Sensemaking in the Wild: A Review of Practitioner Collected Geospatial Data and its Synthesis within Protected Areas for Poaching Mitigation

ABSTRACT A key challenge for mitigating poaching within protected areas is to understand the geospatial data that are collected by practitioners in protected areas and to characterize the ability to synthesize those data with landscape-level data to form a holistic picture of the movement patterns of humans and animals. Literature reviewed from the past 15 years on geospatial data collected by practitioners to mitigate wildlife poaching reveals a gap in our knowledge on how protected area practitioners make sense of geospatial data that are collected within protected areas. Geospatial data collected within protected areas provide an understanding of movement patterns of humans and animals, which can provide insight on best practices for poaching mitigation, to include where to emplace new geospatial sensors. We classify these data as device-based and human-generated, and their potential to provide geospatially referenced information that forms patterns of poaching activity. This article examines two primary types of geospatial data collected in protected areas, highlights the challenges associated with this data, and discusses knowledge gaps regarding how protected areas make sense of spatial data. We conclude with recommendations for future research on characterizing how geospatial data is represented in protected areas, and filling knowledge gaps on how protected area personnel use those data.


Introduction
Poaching is a global environmental issue, which threatens keystone species and can also threaten human health through the spread of zoonotic diseases from poached species (Monroe et al. 2015;Scanlon, 2021;Wolfe et al. 2005). This is a complex context, with enforcement occurring at the tactical, operational, and strategic levels (Ferreira and Dziba 2021;Gustafson, Sandstrom, and Townsend 2018;Holden et al. 2019). Multiple solution spaces currently exist to tackle the poaching challenge, ranging from community education to advanced applications in geospatial technologies (Koster et al. 2016;Massé et al. 2017;Reuter and Bisschop 2016). In recent years, research investigating anti-poaching methods has included supply-side and demand-side strategies to include the implementation of conservation buffer zones, increase in ranger patrols, community education, and in the development of technological solutions to poaching crime (Ferreira, Pfab, and Knight 2014;Gorgun 2020;Mamba, Randhir, & Fuller, 2020;Rubino & Pienaar, 2018). Kamminga et al. (2018) concludes through a review of literature that 'no Anti-Poaching System (APS) or sensor technique is able to meet all the requirements for effective poaching detection by itself'. It is with this understanding that the global community of conservation practitioners must look towards the synthesis of all available data to form a holistic picture of poaching within their own protected area borders, as well as to understand poaching patterns and trends in other nearby protected areas, as poachers do not stop at protected area boundaries or international borders.
This review focuses on the types of geospatial data collected by protected area practitioners, which forms the understanding of movement of humans (rangers and poachers) and animals across a protected area landscape. We review recent work at the intersections of protected area management, geospatial sensor networks, and information synthesis in GIScience to identify research challenges and opportunities facing those who wish to mitigate poaching. In the sections that follow, we discuss the two main types of geospatial data collected by practitioners in protected areas and the challenges associated with each type of data. Finally, we provide recommendations for future research.
For the initial Scopus search, we focused on journal articles, book chapters, and conference proceedings in the following subject areas to limit the scope: Environmental Sciences, Earth, & Biological Sciences, Social Sciences, Computer Science, and Engineering. Articles for this literature review were then identified and collected using Scopus (Elsevier B.V., Netherlands) and Google Scholar (Google LLC, Mountain View, CA, USA). Additionally, citation snowballing methods were used (backwards and forwards) to discover additional literature relevant to the discussion. The following keywords were used in the refined literature search: (i) poaching, (ii) wildlife, (iii) protected areas, (VII) geospatial, (VIII) synthesis, (IX) data, and(X) sense making.
Of the approximately 3300 articles discovered in the initial Scopus search, we then selected a three per cent subset (N = 100) of all articles in order to examine geospatial data synthesis in poaching-related work. We found that current work conducted in protected areas using geospatial data primarily focuses on one data source, and research using multiple data sources to mitigate poaching was sparse.
Geospatial data collected by protected area practitioners could be classified into two key types: devicebased and human-generated. Device-based data are any data that are collected automatically via a sensor. Human-generated data in the context of this review include all data that are produced with direct human intervention. Common sources of human-generated data include field reports through cell phone applications, radio communications from rangers back to operation centres, or handwritten reports (Reuter and Bisschop 2016).
To develop our review, we collected recent articles that are categorized into two themes: types of collected geospatial data and approaches for sensemaking with those data in protected area management. The first portion of this review focuses on the types of geospatial data that are collected within protected areas for species protection and poaching detection and the challenges associated with these data. The latter portion of this review focuses on the importance of understanding the movement of humans and animals across the landscape to mitigate poaching, and knowledge gaps on protected area practitioners to make sense of poachingrelated data.

Common geospatial data and data challenges in protected areas
Prior to understanding how a protected area synthesizes its geospatial data, a key objective is to first characterize existing data and understand the challenges that are associated with that data. Our literature review revealed that protected area practitioners collect two types of geospatial data: device-based and human-generated. Our discussion of specific forms of each data type is not an exhaustive list, but rather a representation of the common forms of geospatial data currently collected by practitioners in protected areas worldwide. We also found that the current research, which applies these data towards antipoaching, typically focuses on one of the two types, although some research does overlay practitionercollected data with other forms of geographic data to include road networks, terrain, cell towers, etc. Finally, we found that recent research on poachingrelated activities using geospatial data is primarily focused in the African continent ( Figure 2). The first section of this review focuses on devicebased and human-generated geospatial data separately in order to describe these data types, as well as discuss challenges associated with them. This review found that the majority of recent literature focuses on device-based data, with less attention placed on human-generated data in the context of poaching mitigation. We expect that specific forms of device-based technologies will continue to evolve with technological innovation. However, implementation by practitioners may still lag due to a lack of funding .

Device-based data
For the purposes of this review, a device-based sensor is defined as any digital device that collects data independently or with little direct human intervention. Devicebased data can be static or dynamic and can transmit data in near real time (via Wi-Fi or cellular transmission) to the cloud, a locally operated database, or to internal storage from which data can be retrieved in the future (Mulero-Pázmány et al. 2014;Park, Serra, Snitch, & Subrahmanian, 2015;Rowcliffe and Carbone 2008;Tobler et al. 2008). These data include still images, video, audio or location as a point in time from radio frequency or GPS positioning systems. Common devicebased data sources used in poaching mitigation include, but are not limited to: Unmanned Aerial Systems (UAS, also colloquially referred to as drones), camera traps, acoustic devices, telemetry or radio collar devices. This section of the article will focus on the two most frequently used devices (Unmanned Aerial Systems and Camera Traps), with a short discussion on other devices used by practitioners to collect geospatial data in protected areas.

UAS
UAS (also referred to in the literature as unmanned aerial vehicle, remotely piloted aircraft system, or drones), have been the focus of ongoing conservation-related research for over a decade (Garrett and Anderson 2018;Linchant et al. 2015). UAS have been described as a system that include a ground segment, an aerial vehicle that is capable of carrying various remote sensors (Bhardwaj et al. 2016, 197).
A unique aspect to this system is that the infrastructure of a UAS becomes more complex based on the size of the airframe. The platform (or unmanned aerial vehicle) within a UAS is classified into five categories (Terwilliger et al. 2017;Winnefeld and Kendall 2011). Each of these categories is based on the size (weight), maximum altitude, and maximum speed of the airframe. As an airframe increases in size, its potential capability for distance range from the pilot, dwell time and more complex sensor integration increases due to the availability of additional energy systems, communication networks and an increased ability to carry larger sensors (Gupta, Ghonge, and Jawandhiya 2013;). As UAS proliferate and prices decrease, recent work has investigated the effectiveness of different platforms for conservation research (Linchant et al. 2015;Paul, Yuvaraj, and Gundepudi 2020). Within the context of poaching mitigation, a team of researchers investigated the effectiveness of UAS for identifying poachers within a protected area by utilizing three types of UAS platform with an electro-optical (EO) sensor to investigate poaching presence using three hiding postures during the day (van Vuuren et al. 2019). The study revealed that the ability to spot poachers with a UAS varied based on the hiding posture of the poacher. The first trend using UAS for poaching-related work revealed interest in investigating multiple sensor types. An example of this is work by Hambrecht et al. (2019), which used a UAS carrying RGB and thermal infrared (TIR) video payloads to determine the probability of detecting humans in woodland landscape in Tanzania. This work found that the canopy density and the image analyst had the most impact on the ability to detect humans in both RGB and TIR imagery. An interesting aspect of this research is that the authors chose to extract still-frame images from the video for analysis, rather than analyse the video to determine if motion in the fast-moving frames (30 frames per second) had an impact on detection. In another study, Mulero-Pázmány et al. (2014) studied sensor data for poaching enforcement, performing UAS flight surveys utilizing multiple imagery modes (RGB still frame, RGB video, 'white hot' and 'black hot' thermal infrared video). Similar to the Hambrecht et al. (2019) study, the authors also chose to extract still imagery frames from the full motion video data.
We found a second growing trend in research that utilizes UAS with other sensors such as satellite imagery or emerging technologies such as acoustic devices. Wang, Shao, and Yue (2019) recently conducted a review of UAS, manned aerial, and spaceborne research and proposed additional research on the fusion of satellite and UAS data. Additionally, Marvin et al. (2016) advocate for the integration of multiple technologies. Suggesting this may aid conservation management of endangered species. Recent work by Penny et al. (2019) investigated the effectiveness of UAS, olfactory, and acoustic methods as an anti-poaching tool. The study revealed that UAS provided a positive response for movement in rhinoceros (Ceratotherium simum).

Camera traps
The biological sciences have embraced the use of camera traps as a less-invasive method of data collection, with a primary focus on their use to enumerate a particular species in a defined geographic area or as a means to collect information on wildlife behavioural traits (Buxton et al. 2018;Caravaggi et al. 2020;Fontúrbel et al. 2020;Khwaja et al. 2019;Murphy et al. 2018;Soria-Díaz & Monroy-Vilchis, 2015). Camera traps use a digital sensor that can collect images and store pixels on a memory card or transmit pixels via cellular/wireless networks (Glover-Kapfer, Soto-Navarro, and Wearn 2019; Meek et al. 2019;Newey et al. 2015). Advantages of camera traps employed by biologists can be applied towards poaching mitigation. These advantages include: 24/7 all-weather capability, produce a visual record which can be reviewed by practitioners and law enforcement officials and the ability to set up an array of sensors to cover multiple locations on a landscape (Fleming et al. 2014;Swann, Kawanishi, and Palmer 2011;Tobler et al. 2008).
Our search for recent articles that combine camera traps and geospatial applications revealed numerous examples from the biological sciences, and an increasing number of studies that use similar methodologies towards poaching mitigation. Ferreguetti et al. (2018) collected 13 months of camera trap images to develop an occupancy model of poaching hot spots in a protected area in Brazil. This work revealed poacher hot spots near water resources and forest edges. Pettorelli et al. (2010) assessed carnivore distribution patterns by combining camera trap imagery with ecological niche factor analysis (ENFA), over an occupancy model. Their study detected a new spatial distribution pattern for the underreported bushy-tailed mongoose (Bdeogale crassicauda). By investigating the movement of animals, we can better understand where poachers may set up traps, snares, and other methods to capture or kill threatened species. Along this line of inquiry, de Matos Dias, Ferreguetti, and Rodrigues (2020) combined occupancy models and heat maps to identify potential poaching hot spot areas in order to utilize limited camera trap devices. The results of their study found that the distance to transportation routes and the density of game species were the most likely factors in estimating poaching distribution. Jenks, Howard, and Leimgruber (2012) conducted a spatial analysis of the poacher presence near ranger stations using camera traps in Thailand. The study suggests that infrastructure improvements (i.e. roads) to provide access to ranger stations may also be used by poachers. Knowledge of animal movement behaviour can be used to ascertain pathways that poachers employ to capture or kill species for profit; this framework is described as combat tracking in the literature (Gustafson, Sandstrom, and Townsend 2018). Further exploration of combat tracking frameworks may increase understanding of poacher movement within protected areas, and could aid practitioners in the design of sensor networks to detect poachers entering areas frequented by targeted species.

Other device-based sensors
Recent literature provides examples of other devicebased methods used by protected area practitioners, although research using the devices specifically for antipoaching is sparse. O'donoghue and Rutz (2016) have investigated the use of geospatially referenced biologgers that create geospatially referenced points as a real-time indicator of poaching activity. Another study used an acoustic grid of autonomous sensors to investigate the soundscape of a protected area by using a gunshot detection algorithm (Astaras et al. 2020). Additionally, one study used GPS collars on elephants to determine that individual elephants' movements at night were likely faster when and where poaching activity was occurring (Ihwagi et al. 2018). Other recent research paired camera traps with acoustic sensors, showing that the pairing of these sensors can capture complementary species and human disturbances (Buxton et al. 2018). Outside of academic research, in the public sector, the company Hawkeye 360 has utilized radio frequency (RF) signal geolocation data from a constellation of small satellites to characterize possible illicit activities within the boundaries of Garamba National Park (undefined).

Challenges related to device-based sensors
Device-based data for poaching mitigation may provide benefits, however there are also challenges associated with these technologies. Our review found that there are five challenges that emerge from the literature. These include (i) human-related challenges, (ii) technical difficulties, (iii) environmental issues, (iv) cost constraints, and (v) ethical/privacy concerns.

Human-related challenges
Human-related challenges for device-based data collection vary from the requirement of certifications to operate equipment, to physical access to a geographic location, as well as local, regional or national laws and policies (Floreano and Wood 2015;Fust & Loos, 2020;Linchant et al. 2015;Mulero-Pázmány et al. 2014;Pettorelli et al. 2010).
With respect to UAS, two challenges that have been cited frequently in the literature include pilot certifications and airspace limitations (Floreano and Wood 2015;Fust & Loos, 2020; Jiménez López and Mulero-Pázmány 2019; Linchant et al. 2015). Within the United States, an FAA UAS pilot certification is required to operate a UAS in US airspace for most research projects. This certification can take additional time that would ordinarily be used towards research design or data collection (FAA 2022;Lorah, Ready, and Rinn 2018). In addition to obtaining FAA certification, differences between the 50 US state statutes can impact the ability of an academic researcher to utilize UAS for image collections due to airspace limitations (Hodgson and Sella-Villa 2021). Outside of the USA, obtaining permission to fly a UAS can also be similarly complex with some countries requiring additional training, certifications, and regulations (Global Drone Regulations Database n.d..).
With camera traps, sensor network design factors can impact results. For instance, camera traps spaced too far apart can lead to inaccurate animal inventories, whereas camera trap arrays in close proximity result in superfluous images that take too much time to parse and are difficult to store (Newey et al. 2015). In some cases, camera trap spacing is limited by the number of camera traps available and/or access to ideal site locations for camera trap emplacement (Evans, Mosby, and Mortelliti 2019;Macaulay, Sollmann, and Barrett 2020). Additionally, many camera trap research sites do not have the ability to transmit images and video in nearreal time due to a lack of network connectivity. Humans are required in the loop to access and retrieve memory cards on a regular basis. This protocol can prove challenging when battery power does not last as long as predicted, access to camera trap sites is limited due to environmental conditions, or if a camera is stolen before someone can retrieve its memory card (Glover-Kapfer, Soto-Navarro, and Wearn 2019; Meek et al. 2019). Finally, the security of camera traps is a challenge on its own. The placement of camera traps in open areas can lead to destruction of property from nefarious actors, particularly in areas where poaching is a concern (Pettorelli et al. 2010;Tobler et al. 2008).

Technical difficulties
Technical difficulties can be a challenge when collecting data in protected areas. We found that though there are many potential technical challenges mentioned in recent literature, two in particular stood out as the most prolific. These include battery life limitations and sensor resolution (Floreano and Wood 2015;Linchant et al. 2015;Mulero-Pázmány et al. 2014;Wearn and Glover-Kapfer 2017;Weaver, Westphal, and Taylor 2021). A recent survey by Glover-Kapfer, Soto-Navarro, and Wearn (2019) found that respondents who used camera traps for research indicated that battery life ranked high as an important constraint on achieving scientific goals. Furthermore, the same survey indicated that respondents were optimistic that this challenge could be overcome in the future with improvements to integrated lithium-ion batteries, as well as the use of solar chargers in camera traps. Within UAS research, there are multiple ongoing studies on methods and design to improve the battery life of small UAS systems in order to prolong flight (Addabbo et al., 2020;Lussier 2019;Sarkar et al. 2022;Sharma and Atkins 2021) In many cases, imagery from UAS has a much higher resolution than previously used remote sensing techniques coming from satellite and manned aircraft platforms (Lu and He 2017;Marcaccio, Markle, and Chow-Fraser 2015;Müllerová et al. 2016). Although UAS have improved the ability to obtain highresolution data, there are still some sensor resolution challenges that researchers currently face when using UAS for research in protected areas, namely the choice between the carrying capacity of the platform and the quality of data (Burke et al. 2019;Linchant et al. 2015). In one study on orangutans, sensor resolution was impacted by the canopy height at the field location. Because orangutans are arboreal species, the UAS had to be flown higher than the calculated maximum height to produce data of a desired resolution to avoid trees (Burke et al. 2019).
There may be other technical limitations that have yet to be documented. Newey et al. (2015) argue that additional issues with camera trap studies may not be published because research is primarily focused on publishing on ecological advancements, not on technological difficulties. Publishing on technical difficulties with camera trap or other device-based methods of collection could aid other researchers to avoid common pitfalls.

Environmental impacts
Environmental impacts are primarily weather-based with wind or fluctuating temperature impacting these types of challenges (Cromwell et al. 2021;Linchant et al. 2015;Oleksyn et al. 2021). For UAS, wind can play a factor in the operation of a UAS as well as the quality of the collected data (Christie et al. 2016;Jeong, You, and Seok 2021;Spence and Mengistu 2016). Additionally, clouds and other weather phenomena can affect the flight profile and image quality (Kramar n.d.;Linchant et al. 2015).
Camera traps can also be affected by environmental conditions. Extreme weather (too hot or too cold) can tax battery life, and very windy conditions can lead to false-positive images taken from the ground cover or tree limbs moving in front of a camera sensor (Linchant et al. 2015;Wearn and Glover-Kapfer 2017). Additionally, Wearn and Glover-Kapfer (2017) highlight delays to data retrieval due to inhospitable weather conditions.

Cost
Operational costs are often a challenge when using device-based methods (Floreano and Wood 2015;Jiménez López and Mulero-Pázmány 2019;Knight, Bailey, and Faulkner 2018;Linchant et al. 2015). Scientists are often restricted by cost factors that decide platform specifications and sensor characteristics. In a review of UAS for wildlife monitoring, Linchant et al.
(2015) discussed the cost variation for a complete UAS package (to include but not limited to airframe, sensor and ground control equipment). Systems varied from do-it-yourself (DIY) systems, which cost as little as $2000, to professional/military-grade systems that can cost $100,000 or more.
For camera traps, cost factors include the number of cameras purchased, functionality, and camera life of higher end cameras (Meek et al. 2019;Pettorelli et al. 2010). Additionally, costs increase when camera traps are stolen or damaged by individuals who do not want a record of illicit activity (Lapuente et al. 2020;Meek, Ballard, and Falzon 2016;Meek et al. 2019;Sandbrook, Luque-Lora, and Adams 2018).

Ethical/Privacy concerns
A final issue that might arise from the use of these systems is ethical or privacy concerns. Ethical concerns arise where these devices may cause harm to either humans that live along the perimeter or within protected areas, and potential disruption or harm to animals due to the presence of these devices (Sandbrook, Luque-Lora, and Adams 2018;Sharma et al. 2020). In recent years, several scholars have identified the need for conservation and ecology communities to discuss the social implications of using surveillance technologies for conservation (Pebsworth and LaFleur 2014;Sandbrook, Luque-Lora, and Adams 2018;Sharma et al. 2020). One systematic review of commercial drone use found that only 3.5% of peer reviewed articles explored ethics associated with commercial drone use (Luppicini and So 2016). Sandbrook (2015) describes human ethical factors such as privacy, safety and psychological wellbeing as an important consideration when conducting research using UAS. Additionally, other scholars argue that the expansion of surveillance technologies for research and commercial endeavours threatens to evaporate society's expectations of privacy (Finn and Wright 2012).
Another ethical dilemma regarding the use of devicebased technologies is their potential effects on local wildlife. This pertains to potential suffering that may be caused to an individual animal by the presence of device-based technologies on the landscape. Caravaggi et al. (2020) discuss the impact that camera traps have on wildlife due to the fact that these devices emit light and sound, carry human scent, and have a tangible and novel presence in the environment. This topic has been a subject of recent UAS research as the presence of a UAS is believed to cause harm to wildlife, but there are still many gaps in our knowledge (Jiménez López and Mulero-Pázmány 2019; Smith et al. 2016). In recent work, the use of Vertical Take-off and Landing (VTOL) UAS contributed to the behavioural response of several terrestrial species in Botswana based on the aircraft flight altitude and distance from the animal (Bennitt et al. 2019). The research team found similar results in species behavioural response when they investigated videos uploaded to YouTube of UAS operations (Rebolo-Ifrán, Grilli, and Lambertucci 2019). Their study found that 26% of species observed were listed as threatened species by the International Union for Conservation of Nature. Finally, a recent study investigated how white rhinoceroses (Ceratotherium simum) respond to three stimuli (acoustic, olfactory, and drone) to determine effectiveness as deterrents that could be used to move rhinoceroses from areas of high poaching risk (Penny et al. 2019). The results of the study revealed that the rhinoceroses could perceive the drone up to at least 100 m in altitude. However, low altitude flights (less than or equal to 20 m) produced the highest rate of deterrence, especially in mother-calf pairs.

Human-generated poaching mitigation data
Human-generated data includes any data that is collected by a human in hand-written form or by manual input into a digital database. Although these data can be analog in nature, any time a location is observed within, they can help characterize the geospatial understanding of poaching patterns and trends. Typically, humangenerated geospatial data is collected by practitioners who work within protected areas, however there have been some research works on the use of these data collected through volunteered geographic information (VGI) (Vitos et al., 2013). Human-generated poaching data collected by practitioners can include, but is not limited to, written reports (such as threatened species observation logs), animal carcass sites, poaching camp locations and access control reports (Milda, Ramesh, Kalle, Gayathri, & Thanikodi, 2020;Moore et al. 2018;Reuter and Bisschop 2016). These data are generally unstructured and may not contain high fidelity geolocation information.
Compared to recent literature on device-based methods for poaching mitigation, the literature that specifically focuses on an understanding of humangenerated data collected by practitioners for poaching mitigation is sparse. Work by Reuter and Bisschop (2016) discussed types of poaching-related data to include: entrance and exit screening, perimeter fence observations, ranger patrols and vehicle reporting within the protected area. Academic research that utilizes human-generated geospatial data to understand poaching patterns is often layered with other environmental or landscape data within Geographic Information Systems (GIS). A recent study that used human-generated poaching data combined ranger reports along with several methods of modelling to predict possible poaching locations (Gholami et al. 2018). Gurumurthy et al. (2018) utilized survey data from protected area subject matter experts to study clusters within a protected area, by creating a 1 km grid over the protected area. Their team then created points based on ranger patrol reports of poaching incidents and constructed a dataset where each datapoint corresponds to a grid cell in a patrol season. Finally, Mailley (2014) discusses the use of intelligencestyle information obtained by informants to provide details on the modus operandi of local tiger poachers. Although this recent work investigates the use of human-generated data, it is not clear whether these data had a geospatial component or for that matter how accurate that geospatial component may have been if collected.
As discussed by Rein and Biermann (2013), one challenge with human-generated data is that it frequently uses natural language, which can come in various forms (synonyms, slang, etc.). Without a common lexicon for a particular issue, observations can be difficult to parse. For example, when reporting on observations of a potential protected area breach, there are multiple possibilities that a practitioner may use to describe the breach. This description can vary based on translation, and local colloquialisms. Therefore, when these observations are sent to a cloud-based database, a machine learning algorithm may have difficulty determining how an observation may be parsed (Rein and Biermann 2013).
One potential challenge for human-generated data is that when collected in analog form (pen and paper versus a digital application), geospatial information may be omitted or incorrectly documented. Unfortunately, there is scant research on this topic; therefore, it is impossible to quantify how often (or how infrequently) humangenerated poaching data utilize geographic information. Furthermore, there is little understanding regarding how accurate locational information may be or if there is consistency in the locational information provided (e.g. use of different forms of geographic coordinates or differences in reference systems).
Human error and consistency in observations present another potential challenge (De Felice and Petrillo 2018;Kirwan 1996). For example, Sarkar et al. (2022) indicated that anti-poaching efforts at one protected area in Uganda were entered into one of the three databases, but data entry was inconsistent.
Finally, access to software, training for software, and data entry into software to track analog, human-generated data may be a challenge. In the Ethnology of Infrastructure (Star 1999), Susan Leigh Star discusses challenges associated with users employing software to its fullest potential despite the best efforts of software engineers to design a system using the principles of participatory design.

Sense making, movement patterns, and data synthesis
Once device-based and human-generated data are collected, a key challenge is to make sense of these disparate data and use to determine patterns and trends of poaching within protected area boundaries. Sensemaking has been defined as the way in which 'people search for, organize, and create new knowledge from source information' (Heer and Agrawala 2008;Pirolli and Card 2005). In a recent work, Krenc et al. (2018) described sense-making as 'the knowledge about context and behaviour related to a data set of sources for a particular situation'. In the context of big data, which is typically derived from geospatial technologies, Lisle et al. (2021) defines sense-making as a 'cognitively difficult task when it involves foraging through large amounts of data to find meaningful items and inferring how those items relate to one another'. When protected area practitioners parse various types of data to determine poaching patterns and trends, they need to be able to see a holistic picture of poaching within their boundaries. This includes overlaying collected device-based and human-generated geospatial data with landscape-level and environmental geospatial data to analyse hot spots of activity (Figure 3).

Movement patterns within protected areas
A key impediment to effective strategies for combating poaching includes the complexity of human and animal movement across a protected area landscape. Incorporating temporal and spatial variables into poaching data encourages trend analysis of previous poaching activity, which can lead to new approaches for patrol routes and sensor placement.
To create effective strategies for mitigating poaching at the local level, one must first understand the local landscape and cognitive spatial processes that are employed by protected area practitioners, as well as the poachers themselves. Furthermore, the spatial distribution of endangered animals changes where a poacher may ingress or egress a protected area or hunt once an illicit entry has been made (Lemieux 2014;van Doormaal, Lemieux, & Ruiter 2018). These spatial processes are often understood by analysis of patterns of movement and previous illicit activity (Haas and Ferreira 2018;van Doormaal, Lemieux, and Ruiter 2018). In order to predict future poaching locations, one must understand the patterns of movement of three objects: the poachers, protected area rangers and targeted animal species (Lemieux 2014). It is the geographic overlap of these three patterns of movement that form the basis for the geographic profiling of poaching within a protected area (Figure 4). Geographic profiling has been used as a method to characterize how activity is tied to geographic space and has been applied by epidemiologists, biologists, and criminal scientists to predict patterns and trends of human movement through geographic space (Rossmo 2012).

Animal movement
Animal movement across time and space has been a popular topic among ecologists. Approximately a decade ago, the movement ecology framework (MEF) was born which 'introduced an integrative theory of organismal movement, linking internal state, motion capacity and navigation capacity to external factors' (Joo et al. 2020). More recent work on rhinoceros movement by Seidel et al. (2019) studied movement using telemetry from GPS enabled radio collars, which investigated movement patterns over daily, weekly, and yearly time periods.

Protected area ranger movement
Research which investigated protected area ranger movement through protected areas includes work by Moore et al. (2018), which quantified poaching-related threats based on Ranger patrol routes. Reuter and Bisschop (2016) discussed ranger movement in the form of hot spot and buffer zone patrolling as part of the Balule Nature Reserve's conservation model. Other work involving protected area rangers includes work by Moreto and Lemieux (2015), who use qualitative methods to investigate ranger perspectives on poaching activity.

Poacher movement
Poacher navigation is the most difficult of the three objects to delineate. A recent study provided an assessment of poaching spatial patterns to produce a hotspot map of potential poaching activity in north-eastern Brazil. In the study, the authors' results revealed an increase in poaching near roads and along protected area edges (de Matos Dias, Ferreguetti, and Rodrigues 2020). In another study, illegally protected area boundary crossings were combined with landscape features to predict hypothesized future crossing sites (van Doormaal, Lemieux, and Ruiter 2018). The most statistically significant result was that illegal boundary crossing sites were related proximity to roads outside the protected area. However, the study suggested that more research on additional protected areas was needed to understand the role of other landscape features in poacher illegal ingress into protected areas.

Synthesizing protected area data
Once movement data is collected by practitioners, a method to combat poaching at the local level is to synthesize these movement patterns with landscapelevel geospatial datasets. Data synthesis has been characterized in various ways in the previous GIScience research ( Gahegan and Brodaric 2002;Robinson 2011;Sui 2017). For the purpose of our review, we define geospatial data synthesis as the process in which practitioners collect, organize, analyse and fuse geospatially referenced digital and analog datasets. This includes the ways in which protected areas make sense of their data and the extent to which that data is integrated between protected areas within countries, as well as across borders for those areas that intersect with international boundaries.
The data synthesis process generates a comprehensive examination of events across time and space, and supports visual representations of the data. When we  synthesize and visualize geographic data, we reveal unknowns ( Figure 5) (MacEachren 1994;Roth 2013). If synthesis of poaching-related geospatial data is desired, a protected area must be capable of data discovery, data retrieval, visualization of data, and notification when new data is ingested. By synthesizing geospatial data and visualizing data trends, protected area practitioners can improve poaching enforcement by adding new sensors or changing ranger patrols to disrupt poacher movement.
The inconsistency in data and management practices of some protected areas further complicates efforts to synthesize data (Bertzky and Stoll-Kleemann 2009;Stoll-Kleemann 2010). Geospatial data can vary within protected areas due to factors, such as cost, capability and availability (Emerton, Bishop, and Thomas 2006;Kamminga et al. 2018). The cost of some persistent geospatial sensor platforms and digitally linked sensor networks often drives capability; the budgets of protected areas which employ these sensors rely on funding from various sources, which may include domestic government support, international assistance, and private donations (Emerton, Bishop, and Thomas 2006;Jiménez López and Mulero-Pázmány 2019;). These sources of data and their dynamics may also vary, resulting in disparities between protected areas (Bertzky and Stoll-Kleemann 2009). Therefore, it is no surprise that sense making of available protected area data can be problematic at best.

Tools for sense making within protected areas
In order to make advances that support work within protected areas, we must first understand the types of geospatial data collected and current methods of sensemaking with geospatial data in the context of antipoaching efforts. Although several geospatial tools are currently in use by practitioners to understand poaching issues, might we be able to enhance these tools by first understanding the disparate data collected within protected areas and how practitioners employ sensemaking to synthesize this data?
One impediment to sensemaking, analysis, and collaboration is the software and infrastructure that allows researchers and practitioners to upload/download data to one central location where collaborative analysis can occur. To date, current research primarily focuses on the design of separate software solutions to visualize and analyse geospatial data (Krishnappa and Turner 2014;Park et al., 2015;Speaker et al. 2021;Yousif et al. 2019). For example, Greenberg, Godin, and Whittington (2019) focus on user interface design patterns for software that supports the classification of camera trap images. A next step would be to investigate methods for combining camera trap data with other geospatial sensor data. Though there have been efforts to accomplish this, most efforts appear to happen outside of the academic space (Percivall, Reichardt, and Taylor 2015;Schuck 2010). The current literature discusses some software efforts that seek to accomplish sense-making within protected areas, but further investigation is needed to develop solutions (Cronin et al. 2021;Kulits et al. 2021). Complicating matters, technology-oriented scientific research in wildlife conservation is often focused on a single sensor or single dataset. One study which did utilize multiple data types (to include camera traps and detailed spatial datasets) found that there was a significant relationship between the number of carnivore species identified in camera trap imagery and the high number of camera traps in an array (Pettorelli et al. 2010). Although these data were not poaching-specific, this methodology can be applied to poaching. The next step could be to investigate the synthesis of diverse datasets that would allow a holistic picture of poachingrelated activity in a protected area.
The ability to visualize more than one dataset at a time is important, because in order to decrease poaching, a protected area must be able to visualize movement patterns of human and animal activity in order to schedule patrols or design the placement of new sensor networks. One commercial system that is designed to accomplish this task is Vulcan's EarthRanger software (https://www.earthranger.com/). Kulits et al. (2021) discusses Earth Ranger's capability to record 'events' which include information such as time, location, and details specific to each event type. The authors integrated EarthRanger with their own tool called 'ElephantBook' as a means to create a semi-automated human-in-the-loop system for elephant identification. In other work, Whytock et al. (2021) integrated EarthRanger capabilities with artificial intelligence-enabled camera traps to send out realtime alerts. Future efforts could investigate how this system is used by protected area practitioners to synthesize geospatial information and support sensemaking by practitioners. Additionally, within the context of wildlife conservation and poaching mitigation, there has been mention of individualized training on software tools, but to date there has been limited published work on the effectiveness of training with software that parses humangenerated data (Lynam, Htun, and Zaw 2015;Wilson et al. 2019). Another tool that is used within the wildlife conservation community to analyse poaching-related information (to include human-generated data) is SMART (https://smartconservationtools.org). SMART has mobile, web and cloud-based solutions that are used by protected area practitioners to collect, visualize, store, analyse, and report on poaching and other conservation-related observations (Wich and Piel 2021). Although the community of SMART users has a robust online portal to discuss their software, it is unclear if practitioners are meeting all of their analytical needs with this platform without results from empirical user studies.

Sense-making and the future of poaching mitigation
Geospatial data synthesis is becoming more important as new sensors are developed and implemented to enhance anti-poaching practices. When asked what direction camera-trapping would take over the next one to two decades, respondents to a recent survey indicated that integration with these sensors with UAS or other new sensors would likely be at the forefront (Glover-Kapfer, Soto-Navarro, and Wearn 2019). Speaker et al. (2021) conducted a recent survey on the state of conservation technology. Their results indicated that conservation practitioners ranked networked sensors third (out of 11 categories) in the capacity to advance conservation science, with UAS and GIS/Remote Sensing in fifth and sixth place, respectively (Speaker et al. 2021). With the influx of new device-based technologies, it is imperative that in addition to collecting and databasing the information, practitioners are able to make sense of the data in order to determine patterns and trends. Although machine learning (ML) and computer vision (CV) may help in this regard, there is still a need for human sense-making with these data. Understanding the practices of geospatial data synthesis in protected areas is a crucial first step prior to beginning the next generation of sensor network and geovisualization tools that might be targeted towards anti-poaching efforts.
A follow-on question could be: How do practitioners make sense of these combined data to design improved sensor networks? Nuñez et al. (2019) discussed the analysis and fusion of camera trap, visual and dung survey data for assessing wildlife species counts in a given location. This research provides an example of a step in the right direction towards making sense of various geospatial data, and a similar methodology could be applied towards anti-poaching data from device-based and human-generated sources.

Recommendations
Endangered species face the daily threat of being killed for meat, horns, or skin, even within the boundaries of areas designated to protect species. As poachers become more technologically advanced, protected areas must evolve by incorporating technological solutions to counter the threat. The advent of device-based geospatial technologies that can send data in real time to protected area personnel can aid in poaching mitigation. Camera trap methodologies developed for wildlife behaviour could be applied to poaching issues. For example, Rowcliffe et al. (2016) investigated animal speed and day range using camera traps. Although this work aims to understand animal movement, future research could be applied to understanding human movement for poaching enforcement. Camera traps could be used to detect the entry or movement of wildlife traffickers in protected areas, based on animal movement patterns.
Looking ahead, as new technologies that create geospatially referenced data become more widely available, new research could investigate the complexities and advantages of the synthesis of multiple geospatial data sets in a poaching mitigation environment. Future research could develop methods that investigate how one sensor mode may be advantageous over another, and how the combination of multiple sensors can fill information gaps in understanding how poachers travel through protected areas in search of targeted species. For example, new studies on the impact of sensor mode selection between still imagery and video or 'black hot' verses 'white hot' TIR could provide new insight into best practices for poacher detection within protected areas.
Despite the growth of technology to meet wildlife conservation, and specifically poaching mitigation needs, there are few examples that document how device-based geospatial data is synthesized with humangenerated data in the protected area context. In a different domain, a NATO-funded study investigated the fusion of structured and unstructured humangenerated data with device-generated data (Rein and Biermann 2013). It is with an eye towards a similar goal that the global community of anti-poacher subject matter experts could be investigating how to improve efficiencies with the synthesis of geospatial data. By investigating the process of sensemaking currently used in protected areas, we can form the baseline of understanding the synthesis of geospatial data at these sites.
Before software and hardware are developed, enhanced, or further implemented, we must develop an understanding of the currently available data, characterize how that data is represented, and know how protected area personnel use those data. It is with this baseline knowledge that new development can advance the capabilities of protected areas to counter illicit activity within their boundaries. The real power behind technologies used for poaching mitigation is how practitioners make sense of these data, particularly when they are synthesized with human-generated geospatial data to understand patterns and trends in human and endangered species movement. The implementation of similar sense making strategies across protected areas to make sense of geospatial data may support improved sharing between protected areas.
We see an expansion in research opportunities that broaden our understanding of geospatial data synthesis for poaching mitigation. For instance, the next steps could further explore practitioners' use of geospatial data for poaching mitigation. We expect to see new research fill knowledge gaps on the intersection of human and animal movement patterns from synthesized data, and the evaluation of current tools used by protected areas to synthesize and visualize these data.