Sit-and-wait foraging snakes (such as the Eastern Massasauga), which are inactive for long periods, might not be efficiently sampled using traps or artificial cover objects [
43,
44]; however, there is some indication that passive sampling can provide supplementary data [
45]. Relatively inactive species are best sampled using time- and/or area-constrained searches. However, determining parameters influencing detectability becomes of paramount importance in the design and implementation of survey and monitoring programs. We found three abiotic parameters (solar irradiance, substrate temperature, mean three-day minimum air temperature), two search parameters (effort per searcher and time of day), and one management related parameter (burns) which significantly influenced detection of Eastern Massasaugas.
4.1. Modeled Effects
Detection probabilities were slightly higher for surveys in burned habitat (either full or partial). Burning is typically used as a management tool in Eastern Massasauga habitat to maintain early successional vegetation [
46]. Eastern Massasaugas do not show alterations in movements, home ranges, or habitat use associated with burns, but do select habitat characteristics favoring thermoregulation [
47]. Thus, they are likely more observable thermoregulating during the spring emergence period from greater distances in burned compared to unburned habitats. Timing or pairing visual encounter surveys (VES) with management burns can be an effective tool for increasing the chances of detection. However, caution must be exercised when burning after snakes emerge, which could cause significant mortality [
46,
47].
Both substrate temperature and mean three-day minimum air temperature influenced detection probability of Eastern Massasaugas in our study system. Such thermal relationships are expected for an ectothermic vertebrate. Harvey [
48] found that Eastern Massasaugas in a Canadian population were more detectable at intermediate body temperatures (20–30 °C) and when they were more visible (>75% body exposure). Furthermore, he found that weather conditions which expressed thermoregulatory behavior (e.g., slightly cooler, sunny days) for intermediate body temperatures also resulted in increased detectability.
In contrast to other studies of snakes in Europe and North America, we did not find a significant positive effect of day of year on detection probability. In the Jura Mountains of Europe (France and Switzerland), Kéry [
4] found a unimodal relationship between day of year and detection probability for three species of snakes (Asp Viper
Vipera aspis; Smooth Snake
Coronella austriaca; Grass Snake
Natrix natrix), and the same unimodal relationship was found in North America for both Banded Watersnakes (
Nerodia fasciata) and Black Swamp Snakes (
Seminatrix pygaea; [
49]). However, the Banded Watersnake and Black Swamp Snake were primarily sampled with passive trapping throughout a full season. Still, it is possible that such a trend may have emerged for Eastern Massasaugas if we sampled throughout the active season. Our data do suggest that fluctuations in environmental temperatures as the day of year progresses can result in successful search periods outside of the spring emergence period, and season and/or day of year might not always be a good proxy for environmental temperatures.
Detection probabilities of Eastern Massasaugas decreased as solar irradiance increased, and a similar pattern was also observed for the time of day. Although the two metrics are not significantly intercorrelated, solar irradiance peaks at astronomical noon, as well as throughout the activity season (strongest during the peak of summer), but it becomes uncoupled in overcast situations. During the spring and early summer, Eastern Massasaugas bask in the morning [
29], which explains our increased detection rates early in the day. As temperatures increase, there is less need to bask; thus, they become more crepuscular and cryptic during the hottest periods of the year. Our decline in detection rates illustrates the reduced need for basking during periods when shaded air temperatures are greater than 24 °C. A similar trend was found by Erb et al. [
7] for Eastern Box Turtles (
Terrapene carolina), which had significantly higher detection rates earlier in the day. However, the pattern does not hold for all reptile species. Common Skink (
Oligosoma polychroma) detections in New Zealand were the highest and least variable at ambient temperatures (12–18 °C) during daytime hours and did not decline until the evening [
50].
We found that detection increased as search effort within a visit increased, but there were diminishing returns. Survey duration within a visit can significantly influence detection values in some, but not all, reptiles [
5]. Furthermore, increasing survey duration was found to increase the power to detect raptor migration trends [
51] and increase the number of species detected in anuran call surveys [
52]. While search effort is incorporated into detection estimates, the effects of search area size are less commonly investigated. For Eastern Massasaugas, we did not find a relationship between search area size and detection probabilities. Our result is similar to Kéry [
4], who found no effect of search area on the detection of three common snake species in Europe, although he did find that detection rates increased as population sizes increased. For relatively common species, one would expect higher detection rates when population sizes are larger since there is a greater chance of encountering the organism. Endangered species often have reduced numbers across the landscape and, therefore, can have small population numbers regardless of patch size. In this instance, if population size remains relatively constant across patches, but patch sizes increase, we would expect detection rates to decrease as the survey area increases [
53]. Therefore, the next step might be to examine the interrelationship between density/abundance, patch size, and detection probabilities. Overall, our results highlight and reinforce the need to maximize search effort for endangered species.
Numerous other survey methods were used across the Eastern Massasauga’s range, including road cruising, drift fences, and cover board arrays. Researchers were effective at encountering enough individuals using drift fence–funnel trap arrays for radio-telemetry studies of Eastern and Western Massasaugas [
46,
47]. The only study to directly compare multiple methods found that effort required (in person-hours) is lower for the more passive methods of carpet squares, cover boards, and drift fences [
45]. The study also found sex-specific behavioral biases with females more likely to be encountered using coverboards and males more likely to be encountered with drift fences [
45]. The overall results do recommend the use of multiple encounter methods for monitoring projects [
45], and further work could be done to examine how detection probabilities differ among methods.
As MacKenzie et al. [
24] pointed out, rare and imperiled species are simultaneously the species most in need of information on state variables (e.g., occupancy, abundance) and vital rates, and the species for which such information is the most difficult to obtain. Imperfect detection probability (i.e., the probability of detecting a species when present is less than one) must be accounted for when estimating state variables [
24,
54]. While detection rates can covary with predictor variables, detection probability often is treated as a nuisance variable in the occupancy or abundance modeling process [
55,
56]. However, by focusing on detection probability, researchers can evaluate the effects of sampling covariates such as abiotic factors, sampling method, and survey type [
6,
57,
58]. Furthermore, this information can be used to determine the minimum number of visits required to a site to declare absence with a given level of confidence [
4,
27], which is a crucial first step in the development of survey protocols and monitoring programs.
The model we constructed provides good predictive capacity, but it was derived from the southern range limit in grassland habitat for a species with a historically broad distribution [
29,
59]. Over their broad range, Eastern Massasaugas occupy habitats ranging from wet to dry grasslands throughout the Midwest, to bogs, fens, and peatlands in the Great Lakes region, to open woodlands in northern Michigan and parts of Canada [
29,
59]. Therefore, not only may clinal variation exist but also variation due to habitat preferences. Thus, our model may only have the predictive capacity for grassland sites in the lower Midwest. We encourage the repetition of the methods and validation of the model across the species’ range.
The only previous study examining detection in Eastern Massasaugas used 54 search events over two years from May–August paired with radio-equipped snakes [
60]. From their 11 detections, they found detection probabilities approached 1.00 when the per searcher effort exceeded 90 min [
60], similar to our study. However, they did not find the strong coupling with temperatures we found, and although they found that detection probabilities approached 0.80 on cooler mornings, confidence intervals were wide [
60]. Future designs should also include stage classes as some stage classes may be more readily detectable than others as males tend to move more, gravid females tend to bask more often during gestation, and neonates/juveniles are extremely cryptic.
4.2. Conservation Applications
Giving a 0.10 spread from the maximal detection probability (in parentheses) up to a 0.95 probability for each variable, our data for Eastern Massasaugas indicate that, to maximize detection efficiency, surveys should be conducted preferably in burned habitat, between 9:30 and 16:00 h (0.43–0.50), between substrate temperatures of 16.7 and 30.2 °C (0.49–0.59), and between mean minimum three-day temperatures of 15.9 and 19.7 °C (0.70–0.80), under overall solar radiation of 373–844 J/s·m2 (0.50–0.60), for 1.8–2.4 h (0.85–0.95) per searcher. Under the Manager model (excluding daRad), the combinations of variables we present above should result in mean detection probabilities from 0.88 to 0.96, but with greater uncertainty.
To simplify the decision-making process for Eastern Massasauga surveys, we developed a spreadsheet-based detection tool as a guide. A researcher or land manager can enter forecasted values for the significant parameters that influenced Eastern Massasauga detection and determine the probability of detection for the upcoming survey, as well as the mean number of searches required. Additionally, one can estimate detection probability after the survey by entering the actual parameter values recorded during the sampling period. Such data could be used to determine if a site should be declared extirpated for management purposes. However, we do caution the use of the model outside of grassland habitats and, potentially, the region itself until it is further validated range-wide for the species. Given the decline of numerous snake species around the world, there is an urgent need to establish current status and distribution states for species of conservation concern. To deploy resources (i.e., time and money) in a cost-effective manner, and avoid wasted effort to establish presence or extirpation at a given site, we anticipate that our spreadsheet-based detection tool can be modified to meet the needs for other species.