S

Journal of Archaeological Method and Theory https://doi.org/10.1007/s10816-024-09680-7 (2025) 32:2 RESEARCH Modeling Marronage: GIS Heuristics of Refuge Affordances in Colonial St. Croix Lauren E. Kohut1 · Steven A. Wernke2 · Justin Dunnavant3 Accepted: 9 September 2024 © The Author(s) 2024 Abstract On the Caribbean island of St. Croix, archival documents reference settlements of runaway enslaved Africans in the mountainous range known as Maroon Ridge. These settlements provide an important record of Afro-Atlantic resistance to enslavement. However, as both intentionally secluded and ephemeral places of refuge, these maroon settlements are difficult to locate in the archaeological record. Geospatial modeling provides one avenue for understanding African geographies of resistance. Building on prior geospatial modeling efforts, this paper uses a GISbased multicriteria suitability analysis to characterize the shifting affordances of marronage on Danish colonial St. Croix across the second half of the eighteenth century. By considering how the island landscape “looked” to those seeking refuge, we trace how possibilities for refuge were distributed through space and over time. In this paper, we develop affordance heuristics to model refuge using digitized historic maps and publicly available LiDAR data. The resulting model suggests shifting maroon refugia sites over time and demonstrates how geospatial approaches, paired with historical archives, can model historic affordances across time. Keywords Caribbean · Slavery · Suitability analysis · Circuit analysis · LiDAR · GIS · Colonialism * Lauren E. Kohut kohutl@winthrop.edu 1 Department of Chemistry, Physics, Geology, and the Environment, Winthrop University, Rock Hill, SC, USA 2 Department of Anthropology, Vanderbilt University, Nashville, TN, USA 3 Department of Anthropology, University of California, Los Angeles, Los Angeles, CA, USA Vol.:(0123456789) 2 Page 2 of 30 L. E. Kohut et al. Introduction From the earliest recorded instances of enslavement in the Americas, captive Africans resisted subjugation, fleeing captivity and establishing their own autonomous communities in some of the most secluded landscapes. Marronage, or flight for liberation, took many forms from petit acts of truancy to grand acts of physical escape (Debien, 1966). Often existing with a degree of secrecy, the material remains of maroon communities have been difficult for historians and archaeologists to locate. Towards this end, geospatial methods have much to contribute. Due to the secretive nature of marronage, and the resulting biases and silences in historical documentation, archaeology provides a unique and important window into many aspects of maroon life. Early attempts by cultural anthropologists to locate and identify maroon sites began in the 1970s (Price, 1973), but it was not until the 1990s that the Maroon Heritage Research Project (MHRP) established the first archaeological excavations of maroon settlements. These initial investigations focused on Nanny Town in Jamaica (Agorsah, 1993). Subsequent studies have examined maroon communities across the Afro-Atlantic world, including the Great Dismal Swamp in Virginia (Sayers, 2016), Hispaniola (Singleton, 2020), Suriname (White, 2006, 2010), the African-Seminole community at Pilaklikaha (Ibarrola, 2022; Weik, 2009), Palmares in Brazil (Orser & Funari, 2001), Palenques in Cuba (La Rosa Corzo, 2005), Veracruz in Mexico (Weik, 2008), and Fort Mose, in St. Augustine, Florida (Deagan & MacMahon, 1995 and pers. comm. Mary Ibarrola). Outside of the Americas, archaeologists have explored East African maroon communities in Kenya (Marshall, 2018) and Mauritius (Colwell-Chanthaphonh et al., 2014). Over the past decades, the goals of maroon archaeology have been fourfold: to locate maroon sites; identify new research areas; survey maroon settlement boundaries and features; and explore variation in settlement patterns (Weik, 1997). These studies have provided insights into maroon subsistence strategies, settlement and placemaking practices, and relationships to neighboring colonial societies. More recently, archaeological studies have begun exploring both the early development of these communities and the contemporary experiences of their descendants. Working alongside maroon-descended communities, archaeologists are developing research to answer critical questions about the past and contribute to current demands on land-claims and acknowledgement (White, 2006; de Moraes, 2012; Romário Melo De Jesus, 2024). Maroon settlements ranged from ephemeral places of refuge to long-term polities. Places such as Palmares and Accompong, for example, may have begun as refuges, but grew to encompass multiple settlements with their own governing structure (Landers, 2005). Some of these larger and more permanent communities were polities recognized by colonial powers, some even signing treaties with colonial settlers. While maroon settlements are most closely associated with enslaved Africans, maroon communities often included both African and Indigenous ethnic groups across Africa and the Americas (e.g., Weik, 2009; Ibarrola, 2022). Far more is known about large maroon communities than the clandestine refuge settlements that were almost certainly more ubiquitous. Refuge landscapes Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 3 of 30 2 are difficult to trace archaeologically, largely due to the challenges of identifying hidden settlements. To be successful, hideouts and temporary camps need to remain invisible to those in power (Hauser, 2022). Thus, not only are the material traces of refuges generally scant, but refuge settlements are often hidden from view and difficult to access (Kusimba, 2004). Known maroon refuge sites are generally located in difficult terrain, distant from population centers, and the material evidence of these ephemeral settlements is scant and often indistinguishable from other “contact-period” archaeology sites (White, 2010). This paper focuses on these hidden and ephemeral maroon refuge settlements. Maroon and enslaved individuals are typically rendered invisible in colonial maps by omission. Reconstructing geographies of maroon refuges thus requires reading historic maps and archival materials against-the-grain. Scholars working in a range of colonial contexts have demonstrated that careful analysis can uncover Indigenous, African, and maroon settlements obfuscated in colonial maps (Brown, 2020; Byrne, 2016; Harley, 1992; Norton, 2020; Panich et al., 2018). In the Caribbean, analysis of historic maps and other archival materials has been used to map Afro-Atlantic fugitivity. Through analysis of historic maps of slave rebellions on St. Jan in 1733, Holly Norton (2020) concluded that rebels maintained encampments near the plantations where they were enslaved. Similarly, Vincent Brown (2020) mapped the expansion of the 1760 Tacky’s Revolt in Jamaica using archival documents. In doing so, he highlighted the geographic scope of the revolt. On the island of St. Croix, earlier research by Norton and Espenshade (2007) first suggested using geospatial methods to locate maroon settlement. Bo Ejstrud (2008) followed, using GIS to model possible maroon settlements with data from historic maps, digital elevation models, and viewshed analysis. Combining data on elevation and sight lines to plantation structures, he identified several areas on and around Mount Eagle on St. Croix with potential for Maroon settlement (Ejstrud, 2008). Here, we build upon this earlier work, and leverage additional colonial maps, advances in geospatial modeling, and higher resolution topographic data to refine these earlier models. In this paper, we use geospatial modeling, LiDAR data, and historic maps to characterize the suitability of the St. Croix landscape for marronage at distinct historical moments in the island’s Danish colonial occupation. We begin by situating our approach to modeling possibilities for marronage within broader theories of landscape affordances. Next, we provide background on the dynamics of Crucian marronage and colonial mapping, particularly focused on the eighteenth century. We then explain how we simulate subject-oriented suitability for maroon settlement using a fuzzy-logic approach that combines models of landscape permeability, visibility, and walking distances from roads, plantation structures, and fresh water. Colonial roads and settlements were digitized from two historic maps, one from 1750 and one from 1799, to model changing landscapes of maroon affordance across nearly 50 years of Danish colonial occupation. Tracing the changing distribution of risk factors enables a dynamic perspective on dwelling in, perceiving, and moving through a landscape beset by potentially mortal risks for enslaved peoples seeking refuge. The results indicate how and where the spatial field of marronage 2 Page 4 of 30 L. E. Kohut et al. suitability diminished through time and, by extension, suggest an account of why St. Croix maroons ultimately opted to take to the sea in their search for liberation. The approach presented here makes two significant contributions to understanding the spatial dimensions of marronage on St. Croix and, by extension, to discussions of human perception of and navigation in their environments. First, we conceptualize marronage suitability in relation to the landscape affordances for marronage at different moments in the Danish colonial occupation of St. Croix. In doing so, we seek to move beyond the perspective of the environment as a collection of independent variables that people either adapt to directly, or as mediated through specific constructs and institutions. Rather, we seek insights into how dwelling in, perceiving, and moving through a landscape are tied to an agent’s goals and means of attaining them. In this way, our approach emphasizes suitability as relative to the purposeful action of a historically and socially situated actor; and, in turn, emergent properties of landscape affordances. The second contribution is to illustrate how indicators of landscape affordances for marronage change markedly across nearly half a century of Danish colonial control of the island. The result is a plausible, granular approximation of marronage suitability as perceived by enslaved people seeking refuge in a landscape increasingly encroached upon by colonial expansion. Marronage Affordance Heuristics In the preceding discussion, “affordances” play a key role in conceptualizing how those seeking liberation through marronage oriented themselves within the colonized landscapes of St. Croix. We employ heuristics for affordances here to model the subject positions of enslaved people seeking refuge—historical agents who have been otherwise nearly invisible or erased from prevailing historical narratives. We ask: how might traces of their experiences be resuscitated? GIS-based modeling of marronage affordances provides a pathway for both informing archaeological investigation and for counter-mapping colonial cartographic and historiographic representations (Dunnavant et al., 2023). By modeling affordances for marronage, we seek to move beyond traditional GIS-based approaches to the environment that conceptualize a set of independent variables to which humans variously react or adapt (see Kosiba & Bauer, 2012 for extended discussion). We draw from broadly phenomenological approaches that seek to account for how people perceive and interact with their environment, and more specifically from discussions of James Gibson’s affordance concept in the ecological psychology literature and its application in archaeological spatial analysis (Gillings, 2009, 2012; Kohut, 2018; Llobera, 1996, 2012; Wernke et al., 2017). Though Gibson’s own formulations remained imprecise, a core insight he developed in his theory is that perception does not happen from the outside-in, nor that meaning is only produced in the brain of the perceiver: “The affordances of the environment are what it offers the animal, what it provides or furnishes, either for good or ill” (for discussion of the evolution of Gibson’s thinking on the affordance concept, see, e.g., Greeno, 1994; Jones, 2003). This conceptualization was important to Ingold’s influential approach to landscape as a place as apprehended by people who Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 5 of 30 2 dwell in it, and to his related concepts such as taskscapes (Ingold, 1992, 1993). But the ontological status of affordances has remained a fundamental issue of debate. If affordances are neither objective properties of the environment, nor subjective qualia, how might they emerge from both? We adopt a derived, relational approach to affordances articulated by Chemero (2003; see also Gillings, 2009, 2012), which holds that affordances emerge as relations between particular endowments or “abilities” of the perceiving agent and specific situational features of the environment. Thus, certain situational features are salient to a perceiving agent given their abilities. In this formulation, there are (at least) two irreducible sides of an affordance relation: situational features of the environment and the abilities of the perceiving agent. “Abilities” in the case of people necessarily involve knowledge. Abilities are thus in part constituted in practice via bodily hexis and habitus (Dunnavant et al., 2023; Gillings, 2012; Llobera, 1996, 2001, 2012; Wernke et al., 2017). As such, abilities are culturally relative and situationally specific. By extension, affordances cannot be observed archaeologically because they are evanescent in the encounter of a perceiver and situationally salient features in the environment (a point raised by Webster, 1999). Nonetheless, heuristics of affordances are within reach, and geospatial modeling of affordance heuristics can spark insights for thinking through subject-oriented approaches to historically situated landscapes as they relate to differentially positioned subjects. These heuristics include both goal-oriented subjects (in this case, those fleeing captivity and seeking refuge) and what a given (historically specific) landscape offers them (echoing Gibson) for good or ill. Below we describe the historical context of Crucian Marronage before outlining the formal dimensions of affordance heuristics and the geospatial modeling we developed for them. Crucian Marronage For centuries of colonial occupation, the island of St. Croix, known by the Taino as Ay Ay, remained sparsely populated by a multi-national group of European plantation owners and enslaved Africans (Fig. 1). The island, while prime for cash crop agriculture, failed to establish a viable plantation economy as it passed through five different colonial powers including the Dutch, Spanish, English, Knights of Malta, and French. In 1733, the Danes, under the auspices of the Danish West India Company (VGK), bought the island and consolidated it with St. Thomas and St. Jan as the Danish West Indies. When the Danes took over the 218-km2 island, enslaved Africans outnumbered the 550 European residents at a rate of ten to one. The majority of the European inhabitants were British colonists who were encouraged to continue their plantation operations (Hall, 1992). While colonists toiled on the island lowlands, an unknown number of runaway Africans established themselves in the northwestern hills. Maroon Ridge—also known as Maronberg or Maroon Hill—was regarded as the dominant maroon settlement on the island. Encompassing approximately 30 km2 of northwest St. Croix, this mountainous area abuts the Caribbean Sea to the north. The most detailed written accounts of Maroon Ridge come from Moravian 2 Page 6 of 30 L. E. Kohut et al. Fig. 1 The island of St. Croix showing topography, streams, and major colonial cities. Inset shows the location of the island within the eastern Caribbean Sea missionary Christian Oldendorp. Sent by the German government to report on the status of Moravian missions, Oldendorp lived on the island from May 1767 to October 1768 (Pope, 1972:153). His reports describe how maroons built fortifications of poison-tipped wooden stakes and used the landscape’s dense vegetation to enshroud their settlements. According to Oldendorp (1994), they subsisted on soursop fruit (Annona murciata) and rainwater collected in basins and rock crevices. While this description provides a glimpse into what may have been a vibrant maroon community, any colonial account of maroon life must be taken with a degree of skepticism. It is unlikely that Oldendorp had first-hand experience with Maroon settlements and the source of this information is unclear. The timing, duration, and extent of occupation at Maroon Ridge are unknown. Some accounts purport that settlement peaked in the mid-eighteenth century immediately before the Danes purchased the island from the French (see Roopnarine, 2010:102). Others argue Maroon Ridge was a transitional habitation zone where maroons camped temporarily before continuing to Puerto Rico, Vieques, Tortola, and other islands (Hall, 1985). Historical accounts mention maroons living in the mountains as late as 1828, suggesting that it may have served as a sanctuary until slavery was abolished in 1848 (Norton & Espenshade, 2007). During this time, maroons remained under constant threat from periodic raids and the expansion of plantations, which cleared forested areas and mangroves reducing potential maroon sanctuaries (Hall, 1985). As a result, maroons increasingly turned to maritime marronage, finding refuge on neighboring islands (Dawson, 2021; Dunnavant, 2021a, 2021b; Simonsen & Christensen, 2023). Maroon Ridge did not appear on Danish maps of St. Croix until 1799. However, the area was known to plantation owners and colonial administrators, and a 1767 German map (Fig. 2) labels the northwestern hills on the island “Maronberg.” Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 7 of 30 2 Fig. 2 Paul Küssner, “Die Insel Sainte Croix mit den Namen der Plantagen die bestændig sind”, 1767 (source: Royal Danish Library, http://www5.kb.dk/maps/kortsa/2012/jul/kortatlas/object65434/en/) Depictions of the northwestern mountains on these earlier colonial maps are far less precise than other parts of the island, likely due to its inaccessibility (Dunnavant et al., 2023). Despite its omission from colonial maps, the area was known as having a maroon population. Colonial surveyors likely feared an encounter with the maroons and may have avoided or only superficially surveyed the area. Alternatively, surveyors may have intentionally omitted cartographic reference to the maroon community to avoid raising concerns from investors and colonial officials in Copenhagen. Far from impartial spatial registry, land survey and cartography were integral to the colonial project, eliding as much as marking out the distribution of people and resources in the landscape. Nonetheless, as we have argued elsewhere (Dunnavant et al., 2023), through critical cartographic and textual inquiry, colonial maps can also unwittingly yield information for counter-mapping, that is, cartographic “efforts to contest or undermine power relations and asymmetries in relation to cartographic products or processes” (Hazen & Harris, 2006:115). We seek a form of anti-colonial counter-mapping by rendering visualizations of a landscape of changing affordances for marronage on St. Croix through its colonial occupation. Mapping Colonial St. Croix Colonial maps were important tools of Danish colonial governmentality, facilitating the division of land into administrative units, the collection of taxes, and quantification of land for plantation agriculture (Dunnavant et al., 2023). Soon after the Danes purchased St. Croix, the newly appointed Governor Frederick Moth commissioned a survey to identify the best lands for cultivation to ready them for sale and taxation (Hopkins, 1992b). The island was divided into nine administrative quarters and a 2 Page 8 of 30 L. E. Kohut et al. new measurement, the Danish Acre, was established to calculate plantation plots. These plantation plots were valued equally, and the unclaimed plots were to be distributed to stockholders through lottery. Various eighteenth century maps provide detailed information about the colonial landscapes and convey implicit colonial interests. The first official Danish map was produced in 1750 by military officials Johann Cronenberg and Johann von Jaegersberg (Fig. 3). In addition to military outposts, the map detailed earlier plantation structures, roads, waterways, fields and wooded areas on the island. Reefs were also noted and the various bays and inlets were demarcated (Hopkins et al., 2011). The map was not widely circulated and may have been marked classified because it disclosed important naval information (Hopkins, 1992a). A second map published in 1754 by Jens Beck was likely used specifically for census, insurance, and taxation. Unlike the Cronenberg-Jaegersberg map, Beck’s map did not indicate areas of cultivated land and wooded areas, nor the location of plantation structures. The boundaries of plantation plots were clearly marked and the map shows the handwritten names of plantation owners. After the release of Beck’s map, the 1750 Cronenberg-Jaegersberg map disappeared into obscurity (Hopkins et al., 2011:89). In 1794, Peter Oxholm, a Danish military engineer, was sent to map the island’s defenses. Published in 1799 (Fig. 4), it is the most detailed map of the island during the Danish colonial period. The colonial government argued it would be too expensive to publish, so Oxholm self-published the map in Copenhagen (Hopkins et al., 2011). Fig. 3 Cronenberg and Jægersberg, 1750: “Charte over Eilandet St. Croix”, manuscript map No. A/18– 49, Nautical Charts Department Archive, Copenhagen Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 9 of 30 2 Fig. 4 Peter L. Oxholm, “Charte over den Danske Øe St. Croix i America,” 1799 (source: Royal Danish Library, http://www5.kb.dk/maps/kortsa/2012/jul/kortatlas/object65449/en?id=%2Fmaps%2Fkortsa% 2F2012%2Fjul%2Fkortatlas%2Fobject65449) Early cartographers faced significant challenges mapping St. Croix, which resulted in gaps and omissions that remain obfuscated by the totalizing perspective of the island conveyed by these historical maps (Dunnavant et al., 2023). While the 1750 and 1799 cadastral maps clearly delineated plantations, surveying the terrain proved problematic. Surveyors had to account for the various settlers and squatters already on the island, particularly in the East End, who had already made land claims. These claims often conflicted with administrative units and standardized plot boundaries envisioned by colonial administrators. Island conditions also hampered surveying. Dense vegetation made access to many areas difficult, and cultivated lands would quickly revert to thick brush if left untended for even short times (Hopkins, 1992a). Frequent hurricanes, labor shortages, disease, and intense heat hampered and periodically halted survey work (Hopkins, 1992a). Furthermore, the veracity of many colonial maps was challenged by inadequate survey equipment, miscalculation, distortion of the map parchment, and errors that arose during the process of reproduction (Weik, 2019:59). These challenges contributed not only to inaccuracies, but also to omissions in the maps produced—particularly in the more remote parts of the island. In Beck’s 1754 map, for example, the northwest part of the island that eventually became known as Maroon Ridge is listed as “Uoptagne Grunde,” or unrecorded/unoccupied grounds. A striking example of the fallibility of colonial maps comes from the neighboring island of St. Jan, where the first map created in 1719 was likely fabricated, as Dutch cartographer Gerard Van Keulen never actually visited the island (Norton, 2020). Therefore, these maps, while useful references, should not be taken as wholly accurate representations of the physical landscape. Still, they provide a viable archive by which to explore changes in colonial society. We focus on the 1750 Cronenberg-Jaegersberg map and the 1799 Oxholm map to chart the successive appropriation of land to sugarcane cultivation and the expansion of plantation society over two decades after the purchase of the island by the Danish, followed by a half century of Danish colonial rule. Reading these maps in tandem and against-the-grain enables a counter cartography of how the landscape may have 2 Page 10 of 30 L. E. Kohut et al. presented itself in relation to possibilities for liberation of the enslaved laborers of the plantations. A Relational Approach to Refuge Affordances A relational approach to affordances of refuge in the St. Croix landscape requires assembly of a multidimensional model that approximates features and processes salient to those seeking to escape enslavement. How might the landscape “look” to those seeking refuge? What does it offer them (recalling Gibson), for good or ill? How are its affordances for refuge perceived and distributed through space and over time? Conversely, how are perceived risks of detection and capture distributed, and how do they change through time? In addressing these questions, we leverage GISbased modeling to approximate a subject-oriented perspective of dwelling in and moving through a landscape, though not from the perspective of any particular agent in the past. While GIS privileges a cartesian view of environments as a set of layers to be modeled, analyzed, and interpreted from a “god’s eye” analytical vantage (Haraway, 1988), its utility is far from restricted to such a perspective when engaged critically (Pavlovskaya & Martin, 2007). Archaeologists increasingly use GIS to explore subject-oriented, relational perspectives by centering human perception and movement in modeling and analytical efforts (Howey & Brouwer Burg, 2017; Llobera, 2012; Llobera et al., 2011; Lock & Pouncett, 2017). Approaching landscapes this way enables more dynamic perspectives on landscapes as ever-emergent from human-nonhuman interaction in spatio-historical context. Our approach to modeling such contingent landscapes builds on GIS-based suitability analysis to simulate culturally and historically contingent affordances of St. Croix maroon landscapes. Suitability analysis involves decomposing “suitability” into constituent criteria that can be modeled geospatially. These criteria are scaled and sometimes weighted before being combined to form a singular, continuous phenomenal field of greater or lesser suitability—or impedance to such suitability—for a given task. When applied to land use or ecological phenomena, these criteria are typically represented as raster-based themes, where the landscape appears as a surface of pixels, each encoding the relative suitability of the location to the given task or scenario. While suitability analysis has been used widely in archaeology, especially in the field of site predictive modeling, much of this work is behaviorist and adaptationist in orientation (see Wheatley, 2004 for a critical perspective on this extensive literature). What we seek is a heuristic model of the affordances in a landscape relative to a specific goal-orientation and not as a general proposition about fitness for habitation. Overview of Affordance Heuristics and Modeling Our methodology thus starts by specifying marronage affordance heuristics by their constituent criteria and considering which could be modeled as continuous fields via raster themes. In seeking clandestine encampment or settlement, visibility, Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 11 of 30 2 remoteness from colonial settlements and roads, and proximity to water sources (for provisioning) would have been important considerations. Visibility, or rather invisibility, was likely a central concern when locating refuge settlements. The visibility of locations on the landscape is shaped by several factors, including changes in the elevation of the terrain, vegetation cover, time of day, weather conditions, and the visual acuity of the viewer. Our analysis focuses on historically contingent and subject-oriented factors, rather than ephemeral (i.e., time of day and weather) or individual (i.e., visual acuity) conditions. The two most important factors influencing visibility in our model would be terrain and vegetation cover. Of these, modeling visibility in relation to vegetation cover—especially in areas of diverse or unknown landcover—is a challenging issue. Our approach, which uses fuzzy logic (discussed further below) partially accounts for the imprecision and uncertainty of measuring this variable, but we recognize that the question of vegetation cover remains unresolved. Standard GIS-based visibility analysis is best equipped to model visibility relative to topography. This type of bare earth viewshed analysis has been justifiably subjected to critique as an unreliable and even spurious simulacrum of visibility affordances in areas with variably occlusive vegetation cover (see, e.g., Gillings, 2009; Llobera, 2007a, 2007b). Here, we chose to focus on visibility at middle-range distances (4 km) which could capture topographic (in)visibility of larger features of settlement, such as shelter. This approach is similar to Gillings’ (2015) model of invisibility, creating a continuous surface where the value of each location is an index of the number of locations from which it can be viewed, rather than the location’s visibility from a fixed point. Remoteness from colonial loci—such as settlements, windmills, animal mills, and roads—could however be more reliably approximated. We conceive of remoteness along two dimensions. The first measures how difficult it is to traverse locations in a landscape; what we call permeability. The second evaluates distance from points and pathways of potential surveillance. Each measure of remoteness was modeled using different approaches. To model landscape permeability, we used circuit analysis, an approach developed in the field of landscape ecology (McRae et al., 2008, 2016; Pelletier et al., 2014) that has increasingly been used in archaeology to model human mobility (Howey, 2011; Howey & Brouwer Burg, 2017; Kohut, 2018). Our specific approach here is multidirectional, providing a measure of landscape permeability that is agnostic of origin, destination, or direction of travel (Howey & Brouwer Burg, 2017; Kohut, 2018; McRae et al., 2016; Pelletier et al., 2014). This technique, described in further detail in the following section, simulates impedance to movement by injecting a simulated current through the terrain model from the four cardinal directions, producing directional permeability rasters, which in turn are combined to an aggregate permeability raster. Remoteness relative to points and pathways of potential surveillance was modeled using cost distance, or the relative simulated travel cost from colonial loci. Cost distance, as opposed to Euclidean distance, was chosen to better simulate the human experience of traversing a heterogeneous landscape. Cost distance modeling and, particularly, least-cost path analysis, where cost surfaces are used to identify 2 Page 12 of 30 L. E. Kohut et al. the most efficient route between two points, have been critiqued for their emphasis on efficiency and inability to capture local knowledge systems that shape how humans move through the landscape (e.g., Cameron, 2013; Herzog, 2016; Howey, 2011; Llobera, 2000; Supernant, 2017). Our aim, however, is not to model movement paths, but instead to use modeled travel costs as a measure of remoteness. Our analysis considers two primary landscape costs: slope, or the steepness of the terrain, and land cover. St. Croix is a heterogeneous landscape that contains both lowlying plains and rugged mountains. Eighteenth-century maps depict this diversity, including a variety of land covers, such as cultivated fields, urban areas, and dense forest. Both slope and land cover shape the difficulty of traversing the landscape and would have had direct bearing on perceived remoteness from colonial infrastructure. In our analysis, cost surface rasters were used to simulate travel costs from colonial settlements and roads as represented in the maps of 1750 and 1799. Digitized settlements (as point themes) and roads (as line themes) were used as cost surface sources. Slope cost was calculated using a model of energy costs developed by Minetti and colleagues (Minetti et al., 2002). A terrain multiplier was applied to the slope cost using land cover estimates and cost multipliers following Soule and Goldman (1972). Land cover for 1750 and 1799 was approximated from the historic maps and contemporary land cover data (explained in detail below). The resulting cost distance surfaces for roads and settlements were generated using the Distance Accumulation tool in ESRI ArcGIS Pro. This resulted in four cost surface rasters— one relative to roads and one relative to settlements for each map, with higher travel costs reflecting greater remoteness from colonial settlements and roads. Lastly, access to water must have been a significant factor for situating even ephemeral encampments, and certainly for more permanent settlements. Stream locations were derived from a USGS flowline hydrography model, which includes all perennial streams. Modern features such as canals, locks, and diversions were removed or edited to correspond to their courses as represented in the colonial maps. Stream courses were then used as cost surface sources using the same cost distance model as for colonial era settlements and roads. Here, however, cost surface values were modeled inverse to those of settlements and roads: low values are considered more suitable to refuge because they are closer to fresh water. The suitability model thus combines these raster layers: visibility, permeability, cost distance from settlements, cost distance from roads, and cost distance from streams, for each of the colonial maps, using a fuzzy overlay workflow, as described below. Digital Elevation Model Preparation The Digital Elevation Model (DEM) used in this study is derived from a high-resolution LiDAR point cloud from an airborne platform in 2013 (Office for Coastal Management, 2013). Initially, a 1-m (nominal) resolution DEM was interpolated from filtered bare earth returns only. In heavily forested areas, this clearly exceeded the actual ground sample distance of the data. We chose to optimize to the lower resolution forested areas rather than the more open land cover areas, down sampling to a Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 13 of 30 2 10-m resolution. This 10-m DEM forms the basis for our models of remoteness and permeability. Still, the DEM brought with it some issues for modeling the historic landscape of St. Croix. In particular, modern roads along with buildings in more developed parts of the island resulted in variations in elevation that would impact mobility modeling. We were particularly concerned that road cuts in the mountainous northwestern portion of the island would be advantaged in mobility modeling. To smooth the DEM, a modern land cover raster was used to identify areas classified as impervious surfaces. Elevations of impervious surfaces were smoothed using the mean elevation of the surrounding 20-m radius. The resulting raster was then resampled (using bilinear interpolation) to 10-m resolution. Visibility The visibility of each raster cell was calculated from the 10-m DEM. First, 5000 random points were generated in an area that included the island and a 10-km buffer surrounding (Supplementary Information Fig. 1). A buffer was used to account for the potential use of boats to look for evidence of maroon settlements. Visibility was calculated from each random point to each raster cell on the island using a 4000-m radius. Visibility was calculated in 16 directions from each raster cell. The resulting raster provides an index of visibility that ranges from zero to one, where one indicates that the cell is visible from all directions and a zero indicates a cell that is not visible from any direction (Supplementary Information Fig. 2). Landscape Permeability Landscape permeability was modeled through multidirectional connectivity modeling. Typical connectivity models trace connectivity between point locations or resource patches (e.g., Howey, 2011; Rayfield et al., 2011; Urban et al., 2009). Here, a modified approach to connectivity modeling using the open-source program Circuitscape is used to model landscape permeability across the whole island of St. Croix. Circuitscape represents the landscape as a raster where the value of each cell represents the relative cost to traverse it (McRae et al., 2008). Resistance is thus a flexible value that can be adapted to the specific travel considerations of a given species or, in our case, a given historical context. Connectivity is calculated by injecting current into source points where it travels across the resistance raster to one or more ground nodes. The current that passes through each cell is determined by both the resistance of the cell and the availability of other cells. In the resulting current raster, the strength of the current of a cell is interpreted as the likelihood that one would traverse the cell when moving from the source to the ground. Following Kohut (2018), we employed a multidirectional approach to model overall landscape permeability, rather than connectivity between discrete origins or destinations (see also Howey & Brouwer Burg, 2017; Pelletier et al., 2014). The resistance surface considers both the slope and land cover. Four separate resistance rasters were calculated: one for each cardinal direction of travel to account for differences in cost relative to uphill, downhill, and lateral travel. 2 Page 14 of 30 L. E. Kohut et al. Slope was calculated by finding the difference in elevation from each cell to the adjacent cell in the direction of travel (rise) and dividing by the cell size (run). Minetti’s (Minetti et al., 2002) metabolic energy cost of walking function was applied to the slope values, using the following equation, where Cw is the walking cost of the slope gradient (i): Cwi = 280.5i5 − 58.7i4 − 76.8i3 + 51.9i2 + 19.6i + 2.5 The resulting slope walking cost was multiplied by a terrain coefficient to create a travel cost that accounted for both slope and land cover. Land cover was approximated from contemporary land cover maps and the georeferenced historical maps. For both time points, the baseline land cover was assumed to be either wetlands or forested, unless otherwise indicated by the historic map. The island of St. Croix is characterized by tropical dry forests and it is estimated that forest land covered 90% of the island prior to Danish colonization (Atkinson & MarínSpiotta, 2015; Brandeis & Oswalt, 2007). Areas depicted as fields, urban areas, or roads, in the Cronenberg and Oxholm maps, were digitized. Streams tend to be small and there are no navigable rivers on the island, so they were considered neither facilitators nor barriers to movement. For the remaining island areas, wetlands were identified using a 2012 land cover raster for the island (Office for Coastal Management, 2012). All remaining unclassified cells were classified as forested (Supplementary Information Fig. 3). Terrain coefficients were assigned to each terrain type following Soule and Goldman (1972) (Table 1). The slope walking cost was multiplied by the terrain coefficient to generate the walking cost. Resulting values were rescaled from 0.0 to 100 by multiplying the raster by 100 and dividing by the maximum cost to represent the relative resistance of each cell. To account for potential edge effects, a minimum bounding envelope of 1 km was generated and ocean values were assigned the mean resistance. Ocean areas were later clipped from the circuit analysis results. A total of eight resistance rasters were generated: one for each cardinal direction of travel for 1750 and 1799. Source and ground nodes were generated for each raster cell along the edge of the analysis region. The circuit analysis was run four times, once for each cardinal direction of travel, each time using the corresponding resistance raster. The resulting current rasters were normalized and summed to create a single current raster for the island. The resulting raster shows the overall range in permeability across the island. High values indicate areas of minimal resistance that facilitated Table 1 Terrain costs following Soule and Goldman (1972) Terrain Coefficient Agriculture/croplands 1.2 Wetlands 1.8 Urban areas (dirt roads) 1.1 Roads (dirt roads) 1.1 Forest 1.5 Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 15 of 30 2 mobility, while low values reflect greater resistance to permeability (Supplementary Information, Fig. 4). Lower currents reflect a lower likelihood of a passerby to traverse a region, thus indicating greater seclusion. Distance to Streams, Roads, and Settlements Distances to streams, roads, and settlements were calculated separately using the distance accumulation tool in ArcGIS Pro from vectorized features. The source data for streams came from a USGS flowline hydrography model (USGS, 2019). Roads and settlements were traced from georeferenced scans of the Cronenberg-Jaegersberg and Oxholm maps. A cost raster was generated following a similar method to that described above for the resistance rasters. First, the slope tool in ArcGIS Pro was used to create a slope raster. Raster calculator was used to generate the slope walking cost following the Minetti et al. (2002) calculation above. Finally, the slope walking cost was multiplied by the appropriate terrain coefficient raster to produce the cost raster. Separate cost rasters were generated for 1750 and 1799 to reflect changes to the terrain due to the expansion of the plantation system. The results were continuous raster surfaces that reflect minimum cost to the nearest feature for each time point. Suitability Analysis with Fuzzy Logic Fuzzy overlay is an approach to suitability analysis that incorporates fuzzy techniques to address inevitable inaccuracies and biases that arise when criteria variables are defined, classified, and combined. This approach allows multiple, exploratory variables to be combined while avoiding the uncertainties created when overlay factors are binned to discrete values (Bonham-Carter, 2014). Fuzzy techniques are thus particularly well-suited to modeling marronage affordances while accounting for the ambiguities of modeling heuristics for the subjective experiences of runaway enslaved Africans. Suitability analysis in GIS relies on overlay techniques in which human and environmental attributes are represented as layers that are classified in terms of their suitability to a particular objective. These layers can then be overlaid to determine the overall suitability. In weighted overlay approaches, values for each variable are assigned to specific classes that reflect the suitability of that variable to the phenomenon being modeled. These are called crisp sets because a value can only ever be a member of a class or not. Each variable is weighted according to its perceived significance to the phenomenon being measured and the variables are combined using map algebra to create a measure of overall suitability. By contrast, fuzzy logic recognizes that there are uncertainties and imprecision when classifying values (Zadeh, 1978). Historical data presents particular uncertainties when compared to modern GIS datasets (Dragicevic et al., 2001). Historical datasets are drawn from diverse sources, reflecting objectives and motives that are distinct from those of the researcher. Moreover, imprecision in the data can have ramifications for how the results are interpreted. To use an example from the present work, while we can be 2 Page 16 of 30 L. E. Kohut et al. quite confident that maroon settlements would be preferentially located further from plantations, any attempt to classify one distance as highly suited versus somewhat suited would be artificial and potentially inaccurate. Furthermore, we suggest here that fuzzy logic approaches are more appropriate for capturing the culturally and historically contingent affordances of the landscape for marronage settlement. Fuzzy logic approaches are preferable for situations like these where the boundaries between classes are unclear or unknowable, and has been applied to a variety of suitability scenarios, including those in archaeology (e.g., Alexakis & Sarris, 2010; Hatzinikolaou et al., 2003; Jarosław & Hildebrandt-Radke, 2009; Ripy et al., 2014). In fuzzy logic approaches, values are not assigned classes, but instead are defined in terms of the degree to which an item likely belongs to a category: in this case, a suitable area for maroon settlement. Fuzzy values range from 0 to 1, where 0 indicates low possibility for use, and 1 indicates high possibility for use. This is achieved by scaling values along a distribution curve that reflects the rate at which the values change in their possibility of belonging to a set. In GIS-based fuzzy overlay, fuzzy logic is implemented in a two-step process. First, each variable is converted into fuzzy membership values. Second, the resulting variables are combined to provide an overall fuzzy value to represent the possibilities for, in this case, marronage. The fuzzy overlay tool provides several options for combining fuzzy values; each of which provides a different suitability outcome. Fuzzy Membership Following fuzzy logic approaches, fuzzy membership transforms variables into fuzzy sets that reflect the likelihood that the particular variable is suitable to the phenomenon being modeled: maroon refuge. Decisions about how to convert each variable into a fuzzy variable were done iteratively. First, we determined the appropriate distribution of values for each variable. All variables were scaled using either a small function (where smaller index values are increasingly more suitable) or a large function (where larger index values are increasingly more suitable). For each variable, we constructed a series of fuzzy membership rasters with varying parameters. We then graphed the crisp values against the fuzzy values and evaluated the curves for each in relation to the phenomenon under evaluation (Supplementary Information Figs. 5–9) to select the one we felt best represented the phenomenon in question (Table 2). Visibility was the only variable that was consistent across the analyses of both the 1750 and 1799 historic maps. Here, fuzzy membership was defined using the Fuzzy Small function (midpoint = 0.5, spread = 1). Fuzzy membership for landscape permeability was also defined using Fuzzy Small, using the dataset median as the midpoint (1750 = 33.5, 1799 = 36) and a spread of 5. The remaining variables were converted using either the MS Small or the MS Large function. In these functions, fuzzy membership is defined based on the mean and standard deviation of the dataset. Thus, the relative suitability of a variable is scaled based on the overall possibilities within the dataset as a whole. This approach was selected to emphasize how Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 17 of 30 2 Fig. 5 Variable inputs (a–j) and results (k, l) from the fuzzy overlay model for 1750 and 1799. All variables and results are shown in fuzzy membership values, which range from 0 to 1 2 Page 18 of 30 Table 2 Variables for suitability and the method used in fuzzy conversion Variable Units Fuzzy conversion Interpretation Visibility Relative visibility Small Midpoint = 0.5 Spread = 1 The relative visibility of the raster cell at a maximum distance of 4 km. Values range from 0 to 1, with 1 indicating visibility from all directions. Lower visibility represents relative invisibility in the landscape Landscape permeability Current Small Midpoint = median Spread = 5 The relative seclusion of a region compared to the island as a whole. High current values reflect high landscape permeability and thus low seclusion. Low current values represent areas of low landscape permeability and greater seclusion Distance from settlements Cost distance MS Large Mean multiplier = 0.5 SD multiplier = 0.5 Cost distance from settlements. Higher values are further from settlements and thus more suitable for marronage Distance from roads Cost distance MS Large Mean multiplier = 0.5 SD multiplier = 0.5 Cost distance from roads. Higher values are further from roads and thus more suitable for marronage Proximity to streams Cost distance MS Small Mean multiplier = 0.5 SD multiplier = 2 Cost distance to streams. Lower values are nearer to streams and thus more suitable for marronage L. E. Kohut et al. Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 19 of 30 2 possibilities for marronage were understood in relation to a historically contingent landscape. Fuzzy Overlay The resulting fuzzy membership rasters (Fig. 5a–j) were combined using the fuzzy overlay function and the “Fuzzy And” type (Fig. 5k, l) to reflect the minimum fuzzy membership value across all the variables. This allowed us to identify areas that were likely suitable for marronage across all variables and prevented exceptionally high values in one variable from skewing overall suitability. Thus, the areas considered suitable to marronage are only those that meet the minimum threshold across all variables. Here, we chose to only consider values greater than 0.5 as possibly suitable for marronage in order to refine the output (Fig. 6). Results The results of the fuzzy overlay analysis provide plausible simulacra of how those seeking refuge perceived and encountered the changing landscape of St. Croix (Fig. 6a–d.). In our analysis, fuzzy membership values reflect the likelihood that a particular location would have been seen as suitable refuge for self-emancipated individuals. Unsurprisingly, the mountainous regions that make up the eastern and northern parts of the island stand out in terms of their relative suitability to refuge. This suitability emerges from the area’s ruggedness, but perhaps more importantly, from the area’s unsuitability to the plantation system. The suitability of this landscape was not fixed, however. As shown in Fig. 6, by 1799 the expansion of colonial infrastructure on St. Croix, especially into the northwestern mountains, significantly reshaped possibilities for refuge. Maroon Landscapes in 1750 In 1750, sugar and cotton plantations concentrated in the flatter, low-lying areas on the southern and eastern portions of the island, leaving the mountainous region to the northwest relatively devoid of settlement (Fig. 6e). Only a few roads are indicated on the Cronenberg-Jaegersberg map, which mainly connected the eastern and western parts of the island (Fig. 6a). In turn, the unsettled northwestern part of the island stands out in both its ruggedness and its remoteness. From the perspective of maroons, this region provided the greatest potential for refuge. The mountainous terrain would have made it difficult to traverse, and without significant settlement in the region, chance encounters with passersby would be unlikely, providing greater seclusion. Other mountainous parts of the island were not as secluded, however. Despite similarly rugged terrain on the eastern side of the island, there were very few areas suitable for maroon settlement (Fig. 6a). In part, this appears to be driven by the density of settlement here, as more plantations would increase the risk of capture (Fig. 6b). The 2 Page 20 of 30 L. E. Kohut et al. Fig. 6 Results of the suitability analysis showing the changing distribution of areas suitable to refuge ▸ between 1750 (a, e) and 1799 (c, f). Differences are particularly stark in the area known as Maroon Ridge (e, f). Smaller inset maps (b, d) show changes in structure density topography likely also played a role. The mountains here form several narrower, discontinuous ridges. The flatter terrain on either side was clearly preferred for early plantations and impinged upon the seclusion of these mountains. Maroon Landscapes in 1799 By 1799, the built landscape of St. Croix had changed dramatically (Table 3). The number and density of settlements increased substantially, along with the extent of roads throughout the island (Fig. 6c, d). The northwestern part of the island in particular saw significant development (Fig. 6f). The growth of colonial infrastructure through this area significantly contracted possibilities for maroon settlement. As shown in Table 3, areas of refuge were already scarce in 1750, totaling 1684.37 ha or only about 8% of the island’s land area. By 1799, suitable area declined to 652.60 ha, or approximately 3% of the island—a 60% decline. In addition to the overall contraction in suitable areas reported in Table 3, the remaining suitable areas were far more discontinuous. The landscape became an archipelago of small “islands” suitable for refuge. The largest contiguous area with a suitability greater than 0.5 was 162 ha in 1750. By 1799, this had dropped to 42 ha—a decline of 74%. Increased patchiness of suitable terrain almost certainly made marronage riskier and could have undermined community cohesion. It could have also endangered continuous occupation of maroon communities, driving marronage towards tactics of temporary refuge. The most persistent area of suitable land in the northwestern hills is the northern face of Maroon Ridge. On the eastern side of the island, possibilities for marronage expanded slightly, likely owing to the abandonment of cotton plantations there. From 1750 to 1799, the character of maroon landscapes changed along with their distribution. Areas most suitable to maroon settlements in 1799 were significantly less secluded, as measured by their proximity to colonial buildings and roads, visibility, and landscape permeability (Fig. 7; Table 4). The character of these suitable areas highlights that maroons in 1799 had to carve out spaces of refuge in areas that were increasingly less secluded. At the same time, suitable areas were further from rivers, adding to the precarity of maroons and increasing their risk of capture. These changes approximate trade-offs that fleeing Africans would have likely faced during this period of colonial settlement. With the expansion of colonial infrastructure across the island, maroons were left to carve out spaces of seclusion in a landscape that was increasingly precarious. Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 21 of 30 2 2 Page 22 of 30 L. E. Kohut et al. Table 3 Metrics of colonial infrastructure and fugitive settlement suitability Year Colonial Roads structures (km) (n) Fugitive settlement suitability Total suitable area Fewer possibilities Greatest possibilities 0.51–0.6 0.61–0.7 0.71–0.8 120.66 ha 22.83 ha 0.58 ha 1684.37 ha 1750 218 76.03 1117.08 ha 423.22 ha 1799 464 246.57 145.39 ha 438.20 ha 62.23 ha 0.81–9 0.91–1 6.30 ha 0.48 ha 652.60 ha Discussion Ejstrud’s (2008) initial predictive model based on the Cronenberg-Jaegersberg map provided an informative snapshot of potential maroon habitation on St. Croix in 1750 by analyzing viewshed and elevation on the island. Since then, advances in geospatial modeling and associated technologies, coupled with the digitization of the Oxholm 1799 map, make it possible to explore the changing affordances of maroon settlement on the island and suggest potential motives for fugitive movement. The implementation of suitability analysis with fuzzy logic is used to develop a heuristic model that reflects possibilities for mobility and settlement across time in relation to colonial expansion. This dynamic approach to historical geospatial analysis encourages us to move beyond the idea of a static, passive landscape to one that is continually transforming. Furthermore, the refined dataset of 10-m resolution LiDAR-derived DEM, compared to the 50-m resolution of Ejstrud’s study, gives us greater precision by which to assess potential settlement. By modeling refuge affordance heuristics across time, it becomes apparent that the viability of maroon settlement in St. Croix, even in the areas most remote from colonial settlements and roads near Maroon Ridge, became increasingly untenable toward the end of the eighteenth century. This period marked a significant growth in sugar plantations and populations on the island, nearly tripling from 10,200 in 1755 to 28,839 in 1797 (Hall, 1992:5). Additionally, the development of the Frederiksted port and increased trade throughout the western part of the island spurred colonial expansion into the northwestern quadrant of the island. Encroaching plantations constricted potential maroon settlement to the north coast and small patches in the northwest mountains. This fragmentation of suitable maroon lands may have contributed to the creation of disarticulated maroon communities along Maroon Ridge, as opposed to a larger, more cohesive group. While the extent of clandestine encampment and settlement on Maroon Ridge remains unknown, declining seclusion and increasingly precarious conditions within Maroon Ridge would have forced many maroons to seek refuge elsewhere, including fleeing to other islands. Archival records discuss a number of attempted and successful maritime maroon voyages from St. Croix, mainly to Puerto Rico, during the mid- to late-eighteenth century (Hall, 1985). It is possible many of these maroons sailed their own canoes from the shores of Maroon Ridge, relying on the ocean’s natural currents. Others may have stowed away on legitimate flagged vessels, commandeered ships, or paid knowing sailors for safe passage (Dunnavant, 2021a). Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 23 of 30 2 Fig. 7 Comparison of the distribution of input variables for suitable areas (suitability > 0.5) in 1750 and 1799 2 Page 24 of 30 L. E. Kohut et al. Table 4 Comparison of variables across suitable areas (suitability > 0.5) in 1750 and 1799. Welch’s two sample t-test indicates statistically significant differences between years for all variables Variable 1750 Mean Visibility 1799 SD Mean t test SD 0.294 0.140 0.345 0.203 20.109 7.695 22.631 9.905 Distance from settlements 19,887.130 9809.332 6579.030 3000.867 239.07** Distance from roads 24,331.490 10,605.093 6501.139 3482.549 294.62** Proximity to streams 4316.389 2594.749 4371.147 2594.749 − 2.9089* Landscape permeability − 40.382** − 38.549** * P < 0.01 ** P < 0.001 Conclusion The methodological approach taken here opens a pathway for those working in other geographic contexts to conceptualize how landscape affordances might vary in relation to differentially situated actors, and how they might change in relation to transformations of its human and non-human constituents. This study demonstrates how a suitability model can be used to generate multicriteria affordance simulacra derived from data extracted from historical maps, contemporary terrain model data, and simulations of human foot traffic. Identifying affordances, however, requires both domain-specific and culturally and historically sensitive knowledge. Like all models, it is partial and a simplified representation of what was surely a much more complex context and set of experiences by diverse actors in the past. Additional variables such as historic tree cover and potential landscape modifications cannot be modeled with present information and technologies. As with any model, ours does not account for all possibilities of maroons as historically situated actors. Affordances for marronage are not directly observable, though we have attempted to account for at least some of the most salient heuristics for considering the features in the landscape that would have set the broad (and changing) parameters for marronage. The maroon refuge affordance model presented here is not meant to totalize maroon landscapes. Instead, our approach provides a framework for developing affordance heuristics that are subject-oriented and historically and culturally contingent. For example, while we generally assume that formerly enslaved Africans were mostly concerned with defensive tactics and strategies, Vincent Brown (2020) challenges us to view marronage as an extension of African warfare and imagine how an offensive posture may have influenced maroon mobility. In the context of Jamaica, the British were forced to sign treaties acknowledging the autonomy of maroon settlements after various “maroon wars” proved too dangerous and costly for the colonial forces to continue. It is likely that such offensive tactics were more important for larger and longer-term Maroon communities than the temporary refuge settlements considered here. Perhaps an offensive heuristic model would reveal Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 25 of 30 2 alternative affordances as these historically situated actors sought to control strategic resources and locations and slow colonial expansion. This approach has broad applicability beyond marronage, particularly in contexts where human-landscape interaction defies expectations of efficiency and optimization. In this way, our approach compliments others who have addressed the limitations of traditional GIS approaches to account for the historically and culturally situated goals and local landscape knowledge. These approaches have been fruitfully applied to questions of movement (e.g., Howey, 2017, 2011; Supernant, 2017; ten Bruggencate et al., 2016) and visibility (e.g., Fábrega-Álvarez & Parcero-Oubiña, 2019; Gillings, 2015; Ogburn, 2006). Importantly, we develop an approach for integrating multiple dimensions of human-landscape interaction. Substantively, this project continues to add historical context to formerly enslaved Africans who have been traditionally written out of colonial archival records. Suitability modeling allows us to enact a form of counter mapping that approximates African geographies of resistance in a colonial Atlantic world context (Dunnavant et al., 2023). It serves to mitigate, even in a small way, the cartographic violence that occurs when colonial territories are deemed fungible and defined by colonial ownership and productivity, turning the cartographic gaze toward an anticolonial worldview (King, 2015; McKittrick, 2006; Perry, 2018). To etch maroon presence onto colonial maps acknowledges their fugitive presence on the island, challenging the inherent dominance implicit in colonial landscapes. If we take seriously Imani Perry’s (2018:180) claim that maps should be viewed as “sites of contestation” and “as places for alternative imaginings and naming of relations and representations,” we need to consider how we can read these maps beyond their intended functions. Whereas the early colonial maps of St. Croix were drafted for purposes of demarcating colonial defenses, plantation plots, and taxing commodities, they can also reveal “rival geographies” (Camp, 2004). As theorized by Camp (2004:7), rival geographies are “alternative ways of knowing and using plantation and southern space that conflicted with planters’ ideals and demands” and serve as “space for private and public creative expression, rest and recreation, alternative communication, and importantly, resistance to planters’ domination of slaves’ every move.” Geospatial models provide a scaffolding for such counter-mapping, enabling insights into the landscapes of opportunity and constraint for those seeking refuge. These techniques can be used to geographically situate maroons within a landscape for which they were once rendered ungeographic. While suitability models necessarily reduce the dimensionality of those landscapes, they may help us identify spaces occupied by those who were able to sustain life outside and beyond the colonial order. Through future community-engaged field survey in the areas most suited to refuge, it may be possible to gain further insights into those otherwise rendered invisible in colonial maps and texts. Given the ephemeral nature of these sites and refuge assemblages generally, such work will require a close attention to the archaeological record (Hauser, 2022). Archaeologists are currently exploring efforts to groundtruth areas within Maroon Ridge (pers. comm. Todd Ahlman) and convert a portion of Maroon Ridge into a culturally protected Maroon Sanctuary (Davis, 2022). Such efforts would shed new light on the materiality of marronage as well as help determine the efficacy of the geospatial model. 2 Page 26 of 30 L. E. Kohut et al. Supplementary Information The online version contains supplementary material available at https://doi. org/10.1007/s10816-024-09680-7. Acknowledgements The research for this article was conducted with the support of an Academic Pathways Postdoctoral Fellowship at Vanderbilt University. This research did not receive other specific grants from funding agencies in the public, commercial, or not-for-profit sectors. Modeling and analysis were conducted using the facilities of the Vanderbilt University Spatial Analysis Research Laboratory and the Winthrop University Geospatial Environmental Modeling Laboratory. Thanks to Samantha Turley for assisting with the georeferencing of historic maps and LiDAR data processing. We gratefully acknowledge Dr. Daniel Hopkins for providing a high-resolution scan of the historic map of St. Croix by Johann Cronenberg and Johann von Jaegersberg, 1750. Finally, we thank the four anonymous reviewers whose constructive feedback improved the paper. Author Contribution L.K: Methodology; Software; Data curation; Formal analysis; Visualization; Writing—original draft; Writing—review & editing. S.W.: Data curation; Methodological concept; Modeling and analysis; Validation; Visualization; Supervision; Writing—original draft; Writing—review & editing. J.D.: Conceptualization; Investigation; Data collection; Writing—original draft; Writing—review & editing. Funding Open access funding provided by the Carolinas Consortium. Academic Pathways Postdoctoral Fellowship at Vanderbilt University Data Availability Data used in this analysis are publicly available from the sources listed in the manuscript. Declarations Conflict of Interest The authors have no relevant financial or non-financial interests to disclose. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/ licenses/by/4.0/. References Agorsah, E. K. (1993). Nanny town excavations: Rewriting Jamaica’s history? Jamaican Geographer, 8, 1–7. Alexakis, D., & Sarris, A. (2010). Environmental and human risk assessment of the prehistoric and historic archaeological sites of Western Crete (Greece) with the use of GIS, remote sensing, fuzzy logic and neural networks. In M. Ioannides, D. Fellner, A. Georgopoulos, & D. G. Hadjimitsis (Eds.), Digital heritage (Vol. 6436, pp. 332–342). Springer Berlin Heidelberg. https://doi.org/10. 1007/978-3-642-16873-4_25 Atkinson, E. E., & Marín-Spiotta, E. (2015). Land use legacy effects on structure and composition of subtropical dry forests in St. Croix, U.S. Virgin Islands. Forest Ecology and Management, 335, 270– 280. https://doi.org/10.1016/j.foreco.2014.09.033 Bonham-Carter, G. F. (2014). Geographic information systems for geoscientists: Modelling with GIS. Elsevier. Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 27 of 30 2 Brandeis, T. J., & Oswalt, S. N. (2007). The status of US Virgin Islands’ forests, 2004. [Resource Bulletin]. U.S. Department of Agriculture, Forest Service, Southern Research Station. https://doi.org/ 10.5555/20083013113 Brown, V. (2020). Tacky’s revolt: The story of an Atlantic slave war. Belknap Press: An Imprint of Harvard University Press. Byrne, D. (2016). Counter-mapping in the archaeological landscape. In B. David & J. Thomas (Eds.), Handbook of landscape archaeology (pp. 609–616). Routledge. Cameron, C. M. (2013). How people moved among ancient societies: Broadening the view. American Anthropologist, 115(2), 218–231. Camp, S. M. H. (2004). Closer to freedom: Enslaved women and everyday resistance in the plantation south. University of North Carolina Press. Chemero, A. (2003). An outline of a theory of affordances. Ecological Psychology, 15(2), 181–195. Colwell-Chanthaphonh, C., Chartier, S. L., & Jacquin-Ng, S. (2014). The search for Makak: A multidisciplinary settlement history of the northern coast of Le Morne Brabant, Mauritius. International Journal of Historical Archaeology, 1–40,. https://doi.org/10.1007/s10761-014-0263-4 Davis, O. (2022). Open forum: Our maroon ancestors deserve a sanctuary on St. Croix. St. Thomas Source, July 18, 2022. https://stthomassource.com/content/2022/07/18/open-forum-our-maroonancestors-deserve-a-sanctuary-on-st-croix/. Retrieved, August 20, 2024. Dawson, K. (2021). A sea of Caribbean islands: Maritime maroons in the Greater Caribbean. Slavery & Abolition, 42(3), 428–448. https://doi.org/10.1080/0144039X.2021.1927509 Deagan, K. A., & MacMahon, D. A. (1995). Fort Mose: Colonial America’s black fortress of freedom. University Press of Florida. Debien, G. (1966). Le Marronage aux Antilles Françaises au XVIIIe siècle. Caribbean Studies, 6(3), 3–43. JSTOR. https://www.jstor.org/stable/25611959 Dragicevic, S., Marceau, D. J., & Marois, C. (2001). Space, time, and dynamics modeling in historical GIS databases: A fuzzy logic approach. Environment and Planning B: Planning and Design, 28(4), 545–562. https://doi.org/10.1068/b2567 Dunnavant, J. P. (2021a). Have confidence in the sea: Maritime maroons and fugitive geographies. Antipode, 53(3), 884–905. Dunnavant, J. P. (2021b). In the wake of maritime marronage. Slavery & Abolition, 42(3), 466–483. Dunnavant, J. P., Wernke, S. A., & Kohut, L. E. (2023). Counter-mapping maroon cartographies: GIS and anticolonial modeling in St. Croix. ACME: An International Journal for Critical Geographies, 22(5), 1294–1319. Ejstrud, B. (2008). Maroons and landscapes. Journal of Caribbean Archaeology, 8, 1–14. Fábrega-Álvarez, P., & Parcero-Oubiña, C. (2019). Now you see me. An assessment of the visual recognition and control of individuals in archaeological landscapes. Journal of Archaeological Science, 104, 56–74. Gillings, M. (2009). Visual affordance, landscape, and the megaliths of Alderney. Oxford Journal of Archaeology, 28(4), 335–356. https://doi.org/10.1111/j.1468-0092.2009.00332.x Gillings, M. (2012). Landscape phenomenology, GIS and the role of affordance. Journal of Archaeological Method and Theory, 19(4), 601–611. https://doi.org/10.2307/23365981 Gillings, M. (2015). Mapping invisibility: GIS approaches to the analysis of hiding and seclusion. Journal of Archaeological Science, 62, 1–14. Greeno, J. G. (1994). Gibson’s affordances. Psychological Review, 101(2), 336–342. APA PsycArticles®. https://doi.org/10.1037/0033-295X.101.2.336 Hall, N. (1985). Maritime maroons: “Grand Marronage” from the Danish West Indies. The William and Mary Quarterly, 42(4), 476–498. Hall, N. (1992). Slave society in the Danish West Indies: St John, and St. Croix. University of the West Indies Press. Haraway, D. (1988). Situated knowledges: The science question in feminism and the privilege of partial perspective. Feminist Studies, 14(3), 575–599. Harley, J. B. (1992). Rereading the maps of the Columbian encounter. Annals of the Association of American Geographers, 82(3), 522–542. https://doi.org/10.1111/j.1467-8306.1992.tb01973.x Hatzinikolaou, E., Hatzichristos, T., Siolas, A., & Mantzourani, E. (2003). Predicting archaeological site locations using GIS and fuzzy logic. In M. Doerr & A. Sarris (Eds.), The digital hertiage of archaeology, archive of monuments and publications (pp. 69–178). Hellenic Ministry of Culture. Hauser, M. W. (2022). The work of boundaries: Critical cartographies and the archaeological record of the relatively recent past. Annual Review of Anthropology., 51, 509–526. 2 Page 28 of 30 L. E. Kohut et al. Hazen, H. D., & Harris, L. (2006). Power of maps: (Counter) mapping for conservation. ACME: An International Journal for Critical Geographies, 46(1), 99–130. Herzog, I. (2016). The potential and limits of optimal path analysis. In A. Bevan & M. Lake (Eds.), Computational approaches to archaeological spaces (pp. 179–211). Left Coast Press. Hopkins, D. (1992a). An 18th-century cadastral audit in the Danish West Indies. Cartography and Geographic Information Systems, 19(2), 69–79. https://doi.org/10.1559/152304092783786627 Hopkins, D. (1992b). An early map and the cadastral survey of St. Croix, Danish West Indies 1734–41/A Cartographic Cul-de-sac. Cartographica: The International Journal for Geographic Information and Geovisualization, 29(3 & 4), 1–19. Hopkins, D., Morgan, P., & Roberts, J. (2011). The application of GIS to the reconstruction of the slaveplantation economy of St. Croix, Danish West Indies. Historical Geography, 39, 85–104. Howey, M. C. L. (2011). Multiple pathways across past landscapes: Circuit theory as a complementary geospatial method to least cost path for modeling past movement. Journal of Archaeological Science, 38(10), 2523–2535. https://doi.org/10.1016/j.jas.2011.03.024 Howey, M. C. L., & Brouwer Burg, M. (2017). Assessing the state of archaeological GIS research: Unbinding analyses of past landscapes. Journal of Archaeological Science, 84, 1–9. Ibarrola, M. E. (2022). Negotiating freedom in Florida: A study of ethnogenesis and resistance at three nineteenth-century Florida sites. International Journal of Historical Archaeology, 26(3), 647–714. https://doi.org/10.1007/s10761-021-00618-3 Ingold, T. (1993). The temporality of the landscape. World Archaeology, 25(2), 152–174. Ingold, T. (1992). Culture and the perception of the environment. In E. Croll & D. Parkin (Eds.), Bush base: Forest farm. Culture, environment and development (pp. 39–56). Routledge. Jarosław, J., & Hildebrandt-Radke, I. (2009). Using multivariate statistics and fuzzy logic system to analyse settlement preferences in lowland areas of the temperate zone: An example from the Polish Lowlands. Journal of Archaeological Science, 36(10), 2096–2107. https://doi.org/10.1016/j.jas. 2009.06.004 Jones, K. S. (2003). What is an affordance? Ecological Psychology, 15(2), 107–114. King, A. D. (2015). Urbanism, colonialism, and the world-economy. Routledge. Kohut, L. E. (2018). A multidirectional model for studying mobility affordance of past landscapes. Journal of Archaeological Science: Reports, 19, 239–247. Kosiba, S., & Bauer, A. M. (2012). Mapping the political landscape: Toward a GIS analysis of environmental and social difference. Journal of Archaeological Method and Theory, 14(1), 61–101. Kusimba, C. M. (2004). Archaeology of slavery in East Africa. African Archaeological Review, 21(2), 59–88. Landers, J. (2005). Leadership and authority in Maroon settlements in Spanish America and Brazil. In J. C. Curto & R. Soulodre-LaFrance (Eds.), Africa and the Americas: Interconnections during the slave trade (pp. 173–84). Africa World Press. Llobera, M. (1996). Exploring the topography of the mind: GIS, social space and archaeology. Antiquity, 70(269), 612–622. Llobera, M. (2000). Understanding movement: A pilot model towards the sociology of movement. In G. R. Lock (Ed.), Beyond the map: Archaeology and spatial technologies (pp. 65–84). IOS Press. Llobera, M. (2001). Building past landscape perception with GIS: Understanding topographic prominence. Journal of Archaeological Science, 28(9), 1005–1014. Llobera, M. (2007a). Modeling visibility through vegetation. International Journal of Geographical Information Science, 21(7), 799–810. Llobera, M. (2007b). Reconstructing visual landscapes. World Archaeology, 39(1), 51–69. Llobera, M. (2012). Life on a pixel: Challenges in the development of digital methods within an “interpretive” landscape archaeology framework. Journal of Archaeological Method and Theory, 19(4), 495–509. https://doi.org/10.2307/23365975 Llobera, M., Fábrega-Álvarez, P., & Parcero-Oubiña, C. (2011). Order in movement: A GIS approach to accessibility. Journal of Archaeological Science, 38, 843–851. Lock, G., & Pouncett, J. (2017). Spatial thinking in archaeology: Is GIS the answer? Journal of Archaeological Science, 84, 129–135. https://doi.org/10.1016/j.jas.2017.06.002 Marshall, L. W. (2018). Maroon archaeology beyond the Americas: A view from Kenya. Historical Archaeology, 52(4), 717–740. https://doi.org/10.1007/s41636-018-0146-3 McKittrick, K. (2006). Demonic grounds: Black women and the cartographies of struggle. University of Minnesota Press. Modeling Marronage: GIS Heuristics of Refuge Affordances… Page 29 of 30 2 McRae, B. H., Dickson, B. G., Keitt, T. H., & Shah, V. B. (2008). Using circuit theory to model connectivity in ecology and conservation. Ecology, 10, 2712–2724. McRae, B. H., Popper, K., Jones, A., Schindel, M., Buttrick, S., Hall, K., Unnasch, B., & Platt, J. (2016). Conserving nature’s stage: Mapping omnidirectional connectivity for resilient terrestrial landscapes in the Pacific Northwest. The Nature Conservancy. http://nature.org/resialienceNW. Accessed 8 Nov 2022. Minetti, A. E., Moia, C., Roi, G. S., Susta, D., & Ferretti, G. (2002). Energy cost of walking and running at extreme uphill and downhill slopes. Journal of Applied Physiology, 93(3), 1039–1046. https://doi. org/10.1152/japplphysiol.01177.2001 de Moraes, I. P. (2012). Do tempo dos Pretos d’antes aos Povos do Aproaga: Patrimônio arqueológico e territorialidade quilombola no vale do rio Capim (PA) [Dissertação de Mestrado]. Universidade Federal do Pará. Norton, H. K. (2020). Finding the spaces betwixt and between: GIS of the 1733 St. Jan Slave Rebellion. International Journal of Historical Archaeology, 1–20. https://doi.org/10.1007/ s10761-019-00526-7 Norton, H. K., & Espenshade, C. T. (2007). The challenge in locating maroon refuge sites at Maroon Ridge, St. Croix. The Journal of Caribbean Archaeology, 7, 1–17. Office for Coastal Management. (2012). C-CAP land cover, United States Virgin Islands, St. Croix . https://www.fisheries.noaa.gov/inport/item/48361. Accessed 28 July 2020. Office for Coastal Management. (2013). 2013 NOAA topographic lidar: U.S. virgin islands (St. Croix, St. John, St. Thomas). https://www.fisheries.noaa.gov/inport/item/48219. Accessed 28 July 2020. Ogburn, D. E. (2006). Assessing the level of visibility of cultural objects in past landscapes. Journal of Archaeological Science, 33, 405–413. Oldendorp, C. G. A. (1994). History of the mission of the Evangelical Brethren on the Caribbean islands of St. Thomas, St. Croix, and St. John. In G. F. Tyson & A. R. Highfield (Eds.), Kamina Folk (pp. 61–97). Virgin Islands Humanities Council. Orser, C., & Funari, P. (2001). Archaeology and slave resistance and rebellion. World Archaeology, 33(1), 61–72. Panich, L. M., Allen, R., & Galvan, A. (2018). The archaeology of Native American persistence at mission San José. Journal of California and Great Basin Anthropology, 11–29. Pavlovskaya, M., & Martin, K. S. (2007). Feminism and geographic information systems: From a missing object to a mapping subject. Geography Compass, 1(3), 583–606. https://doi.org/10. 1111/j.1749-8198.2007.00028.x Pelletier, D., Clark, M., Anderson, M. G., Rayfield, B., Wulder, M. A., & Cardille, J. A. (2014). Applying circuit theory for corridor expansion and management at regional scales: Tiling, pinch points, and omnidirectional connectivity. PLoS ONE, 9(1), e84135. https://doi.org/10.1371/journ al.pone.0084135 Perry, I. (2018). Vexy thing: On gender and liberation. Duke University Press. Pope, P. (1972). A Maroon settlement on St. Croix. Negro History Bulletin, 35(7), 153–154. Price, R. (Ed.). (1973). Maroon societies: Rebel slave communities in the Americas. Anchor Books. Rayfield, B., Fortin, M.-J., & Fall, A. (2011). Connectivity for conservation: A framework to classify network measures. Ecology, 92(4), 847–858. https://doi.org/10.1890/09-2190.1 Ripy, J., Grossardt, T., Shouse, M., Mink, P., Bailey, K., & Shields, C. (2014). Expert systems archeological predictive model. Transportation Research Record, 2403(1), 37–44. https://doi.org/10. 3141/2403-05 Romário Melo De Jesus, H. (2024). Arqueologia por quilombolas no Brasil Uma narrativa Tupinambá do baixo Tapajós, Santarém. Revista de Arqueologia, 37(1), 126–136. https://doi.org/10.24885/ sab.v37i1.1138 Roopnarine, L. (2010). Maroon resistance and settlement on Danish St. Croix. Journal of Third World Studies XXVII, 2, 89–108. La Rosa Corzo, G. (2005). Subsistence of cimarrones: An archaeological study. In S. L. Dawdy, L. A. Curet, & G. La Rosa Corzo (Eds.), Dialogues in Cuban archaeology (pp. 163–180). University of Alabama Press. Sayers, D. O. (2016). A desolate place for a defiant people: The archaeology of Maroons, indigenous Americans, and enslaved laborers in the Great Dismal Swamp. University Press of Florida. Simonsen, G., & Christensen, R. (2023). Together in a small boat: Slavery’s fugitives in the Lesser Antilles. The William and Mary Quarterly, 80(4), 611–646. https://doi.org/10.1353/wmq.2023. a910393 2 Page 30 of 30 L. E. Kohut et al. Singleton, T. (2020). Archaeology of Marronage in the Caribbean Antilles Arqueologia da marronagem nas Antilhas do Caribe. Revista Do Museu De Arqueologia e Etnologia, 35, 1–13. Soule, R. G., & Goldman, R. F. (1972). Terrain coefficients for energy cost prediction. Journal of Applied Physiology, 32(5), 706–708. https://doi.org/10.1152/jappl.1972.32.5.706 Supernant, K. (2017). Modeling Métis mobility? Evaluating least cost paths and indigenous landscapes in the Canadian west. Journal of Archaeological Science, 84, 63–73. https://doi.org/10.1016/j.jas. 2017.05.006 ten Bruggencate, R. E., Stup, J. P., Milne, S. B., Stenton, D. R., Park, R. W., & Fayek, M. (2016). A human-centered GIS approach to modeling mobility on southern Baffin Island, Nunavut, Canada. Journal of Field Archaeology, 41(6), 684–698. https://doi.org/10.1080/00934690.2016.1234897 Urban, D. L., Minor, E. S., Treml, E. A., & Schick, R. S. (2009). Graph models of habitat mosaics. Ecology Letters, 12(3), 260–273. U.S. Geological Survey. (2019). National Hydrography Dataset (ver. USGS National Hydrography Dataset Best Resolution (NHD) for Hydrologic Unit (HU) 4 - 2001 (published 20191002)). https://www. usgs.gov/national-hydrography/access-national-hydrography-products. Accessed 12 Oct 2020. Webster, D. S. (1999). The concept of affordance and GIS: A note on Llobera (1996). Antiquity, 73(282), 915–917. Cambridge Core. https://doi.org/10.1017/S0003598X00065698 Weik, T. (1997). The archaeology of maroon societies in the Americas: Resistance, cultural continuity, and transformation in the African Diaspora. Historical Archaeology, 31(2), 81–92. https://doi.org/ 10.1007/BF03373604 Weik, T. (2008). Mexico’s Cimarron heritage and archaeological record. African Diaspora Archaeology Newsletter, 11(2), 1–12. Weik, T. (2009). The role of ethnogenesis and organization in the development of African-Native American settlements: An African Seminole model. International Journal of Historical Archaeology, 13(2), 206–238. https://doi.org/10.1007/s10761-009-0079-9 Weik, T. (Ed.). (2019). The archaeology of removal in North America. University Press of Florida. Wernke, S. A., Kohut, L. E., & Traslaviña, A. (2017). A GIS of affordances: Movement and visibility at a planned colonial town in highland Peru. Journal of Archaeological Science, 84, 22–39. Wheatley, D. (2004). Making space for an archaeology of place. Internet Archaeology, 15. http://intarch. ac.uk/journal/issue15/wheatley_toc.html. Accessed 25 Mar 2021. White, C. (2006). Maroon Archaeology. African Diaspora Archaeology Newsletter, 9(1), 1–12. White, C. (2010). Maroon archaeology is public archaeology. Archaeologies, 6(3), 485–501. https://doi. org/10.1007/s11759-010-9154-4 Zadeh, L. A. (1978). Fuzzy sets as a basis for a theory of possibility. Fuzzy Sets and Systems, 1(1), 3–28. https://doi.org/10.1016/0165-0114(78)90029-5 Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.