1.1 Proactive engagement and benefit-sharing in research development and data collection

Researchers must invest time and resources to develop relationships with community members to gain support and permission to access sites and obtain samples. Access should be obtained legally and ethically, such as those outlined by the Nagoya Protocol or following community protocols. Community partners should provide input and be a part of the co-development of project goals throughout the life of the study. It is critical to be transparent about initial project goals and benefits with community partners and the risks and benefits of storing samples away from the community's residence (if applicable). When curating metadata, researchers should redact specific sensitive metadata fields congruent with Dublin Core and Darwin Core (Wieczorek et al., 2012).

1.2 Data governance and storage

Researchers should understand and respect the Indigenous communities' cultural sensitivities, customs and protocols surrounding the governance of the data life cycle. A responsible data management plan should be developed to prioritize long-term sustained community access and perspectives.

1.3 Research communication and dissemination

Researchers should design a research communication and dissemination plan that considers a breadth of appropriate audiences, such as community members and partners, land managers, or researchers. Researchers should consider further project and partnership continuity through funding opportunities if possible and mutually desired.

2.1 Site selection and arthropod sampling

We selected six farm sites along a diversification gradient, which was based on the presence of on-farm crop diversity, size (area in production), elevation, and similar utilization of inputs such as pesticides and fertilizer. Due to our interest in comparing farm arthropod community composition and structure to that in forested areas, two forest sites were sampled. Forest sites were selected based on the degree of disturbance and elevation: primary forests facing degradation from invasive plants and arthropods located between 792 and 914 m in elevation. Therefore, forest sites had a mix of native and introduced vegetation. We also ensured that all sites were at least 800 m apart.

2.2 Sample collection

We collected arthropods using timed vegetation beating (40 seconds) at five points (2 m radius) along a 25-meter transect. Before beat sampling, we collected and sifted leaf litter from a 1 × 1 m plot at each transect point. Arthropods from litter samples were then collected using a Berlese funnel. All arthropod samples were stored in 95% ethanol at −20°C until we conducted DNA extractions.

2.3 DNA extraction

DNA extraction of size-sorted arthropod-plant community samples was performed using the Tissue protocol described in the Qiagen Puregene kit modified for automation (Lim et al., 2022).

2.4 Library preparation and sequence analysis

We used a primer combination (ArF1 - Fol-degen-rev) which targets a 418 bp fragment in the barcode region of the Cytochrome Oxidase I (COI) gene (Lim et al., 2022) in triplicate amplifications using the Qiagen Multiplex PCR kit (Qiagen). The three sample replicates were pooled, and the quality of all pooled PCR products was ensured through bead cleanup and fragment length analysis. Final amplicon libraries consisted of pooled equimolar samples and were sequenced on an Illumina® MiSeq.

Sequences were demultiplexed on Illumina® BaseSpace. PCR primers were trimmed using Cutadapt (Martin, 2011). Sequences were merged, filtered, and denoised to amplicon sequence variants (ASVs) using DADA2 (version 1.14.1.; Callahan et al., 2016; Brandt et al., 2021). ASVs were then clustered to 3% radius (97%) OTUs using DECIPHER (2.14.0; Wright et al., 2012). A curated OTU table was created using LULU (version 0.1.0; Frøslev et al., 2017; Brandt et al., 2021). All remaining OTU sequences were compared to Genbank using ElasticBLAST on Amazon Web Services.

2.5 Native/introduced assignment

To assign a native or introduced status to all OTUs, we utilized NIClassify (https://github.com/tokebe/niclassify), a software tool that implements a machine-learning strategy based on the principles of Andersen et al. (2019). The tool has been used to accurately assign status to several arthropod datasets from Hawai'i (Graham et al., 2022; Kennedy et al., 2022).

2.6 Agricultural diversification index

To create the agricultural diversification index, we used the first principal component of a PCA matrix that included the scaled values of all management attribute variables (coffee cover, crop diversity, non-crop vegetation, canopy layers, litter depth, and distance to natural habitat) (Lu et al., 2022; Armengot et al., 2011). The index allowed us to explore the overall effect of agricultural diversification.

2.7 Statistical analysis

We assessed the alpha-diversity (observed ‘richness’) and beta-diversity (‘composition’ based on Bray–Curtis dissimilarities of Hellinger-transformed community matrices) of native and introduced arthropods in two ways. First, we examined site-level differences between richness and composition to address farmer's interests, including the individual environmental and management attributes that drive these differences and the individual arthropod taxa that contribute to site-level differences. Next, we tested the effect of agricultural diversification (combining all management attributes into a singular index) and distance to natural habitat (‘close’ and ‘far’) to examine the ecological mechanisms that drive the richness and composition of arthropods.

2.8 Site-level differences in arthropod communities

2.9 Agricultural diversification and landscape effects on arthropod communities

2.10 Proactive engagement in research development and data collection

2.11 Data governance and storage

2.11.1 Contextualizing community data

Due to colonial practices and policies, the research enterprise has resulted in most Indigenous data being generated and analyzed away from the origin. Systemic inequities and power imbalances perpetuate unjust disconnections between Indigenous communities, their samples, and data. In Hawai'i, previous and ongoing biopiracy projects, plant patenting, and human genome projects have caused a rightful hesitancy among the ‘ōiwi community concerning the genomic research enterprise (Goodyear-Kaʻōpua et al., 2014). In 2003, in response to the increasing commercialization, commodification, and exploitation of Indigenous resources, such as kalo (taro), ʻōiwi elders and cultural leaders crafted the Paoakalani Declaration (2003). This foundational document outlined ʻōiwi perceptions of traditional knowledge, genetic and biological material stewardship, and a governance framework (Figure 3).

In this project, the arthropods we collected are biological and genetic material from Hawai'i. Therefore, they are protected under the Paoakalani Doctrine. Moreover, culturally, arthropods are kin to ʻōiwi. Several species—both native and non-native—are mentioned in the Kumulipo (Figure 1b), the Hawaiian life origin story. The presence of arthropods in the Kumulipo ties them genealogically to the ʻōiwi community, as the creation of all life (from plants to insects to human beings) is recounted in the epic story and weaved together in succession—just like a phylogeny in the field of genetics. Arthropods are also discussed in moʻolelo and kaʻao (two types of storytelling) (Paglinagwan, 2022). How these arthropods are described varies from revered cultural beings, such as having the designation of ‘aumakua (guardian), to agricultural pests. Therefore, there are many layers of traditional knowledge associated with arthropods, thus imbuing them with kinship.

Many of the arthropods in the Kumulipo were present in our data set. Therefore, sensitivity around the data from these Orders is elevated. However, since these orders represent dozens of families and species of arthropods, further discussion is required to untangle if each species is treated in the same way. Consequently, in the instance of the ant, it is a known introduced arthropod with significant negative impacts on ecosystems. How do you reconcile its presence in the Kumulipo with taxonomic origin and impact? These questions must be addressed by relevant cultural leaders, which we describe further in section 2.11.3 below.

2.11.2 Operationalizing and embedding Indigenous Data Sovereignty

As with the CARE principles, Paoakalani offers generalized, theoretical models for research governance and conduct in partnership with Indigenous communities. However, the research team is responsible for appropriately operationalizing these guidelines, specifically in the context of their research project. To operationalize the wishes of Paoakalani, our research team sought innovative modalities to safeguard IDS across all samples collected and data generated. For this, we utilized The Biocultural (BC) and Traditional Knowledge (TK) Notices developed by Local Contexts, an Indigenous-led organization, that are designed to provide Indigenous context and agency over Indigenous resources (Anderson & Christen, 2019). These Labels and Notices provide an extra-legal system of interest disclosure that creates space for community voices to be heard and address a pitfall in current Intellectual Property regimes that only recognize individual rights. To utilize the Label and Notices disclosure system, the project team created a researcher account through an online hub managed by Local Contexts. The application of these notices can be both visible as an icon in this publication (Figure 4) and as added fields in metadata tables with a specific project identifier (Liggins et al., 2021). However, commonly utilized genomic metadata standards still needed to be developed to include the disclosure of Indigenous rights, interests, and provenance information. Therefore, we added our own fields to the iBOL metadata manifest and filled them according to Mc Cartney, Anderson, et al. (2022); Mc Cartney, et al. (2023) using language from the Local Contexts Hub (Table 2).

3.1 Overview of the data

In total, 222 OTUs were collected among all sites, of which 183 OTUs had introduced taxonomic status while 39 had native status. The orders Araneae, Coleoptera, and Lepidoptera represented many native OTUs (Figure S1). Conversely, most introduced OTUs held an equal proportion with some increase in Araneae and Coleoptera.

3.2 Exploring farmer interests

During conversations with farmers, we encountered two main interests regarding arthropods on their farms. Farmers asked: first, what arthropods are present on my farm? Second, how does this compare to other farms in the region? Some farmers also had specific questions about the Coffee Borer Beetle (‘CBB’; Hypothenemus hampei), which is plaguing their farms (Aristizábal et al., 2016). We could address the first and second questions with the data we generated from this project (Figure 4a–e). However, we did not detect any CBB despite the presence of beetles from the same family (Curculionidae) as CBB in our samples. The lack of CBB is potentially due to seasonality and sampling methodology. CBB tend to be more abundant with the development of cherries and collected by extracting them from cherries or beetle-specific traps (Follett et al., 2016). Therefore, the data could not address any CBB-related questions.

To address the farmer's questions, we analyzed the observed richness of both introduced and native arthropods at the site level (1–8), with forest sites included as a reference baseline. We examined environmental and management attributes known in other studies to alter the composition of arthropods across individual sites (Table 1). All sites varied in environmental and management attributes, which included measurements on on-farm vegetation (e.g. coffee cover, crop diversity, and non-crops), canopy structure, the amount of leaf litter on the ground, and the distance to natural habitat (Figure 4a). The variation among sites captures that agricultural land use is heterogenous, and seemingly similar sites (e.g. a few numbers of crops) can still have different habitat/structural properties (Benton et al., 2003).

A Tukey HSD showed introduced arthropod richness did vary among some sites, with some farm sites significantly more (e.g. site 6 vs. sites 1–4 and 7) or less (e.g. site 4 vs. sites 5, 6 and 8) than other sites (Figure 4b). In contrast, native arthropod richness appears to be more variable between sites. The forest sites had the highest number of native arthropods. However, diversified farm sites 6 and 7 were similar to the forest sites in richness. Curiously, simplified farm site 5 appears to have a subset of native arthropods on all sites. Diversified farm site 8 had lower observed richness than other diversified sites, with richness on par with simplified sites 3 and 4. This could be described by the landscape surrounding site 8 being dominated by monocultures, while the other sites had a more complex landscape mosaic. An additional factor could be the time in management, which we attempted to control by ensuring each farm had been using simplified or diversified management for at least five years. However, site 8 had not been in diversified management for as long as the other diversified sites.

Changes in arthropod richness can be attributed to some on-farm environmental and management properties. Coffee cover (p = .013) and distance to natural habitat (p < .001) significantly altered the richness (Table S1) and composition of introduced arthropods (Figure 4d). Site 3, a simplified farm with the highest proportion of on-farm coffee cover, had an observed introduced arthropod richness comparable to several other sites (Figure 4b). However, the arthropod community sampled from site 3 differed from other sites (Figure 4c). For native arthropods, richness was impacted by several site features, negatively by greater crop diversity and positively with more canopy layers, but proximity to natural habitat had no effect. Interestingly, diversified farm site 8, which had the highest crop diversity, had one of the lowest numbers of native arthropods (Figure 4d); yet, the composition of this arthropod community was distinct from other farm sites due to its high level of crop diversity, especially Polynesian crops (see below in section 3.3 for further discussion; Figure 4e). Therefore, there is a potential for farms, even those located within an area with little natural habitat, to support a unique composition of arthropods through on-farm management practices such as increasing crop diversification and including more Polynesian crops.

A SIMPER analysis identified the native and introduced OTUs contributing the most to the differences between sites (Figure 5a,b). For introduced arthropods, some of the most important OTUs driving community variation included OTU9 (Entomobryidae, 10.2% variation), OTU3 (Amphipoda, 6.89% variation), OTU15 (Brachymyrmex cordmoyi, 4.52% variation) and OTU7 (Entomobryidae, 4.37% variation) (Figure 5a). OTU9 and OTU7 belong to the springtail family, which are detritivores that thrive in soil, and are some of the most abundant introduced OTUs across all sites. Springtails may be particularly abundant in diversified farm sites due to management strategies that promote the build-up of leaf litter and the use of mulch. Curiously, OTU15, an ant known to thrive in the Neotropics, is mainly abundant in simplified farm sites.

In terms of native arthropods, some of the most important OTUs driving community variation include OTU134 (Polydesmida, 11.69%) OTU187 (Tetragnatha, 10.79%), OTU257 (Psocoptera, 8.73%) OTU165 (Psocoptera, 7.95%) and OTU11 (Tetragnathidae, 6.61%) (Figure 5b). A commonality among all of these OTUs is that they have generalist feeding habits. OTU34 is a detritivore in the millipede family and is abundant at diversified farm sites. Again, this may be explained by the soil and leaf litter enhancement strategies on these diversified farms compared to simplified ones. OTU11 and 187 belong to Tetragnathidae, a well-studied family of spiders in Hawai'i that feed on various prey. These OTUs are well represented in farm and forest sites. Pscoptera (OTU257 and OTU165) were abundant across all sites and feed on lichen, fungi, and plant materials on various plant species. We further explore hypotheses on what mechanisms may be behind the retention of certain native taxa in farm sites below (see Section 3.3).

Drivers of site-by-site differences in arthropod richness and composition appear ambiguous, which could be aided by including more farmer participants/sites in future sampling and, thus, increasing statistical power. Yet, the data we collected could be more meaningful to individual farmers if it is not just aggregated to examine patterns that drive the number or composition of arthropods. Therefore, our farmer communication plan involves sharing a flier with individualized information for each farmer, including comprehensive information on the arthropod community detected on their farm. For example, by providing the trophic assignments for nearly all taxa observed on their farm, farmers can use this information to match what they see on the farm with the species list and hone in on pest species they may be encountering. Then, farmers can decide if they wish to engage in forms of Integrative Pest Management or work with the University of Hawai'i Extension or Natural Resource Defence Council to further inquire about the benefits of the species present (i.e. conservation payment programs). After receiving the flier, if a farmer wishes to engage further and learn more, we invite them to attend online or in-person one-on-one or group meetings. A communication plan ensures that project data will make it back to the community meaningfully, which is sometimes in contrast to project goals. This dual approach allows farmers to engage based on their comfort and interest.

3.3 Addressing our ecological question

After addressing farmer interests, the research team sought to understand our main research question to understand how the degree of agricultural diversification alters the diversity and composition of arthropod communities across sites with distinct distances to natural habitats. There is an increasing understanding that the diversification practices within agroecosystems, rather than just the presence of agriculture, play an instrumental role in local and landscape biodiversity patterns (Esquivel et al., 2021). Yet, diversification practices are heterogeneous––as demonstrated in the above-mentioned site-by-site variation in agricultural practices (Figure 4B). Therefore, we created an agricultural diversification index to assess how the culmination of practices (rather than a singular feature) impacts the diversity and composition of arthropods.

We expected a general positive effect of agricultural diversification on both introduced and native arthropods that would be magnified when sites were closer in distance to natural habitat. Yet, we only partially observed these trends. The observed richness of introduced and native arthropods increased with agricultural diversification (Figure 6a,c). Surprisingly, however, when sites were further from natural habitat, native arthropod richness decreased on more diversified sites (Figure 6c). A possible explanation could be that the native species are predated on or in competition with the high amount of introduced species in these systems. This high amount of introduced species is spread across a diversity of orders (Figure S1A) representing various trophic positions, of which orders such as Coleoptera and Araneae, which commonly hold predator trophic positions, are especially abundant.

We expected a combination of diversification practices, such as the number of different crops and non-crops, could create habitat/opportunities for native arthropods (Figure 6e). With the shorter distance between farm sites and natural habitat, there is potential for repeated colonization from the natural habitat to the farm (i.e. a spillover effect), especially on farm sites with high agricultural diversification. As a result, the combination of agricultural diversification and proximity habitat may reduce competition between native and introduced arthropods. Further, more simplified farm sites also had a greater richness of native arthropods than diversified sites at similar distances. One possible explanation is that the more diversified farm sites were dominated by introduced arthropods (Figure 6a; Figure S1A). This suggests agricultural diversification, especially on sites further from natural habitat, may present opportunities for new, introduced species to establish and, consequently, may increase competition with native arthropods. In contrast, more simplified farms may generally have less habitat for introduced arthropods, thus presenting a less competitive environment for native species.

Despite the reduction in native arthropod richness, the further diversified farm sites harbored a unique composition of introduced and native species (Figure 6b,d). The composition of the native species present was heavily represented by mobile arthropods that feed on plant material or detritus (in the orders Lepidoptera, Diptera and Psocoptera; Figure S1B; Figure 6b). In addition, one particular group of arthropods in the families Crambidae and Chloropidae, known to feed on Polynesian crops, was highly abundant on the furthest site dominated by Polynesian crops (e.g. Maiʻa (banana; Musa acuminata) and kō (sugarcane; Saccharum officinarum) (Swezey, O.H., 1954)).

Unsurprisingly, the forest sites had the highest native richness (Figures 4c and 6c ). However, few species were exclusively found in these sites, as there was much taxonomic overlap with farm sites (Figure 6e). The subset of arthropod species only present on these forest sites belongs to orders with narrower ranges because they co-evolved with specific plant taxa such as Hemiptera (Roderick & Metz, 1997; Figure S1B; Figure 5b). The general lack of unique taxa is likely attributed to lower-elevation forest sites being inundated with introduced flora and fauna. Both forest sites had invasive flora that spread fast, including Yellow Ginger (Hedychium flavescens) and Mickey Mouse plant (Ochna serrulata). Taken together with the Polynesian crop results above, there is room for new management paradigms.

Preserving and restoring native flora in these lowland forests is often labor-intensive and expensive due to the invasive traits of many introduced flora, often leading to forest systems remaining degraded. To combat this trend, hybrid approaches of restoration utilizing Polynesian and non-invasive crops alongside native plants as tools have been proposed (Burnett et al., 2006; Ostertag et al., 2020; Winter et al., 2020), especially as a means to connect and prompt the access of ʻōiwi community members to their lands (Hutchins & Feldman, 2021; Lincoln et al., 2018). Still, many in the conservation field believe native arthropods cannot be found on agricultural sites. However, our findings support the potential of hybrid systems utilizing Polynesian and other crop species to support native arthropod biodiversity.