The hunter and the hunted: Using web‐sourced imagery to monitor leopard (Panthera pardus pardus) trophy hunting

Abstract

Sustainable offtake of any threatened species and objective monitoring thereof relies on data‐driven and well‐managed harvest quotas and permit compliance. We used web‐sourced images of African leopard (Panthera pardus pardus) trophy hunts to determine whether online photographs could assist in monitoring and documenting trophy hunting in Africa. Of 10,000 images examined, 808 (8%) showed leopard trophy hunts and could be contextualized by date and country. From a subset of photos (n = 530), across six countries between 2011 and 2020, we extracted information on the leopards killed and hunter demographics. We found no significant differences in leopard sex, age, or shot wound position between countries, and most trophy leopards were in good physical condition. Most hunters were White (96%) and estimated at over 40 years old (82%), with the proportion of women hunters in younger age classes significantly higher than in older classes. Rifles, bows, and hounds were used in all countries, except Tanzania and Zambia, where rifles were exclusively used. Online images could not be reasonably compared to the CITES trade database, but in South Africa, more than half (57%) of all nationally registered leopard trophy hunts in the last decade (2010–2020) have been posted online. Online images also reveal hunting violations, including non‐permitted hunting of female leopards and illegal hounding. Such monitoring methods may become increasingly useful as social media usage grows and provide valuable insight into this multi‐million dollar industry.

INTRODUCTION

Effective biodiversity conservation, including adaptive monitoring and sustainable harvest, necessitates multiple tools. Trophy hunting applied as such a tool is controversial (Leader-Williams, 2001; Lindsey et al., 2007; Saayman et al., 2018). Understanding the relative conservation costs and benefits of this activity is particularly challenging without comprehensive monitoring data on harvested populations and the assurance of ethical and transparent quota compliance (Bunnefeld et al., 2013; Di Minin et al., 2021; Wanger et al., 2017). Trophy hunting is a highly lucrative industry (Wilkie & Carpenter, 1999), grossing over USD 326.5 million per year throughout Botswana, Ethiopia, Mozambique, Namibia, South Africa, Tanzania, Zambia, and Zimbabwe, and forms a crucial revenue stream for many developing countries (Lindsey et al., 2007; Safari Club International Foundation, 2015).

The African leopard (Panthera pardus pardus; hereafter ‘leopard’) is listed as “Vulnerable” on the IUCN Red List (Stein et al., 2020) and has been eradicated from 63% to 75% of its historical range (Jacobson et al., 2016). Leopards are a particularly desirable big game trophy species, contributing 8%–20% of total national trophy hunting revenue in East and southern Africa (Balme et al., 2010, 2012; Braczkowski et al., 2015; Stein et al., 2020). The species is listed on Appendix I (i.e., a species threatened with extinction and whose trade is permitted only in exceptional circumstances, provided both export and import permits are issued) of the Convention for the International Trade in Endangered Species (CITES). The export of leopard trophies from certain African countries is thus permitted, following CITES-approved quotas for each country (Balme et al., 2012; CMS/CITES, 2018). Leopards are also heavily harvested for use in traditional practices (Harries, 1993; Naude et al., 2020; Williams et al., 2017), and are persecuted in retaliatory conflict due to their real and perceived threat to livestock (Loveridge et al., 2010; Naude et al., 2020). Under these circumstances, trophy hunting may promote tolerance toward leopards and other large carnivores and reduce poaching (Di Minin et al., 2021). Furthermore, the income generated from, and the land allocated to, leopard trophy hunting may improve conservation efforts and aid in the recovery of declining populations (Palazy et al., 2011; Swanepoel et al., 2014). Tolerance may therefore be crucial to their survival in unprotected areas (Power et al., 2021), where growing human population densities may lead to increased human–wildlife conflict (Croes et al., 2011; Cusack et al., 2021; Redpath et al., 2013; Swanepoel et al., 2014). Trophy hunting and its economic benefits might also then serve as an incentive for both the maintenance of trophy species and the protection of their habitats (i.e., as both umbrella and sentinel species), making hunters proponents of conservation (Palazy et al., 2011).

On the other hand, ethical and welfare considerations regarding the killing of animals for recreation and trophies, as well as issues relating to social disruption, artificial selection of particular traits, localized population declines, unsustainable off-takes, and violations of permitted activities have all been pointed to as problematic by trophy hunting opponents. These issues and associated controversies can seriously compromise the potential role trophy hunting could play as an effective tool for biodiversity conservation (Leader-Williams, 2001; Lindsey et al., 2007; Saayman et al., 2018). These issues are amplified due to incomplete demographic, ecological, and genetic monitoring data on harvested populations and concerns around ethical and transparent quota compliance (Bunnefeld et al., 2013; Di Minin et al., 2021; Wanger et al., 2017; Williams et al., 2021). Opportunities for sizeable financial gain may foster corruption within the industry (Lindsey et al., 2007; Palazy et al., 2011), impacting decisions on land allocation for hunting concessions and leading to quotas being exceeded, potentially driving the overexploitation of targeted species (Lindsey et al., 2007; Palazy et al., 2011). Specific quotas are therefore set to ensure the persistence of leopard populations, but limited capacity and diversity of monitoring techniques often constrain their success in ensuring sustainability (Lindsey et al., 2007).

Quotas often fail to account for human–leopard conflict (Swanepoel et al., 2014), and are often violated (Palazy et al., 2011; Trouwborst et al., 2020), which contributes significantly to the decline in leopard populations (Swanepoel et al., 2014). Quota violations can also have severe repercussions within the industry. For instance, the poor reporting compliance by some South African hunters and outfitters (i.e., licensed businesses that employ guides to take care of hunters during hunting expeditions) in 2015 resulted in a year-long leopard hunting moratorium in 2016, which was extended to 2017 (DEA, 2017; CMS/CITES, 2018). If quotas are not well regulated, regularly re-evaluated, and strictly enforced, the effective conservation of leopards and the integrity of the trophy hunting industry are at risk (Balme et al., 2010; Milner et al., 2007; Swanepoel et al., 2014). However, responsibly determined data-driven quotas allow for sustainable harvest and may contribute to species and habitat conservation in the long-term, where leopard conservation also conserves other species that share the same ecosystems and are indicators of ecosystemic functionality (Baker, 1997; Di Minin et al., 2016, 2021; Lindsey et al., 2007).

Leopard trophy hunting regulations in South Africa require that hunted leopards are males and older than 7 years of age (DFFE, 2017). In theory, this ensures that females contribute reproductively throughout their lives. In contrast, harvested males should have already reached their reproductive peak, and their removal should have minimal impact on the local population (Balme et al., 2012). This approach requires professional hunters to accurately sex and age leopards, which can be achieved visually by examining overall leopard size, ear condition, facial scarring, and dewlap size (Balme et al., 2012). Therefore, these regulations featuring age and sex restrictions require extensive, rigorous, and costly monitoring to ensure compliance (Balme et al., 2012; Swanepoel et al., 2016).

Social media hav the potential to become a useful tool for such environmental monitoring, especially with its projected increase in user volume and public accessibility of data (Jarić et al., 2020; Monkman et al., 2018). Such data contain various integrated digital signatures, allowing relevant information to be easily filtered and collected via automated “scraping” platforms (Di Minin et al., 2019, 2021; Tsou, 2015). Several studies have demonstrated the value of web-sourced photographs in tackling ecological and evolutionary questions. For example, Leighton et al. (2016) showed that Google Images can be used, in a relatively unbiased way, to describe the spatial distribution of animal phenotypes for a variety of taxa. The authors also developed a web application, Morphic (https://morphs.io), which simplifies the extraction of information from Google Image searches. The use of online images has since been adapted to various novel ecological applications (e.g., Atsumi & Koizumi, 2017; Bahlai & Landis, 2016; Berryman & Kirwan, 2021; Dylewski et al., 2017; Maritz & Maritz, 2020; Naude et al., 2019; Panter & Amar, 2021) and to monitor illegal wildlife trade (Sonricker-Hansen et al., 2012; Di Minin et al., 2019; Xu et al., 2019; Panter & White, 2020). However, the method has limitations and challenges (Di Minin et al., 2019; Jarić et al., 2020); privacy policies and terms of use required by various platforms need to be considered, and data may prove unreliable (Monkman et al., 2018; Di Minin et al., 2019, 2021). Hence, technical control and data validation steps have been established to record reliable and relevant information through these platforms (Leighton et al., 2016).

Many trophy hunters upload “posed” images of their hunts online and these images are generally accessible under most Creative Commons (CC) licenses. Acknowledging that the method only captures a snapshot of the full extent of the trophy hunting of leopards, as not all hunters will post pictures of their trophy online and that hunters might not post pictures of hunted leopards that they know fall outside regulations, the potential of this monitoring technique has merit and has yet to be tested. Moreover, a standardized technique of leopard sexing and aging from images has already been established (Balme et al., 2012). Thus, an objective online monitoring technique may prove to be a useful and cost-effective tool for statutory authorities, professional hunting associations and conservation organizations in assessing various aspects of the trophy hunting industry.

In this study, we use the internet application Morphic (Leighton et al., 2016) to collate information from photographs posted online, of hunters posing with leopard trophies. Our aim was to explore the efficacy of image-based online monitoring for leopard trophy hunting across Africa. We contextualize the value of this approach relative to current international trade monitoring and hunting records (i.e., the CITES trade database and available national professional hunting registries) and expand on the value of this image-based technique by exploring leopard trophy and hunter characteristics and identifying potential permitting violations.

METHODS Data extraction from photographs

During September–December 2020, information on individual hunters posing with leopard trophies was collated from online photographs scraped from Google Images using the Morphic web application (Leighton et al., 2016). Morphic automatically removes duplicate images through a perpetual hashing algorithm (Farid, 2021; Niu & Jiao, 2008) and uses Google.com rather than any of its national variants to avoid possible biases based on the browser location through Google's Hummingbird relevance algorithm (Chauntelle & Yazdanifard, 2014). “Photographs” were specified in the search to filter out irrelevant images (e.g., paintings and graphic interchange format files). Unique, non-exhaustive search queries were created to obtain as many relevant online images (OIs) relating to leopard trophy hunting as possible (Table S1). Search combinations included terms directly relating to the sport (e.g., safari, dangerous game, and hounds) and identifiers of potential social media platforms on which such images could be posted (e.g., Facebook, Instagram, and Twitter). All web pages were manually examined to identify the original image source and to remove multiple photographs of the same hunting event.

Photographs were classified as “relevant” if they contained a hunter posing with a leopard trophy and “usable” if the date of the hunt and country-level location of the image could be determined (Table S2). Four experienced leopard researchers then independently examined the resulting images (Table S3). Where available, the following data were collected from each usable photograph: leopard sex (male or female); estimated leopard age (adult, juvenile, or sub-adult) following Balme et al. (2012); estimated leopard condition (i.e., a 5-point Likert scale: very poor, poor, fair, good, or very good); shot wound position (head, neck, shoulder, chest, or abdomen); estimated hunter gender (man or woman); estimated hunter age class (<20, 21–30, 31–40, 41–50, or >50 years); estimated hunter race (Asian, Black, Hispanic, White, or Other); visible or documented type of weaponry and hunting hounds used; hunter nationality; and professional hunting outfitter (Table S4). Acknowledging the subjectivity of such estimates, final data classification from each usable photograph required consensus between independent assessors. Photographs were reverse image searched using either TinEye Reverse Image Search (https://tineye.com) or Google Search by Image (https://images.google.com) to obtain information for the original image. The date and location of each hunting event were determined via the associated webpage or extracted from image captions. If this was not readily available, the metadata embedded within the image was derived from the Exchangeable Image File Format (EXIF) using Metapicz (https://metapicz.com) to retrieve any outstanding information.

Trade database comparisons

As there is no formally recognized and globally standardized record of trophy hunting, we do not know the true number of leopards hunted and traded as trophies in southern Africa, and therefore cannot directly compare the objective veracity of OI monitoring. Instead, we rely on the international CITES trade database (CTD) to list the maximum number of trophies that could be legally exported by party countries (assuming that these are disclosed and that permit applications are correctly reported) and on national-level professional hunting registries (PHRs; when available and comprehensive), to estimate the number of hunts per annum.

We obtained the national export quotas (NEQ) for leopard trophies from CITES (https://cites.org/eng/resources/quotas/export_quotas) and the record of the number of leopard trophies people applied to export from the CTD (https://trade.cites.org) for countries with extracted images. As only South Africa has a national PHR, our regional comparative estimate of how many registered leopard trophy hunts occurred during the period for which OI-based monitoring data could be extracted was limited.

As CITES permit records only began in 1975, we excluded usable images occurring before this date and before the relevant country joined and entered CITES “in force” (Table 1). Hence, we extracted export quotas and the number of registered permits from each country with usable images from 1980 to 2020. CTD search parameters were set as follows: year range: 1980–2020; exporting countries: all countries identified in the OIs of leopard trophy hunting were selected, namely Ethiopia (ET), the Central African Republic (CF), the Democratic Republic of the Congo (CD), Kenya (KE), the United Republic of Tanzania (TZ), Zambia (ZM), Mozambique (MZ), Zimbabwe (ZW), Namibia (NA), Botswana (BW), and South Africa (ZA); importing countries: all countries were considered; source: both “wild” (W) and “unknown” (U) source codes were considered; purpose: set at “hunting trophy” (H) purpose code only; trade terms: set to “trophies” (TRO) and “body” (BOD); and taxon: “Panthera pardus”.

TABLE 1 An analysis-based subsetting of the 10,000 images of individual hunters posing with leopard trophies in the wild scraped from Google images through the Morphic web application (Leighton et al., 2016) using 25 unique search terms (Table S1) during September–December 2020

Country	Relevant	Usable	Comparison	Characteristics
	1920–2020	1920–2020	1980–2020	2011–2020
Not images of leopard trophy hunts	9192 (92%)	-	-	-
Unidentified date or country	-	55 (7%)	-	-
Pre-1975 CITES trade database (CTD)	-	-	72 (10%)	-
Pre-2011 or no image-derived data	-	-	-	151 (22%)
Botswana (BW) [1977]a	-	8 (1%)	8 (1%)	-
Central African Republic (CF) [1980]a	-	10 (1%)	10 (1%)	-
Democratic Republic of the Congo (CD) [1976]a	-	2 (<1%)	2 (<1%)	-
Ethiopia (ET) [1989]a	-	1 (<1%)	1 (<1%)	-
India (IN) [1976]a	-	4 (<1%)	3 (<1%)	-
Kenya (KE) [1978]a	-	1 (<1%)	1 (<1%)	-
Mozambique (MZ) [1981]	-	66 (8%)	51 (7%)	47 (7%)
Namibia (NA) [1990]	-	207 (26%)	186 (25%)	148 (22%)
Republic of South Africa (ZA) [1975]	-	184 (23%)	173 (23%)	136 (20%)
United Republic of Tanzania (TZ) [1979]	-	64 (8%)	60 (8%)	55 (8%)
Zambia (ZM) [1980]	-	30 (4%)	27 (4%)	23 (3%)
Zimbabwe (ZW) [1981]	-	176 (22%)	159 (21%)	121 (18%)
Total	808 (8%)	753 (93%)	681 (90%)	530 (78%)

Notes: Indicated are the number of “relevant” images of African leopard (Panthera pardus pardus) trophy hunts retrieved (n = 808, 8% of 10,000 scraped images), the number of “usable” images for which hunt date and country could be determined (n = 753, 93% of 808 relevant images; Table S2), the number of images for hunts occurring post 1975 and therefore comparable to the available permitting and trade databases (n = 681, 90% of usable images), and the number of images of hunts occurring post 2011 for which there is sufficient image-derived data to determine leopard trophy and hunter characteristics (n = 530, 78% of permitting and trade database comparable images; Table S4). Data are summarized by overall count and proportion (%) of the previous subset (left) after removing images per criterion (italicized). Listed also are the years in which each relevant country joined and entered CITES “in force”.

^aCountries with limited data.

The CTD is often mischaracterized (Challender et al., 2021), as its purpose is not to formally monitor and thereby potentially enforce trade compliance among its party members, instead, it is an extensive record (i.e., >38,000 species) of standardized information collected when a person applies for a permit to trade internationally in endangered species. As the CTD is not curated for such quantified monitoring purposes, many transactions only contain data within the “importer reported quantity” and “exporter reported quantity”, making use of these inconsistencies and any interpretation of these data challenging (Chan et al., 2021). For instance, a reliance on raw ‘exporter reported quantity’ is not recommended as it tends to lead to an underestimation of the sum of quantities being traded. Moreover, there is an administrative lag in reporting; hence, trophy export permit years recorded in the CTD may not match the time (at least by year) of the hunt itself. Thus, the CTD cannot provide the total number of leopard hunts per country per annum, but gives a maximum figure for the number of leopard trophies (i.e., listed on permits) that people applied to have exported or imported (noting that there may be more animals listed on those permits than were actually exported and imported, as it is common practice to apply for a permit listing more individuals than are intended for trade). Following the methods established by Vall-llosera and Cassey (2017), Martin (2018), and Chan et al. (2021), we “corrected” for these missing data by using the higher and removing the lower of these two records to account for potential duplication and calculated the maximum number of traded trophies (i.e., where singular body parts were batched by minimum number of possible whole leopard trophies per application) per transaction (Table S5). Where the importer country was listed as “XX” (i.e., unknown; n = 2), these records were removed from further analyses and no relevant geopolitical changes occurred within the studied time period. The resulting database (Table S6) provided the national export quota (NEQ) for leopard trophies and a record of the number of leopard trophies people applied to export from the CTD (i.e., as both a maximum permitted for export and a minimum for the total hunted) for all relevant (i.e., with extracted OIs) countries between 1980 and 2020.

As few countries successfully maintain national PHRs to corroborate national hunting records, it remains challenging to control for these limitations and provide an accurate estimate of leopard trophy hunts per country by year. However, the South African Department of Forestry, Fisheries, and the Environment (DFFE) has collated and maintained a national PHR since 2010, which contains verified provincial-level records of leopard trophy hunts conducted and reported by professional hunters across South Africa (Table S5). In this limited context, we assess the efficacy of image-based online monitoring and consider the relative value of the CTD and PHR (in South Africa) for leopard trophy hunting across Africa (Figure S1).

Statistical analyses

All statistical analyses were run using R statistical software (R Core Team, 2020 [v. 4.0.3]). The yield of New and Relevant Images (NRI) per search effort was determined as a function of the count, and the cumulative count of NRI per total number of searches (n = 25) assessed (Table S1; Figure S2). Online image, CTD, and PHR trophy hunts were calculated and presented by country. We then explored various differences in leopard trophy and hunter characteristics between countries by fitting multinomial models using the “multinom” function in the “nnet” package (Venables & Ripley, 2002). The response variables included leopard sex, estimated leopard age, estimated leopard condition (i.e., on a 5-point Likert Scale), shot wound position, hunter gender, estimated hunter age class, hunter race, and visible or documented use of weaponry, and hunting hounds (Table S3), with focal country as the explanatory variable. Additionally, both hunter gender by estimated age class and OI representation of CTD and PHR were assessed per decade. The overall significance of these factors was determined using Type III ANOVAs in the “car” package (Fox & Weisberg, 2011). The relative effect of each variable of interest (i.e., leopard or hunter characteristic and online representation) by country, hunter gender, and age class was plotted using the “effects” package (Fox & Hong, 2009). Tukey post-hoc analyses were then conducted using the “lsmeans” package (Lenth, 2016) to test for pairwise differences in the proportion of each comparison by country.

RESULTS Online photographs

Of the 10,000 photographs extracted from the 25 unique searches submitted to Morphic (Table S1), 808 featured individual hunters posing with leopard trophies (8%) from 1920 to 2020, and were thus classified as “relevant” (Table 1). The cumulative number of New and Relevant Images (NRIs) per additional search term showed that, while the total number of relevant images continued to increase (Figure S2), NRI yield per effort attenuated significantly (Pearson correlation, R² = 0.37, p < .05) across the 25 searches (from ${\overline{x}}_{\mathrm{NRL}}$ = 53 ± 26 [SE] per 2000 images [46%] in the first five search terms to ${\overline{x}}_{\mathrm{NRL}}$ = 26 ± 9 in the last five [8%]).

There was also an increase in the number of NRIs post-2004, corresponding with the commercialization of digital cameras and technological advances in cellular communications, as well as burgeoning social media applications (Figure 1). Of the 808 relevant images, we were able to extract information on hunt date and country for 753 (93%) images spanning over a century (1920–2020) of leopard trophy hunts across 12 countries (Table S2). Of those usable images, 681 (90%) showed hunts occurring post-1975, allowing for comparison with the CTD and PHR (Figure S1), 530 (78%) of which occurred between 2011 and 2020 and provided both hunter and leopard image-derived data that were used in the subsequent analyses (Table S3).

FIGURE 1. Accumulation curve of new and relevant images (Acc. NRI; black line and gray circles) from the 10,000 images of individual hunters posing with African leopard (Panthera pardus pardus) trophies in the wild scraped from Google images through the Morphic web application (Leighton et al., 2016) using 25 unique search terms (Table S1) during September–December 2020 between 1920 and 2020. Here Acc. NRI is shown relative to significant technological advances and commercialization of digital photography, contextualized by the global user growth and spread of prominent social media platforms (various colored lines and circles; our world in data 2019)

Trade and hunting database comparisons

The 681 usable images came from Ethiopia (ET), the Central African Republic (CF), the Democratic Republic of the Congo (CD), Kenya (KE), the United Republic of Tanzania (TZ), Zambia (ZM), Mozambique (MZ), Zimbabwe (ZW), Namibia (NA), Botswana (BW), and South Africa (ZA) between 1980 and 2020. There were no records of CITES permitted leopard trophy exports for India (IN), the Democratic Republic of the Congo (CD), and Kenya (KE) during this period. Therefore, we excluded these countries from further analyses. The number of leopard trophies listed annually on the CITES permits applied for (i.e., listed on the CTD) approximated the official NEQ for leopard trophies across these nine remaining countries (Figure S1), while the South African PHR recorded 36% (n = 254/707) of leopard trophy hunting permits applied for in the CTD between 2011 and 2020 (Table S6).

Online database comparison

OI, CTD and PHR (for ZA) records were attained for nine relevant countries annually between 1980 and 2020 (Table S6). Overall, OI proportional coverage of the South African PHR was 57% (n = 145/254), rising non-significantly (χ² = 1.846, df = 2, p = .215) from 52% (n = 33/64) between 2001 and 2010 to 72% (n = 136/190) between 2011 and 2020 (Figure 2).

FIGURE 2. Online image (OI) coverage of individual hunters posing with African leopard (Panthera pardus pardus) trophies in the wild relative to the CITES trade database (CTD) and record of hunts in the national professional hunting register (PHR) records (where available) overall and per decade (1981–1990; 1991–2000; 2001–2010; 2011–2020) for all relevant countries (including the United Republic of Tanzania [TZ]; Zambia [ZM]; Mozambique [MZ]; Zimbabwe [ZW]; Namibia [NA]; and Republic of South Africa [ZA]; Table S5)

As the overall annual number of New and Relevant Images (NRI) was relatively low before 2009 (Table S6) and was particularly low for ET (n = 5), the CA (n = 26), and BW (n = 25), further analyses focused only on images from the remaining six countries (TZ, ZM, MZ, ZW, NA, and ZA) in the southern and east African regions from 2011 to 2020 (Figure 3). Despite there being no records meeting our criteria in the CTD for IN, CD, and KE during this period, online records were reported for IN (n = 2) and CD (n = 2). Estimated OI, CTD, and PHR proportions varied considerably during this time (Figure 3).

FIGURE 3. The number of leopards listed on permits applied for that are documented on the CITES trade database (CTD; green) and record of hunts in the national professional hunting registry (PHR; gold) records (where available) for African leopard (Panthera pardus pardus) and online images (blue) of individual hunters posing with leopard trophies in the wild (Table S3). Where no chart is given, there is no CTD or PHR record for this decade, but there is online evidence (OI). Gray shading indicates countries for which there is no (white) or limited (light gray) data and those with sufficient data for image-based leopard- and hunter-based characteristic analyses (dark)

Image-derived leopard information

Of the 530 usable images, within the six focal countries, from which leopard trophy characteristics could be derived between 2011 and 2020 (Table S4), 99% (n = 412/418) were classified as male, with only six females (1%) hunted in ZW, ZM, and ZA (Figure 4a). The sex of 112 (21%) leopards could not be determined, and given the small proportion of females, there were unlikely to be any significant differences in the leopard sex ratios hunted between countries. We were able to age most leopards (n = 495), with only 35 (7%) leopards not aged confidently. From these aged individuals, 95% were classified as adults (≥7 years old), while only nine were sub-adults (2%; in TZ, ZW, NA, and ZA; Figure 4b), with no significant difference in hunted leopard age class found between countries (χ² = 2.95, df = 5, p = .708). We were able to assess the condition of 388 (73%) leopards, whereas we were unable to do so for 142 leopards, and most of these were considered to be in “good” condition (Figure 4c), with no leopards estimated to be in “poor” or “very poor” condition. Leopard condition differed significantly between countries (χ² = 20.13, df = 10, p = .028), with a significantly greater proportion of leopards characterized as being in “very good” rather than ‘good’ condition in ZA compared to NA (F = 6.012, df = 12, p = .015). Shot wound positions were difficult to identify from photographs and could only be determined for 198 (37%) leopards, with most occuring on the abdomen (61%) or shoulder (23%), but this did not differ significantly (χ² = 21.08, df = 20, p = .392) between countries (Figure 4d).

FIGURE 4. Proportional effects of African leopard (Panthera pardus pardus; a–d) and hunter (e–h) characteristics derived from ‘usable’ online images of individual hunters posing with trophies in the wild (n = 530; Table S5) overall and per country (United Republic of Tanzania [TZ]; Zambia [ZM]; Mozambique [MZ]; Zimbabwe [ZW]; Namibia [NA]; and Republic of South Africa [ZA]) from 2011 to 2020. Indicated are leopard sex (a); estimated leopard age (b) following Balme et al. (2012); estimated leopard condition (c); shot wound position (d); estimated hunter gender (e); estimated hunter age class (f); estimated hunter race (g); visible or documented type of weaponry and hunting hounds used (h; Table S4)

Image-derived hunter information

From the 530 usable images from which hunter characteristics within the six focal countries could be derived between 2011 and 2020 (Table S4), 6% (n = 30/530) of images either did not include a hunter or hunter gender could not be determined from the image. Of the remaining 94% (n = 500/530), only 6% (n = 31/500) of hunters were women and 94% (n = 469/500) men (Figure 4e). Regardless, there were no significant differences in hunter gender between the six countries (χ² = 3.72, df = 5, p = .590). Most hunters (77%; n = 348/423) were within the upper age classes (41–50 and >50 years), with few hunters estimated at 30 years old or younger (Figure 4f). There were no significant differences in the proportion of hunters in each age class between countries (χ² = 25.07, df = 20, p = .200). However, there were significant differences in gender proportion within age class (χ² = 37.58, df = 4, p < .001), with more women in the lower age classes (<20 or 21–30 years) than in older age classes (Figure 5). Hunter race could be determined in 508 (96%) of the 530 images, 488 (96%) of which were White (Figure 4g) and there was no significant difference in pairwise hunter race proportions by country (χ² = 13.60, df = 10, p = .1919). Hunter nationality could only be derived in 14 (3%) images (Table S4) and was therefore excluded from further analyses. Additionally, whether hunting hounds were used was also insufficient to analyze alone (Table S4) and was incorporated within the hunting method (Figure 4h), which showed a significant difference in the pairwise proportion of hunting methods by country (χ² = 46.99, df = 15, p < .001). Overall, there were 48 (9%) instances of leopard hunting with hounds in MZ, ZW, NA, and ZA. Finally, hunting outfitter information was obtained from 449 (84%) images, and 167 individual outfitters were identified (Table S4). Of these, one outfitter conducted the most recorded hunts (4%; n = 20), and the top 25 outfitters accounted for over 42% (n = 225/447) of all hunts captured by online photographs (Figure S3).

FIGURE 5. Proportional effect of hunter gender by estimated age class derived from ‘usable’ online images of individual hunters posing with African leopard (Panthera pardus pardus) trophies in the wild (Table S4) overall and per estimated age class for all relevant countries (including the United Republic of Tanzania [TZ]; Zambia [ZM]; Mozambique [MZ]; Zimbabwe [ZW]; Namibia [NA]; and Republic of South Africa [ZA]) from 2011 to 2020 (Table S5)

DISCUSSION

Online images of hunters posing with leopard trophies as a means of independently and remotely monitoring national hunting quota compliance could not, ostensibly, be tested against the CITES trade records of India, Ethiopia, the Central African Republic, the Democratic Republic of the Congo, Kenya, Tanzania, Zambia, Mozambique, Zimbabwe, Namibia, Botswana, and South Africa over the last 40 years (1980–2020). This is primarily because there are, to our knowledge, currently no formal or traditional methods of monitoring the trade in hunting trophies in southern Africa beyond the national professional hunting registry of South Africa and the CTD. The latter of which has many shortcomings (i.e., fundamentally, it is a record of permit applications and not actual exports) and is often misrepresented as a trade monitoring tool (Challender et al., 2021; Chan et al., 2021); instead what it monitors, is what people wanted to export when they applied for a permit. Outside of the PHR of South Africa, there is no way of knowing what leopard trophy hunting and trade actually occurred, except that actual legal import numbers will be less than the numbers indicated on the CTD (Bertola et al., 2022; Williams et al., 2021). This finding is particularly revealing as it suggests that overapplication for CITES permits, the administrative lag in reporting, and ultimately the fundamental application record, not trade monitoring purpose of the CTD limits the use of this international trade repository in this context.

The national leopard hunting registry of South Africa is the only country, as far as we are aware, for which a reliable hunting registry exists and is available to the public for direct comparison; here, online monitoring captured 136 out of 190 recorded leopards trophy hunts (72%) between 2011 and 2020. Whilst online monitoring may be limited in its ability to monitor the total number of hunts throughout Africa, such an approach still revealed potential hunting violations and also provided unique insight into leopard trophy and hunter demographics throughout the region. Online image representation has also improved five-fold (7%–35%) over the last decade (2001–2010 vs. 2011–2020) and is likely to continue improving rapidly, given the rise in social media use globally (Monkman et al., 2018).

While it remains difficult to monitor the number of leopards hunted with online images, and indeed with most other monitoring data currently available, we are however able to effectively asses compliance relating to leopard and hunter characteristics. We report three cases in which female leopards were illegally hunted: two in South Africa (DEA, 2018) in 2018 and 2019 and one in Zambia (MTA/DNPW, 2018) in 2020. We also detected the hunting of leopards under 7 years of age in South Africa in 2014 and 2016. However, these occurred before the enforcement of the age-based hunting regulations in 2017 (DEA, 2017) and the latter was a violation of the national leopard hunting moratorium of 2016 (CMS/CITES, 2018). We were unable to sex 23% and age 8% of the leopards due to poor picture quality or the individual's position in the picture. Thus, we may have underestimated the number of hunting violations. Most leopard trophies were in good condition and shot wounds positioned either in the abdomen (61%) or shoulder (23%). Although some outfitters and hunting associations provide their own recommendations or guidelines on the location where the shot should be placed on the body (i.e., the heart or lung area), we could not find any internationally recognized guidelines based on a behavioral and physiological scientific assessment of pain (e.g., as in live-trapping of wildlife) against which such practices could be compared. In South Africa, professional hunters are required to take a once-off assessment to determine their competency in correctly judging the age of male leopards and show that they are familiar with legislation (DFFE, 2017). While we did discover violations, which suggests that either this training is inadequate or there is a lack of enforcement, these images were of female leopard hunts, not age-based violations. We did not find leopard trophies estimated younger than 7 years old being hunted in South Africa after the age-based regulation was implemented in that country (DFFE, 2017), suggesting that this new regulation might have been respected and enforced. Nevertheless, the violations identified could have ramifications on local leopard populations, as female leopards not only have greater reproductive value than males (Balme et al., 2012), but also have smaller home ranges than males, making it more challenging for hunted individuals to be replaced in an area (Braczkowski et al., 2015; Naude et al., 2020).

We also discovered the illegal use of hounds to hunt leopards in Namibia (CITES, 2018) and South Africa (DEA, 2018; NEMBA, 2007). In addition, our search tool revealed images of leopards that may have been killed during the moratoria on their hunting in Zambia (2013 and 2015) and South Africa (2016 and 2017). Quotas were set at zero to aid in the recovery of declining populations and to implement better management strategies (DEA, 2018; MTA/DNPW, 2018). Such violations could impede these national conservation efforts and result in extensions of moratoria, further jeopardizing the potential role of trophy hunting as a conservation tool, and impacting communities that have become financially dependent on the tourism income generated by trophy hunting (Di Minin et al., 2016; Mbaiwa, 2018). When financial benefits in protecting these large mammals are lost, it may result in increased human-wildlife conflict (Mkono, 2018), as evidenced by the moratoria in Zambia (MTA/DNPW, 2018). Such instances of non-compliance should be of interest to statutory authorities and conservation organizations, as well as the trophy hunting industry, as it may directly jeopardize the reputation and viability of the industry (Lindsey et al., 2007). Thus, regulation compliance ensures sustainability for both felid populations and the hunting industry (Loveridge et al., 2007; Packer et al., 2011).

We showed that leopard trophy hunters tend to be older (>50 years) White men, which is consistent with existing literature: women tend to show less interest in trophy hunting than men (Adams & Steen, 1997; Andersen et al., 2014). Furthermore, older men are more supportive of trophy hunting (Byrd et al., 2017) and are more likely to share images of their trophies online (Darimont et al., 2017). However, “Cybermovements” (Mkono, 2018) against trophy hunting, such as the one after the illegal killing of “Cecil the lion”, could discourage hunters and outfitters from posting their hunting photos online, and therefore, reduce the efficacy of our method, even with an expected increase in social media use (Monkman et al., 2018).

There is limited research on women in hunting. However, our study revealed the arrival of young (<20 and 21–30 years) women hunters into the industry. This demographic shift is relatively recent but has been recorded for sport hunting in North America and Europe (Hansen et al., 2012; Mcfarlane et al., 2003; Metcalf et al., 2015). Mcfarlane et al. (2003) reported that this was often due to women entering hunting license lotteries to increase the chances of their male partners receiving a license. However, the online images reflect the reality of women hunter representation, as they are seen posing with their trophies. Though poorly studied, some of the possible explanations for the increase in relatively younger, women hunters include their desire to provide a healthier, local and tastier source of food to their family (i.e., when hunting wild game), the sense of purpose, independence and empowerment that comes from possessing the skills to hunt, the opportunity to connect more closely to nature, family, and friends, and the need to challenge themselves (Houtman, 2020; Metcalf et al., 2015). Thus, the strength of online imagery could be in investigating social trends in the industry and examining the typology of women hunters (Metcalf et al., 2015). The hunting industry uses typologies to create tailored programs that meet the needs of their recreationists (Schroeder et al., 2006) and develop plans to recruit and retain women or other social strata (Gigliotti & Metcalf, 2016). Conservation organizations and government agencies should consider this new audience when making informed decisions about management communication, policy, conservation messaging, and training materials (Jones et al., 2019).

The efficacy of online monitoring methods is dependent on the accuracy and reliability of the information obtained from online photographs. This is challenging and subjective, as image quality is often inadequate, or information is unreliable or inaccessible, limiting the amount of extractable data for research purposes. Additionally, an accurate record of the year the hunt took place is vital to compare the number of permit applications recorded by CITES and those captured by the images. These dates are often embedded in the image metadata but can be easily manipulated, potentially making them unreliable (Wijayanto et al., 2016). Occasionally, professional hunters from different outfitters collaborate on hunts, making it difficult to trace where the hunt was performed, as photographs may appear on both websites. Duplicate photographs, or duplicate individual leopards in separate photographs, also constrain the number of new trophy hunts discovered in images, impacting the value of this potential monitoring tool. Trophy hunts posted online may also be missed because Morphic cannot access images posted to private groups on social media (Leighton et al., 2016). Furthermore, we only used English search terms, potentially restricting the number of usable images obtained, particularly throughout west and central Africa. Most importantly, this study assumes that (i) leopard trophies are posted online, (ii) leopard trophies are actually traded and traded across international borders, and (iii) hunts are disclosed and reported to relevant authorities. Therefore, as not all hunters post pictures with their trophy online (i.e., out of preference or experience in both hunting and social media usage), this method only captures a snapshot of the full extent of the exportation and trade in leopard trophies. As the CTD compiles trade records for specimens of CITES-listed species traded across international borders, it does not account for the majority of specimens that are traded within local markets (e.g., shot on a South African game reserve and sold to a South African buyer) or for specimens that are not traded at all and remain in-country or there are possibly instances where a non-hunting overseas customers wanted a trophy hunted leopard skin that a taxidermist could export to them, for instance. This may also explain why there were no CITES records for India, the Democratic Republic of the Congo or Kenya between 1980 and 2020. Alternatively, this may be related to the ratification date of CITES parties to the Convention (i.e., certain countries may not be members of CITES during the studied time periods, joining in more recent years, and therefore will not contribute any trade data within these periods). There is also no doubt that leopards, like many other game species, are hunted illegally and these hunts go unreported to relevant authorities; therefore, these web-sourced imagery samples would not contain data for these hunts. For instance, in Zimbabwe, excess skins from illegal hunts are reportedly smuggled into Zambia and Mozambique to be exported under the quotas of those countries (Balme et al., 2010).

This study determined whether systematically scraped web-sourced images would reflect the number of hunts conducted in southern Africa and Tanzania. Unfortunately, the CTD is not an appropriate baseline with which to compare OI-based monitoring and few countries manage a national PHR with which to compare these findings. Records of endorsed CITES export permits (i.e., a record of the actual number of leopard trophies exported) and verified national professional hunting records could provide better insight into the industry, quota compliance, and determine whether quotas should be adjusted, and in turn, improve the efficacy of our online monitoring method, by providing a more appropriate baseline to which independent monitoring techniques can be compared (Robinson & Sinovas, 2018). Furthermore, it may aid in providing information on hunted wildlife populations by providing hunting effort and yield trends, which could be used to assess the sustainability of harvested populations and the industry (Palazy et al., 2011). If accurate country-level data are maintained by local conservation authorities, such as in South Africa, where more than half of all registered (PHR) leopard trophy hunts in the last decade (2010–2020) have been posted online, our method could be used to help monitor compliance with national legislation at a fraction of current cost and effort (provided trophy images continue to be posted online). Online monitoring could initially be achieved with relatively limited resources (i.e., intern program access for image screening, internet access, computer literacy, and basic database management) and scientific input (i.e., data validation and legislative violation), and once refined, much of this process may become automated. Such applications should not come at the expense of traditional policing and permitting, but should rather provide a complimentary assessment to support ongoing enforcement efforts. Given the lack of formalized methods and complexity of trophy hunt and trade monitoring, such alternative techniques are going to become increasingly valuable and perhaps in some circumstances, a more economical option than current methods or where reliable records are non-existent. It could also expand upon leopard trophy and hunter characteristics that may be of value to statutory authorities, conservation agencies and the professional hunting industry (Gigliotti & Metcalf, 2016).

Using online images to monitor leopard trophy hunting is not a method that can reliably be used to inform sustainable quotas, since photographs collected online will not provide an accurate estimate of the total number of leopard trophy hunts conducted or provide ecological data such as population estimates or range distribution (Trouwborst et al., 2020). However, this method can aid in the detection of hunting violations and deliver a typology of trophy hunters and their changing demographics in a simple, cost-effective manner over a large spatial area (Schroeder et al., 2006). These data are crucial for managing trophy hunting globally and provide insight into potential conservation and welfare concerns for harvested species (Mkono, 2018). The ability to extract useful information from online images enables the application of this method to a multitude of species and systems, increasingly contributing to a field termed “iEcology” (Jarić et al., 2020). For example, Hansen et al. (2012) developed an online surveillance system to monitor the illegal wildlife trade. Combining our method with their tools may provide information about trader demographics and which platforms are frequently used for trade. Although the online images could not accurately estimate the full extent of leopard trophy hunts, this method has been successfully applied in other ecological and conservation studies (e.g., Amar et al., 2019; Atsumi & Koizumi, 2017; Bahlai & Landis, 2016; Berryman & Kirwan, 2021; Dylewski et al., 2017; Maritz & Maritz, 2020; Naude et al., 2019; Panter & Amar, 2021; Panter & White, 2020; Xu et al., 2019). Furthermore, the increasing use of social media provides promising advocacy for the future use of this method as a rapid and increasingly effective monitoring tool.

AUTHOR CONTRIBUTIONS

Jessica R. Muller: Data collection, analysis, manuscript drafting, editing, and reviewing. Sarah-Anne Jeanetta Selier: Data collection, manuscript editing and reviewing. Marine Drouilly: Data collection, manuscript editing and reviewing. Joleen Broadfield: Data collection, manuscript editing and reviewing. Gabriella R.M. Leighton: Conceptualization, study design, manuscript editing and reviewing. Arjun Amar: Manuscript editing and reviewing. Vincent N. Naude: Conceptualization, study design, data collection, analysis, manuscript drafting, editing, and reviewing.

ACKNOWLEDGMENTS

This work was funded by the Robert Niven Trust and the University of Cape Town Postgraduate Funding Office. We would also like to acknowledge G. Groenewald for assistance with statistics and Z. Zondi for clarification with EXIF. We especially thank three independent reviewers for their support and insightful commentary, ultimately contributing to a stronger study overall.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT

All data are available in an online UCT-based Figshare repository at: https://doi.org/10.25375/uct.13726915

SUPPORTING INFORMATION

Additional supporting information may be found in the online version of the article at the publisher's website.