A vision for participatory models of animal movement: a case study with Moose

A vision for participatory models of animal movement: a case study with Moose⋆ Jugal Patel1,3and Niiyokamigaabaw Deondre Smiles2,3 1Department of Geography, McGill University, Montreal, Quebec [email protected] 2Institute for Resources, Environment, and Sustainability, University of British Columbia, Vancouver, British Columbia [email protected] 3Geographic Indigenous Futures Collaboratory Abstract. We envision an indigenous geographic approach to co-modeling animal movement in tightly coupled socioecological systems. This paper outlines our vision for a movement-place model of animal movement. Applicable to coupled systems, this mixed-methods approach prioritizes situated knowledge and data sovereignty. Our vision requires three movement-focused modelling components: movement-trace, movementspace, and movement-place. Together with community members, we encourage iteration through each component, while making distinctions for data sovereignty along the way. We outline a GeoAI approach, which addresses concerns related to data availability and data sovereignty, often expressed by people living within study systems. First, this method operates on the presumption that local animal trajectory data is unavailable. Lack of animal trajectory data is common across socioecological systems, and in some cases collecting new data is discouraged by people living within coupled study systems; posing an interesting geospatial data challenge. Second, the approach prioritizes data sovereignty of community-held situated knowledge by outlining important distinctions between movement-space and movement-place modelling components. That is, whether a model incorporates situated knowledge or not. In our work, traces are modelled using moose (Alces alces) movement data from three source sites; then traces are related to environmental spaces; which together form the basis upon which we initiate movement-place co-modelling with local stakeholders, bands, and tribes across Minnesota and northern Ontario. By involving people embedded within study systems, our vision of a movement place model can be realized by co-developing a spatiallyexplicit and locally enriched agent-based model of a coupled socioecological system. Movement place models - models which are co-developed to integrate local situated knowledge into data-driven representations of movement - aim to enable or otherwise enhance socioecological resilience by improving how local policy decisions are made (i.e., either with or without people living within said systems; either with or without important socioecological relations and interactions being considered). ⋆DOI: https://doi.org/10.5281/zenodo.17686018 2 J. Patel et al. Keywords: GIScience ·Indigenous Geography ·agent-based modelling ·animal movement 1 Vision We envision a movement-place model of animal movement for tightly coupled socioecological systems, which incorporates and extends notions of movement traces and spaces found in computational movement analysis literature. Here, trajectories are understood as movement traces occurring in a given environmental space [1] [2] [3]. Extending this, we introduce movement-place as an additional aspect focused on explicitly considering situated knowledge of people embedded within coupled study systems. In this vision paper, we describe how we are approaching participatory agentbased modeling of moose (Alces alces) ecology across Minnesota and northern Ontario. By combining computational movement analysis with participatory mapping and software engineering principles, the method offers a means for communities to develop policy, scientific geospatial visualizations, and insight as they see fit. The envisioned three component method – trace, space, and place – is iterative, and intended to enhance local decision-making related to animal movement. To emphasize data sovereignty, movement trace and space aspects of modelling animal movement are distinguished from movement-place, in part, by privacy, consent, and community-determined objectives. The remainder of this section is organized along these three movement modelling components: movement traces, spaces, and places; and is followed by brief background on related concepts before concluding with a summary of the envisioned movement-place model and approach to co-modelling animal movement in coupled socioecological systems. 1.1 Movement traces Movement traces, in this case study, are representations of moose movement across three source sites: 1. the Alberta-BC border north of Prince Rupert [4]; 2. near Fort McMurray [5]; and 3. the Upper Koyukuk River in Alaska [6]. These data are available via MoveBank - an open access repository for animal movement data, and we use these to develop a general moose movement model [7]. To start, we process trajectories using MovingPandas [8]. We then fit Continuous Time Correlated Random Walks (CTCRW) for each individual [9]. CTCRWs offer a distinct advantage: the capacity to model uniformly sampled trajectories, using non-uniformly sampled input trajectories [9]. Generated CTCRWs form the movement-trace component of modelling moose ecology in coupled systems across Minnesota and northern Ontario. Simultaneously, a preliminary spatial agent-based model is prepared with ’moose-agents’ that use CTCRW to model agent movement in simulated spaces. For the purpose of our vision, it is not strictly necessary to use CTCRWs to model movement traces. Any workflow which results in type II regular trajecto- A vision for participatory models of animal movement 3 ries is encouraged to maintain simplicity in communication, collaboration, and downstream co-development of spatial agent-based models. 1.2 Movement spaces To demonstrate development of a movement space model, we relate trace objects to common environmental features. We conduct informal interviews and initiate conversation with biologists and moose specialists to help decipher regional environmental variables or other spatial features thought to be affecting moose behaviour. Thus far we have collected road and trail network data from OpenStreetMap, and indices for anthropogenic heat flux (1KM resolution [10]), precipitation, and tree cover processed and availed by Google Earth Engine. Additional relevant environmental features may be identified by the community at a later date. This information is then overlaid with movement traces to extract values along movement traces. Using these environmentally-enriched movement traces, we train a random forest classifier on path segmentation labels. We apply a simple threshold of >2m/s to define moving behaviours. We aggregate outputs from the random forest classifier, developing a movement selection surface [11]. The moose agent based model is adapted to incorporate our movement-selection surface, selecting points most likely to suit movement behaviours. A variety of methods can be deployed to develop behaviour selection surfaces. These are simply outputs of resource, path, or behaviour selection methods [12], [13]. However, the method must be compatible (i.e., able to integrate) with downstream participatory mapping outputs. These may include digitizations of drawn paths, rough polygons with uncertain placement in space and time, emotional assessments of places, or a wide variety of not commonly modeled spatial features [14] [15]. These require explicit integration when co-developing the movementplace model. Although integration in this envisioned context is ultimately quantitative and computational, agent-based models are unique technologies well positioned to ascribe simulated objects and spaces a variety of mutable attributes, each with potential to be explicitly considered during simulation. Overall, practitioners should prepare their methodological approaches for explicit integration of situated knowledge. Our deployed movement-space method can be adapted to include new spatial features as determined by the community [11] [16]. Given the above, we initiate the place aspect of modelling movement in coupled systems. We present our findings and generalized movement trace and space models to local stakeholders, bands, and tribes across Minnesota and northern Ontario that have potential interest in co-developing a simulation of moose ecology. 1.3 Movement places Our movement trace and movement space modelling components have contributed to a preliminary and baseline spatial agent-based model of moose ecology. The next step, movement-place modelling, is focused on defining agent- 4 J. Patel et al. based modelling objectives and integrating situated knowledge, together with community members. Community members will be encouraged to iterate through movement modeling components (i.e., trace, space, place), identifying relevant features or socioecological practices (e.g., seasonality, hunting, harvest, care etc.,). Changes made (e.g., new features being identified) are integrated into a community held and maintained movement-place moose agent-based model. This community held model is essentially (and literally) a fork of a public ‘movement-space’ model repository plus local changes. For each iteration, during the movement-place modelling step, research practitioners should pose the question: Should these changes be shared publicly? If the answer is no, for a given change, it is exclusively integrated (i.e., pushed) into the community-held movement-place model. If it can be shared, the community may opt to help integrate their changes into a researcher-held movement-space model. In most respects, the above is not enough information to adequately model animal movements in an untrained or unseen target destination (i.e., having disparate source and target sites). However, in the places we envision our approach for (i.e., coupled socioecological systems) people hold situated knowledge of moose relations, behaviour, ecology, and environment [17]. Our approach asks: Can we develop an adequate simulation of a moose socioecological system by combining a generalized movement model (trained on unrelated source sites) with situated knowledge of local participants. To clarify, the generalized movement model is based on the CTCRW described above in ’Movement traces’, and the movement-selection surface described in ’Movement spaces’. While their resulting integration into an agent-based model forms the community-held movementplace model. Fig. 1. Workflow of a movement-place model, following three movement modelling components - trace, space, and place. A vision for participatory models of animal movement 5 1.4 Movement-place Model We combine computational movement analysis with participatory software engineering and participatory mapping methods to co-develop animal movement models with communities embedded within coupled study systems. In this context (i.e., tightly coupled socioecological systems), we believe GeoAI and complex systems approaches may enable a way to: first generalize a moose movement model; then apply, in collaboration with situated knowledge holders and community members, information into a spatial agent-based model. The outputs of such a model are geo-visual, easy-to-understand, and can be powerful participatory models, useful for informing local environmental policy [18] [19]. Our approach is premised upon the understanding that people embedded within socio-ecological complex adaptive systems are better equipped to integrate information regarding their own relations with land, animals, etc., into ecological models [17]. With two outcomes, helping inform more efficacious environmental policy, and developing a means to preserve socio-ecological relations. By definition, the better socio-ecological relations are maintained, the more resilient a coupled system can be to environmental disruptions (e.g., climate change, rapid anthropogenic development, etc.). The adaptive capacity of socio-ecological systems, that is the ability to maintain critical information flows, are particularly relevant to avoid global ecosystem collapse and various related catastrophes [20] [21] [22] [23]. Collecting new animal movement data can be resource intensive. Similarly, communities may be unwilling or otherwise incapable of collecting trajectory data on animals in their environment. Often in tightly coupled socioecological systems, peoples embedded in study systems discourage animal monitoring. These situations provide opportunities for exciting innovations in GIScience and GeoAI to solve such socioecological problems [24]. 1.5 Indigenous Places As researchers who work as part of a collective that has built relationships with Indigenous communities and organizations [16], we emphasize the importance of ensuring that this proposed vision adheres to established best practices of Indigenous data sovereignty and meaningful Indigenous participation in the work being done. Although historically, geospatial work with Indigenous communities has run the risk of being extractive in scope and practice, with no benefit or data returning to communities [25], contemporaneous best practices for participatory work demand at the bare minimum active collaboration and participation by Indigenous communities, and the adherence by researchers to key frameworks of ethical research, such as CARE (collective benefit, authority to control, responsibility, and ethics) and the Principles of OCAP (ownership, control, access, possession) [26] [27] [28] [29]. Our proposed methodology operates under the assumption that all data that is collected with Indigenous communities belong to those communities. Once 6 J. Patel et al. movement-place modelling is initiated, the community has full input and oversight over the work done, and maintains data sovereignty. Additionally, the model we envision in this paper is iterative, meaning we assume there will be conversations with communities about the nature of modelling, and how community inputs and worldviews can be placed at the centre of the work. We view the community’s participation as a framework through which models will change, adapting to local needs [30]. This means that the model itself becomes tied to the community, and multiple movement-place models may originate from a generalized animal movement model. While this affects reproducibility, the lack of a ready stand-alone reproducible workflow upholds Indigenous community-based research data sovereignty; in that community input is a critical part of the model, and without which the model cannot meaningfully simulate local context. 2 Background The remainder of this paper shares background on key background topics: Participatory GIScience and Agent-based Modelling Animal Movement before concluding with a summary of our approach and the envisioned ’movement-place’ model. 2.1 Participatory GIScience Participatory techniques offer an alternative approach to informing policy: collaboratively with communities. By co-developing movement models with local stakeholders or community members, participatory movement models offer a means to disrupt how insight from computational movement analysis is operationalized. Popular statistical techniques for modeling animal movement, resource selection methods, have demonstrated improved accuracy when integrated with traditional or local ecological knowledge [31]. Like selection methods, ABMs are often calibrated using expert opinion. For both techniques people living within study systems can better represent their socio-ecological relations, enabling improved computational models of movement at, as Buchholtz et al., notes, local scales [31]. Ramanath and Gilbert suggest techniques for effective participatory agentbased modelling drawing on software engineering literature [32]. Filatova, Verburg, Parker, and Stannard outline challenges for agent-based modelling socioecological systems; highlighting design, validation, spatial representation, and integration with existing theoretical models as key areas for future contributions [33]. Crooks, Castle, and Batty outline seven challenges for agent-based modelling complex adaptive spatial systems, including validation, defining the purpose of the model, and the extent to which independent theory informs model specifications and parametrization [34]. Our approach considers these emerging methodological trends in agent-based models, and envisions co-developing and co-modelling simulations of socio-ecological systems using locally valid situated A vision for participatory models of animal movement 7 knowledge, using participatory software engineering principles to develop spatial agent-based models of animal movement. 2.2 Agent-based Modelling Animal Movement Agent or individual based models (ABMs) are object-oriented programming techniques used in social and ecological sciences to simulate and link outcomes with policy, behaviour, or other input criteria [35]. They can also be co-designed with communities, re-orienting the purpose and outputs of participatory agentbased models to align with local objectives, values, and knowledge systems [18]. Agent-based approaches to modelling movement are particularly relevant for animals roaming large distances each day, often passing through multiple liminal socioecological spaces [36] [37] and for balancing animal conservation and human livelihood in coupled socioecological systems [38] [39]. The simulated space an animal-agent moves within can be geographically explicit, making spatial ABMs well-suited to visually represent and simulate outcomes related to animal movement [40] [41]. By programming autonomous objects within a model to interact, simulation approaches enable observation and experimentation of how information may behave in dynamic systems [42]. 3 Conclusion We build upon established ideas of movement traces and spaces by introducing movement-place models for tightly coupled socioecological systems. When people embedded in coupled systems are willing, interested in, and able to integrate their situated knowledge into generalized models of animal movement, more efficacious and ethical environmental policy can be developed. This vision paper outlines this approach in our work thus far with bands and Indigenous groups across Minnesota and northern Ontario. In comparison to traditional movement ecology methods (e.g., radio telemetry, GPS collars, etc.), our approach has a number of important advantages: 1. it is less resource intensive, requiring no new local monitoring procedures be set up; 2. data sovereignty remains unambiguously with the local community; and 3. the method explicitly incorporates situated knowledge. There are also disadvantages: 1. generalized models are usually not applicable to new regions or animal groups and therefore will face resistance from advocates of collecting more local animal movement data; and 2. the method is dependent upon traditional animal monitoring practices; if not here, elsewhere to help develop the initial generalized movement model. Without explicit considerations for socioecological relations, we cannot conserve dynamic and complex coupled systems. Movement-place models offer a theoretical approach to incorporating situated knowledge; are critical and timely given environmental disruption and a looming crisis in (socio-)ecological resilience; and are relevant as systems approaches enjoy a renewed emphasis in GIScience for education [43]. 8 J. Patel et al. We hope this vision paper enables further discussion on the topics of Indigenous Geographies, tightly coupled socioecological systems, GIScience, and how they relate to modelling animal movement in systems balancing human and animal interests. Disclosure of Interests. The authors have no competing interests to declare that are relevant to the content of this article. References 1. Gudmundsson, J., Laube, P., Wolle, T.: Computational movement analysis. In: Springer Handbook of Geographic Information, pp. 423–438. Springer, Berlin, Heidelberg (2012) 2. Laube, P.: Computational Movement Analysis. Springer, Cham (2014) 3. Dodge, S., Weibel, R., Ahearn, S.C., Buchin, M., Miller, J.A.: Analysis of movement data. International Journal of Geographical Information Science 30(5), 825–834 (2016) 4. Peters, W., Hebblewhite, M., DeCesare, N., Cagnacci, F., Musiani, M.: Resource separation analysis with moose indicates threats to caribou in human altered landscapes. Ecography 36, 487–498 (2013). https://doi.org/10.1111/j.16000587.2012.07733.x 5. Bohm, H., Neilson, E., de la Mare, C., Boutin, S.: Wildlife habitat effectiveness and connectivity: moose ecology project summary report 2010–2012. Final report, 41 pp. Alberta Environment and Sustainable Resource Development, Edmonton (2014) 6. Joly, K., Craig, T., Sorum, M.S., McMillan, J.S., Spindler, M.A.: Variation in finescale movements of moose in the Upper Koyukuk River Drainage, northcentral Alaska. Alces 51, 97–105 (2015) 7. Kranstauber, B., Cameron, A., Weinzerl, R., Fountain, T., Tilak, S., Wikelski, M., & Kays, R. (2011). The Movebank data model for animal tracking. Environmental Modelling & Software, 26(6), 834-835. 8. Graser, A. (2019). Movingpandas: efficient structures for movement data in python. GIForum, 1, 54-68. 9. Johnson, D.S., London, J.M., Lea, M.A., Durban, J.W.: Continuous-time correlated random walk model for animal telemetry data. Ecology 89(5), 1208–1215 (2008) 10. Jin, K., Wang, F., Chen, D., Liu, H., Ding, W., Shi, S.: A new global gridded anthropogenic heat flux dataset with high spatial resolution and long-term time series. Scientific Data 6(1), 139 (2019) 11. Patel, J., Katan, J., Perez, L., Sengupta, R.: Transferring decision boundaries onto a geographic space: Agent rules extracted from movement data using classification trees. Transactions in GIS 25(3), 1176–1192 (2021) 12. Michelot, T., Blackwell, P. G., & Matthiopoulos, J. (2019). Linking resource selection and step selection models for habitat preferences in animals. Ecology, 100(1), e02452. 13. Wang, Y. S., Blackwell, P. G., Merkle, J. A., & Potts, J. R. (2019). Continuous time resource selection analysis for moving animals. Methods in Ecology and Evolution, 10(10), 1664-1678. 14. Corbett, J. (2009). Good practices in participatory mapping: A review prepared for the International Fund for Agricultural Development (IFAD). 15. Patel, J., & Parent, N. (2024). Participatory Mapping for Public Health. A vision for participatory models of animal movement 9 16. Williams, S. K., Georgeson, J., Gowland, A., Patel, J., Weeks, M., & Smiles, N. D. (2024). 2023 geographic Indigenous futures of the Salish Sea symposium. Western Geography, 9(3). https://wdcag.journals.uvic.ca/index.php/wg/article/view/9/3 17. Haraway, D.: Situated knowledges: The science question in feminism and the privilege of partial perspective. In: Women, Science, and Technology, pp. 455–472. Routledge (2013) 18. Eitzel, M.V., Solera, J., Wilson, K.B., Neves, K., Fisher, A.C., Veski, A., Hove, E.M.: Indigenous climate adaptation sovereignty in a Zimbabwean agro-pastoral system: exploring definitions of sustainability success using a participatory agentbased model. Ecology & Society 25(4) (2020) 19. Castella, J. C., Trung, T. N., & Boissau, S. (2005). Participatory simulation of landuse changes in the northern mountains of Vietnam: the combined use of an agentbased model, a role-playing game, and a geographic information system. Ecology and Society, 10(1). 20. Altizer, S., Nunn, C.L., Lindenfors, P.: Do threatened hosts have fewer parasites? A comparative study in primates. Journal of Animal Ecology 76(2), 304–314 (2007) 21. Bradley, C.A., Altizer, S.: Urbanization and the ecology of wildlife diseases. Trends in Ecology & Evolution 22(2), 95–102 (2007) 22. Jenkins, E.J., Simon, A., Bachand, N., Stephen, C.: Wildlife parasites in a One Health world. Trends in Parasitology 31(5), 174–180 (2015) 23. Semeniuk, C.A., Haider, W., Cooper, A., Rothley, K.D.: A linked model of animal ecology and human behavior for the management of wildlife tourism. Ecological Modelling 221(22), 2699–2713 (2010) 24. Nelson, T., Frazier, A.E., Kedron, P., Dodge, S., Zhao, B., Goodchild, M., Wilson, J.: A research agenda for GIScience in a time of disruptions. International Journal of Geographical Information Science 39(1), 1–24 (2025) 25. Louis, R. P., & Grossman, Z. (2009, March). Discussion paper on research and Indigenous peoples. In Association of American Geographers. 26. Carroll, S. R., Garba, I., Plevel, R., Small-Rodriguez, D., Hiratsuka, V. Y., Hudson, M., & Garrison, N. A. (2022). Using indigenous standards to implement the CARE principles: Setting expectations through tribal research codes. Frontiers in Genetics, 13, 823309. 27. Carroll, S. R., Plevel, R., Jennings, L. L., Garba, I., Sterling, R., Cordova-Marks, F. M., ... & Garrison, N. A. (2022). Extending the CARE Principles from tribal research policies to benefit sharing in genomic research. Frontiers in genetics, 13, 1052620. 28. Jennings, L., Anderson, T., Martinez, A., Sterling, R., Chavez, D. D., Garba, I., ... & Carroll, S. R. (2023). Applying the ‘CARE Principles for Indigenous Data Governance’to ecology and biodiversity research. Nature ecology & evolution, 7(10), 1547-1551. 29. Schnarch, B. (2004). Ownership, control, access, and possession (OCAP) or selfdetermination applied to research: A critical analysis of contemporary First Nations research and some options for First Nations communities. International Journal of Indigenous Health, 1(1), 80-95. 30. Smiles, D. (2023). Decolonial methods. Key methods in geography, 116-32. 31. Buchholtz, E.K., Fitzgerald, L.A., Songhurst, A., McCulloch, G.P., Stronza, A.L.: Experts and elephants: local ecological knowledge predicts landscape use for a species involved in human-wildlife conflict. Ecology & Society 25(4) (2020) 32. Ramanath, A.M., Gilbert, N.: The design of participatory agent-based social simulations. Journal of Artificial Societies and Social Simulation 7(4) (2004)