Comparing Effectiveness of AHDriFT Systems and Sherman Traps for Surveying Small Mammals in Northeastern Indiana

Small mammals contribute a great deal to the health and continuity of terrestrial ecosystems. They provide a reliable and important prey base (McCleery et al. 2014), distribute seeds across landscapes (Nelson et al. 2019), and affect local soil structures (Rosenstock 1996; Leis et al. 2007). This highlights the need for conservation and management of small-mammal communities, which are in decline in many landscapes (Gibbs 2000). Survey methods, primarily through forms of trapping, aid in the monitoring and assessment of small-mammal community health (Di Cerbo and Biancardi 2012; Villette et al. 2015; Wearn and Glover-Kapfer 2019). Unfortunately, small-mammal surveys often involve intense effort and resources, requiring funding that is becoming increasingly difficult to obtain.

The trapping methods that surveyors use depend upon the objectives of a given study. Traditional methods of surveying small mammals include pitfall traps (Umetsu et al. 2006), snap traps (Sealander and James 1958; King and Edgar 1977), and metal-box traps (Sherman traps; Rosenstock 1996; Francl et al. 2002). These trapping methods allow for physical assessment of captured individuals and the assessment of population and community structures through the application of tags and other markers. The use of Sherman traps is common among studies that investigate species morphometries (Umetsu et al. 2006), monitor parasites/pathogens (LoGiudice et al. 2003), monitor general demographic parameters (King and Edgar 1977; Rosenstock 1996), and survey small mammal communities (Webala et al. 2006; Bissonette and Rosa 2009). Depending on the landscape and target species, Sherman traps are more successful at capturing small mammals than some types of live traps (e.g., Victor traps and pitfall traps; Sealander and James 1958; Torre et al. 2010) but less successful than others (e.g., mesh and Longworth traps; O'Farrell et al. 1994; Jung 2016). Additionally, Sherman traps can show a greater ability to capture a wide array of species than other traps, including larger and highly mobile species (Beckmann 2022). Sherman trap success can vary on the basis of trap characteristics (Slade et al. 1993) and remnant scents from previous tenants (Wolf and Batzli 2002). Although beneficial to studying individuals, active trapping methods like Sherman traps can stress animals (Enge 2001; Bosson et al. 2012) and have demonstrated a 20–50% mortality rate for some species (Stephens and Anderson 2014). This places an upper limit on the effectiveness of the trap at surveying populations and communities (Amber et al. 2021). For example, researchers worried about unintentional mortality may seek to decrease the number of trap nights (one trap set on one calendar night) they conduct. However, survey results are limited when trapping nights are low, as only one individual has access to a trap during each survey (Trapp and Flaherty 2017; Wearn and Glover-Kapfer 2019; although see Bergstrom and Sauer 1986; Feldhamer et al. 2008 for exceptions). Low trap nights can also reduce a surveyor's ability to detect rare or uncommon species when using traditional trapping methods. For example, Palmeirim et al. (2020) indicate that 8 calendar nights of trapping are necessary to capture most common small mammal species, whereas more than 10 nights were needed to capture rare species in fragmented Brazilian forests. Additionally, the number of traps deployed is limited by the bulkiness of the traps themselves, as many traps can be difficult to transport to field sites (Sweetapple and Nugent 2011). Some species or individuals may also actively seek out or avoid traps, causing results to be skewed toward increased recaptures or reduced community diversity (Parsons and Bondrup-Nielsen 1996; Wolf and Batzli 2002).

As an alternative to live traps, camera trapping has evolved and flourished as an effective manner to passively survey fauna since its early 20th-century introduction (Cutler and Swann 1999; Srbek-Araujo and Chiarello 2005; Burton et al. 2015). Camera traps have been used to study nest predation (Cutler and Swann 1999), feeding rates (Di Cerbo and Biancardi 2012; Trapp and Flaherty 2017), overall species presence (Di Cerbo and Biancardi 2012; McCleery et al. 2014), and seasonal activity patterns (Cutler and Swann 1999; Srbek-Araujo and Chiarello 2005). A significant benefit of camera traps is their lack of required time in the field, making them a better option for researchers lacking time or funds (Cutler and Swann 1999). Camera traps also help reduce stress to animals and species and observer trapping bias by reducing encounters in which researchers directly interact with individuals in the ecosystem (Cutler and Swann 1999). In part, by reducing stress, camera trapping also reduces the risk of individual mortality (Littlewood et al. 2021), which is especially important when working with at-risk species. However, camera trapping does not allow for the physical assessment of individuals without the addition of scaling objects within view of the camera and morphometric analyses (Pérez-Flores et al. 2016; Tarugara et al. 2019). Limitations to camera trap surveys include high start-up costs based on camera prices (Meek et al. 2015), the difficulty of identifying some species in images (McCleery et al. 2014), and potential problems with repeatability due to camera model discontinuations. Despite these limitations, the continuous collection of data appeals to many researchers. Unfortunately, camera traps are a typically unreliable method of surveying small mammals because of the difficulties in capturing quality pictures of such small individuals (Di Cerbo and Biancardi 2012; Littlewood et al. 2021). However, the recently developed adapted-Hunt drift fence technique (AHDriFT) has been used to successfully survey a variety of small vertebrates (Martin et al. 2017; Amber et al. 2021).

The AHDriFT system consists of a drift fence (multiple design options) used to funnel small and medium-sized terrestrial animals toward one of two Hunt traps equipped with passive-infrared sensor cameras. Hunt traps (McCleery et al. 2014) house a camera in the top of overturned buckets with opposite entry and exit openings to allow animals to move through camera views freely. In one of the first studies using this method, Martin et al. (2017), using the AHDriFT system with linear-style drift fences, successfully identified vertebrates (including small mammals) within the coastal dunes of Florida. They also reduced researcher time in the field by 95%, compared with a previously conducted active trapping method in the same area. Amber et al. (2021) used the AHDriFT system (Y-shaped drift fences) in numerous wet meadow sites and noted a greater overall biodiversity and a greater detection rate of rare mammal and reptile species than that of previous studies that use only traditional trapping techniques (e.g., Sherman traps, snap traps, pitfall traps). Although the system seems to be an effective replacement for traditional active trapping methods, it may not be effective for all communities in all areas. Additionally, no study that we are aware of has compared the effectiveness of AHDriFT systems and traditional trapping methods used simultaneously. To further evaluate the AHDriFT system's effectiveness, we used the system to survey small mammal communities in a wetland complex restored from crop agriculture in Fort Wayne, Indiana. Specifically, our objective was to compare the AHDriFT system's efficiency and effectiveness at surveying small-mammal community diversity with that of traditional Sherman trap surveys. We compared community structure indices, costs, and time spent between the two methods.

We conducted this study at the Eagle Marsh Nature Preserve (41°2′16.8″N, 85°13′35.76″W; WGS84) in Allen County, Indiana. Eagle Marsh is a 336-ha natural area in southwestern Fort Wayne surrounded by urban and agriculturally dominated rural areas. The marsh is bordered by medium-density urban land cover to the north, a landfill to the east, a forest to the south, and Interstate 69 to the west. A water treatment plant is located on the north border. The southern border runs parallel to another natural area (Fox Island County Park), but the two areas are separated by a railroad track (Barton et al. 2020).

Eagle Marsh is a reclaimed agricultural field that was purchased by the Little River Wetland Project in 2005. The restoration of the area began in 2006; agricultural drain tiles and pumps were removed, and native plants and trees were planted (Ruch et al. 2016). Since then, the Little River Wetlands Project continues to protect and manage the area through invasive plant control and prescribed burns. The area also has an active trail system and an education center that receives variable use throughout the year. We surveyed small mammal communities at four sites within Eagle Marsh, each within varied habitat types. Sites were within a young, planted bur oak Quercus macrocarpa forest, a mesic broad-leaf forest next to a temporal wetland, a mesic prairie dominated by pokeweed Phytolacca americana, and a tallgrass prairie (e.g., reed canary grass and big bluestem; Phalaris arundinacea and Andropogon gerardii, respectively) bordered by a cattail Typha × glauca hybrid marsh. Sites were located approximately 0.51 ± 0.36 km apart (mean ± SD; range 0.13−0.85 km). The mean daily temperature during the project period (May to August 2020 and 2021) was 21.0°C in 2020 and 20.9°C in 2021. The sum precipitation from May through August was 402 mm in 2020 and 548 mm in 2021 (Fort Wayne International Airport Weather Station; KFWA).

We collected data from both AHDriFT systems and Sherman traps from 23 February to 25 July 2020 and 28 May to 27 August 2021. We installed AHDriFT systems to three of the four sites in early February 2020 and collected images continuously until the end of the study period. We abandoned our tallgrass prairie site in April 2020, after a prescribed burn destroyed the drift fence. We added a system to a fourth site (mesic prairie) in May 2021. Two or three sites were active at a time.

We designed AHDriFT systems on the basis of Martin et al. (2017). They consisted of a single staked, 10-m-long × 1-m-high drift fence (plastic silt fence buried 15 cm into the ground to prevent burrowing) with an overturned 5-gallon bucket containing a trail camera (model BTC-4P Command Ops Pro; Browning, Morgan, UT) at either end (Figure 1). We affixed the cameras to the roof of the overturned buckets (machine bolts affixed to L brackets superglued to cameras; wing nuts affixed bolts to bucket) and aligned their lenses to the center of the floor. We affixed two 1 × 2 in. pine boards to the inside of the buckets to funnel animals under the lenses. To prevent animals from escaping through the openings between the drift fence and the buckets, we affixed two 1 × 4 in. boards to the outside of the buckets. Camera settings included 8-MP image size, 10-s capture delay between triggering events, and 0.5-s trigger speed. We set cameras to take three-image bursts whenever they were triggered. We affixed 2.75−3.75 diopter reading glass lenses to the camera lenses using electrical tape to reduce the cameras' focal distances to the bottom of the bucket where the animals were located (approximately 30 cm from the camera lens; Figure S1, Supplemental Material). We used one AHDriFT system, consisting of two cameras inside buckets with one drift fence stretching between them, per site. We checked our AHDriFT systems weekly to ensure proper function and collected image data monthly. We considered cameras at either end of the drift fences to be part of the overall AHDriFT system; thus we pooled image data from both cameras.

We conducted Sherman trap surveys at each site approximately once a week throughout the project period. At each site, Sherman traps (7.6 × 7.6 × 25.4 cm nonfoldable traps) were deployed in a 10-trap, 50-m straight transect that ran parallel to, but 5 m to the side of, the site's AHDriFT system. We did not combine Sherman traps with drift fences, as five drift fences would be required to obtain the same number of trap nights. We baited each Sherman trap with a mixture of peanut butter, rolled oats, and wild bird seed mix (only sunflower seeds were used in 2020) and added a fist-sized ball of polyester filling (more during colder weather) to the traps before setting in the field. Once set in the field, we placed a roofing shingle upon each trap to provide weatherproofing. We deployed traps in the evening (approximately 1600 hours) and checked them the next morning (approximately 0900 hours). We determined the species, sex, age, and reproductive status of each captured small mammal (Petit and Waudby 2012; Orrock 2021). We washed and added new bait and filling to traps before each survey to avoid adding scent bias (Boonstra and Krebs 1976). All handling of captured mammals was done in compliance with the guidelines of the American Society of Mammalogists for the use of wild mammals in research (Sikes 2016), Indiana State Collections Permits (permits 3203 and 2667), and institution animal care and use permits (1902001857).

We distributed images downloaded from cameras among a group of data collectors who then extracted data from each image. Image data included presence of animal, species of animal if applicable, date and time of image, and internal AHDriFT system temperature. We ensured that two observers viewed each image; interobserver agreement on species identification was 87 ± 10%. When we only used images containing enough detail for identification purposes, percent agreement reduced to 74 ± 22%. Because we expected this low interobserver agreement in identification, we cross-checked images with disagreement to ensure the quality of results. To avoid the inclusion of multiple observations of the same individuals in our analyses of AHDriFT system images, we calculated the number of unique capture events (UCEs) of each species for each survey night. We considered an image observation to be a UCE if the species depicted in the image had not been observed at the same site in the last 60 min (Amber et al. 2021).

We pooled data between years, despite applying different types of baits to Sherman traps each year. This likely did not affect the results of our analyses, as species richness (Z = −0.49, P = 0.63), species evenness (Z = −0.20, P = 0.84), and Shannon diversity index (Z = −0.20, P = 0.84) calculated from Sherman trap catches per trap night did not differ by year (Wilcoxon sign-ranked tests). Because Sherman trap surveys were conducted approximately once a week, we pooled AHDriFT system UCE and Sherman trap survey data by survey period. A single survey period consisted of all data collected between Sherman trap transect deployments, starting with the day Sherman traps were checked (6.2 ± 2.3 d; range 3−12 d) at a single site. We did not include the AHDriFT system data collected during the night that Sherman traps were deployed in our analyses to avoid bias resulting from animals being contained within the traps rather than traversing the landscape. To make survey types more comparable, we filtered AHDriFT system UCE data into nighttime data (images captured between 1600 and 1000 hours), as this was the time that Sherman traps were accessible to small mammals. We calculated survey period species richness (R), evenness (J), and Shannon diversity index (H) for Sherman trap data and AHDriFT system data collected throughout the entire day and at night. To remove variation caused by variable trapping effort between the methods, we weighted community diversity characteristics by the number of trap nights conducted during survey periods. Specifically, we weighted Sherman trap community characteristics by the number of traps set during the survey period (10 traps per transect, one transect per survey period) and AHDriFT system characteristics by the number of days within the survey period minus the day the Sherman traps were deployed.

The data were positively skewed and we were unable to normalize it through logarithmic transformations; therefore we used paired Wilcoxon sign-ranked tests to compare survey period community diversity characteristics between survey methods. We compared data both weighted and unweighted by trap nights and compared Sherman trap data with AHDriFT data collected throughout the day and during nighttime only. We also identified the survey method's ability to capture specific small mammal species (deer mouse Peromyscus spp., vole Microtus spp., meadow jumping mouse Zapus hudsonius, eastern chipmunk Tamias striatus, northern short-tailed shrew Blarina brevicauda, common shrew Sorex cinereus, and long-tailed weasel Neogale frenata) by comparing the proportion of survey periods in which they were detected by the two survey methods using Fisher's exact tests. We chose these specific species because they were the only species observed by the AHDriFT systems that could also reasonably be captured by Sherman traps. We generated rarefied species accumulation curves for each survey method, pooling site data to compare how many trap nights each method required to capture rarer species (Gotelli and Colwell 2001). All averaged data are reported as mean ± SD.

We conducted our study for 244 d, 153 d in 2020 and 91 d in 2021. We conducted 640 trap nights of Sherman trapping surveys and captured a total of 192 small mammals of three distinct species (0.3 catches per trap night; Table 1; Data S1, Supplemental Material). We implemented the AHDriFT system for 551 trap nights, recording 18,715 images. We filtered multiburst images into one representative image, resulting in 7,825 images. Of these images, 5,153 (65.8%) consisted of false triggers (images not containing animals or parts of animals; Amber et al. 2021), 858 (10.9%) of the images showed small mammal presence but to an unidentifiable degree, and 1,814 images (23.2%) contained identifiable small mammals of seven distinct species (3.29 observations per trap night; Table 1; Data S2, Supplemental Material). We can attribute the large number of false-trigger images to shifts in daylight or precipitation entering the system. Our AHDriFT systems resulted in 706 UCEs (1.28 UCEs per trap night) when monitoring for a full 24-h period and 532 UCEs (0.97 UCEs per trap night) when data were filtered to nighttime observations. Because of battery failure and destruction by a fallen tree, we lost 49 trap nights of AHDriFT system data.

Analyses resulted in varied outcomes when comparing the effectiveness of survey methods at detecting specific small mammal species. Both survey methods were equally likely to detect deer mice (P = 0.24) and voles (P = 0.32; Figure 2). However, AHDriFT systems were significantly more effective in detecting meadow jumping mice (P < 0.01), long-tailed weasels (P = 0.01), eastern chipmunks (P < 0.01), common shrews (P < 0.01), and northern short-tailed shrews (P < 0.01). Although voles were detected using the AHDriFT system, images did not include enough of the individuals (e.g., ventral pelage) to differentiate between the possible species in the area (eastern meadow vole Microtus pennsylvanicus and prairie vole M. ochrogaster). These species could be differentiated upon capture by Sherman traps; however, we captured no prairie voles during this study. Our AHDriFT systems were also able to detect other vertebrate animals that would not have been observable using Sherman traps (Table 1).

During night hours, comparing catches per trap night, AHDriFT systems recorded two times greater species richness (Z = −6.21, P < 0.01), 16 times greater evenness (Z = −4.83, P < 0.01), and 23 times greater diversity (Z = −4.87, P < 0.01) than Sherman traps at the same sites and during the same survey periods (Table 2). We obtained similar results when comparing community diversity characteristics unweighted by trap nights and when comparing Sherman trap data to AHDriFT system data collected throughout the entire day (Table 2). Species accumulation curves showed that AHDriFT systems detected more species per trap night than Sherman trap surveys (Figure 3). Even after 600 trap nights, Sherman traps never detected the same number of species as AHDriFT systems. The gradual inclined shape and the lack of an asymptote of the Sherman trap curve suggests that additional species may be detected with much more effort. The AHDriFT systems rapidly accumulated species until approximately 100 trap nights and then reached a stable asymptote; suggesting that the method can quickly detect most species in the area.

We spent 2,910 min (108 min per survey period; 36 min per survey period per site; 5 min per trap night) in the field conducting Sherman trap surveys, including time spent baiting, setting, and checking the traps. Although time allotted for transportation will vary per each study, our sites were 40 min away from our lab. Taking travel into account, we spent a total of 4,030 min (149 min per survey period; 49 min per survey period per site; 6 min per trap night) on our Sherman trap surveys. Weighting this number by the total number of identifiable small-mammal observations (192 captured animals), we spent 21 min per observation using Sherman traps, 100% of which was time in the field. We assumed that time entering and managing Sherman trap data are minimal and, thus, did not include it in our analysis.

Although we checked AHDriFT systems during Sherman trap procedures, we recorded times independently. It took 56 min to check AHDriFT cameras at all three sites, totaling 224 min of fieldwork (8 min per survey period; 3 min per survey period per site; 0.4 min per trap night) for the duration of the study. Adding in the time spent on travel, we spent a total of 384 min (14 min per survey period; 5 min per survey period per site; 0.7 min per trap night) collecting our AHDriFT data. It took us 36.8 s to completely process each image. We calculated this number as mean time taken for individuals to analyze an image (24.5 ± 9.3 s per image on the basis of trials conducted on a batch of images) plus 1.5 times additional time to quality control each image. In total, we processed and rechecked 18,715 images from the AHDriFT system, resulting in 11,447 min. This number is likely an overestimate, as we did not quality control images that did not contain animals. Adding in the time spent collecting the data, we spent a total of 11,831 min (438 min per survey period; 146 min per survey period per site; 22 min per trap night) collecting our AHDriFT system data. Weighting this number by the total number of identifiable small-mammal observations (total images = 1,814, UCEs = 532), we spent 7 min per observation and 22 min per UCE using AHDriFT systems. Additionally, it is important to note that we spent only 2% of this time in the field.

Our Sherman trap surveys cost approximately US$2,330 ($86 per survey period; $29 per survey period per site; $3.60 per trap night), including funds for 30 Sherman live traps, bait, polyester filling, roofing shingles, and people hours (Table 3). Unpaid student volunteers conducted our study, so people hours were not part of our overall costs. However, to provide a full comparison on the cost of the methods, we estimated the cost of people hours using the minimum wage at the time of study ($7.25). With 192 unique small mammals captured throughout the duration of the study, we spent approximately $12.14 per capture. The three AHDriFT systems (materials from tallgrass prairie site moved to mesic prairie site) used in this study cost approximately $4,351 ($1,450 per system; $161 per survey period; $54 per survey period per site; $7.90 per trap night; Table 3). Most of this cost was associated with time spent analyzing images. Considering the number of small mammals identified, costs per observation equated to $2.40 per observation and $8.18 per UCE.

This study was designed to be one of the first to compare the recently developed AHDriFT system with the traditional Sherman trap survey method. We detected 23 times higher community diversity using AHDriFT systems than Sherman traps in our midwestern site. Additionally, AHDriFT system surveys were approximately five times less expensive per observation than Sherman trap surveys and 1.5 times less expensive per UCE. Overall, AHDriFT systems reduced time per observation threefold, but when calculated using UCEs the time was similar between techniques. Despite this, AHDriFT systems were still more time efficient when considering time spent in the field. Specifically, we spent 10 times less time in the field conducting AHDriFT system surveys than Sherman traps surveys (90% reduction in field time). This suggests that AHDriFT systems could present a valuable alternative to live trapping for research groups with remote field sites or constraints on available field time.

Sherman trap surveys resulted in reduced community diversity compared with AHDriFT systems. This lower diversity was a result of low estimated species richness and high relative abundance of deer mice (97% of total Sherman trap captures). Deer mouse bias in Sherman trap surveys has been documented in other studies and is caused either by surveying communities dominated by deer mice (Anthony et al. 1981; Phelps and McBee 2009) or a bias in the survey method (Taylor and Lowman 1996; Stephens and Anderson 2014). Although AHDriFT systems did record a high abundance of deer mice, it is unlikely that the species truly dominates Eagle Marsh's small-mammal community to the extent indicated by Sherman trapping. The eastern meadow vole was the most captured species in the area in 2014 and 2017 (Ruch et al. 2016; Hopkins 2018). The relative abundance of eastern meadow voles should not have changed between 2017 and the time of this study, as the availability of habitats has not varied within Eagle Marsh. A crash in eastern meadow vole population between these time periods could have resulted from factors not related to changes in available habitat (e.g., disease, predators, food availability). However, these, primarily regional, factors would have likely affected the deer mouse population as well. Deer mice also likely did not competitively exclude eastern meadow voles from the area, as voles tend to be the better competitors in sympatric landscapes (Redfield et al. 1977; Abramsky et al. 1979). Our site selection may have inflated captures of deer mice, as two of our four sites were in forested areas, habitat preferred by deer mice in our region (Wolf and Batzli 2002), whereas we only collected one summer's worth of data from our prairie site, habitat that eastern meadow voles tend to prefer (Grant 1970; Anthony et al. 1981). However, our site likely did not affect trap bias, as 98.5% of Sherman trap captures from the prairie site were deer mice. If characteristics of the small-mammal community in Eagle Marsh are not the driving force, the inflated presence of deer mice in our Sherman trap surveys may instead be an artifact of inherent trap bias.

A variety of trap-specific factors may inflate or deflate the relative abundance of certain species. For example, scents left behind in Sherman traps after the capture of individuals can be a source of attraction or deterrence. Both deer mice and voles are more likely to be captured by traps that previously held conspecifics and less likely by traps that previously held potential predators (Stoddart 1982; Heske 1987; Parsons and Bondrup-Nielsen 1996; Wolf and Batzli 2002). Although we cleaned all traps after each survey period, we cannot rule out that there may have been remnant scents in the traps that could have biased our catch. Alternatively, the addition of human scent can increase the probability of bait removal by certain rodent species (Duncan et al. 2002). Because we did not use gloves during deployment (we did use gloves during bait application), human scent was applied to our Sherman traps. White-footed field mice Peromyscus leucopus exhibit a strong attraction to new stimuli (Sheppe 1966), whereas voles show variation in attraction to stimuli (Shillito 1963; Boonstra and Krebs 1976). Thus, deer mice may have more readily entered traps than some other species in our site, thereby reducing the total number of traps available to other species and the abundance of other species in our overall catch. Comparatively, AHDriFT systems likely reduced scent bias associated with both wild mammal and human scent. Since mammals spent less time within the system and were less stressed (Eisenberg and Kleiman 1972), they left little scent in the traps. There was also likely less human scent associated with the AHDriFT system, as we only manipulated these traps approximately once a month.

The specific characteristics of our Sherman traps may have also led to deer mouse bias. We used 25.4-cm-long Sherman traps, which are larger than others available. Studies from Virginia (Dalby and Straney 1976) and Wisconsin (Anthony et al. 2005) determined that shorter Sherman traps (e.g., 17 cm) are better at capturing smaller organisms such as deer mice and shrews. Conversely, Vanek et al. (2021) captured almost twice as many deer mice using longer Sherman traps in (22.9 cm) in northeastern Illinois, suggesting that the effect of trap size may be landscape dependent. Longer Sherman traps (e.g., 23 cm) tend to capture greater species richness than shorter traps (Slade et al. 1993; Anthony et al. 2005; Vanek et al. 2021). Trap tripping weights can also bias results. For example, Grant (1970) captured a greater proportion of deer mice in Longworth live traps (similar trapping mechanism to Sherman traps) set with low tripping weights (i.e., easier to trigger with less mass) and a greater proportion of eastern meadow voles when tripping weights were set high. We may have unintentionally increased our catch of deer mice by intentionally reducing our Sherman trap tripping weights in the hopes of increasing overall trap success. Because of this intratrap-type variation, active trapping surveys are most effective when researchers deploy a variety of traps with a variety of characteristics (e.g., different lengths, different trigger weights, different traps; Francl et al. 2002; Stephens and Anderson 2014). However, the addition of more types of traps (e.g., Longworth traps or pitfall traps) to the survey would increase the costs and time requirements of the study, making it even less efficient than AHDriFT systems, which seem to be more robust to trap bias.

The AHDriFT systems in our study captured smaller proportions of deer mice (daytime proportion = 0.03, nighttime proportion = 0.53) and greater proportions of voles (daytime = 0.05, nighttime = 0.03) and T. striatus (daytime = 0.14, nighttime = 0.05) during daytime hours (Table 1). This resulted in amplified differences in diversity characteristics when comparing Sherman trap data with unfiltered AHDriFT system data (Table 2). If we included daytime Sherman trapping in our study, we may have recorded higher diversity in our Sherman catch as well. Daytime trapping would also decrease the proportion of hours that traps are closed because of disturbance from nontarget wildlife (e.g., raccoons Procyon lotor and opossums Didelphis virginiana), which occasionally occurred during nighttime trapping in our study, as in others (Layne 1987; Atkinson 1997). However, the increased exposure to heat poses an elevated risk of mortality for individuals captured with metal Sherman traps (Shonfield et al. 2013). This mortality risk, scheduling conflicts, and a lack of personnel were the factors that prompted our decision to not include daytime hours in our Sherman trap surveys. Contrarily, heat mortality risk is not an issue with AHDriFT systems, which averaged 26.5 ± 0.4°C in temperature during daytime hours (17.9 ± 0.2°C during nighttime hours, on the basis of internal camera trap temperature sensor data pooled between both years and all sites), as individuals are able to freely move through the trap. Additionally, nontarget wildlife rarely disturbed our AHDriFT systems. We did observe predators entering the AHDriFT systems (raccoons, opossums, American mink Neogale vison, weasels) and observed a barred owl Strix varia attack an unknown animal moving along an AHDriFT system drift fence.

We obtained slightly different results compared with previous studies that investigated the effectiveness of AHDriFT systems (Martin et al. 2017; Amber et al. 2021). Our total AHDriFT system catch per unit effort was approximately two times greater (0.97 UCEs per trap night) than those reported in these previous studies (0.49; Amber et al. 2021). However, broken down, only our Peromyscus spp. catch per unit effort (0.51 UCEs per trap night) was greater than reported in other studies. We captured fewer voles (0.03 UCEs per trap night), shrews (0.09 UCEs per trap night), and snakes (0.07 UCEs per trap night) per trap night (see Amber et al. 2021 for the other studies' data). The differences in success between these studies and our own may be due to differences in sampling communities, trap nights sampled, and the characteristics of the AHDriFT systems deployed. Both Martin et al. (2017) and Amber et al. (2021) conducted their studies in areas known for their high species diversity. Conversely, our site was in a heavily disturbed wetland that inherently had a reduced mammal diversity compared with these other two sites (Ruch et al. 2016). More trap nights are expected to lead to higher probabilities of detecting rarer species (Tobler et al. 2008). Because they sampled in more locations and over more calendar days, both previous studies conducted at least eight times more trap nights than we did. This potentially led to the detection of a greater number of species. We did allow our AHDriFT systems to sample continuously throughout 2 y of the study, including in winter and spring. However, we did not include those data here for the sake of comparison with our Sherman trap surveys. If we were to include all 2 y worth of data, we would have added the American mink to our list of detected mammals, as we detected the species in the fall of 2020. Amber et al. (2021) was the only study that used a Y-shaped fence array, which may have provided more surface area for mammals to encounter while traversing the terrain. However, such an array also increases the cost of each AHDriFT system, as an additional Hunt trap (bucket and camera trap) is required.

Both Martin et al. (2017) and Amber et al. (2021) used professional-grade cameras in their AHDriFT systems, although the latter used less expensive cameras than the former. Our AHDriFT system used low-cost, consumer-grade cameras, but still obtained valuable community diversity data and detected the presence of rarer vertebrate species (e.g., meadow jumping mouse, long-tailed weasel; Figure 4). Our cameras did, however, accrue numerous false-trigger images. Upgrading to professional-grade cameras may decrease the frequency of false-trigger images (Glen et al. 2013; Amber et al. 2021). However, our cameras returned a 3:1 false-trigger : species image ratio, which is similar to what Amber et al. (2021) reported (2:1) but much lower than the 50:1 ratio reported by Martin et al. (2017). We suggest that labs upgrade their cameras if they have the funds available. However, our study suggests that valuable data and community estimates can still be collected using lower-quality cameras.

In our study, we used consumer-grade trail cameras with lenses that cannot be adjusted to account for different focal distances. In response to this, we strapped 2.75–3.75 diopter reading glass lenses to the front of each camera lens to reduce their overall focal distance. This method worked well for our purposes and did not reduce the ability of cameras to produce quality images (Figure S1, Supplemental Material). Concurrent with our results, Meek and Cook (2022) found that 0.5–6 diopter reading glass lenses constrain camera focal distance points to 200–20 cm, which they used to effectively identify small mammal species using typical trail camera setups. The results from both our study and that of Meek and Cook (2022) indicate that simple reading glasses will allow researchers to survey smaller animals or close-up locations without the need to purchase expensive trail cameras with modified focal distances. Additionally, this method provides some flexibility in camera use, as researchers can use the same camera to survey different focal distances by simply switching lens diopters.

Much more time (7 min per observation) was spent conducting our AHDriFT system surveys than other studies testing the effectiveness of the method (0.29–1.19 min per observation; Martin et al. 2017; Amber et al. 2021). We manually sorted photos and generated species record tables rather than using data management (e.g., automatic image sorting; camtrapR; Niedballa et al. 2016) and automated image processing software (Norouzzadeh et al. 2018), as other studies did (Martin et al. 2017; Amber et al. 2021). The reduction in time spent managing and analyzing images that these programs would have provided (Yu et al. 2013; Beery et al. 2019) would have greatly reduced the cost per observation for AHDriFT systems, making them even more cost-effective than Sherman traps. Image analysis software that automatically identifies species has also proven to be just as accurate as manual vetting (Norouzzadeh et al. 2018). Thus, effectiveness would not be sacrificed for increased efficiency. As in Martin et al. (2017), our study did show a notable decline in time in the field compared with an active trapping method conducted in the same area. This reduction in field time may make up for the time needed to process images, assuming data software is not available.

The AHDriFT system is a useful yet developing tool in the new age of wildlife research. Our study demonstrates the increased ability of the system to document species diversity despite a low effort required, when compared with traditional Sherman trapping. Given the strengths demonstrated, the AHDriFT system would work well in preliminary surveys, long-term studies, or community-level surveys such as in this study. The AHDriFT system does have limitations relative to other trapping methods. The inability to differentiate between individuals removes its use from mark–recapture analyses and difficulty in assessing individual physical traits removes it from demographic studies. Additionally, live-trapping methods such as Sherman trapping are currently the better option for studies requiring the handling of specimens. However, creative ingenuity, such as was used in McCleery et al. (2014) to measure individuals as they enter the trap, or the addition of extra features (e.g., bait, passive integrated transponder tag readers) may increase the uses of the AHDriFT system. The AHDriFT systems also have not been compared with other methods or combinations of methods (e.g., Sherman traps + pitfall traps), they have not been tested in multiple weather conditions (we noticed increased false triggers during rain and snow events in our study), nor have they had their specifications (e.g., fence material, presence of shading material, floor substrate) evaluated. Even with room to grow, this new method for passively surveying wildlife demonstrates an ability to capture the diversity of a study site more so than a traditional trapping method. The AHDriFT system collected valuable data for our research purposes and we believe it could be modified in many ways to diversify its potential areas for use.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Data S1. Data archive (comma-separated value spreadsheet) of summarized data from Sherman trap surveys conducted at four sites in Eagle Marsh Nature Preserve, Fort Wayne, Indiana. We collected data once a week from 23 February to 25 July 2020 and 28 May to 27 August 2021. Surveys consisted of a single 10-trap transect deployed parallel with the AHDriFT systems at the same sites. We baited traps with a mixture of peanut butter, rolled oats, and wild bird seed mix; we placed a polyester filling inside each trap and placed a single roofing shingle on top of each trap for weather resistance. We set traps at approximately 1600 hours and checked at approximately 0900 hours the next day. We provide only positive capture data in this spreadsheet; misfires or no-capture data are not reported.

Available: https://doi.org/10.3996/JFWM-22-042.S1 (1 KB XLSX)

Data S2. Data archive (comma-separated value spreadsheet) of raw camera images collected by the four AHDriFT systems deployed in this study. We deployed two to three systems from 23 February to 25 July 2020 and 28 May to 27 August 2021. We collected data from Eagle Marsh Nature Preserve, Fort Wayne, Indiana. We set camera traps to take three-image bursts whenever triggered. We coded vertebrates within images to the species level whenever possible.

Available: https://doi.org/10.3996/JFWM-22-042.S2 (100 KB XLSX)

Figure S1. Images of a deer mouse Peromyscus spp. specimen (affixed to a meter stick) taken from an AHDriFT system containing cameras (model BTC-4P Command Ops Pro; Browning, Morgan, UT) without (A) and with (B) a 2.0 diopter reading glass lens. This glass lens was affixed to the camera lens using electrical tape. The camera lens was approximately 30 cm from the surface of a lab table. Photos were taken in the same lab with the same lighting.

Available: https://doi.org/10.3996/JFWM-22-042.S3 (31 KB PDF)

This work was funded by Purdue University Fort Wayne–University Research and Innovation as well as the Environmental Resources Center. We thank the numerous volunteers that helped collect data and process images and The Little River Wetland Project for access to sites. We also thank anonymous reviewers and editors for comments that helped improve the manuscript.

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.