Printer Friendly

Using infrared cameras and skunk lure to monitor swift fox (Vulpes velox).

Mark-recapture techniques have been used to determine the status of swift foxes (Vulpes velox) in Colorado (Schauster et al., 2002; Finley et al., 2005; Martin et al., 2007) and elsewhere (Moehrenschlager et al., 2003; Olsen et al., 2003; Ausband and Foresman, 2007). Although mark-recapture will yield population estimates, this technique can be labor intensive, increases the risk of injury to animals, and is often precluded by budgetary constraints on a large scale. Thus, state agencies have explored more efficient and less expensive means to monitor the status of the species. Various noninvasive survey methods have been developed to monitor swift foxes across large areas to reduce time and labor costs.

For swift foxes, noninvasive techniques including scent stations, scat collection, track plates, spotlighting, and calling, among others, have been used to estimate relative and absolute abundance with mixed results (Harrison et al., 2002; Sargeant, et al., 2003; Knox and Grenier, 2010; Wilson, 2011). Harrison et al. (2002) compared multiple noninvasive survey methods and found scent-station surveys to be the most reliable method for detecting swift foxes. Another study (Schauster et al., 2002) evaluated six methods including catch-per-unit-effort, mark-recapture, scent-post surveys, spotlight counts, scat deposition rates, and an activity index and concluded that all methods, except spotlight counts, were reliable and consistent for detecting swift fox presence. Generally, animal sign surveys and scat collection are the most common methods used by states within swift fox range (Harrison et al., 2004; Sargeant et al., 2005; Bly, 2011).

Knox and Grenier (2010) evaluated the use of hair snares, live trapping, and infrared cameras at scent stations as potential survey methods and concluded that infrared cameras were the most efficient method for determining swift fox presence. They noted that infrared cameras detected swift foxes when other methods failed to do so. With the advancements in photo quality and movement detectability, infrared cameras continue to improve our ability to identify the species present and the distance from which they can be detected.

Recent advances in estimation methods have emphasized that reliable inferences can be made from occupancy studies including index surveys (i.e., evidence left by an animal [e.g., hair, scat, tracks] that the species was present), provided that both detection and occupancy probabilities are simultaneously estimated (Stanley and Royle, 2005; MacKenzie, 2006; Bailey et al., 2007). Occupancy modeling using detection-nondetection (presence-absence) data has become a preferred alternative to mark-recapture to efficiently monitor species on a landscape scale. Occupancy modeling is predicated on the probability of detecting animals. Because the absence of a species is difficult to determine, imperfect detectability (i.e., detectability is <1) must be accounted for in the modeling process (MacKenzie and Bailey, 2004; MacKenzie, 2006). The bias associated with a "false absence" (i.e., the species was present but undetected) can be minimized through changes in study design, such that sufficient surveying effort is expended so that the probability of not detecting the species is negligible (MacKenzie, 2005). Therefore, the desire to improve the probability of detecting an animal has direct implications on survey design and ultimately on the level of effort needed to obtain robust and unbiased estimates. A variety of survey techniques have been explored to assess the tradeoff between improving the probability of detecting an animal and obtaining some practical level of efficiency.

Occupancy modeling using scent-station survey data is an increasingly viable option for surveying canids, such as swift foxes, because the detection of animals within a given area is improved by the aid of an attractant. However, the effectiveness of scent-station surveys is ultimately dependent on the ability of the lure to attract the target animals. Previous studies have shown that a wide variety of both natural and artificial products have been used to attract and survey swift fox and other canids (Kitchen et al., 1999; Harrison et al., 2002; Olsen et al., 2003; Sargeant et al., 2003; Finley et al., 2005; Martin et al., 2007). Because the density of swift foxes varies across the species range and in most cases is unknown, researchers have predominantly used attractants consisting of fish oils or commercial products for their aromatic scent, which may persist in varying environmental conditions (Schauster et al., 2002; Olsen et al., 2003; Sargeant et al., 2003; Harrison et al., 2004; Bly, 2011; Wilson, 2011). However, Cudworth et al. (2011) reported a significant increase in detection probabilities, using a combination of fish oil and skunk essence as the attractant, compared to their previous surveys when only fish meat was used. Although this provides some insight that skunk scent may help in attracting swift foxes, information is lacking on the effectiveness of skunk scent as the exclusive attractant.

Swift fox surveys using mark-recapture sampling were conducted in 1995-1997 (Finley, 1999; Finley et al., 2005), and again in 2004-2005 (Martin et al., 2005; 2007) to estimate swift fox occupancy and their probability of detection in eastern Colorado. The attractants used in those surveys were a combination of poultry meat (turkey poult or chicken parts), commercial bait, canned mackerel, or cat food. In 2011, we developed a new survey design to evaluate the use of infrared cameras at skunk scent stations to estimate detection and occupancy rates of swift foxes in eastern Colorado. Because capture-recapture sampling is widely accepted as an accurate technique for estimating detection probabilities (Seber, 1982; Pollock et al., 1990; Williams et al., 2002; Stanley and Royle, 2005), we wanted to use the results from the previous survey as a comparative benchmark for determining the effectiveness of using infrared cameras with skunk scent as the sole attractant for estimating occupancy and detection probabilities. Our objectives were to 1) estimate detection probabilities and swift fox occupancy rates of 4.8 x 6.4 km (3 x 4-mi) grids; 2) compare parameter estimates to those from the previous mark-recapture survey (Martin et al., 2007) to determine if the use of cameras and skunk scent was a viable survey alternative; and 3) evaluate the efficiency of the design relative to time, labor, survey costs, and results.

MATERIALS AND METHODS--Study Area--The survey area included portions of 23 counties in eastern Colorado, primarily east of Interstate 25, encompassing nearly 80,000 [km.sup.2]. The eastern plains are dominated by shortgrass and midgrass prairies, Conservation Reserve Program plantings, and agricultural development. The terrain varies widely from flat to rolling upland plains in the eastern and central portions to high plains and canyons in the southeastern part of the state. Both irrigated and dryland corn and wheat have been the primary agricultural crops in production on the eastern plains (United States Department of Agriculture, 2009). Cattle production is common throughout the region and grazing intensity varies greatly.

Shortgrass prairie is dominated by blue grama (Bouteloua gracilis), buffalo grass (Buchloe dactyloides), scarlet globemallow (Sphaeralcea coccinea), prickly-pear cactus (Opuntia polyacantha), rabbitbrush (Chrysothamnus nauseosa), broom snakeweed (Gutierrezia sarothrae), and spreading buckwheat (Eriogonum effusum). In eastern Colorado, Conservation Reserve Program plantings contain a variety of native and nonnative vegetation. Although the vegetative composition varies by location, generally Conservation Reserve Program plantings consist of western wheatgrass (Pascopyrum smithii), switchgrass (Panicum virgatum), blue grama, sand bluestem (Andropogon hallii), yellow Indiangrass (Sorghastrum nutans), prairie sandreed (Calamovilfa longifolia), and green needlegrass (Nassella viridula). Pinyon pine (Pinus edulis) and one-seed juniper (Juniper monosperma) are common within and along canyon breaks, bluffs, and mesas in the southeastern part of the state.

The climate on the eastern plains is generally semiarid and uniform across the region. It is characterized by low humidity, infrequent rains and snow, moderate-high wind movement, and a large daily and seasonal range in temperatures (Pielke et al., 2003). Winter precipitation is light and infrequent and most of the annual precipitation (70-80%) falls during the growing season (April-September). Annual precipitation ranges from <30 cm in the Arkansas Valley to nearly 46 cm in extreme northeastern and southeastern corners of the state (Pielke et al., 2003). Mean temperature during the fall survey season (September-November) for the state was 7.0[degrees]C and the mean precipitation was 9.7 cm (1991-2011 data; National Climatic Data Center, cag3/co.html).

Data Collection--Finley et al. (2005) demonstrated that the composition of shortgrass prairie in a grid was a reliable predictor for the probability of occupancy ([psi]) and probability of detection (p) for swift foxes in eastern Colorado. Thus, in 2004, Martin et al. (2007) identified 2,566 available grids in eastern Colorado, and 51 grids were randomly selected to survey, which approximately represented the percentages of shortgrass prairie proportional to its availability in the study area and was based on power calculations provided (Finley et al., 2005). To compare changes in occupancy and detection over time, we maintained the same grid size (4.8 x 6.4 km) as previous surveys (Finley et al., 2005; Martin et al., 2007) and used the same 51 grids surveyed in 2004-2005 as our initial sampling frame. Using a spatially-balanced sampling process employing a Reversed Randomized Quadrant-Recursive Raster algorithm (Theobald et al., 2007), we randomly selected an additional 51 grids to be used as alternative survey sites in case landowners denied access to the primary grids. The Finley study (Finley et al., 2005) reported that swift fox detections varied throughout the year and were highest in the fall. Therefore, we conducted surveys from August-October 2011 to maintain consistency with previous surveys and maximize detection probabilities (Martin et al., 2007; Knox and Grenier, 2010).

Camera Station Design--For each 30.7-[km.sup.2] grid, we used an array of eight Reconyx[TM] infrared cameras (Model PC800; Reconyx Inc., Holmen, Wisconsin). Three cameras were spaced at 3.2-km intervals across the top and bottom of each grid and two cameras were centered north to south, spaced 3.2 km apart and 1.6 km from the east and west edge of the survey grid, representing the most even dispersion. When necessary, we moved cameras within a grid to accommodate landowners who denied access. In most cases, we moved cameras [less than or equal to] 0.8 km. We placed cameras along fences, power lines, and trails which are common travel routes for canids, including swift fox.

We attached cameras to either light duty "U" posts measuring 0.91 m using a single screw or to 1.27-cm rebar measuring 76 cm using a rubber-lined pipe clamp. The "U" posts were equipped with predrilled holes spaced evenly along the shaft, which provided for quick attachment and consistent height reference. We used rebar primarily in areas adjacent to public roads to conceal cameras from public view and reduce potential theft. We placed a wooden stake (61 cm) approximately 3 m in front of each camera to serve as a base for the lure and a focal point for the camera. We placed both the camera and survey stake at a height of 38-40 cm. We created a skunk-based lure by heating 385 ml of petroleum jelly to liquid form, adding 8 ml of skunk essence (Schmitt Enterprises, Inc., New Ulm, Minnesota), and allowed the lure to solidify (Cudworth et al., 2011). We applied 5-10 ml of lure to the top of each stake as the attractant.

We programmed the cameras to take three consecutive photos at 1-s intervals each time the camera was triggered. We set cameras to take pictures from 1 h before sunset to 1 h after sunrise to take advantage of peak swift fox activity (Kitchen et al., 1999; Moehrenschlager et al., 2003; Finley et al., 2005) and to reduce extraneous, nontarget photos (e.g., livestock and vegetation movement). Because swift fox are primarily active at night, we do not believe that the limited daytime activity that may have been missed would bias estimates of p and W We programmed photos to be stamped with the date, time, temperature, camera number, and grid number. Cameras were active for five consecutive nights on all grids (Cudworth et al., 2011). We recorded the number of swift foxes and nontarget species observed from each camera and survey grid. We used a Global Positioning System set to North American Datum 1983 to collect Universal Transverse Mercator coordinates for each camera location.

Statistical Analyses--We combined data from the eight cameras to develop an encounter history for each grid and estimated the probability of occupancy (W) and detection (p) using Program PRESENCE (Hines, 2011). Because there is evidence that shortgrass prairie may influence occupancy ([psi]) and detection probabilities (p) (Finley et al., 2005; Martin et al., 2007), we considered a set of a priori models that incorporated the percentage of shortgrass prairie within each survey grid to model [psi] and p. Because the distance and direction swift foxes came from was unknown, after completion of the survey we used ArcMap, version 10.0 (Environmental Systems Research Institute, Redlands, California) to buffer each grid by 1.6 km, which was one half the average intercamera distance to account for swift foxes that may have originated outside the sampled grid (Martin et al., 2007). We recalculated the percent shortgrass prairie for each 8.0 x 9.6-km (buffered) grid in ArcMap 10.0 using Southwest ReGap Vegetation Classification data. We standardized the covariate (shortgrass prairie [SGP]) before inclusion in the models (Franklin, 2001). We modeled both p and [psi] for the complete survey period and for just the first 3 nights for comparison with previous surveys. We calculated two estimates for [psi]: 1) when SGP was held constant across all grids, and 2) when SGP was allowed to vary by grid using the actual percentages within each grid. We report model outputs which include two estimates of [psi] and up to five detection probabilities (p).

We evaluated occupancy models using Akaike's Information Criterion adjusted for small sample sizes ([AIC.sub.c]) to perform model selection in an information-theoretic framework (Burnham and Anderson, 2002). We considered models with [DELTA][AIC.sub.c] values [less than or equal to] 2 to be equally parsimonious (Burnham and Anderson, 2004) and used Akaike weights ([w.sub.i]) to assess relative support for different models. For the top models selected, we performed a MacKenzie-Bailey goodness-of-fit test (MacKenzie and Bailey, 2004) to determine whether there was evidence that [psi] and p were overestimated based on the percentage of SGP in each grid, indicating a lack-of-fit of the models. To estimate the overdispersion parameter ([??]), we grouped the data by grid, based on the covariate SGP, and used a parametric bootstrap in Program PRESENCE to generate modeled estimates of the expected number of grids for each possible detection history. We considered [??] values [less than or equal to] 1.1 as evidence of adequate model fit (MacKenzie and Bailey, 2004; Hines, 2011). We estimated occupancy and detection probabilities from the minimum AICc model and used model averaging when more than one model was supported (Burnham and Anderson, 2002; MacKenzie et al., 2006).

We estimated the proportion and number of grids occupied by swift foxes in eastern Colorado using the logistic equation [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (MacKenzie et al., 2006; Hines, 2011). We used the estimates from the minimum [AIC.sub.c] model with the grid-specific covariate SGP, where the estimates from Program PRESENCE are the intercept ([[??].sub.0]) and slope parameter ([[??].sub.1]) and the covariate value of SGP for grid i (i = 1... 2,566) is [x.sub.i]. The number of occupied grids was estimated by [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] with the variance estimated by the sum of the elements of the 2,566 x 2,566 estimated variance-covariance matrix of the [[??].sub.i] (Martin et al., 2007).

RESULTS--Survey Effort--We initially surveyed four grids from 1-9 August 2011 as a pilot to estimate detection probabilities (p) to determine if the sampling frame was adequate for comparison with previous mark-recapture survey results. We detected swift foxes on three of four grids; therefore, no changes were made to the sampling frame. We surveyed a small number of alternative grids (n = 5) because adequate landowner permission could not be obtained on four of the original grids, and permission was obtained for one additional alternative grid. Therefore, we surveyed the remaining 48 grids from 29 August-28 October 2011 for a total sampling frame of 52 grids.

We completed the survey with 95 camera nights (CN) in which no data were collected. The initial pilot survey in early August accounted for 26% of the inoperable camera nights, all from livestock interference. The remainder resulted from battery failure (20 CN), camera theft (15 CN), livestock interference (15 CN), camera destruction (10 CN), and human error (10 CN). We collected 331 swift fox detections during the remaining 1,985 CN. We detected [greater than or equal to] 1 swift fox on 45 of the 52 survey grids and the number varied from 1-18 detections per grid (Fig. 1). Of those 45 grids, we detected swift foxes on 80% of the grids (36 of 45) in the first night. After the second night, 91% (41 of 45) of the grids had obtained a swift fox detection and 98% (44 of 45) of the grids had a confirmed detection by the end of the third night. We detected a total of nine species of birds and 21 mammals. In addition to swift fox, the most common predators detected were badger (Taxidea taxus), coyote (Canis latrans), raccoon (Procyon lotor), and striped skunk (Mephitis mephitis) (Table 1).

Detection and Occupancy Estimation--Detection probabilities were relatively consistent across most nights with the first night having the highest probability at p = 0.799 (SE = 0.060, 95% confidence interval [CI] 0.681-0.916) (Fig. 2). The average probability of detecting a swift fox across all nights was p = 0.692 (SE = 0.031, 95% CI 0.631-0.753). To compare with previous surveys, the detection probability across the first three nights was p = 0.700 (SE = 0.043, 95% CI 0.617-0.783). There was no evidence of lack-of-fit in the top models ([??] = 1.081; P = 0.324) based on 1,000 iterations, confirming that the estimates of p and w, as a function of SGP within the grids, were unbiased. This also validated that the distribution of percentages of SGP across the sampling frame was sufficient for determining detection rates and occupancy of swift foxes in eastern Colorado. Compared to the top model with constant detection probabilities, allowing detection rates to vary based on the percentage of SGP did not improve model fit of detection probabilities, although there was evidence ([w.sub.i] = 0.269) to suggest that it does have some influence on detection (Table 2). However, the percentage of SGP did provide an important predictor of occupancy (Fig. 3) from the logit predictive equation where [[??].sub.0] = -1.63 (SE = 1.12) and [[??].sub.1] = 9.76 (SE = 3.96) based on model averaging.

The overall occupancy rate using the average amount of SGP found on the 52 grids sampled was [??] = 0.868 (SE = 0.048). When we estimated occupancy across the 52 grids using the observed amount of SGP on each grid, [??] = 0.867 (SE = 0.051). When we estimated the occupancy rate from the first 3 nights of the survey, [??] = 0.872 (SE = 0.053) when we used the average amount of SGP on the grids and [??] = 0.861 (SE = 0.057) when we used the observed amounts of SGP. Finally, based on the set of 2,566 grids from which the 52 surveyed grids were selected, we estimated the number of grids in eastern Colorado occupied by swift foxes at 1,963.1 grids (SE = 87.5, 95% CI 1,792-2,134). Thus, we estimated the overall proportion of occupied grids in eastern Colorado to be [??] = 0.765 (SE = 0.034, 95% CI 0.698-0.832).

DISCUSSION--After the first night, detection probabilities declined an average of 14% for the remaining survey nights. In Wyoming, Cudworth et al. (2011) found nearly a 10% decline after the first detection and it was hypothesized this was due to a lack of curiosity in the lure after the initial swift fox investigation. In our study, the detection of swift foxes tended to show a pattern of visitation over time, with numerous detections occurring on a 2- or 3-d cycle. The influence of shortgrass prairie on detection was evident in our study because cameras placed within or on the edge of shortgrass prairie habitat accounted for all swift fox detections, which was reflected in our analysis (second highest model; Table 2). Our results showed a stronger relationship between the percentage of shortgrass prairie and probability of detection ([w.sub.i] = 0.269) compared to previous mark-recapture surveys ([w.sub.i] = 0.211, Finley et al., 2005; [w.sub.i] = 0.138, Martin et al., 2007), demonstrating that our study design was robust in capturing the habitat association with the species. In Wyoming, detection probabilities were also influenced by the percentage of grassland in the grids, which included shortgrass prairie (Cudworth et al., 2011).


Comparing detection probabilities with the previous mark-recapture surveys, detection averaged 0.700 over the first 3 d compared to 0.680 in 2004-2005 (Martin et al., 2005) and 0.720 in 1995-1997 (Finley et al., 2005). Regardless of the possible demographic changes that may have occurred over time, detection was high and consistent with previous surveys demonstrating that using skunk lure as a sole attractant was effective in attracting swift foxes without the added expense and effort to handle additional attractants like those used in other surveys.

Previous surveys used up to five attractants, which invariably increased the expense, logistics, and handling time over the use of a single attractant. In addition, using an attractant that can be removed from the station, (e.g., poultry, fish, or meat) requires additional labor because each station must be revisited regularly during the survey period to maintain consistent detectability across time. Furthermore, other important information, such as number of animals visiting a station, may be biased because the probability of detecting additional animals is reduced once the attractant is removed. Based on our observations throughout the study, there was no noticeable change in the pungent scent of the skunk lure over time, and the lure was never washed away following rain events that periodically occurred; thus, reapplication was never conducted. This eliminated the need to revisit each station, thereby reducing the man-hours needed to survey a single grid from the 20-25 h reported for previous surveys (Finley et al., 2005; Martin et al., 2005) to 5-7 h in our study, depending on access. The average cost of our lure was US$1.50/grid, which is a substantial cost savings compared to US$10/grid estimated for previous surveys using multiple attractants (Finley et al., 2005; Martin et al., 2005).

The skunk-based lure used in this and other studies (Cudworth et al., 2011) was very effective at attracting swift foxes because of its aromatic potency, which persisted over time. We believe the use of a skunk-based lure was the primary reason our results were robust in relation to previous mark-recapture surveys. Martin et al. (2007) reported 2.7 detections per grid using 20 cage traps and multiple attractants compared to 4.1 detections per grid for the first three nights in our study using skunk lure and 12 fewer survey stations per grid. Across all five nights, we averaged 6.4 detections per grid. The Wyoming survey (Cudworth et al., 2011) reported 4.2 detections per grid across five survey nights using fish oil and skunk essence. The positive results from our study and those reported for Wyoming (Cudworth et al., 2011) provide strong support that the use of cameras and skunk-based lure is a robust technique for detecting swift foxes and estimating occupancy rates.

Martin et al. (2007) estimated overall [??] = 0.711 (SE = 0.069) from 3 d of mark-recapture for the 2,566 grids, whereas our study found [??] = 0.765 (SE = 0.034) from the first 3 d using cameras and skunk lure. Our results represent an increase of 138.5 additional grids (4,255 [km.sup.2]) estimated to be occupied by swift foxes. However, the difference is within the sampling variation of the estimate; thus, a significant change in swift fox occupancy in eastern Colorado was not detected. Our results also show an improvement in the precision of our estimate over previous results providing strong support for the use of cameras and skunk-based lure as a robust technique for estimating swift fox occupancy rates. In addition, two studies (Finley, 1999; Martin et al., 2005) reported that their detection probabilities may have been biased because swift fox sign was recorded at stations in which no swift foxes were captured. Although changes in population size and distribution may have been a factor, we believe the increase in occupancy estimates was largely the result of changing to a more effective survey methodology, which resulted in more accurate estimates of occupancy than previously reported.


Cameras provided a less intrusive mode of detection because swift foxes could be detected more easily than they would have been using cage traps or sign surveys. Generally there is some degree of bias associated with cage traps, animal sign surveys, and surveys using cameras when animals are present but never captured, which results in underestimation of both detection and occupancy (Finley, 1999; Martin et al., 2005; Mackenzie, 2006). However, using cameras may reduce some of the negative trap response associated with other techniques because animals are not required to be captured or leave identifiable evidence to verify their presence. The cameras we used provided excellent quality photos with high resolution, which made for quick and easy identification of all species photographed out to distances well beyond the reference survey stake. We estimate that swift foxes were detected up to 15 m away from the scent station, providing a substantially larger detection zone that greatly reduced any potential response bias.

Other forms of bias may be introduced with cameras if animals are missed or misidentified due to poor photo quality or photos taken from long distances. Because animal movements can be detected at extended distances, cameras should be tested prior to surveying to establish a distance threshold that maximizes photo quality for accurate species identification. Distance markers can then be used to standardize photo inclusion in the dataset. Likewise, obstruction of the camera lens due to cold, rain, wind, vegetation, or battery failure, among others, is inherent and may also introduce bias if animals are present but never identified. Because occupancy and detection were based on the grid and not on individual cameras, we minimized this source of bias by using multiple detection sites within each grid to compensate for the failure or restrictions that periodically occurred to one or more cameras.

Martin et al. (2007) reported the capture of 73 nontarget animals that made the traps inoperable for detecting additional swift foxes for the remainder of those survey nights In contrast, cameras can provide continuous detection throughout the night, irrespective of the presence of nontarget species. The use of cameras was also less labor intensive than the use of cage traps. Within the first 3 d of our study, we documented 78 more swift fox detections on 25% more grids. Despite surveying two additional days on each grid, we also reduced the total survey time by 100 days using a two-person crew, which was half the manpower used in previous surveys. Overall, the improvement in survey efficiency with our design resulted in an 80% reduction in labor costs to complete our survey as compared to the trapping efforts used in the previous survey for estimating occupancy and detection probabilities.

The ability to accurately assess changes in swift fox occupancy requires a determination of what areas are actually occupied across the landscape. The use of cameras and skunk lure was successful in refining detection and swift fox occupancy estimates in Colorado compared to previous surveys that used mark-recapture techniques. Our estimates showed improvements in precision and accuracy for detecting swift foxes and estimating occupancy across eastern Colorado. The use of cameras and skunk lure proved to be an efficient and accurate tool for monitoring swift foxes on a landscape scale by reducing labor costs, survey time, and potential biases that may occur using other survey techniques.

Scent-station surveys are ultimately dependent on the success of the lure to attract animals. We demonstrated that a skunk-based lure is very effective in attracting swift foxes and is recommended because of its aromatic longevity, which minimizes the need for reapplication. Although our survey was specifically designed for swift foxes, we believe this lure would also be effective for surveys on other mesocarnivores (Table 1). Because the dynamics of swift fox populations vary across the species range, further studies are needed to determine if the use of a skunk-based lure improves swift fox detections in other regions of the species range where abundance and preferred swift fox habitat likely differs.

The capability of the cameras to take multiple, high-resolution pictures in succession each time movement was detected provided the means to positively identify all mammalian species, along with important information such as date and time of visitation, which can be subjective or unknown with other survey techniques. Likewise, our use of camera stations was an efficient mode for detecting swift foxes while providing reliable estimates of occupancy and improving our knowledge of the species distribution. Although we surveyed for five nights, managers should evaluate the trade-off between survey effort and their desired level of precision. Based on our detection probability, surveying for three consecutive nights would have been adequate for determining the presence of swift foxes. This change would have further improved survey efficiency and provided opportunities to survey additional areas within the same timeframe.

We thank J. Wills for his efforts collecting and compiling data. We also thank the many District Wildlife Managers and Biologists from across the eastern plains for their help obtaining landowner permission and in the field. We thank C. Woodward for assistance with Geographic Information System processing and P. Lukacs and J. Runge for their statistical support. We extend a special thanks to all the landowners for their continued cooperation in allowing access to their lands. This project was funded by the Colorado Division of Parks and Wildlife.


AUSBAND, D. E., AND K. R. FORESMAN. 2007. Dispersal, survival, and reproduction of wild-born, yearling swift foxes in a reintroduced population. Canadian Journal of Zoology 85:185-189.

BAILEY, L. L., J. E. HINES, J. D. NICHOLS, AND D. I. MACKENZIE. 2007. Sampling design trade-offs in occupancy studies with imperfect detection: examples and software. Ecological Applications 17:281-290.

BLY, K. (editor). 2011. Swift fox conservation team: report for 2009-2010. World Wildlife Fund, Bozeman, Montana, and Montana Department of Fish, Wildlife and Parks, Helena.

BURNHAM, K. P., AND D. R. ANDERSON. 2002. Model selection and inference: a practical information-theoretic approach. Springer-Verlag, New York.

BURNHAM, K. P., AND D. R. ANDERSON. 2004. Multimodel inference: understanding AIC and BIC in model selection. Sociological Methods and Research 33:261-304.

CUDWORTH, N., L. VAN FLEET, AND M. GRENIER. 2011. Monitoring trends and documenting distribution of swift fox in Wyoming. Wyoming Game and Fish Department, Nongame Program, Lander.

FINLEY, D. J. 1999. Distribution of the swift fox (Vulpes velox) on the eastern plains of Colorado. M.S. thesis, University of Northern Colorado, Greeley.

FINLEY, D. J., G. C. WHITE, AND J. P. FITZGERALD. 2005. Estimation of swift fox population size and occupancy rates in eastern Colorado. Journal of Wildlife Management 69:861-873.

FRANKLIN, A. B. 2001. Exploring ecological relationships in survival and estimating rates of population change using program MARK. Pages 290-296 in Wildlife, land, and people: priorities for the 21st century (R. Field, R. J. Warren, H. Okarma, and P. R. Sievert, editors). Proceedings of the second international Wildlife Management Congress. The Wildlife Society, Bethesda, Maryland.

HARRISON, R. L., D. J. BARR, AND J. W. DRAGOO. 2002. A comparison of population survey techniques for swift foxes (Vulpes velox) in New Mexico. American Midland Naturalist 148:320-337.

HARRISON, R. L., P. S. CLARKE, AND C. M. CLARKE. 2004. Indexing swift fox populations in New Mexico using scats. American Midland Naturalist 151:42-49.

HINES, J. E. 2011. Program PRESENCE. Version 4.1. Patuxent Wildlife Research Center, United States Geological Survey, Laurel, Maryland.

KITCHEN, A. M., E. M. GESE, AND E. R. SCHAUSTER. 1999. Resource partitioning between coyotes and swift foxes: space, time and diet. Canadian Journal of Zoology 77:1645-1656.

KNOX, L., AND M. GRENIER. 2010. Evaluation of swift fox survey techniques, completion report. Pages 205-212 in Threatened, endangered, and nongame bird and mammal investigations (A. C. Orabona, editor). Wyoming Game and Fish Department, Nongame Program, Lander.

MACKENZIE, D. I. 2005. What are the issues with presenc-eabsence data for wildlife managers? Journal of Wildlife Management 69:849-860.

MACKENZIE, D. I. 2006. Modeling the probability of resource use: the effect of, and dealing with, detecting a species imperfectly. Journal of Wildlife Management 70:367-374.

MACKENZIE, D. I., AND L. L. BAILEY. 2004. Assessing the fit of site-occupancy models. Journal of Agricultural, Biological, and Environmental Statistics 9:300-318.

MACKENZIE, D. I., J. D. NICHOLS, J. A. ROYLE, K. H. POLLOCK, L. L. BAILEY, AND J. E. HINES. 2006. Occupancy estimation and modeling: inferring patterns and dynamics of species occurrence. Elsevier, Inc., Oxford, United Kingdom.

MARTIN, D. J., G. C. WHITE, AND F. M. PUSATERI. 2005. Population monitoring of swift fox in eastern Colorado. Study Report. Colorado Division of Wildlife, Fort Collins.

MARTIN, D. J., G. C. WHITE, AND F. M. PUSATERI. 2007. Occupancy rates by swift foxes (Vulpes velox) in eastern Colorado. Southwestern Naturalist 52:541-551.

MOEHRENSCHLAGER, A., D. W. MACDONALD, AND C. MOEHRENSCHLAGER. 2003. Reducing capture-related injuries and radio-collaring effect on swift foxes. Pages 107-113 in The swift fox: ecology and conservation of swift foxes in a changing world (M. A. Sovada and L. Carbyn, editors). Canadian Plains Research Center, University of Regina, Regina, Saskatchewan, Canada.

OLSEN, T. L., J. S. DIENI, F. G. LINDZEY, AND S. H. ANDERSON. 2003. swift fox detection probability using tracking plate transects in southeast Wyoming. Pages 93-98 in The swift fox: ecology and conservation of swift foxes in a changing world (M. A. Sovada and L. N. Carbyn, editors). Canadian Plains Research Center, University of Regina, Regina, Saskatchewan, Canada.

POLLOCK, K. H., J. D. NICHOLS, C. BROWNIE, AND J. E. HINES. 1990. Statistical inference for capture-recapture experiments. Wildlife Monographs 107:3-97.

PIELKE, R. A., SR., N. DOESKEN, AND O. BLISS. 2003. Climate of Colorado. Climatology report 60. Department of Atmospheric Science, Colorado State University, Fort Collins.

SARGEANT, G. A., M. A. SOVADA, C. C. SLIVINSKI, AND D. H. JOHNSON. 2005. Markov chain Monte Carlo estimation of species distributions: a case study of the swift fox in western Kansas. Journal of Wildlife Management 69:483-497.

SARGEANT, G. A., P. J. WHITE, M. A. SOVADA, AND B. L. CYPHER. 2003. Scent-station survey techniques for swift and kit foxes. Pages 99-105 in The swift fox: ecology and conservation of swift foxes in a changing world (M. A. Sovada and L. N. Carbyn, editors). Canadian Plains Research Center, University of Regina, Regina, Saskatchewan, Canada.

SCHAUSTER, E. R., E. M. GESE, AND A. M. KITCHEN. 2002. An evaluation of survey methods for monitoring swift fox abundance. Wildlife Society Bulletin 30:464-477.

SEBER, G. A. F. 1982. The estimation of animal abundance and related parameters. 2nd ed. Charles Griffin, London, United Kingdom.

STANLEY, T. R., AND J. A. ROYLE. 2005. Estimating site occupancy and abundance using indirect detection indices. Journal of Wildlife Management 69:874-883.

THEOBALD, D. M., D. L. STEVENS, JR., D. WHITE, N. S. URQUHART, A. R. OLSEN, AND J. B. NORMAN. 2007. Using GIS to generate spatially-balanced random survey designs for natural resource applications. Environmental Management 40:134-146.

UNITED STATES DEPARTMENT OF AGRICULTURE. 2009. 2007 census of agriculture: United States summary and state data. Volume 1. Geographic Area Series, Part 51. National Agricultural Statistics Service, United States Department of Agriculture, Washington, D.C.

WILLIAMS, B. K., J. D. NICHOLS, AND M. J. CONROY. 2002. Analysis and management of animal populations. Academic Press, San Diego, California.

WILSON, S. 2011. Nebraska swift fox report, 2009-2010. Page 18 in Swift fox conservation team: report for 2009-2010 (K. Bly, editor). World Wildlife Fund, Bozeman, Montana, and Montana Department of Fish, Wildlife and Parks, Helena.

Submitted 26 December 2013.

Acceptance recommended by Associate Editor, Troy A. Ladine, 24 March 2014.


Colorado Division of Parks and Wildlife, 122 E. Edison Street, Brush, CO 80723 (MRS)

Colorado Division of Parks and Wildlife, 0722 S. Road 1 East, Monte Vista, CO 81144 (JAA)

* Correspondent:
TABLE 1--Frequency of the most common predator
species detected from cameras set for swift
foxes in eastern Colorado, 29 August-28 October 2011.

                     Species         No.      No.           No.
                                   grids   stations   camera-nights

Badger            Taxidea taxus      27        46           51
Coyote            Canis latrans      50       164           207
Raccoon           Procyon lotor      12        28           46
Striped Skunk   Mephitis mephitis    24        52           89
Swift Fox         Vulpes velox       45       149           280

TABLE 2--Model selection results for swift fox occupancy
on 52 grids surveyed in eastern Colorado, August-October
2011. Variable definitions are: p = detection probability,
[psi] = occupancy probability, SGP = percent shortgrass
prairie in the grid, Day = detection
varied by day.

Model                       [AIC.sub.c]     [DELTA]     [w.sub.i]
                                (a)       [AIC.sub.c]      (b)

{[psi](SGP) p(.)}             307.960        0.000        0.684
{[psi](SGP) p(SGP)}           309.826        1.866        0.269
{[psi](SGP) p(Day)}           313.925        5.965        0.035
{[psi](SGP) p(Day x SGP)}     316.230        8.270        0.011
{[psi](.) p(.)}               322.455       14.495        0.000
{[psi](.) p(SGP)}             324.070       16.110        0.000
{[psi](.) p(Day)}             328.027       20.067        0.000
{[psi](.) p(Day x SGP)}       330.065       22.105        0.000

Model                       Likelihood   k (c)   Deviance

{[psi](SGP) p(.)}             1.000        3     301.480
{[psi](SGP) p(SGP)}           0.393        4     301.010
{[psi](SGP) p(Day)}           0.051        7     297.490
{[psi](SGP) p(Day x SGP)}     0.016        8     297.030
{[psi](.) p(.)}               0.001        2     318.220
{[psi](.) p(SGP)}             0.000        3     317.590
{[psi](.) p(Day)}             0.000        6     314.240
{[psi](.) p(Day x SGP)}       0.000        7     313.630

(a) Akaike Information Criterion for small samples.

(b) Akaike weight.

(c) Number of parameters.

FIG. 2--Probability of detecting swift foxes
by camera night on 31-[km.sup.2] survey grids
in eastern Colorado, August-October 2011.
Error bars represent [+ or -] 1 SE.

Camera Night   Probability of Detection

1                        0.799
2                        0.621
3                        0.666
4                        0.710
5                        0.666

Note: Table made from bar graph.
COPYRIGHT 2014 Southwestern Association of Naturalists
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2014 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Stratman, Marty R.; Apker, Jerry A.
Publication:Southwestern Naturalist
Article Type:Report
Date:Dec 1, 2014
Previous Article:Late pleistocene shrews and bats (Mammalia: Soricomorpha and Chiroptera) from Terapa, a neotropical-nearctic transitional locality in Sonora, Mexico.
Next Article:Seasonal shifts in the diet of the big brown bat (Eptesicus fuscus), Fort Collins, Colorado.

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters