Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2026 Apr 13;40(5):e70283. doi: 10.1111/cobi.70283

Monitoring wildlife health for diseases with visible signs by integrating camera traps with marked individuals

Jonathan Tichon 1,✉, Roi Lapid 2, Roni King 3, Elyasaf Freiman 1, Orr Spiegel 4, Erez Baruchi 3, Aviam Atar 3, Shirli Bar‐David 1, David Saltz 1,3
PMCID: PMC13613544  PMID: 41972464

Abstract

Monitoring wildlife health is essential for conservation and management, wildlife and livestock welfare, and public health in a One Health framework. Yet, wildlife health monitoring often requires long‐term fieldwork and intensive sampling, which can be costly or logistically challenging, especially for remote, rare, or elusive populations. To address these challenges, we evaluated the feasibility and utility of integrating camera traps with a marked subset of Nubian ibex (Capra nubiana) to monitor five diseases that produce visible signs. Over six consecutive years, we deployed 51 camera traps and monitored 43 uniquely identifiable individuals (25 males and 18 females). We estimated apparent prevalence annually and modeled temporal trends with generalized additive models, which allowed flexible nonlinear relationships between disease prevalence and time. To assess recurrence dynamics and mortality risk, we fitted joint frailty models that simultaneously estimated rates of recurrent disease events and survival while accounting for individual heterogeneity. Apparent prevalence ranged from 2% to 38% depending on disease and year. For the most prevalent disease, caseous lymphadenitis (CLA), prevalence was higher in males (β = 2.08, p = 0.041) and did not vary across years. Despite small sample size (n = 9), joint frailty models yielded epidemiologically meaningful inference. Recurrence of CLA was positively associated with mortality (association parameter α = 2.32, p = 0.012), and male sex was a significant prognostic factor for terminal events (hazard ratio = 0.07, p = 0.019). Our proof of concept demonstrates that combining camera traps with marked individuals can provide a practical auxiliary or stand‐alone tool for wildlife health monitoring. This approach is particularly suited to rare, elusive, or logistically inaccessible species and will gain further strength once its diagnostic specificity and sensitivity are validated.

Keywords: Capra nubiana, epidemiology, frailty, joint frailty models, Nubian ibex, One Health, WHM, Cabra montés del Nilo, Capra nubiana, epidemiología, fragilidad, modelos de fragilidad conjunta, One Health, WHM

INTRODUCTION

Monitoring wildlife health is a crucial aspect of conservation and management programs and a direct objective of public health within a One Health framework (Barroso et al., 2023). Nevertheless, wildlife health monitoring (WHM) often requires longitudinal fieldwork and intensive sampling procedures, typically associated with high costs, especially when populations inhabit remote or inaccessible areas or consist of rare and elusive species. Accordingly, development of additional methods for monitoring wildlife health with demonstrated feasibility and utility is essential (Barroso et al., 2024; Ryser‐Degiorgis, 2013).

Camera traps are exceptional tools for wildlife research and monitoring. They provide versatile functions at relatively low costs and are easy to use (Moore et al., 2023). Consequently, their use has increased substantially in recent years (Wearn & Glover‐Kapfer, 2019), including for detection of specific diseases, such as sarcoptic mange in Iberian wolves (Canis lupus signatus) (Olaega et al., 2011) and in coyotes (Canis latrans) (Murray et al., 2021); contagious ecthyma (ORFV) in Bezoar ibex (Capra aegagrus) (Gündoğdu et al., 2018); leprosy in chimpanzees (Pan troglodytes) (Hockings et al., 2021); and giraffe skin disease in giraffes (Giraffa camelopardalis) (Muneza et al., 2019). Researchers have used camera traps to detect the presence of a focal disease (Gündoğdu et al., 2018; Hockings et al., 2021; Olaega et al., 2011), quantify its occurrence with occupancy models (Murray et al., 2021), and determine the severity of the infection (Muneza et al., 2019). To the best of our knowledge, however, no one has used camera traps as a tool for routine WHM (i.e., to scan simultaneously for a suite of diseases, as opposed to one focal disease).

Although it is possible to obtain an index of disease prevalence in a population from unmarked individuals with, for example, camera traps, such a camera‐based index would not be robust because repeated detections of unmarked individuals in the camera frame cannot be distinguished (i.e., sampling with replacement). Likewise, unmarked individuals do not allow monitoring of recovery progression because they cannot be uniquely identified, which prevents follow‐up. In contrast, when a sufficiently large subset of the population is individually marked (artificially or naturally) and constitutes a representative sample (by virtue of sampling procedure), then prevalence of diseases in the general population should be able to be robustly estimated based on that marked subset. This is the method we propose for WHM with camera traps and marked individuals.

Longitudinal monitoring of marked individuals enables recording of disease onset, progression, recovery, recurrence, and, under certain conditions, terminal events. Because longitudinal camera trap monitoring of marked individuals yields such data, joint frailty models that account simultaneously for these processes (Rondeau & Gonzalez, 2005) provide a powerful analytical complement to our proposed method. Frailty models are regressions for survival data that incorporate a random effect (i.e., frailty) that represents unmeasured covariates and exerts a multiplicative effect on the baseline hazard function (Balan & Putter, 2020). Hazard functions describe, among other things, the distribution of the time to event. As a result, the frailty provides (through its multiplicative effect) a unique value for the hazard function per individual (when recurrent events are grouped into a cluster of observations of the same individual) or among related individuals (when they are grouped into a cluster). Consequently, frailty accounts for unobserved heterogeneity across individuals or clusters while inducing dependency among observations in each cluster. Through the integration of covariates, the model simultaneously accounts for observed heterogeneity (Balan & Putter, 2020).

By extension of the frailty models concept, the joint frailty models for recurrent events and a terminal event combine two regression models (hence joint): one evaluates overall survival, and the other evaluates gap times between consecutive recurring events (e.g., disease recurrence or relapse) (Huang & Liu, 2007). A series of recurrent disease events (especially when serious) often correlates with a subsequent mortality event that terminates the observation series, thereby censoring the gap time process and inducing inherent dependencies between recurrent and terminal events (Huang & Liu, 2007). Joint frailty models simultaneously account for these events and their dependencies by treating the terminal event as informative censoring and estimating the distributions of survival and gap times along with the covariate effects that influence them (Balan & Putter, 2020). These models share a frailty term that links the hazard functions of the survival and gap time processes. Consequently, this framework enables simultaneous modeling of two survival processes, which allows examination of the joint interaction between recurrent events (e.g., disease relapses), a subsequent terminal event, and the effects of covariates over time (Balan & Putter, 2020).

We aimed to evaluate the feasibility and utility of integrating camera trap with a marked subset of individuals to monitor diseases with visible signs in wildlife populations and to assess the compatibility of joint frailty models as an analytical complement to this method. To that end, we used a dataset compiled from six consecutive years of camera trap dataset from an ongoing monitoring project of a Nubian ibex (Capra nubiana) (hereafter ibex) population in the Judean Desert, Israel. The dataset included camera trap footage of uniquely marked individuals and the unmarked population. Our objectives were to monitor multiple diseases with visible signs simultaneously; provide robust estimates of apparent disease prevalence and temporal trends inferred from the marked subset; and evaluate the applicability of joint frailty models to identify key covariates affecting epidemiological and demographic processes, even with small sample sizes (as is often the case for threatened species). Once validated as a proof of concept, this method could strengthen existing WHM programs that currently rely solely on biological samples, such as tissue, blood, and fluids (Ryser‐Degiorgis, 2013).

METHODS

Study area

The Judean Desert in Israel is a hyper‐arid rain‐shadow desert, with mean summer (June–August) maximal temperatures of ∼40°C and mean annual rainfall of ∼90 mm (Dayan & Morin, 2006). The study area was in the northeastern mountainous region of the Judean Desert and consisted of a linear cliff‐line formation, locally known as Matzok Ha'Etekim (Figure 1). Along this cliff line were 10 focal perennial water sources (natural and artificial), mostly arranged below the cliff line, and three perennial flowing streams. These were the only sources of drinking water for wildlife during the harsh dry period (May–August). The study area encompassed Matzok Ha'Etekim Nature Reserve (31.66°N, 35.42°E) and En‐Gedi Nature Reserve (31.46°N, 35.39°E).

FIGURE 1.

FIGURE 1

Open in a new tab

Study area in the north Judean Desert, Israel, where the health of the Nubian ibex (Capra nubiana) population was monitored from 2017 to 2022.

Focal species, capturing, marking, and monitored diseases

The Nubian ibex is a sexually dimorphic, gregarious, mountain ungulate, exhibiting sexual segregation outside of the rut season (September–December) (Aronson, 1982; Greenberg‐Cohen et al., 1994; Gross et al., 1995; Tichon et al., 2022, 2024). The ibex is an obligate drinker (Shkolnik, 1981), requiring water every few days (Aronson, 1982; Habibi, 1994) and restricting its movements to the vicinity of cliffs (Hochman & Kotler, 2007), which serve as refuge (Iribarren & Kotler, 2012). Accordingly, cliffs and water sources delimit the spatial distribution of ibex (Tichon et al., 2024).

All handling procedures for live animals received legal approval and were performed by authorized veterinarians of the Israel Nature and Parks Authority (INPA) (permit number 2017/41747). From 2016 to 2019, we darted ibex in the north Judean Desert, targeting adult individuals (i.e., males ≥6 years old and females ≥1.5) and tagging them with unique combinations of color‐ and shape‐coded ear tags (Leader Large Aussie tags; Bayer AG). We darted from blinds positioned on all focal perennial water sources and from moving vehicles along Road 90 (Figure 1), thus covering the entire study area. We designed the capturing effort to be evenly distributed spatiotemporally and between the sexes, so the marked subset would constitute a representative sample of the population.

Because no research has been done to date on visually identifiable diseases in Nubian ibex, we turned to the closely related Iberian ibex (Capra pyrenaica), whose diseases were reviewed (Valldeperes et al., 2023) and include sarcoptic mange via Sarcoptes scabiei, ORFV via Orf virus, infectious keratoconjunctivitis (IKC) via Mycoplasma conjunctivae (Aguilar et al., 2019), caseous lymphadenitis (CLA) via Corynebacterium pseudotuberculosis, lumpy jaw (LJW) via Actinomyces spp. or other pathogens or trauma to the jaw that cause mandibular osteomyelitis (Fagan et al., 2005), and abortions, which are often visually identifiable by bloody stains on the breech wool and hips from serosanguineous vulvar discharges (Underwood et al., 2015) and frequently caused by abortion‐inducing pathogens (e.g., Chlamydia abortus and Coxiella burnetii [Q‐fever]). Detection of signs, such as bloody stains on the breech wool and hips, is likely infeasible with camera traps during nighttime because they use infrared lighting that does not allow differentiation of colors. Accordingly, we monitored all aforementioned diseases in the marked subset (excluding sarcoptic mange, which has never been documented in wild ibex in Israel [R.L. and R.K., unpublished data]).

Data collection, diagnosis, and categorizing events

To monitor disease onset, progression, recovery, and recurrence in the marked subset, we deployed 51 camera traps (Trophy Cam 119774C/119876C; Bushnell) in an array that covered the entire study area (Figure 1). Because ibex are obligate drinkers, it would have sufficed to deploy cameras only on the perennial water sources in the study area during the dry season (as previously proposed for bighorn ship [Ovis canadensis] by Sanchez et al. [2022]). However, camera coverage of the entire study area enabled disease monitoring throughout the year (not limited to the dry season) and was applicable regardless of the presence or absence of perennial water sources (similar to mark–resight surveys, e.g., Bengsen et al., 2022).

We retrieved data weekly from cameras positioned in locations used intensively by wildlife (e.g., water sources) and monthly from cameras in less frequented locations (e.g., on cliff‐spanning trails). In each visit to the cameras, we replaced SD cards and batteries prior to depletion. The data were sorted into folders per camera (indicating spatial location), year, and date of the data accumulated on the SD cards. Any photo of marked ibex with some physical anomaly (e.g., injuries, disease signs, broken limbs) was stored separately for later inspection by wildlife veterinarians that provided the best available visual diagnosis (under the inherent limitations of the method). Because marked individuals (unlike unmarked) allow for robust estimates of disease apparent prevalence, trends, and individual follow‐up, we focused solely on their footage. For each marked individual with visual diagnosis, the disease name and date of onset were recorded. Once this individual reappeared in camera footage without those signs, a recovery was recorded. If the same disease signs reappeared in the same individual later, it was recorded (in addition to disease onset) as a recurrence.

We consider a special case in which uniquely marked individuals belong to a species exhibiting range fidelity and stable home ranges are being monitored with several camera traps located within each home range. These cameras regularly capture footage of each individual longitudinally. In such cases, the absence of sightings of marked individuals for an extended period (e.g., 1 year) in all cameras in their home ranges is a strong indication of mortality or emigration. Accordingly, under similar circumstances, camera traps can be considered reasonable tools for mortality inference (although not definitive) once emigration is ruled out.

We inferred mortality from camera traps based on the rationale described above. This inference was further warranted by the characteristics of our study system. First, ibex exhibit strong philopatry to core areas within their home ranges year‐round (except for males during the rut), demonstrating range fidelity and stable home ranges (Aronson, 1982; Tichon et al., 2024). Second, the perennial water sources in the study area constitute the only available drinking water during the dry season. Because all focal perennial water sources were covered by camera traps throughout the study, the probability of undetected survival was negligible. Marked ibex might go undetected due to mortality or emigration (from the study area); however, no emigration events were observed during the study period. Third, the study area was at the edge of the natural ibex distribution in Israel. In this region, undetected dispersal in three of the four cardinal directions would be unlikely (north and west have no ibex habitat, east is the Dead Sea), and the passage southward was covered with cameras. Therefore, the absence of a marked individual from all camera traps throughout an entire dry season (and particularly throughout a full year) was considered indicative of mortality. Accordingly, the month of an individual's last camera trap detection was recorded as the time of its terminal event. Marked individuals whose carcasses were directly observed (e.g., via GPS or VHF collar tracking, or opportunistically) were likewise recorded as terminal events. We estimated the timing of these events based on camera trap footage, lack of movement indicated by GPS collars, VHF mortality signals, or carcass decomposition state.

Assessing temporal trends and sex‐specific differences

To make biologically meaningful interpretations, we analyzed the data to determine whether annual trends in disease prevalence and sex‐specific differences were present. We calculated annual disease prevalence in the marked subset as the proportion of infected, marked individuals relative to the total number of living marked ibex (per year and per sex) (Table 1). Because we anticipated small sample sizes (as is common for threatened and elusive species), we estimated prevalence with 95% confidence intervals based on Wilson score interval, which is well suited for small sample (Wilson, 1928). Because nonlinear temporal patterns (e.g., year effects or seasonality) could influence prevalence trends over time and between sexes, we fitted a generalized additive model (GAM) in the R package mgcv (Wood, 2011, 2017). From Table 1, we constructed an aggregated dataset suitable for analysis of sex‐specific prevalence, with one row per sex per year. Each row contained the number of infected marked ibex (by sex) and the total number of marked ibex alive that year. GAMs allow flexible regression modeling and identification of nonlinear relationships between outcomes and explanatory variables by estimating smoothing parameters for selected predictors. This approach yields smooth curves well suited for ecological and epidemiological processes in which trends evolve nonlinearly over time (Hastie & Tibshirani, 1986; Wood, 2025). Specifically, the model included sex as a fixed effect and a smoothing term for year interacting with sex, thereby producing one smooth curve per sex to evaluate longitudinal differences between males and females.

TABLE 1.

Occurrences and outcome of diseases with visible signs in a subset of marked ibex individuals.

Disease onset (O) and recovery (R)
Individual name Sex 2017 2018 2019 2020 2021 2022
Ezer M CLA (10 Aug) O Dead Dead Dead Dead Dead
Golda F SAB (02 Apr) O SAB (20 Mar) O

CLA (27 Aug) O

CLA (20 Nov) R

CLA (11 Jan) O

CLA (05 Oct) R

 None None
Herzel M None None IKC (27 Oct) O

IKC (05 Apr) R

(scar remain)

IKC

(scar remain)

IKC

(scar remain)
Kokhba M CLA (22 Feb) O

CLA (01 Jun) R

CLA (12 Aug) O

 Dead Dead Dead Dead
Mendelson M None None N one None CLA (25 Sep) O CLA (01 Apr) R
Mordi M N/A N/A None

CLA (07 Jun) O

CLA (12 Jul) R

 Dead Dead
Peres M IKC (scar remain) IKC (scar remain) Dead Dead Dead Dead
Ruth F N/A None None None

CLA (21 Jun) O

CLA (28 Jul) R

 None
Shimi M N/A N/A None None CLA (24 Apr) O

CLA (18 May) R

ORFV (15 Jan) O
Shoshi F SAB (12 Dec) O None Dead Dead Dead Dead
Tuvya M N/A CLA (10 Jun) O Dead Dead Dead Dead
Yadin M N/A N/A None

CLA (13 May) O

LJW (19 Jun) O

LJW (08 Sep) R

LJW (08 Jun) O

LJW (09 Aug) R

CLA (16 Sep) R

 Dead
Total living marked individuals per year and sex 18 (11 M, 7 F) 33 (17 M, 16 F) 35 (18 M, 17 F) 26 (11 M, 15 F) 23 (10 M, 13 F) 21 (9 M, 12 F)
Open in a new tab

Note: Subsequent onset categorized as recurrence. Date of the event (O, R, or dead) in parentheses. Residual tissue scarring can have physiological implications.

Abbreviations: CLA, caseous lymphadenitis (abscess); IKC, infectious keratoconjunctivitis (eye); LJW, lumpy jaw; N/A, individuals not marked; ORFV, contagious ecthyma (mouth); SAB, suspected abortion caused by disease (e.g., Coxiella burnetii [Q‐fever] or Chlamydia abortus).

We fitted the GAM by maximizing a penalized likelihood for the smooth terms. For a given set of smoothing parameters, model coefficients were estimated using penalized iteratively reweighted least squares (IRLS), whereas smoothing parameters were selected by restricted maximum likelihood (REML) criterion, the recommended approach for smoothing parameter estimation in GAMs (Wood, 2017, 2025). We used thin plate regression splines bases (mgcv package's default) because they require no manual knot placement, are well suited for small sample sizes, avoid arbitrary design choices (thus reducing bias), and provide low‐biased estimates of the underlying smooth function (Wood, 2017). The basis dimension (k) was set to five for each sex‐specific smooth. This parameter sets an upper limit on the potential flexibility of the smooth (actual flexibility is determined by the smoothing parameters) and should not exceed what is justified by the number of unique covariate values (Wood et al., 2016; Wood, 2017). With 6 years of data per sex, setting k = 5 (i.e., n − 1  degrees of freedom) was appropriate. The response followed a binomial distribution with a logit link.

To evaluate whether repeated observations of the same ibex violated the independence assumption of the GAM, we also fitted a GAM with a random effect for individual (GAM_re) as a sensitivity analysis (for the individuals with observed disease occurrences). The GAM_re requires Bernoulli individual‐per‐year observations (to model within‐individual dependencies) with one row per ibex per year. Each row indicates (binary) whether the ibex showed signs of one of the five monitored diseases that year. This disaggregated data structure is suitable for assessing pseudoreplication but less suitable for estimating annual disease prevalence. For the GAM_re, we used the same fixed effects and smooth structure as in the GAM, with the addition of an individual‐level random effect. Likewise, for the GAM_re, we used a binomial distribution with a logit link, REML for smoothing parameter estimation, thin plate regression spline bases, and a basis dimension of k = 5 for each sex‐specific smooth. For parsimony, we assessed whether the more complex random‐effect model was justified for final inference by conducting model selection with the Akaike information criterion corrected for small sample size (AICc) (Burnham & Anderson, 2004). If the random‐effect variance in the GAM_re approached zero, this indicated that between‐individual heterogeneity was negligible given the available data and that repeated observations did not introduce meaningful pseudoreplication. If, additionally, model selection indicated that the GAM_re was not the leading model, this would further suggest that the more complex model was unnecessary for inference and presentation.

Joint frailty models for recurrent events and a terminal event

We used event data from the marked individuals, including disease onset, recovery, recurrence, and terminal events, to fit joint frailty models for recurrent events and a terminal event via maximum penalized likelihood estimation (Rondeau et al., 2003, Rondeau 2007). We recorded the data with a gap timescale (i.e., after each recurrent event, time to the next event restarted from 0). For the analysis, we used the R package frailtypack (Rondeau & Gonzalez, 2005; Rondeau et al., 2012) following the methodology described in Rondeau et al. (2012) and Król et al. (2017). We clustered the observations of recurrent events at the individual level (rather than group level). Joint frailty models account for dependencies among recurrent events of each individual (with each individual treated as a cluster) and for dependence between recurrent events and the terminal event (Rondeau et al., 2012). In essence, for each of the five monitored diseases, a separate joint frailty model was needed because the models account for recurrences (i.e., relapses of the same disease per individual) and the associated terminal event.

Joint frailty models estimate several parameters, of which α^ and θ^ characterize the dependency between recurrent events and the terminal event. The parameter α^ represents the association between recurrences and terminal event frailties (e.g., to what extent the frailty affecting recurrences also affects the terminal event). When α^ = 0, the dependency between recurrent events and the terminal event is fully explained by the observable covariates (and not by any frailty). When α^ > 0, the frailty of recurrences affects the frailty of terminal event (with larger values of α^ indicating stronger dependency between the two). When α^ = 1, the frailty effects are the same for both recurrent and terminal events; when α^ > 1, higher frailty increases risk for both recurrence and death rates; and when α^ < 1, vice versa. The parameter θ^ represents the variance of frailties, which measures the heterogeneity of the observations. If θ^ = 0, it means that none of the heterogeneity between individuals is explained by non‐observed covariates (i.e., frailties). When θ^ > 1, it implies that there is heterogeneity between individuals that is explained by frailties, and the larger θ^ is, the greater the heterogeneity (i.e., the variance) (Rondeau et al., 2012).

The joint frailty models also allowed inclusion of covariates. These covariates could be the same or different for the recurrent events and the terminal event processes, and the models yielded separate coefficients and hazard ratios for each of these processes (i.e., for recurrences and for terminal events). The coefficient sign (negative or positive) indicates the direction of its effect relative to the baseline hazard. The hazard ratios quantify the magnitude of the multiplicative change relative to the baseline hazard. Covariates of biological interest that we initially considered were sex, age, and season (dry, rut, and wet).

RESULTS

During 2016–2019, we darted, immobilized, and marked 43 ibex (25 males, 18 females); all were adults except for one male that was 5.5 years old (i.e., intermediate age) and one female that was <1 (i.e., a kid). Sampling was evenly distributed both spatiotemporally and sex wise, making it likely that the marked sample was indeed representative of the adult population.

Disease onset, recovery, and recurrence

The cameras took 211,129 photos or videos in which one or more of the 43 marked individuals were observed (mean = 4910 observations per individual). In 6 years of monitoring, there were 12 different marked individuals (out of 43 in total) that exhibited visible disease signs of one (or more) of the five monitored diseases (Table 1; Figure 2), six of which exhibited signs in more than 1 year and three of them had more than one disease (not necessarily concurrently). The mean was 3.5 different individuals with detected diseases per year, out of a mean of 26 total living marked individuals per year (Table 1).

FIGURE 2.

FIGURE 2

Open in a new tab

Examples of marked Nubian ibex (Capra nubiana) individuals with visible signs of disease taken from 2017 to 2022 from a camera trap array used to monitor the population of the north Judean Desert, Israel: (a) abscess from caseous lymphadenitis (CLA) on adult male (Herzel), (b) CLA in adult male (Kokhba), (c) CLA in adult female (Ruth), (d) Ruth recovered from CLA, (e) infectious keratoconjunctivitis (IKC) in adult male (Herzel), (f) Herzel recovered from IKC with scarring to the eye tissue (white spot), (g) lesions from contagious ecthyma (Orf virus) on adult male (Shimi), (h) lumpy jaw lesion on adult male (Yadin), (i) pregnant adult female (Golda) on 21 March 2017, (j) suspected abortion (SAB) in Golda (2 April 2017) as indicated by blood stains and dwindled belly (in 2016 Golda tested positive for past exposure to Coxiella burnetii [Q‐fever]), (k) Golda in another SAB (photo 20 March 2018), and (l) SAB in adult female Shoshi, indicated by blood stains and dwindled belly (photo 12 December 2017) (Shoshi tested positive in 2017 for past exposure to Chlamydia abortus).

There were 17 recorded cases of marked individuals with diseases (all five monitored diseases were represented). The most prevalent disease was CLA, manifested in visible abscesses, with 11 documented cases of disease onset (two of which were recurrences, hence in total nine individuals). To corroborate the visual diagnosis, we took a sample from the abscess of one of the marked individuals exhibiting CLA signs. It was found positive for Corynebacterium pseudotuberculos is on aerobic culture.

Annual CLA prevalence in the marked subset varied between years and sexes (Table 1). In 2017, prevalence was 0.18 (95% CI 0.05–0.44) in males and 0.00 (0.00–0.35) in females. In 2018, prevalence was 0.12 (0.03–0.34) in males and 0.00 (0.00–0.19) in females. In 2019, prevalence was 0.00 (0.00–0.18) in males and 0.06 (0.01–0.27) in females. In 2020, prevalence was 0.18 (0.05–0.48) in males and 0.07 (0.01–0.30) in females. In 2021, prevalence was 0.30 (0.11–0.60) in males and 0.08 (0.01–0.33) in females. In 2022, prevalence was 0.22 (0.06–0.55) in males and 0.00 (0.00–0.26) in females. Prevalence was not calculated for the remaining four monitored diseases because there was an insufficient number of observations over the 6 years (total cases across both sexes: SAB = 3, IKC = 1, LJW = 2, and ORFV = 1), which precluded meaningful estimation of annual prevalence or confidence intervals.

Temporal trends and sex‐specific differences

Apparent CLA prevalence was higher in males (eight cases out of 25 marked males = 32%) than in females (three cases out of 18 marked females = 17%). Likewise, in the GAM, males had a significantly higher baseline CLA prevalence than females after accounting for the effect of years (sex: male β = 2.08 [SE 1.01], p = 0.041). In contrast, there was no evidence for temporal trends in prevalences from 2017 to 2022 for either sex (smooth term for year: edf = 1.36, χ2 = 1.94, p = 0.40 for males and edf = 2.01, χ2 = 1.34, p = 0.60 for females). Figure 3 shows apparent (observed) prevalences (point estimates with error bars) overlaid with GAM‐predicted prevalences (curves with colored ribbons). These results should be interpreted cautiously given the small annual sample sizes and the wide confidence intervals around both observed and model‐predicted prevalences.

FIGURE 3.

FIGURE 3

Open in a new tab

Annual trends in caseous lymphadenitis (CLA) abscess prevalence in the marked subset of Nubian ibex (Capra nubiana) in the north Judean Desert, Israel, from 2017 to 2022, shown separately for males and for females. Apparent annual prevalence (calculated as the proportion of infected marked individuals relative to the total number of living marked individuals per sex and year) is shown as point estimates (red square, females; blue triangles, males) with Wilson score 95% confidence intervals (error bars). Generalized additive model (GAM)‐predicted prevalence is shown as sex‐specific smooth curves with associated 95% confidence intervals (shaded red and blue ribbons, for females and males, respectively).

The sensitivity analysis with GAM_re indicated that the estimated standard deviation of the individual random effect was near zero (SD = 7.239 × 10−5), suggesting no detectable between‐individual heterogeneity attributable to repeated observations given the available data. For the model selection, both GAM and the GAM_re were fitted to the same Bernoulli‐based (disaggregated) dataset to ensure comparability. Model selection with AICc showed that the simpler GAM was equally supported but more parsimonious (AICc = 49.3, LL = −19.9, k = 4, weighting = 0.50). The GAM_re differed only marginally from the GAM (ΔAICc = 0.000074) because it effectively collapsed to the same model as the random‐effect variance approached zero. Combined, these results indicated that pseudoreplication was unlikely to have biased prevalence estimates, further supporting the use of the simpler, more parsimonious aggregated binomial GAM for inference and visualization (Figure 3). The GAM_re was therefore retained solely as a sensitivity analysis. Datasets and the full R code for GAM and GAM_re analyses are in Appendices S1 and S2, respectively.

Other than CLA, the recorded diseases in the marked subset were one IKC, two LJW, one ORFV, also corroborated in the lab), and three suspected abortions (SABs). The three SABs were in two different females (one of which had two SABs in two consecutive years), and both had serological evidence (via blood samples taken when captured) for past exposure to Chlamydia abortus or Coxiella burnetii, known agents associated with abortions in populations of small ruminants (Concha‐Bermejillo & Romano, 2021).

Detected diseases in the unmarked population included (Figure 4) CLA (most prevalent, in agreement with the marked subset), IKC, necrotic lesion, nodular lesion, and two cases of healing tissues from infestations (one probably by Hypoderma bovis, whereas the other likely by Oestrus ovis). The occurrence of additional diseases that were not observed in the marked subset highlights the further potential of camera traps as a monitoring tool for WHM.

FIGURE 4.

FIGURE 4

Open in a new tab

A sample of diseases with visual signs in the Nubian ibex (Capra nubiana) unmarked population of north Judean Desert, Israel, from 2017 to 2022: (a) abscess from caseous lymphadenitis (CLA) in an adult male, (b) two CLA abscesses in an adult female, (c) healing tissue, probably from sheep bot fly (Oestrus ovis) infestation in an adult female, (d) nodular lesion on an adult female, (e) nodular lesion on the female in (d) 2 weeks later, (f) necrotic lesion on young male, (g) infectious keratoconjunctivitis (IKC) in adult female, (h) IKC in adult male, and (i) healing tissue, probably from warble fly (Hypoderma bovis) infestation in an adult female.

Joint frailty model for recurrent events and a terminal event

Because CLA was the most prevalent disease, we fitted a joint frailty model for recurrent events and a terminal event with data on disease onset, recurrence, and terminal events from nine marked individuals (seven males, two females) that exhibited CLA signs between 2017 and 2022 (Table 1). The data used to fit the model and our associated R code are available in Appendices S3 and S4, respectively. During this period, among these nine individuals, there were four terminal events and two recurrent events (CLA relapses). Given our small sample size (9) and the limited number of events, we included only one covariate (sex) in the model to maintain parsimony. Results (Table 2) showed statistically significant effects for the terminal event coefficient and for the association parameter α (i.e., association between recurrence and terminal rates), whereas the recurrence coefficient and the variance parameter θ (i.e., variance of frailties) were not statistically significant. The estimate of α^ = 2.32 (SE 0.92, p = 0.012) indicated a positive association between recurrence and death rates. Male sex was a significant prognostic factor for terminal events (coefficient = −2.60, hazard ratio = 0.07, SE 1.11, p = 0.019), indicating that among the nine marked individuals with CLA, male ibex had lower hazard of terminal events relative to females. Specifically, the hazard for males was approximately 7% of that for females.

TABLE 2.

Results of joint frailty models for recurrent events and a terminal event (with one covariate [sex]), in which each cluster contains all observations of one individual ibex.

Coefficient and parameter estimates a Hazard ratio [exp(coefficient)] b SE p
Covariate
Recurrent disease events
Male sex −0.96 0.38 1.44 0.507
Terminal disease event
Male sex −2.60 0.07 1.11 0.019
Parameter
α^ c (association between recurrence and terminal rates) 2.32 N/A 0.92 0.012
θ^ (variance of frailties) 0.92 N/A 0.61 0.064
Open in a new tab
a

Coefficient for terminal event shows the shift from baseline hazard function is negative.

b

Magnitude of shift from the baseline (i.e., 7% of the baseline).

c

Signifies a positive association between recurrent and terminal events.

We did not fit joint frailty models for the remaining four monitored diseases because the data were insufficient for such analyses. Specifically, these diseases lacked sufficient numbers of recurrent and terminal events required for joint frailty modeling. ORFV and IKC had no recurrent events; SAB had one recurrence but no terminal events; and LJW had one recurrence and one terminal event. In contrast, CLA had two recurrences and four terminal events, which, although limited, were sufficient to permit model estimation.

DISCUSSION

We demonstrated the feasibility and utility of a novel method for monitoring wildlife diseases with visible signs. By integrating camera trap data with data with uniquely identifiable marked subset of individuals (with artificial or natural marking), we obtained robust estimates of apparent disease prevalence and temporal trends at the population level. In addition, we established the applicability of joint frailty models to these data, showing that this framework can yield further epidemiological insight into recurrence dynamics and their association with mortality.

At least 12 of the 43 marked individuals showed signs of one or more of the five diseases we considered. CLA was the most prevalent disease and the only one for which sufficient data were available to estimate apparent prevalence and temporal trends reliably. Despite the small number of individuals with CLA (n = 9), joint frailty models enabled meaningful and robust inference on epidemiological and demographic processes, while identifying sex as a key covariate influencing terminal event risk. The ability of joint frailty models to operate effectively even with small sample sizes makes them a compelling complement to this monitoring method. With larger sample sizes, these models could yield more precise estimates and facilitate evaluation of additional covariates (Rondeau et al., 2003, Rondeau 2007), such as age and season, both well‐established factors affecting CLA in domestic goats (Yitagesu et al., 2020).

Benefits of the method

Developing methods that avert some of WHM data collection challenges is necessary (Barroso et al., 2024; Ryser‐Degiorgis, 2013). Considering the prevalence, low cost, ease of application, and ability to operate in remote and harsh environments (also for long durations), camera traps are exceptionally useful tools for WHM. Their advantages are particularly evident in situations where collecting diagnostic samples is difficult, often requiring culling, extensive capture surveys, or, most challenging, spotting sick individuals and subsequently trapping or darting them. Clearly, capturing and marking individuals in wild populations is no small feat, and obtaining larger samples can prove labor intensive and costly (Festa‐Bianchet et al., 2017), adding to the method's challenges. However, many species possess naturally occurring individually identifiable marks (e.g., tigers [Panthera tigris] stripe pattern), rendering, in such cases, the application of this method more resource efficient (no need for capturing) and potentially more accurate (because of bigger sample size). Thus, camera traps combined with marked individuals can serve as an auxiliary method for ongoing WHM programs or operate as a stand‐alone approach when more intensive surveillance methods are not feasible.

Limitations of the method

First, this method requires a camera trap array that adequately covers the area of interest (at the very least, dozens of cameras) for extended periods (e.g., multiple years). Accordingly, the footage volume will be large (sometimes huge, in the order of millions of files), requiring many hours of manual sifting by trained professionals. One way to address this obstacle is to use artificial intelligence (AI) to sift through the footage, identify species (and marked individuals), and produce an organized database (Mulero‐Pázmány et al., 2025; Vélez et al., 2023). Albeit efficient, AI has its limitations. For example, it may yield a higher percentage (vs. expert eyes) of false positives, including misidentifications (e.g., classifying a male ibex as a female or even as a gazelle) and false identifications (e.g., classifying an artifact as an ibex), which is particularly problematic when dealing with threatened species. We manually scanned the footage because it allowed us to learn and familiarizing ourselves with ibex behavior on several levels: individual (via uniquely marked individuals), social (via intraspecific interactions), and community (via interspecific interactions). Nevertheless, manual scanning is undoubtedly resource‐ and time‐consuming, and in many cases, alternatives (e.g., AI) might prove more beneficial.

Second, validating the sensitivity and specificity of diagnostic methods in wild populations is mostly unfeasible, especially in species inhabiting inaccessible habitats. A complete validation of the method's performance will require captive settings. For instance, future studies could deploy camera traps in large enclosures of captive ibex (or similar species, e.g., fallow deer [Dama dama]). For every visually diagnosed sign (by veterinarians) from footage, the sign‐bearing individual is captured, and samples are taken for lab diagnosis. This way a precise ratio between veterinarian visual diagnoses can be compared with verified results from the lab and ultimately to a gold standard to calculate the method's accuracy. Although such validation would undoubtedly contribute to the standing of our method, it was outside our study scope.

Third, another source of uncertainty comes from the detection probability of the visible signs of the diseases (or rather the probability that visible signs would go undetected). Estimating this probability would require a population that is entirely monitored and accessible (e.g., captive) with a known number of visible signs. Then, a mark–resight analysis, heuristically equating known signs to uniquely marked individuals, must follow, in which the abundance of visible signs (vs. a known number) is assessed and the probability of sign detection is estimated. Such validation would strengthen our method, yet it exceeded the study scope. Nonetheless, by positioning multiple camera traps in individual home ranges and configuring them to also take videos, we sought to increase the likelihood of detecting visible disease signs (i.e., increasing the detection probability). The increased camera density and video recording provided more opportunities to detect signs (e.g., from different angles, time of the day, distances, and lighting conditions).

Finally, our method is more suitable for certain species. For example, small species (e.g., all invertebrates, many reptiles, amphibians, and very small mammals) would be a bad fit because tagging them with individually identifiable marks implies very small markings, which are likely impossible to differentiate individually in camera trap footage. Similarly, most bird species would not be a good fit, especially nomadic species or species that spend most of their time in the air. For camera traps with infrared‐based motion sensors, species whose body temperatures approximate ambient temperatures (e.g., many nonhomeothermic taxa) would mostly go undetected (but see Welbourne et al. [2017] for advances in this regard). Essentially, a method that combines camera traps and marked individuals will predominantly apply to terrestrial species that exhibit range fidelity, have defined home ranges (that are not vast, due to camera quantities and logistics limitations), and have body sizes that are sufficiently large (e.g., ≥1 kg) such that individual markings are identifiable in camera footage (preferably in day‐ and nighttime). Despite these limitations, the method is relevant for a substantial proportion of terrestrial mammal species (∼25% or 1600 species) and some bird species.

Conservation implications

Overall, our monitored ibex population did not exhibit high prevalence of detectable illness; 28% of marked individuals showed visible sign of one or more of the five monitored diseases over the course of six consecutive years. This proportion likely underestimates true occurrence because individuals with visible visual signs may have gone undetected if camera traps failed to capture them at appropriate angles (which we mitigated, by adding video tracking and additional cameras to increase detection probability, though we cannot quantify by how much). Further, because camera traps only allow detecting diseases with visible signs, this statistic is not an indication of the overall health of the population but rather of the five specific monitored diseases. Using this method, we attained needed data on Nubian ibex health, a topic that is currently lacking in the literature (Ross et al., 2020).

As a proof of concept, our study substantiates this new method as a valuable tool for WHM programs. Even in the absence of formal accuracy validation (e.g. sensitivity and specificity), the method can provide useful monitoring capacity, particularly for detecting disease occurrence and temporal or spatial trends at broad scales. Given camera traps are limited to detecting diseases with visible signs, they cannot provide a comprehensive diagnostic solution for WHM programs and should therefore function either as an auxiliary component of existing surveillance programs or as a stand‐alone approach where more intensive alternatives are not feasible. Combined with the rising popularity of camera traps as a research tool and their growing numbers in the field, this method may boost the amount of data available in WHM, especially of rare and elusive species and populations inhabiting remote and rugged habitats. Although providing incomplete coverage, camera traps enable monitoring of a wide range of diseases with visible signs that are currently not monitored in wild populations.

Supporting information

Supporting information

COBI-40-e70283-s003.csv (563B, csv)

Supporting information

COBI-40-e70283-s001.docx (28.8KB, docx)

Supporting information

COBI-40-e70283-s002.xlsx (9.5KB, xlsx)

Supporting information

COBI-40-e70283-s004.docx (22.3KB, docx)

ACKNOWLEDGMENTS

We thank all those who contributed to this research, including fieldwork, ibex capturing, camera trap maintenance, and brainstorming. We thank INPA rangers and volunteers, S. Kfir, Y. Arri, L. Lev, A. Mizrahi, L. Reuveni, E. Hazan (and Einot Tzukim staff), and M. Plattner. We also thank our lab members T. Halevy, O. Golan, I. Goodman, L. Zecherle, and Y. Cohen. Finally, we are grateful to the handling editor G. Hosack and two anonymous reviewers for their insightful comments and ideas, which greatly improved this manuscript.

Tichon, J. , Lapid, R. , King, R. , Freiman, E. , Spiegel, O. , Baruchi, E. , Atar, A. , Bar‐David, S. , & Saltz, D. (2026). Monitoring wildlife health for diseases with visible signs by integrating camera traps with marked individuals. Conservation Biology, 40, e70283. 10.1111/cobi.70283

Article impact statement: Integrating camera traps and a subset of marked individuals to monitor wildlife health can boost available data worldwide.

REFERENCES

  1. Aguilar, X. F. , López‐Olvera, J. R. , Ribas, M. P. , Begovoeva, M. , Velarde, R. , Cardells, J. , & Cabezón, O. (2019). Mycoplasma conjunctivae in insect vectors and anatomic locations related to transmission and persistence. Veterinary Microbiology, 228, 7–11. [DOI] [PubMed] [Google Scholar]
  2. Aronson, L. (1982). The ibex—King of the cliffs. Masada Press. [Google Scholar]
  3. Balan, T. A. , & Putter, H. (2020). A tutorial on frailty models. Statistical Methods in Medical Research, 29(11), 3424–3454. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Barroso, P. , López‐Olvera, J. R. , Kiluba, T. K. , & Gortázar, C. (2024). Overcoming the limitations of wildlife disease monitoring. Research Directions: One Health, 2, Article e3. [Google Scholar]
  5. Barroso, P. , Relimpio, D. , Zearra, J. A. , Cerón, J. J. , Palencia, P. , Cardoso, B. , Ferreras, E. , Escobar, M. , Cáceres, G. , López‐Olvera, J. R. , & Gortázar, C. (2023). Using integrated wildlife monitoring to prevent future pandemics through one health approach. One Health, 16, Article 100479. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bengsen, A. J. , Forsyth, D. M. , Ramsey, D. S. L. , Amos, M. , Brennan, M. , Pople, A. R. , Comte, S. , & Crittle, T. (2022). Estimating deer density and abundance using spatial mark‐resight models with camera trap data. Journal of Mammalogy, 103(3), 711–722. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Burnham, K. P. , & Anderson, D. R. (2004). Multimodel inference: Understanding AIC and BIC in model selection. Sociological Methods & Research, 33(2), 261–304. [Google Scholar]
  8. Concha‐Bermejillo, A. , & Romano, J. (2021). Pregnancy loss in small ruminants. Clinical Theriogenology, 13(3), 194–205. [Google Scholar]
  9. Dayan, U. , & Morin, E. (2006). Flash flood–producing rainstorms over the Dead Sea: A review. In Enzel Y., Agnon A., & Stein M. (Eds.), New frontiers in Dead Sea paleoenvironmental research (pp. 53–62). Geological Society of America. [Google Scholar]
  10. Fagan, D. A. , Oosterhuis, J. E. , & Benirschke, K. (2005). “Lumpy jaw” in exotic hoof stock: A histopathologic interpretation with a treatment proposal. Journal of Zoo and Wildlife Medicine, 36(1), 36–43. [DOI] [PubMed] [Google Scholar]
  11. Festa‐Bianchet, M. , Douhard, M. , Gaillard, J. M. , & Pelletier, F. (2017). Successes and challenges of long‐term field studies of marked ungulates. Journal of Mammalogy, 98(3), 612–620. [Google Scholar]
  12. Gündoğdu, E. , Arpacık, A. , Sari, A. , & Başkaya, Ş. (2018). Orf virus infection in wild goats (Capra aegagrus Erxleben 1777) of Saricicek mountain. Applied Ecology and Environmental Research, 16(4), 4515–4521. [Google Scholar]
  13. Greenberg‐Cohen, D. , Alkon, P. U. , & Yom‐Tov, Y. (1994). A linear dominance hierarchy in female Nubian ibex. Ethology, 98(3‒4), 210–220. [Google Scholar]
  14. Gross, J. E. , Alkon, P. U. , & Demment, M. W. (1995). Grouping patterns and spatial segregation by Nubian ibex. Journal of Arid Environments, 30(4), 423–439. [Google Scholar]
  15. Habibi, K. (1994). The desert ibex. Immel Publishing. [Google Scholar]
  16. Hastie, T. , & Tibshirani, R. (1986). Generalized additive models. Statistical Science, 1(3), 297–310. [DOI] [PubMed] [Google Scholar]
  17. Hochman, V. , & Kotler, B. P. (2007). Patch use, apprehension, and vigilance behavior of Nubian ibex under perceived risk of predation. Behavioral Ecology, 18(2), 368–374. [Google Scholar]
  18. Hockings, K. J. , Mubemba, B. , Avanzi, C. , Pleh, K. , Düx, A. , Bersacola, E. , Bessa, J. , Ramon, M. , Metzger, S. , Patrono, L. V. , Jaffe, J. E. , Benjak, A. , Bonneaud, C. , Busso, P. , Couacy‐Hymann, E. , Gado, M. , Gagneux, S. , Johnson, R. C. , Kodio, M. , … Leendertz, F. H. (2021). Leprosy in wild chimpanzees. Nature, 598, 652–656. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Huang, X. , & Liu, L. (2007). A joint frailty model for survival and gap times between recurrent events. Biometrics, 63(2), 389–397. [DOI] [PubMed] [Google Scholar]
  20. Iribarren, C. , & Kotler, B. P. (2012). Foraging patterns of habitat use reveal landscape of fear of Nubian ibex Capra nubiana . Wildlife Biology, 18(2), 194–201. [Google Scholar]
  21. Król, A. , Mauguen, A. , Mazroui, Y. , Laurent, A. , Michiels, S. , & Rondeau, V. (2017). Tutorial in joint modelling and prediction: A statistical software for correlated longitudinal outcomes, recurrent events and a terminal event. Journal of Statistical Software, 81(3), 1–52. [Google Scholar]
  22. Moore, H. A. , Dunlop, J. A. , Geyle, H. M. , Greenwood, L. , & Nimmo, D. G. (2023). First you get the money, then you get the power: Comparing the cost and power of monitoring programs to detect changes in occupancy of a threatened marsupial predator. Conservation Science and Practice, 5(2), Article e12881. [Google Scholar]
  23. Mulero‐Pázmány, M. , Hurtado, S. , Barba‐González, C. , Antequera‐Gómez, M. L. , Díaz‐Ruiz, F. , Real, R. , Navas‐Delgado, I. , & Aldana‐Montes, J. F. (2025). Addressing significant challenges for animal detection in camera trap images: A novel deep learning‐based approach. Scientific Reports, 15, Article 16191. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Muneza, A. B. , Ortiz‐Calo, W. , Packer, C. , Cusack, J. J. , Jones, T. , Palmer, M. S. , Swanson, A. , Kosmala, M. , Dickman, A. J. , Macdonald, D. W. , & Montgomery, R. A. (2019). Quantifying the severity of giraffe skin disease via photogrammetry analysis of camera trap data. Journal of Wildlife Disease, 55(4), 770–781. [PubMed] [Google Scholar]
  25. Murray, M. H. , Fidino, M. , Lehrer, E. W. , Simonis, J. L. , & Magle, S. B. (2021). A multi‐state occupancy model to non‐invasively monitor visible signs of wildlife health with camera traps that accounts for image quality. Journal of Animal Ecology, 90(8), 1973–1984. [DOI] [PubMed] [Google Scholar]
  26. Olaega, Á. , Casais, R. , Balseiro, A. , Espí, A. , Llaneza, L. , Hartasánchez, A. , & Gortázar, C. (2011). New techniques for an old disease: Sarcoptic mange in the Iberian wolf. Veterinary Parasitology, 181(2‐4), 255–266. [DOI] [PubMed] [Google Scholar]
  27. Rondeau, V. , Commenges, D. , & Joly, P. (2003). Maximum penalized likelihood estimation in a gamma‐frailty model. Lifetime Data Analysis, 9(2), 139–153. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Rondeau, V. , & Gonzalez, J. R. (2005). frailtypack: A computer program for the analysis of correlated failure time data using penalized likelihood estimation. Computer Methods and Programs in Biomedicine, 80(2), 154–164. [DOI] [PubMed] [Google Scholar]
  29. Rondeau, V. , Mathoulin‐Pelissier, S. , Jacqmin‐Gadda, H. , Brouste, V. , & Soubeyran, P. (2007). Joint frailty models for recurring events and death using maximum penalized likelihood estimation: Application on cancer events. Biostatistics, 8(4), 708–721. [DOI] [PubMed] [Google Scholar]
  30. Rondeau, V. , Mazroui, Y. , & Gonzalez, J. R. (2012). frailtypack: An R package for the analysis of correlated survival data with frailty models using penalized likelihood estimation of parametrical estimation. Journal of Statistical Software, 47(4), 1–28. [Google Scholar]
  31. Ross, S. , Elalqamy, H. , Al Said, T. , & Saltz, D. (2020). Capra nubiana. The IUCN Red List of Threatened Species, 2020, Article e.T3796A22143385. [Google Scholar]
  32. Ryser‐Degiorgis, M. P. (2013). Wildlife health investigations: Needs, challenges and recommendations. BMC Veterinary Research, 9, Article 223. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Sanchez, J. N. , Munk, B. A. , Colby, J. , Torres, S. G. , Gonzales, B. J. , DeForge, J. R. , Byard, A. J. , Konde, L. , Shirkey, N. J. , Pandit, P. S. , Botta, R. A. , Roug, A. , Ziccardi, M. H. , & Johnson, C. K. (2022). Pathogen surveillance and epidemiology in endangered peninsular bighorn sheep (Ovis canadensis nelsoni). Conservation Science and Practice, 4(11), Article e12820. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Shkolnik, A. (1981). Physiological Adaptation for Milk Production in Desert Ungulates (Report No. IAEA‐R‐2138‐F). International Atomic Energy Agency, Vienna, Austria. https://inis.iaea.org/records/ztq5z‐n4c36 [Google Scholar]
  35. Tichon, J. , Freiman, E. , Spiegel, O. , Baruchi, E. , Atar, A. , Lapid, R. , King, R. , Bar‐David, S. , & Saltz, D. (2022). Population dynamics and the effect of drought in the threatened Nubian ibex. Wildlife Society Bulletin, 47(1), Article e1397. [Google Scholar]
  36. Tichon, J. , Freiman, E. , Spiegel, O. , Baruchi, E. , Atar, A. , Lapid, R. , King, R. , Bar‐David, S. , & Saltz, D. (2024). Species behavioral characteristics lead to unique fragmentation threats: The Nubian ibex as a case study. Journal of Wildlife Management, 89(2), Article e22698. [Google Scholar]
  37. Underwood, W. J. , Blauwiekel, R. , Delano, M. L. , Gillesby, R. , Mischler, S. A. , & Schoell, A. (2015). Biology and diseases of ruminants (sheep, goats, and cattle). In Fox J. G., Anderson L. C., Otto G., Pritchett‐Corning K. R., & Wharry M. T. (Eds.), Laboratory animal medicine (3rd ed., pp. 623–694). Academic Press. [Google Scholar]
  38. Vélez, J. , McShea, W. , Shamon, H. , Castiblanco‐Camacho, P. J. , Tabak, M. A. , Chalmers, C. , Fergus, P. , & Fieberg, J. (2023). An evaluation of platforms for processing camera‐trap data using artificial intelligence. Methods in Ecology and Evolution, 14(2), 459–477. [Google Scholar]
  39. Valldeperes, M. , Yerro, P. P. , López‐Olvera, J. R. , Fandos, P. , Lavín, S. , Escofet, R. C. S. , Mentaberre, G. , León, F. J. C.‐M. , Espinosa, J. , Ráez‐Bravo, A. , Pérez, J. M. , Tampach, S. , Estruch, J. , Velarde, R. , & Granados, J. E. (2023). Diseases of Iberian ibex (Capra pyrenaica). European Journal of Wildlife Research, 69(3), Article 63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Wearn, O. R. , & Glover‐Kapfer, P. (2019). Snap happy: Camera traps are an effective sampling tool when compared with alternative methods. Royal Society Open Science, 6(3), 181748. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Welbourne, D. J. , Paull, D. J. , Claridge, A. W. , & Ford, F. (2017). A frontier in the use of camera traps: Surveying terrestrial squamate assemblages. Remote Sensing in Ecology and Conservation, 3(3), 133–145. [Google Scholar]
  42. Wilson, E. B. (1928). Probable inference, the law of succession, and statistical inference. Journal of the American Statistical Association, 22(158), 209–212. [Google Scholar]
  43. Wood, S. N. (2011). Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society, 73(1), 3–36. [Google Scholar]
  44. Wood, S. N. (2017). Generalize additive models: An introduction with R (2nd ed.). CRC Press. [Google Scholar]
  45. Wood, S. N. (2025). Generalized additive models. Annual Review of Statistics and Its Application, 12, 497–526. [Google Scholar]
  46. Wood, S. N. , Pya, N. , & Säfken, B. (2016). Smoothing parameter and model selection for general smooth models (with discussion). Journal of the American Statistical Association, 111(516), 1548–1563. [Google Scholar]
  47. Yitagesu, E. , Alemnew, E. , Olani, A. , Asfaw, T. , & Demis, C. (2020). Survival analysis of clinical cases of caseous lymphadenitis of goats in North Shoa, Ethiopia. Veterinary Medicine International, 17, Article 8822997. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting information

COBI-40-e70283-s003.csv (563B, csv)

Supporting information

COBI-40-e70283-s001.docx (28.8KB, docx)

Supporting information

COBI-40-e70283-s002.xlsx (9.5KB, xlsx)

Supporting information

COBI-40-e70283-s004.docx (22.3KB, docx)

Articles from Conservation Biology are provided here courtesy of Wiley

RESOURCES