Refining Camera Trap Surveys for Mammal Detection and Diversity Assessment in the Baviaanskloof Catchment, South Africa

Simple Summary

Biodiversity conservation is essential, especially in areas where natural habitats and agricultural land coexist. In South Africa, a large proportion of wild species live outside protected reserves, making wildlife monitoring in mixed-land-use landscapes crucial. This study used 131 camera traps over two years to survey medium- to large-sized mammals in the Baviaanskloof catchment, Eastern Cape, South Africa, recording 34 mammal species across 21,020 trap days. Species diversity varied across locations, and detection rates were influenced by differences in camera setups and placements. Optimal setups included elevations of 40–70 cm, angles between 50 and 90° and north- or south-facing orientations. Cameras placed higher than 100 cm, at extreme angles, or facing west were less effective. These insights provide a valuable baseline for mammal diversity in the Baviaanskloof and offer practical guidelines to enhance wildlife monitoring and conservation efforts in similarly diverse landscapes.

Abstract

Conserving biodiversity in mixed-land-use areas is essential, as nearly 80% of South Africa’s wild species exist outside protected areas. This study investigated mammalian diversity within the Baviaanskloof catchment, a mixed-use landscape in the Eastern Cape, South Africa. It also evaluated how camera setup parameters impact species detectability. Using 131 camera traps over four survey sessions from January 2020 to April 2022, 34 mammalian species were recorded over 21,020 trap days. Biodiversity indices revealed high species diversity with substantial variability across camera locations. Species discovery reached an asymptote at approximately 153 sampling days, though extended monitoring detected rarer species. Cameras positioned at heights of 40–70 cm improved detection rates, while heights above 100 cm reduced captures. However, elevation effects varied across species, highlighting the need for species-specific optimization. Optimal detection angles ranged from 50 to 90°, with extreme angles decreasing capture frequency. North- and south-facing cameras yielded better detection rates, while west-facing orientations introduced glare and reduced visibility. These findings underscore the biodiversity significance of the Baviaanskloof and emphasize the need to optimize camera configurations to enhance wildlife monitoring and conservation strategies in complex, mixed-use landscapes.

1. Introduction

Biodiversity, defined as the variety and abundance of species within a specific area, is a key metric in conservation biology, essential for tracking changes over time and understanding the underlying ecosystem dynamics [1]. Effective monitoring provides critical baseline data needed to measure progress in reducing biodiversity loss [2]. From a management perspective, monitoring assesses ecosystem health and guides conservation interventions, while scientific objectives focus on understanding species behaviour, distribution patterns, and ecological interactions within their environments [3,4]. Such monitoring efforts are crucial for providing actionable data to policymakers and conservationists, aiding in the development of effective conservation strategies [2,4].

Monitoring efforts are most effective when carefully designed, as the metrics chosen and survey methodologies employed directly influence the reliability and relevance of collected data [5]. Understanding mammalian distributions is particularly valuable for identifying biodiversity patterns and informing conservation strategies [6,7]. However, expanding and intensifying land use globally has placed unprecedented pressure on ecosystems, with many of the resulting impacts on ecological processes still poorly understood [7]. Habitat loss, compounded by the fragmentation of remaining habitats, significantly alters ecosystems. Species that persist in fragmented environments face challenges such as reduced habitat availability, increased isolation, and exposure to novel ecological boundaries [8,9]. Understanding how these driving forces affect biodiversity over time is essential to guide effective conservation efforts.

Survey design is a critical component of biodiversity monitoring, particularly when assessing changes in species occupancy and diversity across landscapes. Species richness, overall abundance, and evenness are commonly used metrics for biodiversity monitoring [4]. When species vary significantly in size or behaviour, occupancy models or relative abundance indices may be more appropriate than simple abundance measures [10]. These metrics offer valuable insights into the distribution of species across fragmented or modified habitats, highlighting areas of concern where biodiversity may be declining [4]. However, many biodiversity surveys fail to adequately address key questions regarding survey objectives and design, which can undermine the utility of the data collected [5]. Careful consideration of survey duration and spatial coverage is essential, as short-term or narrowly focused surveys may not capture the full extent of species dynamics or the long-term risks to populations [9].

To improve the accuracy of biodiversity assessments, monitoring programs must account for spatial variation and species detectability [11]. Detectability can vary widely between species, influenced by factors such as group size, mobility, and habitat preference [11]. It is crucial to incorporate detection probability into survey designs to accommodate differences in species detectability [12]. Incorporating these factors enables more precise estimates of species occupancy and diversity, particularly when using tools like camera traps that are prone to detection bias [13].

Camera traps have become an invaluable tool for monitoring medium- to large-sized mammals, offering a non-invasive method to observe species across diverse habitats and enabling continuous data collection over extended periods [14,15,16]. They are particularly effective for assessing species occupancy and diversity, as they can capture data on elusive or nocturnal species that might otherwise go undetected [16,17]. However, while camera traps have revolutionized biodiversity monitoring, their effectiveness is highly dependent on survey design and setup parameters, which require careful consideration to ensure reliable results [18,19]. Factors such as camera trap placement, deployment duration, and habitat type can influence detection rates, and survey designs must be robust enough to account for species-specific differences in detection probability [4,20]. Key camera setup parameters, including camera model, elevation, angle, aspect, and placement on or off trails, play a critical role in detection efficiency. Moderate camera elevation, typically between 40 and 70 cm, has been shown to optimize detection rates, while higher placements may reduce detection probability, particularly for smaller species [21]. Similarly, camera angle affects the field of view and species detectability, with moderate angles (50–80°) enhancing captures and extreme angles leading to missed detections or image distortion [22]. Camera aspect, or the directional orientation of the camera, can also influence detection rates, with cameras orientated away from sunrise or sunset often yielding higher capture success due to more consistent lighting conditions [23].

Camera placement is another critical factor in camera trap surveys, as the positioning of cameras on established trails and roads have been shown to increase the likelihood of detecting large-bodied species that frequently use these pathways, such as ungulates and carnivores. In contrast, off-trail placements provide a more comprehensive assessment of species that avoid human or predator-associated pathways [24]. Additionally, the choice of camera model, which determines factors such as trigger speed, detection range, and image quality, significantly affects survey performance and data accuracy [25]. Ensuring the optimal configuration of these parameters is essential for maximizing detection rates and obtaining reliable data for wildlife research and conservation planning. Therefore, to achieve the most accurate representation of species diversity and occupancy, camera trap studies must balance these setup variables carefully and adapt to the ecological context of the study area.

2. Materials and Methods

2.1. Study Area

The study area encompasses the 1234 km² Baviaanskloof catchment in the Eastern Cape province, South Africa (Figure 1). This region includes the Baviaanskloof River and surrounding semi-arid, mountainous habitats [26]. It comprises the Baviaanskloof Nature Reserve, part of a UNESCO World Heritage Site, and the privately owned farmlands of the Baviaanskloof Hartland. Renowned for its exceptional biodiversity, the area features seven of South Africa’s eight biomes, including the Cape Floristic Region and subtropical thicket, both recognized as global biodiversity hotspots [27,28]. The region’s diverse geology, topography, and climate have created a wide variety of ecosystems that support an array of wildlife species [29].

The region’s climate is characterized by highly variable rainfall ranging from 100 to 500 mm annually, depending on elevation, and temperatures ranging 5–32 °C [30]. The Baviaanskloof River is largely ephemeral, with water flows significantly influenced by local land use, leading to issues such as erosion and sedimentation [26]. Vegetation biomes within the region include fynbos, subtropical thicket, savanna, forest, and grassland [31]. Overgrazing by livestock, particularly goats (Capra aegagrus hircus), has caused widespread degradation of thicket vegetation [31]. To address this, restoration projects focus on replanting indigenous vegetation and improving land management practices [30], with Spekboom (Portulacaria afra) playing a key role in thicket restoration efforts [32]. Efforts to balance biodiversity conservation with sustainable land use include protected area management, community-based conservation initiatives, and ecological restoration projects [32,33].

2.2. Survey Design

This study employed a stratified, random placement design to ensure comprehensive spatial coverage across various landscape types between January 2020 and April 2022 (Figure 1). A total of 131 unique cameras were deployed over four survey sessions. Here, a deployment is defined as a continuous sampling period during which a single or pair of cameras was placed at a given location with specific setup and environmental characteristics.

Camera traps were deployed in both the nature reserve and farmlands spanning diverse monitoring units defined by topography (high plateau, hillslope, valley bottom, alluvial fan, and narrow valley), vegetation type (thicket, fynbos, forest, and savannah), agricultural activities (crop lands, rangelands, and resting farmland), and varying levels of degradation (transformed land, severely degraded, moderately degraded, and intact vegetation). Stratified random points were generated using the Geospatial Modelling Environment (GME) 0.7.2 RC2 software) [19], ensuring representation across the identified monitoring units. While a true grid-based design was not employed, this stratified random approach allowed for comprehensive spatial coverage and ensured that a wide range of habitats and conditions representative of the region were sampled [19].

The camera setups consisted of three models: Cuddeback (119), Bushnell (9), and Acorn (3). The cameras were placed across three distinct trail types, with the majority on animal trails (90), followed by roads (36), and riverbeds (5) [20,34]. Mounting methods included attaching cameras to trees (84), metal stands (32), and posts (15). Camera angles ranged from 40° to 110°, with the most common angle being 90° (114). Camera heights ranged from 40 to 1200 cm, with the majority of cameras positioned between 40 and 70 cm (109) [35]. Bearings were set in 12 different directions, with the highest number facing south (54) [36]. Additionally, 39 camera sites were equipped with paired cameras (two cameras per location) specifically to improve detection and identification of leopards, while 92 sites were monitored with single cameras. The camera setups were determined by and adjusted according to the requirements of each location. For example, cameras in flood-risk areas were placed at higher elevations, and those on slopes were angled either upward or downward. A comprehensive breakdown of setup parameters per camera deployment is provided in Table S1.

Each survey session lasted approximately 180 days, following recommendations for survey designs targeting species with both low and high encounter rates [37]. The camera traps were operational 24 h a day and were configured to take a single image per trigger event, followed by a 30 s delay before being able to trigger again. This delay was implemented to reduce duplicate captures and ensure independent detection events. Camera trap days were calculated as the number of days each camera was functional [38].

To improve species richness assessments and reduce false triggers, cameras were placed in areas where two or more signs of animal activity (e.g., tracks, scat, or foraging evidence) were observed [38]. Vegetation directly obstructing the camera’s field of view within two meters was trimmed [39]. No bait was used at the camera trap sites to avoid influencing natural animal behaviour [38]. The camera traps were checked every 30 to 45 days to download images, replace batteries if needed, ensure functionality, and clear any vegetation obstructing the camera’s view [40]. A field guide [41] was used to identify all medium to large mammals (over 1 kg) captured in the photographs [39]. Multiple photographs of the same species taken within a short timeframe were considered a single capture event [42]. Photographs were renamed based on the camera station code, time, and date of capture. Where possible, additional information such as species, sex, and the number of animals were tagged in the images. The photographic database was managed using the Timelapse image analysis system [43].

2.3. Data Analysis

All analyses were conducted using R version 4.3.0 [44]. Data cleaning involved removing erroneous records and ensuring date, time, and other metadata formats were standardised and consistent. Survey effort was assessed by calculating the number of operational trap days per camera, representing the period each camera actively recorded data [45]. Camera operation days were determined by subtracting the setup date from the removal date for each camera deployment and was summarized across stations, calculating the mean, minimum, maximum, and total operation days to offer a detailed overview of deployment effort [46].

To ensure temporal independence in observations of medium- and large-sized mammals, a 30 min interval between consecutive detections of the same species at each camera station was applied [47,48]. This threshold was validated through Autocorrelation Function (ACF) plots, using the forecast package, which identified the lag time at which autocorrelation approached zero, suggesting that subsequent detections were independent of prior ones [49]. This approach allowed for the assessment of temporal clustering in detection events across species, ensuring that detected activity patterns were not artificially clustered due to closely spaced detections [39,42]. The analysis primarily focused on shorter lags to determine whether detections were clustered within narrow timeframes or displayed temporal independence at a specified interval. Species with high autocorrelation values at shorter lags, with ACF values around 0.1, indicated clustered activity patterns, and species with consistently low or negative ACF values at initial lags indicated sporadic detections [49].

Species were classified into trophic guilds (Carnivore, Insectivore, Large Herbivore, and Small Herbivore) and by weight class, with species under 50 kg classified as small mammals [50]. This classification was used to identify how body size and ecological roles influence detectability and occupancy patterns across camera trap deployments. The number of captures per species was calculated using independent capture events at 30 min intervals. Capture frequency per species and deployment was calculated by dividing the total independent captures of each species by the total number of independent captures across all species at that deployment and then multiplying by 1000. This standardization step expresses capture frequency as a rate per 1000 captures, allowing for more consistent comparison of relative capture rates across deployments with varying total capture counts [51].

Species richness was defined as the number of unique species per deployment [52], while abundance represented the total number of individuals per species. Mammal diversity was quantified using the Shannon index, which accounts for both species richness and relative abundance [53], and the Simpson index (1-D), which measures the probability that two randomly selected individuals belong to the same species, with values near 1 indicating higher diversity [54]. Evenness assessed species distribution across sites, ranging from 0 (uneven) to 1 (even) [55], and the Jaccard similarity index measured species composition similarity between sites, with values closer to 1 indicating greater similarity [56].

Camera running times were converted into monitoring days, and the cumulative number of unique species observed per day was aggregated to generate a species accumulation curve [52]. A logistic growth model was fitted to estimate the point at which species discovery would reach an asymptote, indicating that further effort would yield diminishing returns [57]. This curve was modelled using a logistic growth function with the following parameters: initial species observed (P0), maximum species capacity (K), and accumulation rate (r). These were estimated via the Non-linear Least Squares Method (NLSM) using the nlsLM function [58]. Bootstrapping with 1000 iterations assessed prediction reliability by resampling data, fitting the model to each resample, and calculating 95% confidence intervals for the asymptote day. Model predictions extended the species accumulation curve beyond observed data, providing insights into future species discovery trends and the point at which additional monitoring would have minimal returns [52].

A Bayesian hierarchical model was implemented via JAGS (Just Another Gibbs Sampler) to estimate occupancy (ψ), the probability that a species is present at a given site, and detection (p), the probability of detecting a species given its presence [59]. Detection history data were compiled as binary data where 1 indicates detection and 0 indicates non-detection. Covariates were prepared and scaled to facilitate model convergence. Continuous covariates, such as camera elevation (height) and camera angle, were standardized using z-score normalization to improve convergence and ensure numerical stability. Categorical variables, including camera model, camera bearing, vegetation type, land use, topographical unit, and species, were converted to dummy variables for compatibility with the model. Occupancy was modelled as a Bernoulli random variable, with the following:

$$z[i,s]\sim Bernoulli\left(\psi \right[i,s\left]\right)$$

where z[i,s] indicates the presence or absence of species s at site i. Occupancy probability (ψ[i,s]) was expressed as a logistic function:

$${\psi }_{i,s}=\mathrm{ilogit}\left({\beta }_0\left[s\right]+\sum _{k=1}^{n_{\mathrm{covariates}}}{\beta }_{\mathrm{cov}}\left[s,k\right]\cdot \mathrm{covariates}\left[i,k\right]\right)$$

where β₀[s] is the intercept for species s and β_cov[s,k] represents the effect of covariate k on species s. Detection probability (p[i,s]) was assigned a Beta prior to constrain values between 0 and 1 [10]:

$$p[i,s]\sim Beta\left(\mathrm{2,2}\right)$$

Intercepts β₀[s] and covariate effects β_cov[s,k] were assigned normal priors with mean 0 and precision 0.001. ψ was expressed as a logistic function of site-specific covariates to account for spatial variation and p conditional on species presence [10]. The model was run using Markov Chain Monte Carlo (MCMC) sampling with 10,000 iterations, a 1000-iteration burn-in, and thinning every 10th sample to reduce autocorrelation. Posterior distributions were generated for all parameters, accounting for both site- and species-specific variations, providing mean estimates, standard deviations, and 95% credible intervals, quantifying uncertainty in estimates [60]. A breakdown of all variables and covariates used in the model is provided in Supplementary Table S2.

Factor Analysis of Mixed Data (FAMD) and Generalized Linear Models (GLM) were used to examine how camera setup variables influenced capture frequency. FAMD, using FactoMineR for handling categorical and continuous variables and factoextra for visualizations, revealed how variables contributed to detection [61]. The GLMs were run separately for occupancy (quasi-binomial model, as occupancy is proportion-based) and capture frequency (Poisson model, as it represents count data) [62]. The occupancy model assessed the impact of camera setup and trophic levels on species detection probability. The capture frequency model incorporated camera placement (e.g., trails vs. roads) and species richness to explore detection rates. Categorical variables were converted to factors for GLM modelling, and p-values were used to assess significant relationships [63]. The FAMD and GLMs thus provided complementary insights into how camera setup variables influenced capture frequency and occupancy.

Akaike Information Criterion (AIC) was used to compare models predicting species occupancy based on camera setup variables and body size, balancing model complexity and fit [61]. Lower AIC values indicated better models, helping assess key predictors while avoiding overfitting. Stepwise models excluded one variable at a time to evaluate its impact. Additionally, interaction terms between body size and camera setup were tested to examine potential influences on species detection. AIC values were calculated using the AIC function, with ΔAIC scores measuring model differences. Models with ΔAIC < 2 were considered equally competitive, while larger values indicated weaker performance [64]. The best-fitting models were identified, revealing the importance of camera setup and body size in predicting occupancy, and results were compiled into species-specific tables for comparison [61].

3. Results

3.1. Survey Design and Detection Patterns

The camera trap survey consisted of 131 camera trap stations with varying operation durations, ranging from 6 to 250 days (median: 179; SD: 49.12), resulting in a total survey effort of 21,020 trap days (Figure S1). Autocorrelation analysis of species-specific activity patterns supported a 30 min threshold for reducing dependent detections across a range of species (Figure S2). Species exhibiting positive autocorrelation, indicating frequent detections, included bushbuck (Tragelaphus sylvaticus; lag 5: 0.1) and buffalo (Syncerus caffer; lag 2: 0.05). In contrast, aardvark (Orycteropus afer: lag 1: −0.18; lag 2: −0.41) and impala (Aepyceros melampus: lag 1: −0.05; lag 2: −0.05) exhibited low or negative autocorrelation, reflecting sporadic activity patterns. Species with moderate activity, such as leopard (Panthera pardus: lag 1: 0.36; lag 4: 0.11) and grysbok (Raphicerus melanotis: lag 1: 0.11; lag 2: 0.05), showed an initial positive autocorrelation that declined rapidly.

3.2. Capture Frequency and Species Occupancy

Of the 34 unique wild mammal species detected (Table 1), chacma baboon (Papio ursinus) was the most widespread species recorded at 126 sites and had the highest detection rates, while grey rhebok (Pelea capreolus) had the most limited distribution. The most abundant large herbivore was kudu (Tragelaphus strepsiceros), detected at 112 sites. Capture frequency varied widely, ranging from 5 to 655 captures per deployment (mean: 161.70). The distribution of capture events across trophic levels further indicated significant variation in species detectability where large herbivores accounted for the largest proportion of total captures (28.3%), followed by small herbivores (21.5%), omnivores (17.2%), carnivores (15.5%), and domestic animals (13.1%), while insectivores were the least detected, comprising only 4.4% (Figure 2).

Occupancy modelling (ψ) and detection probability (p) provided valuable insights into species presence across various deployments with respect to camera setup parameters (Table 1; Figure S3). Black-backed jackal (Lupulella mesomelus), honey badger (Mellivora capensis), and baboon exhibited the highest occupancy with narrow confidence intervals, suggesting that these species are widely distributed and highly detectable within the study area and camera setup parameters. Other species, such as kudu, small grey mongoose (Herpestes pulverulentus), and African wild cat (Felis silvestris lybica), also demonstrated high occupancy, though their wider confidence intervals indicate variability. Several species, including bushpig (Potamochoerus larvatus), porcupine (Hystrix africaeaustralis), grey duiker (Sylvicapra grimmia), and klipspringer (Oreotragus oreotragus), show moderate occupancy estimates. Species with very low occupancy values include aardwolf (Proteles cristata), bat-eared fox (Otocyon megalotis), and red rock hare (Pronolagus saundersiae). Detection probability (p) varied across deployments (Figure 3), with leopard and kudu demonstrating higher detection rates, while otter and red rock hare exhibited high variability in detection. Substantial variability in confidence intervals highlights fluctuations in species presence and detectability across different habitats and camera setup parameters.

Table 1. The distribution and detection patterns of various species, categorized by their common and Latin names, taxonomic order, and trophic classifications (taxonomic order). This table provides a summary of species detection metrics across multiple camera deployments, including the number of capture events, number of sites detected, capture rate, naïve occupancy (Naïve ψ), mean occupancy estimates (psi), and detection probability (p) in relation to camera setup parameters with 95% confidence intervals (in brackets).

Latin Name	Common Name	Taxonomic Order	Captures	Sites	CaptureRate	Naïve(ѱ)	Occupancy (psi)	Detection (p)
Orycteropus afer	aardvark	Insectivore	6	4	0.05	0.03	0.29 (−0.51, 1.08)	0.08 (−0.54–0.70)
Proteles cristata	aardwolf	Insectivore	40	10	0.31	0.08	0.07 (−0.02, 0.15)	0.28 (−0.07–0.62)
Felis silvestris lybica	African wild cat	Carnivore	14	7	0.11	0.05	0.75 (0.05, 1.45)	0.27 (−0.04–0.57)
Papio ursinus	baboon	Omnivore	6327	126	48.3	0.96	0.94 (0.85, 1.03)	0.37 (0.04–0.79)
Otocyon megalotis	bat-eared fox	Insectivore	74	12	0.56	0.09	0.08 (−0.03, 0.19)	0.30 (0.01–0.58)
Canis mesomelas	black backed jackal	Carnivore	21	16	0.16	0.12	0.96 (0.78, 1.14)	0.26 (−0.01–0.53)
Damaliscus pygargus	bontebok	Large Herbivore	28	5	0.21	0.04	0.03 (−0.03, 0.09)	0.19 (−0.20–0.60)
Tragelaphus sylvaticus	bushbuck	Large Herbivore	2353	58	17.96	0.44	0.42 (0.24, 0.60)	0.31 (0.05–0.55)
Potamochoerus larvatus	bushpig	Large Herbivore	357	49	2.73	0.37	0.36 (0.19, 0.53)	0.29 (0.07–0.55)
Syncerus caffer	cape buffalo	Large Herbivore	622	39	4.75	0.30	0.29 (0.13, 0.46)	0.36 (0.07–0.70)
Caracal caracal	caracal	Carnivore	43	25	0.33	0.19	0.46 (0.14, 0.79)	0.23 (−0.04–0.52)
Taurotragus oryx	eland	Large Herbivore	47	5	0.35	0.06	0.18 (−0.23, 0.56)	0.05 (−0.03–0.13)
Oryx gazella	gemsbok	Large Herbivore	231	4	1.76	0.09	0.49 (−0.48, 1.47)	0.44 (−0.13–0.62)
Sylvicapra grimmia	grey duiker	Small Herbivore	182	25	1.39	0.19	0.18 (0.04, 0.32)	0.30 (0.06–0.56)
Herpestes pulverulentus	grey mongoose	Insectivore	33	22	0.25	0.17	0.82 (0.27, 1.38)	0.31 (0.07–0.57)
Pelea capreolus	grey rhebok	Small Herbivore	15	2	0.11	0.02	0.02 (−0.04, 0.07)	0.22 (−0.28–0.68)
Raphicerus melanotis	grysbok	Small Herbivore	248	43	1.89	0.33	0.31 (0.14, 0.48)	0.26 (−0.01–0.54)
Mellivora capensis	honey badger	Carnivore	36	24	0.27	0.18	0.95 (0.80, 1.10)	0.30 (0.07–0.55)
Aepyceros melampus	impala	Large Herbivore	30	6	0.23	0.05	0.05 (−0.04, 0.14)	0.20 (−0.13–0.53)
Oreotragus oreotragus	klipspringer	Small Herbivore	150	25	1.15	0.19	0.17 (0.03, 0.30)	0.29 (0.06–0.57)
Tragelaphus strepsiceros	kudu	Large Herbivore	3886	112	29.66	0.86	0.84 (0.71, 0.96)	0.31 (0.07–0.58)
Genetta tigrina	large-spotted genet	Carnivore	58	23	0.44	0.18	0.19 (0.02, 0.37)	0.29 (0.05–0.56)
Panthera pardus	leopard	Carnivore	400	68	3.05	0.52	0.51 (0.32, 0.71)	0.32 (0.09–0.57)
Redunca fulvorufula	mountain reedbuck	Small Herbivore	142	22	1.08	0.17	0.08 (−0.02, 0.18)	0.26 (−0.03–0.55)
Equus zebra	mountain zebra	Large Herbivore	91	13	0.69	0.09	0.08 (−0.02, 0.18)	0.25 (−0.03–0.55)
Aonyx capensis	otter	Carnivore	5	3	0.04	0.02	0.08 (−0.27, 0.44)	−0.01 (−0.74–0.63)
Ictonyx striatus	polecat	Carnivore	10	4	0.08	0.03	0.03 (−0.09, 0.15)	0.18 (−0.28–0.61)
Hystrix africaeaustralis	porcupine	Small Herbivore	213	43	1.63	0.33	0.33 (0.13, 0.53)	0.31 (0.08–0.57)
Alcelaphus buselaphus	red hartebeest	Large Herbivore	14	7	0.11	0.05	0.73 (0.12, 1.34)	0.19 (−0.29–0.62)
Pronolagus saundersiae	red rock hare	Small Herbivore	23	5	0.18	0.04	0.04 (−0.05, 0.12)	0.17 (−0.27–0.60)
Procavia capensis	rock hyrax	Small Herbivore	59	12	0.45	0.09	0.09 (−0.01, 0.19)	0.29 (0.04–0.57)
Lepus saxatilis	scrub hare	Small Herbivore	514	49	3.92	0.37	0.36 (0.19, 0.54)	0.29 (0.07–0.54)
Antidorcas marsupialis	springbok	Small Herbivore	215	8	1.64	0.06	0.05 (−0.03, 0.12)	0.17 (−0.24–0.60)
Chlorocebus pygerythrus	vervet monkey	Omnivore	785	62	5.99	0.47	0.44 (0.27, 0.61)	0.30 (0.07–0.56)

Table 1. The distribution and detection patterns of various species, categorized by their common and Latin names, taxonomic order, and trophic classifications (taxonomic order). This table provides a summary of species detection metrics across multiple camera deployments, including the number of capture events, number of sites detected, capture rate, naïve occupancy (Naïve ψ), mean occupancy estimates (psi), and detection probability (p) in relation to camera setup parameters with 95% confidence intervals (in brackets).

Latin Name	Common Name	Taxonomic Order	Captures	Sites	CaptureRate	Naïve(ѱ)	Occupancy (psi)	Detection (p)
Orycteropus afer	aardvark	Insectivore	6	4	0.05	0.03	0.29 (−0.51, 1.08)	0.08 (−0.54–0.70)
Proteles cristata	aardwolf	Insectivore	40	10	0.31	0.08	0.07 (−0.02, 0.15)	0.28 (−0.07–0.62)
Felis silvestris lybica	African wild cat	Carnivore	14	7	0.11	0.05	0.75 (0.05, 1.45)	0.27 (−0.04–0.57)
Papio ursinus	baboon	Omnivore	6327	126	48.3	0.96	0.94 (0.85, 1.03)	0.37 (0.04–0.79)
Otocyon megalotis	bat-eared fox	Insectivore	74	12	0.56	0.09	0.08 (−0.03, 0.19)	0.30 (0.01–0.58)
Canis mesomelas	black backed jackal	Carnivore	21	16	0.16	0.12	0.96 (0.78, 1.14)	0.26 (−0.01–0.53)
Damaliscus pygargus	bontebok	Large Herbivore	28	5	0.21	0.04	0.03 (−0.03, 0.09)	0.19 (−0.20–0.60)
Tragelaphus sylvaticus	bushbuck	Large Herbivore	2353	58	17.96	0.44	0.42 (0.24, 0.60)	0.31 (0.05–0.55)
Potamochoerus larvatus	bushpig	Large Herbivore	357	49	2.73	0.37	0.36 (0.19, 0.53)	0.29 (0.07–0.55)
Syncerus caffer	cape buffalo	Large Herbivore	622	39	4.75	0.30	0.29 (0.13, 0.46)	0.36 (0.07–0.70)
Caracal caracal	caracal	Carnivore	43	25	0.33	0.19	0.46 (0.14, 0.79)	0.23 (−0.04–0.52)
Taurotragus oryx	eland	Large Herbivore	47	5	0.35	0.06	0.18 (−0.23, 0.56)	0.05 (−0.03–0.13)
Oryx gazella	gemsbok	Large Herbivore	231	4	1.76	0.09	0.49 (−0.48, 1.47)	0.44 (−0.13–0.62)
Sylvicapra grimmia	grey duiker	Small Herbivore	182	25	1.39	0.19	0.18 (0.04, 0.32)	0.30 (0.06–0.56)
Herpestes pulverulentus	grey mongoose	Insectivore	33	22	0.25	0.17	0.82 (0.27, 1.38)	0.31 (0.07–0.57)
Pelea capreolus	grey rhebok	Small Herbivore	15	2	0.11	0.02	0.02 (−0.04, 0.07)	0.22 (−0.28–0.68)
Raphicerus melanotis	grysbok	Small Herbivore	248	43	1.89	0.33	0.31 (0.14, 0.48)	0.26 (−0.01–0.54)
Mellivora capensis	honey badger	Carnivore	36	24	0.27	0.18	0.95 (0.80, 1.10)	0.30 (0.07–0.55)
Aepyceros melampus	impala	Large Herbivore	30	6	0.23	0.05	0.05 (−0.04, 0.14)	0.20 (−0.13–0.53)
Oreotragus oreotragus	klipspringer	Small Herbivore	150	25	1.15	0.19	0.17 (0.03, 0.30)	0.29 (0.06–0.57)
Tragelaphus strepsiceros	kudu	Large Herbivore	3886	112	29.66	0.86	0.84 (0.71, 0.96)	0.31 (0.07–0.58)
Genetta tigrina	large-spotted genet	Carnivore	58	23	0.44	0.18	0.19 (0.02, 0.37)	0.29 (0.05–0.56)
Panthera pardus	leopard	Carnivore	400	68	3.05	0.52	0.51 (0.32, 0.71)	0.32 (0.09–0.57)
Redunca fulvorufula	mountain reedbuck	Small Herbivore	142	22	1.08	0.17	0.08 (−0.02, 0.18)	0.26 (−0.03–0.55)
Equus zebra	mountain zebra	Large Herbivore	91	13	0.69	0.09	0.08 (−0.02, 0.18)	0.25 (−0.03–0.55)
Aonyx capensis	otter	Carnivore	5	3	0.04	0.02	0.08 (−0.27, 0.44)	−0.01 (−0.74–0.63)
Ictonyx striatus	polecat	Carnivore	10	4	0.08	0.03	0.03 (−0.09, 0.15)	0.18 (−0.28–0.61)
Hystrix africaeaustralis	porcupine	Small Herbivore	213	43	1.63	0.33	0.33 (0.13, 0.53)	0.31 (0.08–0.57)
Alcelaphus buselaphus	red hartebeest	Large Herbivore	14	7	0.11	0.05	0.73 (0.12, 1.34)	0.19 (−0.29–0.62)
Pronolagus saundersiae	red rock hare	Small Herbivore	23	5	0.18	0.04	0.04 (−0.05, 0.12)	0.17 (−0.27–0.60)
Procavia capensis	rock hyrax	Small Herbivore	59	12	0.45	0.09	0.09 (−0.01, 0.19)	0.29 (0.04–0.57)
Lepus saxatilis	scrub hare	Small Herbivore	514	49	3.92	0.37	0.36 (0.19, 0.54)	0.29 (0.07–0.54)
Antidorcas marsupialis	springbok	Small Herbivore	215	8	1.64	0.06	0.05 (−0.03, 0.12)	0.17 (−0.24–0.60)
Chlorocebus pygerythrus	vervet monkey	Omnivore	785	62	5.99	0.47	0.44 (0.27, 0.61)	0.30 (0.07–0.56)

3.3. Species Richness and Accumulation

Biodiversity indices revealed moderate species richness with considerable variability in species distribution across deployments (Figure 3). Species richness per deployment ranged from 1 to 18 (mean: 7.14). The Shannon index varied from 0 to 2.28 (mean: 1.25), indicating uneven species distribution, with some sites highly diverse and others dominated by one or two species. The Simpson index ranged from 0 to 0.84 (mean: 0.61), reflecting variation in species dominance. The Jaccard Similarity Index (mean: 0.30; median: 0.31) showed a moderate overlap in species composition, with similarity ranging from 12% to 39.9%, with most deployments falling between 27.21% and 34.24%. Evenness scores ranged from 0.09 to 1 (mean: 0.69), highlighting variation, with some areas dominated by a few species and others supporting more balanced communities.

The species accumulation curve estimated an asymptote at 153 days (95% CI: 128–179), indicating sufficient monitoring duration for most species. However, rare or cryptic species such as the African wild cat and grey rhebok required extended monitoring periods (Figure 4). While most species were detected within the average monitoring period, site-level variability was high, with asymptotes per deployments ranging from 10 to 401 days and an average species count of 9.89 (range: 2–22; SD: 3.76). Common species like baboon and kudu were detected early, driving the initial rise in the accumulation curve, while rarer species like aardvark and grey rhebok contributed to slower detection rates.

3.4. Influence of Camera Setup

FAMD results showed that the first five dimensions explained 43.63% of the variance in capture frequency (Figure S4). Dim.1 (13.85%) was strongly influenced by large herbivores (21.93%) and species richness (21.78%). Dim.2 (9.65%) was driven by camera elevation, particularly for smaller species, while Dim.3 (7.40%) reflected the effects of camera angle and survey duration. Camera placement on roads increased capture frequency but reduced detection probability. For occupancy, low cos2 values (0.002) were observed, indicating limited contribution of camera setup parameters to occupy estimates.

The GLM analysis showed significant influences of trophic levels and camera setup parameters on capture frequency. Specifically, carnivores were associated with increased capture rates (0.08; p < 0.00), while insectivores showed a decline (−0.03; p < 0.00). Large herbivores, omnivores, and small herbivores positively influenced detection rates. Camera placement on roads (−1.21; p < 0.00) negatively impacted species richness. The choice of camera model also influenced detection rates, with Bushnell and Acorn models resulting in lower capture rates compared to Cuddeback cameras. Camera elevation had a strong influence, with heights set between 40 and 70 cm increasing capture rates, while heights above 100 cm reduced them, particularly for small-bodied species. Optimal detection occurred at cameras angled between 50° and 80°, with extreme angles reducing capture rates. North- and south-facing cameras performed best, while west-facing orientations were less effective.

The Akaike Information Criterion (AIC) results varied across species (

Table 2. Summary of AIC values and ΔAIC (in brackets) for species occupancy models evaluating the influence of different camera setup variables (mounting, angle, elevation, and trail) and body size on species detection. Models compared include the following: Model 1 (all variables included), Model 2 (excluding camera elevation), Model 3 (excluding camera angle), Model 4 (excluding camera trail), Model 5 (excluding camera mounting), and Model 6 (focusing on body size).

Latin Name	Common Name	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Orycteropus afer	aardvark	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Proteles cristata	aardwolf	42.55 (16.00)	41.21 (14.65)	26.55 (0.00)	40.69 (14.13)	38.71 (12.16)	42.55 (16.00)
Felis silvestris lybica	African wild cat	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Papio ursinus	baboon	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Otocyon megalotis	bat-eared fox	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Canis mesomelas	black backed jackal	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Damaliscus pygargus	bontebok	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Tragelaphus sylvaticus	bushbuck	30.00 (16.00)	78.44 (64.44)	14.00 (0.00)	83.14 (69.14)	111.64 (97.64)	30.00 (16.00)
Potamochoerus larvatus	bushpig	47.06 (15.87)	45.40 (14.21)	31.19 (0.00)	43.74 (12.55)	42.32 (11.13)	47.06 (15.87)
Syncerus caffer	cape buffalo	62.72 (10.28)	69.80 (17.35)	52.44 (0.00)	61.66 (9.21)	57.90 (5.46)	62.72 (10.28)
Caracal caracal	caracal	45.95 (14.37)	68.18 (36.60)	31.58 (0.00)	42.36 (10.77)	46.96 (15.37)	45.95 (14.37)
Taurotragus oryx	eland	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Oryx gazella	gemsbok	30.00 (16.00)	31.82 (17.82)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Sylvicapra grimmia	grey duiker	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Herpestes pulverulentus	grey mongoose	54.90 (15.61)	59.68 (20.39)	39.29 (0.00)	51.51 (12.22)	50.14 (10.86)	54.90 (15.61)
Pelea capreolus	grey rhebok	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Raphicerus melanotis	grysbok	822.96 (796.07)	79.41 (52.52)	26.89 (0.00)	69.68 (42.79)	672.79 (645.89)	822.96 (796.07)
Mellivora capensis	honey badger	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Aepyceros melampus	impala	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Oreotragus oreotragus	klipspringer	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	34.77 (20.77)	26.77 (12.77)	30.00 (16.00)
Tragelaphus strepsiceros	kudu	55.41 (7.47)	53.43 (5.49)	47.94 (0.00)	56.76 (8.82)	55.29 (7.35)	55.41 (7.47)
Genetta tigrina	large-spotted genet	30.00 (16.00)	31.82 (17.82)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Panthera pardus	leopard	82.75 (8.82)	81.82 (7.90)	73.93 (0.00)	152.43 (78.50)	78.13 (4.20)	82.75 (8.82)
Redunca fulvorufula	mountain reedbuck	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Equus zebra	mountain zebra	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	45.04 (31.04)	24.00 (10.00)	30.00 (16.00)
Aonyx capensis	otter	35.33 (14.64)	34.28 (13.59)	20.69 (0.00)	36.42 (15.73)	32.81 (12.12)	35.33 (14.64)
Ictonyx striatus	polecat	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Hystrix africaeaustralis	porcupine	60.27 (14.30)	68.25 (22.28)	45.97 (0.00)	56.59 (10.62)	55.44 (9.47)	60.27 (14.30)
Alcelaphus buselaphus	red hartebeest	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Pronolagus saundersiae	red rock hare	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Procavia capensis	rock hyrax	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	36.27 (22.27)	26.77 (12.77)	30.00 (16.00)
Lepus saxatilis	scrub hare	32.77 (16.00)	34.28 (17.51)	16.77 (0.00)	34.57 (17.80)	34.66 (17.89)	32.77 (16.00)
Antidorcas marsupialis	springbok	30.00 (16.00)	28.00 (14.00)	14.00 (0.00)	26.00 (12.00)	24.00 (10.00)	30.00 (16.00)
Chlorocebus pygerythrus	vervet monkey	84.00 (4.54)	82.76 (3.31)	79.46 (0.00)	139.20 (59.74)	108.76 (29.30)	84.00 (4.54)

). Model 1, which included all parameters, fit well for most species but not optimally for aardvark and aardwolf, which had higher AIC values. Excluding camera elevation (Model 2) had minimal impact on bushbuck and buffalo, indicating low relevance for this parameter. For leopard, excluding camera trail (Model 4) significantly reduced model performance, highlighting its importance. Model 3, which excluded camera angle, performed well overall. Excluding camera mounting (Model 5) had little effect on most species but significantly reduced model performance for grysbok. Model 6, which incorporated interaction terms between body size and camera setup variables, improved accuracy for species like vervet monkey, grysbok, and porcupine.

4. Discussion

4.1. Biodiversity in the Baviaanskloof Catchment

The detection of 34 mammal species in the Baviaanskloof catchment highlights the ecological value of this region, despite its mixed-use landscape. Species richness recorded here is comparable to other South African regions, such as KwaZulu-Natal (38 species) [65] and the Fish-Kowie corridor in the Eastern Cape (33 species) [40]. However, fewer species have been reported in the fynbos biome (13–27) [38,66]. While the relationship between vegetation type and diversity was not explicitly analysed in this study, camera traps were deployed across the major habitat types within the Baviaanskloof, including thicket, fynbos, forest, and savanna. The species richness observed reflects the region’s ecological diversity, which spans seven of South Africa’s eight biomes, including the globally recognized biodiversity hotspots of the Fynbos and subtropical thicket [29].

High species richness at some sites likely reflects more complex habitat structures and favourable conditions, whereas lower richness may indicate habitat degradation or reduced complexity. Furthermore, studies suggest that specific agricultural practices, such as low-intensity cattle (Bos taurus) grazing, can sustain mammal species richness comparable to that of natural areas [67]. Notably, carnivores such as the leopard, caracal, and honey badger were positively associated with higher capture frequency, suggesting robust ecosystem functionality through their influence on prey populations and roles within complex food webs [68,69]. This is consistent with previous findings that predator presence correlates with higher mammalian species richness [70]. Conversely, insectivores were associated with lower capture frequencies, which may reflect a preference for less-disturbed environments with reduced competition and predation pressures [67]. This pattern highlights the potential influence of habitat disturbance and modification on detection rates, particularly for species with specialized habitat requirements. Chacma baboon and greater kudu exhibited high occupancy and capture rates, consistent with their adaptability and wide-ranging behaviour [71]. In contrast, elusive species, such as the African wild cat, polecat, and grey rhebok, demonstrated low occupancy and detection rates, likely reflecting their cryptic habits and specific habitat requirements [72]. These findings emphasize the value of using occupancy models that account for imperfect detection [73,74].

The use of a 30 min interval between detections effectively balanced data independence with temporal resolution, reducing the risk of overestimating detections from clustered events [75,76]. However, for species exhibiting high site fidelity or repeated movements within short timeframes, such as the bat-eared fox, a 60 min interval may be more appropriate. This suggests that customizing intervals, based on species-specific movement patterns, could enhance occupancy and detection estimates for certain taxa [47].

Biodiversity indices, designed to capture key characteristics of communities, enabling comparisons across different regions, species groups, and trophic levels [77], revealed moderate species richness and substantial variability in species evenness across the Baviaanskloof catchment. High species evenness and richness appear to support more balanced communities, whereas areas with lower diversity may indicate habitat limitations or greater disturbance [1]. Landscape heterogeneity and human disturbance likely influence species distributions, particularly for herbivores and smaller carnivores that may be more sensitive to habitat modifications.

4.2. Influence of Camera Trap Configurations

This study underscores the importance of carefully configured camera trap setups to enhance wildlife species detection and monitoring efficiency. Our findings demonstrate that camera elevation, angle, orientation, and placement significantly influence detection rates [78]. Consistent with previous research, our study reaffirms that excessive camera heights diminish detection rates and compromise data quality [79,80]. Camera elevations between 40 and 70 cm produced higher capture rates, whereas placements exceeding 100 cm reduced detections, particularly for smaller-bodied species [81]. This supports established guidelines recommending lower camera placements to enhance detection probabilities for a broader range of species or tailored heights to specific target species [82,83]. Optimal detection angles ranged from 50° to 90°, with steeper angles, whether upward or downward, reduced detections by limiting the effective field of view. Aligning sensors with the anticipated body mass centre of the target species and ensuring the lens remains perpendicular to the ground, improves detection probability [84,85,86]. However, the optimal camera angle is highly dependent on the study’s objectives and should be adjusted according to the target species [85].

Camera orientation also plays a critical role in detection efficiency, particularly when considering sunlight interference. North- and south-facing cameras yielded higher capture rates, likely due to more consistent lighting conditions and reduced glare during sunrise or sunset. This observation is consistent with previous studies suggesting that orientation can significantly influence detection probability by minimizing false triggers and overexposure [84,87]. In contrast, west-facing cameras were less effective, particularly during sunset when the glare was most pronounced [84,87].

Cameras positioned along roads captured more frequent but less diverse detections, particularly for species that frequently traverse these pathways, such as kudu, bushbuck, baboon, and leopard, indicating movement hotspots rather than true presence [67,88]. In contrast, off-road camera placements on animal trails provided a broader representation of species, including cryptic carnivores such as the honey badger, that tend to avoid areas of high activity [67,88]. This highlights the importance of balancing road and off-road camera placements on animal trails to capture a wider range of species behaviours and habitat preferences, particularly in ecosystems characterized by high species richness and ecological variability [24,89]. The choice of camera model also influenced detection rates, with the Bushnell and Acorn models yielding lower capture frequencies than the Cuddeback model. This discrepancy may be attributed to the uneven distribution of deployments and differences in technical specifications, such as trigger speed, detection range, and image quality [21,90]. However, it highlights the need for standardizing camera models across survey sites to ensure consistency in detection rates [22,85].

Our stratified, random placement design aimed to provide broad spatial coverage across diverse vegetation types, including thicket, fynbos, forest, and savanna. However, we acknowledge that our study did not employ a systematic gradient-based testing approach where camera height, angle, and orientation were varied at the same site to evaluate their effects independently. Although our approach allowed us to evaluate detection rates across various habitat types, it does not fully isolate the effects of camera placement parameters from habitat-related variability [21,23,91]. The distribution of camera angles and installation heights was also uneven, primarily due to practical constraints during deployment, such as the availability of flat surfaces or the necessity to angle cameras to avoid flood risk and optimize views in challenging terrain. While our statistical models incorporated camera setup parameters, habitat type, and landscape units as covariates to account for their influence on detection rates, a more balanced sampling design would provide stronger evidence for optimal camera placement practices. Future studies would benefit from a controlled experimental design in which all camera setup parameters are systematically varied within specific habitat types. Such an approach would enhance the ability to isolate the influence of camera configuration and improve the reliability of findings [34,92].

Species accumulation curves indicated that 153 sampling days were sufficient to detect most species, but extended durations improved detection rates for elusive or low-density species. Sites with lower detection variability in common species, such as kudu and baboon, reached asymptotes relatively quickly, whereas rarer species, like the grey rhebok, required longer observation periods to ensure reliable detection. This underscores the need for flexible monitoring durations tailored to species-specific detection probabilities [22]. Alternative guidelines for optimizing species richness, occupancy, and detection estimates with camera trap arrays recommend a standardized approach involving a minimum of 40 camera deployments, with each camera operational for approximately five weeks to achieve precise and efficient monitoring outcomes [19,93,94]. While standardization provides practical guidance, our findings suggest that combining adequate spatial coverage with extended deployment durations where necessary enhances species detection, particularly for elusive or low-density species. Effective monitoring frameworks should be adaptable, considering factors such as the study objective’s target species, habitat complexity, and available resources to optimize the effectiveness and relevance of monitoring efforts [22,76].

5. Conclusions

Our study provides empirical evidence demonstrating the effectiveness of integrating camera setup parameters, habitat variability, and monitoring duration to optimize detection efficiency across the diverse and mixed-use landscape of the Baviaanskloof catchment. This ecologically rich region supports 34 mammal species spanning thicket, fynbos, forest, and savanna habitats, comparable to other South African regions with similar ecological complexity. By examining detection rates across varied habitats, we identified optimal camera configurations to enhance detection efficiency. Camera elevations between 40 and 70 cm produced higher capture rates for medium- to large-sized mammals, while heights exceeding 100 cm reduced detections, especially for smaller-bodied species. North- and south-facing cameras yielded higher capture rates due to consistent lighting conditions, while west-facing orientations were less effective due to solar glare. Cameras positioned along animal trails captures a wider range of species compared to those placed along roads. Species accumulation curves suggest that approximately 153 days of monitoring effectively detect most species, though extended durations improve detection of rare or elusive species. Standardizing camera trap methodologies, while allowing flexibility for species-specific adaptations, can improve monitoring outcomes. Considering seasonal variation may further refine monitoring protocols and reveal temporal shifts in wildlife activity and habitat use. By integrating findings across diverse landscapes and tailoring methodologies to specific species and habitats, researchers can enhance data accuracy and contribute to more effective conservation planning within biodiversity-rich but human-modified environments.