faxon, H. & Chapman, M. Beyond spatial bias: Understanding the colonial legacies and contemporary social forces shaping biodiversity data. Environ. Res. Lett. https://doi.org/10.1088/1748-9326/add6b6 (2025) doi:10.1088/1748-9326/add6b6. Park, D. S. et al. The colonial legacy of herbaria. Nat. Hum. Behav. 7 , 1059–1068 (2023). Tomasini, S. Unpacking the Red List: Use (and Misuse?) of Expertise, Knowledge, and Power. Conserv. Soc. 16 , 505 (2018). Leung, B., Hudgins, E. J., Potapova, A. & Ruiz-Jaen, M. C. A new baseline for countrywide α-diversity and species distributions: illustration using >6,000 plant species in Panama. Ecol. Appl. 29 , e01866 (2019). dos Santos, J. G., Malhado, A. C. M., Ladle, R. J., Correia, R. A. & Costa, M. H. Geographic trends and information deficits in Amazonian conservation research. Biodivers. Conserv. 24 , 2853–2863 (2015). Laureto, L. M. O. & Cianciaruso, M. V. Palm economic and traditional uses, evolutionary history and the IUCN Red List. Biodivers. Conserv. 26 , 1587–1600 (2017). Possingham, H. P. et al. Limits to the use of threatened species lists. Trends Ecol. Evol. 17 , 503–507 (2002). Bebbington, D. H. Extraction, inequality and indigenous peoples: Insights from Bolivia. Environ. Sci. Policy 33 , 438–446 (2013). Niesten, E., Zurita, P. & Banks, S. Conservation agreements as a tool to generate direct incentives for biodiversity conservation. Biodiversity 11 , 5–8 (2010). Brush, S. B. Bioprospecting the Public Domain. Cult. Anthropol. 14 , 535–555 (1999). Fernández-Llamazares, Á. et al. No basis for claim that 80% of biodiversity is found in Indigenous territories. Nature 633 , 32–35 (2024). Tengö, M., Brondizio, E. S., Elmqvist, T., Malmer, P. & Spierenburg, M. Connecting Diverse Knowledge Systems for Enhanced Ecosystem Governance: The Multiple Evidence Base Approach. AMBIO 43 , 579–591 (2014). GBIF. GBIF Occurrence Download. (2025). UNEP-WCMC and IUCN. Protected Planet: The World Database on Protected Areas (WDPA) [Online]. UNEP-WCMC and IUCN (2025). Dinerstein, E. et al. An Ecoregion-Based Approach to Protecting Half the Terrestrial Realm. BioScience 67 , 534–545 (2017). Van Rossum, G. & Fred L. Drake Jr. Python 3 Reference Manual. Python Software Foundation (2009). Beech, E., Rivers ,M., Oldfield ,S. & and Smith, P. P. GlobalTreeSearch: The first complete global database of tree species and country distributions. J. Sustain. For. 36 , 454–489 (2017). Weigelt, P., König, C. & Kreft, H. GIFT – A Global Inventory of Floras and Traits for macroecology and biogeography. J. Biogeogr. 47 , 16–43 (2020). Additional Declarations No competing interests reported. Supplementary Files KockEtAlSupplementaryMaterial.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 02 Jun, 2026 Reviews received at journal 01 Jun, 2026 Reviewers agreed at journal 30 May, 2026 Reviews received at journal 27 Mar, 2026 Reviewers agreed at journal 06 Mar, 2026 Reviewers agreed at journal 16 Feb, 2026 Reviewers invited by journal 13 Feb, 2026 Editor assigned by journal 11 Feb, 2026 Submission checks completed at journal 09 Feb, 2026 First submitted to journal 04 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information

Related documents

• Replication Data for "Strategies for Dry Forest Recovery: Fifteen years of Caribbean experimentation across scales "#205092
• Historical tree phenology data across contrasting sites in the Congo Basin#280290
• Introduction to Tribal Forestry#622006
• How much information is there for inferring species trees?#151325
• How much information is there for inferring species trees?#148840