Jaguars (Panthera onca) are a landscape species of conservation importance and our understanding of environmental and anthropogenic drivers of jaguar occurrence is necessary to improve conservation strategies. We reviewed available literature to simply describe environmental and anthropogenic variables used and found to be significant in occurrence modeling. We reviewed 95 documents published from 1980 to 2021 that focused on jaguar occurrence and that used 39 variable types (21 anthropogenic, 18 environmental) among different techniques, scales, and approaches. In general, these variables included both anthropogenic (roads, land use, human activities, and population) and environmental (climate, vegetation, ecological interactions, topographic, water, and others) factors. Twelve variables were identified as affecting jaguar occurrence overall, eleven at local scale and seven at broad scales (regional and continental). Focusing more specifically on the variables that correlate with occurrence should help researchers to make better predictions in areas without quantitative jaguar data.

habitat, humans, landscapes, Panthera onca, prey, roads habitat, humans, landscapes, Panthera onca, prey, roads

Understanding drivers of species distribution under global change scenarios, whether directly anthropogenic or indirectly climatic, is crucial for the development of nature conservation strategies (Kareiva and Marvier 2015). More specifically, resource abundance should determine potential species abundance and distribution, and direct mortality factors should constrain species occurrence (Manly et al. 2002). Thus, documenting and understanding how environmental and anthropogenic factors allow/limit species distribution is essential for species conservation (Morrison et al. 2006; Paschoaletto and De Barros Ferraz 2012).

There are different approaches and interpretations to estimating species distribution, and selection of proper state variables that have causal effects may influence inferences over time and space (Mackenzie and Nichols 2004; Mackenzie et al. 2017). Similarly, the ways in which data are collected (methods and techniques), scale considerations (from fine to broad scale), and statistical approaches (analysis and extrapolation techniques) also may influence findings and predictions. Nevertheless, there often is a lack of systematic classification of common environmental and anthropogenic factors related to species modeling approaches.

Populations of jaguars (Panthera onca), the largest felid on the American continent (Seymour 1989), have been gradually extirpated (Ceballos et al. 2005; Ripple et al. 2015) and now occur in only 54% of their historic geographic range (Sanderson et al. 2002). The species is classified as “Near threatened” (Quigley et al. 2017) and previous jaguar population assessments at the continental scale also show a continued rate of decline (Rabinowitz and Zeller 2010; De la Torre et al. 2017) due to trophy hunting, killing as retaliation by livestock predation, habitat loss, human expansion, and poaching of prey (Quigley et al. 2017). Jaguars are landscape species with large home ranges extending beyond protected areas and across a variety of ecosystems under a gradient of anthropogenic pressures (Silver et al. 2004; Thornton et al. 2016). As apex predators they functionally maintain ecosystem balance and structure, regulating populations at lower trophic levels to more stable states (Estes et al. 2011; MacBride and Thompson 2018; Thompson et al. 2020). Studying free-ranging jaguars can be logistically demanding and expensive due to their elusive behavior, large home range sizes, and low population densities, often in places that are isolated and difficult to access (Salom et al. 2007; Carrillo et al. 2009). As a result, jaguar occurrence across the Americas is fairly well known (Rabinowitz and Nottingham 1986; Rabinowitz and Zeller 2010), but questions about their more precise occurrence trends at different scales, and particularly factors directly influencing distribution, are still relevant as research priorities (Sanderson et al. 2002; De la Torre et al. 2017; Sanderson et al. 2022).

Here we summarize and examine the most-used anthropogenic and environmental predictor variables, and modeling and data collection approaches, cited in peer reviewed literature that best described jaguar occurrence. If a relatively small number of factors were consistently identified as important, then the need for additional such studies would be less. The outcome of this assessment should allow the reconsideration of meaningful (and meaningless) predictor variables in future modeling of jaguar occurrence, and thus make the future application of model results more useful and successful.

A comprehensive literature review of factors influencing jaguar distribution was conducted using two Internet search engines (Web of Science and Google Scholar). A systematic search was temporally delimited from 1980 to 2021 and used the following combination of words: “Jaguar” + “Distribution” + “Environmental variables” + “Prey abundance” + “Panthera” + “Occurrence”. For each publication identified as relevant, we identified the methods of analysis used to inform jaguar occurrence, the geographic scale of the assessment, and a list of variables included in the assessment. We sort the data gathering methods into seven categories: telemetry, camera trap, genetics, historic records, sign counts, interviews, and data derived from geographic information systems (GIS). The analysis methods were separated into four categories: Occupancy (specific modeling approach base on the proportion of areas or sample units occupied), Niche modeling (species-distribution models that used presence data to infer ecological requirements to elucidate potential distributions), deductive (base on previous knowledge and species-environment associations from expert opinion to envisage distribution), and basic statistical empirical approaches (included a number of methods such as comparison tests, generalized lineal models, and analysis of covariance). We also classified the range extent of each study into four scale categories: continental, regional, country, and local. Similar variables with different names were classified into one-name variables, and these were subsequently sorted into sub-categories within the broader categories of anthropogenic and environmental factors in a Microsoft Excel spreadsheet (Suppl. materials 1, 2).

Once we collected the entire range of predictor variables of jaguar occurrence, we identified those that were reported as having a statistically significant influence on occurrence/distribution. We assessed which were most identified, and the degree to which these were related to geographic scale.

We identified 165 peer reviewed documents in our search, but only 95 either tackled issues of jaguar distribution or correlated distribution with anthropogenic or environmental factors. Among these studies we found that the number of jaguar distribution studies recently has increased, with almost 98% of the literature being published after 2000 (Fig. 1A). Most studies took place in Brazil (n = 28), Mexico (n= 16), and Belize (n = 10; Fig. 1B).

Figure 1. 

A Annual number of peer-reviewed documents, and B country(region)-specific number of peer reviewed documents assessing distribution of jaguars.

Among the studies there were four main modeling approaches (Table 1). The most widely used were basic statistic empirical models (n = 41; Suppl. material 1) which usually analyze or describe summaries of empirical data based on correlation among variables (Morrison et al. 2006). Niche or presence-only models (n = 22) rely on occurrence records together with environmental variables to represent the ecological-niche of a species (Phillips et al. 2017). Occupancy models (n = 20) use detection/no-detection records with a set of different covariates combination to choose the best models that explain species occupancy across the sites (Mackenzie et al. 2017). Deductive approaches (n = 12) rely on previous knowledge of species-habitat relationships based on literature or expert opinion (Morrison et al. 2006).

A variety of research techniques used to gather data for assessments of jaguar distribution (Table 1; Suppl. material 1). Data from camera trapping was used most often (n = 38), but historic records (n = 21), telemetry studies (n = 16), and data derived from geographic information systems (GIS; n = 12) were also commonly relied on. Data from sign counts, interviews and genetic studies were used less often.

There also were multiple geographic scales used in modeling efforts (Table 1). Most were local or study area-specific (n = 62), but a number of papers assessed jaguar distribution at continental (n = 13), regional (n = 12), or country (n = 8) scales.

Our summation of different qualitative and quantitative variables types used to model jaguar distribution identified a total of 39, including 21 classified as anthropogenic and 18 as environmental (Table 2). The anthropogenic variables were sorted into four subcategories: road, land use, human activities, and population. Environmental variables were sorted into six subcategories: climatic, vegetation, wildlife, topographic, water, and other.

Early jaguar distribution research was limited by available techniques and technologies, making it difficult to understand important influential variables. With the development of techniques such as camera trapping in India for tigers (Panthera tigris) (Karanth 1995), its use for informing jaguar distribution in the Americas (Silver et al. 2004) increased. Reliable and satellite telemetry equipment furthered research capacity (e.g., Morato et al. 2016). Also, the development of higher computer hardware capacity led to increasingly sophisticated analysis techniques, such as deductive GIS modeling (Sanderson et al. 2002; Zárrate-Charry et al. 2018; Craighead et al. 2019), occupancy modeling (Mackenzie et al. 2017), and niche modeling (Phillips et al. 2017), that has accelerated the efficiency with which jaguar data of various kinds have been used to provide insights into jaguar distribution.

Occurrence model reliability likely is affected by scale, survey technique used, and the anthropogenic and environmental metrics available to be included (Boydston and Gonzàles 2005; Torres et al. 2008; Bitetti et al. 2010; Sollmann 2011; Sollmann 2012; De la Torre et al. 2017; Gese et al. 2018). Most of the studies we surveyed were conducted at a local scale and utilized precise data mostly from camera trap surveys (Michalski et al. 2015; Watkins et al. 2015; Fort 2016; Jordan et al. 2016; Astete et al. 2017a). Nevertheless, local-scale camera-trap modeling studies, for example, may sometimes have scale mismatch issues because they only have available coarse, countrywide geographical layers to apply to ecological processes evaluated at fine scale (e.g., Quiñones et al. 2018); this is a common issue across modeling approaches independent of particular taxa (MacGarigal et al. 2016). Also, camera traps do not capture location information when jaguars or other species are resting, but those places might also be important determinants of distribution and density.

Relevant evidence of road-based metrics affecting jaguar distribution were observed in a few studies (Colchero et al. 2010; Torres et al. 2012; Zeilhofer et al. 2014; Borrego 2015; Dueñas-López et al. 2015; DeMatteo et al. 2017), presumably as consequence of better access routes that result in increased poaching (Sanderson et al. 2002; Petracca 2010). Distance to roads was a common metric in reviewed documents, perhaps because this variable can be easily built with any basic GIS (geographic information system) software (DeMatteo et al. 2017; Gese et al. 2018), but when included it most often was not identified as a significant variable.

Land use metrics should reflect both exposure to negative human interactions and a limitation of prey resources (Junior et al. 2013; Cuyckens et al. 2017). Land cover was identified as a significant metric in many, but not a majority, of studies in which it was assessed, but showed discrepancies in terms of pixel resolution across the studies (Zeller and Rabinowitz 2011; Cuervo-Robayo and Monroy-Vilchis 2012; Cullen et al. 2013; Morato et al. 2014; Torres 2021). Though additional exploratory correlation of urban development and jaguar density was not the most significant, this may be because most jaguar distribution studies used national or global land cover layers due to the high expenses incurred getting higher resolution data at local scale (Hansen et al. 2013).

Human activities may affect jaguar presence or biological processes as a result of anthropogenic recreation or economic activities, including poaching or hunting of jaguars and their prey (Sandoval et al. 2011; Jordan et al. 2016; Jędrzejewski et al. 2017; Silva et al. 2018; Portugal et al. 2019). The metrics of distance to protected areas and level of protection were significant in only half of the studies where they were assessed, and though these two metrics can be easily built, they do not always reflect the intensity and efficiency in law enforcement which we assume to contribute importantly to wildlife occurrence. Also, hunting pressure was identified as significant in 2 or 3 studies, and though this makes clear sense, it is a variable that is difficult to adequately measure and map.

Metrics identified in the population subcategory such as human population density and distance to settlements were sometimes identified as significant, perhaps magnifying the importance of other factors assessed but also indicating that jaguars can co-exist adjacent to areas where people, and perhaps particularly livestock owners, live (Jędrzejewski et al. 2018).

Environmental variables were widely used and mostly described biotic and abiotic factors essential for species subsistence (Ashcroft et al. 2011; Luja et al. 2017; Thompson et al. 2021). Within the subgroups of variables, it seems likely that some metrics are autocorrelated. For example, the climate group variables of seasonality, precipitation, and temperature were all significant in some studies, but seasonality is influenced by the interaction of precipitation and temperature, where high temperatures and low precipitation increase droughts that may also increase mortality because when a drought comes, it also diminishes available food (Sirot et al. 2016). Given the jaguar’s wide geographic distribution encompassing many different biome types, it is not surprising that their occurrence is not strictly dictated by a specific climate regime, and inclusion of such variables in local occurrence modeling may not be necessary.

Vegetation variables were the most used across jaguar studies (Sanderson et al. 2002; Weckel et al. 2006), vegetation type being significant in most. Vegetation type may represent refuge (similar to a forested land cover metric, or an inverse to anthropogenic land cover), a source of prey, and stalking or hunting habitat (Zeilhofer et al. 2014; Booker 2016; Olsoy et al. 2016; Dobbins et al. 2017; Souza et al. 2017).

Wildlife interactions, when they can be identified and mapped, are both common and highly significant factors influencing jaguar distribution. Prey occurrence and abundance are important to jaguars not only because of their high demand relative to other mammals (Sunquist and Sunquist 2002; Guilder et al. 2015), but also because prey has such an influence on carnivore demography (Fuller and Sievert 2001). In places with high prey availability jaguar density is positively correlated as López and Miller (2002) hypothesized. Both prey and competing predator distribution and abundance are often simultaneously collected using camera traps, and are thus both available and reasonable metrics to include in models (Weckel et al. 2006; Azevedo and Murray 2007; Harmsen et al. 2009; Davis et al. 2010; Petracca 2010; Harmsen et al. 2011; Rodríguez-Soto et al. 2011; Petracca et al. 2013; Arroyo-Arce et al. 2014; Gutiérrez-González and López-González 2017; Dallas and Hastings 2018; De la Torre and Rivero 2019).

Topographic variables may affect hunting opportunities (Kruuk 2006), but more likely they are also correlated with other variables such as distribution of humans, protected areas, and land/vegetation cover that are more directly correlated with factors affecting jaguar distribution. Still, elevation may be widely used by researchers who can easily get this information without advanced training in geographic information technologies.

Even though some carnivores can partially fulfill their nutritional water requirements with prey, hunting places near water could increase predator encounters, especially in seasonal environments (Sirot et al. 2016). Distance to water is a commonly used metric, likely also because researchers can easily get this information without advanced training. Though we did not find evidence suggesting fresh water as driver of jaguar density at local scale, we hypothesize in seasonal ecosystems water might be related to prey and therefore to higher jaguar densities.

Distance to beach was identified once as a significant variable in a place where nesting sea turtles are seasonally abundant, and thus a variable reflecting peaks of prey availability (Carrillo et al. 2009). This suggests that modeling considerations should take into account the special circumstances of the study site.

Many variables were not identified as significant, though it seems like they could be important constrainers of jaguar distribution. It is likely that the metrics assessed are constrained by a variety of issues, including the types of variables available (Jędrzejewski et al. 2017; Silva et al. 2018; Anile et al. 2020) or the lack of ease to build them (Colchero et al. 2010; Petracca 2010). Also, variables cannot always be based on or derived for specific effects for which human activities or environmental conditions limit or enhance jaguar presence. Finally, some assessments are constrained by the kinds and/or amounts of data used in modeling. Sample sizes may limit, for example, assessment of sex-, age-, or behavior-specific influences on distribution.

Elucidating jaguar occurrence within its range with a limited number of meaningful predictors is not an easy task. Although jaguars can persist in a wide variety of ecosystems, if the set of variables selected are not causally related to jaguar density at a local scale, occurrence data likely does not directly infer density or population trends, hence species occurrence does not mean that jaguar populations are thriving. Because even the simple use of photo rates of jaguars does not seem to correlate well with jaguar density (Maffei et al. 2011), more focus on monitoring factors directly influencing density is warranted.

Thoughtful assessment of variables potentially affecting jaguar distribution should direct researchers to better identify and then quantify specific casual factors affecting jaguar distribution, rather than simply describe it, especially in terms of jaguar reproduction, survival, and dispersal. Habitat descriptors are useful in understanding a species’ niche (Hutchinson 1957), and habitat quality is often inferred from the distribution of species (McLoughlin et al. 2010). Habitat use patterns may provide a link to population dynamics (Bernal-Escobar et al. 2015; Boyce et al. 2016; Morato et al. 2019), but such links have not been well identified for jaguars. So, even though linking demographic rates to habitat use is logistically and financially challenging, doing so will provide the demonstrated relationships that are needed to best conserve jaguar populations into the future. Jaguar habitat modeling provides a plethora of hypotheses to test, and demographic data will unveil the mechanisms providing for jaguar population viability.

We thank the Universidad Nacional de Costa Rica and the Ministerio de Ciencia y Tecnología, for partial support of this study, as well as the Katie Adamson Conservation Fund.