Abstract

Surveying for rare animals can be difficult but using models to predict suitable habitat can guide sampling efforts. We used models to predict suitable habitat for the Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes (diving beetle hereafter), a dytiscid beetle that is known from 10 streams in central Wyoming. The diving beetle was a category-2 Candidate species for listing as Threatened or Endangered in the Endangered Species Act between 1989 and 1996, and was petitioned for listing in 2007, 2008, and 2013. Suitable habitat for the diving beetle was predicted using Maximum Entropy and Random Forest models in Wyoming. Both models predicted that the diving beetle was more likely to occur in intermittent streams with a gentle gradient, shallow water table, variable precipitation pattern, and high soil electrical conductivity, and in the warmest areas of Wyoming. We conducted surveys for the diving beetle at sites where the species had previously been found, and in new areas that were predicted suitable by our models to evaluate whether it is more widely distributed than indicated by previous estimates. We sampled beetles using dip nets and bottle-traps, and assessed water quality at each site. We collected the diving beetle at three sites in central Wyoming between 2010 and 2012 in small, alkaline, intermittent streams with disconnected pools. The aquatic habitat of the diving beetle is dynamic and our results suggest that annual precipitation patterns have a strong influence on the biogeography of this habitat. Our results also suggest that maintaining the hydrologic integrity of prairie streams in Wyoming is vital to the conservation of the diving beetle.

Introduction

Insects comprise ∼80% of the named and described animals of the world (Marshall 2006); however, only ∼11% of the animals listed as Threatened or Endangered under the U.S. Endangered Species Act (U.S. Endangered Species Act 1973, as amended [ESA]) are insects (NatureServe 2016). Invertebrate species are increasingly being petitioned for listing as Threatened or Endangered under the ESA, but basic information sufficient to decide if protection is necessary is often lacking. Basic information such as the geographic distribution of species, especially invertebrates, is often unknown. Surveying new areas for rare animals can be difficult, but using models that predict suitable habitat can be used to guide field efforts (Crawford and Hoagland 2010). Here, we use two types of models to direct field surveys for an apparently rare aquatic beetle.

The Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes (diving beetle hereafter) is an aquatic beetle known from 10 streams in central Wyoming (Leech 1966; Dahlem 1985; Keenan and Howard 1995; Miller 2002). The diving beetle lives in small intermittent streams with disconnected pools and a high concentration of salts (L.M. Tronstad, personal observation). The U.S. Fish and Wildlife Service has received formal requests since 2007 to consider whether the diving beetle warrants protection as Threatened or Endangered under the ESA (Forest Guardians 2007; WildEarth Guardians 2008, 2013) and is currently determining whether protection is warranted (U.S. Fish and Wildlife Service 2016).

We used species distribution models (SDMs) to evaluate suitable habitat for the diving beetle in Wyoming, and to guide field surveys for this species. The information we gathered with field surveys helped us estimate the status and distribution of the diving beetle and provided information for land managers to base decisions on. Our goals were to 1) identify stream reaches in Wyoming predicted to be suitable for the diving beetle, 2) collect aquatic beetles from suitable stream reaches across the state, and 3) estimate the current distribution of the diving beetle in Wyoming based on survey results.

Study Site

The known range of the diving beetle is in central Wyoming in the Powder River Basin and one site in the Wind River Basin (tributaries of the Missouri River). The area is semiarid with average annual precipitation ≤300 mm. For this study, we surveyed suitable stream reaches for the diving beetle across all the basins of Wyoming. The Wyoming basins are relatively high elevation (>1,000 m) shortgrass prairie ecosystems that receive <400 mm of precipitation each year, much of which is snow during spring. We surveyed in the Columbia, Missouri, and Colorado river basins, all of which have at least part of their headwaters in Wyoming. Within these basins, stream reaches for field surveys were selected based on relative probability outputs from two SDMs (see Species distribution modeling). Generally, we surveyed stream reaches with alkaline soils dominated by sagebrush in the uplands, which is typical of Wyoming basins.

Methods

Species distribution modeling

We fit SDMs using Maximum Entropy (Maxent; Phillips et al. 2006) and Random Forest (Breiman 2001; Cutler et al. 2007) to predict potential stream reaches suitable for the diving beetle throughout the state of Wyoming. Maxent models only use locations where the species has been detected and can be useful for predicting suitable habitat of species that are difficult to collect (Pearson et al. 2007). Random Forest models use locations where the species has been both detected and not detected to predict suitable habitat, and the model uses a different algorithm that does not require variable selection (Fox et al. 2017). Our sampling units were stream reaches from the Enhanced 100k Digital Line Graph stream data generated for the Wyoming Gap Analysis Project (Merrill et al. 1996). We matched previously surveyed stream reaches where the diving beetle was detected or not detected to the appropriate stream reaches from the Digital Line Graph data set using ArcGIS. We added a detected or not detected attribute to each of the associated reaches as the response variable. Eleven reaches where the diving beetle was detected and 27 reaches where the diving beetle was not detected were available for generating the initial models (Miller 2002) using Random Forest; the Maxent model used the same 11 reaches where the diving beetle was detected, with 10,000 randomly selected stream reaches used as background sites.

We chose 57 potential predictors to describe the stream reaches and conditions on the surrounding landscape. The Digital Line Graph data set included attributes for both Strahler stream order (Strahler 1952) and stream flow (i.e., perennial, intermittent, or ephemeral). Additionally, we attributed each of the stream reaches with 55 potential predictors representing various aspects of climate, topography, substrate, and vegetation (Table S1, Supplemental Material). We calculated stream slope as the maximum stream-reach elevation minus the minimum stream-reach elevation divided by the stream reach length. We calculated depth to water table, soil electrical conductivity, soil organic matter, and soil pH using Soil Data Viewer Tool 5.2 (USDA-NRCS 2008) and STATSGO data (Soil Survey Staff, no date). All other variables came from the predictor set generated by Keinath et al. (2010). Except for predictors pertaining to a particular stream reach, we summarized predictors using the Line Raster Intersection Statistics with the Length-Weighted Mean option within Hawth's Tools for each stream reach in Wyoming (Beyer 2004). For example, we summarized stream slope for each reach by calculating the change in elevation divided by the length of the stream reach.

We generated a Maxent model using ten-fold cross-validation with all 57 of the potential predictors to identify the most powerful variables that described the 11 collection locations for the diving beetle. We identified the top 12 predictors by summarizing the scores for overall contribution and jackknife contribution and eliminating variables too highly correlated (R2 > 0.8; Menard 2002) with other, higher ranked variables. We generated a final Maxent model using the top 12 predictors (depth to water table, stream flow, hottest-month mean maximum temperature, stream order, herbaceous cover, sagebrush cover, cottonwood cover, deciduous cover, pinion–juniper Pinus spp.–Juniperus spp. cover, variation of monthly precipitation, slope, and conifer cover). We created a binary prediction by applying the minimum training-presence threshold (Phillips et al. 2005) to the logistic prediction output for all stream reaches. We added logistic and binary predictions for both the detected training and background locations as attribute fields to the Digital Line Graph layer, based on the model.

We generated an initial Random Forest model using the RATTLE package (Williams 2009) in Program R (Venables et al. 2010) with the existing 11 detected points and 27 nondetected points. We used all 57 potential predictors, because Random Forest models perform best without variable selection and the model is robust to including many variables of moderate to low importance and correlated variables (Fox et al. 2017). Random Forest is particularly robust to a large number of predictors (Cutler et al. 2007), so we did no subsequent variable reduction. At each split, we evaluated seven predictors, with a total of 500 trees built. As with Maxent predictions, we wrote Random Forest predictions to each stream reach for both logistic and binary output (minimum training presence).

We chose initial sampling sites for major drainage basins in Wyoming by ranking stream reaches in each basin according to a score calculated as follows:

formula

where BinRF and LogRf were the binary and logistic predictions, respectively, of the initial Random Forest model; BinME and LogME were the binary and logistic predictions, respectively, of the Maxent model; Road was a binary value indicating whether a stream reach did (1) or did not (0) intersect a public road; and Public was assigned a value of 1.0 when the road–stream intersection occurred on public land or a value of 0.5 when the road–stream intersection did not occur on public land. The scoring was intended to prioritize stream reaches on public land accessible by public roads that were predicted as present by both models and had the highest probability according to both models. We selected the top 20 stream reaches in each basin according to the score as priority sampling sites for the initial surveys in early summer 2011 to equally sample in each area.

After we conducted the initial surveys during May and June of 2011 (see Field sampling section below), we generated a revised Random Forest model by incorporating 23 additional locations where we did not detect the diving beetle, using the same settings as the initial model. We did not make a new Maxent model, because we discovered no new detected points of the diving beetle. We generated a new score field using

formula

where LogRF2 was the logistic score for the revised Random Forest model. We used the logistic prediction to rank the top 20 stream reaches with the highest probability of predicting the diving beetle presence because we only used one model. For basins not surveyed during the initial survey period, we prioritized sampling sites based on a review of the original and revised scores by basin.

Field sampling

We collected aquatic beetles in 2010, 2011, and 2012 to estimate the current distribution of the diving beetle in Wyoming. We surveyed three streams in the Powder River Basin in July 2010 where the diving beetle was previously documented: Dugout Creek, Dead Horse Creek, and two locations in Cloud Creek. For each of the sites, we collected beetles 1 mile (∼1.6 km) up and downstream or until the stream dried. We recorded site location with a Global Positioning System unit (Garmin eTrex, ±10-m accuracy), described conditions, and took photos at all sites. We collected aquatic beetles using a D-frame dip net.

We sampled stream reaches throughout Wyoming in 2011 and 2012 predicted to have suitable habitat for the diving beetle. We attempted to visit 20–40 stream reaches with the highest predicted probability of containing the diving beetle in each basin in May through August of 2011. We took a broader approach in 2012 and we sampled in any publicly accessible stream with water within the vicinity of streams predicted suitable by the models. Additionally, we attempted to sample the 10 streams where the diving beetle had been collected previously. At sites with water, we recorded location using a Global Positioning System unit, described conditions, and took photos. We measured dissolved oxygen, pH, specific conductivity, temperature, and oxidation-reduction potential of the stream water using a Professional Plus made by Yellow Springs Instruments, which we calibrated daily. We collected aquatic beetles using either a D-frame dip net or bottle-traps (Aiken and Roughley 1985) left in the stream overnight. We sampled pools by pulling the dip net through the water multiple times depending on the pool size to collect invertebrates living there. We swept near the sediment surface, but we avoided collecting sediment. We preserved aquatic invertebrates in 75% ethanol, and pinned or stored them in ethanol. We identified dytiscid beetles using Larson et al. (2000), and we consulted with Dr Boris Kondratieff (entomologist at Colorado State University) and voucher specimens to verify identification. The diving beetle is a small, light-colored beetle (Figure 1a) identified by the sinuate profemur of males (Figure 1b; Anderson 1983).

Figure 1.

(a) An adult Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes captured in Wyoming during summer 2010 is distinctive with its light-colored elytra (dorsal view); and (b) males can be identified with their sinuate profemur.

Figure 1.

(a) An adult Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes captured in Wyoming during summer 2010 is distinctive with its light-colored elytra (dorsal view); and (b) males can be identified with their sinuate profemur.

Results

Species distribution models

The area under the Receiver Operating Characteristic Curve (Fielding and Bell 1997) for training data was 0.985 (Table S2, Supplemental Material). The Maxent model indicated that the distribution of the diving beetle was most closely associated with various aspects of water regimes, climate, and vegetation cover (Figure 2a). Specifically, the diving beetle was predicted to be more likely to occur in intermittent streams with a Strahler order of 3–4, a gentle stream gradient, and in areas with a shallow water table (distance to ground water). The model also indicated that the species was associated with sparse forest and herbaceous cover, moderate sagebrush cover, and in warm areas of the state with highly variable seasonal precipitation patterns.

Figure 2.

Species distribution models for the Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes in Wyoming using (a) the Maximum Entropy (Maxent) algorithm, (b) the initial Random Forest algorithm, and (c) the revised Random Forest algorithm. Maxent models only use reaches where the diving beetle was detected prior to 2011 and the top 12 predictors. Conversely, Random Forest models use reaches where the diving beetle was both detected and not detected prior to 2011, and all predictors. The initial Random Forest model used all data we had before beginning surveys (May 2011) and the revised Random Forest model combined the initial data and data from the first 2 mo of surveys (prior to July 2011). All maps are projected in Wyoming Lambert.

Figure 2.

Species distribution models for the Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes in Wyoming using (a) the Maximum Entropy (Maxent) algorithm, (b) the initial Random Forest algorithm, and (c) the revised Random Forest algorithm. Maxent models only use reaches where the diving beetle was detected prior to 2011 and the top 12 predictors. Conversely, Random Forest models use reaches where the diving beetle was both detected and not detected prior to 2011, and all predictors. The initial Random Forest model used all data we had before beginning surveys (May 2011) and the revised Random Forest model combined the initial data and data from the first 2 mo of surveys (prior to July 2011). All maps are projected in Wyoming Lambert.

The initial Random Forest model (Figure 2b) had an out-of-bag error estimate (Cutler et al. 2007) of 13.2%, though the sensitivity based on out-of-bag samples was just 63.6%, indicating relatively poor identification of detected locations. The predictors identified as most important to the Random Forest model were similar to those selected by the Maxent model, with the addition of several soil chemistry and soil texture predictors. Particularly, electrical conductivity (a measure of the concentration of salts in soil), soil texture, and water-holding capacity were selected as the most important variables to model accuracy. Electrical conductivity was much higher at training presence points than at training absence points.

The revised Random Forest model (Figure 2c) had a better out-of-bag error estimate of 11.5%, but a worse sensitivity score of 45.5%. The most important variable in the revised model was soil percent clay, with stream reaches at training presence sites having a higher mean clay percentage (28.0%, SD = 8.8%) than stream reaches at training absence sites (19.9%, SD = 5.4%). As with the Maxent model, hottest-month mean maximum temperature was identified as important in the revised Random Forest model; however, this variable was relatively unimportant to the initial Random Forest model. Both the Maxent model and the revised Random Forest model suggested that the diving beetle is more likely to occur in areas with the highest hottest-month mean temperatures in Wyoming.

Field sampling

We collected the diving beetle in two streams—Dead Horse Creek and Cloud Creek—in 2010 (Table S3, Supplemental Material). We collected 24 specimens of the diving beetle: 23 in multiple pools in Cloud Creek and 1 in Dead Horse Creek. We did not collect the diving beetle from Dugout Creek, the original location from which the species was described.

We collected invertebrates in the Green, Wind, Yellowstone, Bighorn, Great Divide, Little Snake, North Platte, Powder, Little Missouri, Belle Fourche, and Cheyenne river basins in 2011, because the models suggested these basins contained suitable habitat for the diving beetle. Of the 305 sites we attempted to visit, 134 sites were dry, 131 sites were inaccessible (locked gate, impassible road, etc.), and 40 sites contained water and were sampled (Figure 3a). We only collected two specimens of the diving beetle at one location—Dead Horse Creek—from the 40 sites sampled. We did not collect the diving beetle at Dugout Creek.

Figure 3.

We surveyed for the Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes between 2010 and 2012 in Wyoming. (a) The map shows all the sites we surveyed in 2010 and 2011, and Miller's 2002 survey. We selected stream reaches that had the highest probability of having suitable habitat in each river basin in 2011, but many stream reaches were dry or inaccessible. (b) We drove public roads through areas predicted to have suitable habitat for the diving beetle and sampled inundated streams on public land to sample more stream reaches in 2012. Both maps are projected in Wyoming Lambert.

Figure 3.

We surveyed for the Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes between 2010 and 2012 in Wyoming. (a) The map shows all the sites we surveyed in 2010 and 2011, and Miller's 2002 survey. We selected stream reaches that had the highest probability of having suitable habitat in each river basin in 2011, but many stream reaches were dry or inaccessible. (b) We drove public roads through areas predicted to have suitable habitat for the diving beetle and sampled inundated streams on public land to sample more stream reaches in 2012. Both maps are projected in Wyoming Lambert.

We collected invertebrates from 206 sites in 2012, and aquatic beetles lived at 74% of these locations (Figure 3b). Seven of the 11 known sites were accessible and held water in 2012. We sampled Dugout Creek (3 dates), Dead Horse Creek (3 dates), Cloud Creek (2 locations on 3 dates), Government Creek (1 date), Teapot Creek (1 date), and Conant Creek (1 date). We did not collect the diving beetle at any of the known sites. We collected four female diving beetles in a tributary stream of Murphy Creek in May. The diving beetle is sexually dimorphic and females lack the diagnostic sinuate profemur that males possess. Therefore, female diving beetles are more difficult to identify and their diagnosis is less certain than male specimens. The diving beetle had not been collected at this site in the Powder River Basin previously. The stream flowed at a slow velocity and alkaline salts were deposited on the stream banks. The stream consisted of soft, fine sediments with algal mats floating on the water surface. Beetles (Order Coleoptera) were the dominant invertebrates in the stream and we collected three other beetles along with the diving beetle (H. patruelis, H. nubilus, and Stictotarsus griseostriatus; Family Dytiscidae). The tributary of Murphy Creek increased the list of sites known to be occupied by the diving beetle to 12.

Basic water quality from the sampled reaches varied (Table 1). Generally, most aquatic ecosystems had ample dissolved oxygen for aquatic invertebrates. The mean specific conductivity of reaches we visited was well above the average conductivity of rivers worldwide (∼250 μS/cm; Wetzel 2001). pH was basic (>7), as is commonly found around Wyoming. The aquatic ecosystems we sampled varied between oxidizing (e.g., >200 mV) and reducing (e.g., <200 mV) conditions. The water quality of Dead Horse Creek was similar to the other streams sampled across Wyoming, except pH and specific conductivity were above average (Table 1). In contrast, most water quality parameters from the tributary stream to Murphy Creek differed from averages across the basins of Wyoming.

Table 1.

Basic water quality at all sites sampled for the Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes in Wyoming during 2011 and 2012. We report minimum, median, mean, and maximum water temperature; dissolved oxygen (DO); specific conductivity (SPC); pH; and oxidation–reduction potential (ORP) for all sites sampled (n = 246). The water quality at Deadhorse Creek and tributary to Murphy Creek, where we collected the diving beetle, were within the range of all the sites we sampled.

Basic water quality at all sites sampled for the Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes in Wyoming during 2011 and 2012. We report minimum, median, mean, and maximum water temperature; dissolved oxygen (DO); specific conductivity (SPC); pH; and oxidation–reduction potential (ORP) for all sites sampled (n = 246). The water quality at Deadhorse Creek and tributary to Murphy Creek, where we collected the diving beetle, were within the range of all the sites we sampled.
Basic water quality at all sites sampled for the Narrow-footed Hygrotus Diving Beetle Hygrotus diversipes in Wyoming during 2011 and 2012. We report minimum, median, mean, and maximum water temperature; dissolved oxygen (DO); specific conductivity (SPC); pH; and oxidation–reduction potential (ORP) for all sites sampled (n = 246). The water quality at Deadhorse Creek and tributary to Murphy Creek, where we collected the diving beetle, were within the range of all the sites we sampled.

Observations of vegetation and stream characteristics from the sampled reaches were similar to the sites where the diving beetle was collected. The majority of sampled sites were used as rangeland and a few had energy development, but we do not have enough data to address the effects of grazing or development on the diving beetle. The dominant landscape vegetation was sagebrush Artemisia spp. (94% of sites), greasewood Sarcobatus spp.; (42% of sites), shortgrass prairie (71% of sites), rabbitbrush Chrysothamnus spp. and Ericameria spp. (37% of sites), and willows Salix spp. (21% of sites). Estimated stream width varied between 0.15 and 300 m (mean width = 10 m). Stream depth varied between 0.06 and 3 m (mean depth = 0.6 m). The dominant substrate at all sites consisted of clay and/or silt. The dominate vegetation at sites where we collected the diving beetle was sagebrush, greasewood, rabbitbrush, and shortgrass prairie, and the land was used as rangeland with little development. Cloud and Deadhorse creeks were small and formed a series of disconnected pools when we collected the diving beetle (<1 m in width, <0.3 m in depth). Tributary to Murphy Creek had no visible flow and was not reduced to pools. Little to no aquatic vegetation was present in streams, except for the tributary to Murphy Creek, which had floating algae. All streams where we collected the diving beetle had precipitated salts around the margins.

Discussion

The diving beetle has been collected from 10 streams in central Wyoming (9 streams in the Powder River Basin and 1 stream in the Wind River Basin; Leech 1966; Dahlem 1985; Keenan and Howard 1995; Miller 2002); however, our SDMs suggested that the diving beetle may be more widely distributed. Statewide sampling yielded one new stream where the diving beetle was collected and the new stream reach was near known sites. We collected the diving beetle from two sites in 2010, and one site in both 2011 and 2012. The lowlands did not receive the spring snows critical for providing the essential moisture that penetrates the soil and recharges shallow groundwater in 2011 and 2012 (Knight et al. 2014), likely limiting the distribution of the diving beetle. Distance to ground water was an important predictor in our models, and lower ground-water levels would result in fewer intermittent streams with surface water. In fact, many stream reaches were dry, even when we surveyed in May. Protecting shallow ground water is probably the most important strategy to protect these diving beetles. Increasing surface water (e.g., discharge of water produced in conjunction with oil and gas development) or pumping shallow ground water may reduce the available habitat for the diving beetle.

In addition to anthropogenic effects on groundwater, annual variability of precipitation may affect the distribution of the diving beetle. The diving beetle appeared to be captured at more sites when the Palmer Drought Severity Index (https://www.ncdc.noaa.gov/temp-and-precip/drought/historical-palmers/) suggested near-average precipitation levels (e.g., 1985, 1992, 1994, and 2010) compared with years that were very wet or very dry (e.g., 1993, 2011, and 2012). We suspect that suitable habitat for the diving beetle is quite dynamic in both space and time. Our models predicted the best habitat as intermittent streams that are seasonally wet. Suitable habitat may shift position across days, months, and years, with adult diving beetles likely changing their position in response on similar time scales. Dytiscid beetles moved with the declining water line as water levels decreased in a floodplain (Tronstad et al. 2005) and dytiscids flew to other habitats when water depth decreased in a desert stream (Velasco and Millan 1998). Less habitat is probably available to the diving beetle during drought years, when intermittent streams contain less water. We do not know what refuge or survival strategy the diving beetle use during drought. The beetle Sclerocyphon (Family Psphenidae) dispersed to perennial water to survive drought periods in Grampians National Park, Australia (Chester et al. 2015). Conversely, the diving beetle may find abundant habitat during wet years when more intermittent streams contain surface water. Thus, the suitability of intermittent streams as habitat for the diving beetle may vary daily to annually, depending on local precipitation patterns, leading to complex and unpredictable patterns of occupancy by the diving beetle.

Surveying for the diving beetle during a wetter year may reveal that the diving beetle is more widespread. Our models predicted that suitable habitat may be located in several basins, including the Bighorn, Green, and North Platte river basins. Given the dynamic nature of habitat for the diving beetle and the number of stream reaches that did not contain water or that we could not access, the diving beetle may be collected from more sites in a larger area of Wyoming during wetter years. However, such surveys could also reveal that the diving beetle may be restricted to a small range. We collected the diving beetle in only two streams where the diving beetle had previously been collected over a 3-y period. The diving beetle was not found in the other known streams over this period, which suggests that the diving beetle has a limited range. Other beetles are known to have small ranges. For example, the most endemic species in Austria occurred within the order Coleoptera, with 80% of these species having ranges <875 km2 (Rabitsch et al. 2016). The federally Endangered Hungerford's Crawling Water Beetle Brychius hungerfordi (Haliplidae) is known from six streams in Ontario and Michigan (ESA 1973; Tansy 2006). Other Hygrotus species also have limited ranges, such as H. curvipes, H. artus, H. fontinalis, and H. pedalis (Anderson 1983).

Most members of the genus Hygrotus have very specific habitat requirements and are usually found in standing water (Anderson 1983; Larson et al. 2000). However, the diving beetle has only been collected from pools of intermittent streams (Miller 2002). Based on available data, the diving beetle lives in intermittent streams that are often a series of disconnected pools. We sampled a variety of aquatic habitats during our statewide surveys including springs, ponds, pools, intermittent streams, and perennial streams; however, we only collected the diving beetle from pools of intermittent streams. The intermittent streams usually had higher concentrations of dissolved salts as indicated by precipitated salts at the water's edge and conductivity measurements.

The diving beetle may have a competitive advantage in certain aquatic ecosystems by being tolerant of dissolved salts. The diving beetle can be the dominant beetle in habitats where they are collected (Miller 2002; K.M. Brown, personal observation). Other Hygrotus species are able to live in or are restricted to waters with relatively high concentrations of salts because they possess special osmoregulatory capabilities. For example, H. salinarius lives in a Canadian lake with higher conductivity than ocean water (Tones 1978). This species survives as an adult by having hyposmotic haemolymph and hyperosmotic urine relative to the water. Fifty percent of H. impressopunctatus individuals survived in seawater for 4 d, indicating that they can tolerate high concentrations of salts for short periods (Minakawa et al. 2001). We collected the diving beetle in pools with conductivity up to 20,000 μS/cm; however, the diving beetle may not live in waters with as high of conductivity as H. salinarius (Miller 2013). Our models predicted that soil conductivity was higher at occupied sites; however, we do not know the maximum concentration of salts in which the diving beetle can survive.

Available information suggests that the diving beetle lives in specialized habitats within a restricted range (∼30,000-km2 range within which 11 occupied streams are known). Modeling the distribution of such spatially restricted species can be difficult because of the small number of detected points which decreases the sensitivity of SDMs (Wisz et al. 2008). Distribution modeling typically uses broad environmental characteristics to predict areas suitable for occupation. Although such modeling is useful to understand major trends in habitat selection, the species' actual occupancy may be limited by environmental gradients and processes that occur at finer scales. Fine-scale variables are not available to use for modeling distribution, but such information can be developed though field studies. For example, information on substrate, pool vegetation, pool depth and width, and canopy cover may be related to habitat suitability for the diving beetle. Understanding both levels of environmental limitations (broad- and fine-scale variables) can lead to a better understanding of the spatial patterns of occupancy for the diving beetle and similar species. Predicting suitable habitat for a rare species is useful to guide field survey, but such predictions should not be interpreted as the species' distribution (Crawford and Hoagland 2010). Our models overpredicted suitable habitat, which is true for many rare species like the diving beetle that live in specialized habitats (Crawford and Hoagland 2010; Campbell and Hilderbrand 2017). We do not interpret the overpredictions as problematic because we are using the models to guide surveys, not to describe their distribution.

We discovered one new site during 2 y of state-wide surveys and collected the diving beetle at two known steams during 3 y of surveys. Despite using similar methods and sampling during similar times of year, the diving beetle was collected at more sites in the past (Leech 1966; Dahlem 1985; Keenan and Howard 1995; Miller 2002), compared with 2010 through 2012. Overall, little basic information is known about the diving beetle, such as life-history information, overwinter strategies, and drought refuges. The morphology of larvae may be described to learn more about the early life-stages of the diving beetle. Currently, adults of the diving beetle appears to be restricted to intermittent streams consisting of disconnected pools with relatively high concentrations of dissolved salts in central Wyoming. Adults may be collected in marginal habitat because they are able to move among streams; however, larval diving beetles are bound to the aquatic habitat where eggs were laid. Identifying larval habitat is vital to managing the species. Similarly, H. curvipes lives in temporary aquatic habitats where adults disperse and recolonize areas, allowing the species to persist across the landscape (Hafernik 1992). Although the diving beetle does not appear to be abundant in the basins of Wyoming, additional surveys are needed to assess the distribution of the diving beetle, especially in wet years when the diving beetle may be more widely collected.

Supplemental Material

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.

Table S1. Potential predictor layers used in building Maximum Entrophy (Maxent) and Random Forest species distribution models for Wyoming during 2011 for the Narrow-footed Hygrotus predaceous diving beetle Hygrotus diversipes. The top 12 variables were included in the Maxent model and all variable were used in the Random Forest model.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S1 (15 KB DOCX).

Table S2. Maximum Entropy model results for a species distribution model of the Narrow-footed Hygrotus predaceous diving beetle Hygrotus diversipes in Wyoming built in 2011.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S2 (133 KB DOCX).

Table S3. The Narrow-footed Hygrotus diving beetle Hygrotus diversipes was discovered in 1964 and is currently known from 12 locations all in Wyoming. All known locations, collectors and dates of collection are listed through 2012. Coordinates are in datum NAD83.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S3 (13 KB DOCX).

Reference S1. Dahlem G. 1985. (No Title). Casper, Wyoming: Report prepared by the Bureau of Land Management.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S4 (492 KB PDF).

Reference S2. Forest Guardians. 2007. A petition to list 206 critically imperiled or imperiled species in the Mountain Prairie Region of the United States as Threatened or Endangered under the Endangered Species Act. Santa Fe, New Mexico: Forest Guardians.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S5; also available at http://www.wildearthguardians.org/support_docs/petition_protection-206-species-r6_7-24-07.pdf (468 KB PDF).

Reference S3. Keenan LC, Howard T. 1995. Status report on the Narrow-foot Diving Beetle Hygrotus diversipes Leech. Report of Professional Entomological Services Technology, Inc. to U.S. Fish and Wildlife Service, Lakewood, Colorado.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S6 (12560 KB PDF).

Reference S4. Keinath DA, Andersen MD, Beauvais GP. 2010. Range and modeled distribution of Wyoming's Species of Greatest Conservation Need. Laramie: Wyoming Natural Diversity Database, University of Wyoming.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S7; also available at http://www.uwyo.edu/wyndd/reports-and-publications/ (602 KB PDF).

Reference S5. Merrill EH, Kohley RW, Herdendorf ME, Reiners WA, Driese KL, Marrs RW Anderson SH. 1996. The Wyoming GAP analysis project: final report. Laramie: University of Wyoming.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S8 (38126 KB PDF).

Reference S6. Miller KB. 2002. Report on the diving beetle Hygrotus diversipes Leech (Coleoptera: Dytiscidae). Ithaca, New York: Cornell University.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S9 (1352 KB PDF).

Reference S7. WildEarth Guardians. 2008. A petition requesting emergency listing of 32 species under the Endangered Species Act. Santa Fe, New Mexico: WildEarth Guardians.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S10; also available at http://www.wildearthguardians.org/support_docs/petition_emergency-listing_6-12-08.pdf (333 KB PDF).

Reference S8. WildEarth Guardians. 2013. Petition to list the Narrow-foot Hygrotus Diving Beetle (Hygrotus diversipes) under the Endangered Species Act. Santa Fe, New Mexico: WildEarth Guardians.

Found at DOI: http://dx.doi.org/10.3996/112016-JFWM-085.S11; also available at http://www.wildearthguardians.org/site/DocServer/Wyoming_diving_beetle_listing_petition.pdf?docID=11022 (1072 KB PDF).

Acknowledgments

We thank Cody Bish, Ken Brown, Kyle Hack, Ben Anson, and Devin Baumer for searching for the diving beetle across much of Wyoming. We thank Aaron Clark and the Wyoming Governor's Office for supporting the research. The late Dr R. E. Roughley at the University of Manitoba provided voucher specimens of the diving beetle, and Scott Shaw and Boris Kondratieff provided valuable comments. We thank Kelly Miller for providing locations of his 2002 survey; and Gary Beauvais, Aaron Clark, Joel Brown, two anonymous reviewers, and the Associate Editor for providing comments that improved the manuscript.

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

References

Aiken
RB.
Roughley
RE.
1985
.
An effective trapping and marking method for aquatic beetles
.
Proceedings of the Academy of Natural Sciences of Philadelphia
137
:
5
7
.
Anderson
RD.
1983
.
Revision of the nearctic species of Hygrotus groups IV, V, and VI (Coleoptera: Dytiscidae)
.
Annals of the Entomological Society of America
76
:
173
196
.
Beyer
HL.
2004
.
Hawth's analysis tools for ArcGIS
.
Available
: www.spatialEcology andDistribution.com/htools (October 2017).
Breiman
L.
2001
.
Random forests
.
Machine Learning
45
:
5
32
.
Campbell
CA.
Hilderbrand
RH.
2017
.
Using maximum entropy to predict suitable habitat for the endangered dwarf wedgemussel in the Maryland Coastal Plain
.
Aquatic Conservation: Marine and Freshwater Ecosystems
27
:
462
475
.
Chester
ET.
Miller
AD.
Valenzuela
I.
Wickson
SJ.
Robson
BJ.
2015
.
Drought survival strategies, dispersal potential and persistence of invertebrate species in an intermittent stream landscape
.
Freshwater Biology
60
:
2066
2083
.
Crawford
PHC.
Hoagland
BW.
2010
.
Using species distribution models to guide conservation at the state level: the endangered American burying beetle (Nicrophorus americanus) in Oklahoma
.
Journal of Insect Conservation
14
:
511
521
.
Cutler
DR.
Edwards
TC.
Beard
KH.
Cutler
A.
Hess
KT.
2007
.
Random forests for classification in ecology
.
Ecology
88
:
2783
2792
.
Dahlem
G.
1985
. (
No Title
).
Casper, Wyoming
:
Report prepared by the Bureau of Land Management
(see Supplemental Material, Reference S1, http://dx.doi.org/10.3996/112016-JFWM-085.S4)
.
Fielding
AH.
Bell
JF.
1997
.
A review of the methods for the assessment of prediction errors in conservation presence/absence models
.
Environmental Conservation
24
:
38
49
.
Forest
Guardians.
2007
.
A petition to list 206 critically imperiled or imperiled species in the Mountain Prairie Region of the United States as Threatened or Endangered under the Endangered Species Act
.
Santa Fe, New Mexico
:
Forest Guardians
.
Fox
EC.
Hill
RA.
Leibowitz
SG.
Olsen
AR.
Thornbrush
DJ.
Weber
MH.
2017
.
Assessing the accuracy and stability of variable selection methods for random forest modeling in ecology
.
Environmental Monitoring and Assessment
189
:
316
. DOI: .
Hafernik
J.
1992
.
Threats to invertebrate biodiversity: implications for conservation strategies
.
Pages
171
195
in
Fielder
PL.
Jain
SK.
editors
.
Conservation biology: the theory and practice of nature conservation, preservation and management
.
New York, New York
:
Chapman & Hall
.
Keenan
LC.
Howard
T.
1995
.
Status report on the Narrow-foot Diving Beetle Hygrotus diversipes Leech
.
Report of Professional Entomological Services Technology, Inc. to U.S. Fish and Wildlife Service, Lakewood, Colorado
(see Supplemental Material, Reference S3, http://dx.doi.org/10.3996/112016-JFWM-085.S6)
.
Keinath
DA.
Andersen
MD.
Beauvais
GP.
2010
.
Range and modeled distribution of Wyoming's Species of Greatest Conservation Need
.
Laramie
:
Wyoming Natural Diversity Database, University of Wyoming
(see Supplemental Material, Reference S4, http://dx.doi.org/10.3996/112016-JFWM-085.S7); also available: http://www.uwyo.edu/wyndd/reports-and-publications/ (October 2017)
.
Knight
DH.
Jones
GP.
Reiners
WA.
Romme
WH.
2014
.
Mountains and plains: the ecology of Wyoming landscapes. 2nd edition
.
New Haven, Connecticut
:
Yale University Press
.
Larson
DJ.
Alarie
Y.
Roughley
RE.
2000
.
Predaceous diving beetles (Coleoptera: Dytiscidae) of the nearctic region, with emphasis on the fauna of Canada and Alaska
.
Ottawa, Ontario, Canada
:
National Research Council of Canada Research Press
.
Leech
HB.
1966
.
The pedalis-group of Hygrotus, with descriptions of two new species and a key to the species (Coleoptera: Dytiscidae)
.
Proceedings of the California Academy of Sciences
33
:
481
498
.
Marshall
SA.
2006
.
Insects: their natural history and diversity
.
New York, New York
:
Firefly Books Ltd
.
Menard
S.
2002
.
Applied logistic regression analysis. 2nd edition
.
Thousand Oaks, California
:
Sage
.
Merrill
EH.
Kohley
RW.
Herdendorf
ME.
Reiners
WA.
Driese
KL.
Marrs
RW
Anderson
SH.
1996
.
The Wyoming GAP analysis project: final report
.
Laramie
:
University of Wyoming
(see Supplemental Material, Reference S5, http://dx.doi.org/10.3996/112016-JFWM-085.S8)
.
Miller
KB.
2002
.
Report on the diving beetle Hygrotus diversipes Leech (Coleoptera: Dytiscidae)
.
Ithaca, New York
:
Cornell University
(see Supplemental Material, Reference S6, http://dx.doi.org/10.3996/112016-JFWM-085.S9)
.
Miller
KB.
2013
.
Notes on flight and respiration at the water surface by Hygrotus salinarius (Wallis) (Coleoptera: Dytiscidae)
.
The Coleopterists Bulletin
67
:
444
446
.
Minakawa
N.
Kurowski
KL.
Yabe
M.
2001
.
Salinity tolerance of diving beetle Hygrotus impressopunctatus (Coleoptera: Dytiscidae) and its implication for insect dispersal
.
Entomological Science
4
:
393
397
.
NatureServe
.
2016
.
NatureServe explorer
.
Available: http://explorer.natureserve.org/ (October 2017)
.
Pearson
RG.
Raxworthy
CJ.
Nakamura
M.
Peterson
AT.
2007
.
Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar
.
Journal of Biogeography
34
:
102
117
.
Phillips
S.
Dudík
M.
Schapire
R.
2005
.
Maxent software for species distribution modeling
. .
Phillips
SJ.
Anderson
RP.
Schapire
RE.
2006
.
Maximum entropy modeling of species geographic distributions
.
Ecological Modeling
190
:
231
259
.
Rabitsch
W.
Graf
W.
Huemer
P.
Kahlen
K.
Komposch
C.
Paill
W.
Reischütz
A.
Reischütz
PL.
Moser
D.
Essl
F.
2016
.
Biogeography and ecology of endemic invertebrate species in Austria: a cross-taxon analysis
.
Basic and Applied Ecology
17
:
95
105
.
Soil Survey Staff
.
No date
.
U.S. General Soil Map (STATSGO2) for Wyoming
.
U.S. Department of Agriculture, Natural Resources Conservation Service
. .
Strahler
AN.
1952
.
Hypsometric (area–altitude) analysis of erosional topography
.
Geological Society of America Bulletin
63
:
1117
1142
.
Tansy
CL.
2006
.
Hungerford's crawling water beetle (Brychius hungerfordi) recovery plan
.
Prepared by the U.S. Fish and Wildlife Service
. .
Tones
PI.
1978
.
Osmoregulation in adults and larvae of Hygrotus salinarius Wallis (Coleoptera, Dytiscidae)
.
Comparative Biochemistry and Physiology
60
:
247
250
.
Tronstad
LM.
Tronstad
BP.
Benke
AC.
2005
.
Invertebrate responses to decreasing water levels in a subtropical river floodplain wetland
.
Wetlands
25
:
583
593
.
[ESA] U.S. Endangered Species Act of 1973, as amended, Pub. L. No. 93-205, 87 Stat. 884 (Dec. 28, 1973)
. .
[USDA–NRCS] U.S. Department of Agriculture-Natural Resource Conservation Service
.
2008
.
Soil data viewer 5.2 user guide
. .
U.S. Fish and Wildlife Service
.
2016
.
Endangered and Threatened wildlife and plants; 90-day findings on 17 petitions. A proposed rule by the Fish and Wildlife Service on 01/12/2016
.
Federal Register
81
:
1368
1375
. .
Velasco
J.
Millan
A.
1998
.
Insect dispersal in a drying desert stream: effects of temperature and water loss
.
The Southwestern Naturalist
43
:
80
87
.
Venables
WN.
Smith
DM.
R Development Core Team
.
2010
.
An introduction to R. Version 2.12.1
. .
Wetzel
RG.
2001
.
Limnology: lake and river ecosystems
.
New York, New York
:
Academic Press
.
WildEarth Guardians
.
2008
.
A petition requesting emergency listing of 32 species under the Endangered Species Act
.
Santa Fe, New Mexico: WildEarth Guardians (see Supplemental Material, Reference S7
,
2018)
.
WildEarth Guardians
.
2013
.
Petition to list the Narrow-foot Hygrotus Diving Beetle (Hygrotus diversipes) under the Endangered Species Act
.
Santa Fe, New Mexico: WildEarth Guardians
.
Williams
G.
2009
.
Rattle: a data mining GUI for R
.
R Journal
1
:
45
55
.
Wisz
MS.
Hijmans
RJ.
Li
J.
Peterson
AT.
Graham
CH.
Guisan
A.
2008
.
Effects of sample size on the performance of species distribution models
.
Diversity and Distributions
14
:
763
773
.

Author notes

Citation: Tronstad LM, Brown KM, Anderson MD. 2018. Using species distribution models to guide field surveys for an apparently rare aquatic beetle. Journal of Fish and Wildlife Management 9(1):330–339; e1944-687X. doi:10.3996/112016-JFWM-085

The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views of the U.S. Fish and Wildlife Service.

Supplementary data