Advertisement

Hydrobiologia

, Volume 808, Issue 1, pp 153–161 | Cite as

Estimating factors influencing the detection probability of semiaquatic freshwater snails using quadrat survey methods

  • Elizabeth L. Roesler
  • Timothy B. Grabowski
Primary Research Paper

Abstract

Developing effective monitoring methods for elusive, rare, or patchily distributed species requires extra considerations, such as imperfect detection. Although detection is frequently modeled, the opportunity to assess it empirically is rare, particularly for imperiled species. We used Pecos assiminea (Assiminea pecos), an endangered semiaquatic snail, as a case study to test detection and accuracy issues surrounding quadrat searches. Quadrats (9 × 20 cm; n = 12) were placed in suitable Pecos assiminea habitat and randomly assigned a treatment, defined as the number of empty snail shells (0, 3, 6, or 9). Ten observers rotated through each quadrat, conducting 5-min visual searches for shells. The probability of detecting a shell when present was 67.4 ± 3.0%, but it decreased with the increasing litter depth and fewer number of shells present. The mean (± SE) observer accuracy was 25.5 ± 4.3%. Accuracy was positively correlated to the number of shells in the quadrat and negatively correlated to the number of times a quadrat was searched. The results indicate quadrat surveys likely underrepresent true abundance, but accurately determine the presence or absence. Understanding detection and accuracy of elusive, rare, or imperiled species improves density estimates and aids in monitoring and conservation efforts.

Keywords

Wetland Spring Endangered species Survey Invertebrates Conservation evaluation 

Introduction

Imperfect detection of elusive, rare, or patchily distributed species is a major impediment to their conservation, as it can generate high levels of uncertainty surrounding estimates of distribution and population density. However, detection is rarely perfect due to observer error, species rarity, species behavior, environmental conditions, etc. (MacKenzie et al., 2002; Martin et al., 2005; Kellner & Swihart, 2014; Bouchet & Meeuwig, 2015). Therefore, not accounting for imperfect detection when analyzing survey data prevents the researcher from evaluating whether variation in the raw count data from a survey is due to factors influencing detectability or actual changes in the distribution or abundance of the targeted species (MacKenzie et al., 2002). Unfortunately, many standardized monitoring protocols do not take these issues into consideration and are generally implemented under the assumption that instances when a species is not observed indicate the absence of the species from that survey point and/or sample (Royle & Link, 2006). In reality, two mutually exclusive conclusions could be drawn from not observing a target species at a survey location: either the species was truly absent from the site, or the species was present but went undetected (Mackenzie et al., 2002; Royle & Link, 2006). Nondetection is not equivalent to the absence; thus, analyses involving survey data, which do not account for detectability effects, can become inflated with false zeros and be negatively biased (MacKenzie et al., 2002; Martin et al., 2005), resulting in underestimates of the population in question (Moilanen, 2002; Martin et al., 2005). Nondetections do not occur at a constant rate and are more likely with rare or cryptic species, which are characterized by small population sizes and/or morphological or behavioral features that may interfere with detection, and when sampling effort is insufficient in time or space (Gu & Swihart, 2004).

When an individual of the target species is detected at a survey point, the probability of its detection has been influenced by an array of factors and variables. Root causes of inconsistent detections within a survey are composed of discrepancies among observers, misidentification of the correct focal species, behavioral patterns of the study organism, and environmental variability. Even though survey methodology is standardized for a particular monitoring effort, surveys generally involve sampling with different observers with highly varying levels of ability, e.g., visual or auditory acuity; skill; experience; and effectiveness, e.g., levels of interest, fatigue, etc. (Anderson, 2001). Variation among observers can affect the probability of detection and potentially has substantial effects on count indexes (Royle & Link, 2006). Similarly, detection probability can be influenced by environmental or habitat features such as precipitation, sunlight, topography, vegetation height and density, disturbance, etc. (Anderson, 2001; Gu & Swihart, 2004; Bouchet & Meeuwig, 2015). Finally, species-specific characteristics affect detectability (Anderson, 2001; Bouchet & Meeuwig, 2015). The issues associated with surveying wildlife and occupancy analyses have been thoroughly addressed in the literature (Anderson, 2001; MacKenzie et al., 2002; Moilanen, 2002; Gu & Swihart, 2004; Royle & Link, 2006), and the overarching conclusion is that understanding detectability, as well as the myriad factors that can influence it, is critical for producing accurate population estimates. Unfortunately, expertise and training to gain accurate quantitative information on population dynamics, including detection, can be costly in terms of time and money for resource managers (Gu & Swihart, 2004), and methods used to estimate detection probabilities at each site and per observer can be time consuming, inefficient, and beyond the scope of the research questions (MacKenzie et al., 2002). Empirically estimating detection can cost extra effort and resources while still improving the findings of count surveys.

Endemic North American freshwater gastropods exemplify the need to account for imperfect detection when designing monitoring surveys. While there are some species that can be accurately and relatively easily sampled quantitatively (e.g., Viviparidae, Pleuroceridae, Semisulcospiridae, Lymnaeidae), others can be challenging because of similarities in size, color, and shape, making accurate identification to the species level difficult without sacrificing individuals for DNA or morphological analyses. Furthermore, most species have a small adult body size, which can render them inconspicuous in their natural habitat and generate uncertainty regarding the efficacy of both historical and contemporary surveys.

For example, monitoring for the assimineid Pecos assiminea (Assiminea pecos Taylor, 1987: Fig. 1a) has proven to be challenging due to difficulty in detecting. It is a freshwater gastropod that reaches an adult size of only 1–2 mm and is golden brown in color, making it easy to overlook in its natural environment (Fig. 1b). It is a semiaquatic snail inhabiting structured litter or saturated soils along the banks of aquifer-fed spring systems and other groundwater-reliant habitats with half the known range found within the spring systems of Bitter Lake National Wildlife Refuge (BLNWR) located near Roswell, New Mexico (NMDGF, 2005; Hershler et al., 2007). The species was listed as federally endangered in 2005 (USFWS, 2005), yet no population estimate exists for Pecos assiminea, in part because a technique to monitor abundance and distribution has only recently been developed. Preliminary quadrat surveys were conducted at BLNWR to gather baseline data; however, the efficacy of the quadrat survey method is an unresolved question. Monitoring efforts require multiple surveyors conducting searches of hundreds of quadrats. In spite of the huge amount of effort, surveyors rarely find more than 3–5 individuals in total over a 3-day monitoring effort. Thus, interpreting these results is difficult as the low numbers could indicate low population density or simply be poor detection due to the small size and cryptic coloration of Pecos assiminea. Furthermore, the influence of observer experience and habitat characteristics on detection probability has not been assessed. Therefore, the objective of this study was to empirically assess factors influencing the detection probability and accuracy of detection of Pecos assiminea to inform survey and monitoring efforts of species, which can improve not only monitoring for this species but other species alike, by obtaining taxa-specific and accurate accounts of detection.
Fig. 1

a, b Photographs of Pecos assiminea (Assiminea pecos) taken in the field at Bitter Lake National Wildlife Refuge near Roswell, New Mexico. The teasing needle is 1.2 mm wide; c. A quadrat during the detection probability experiment in November 2014 at Bitter Lake National Wildlife Refuge near Roswell, New Mexico, after it was searched by 12 observers

Methods

Study site and treatments

Bitter Lake National Wildlife Refuge is 9,929 hectares and is located 10 km east of Roswell, New Mexico, adjacent to the Pecos River. The karst topography of the refuge is rich with numerous springs, seeps, sinkhole lakes, and wetlands (Land & Huff, 2009). Surveys were conducted at a site in BLNWR classified as noncritical habitat where Pecos assiminea were unlikely to occur. The experiment site was chosen to prevent accidental detection of Pecos assiminea and to prevent disturbances on sensitive habitat during the experiment.

Empty phantom springsnail [(Pyrgulopsis texana (Pilsbry, 1935)] shells were used as a surrogate for Pecos assiminea due to their superficial similarities in appearance, size, and shape. Phantom springsnails are not known to occur at BLNWR, thus the risk of biased results due to the detection of live springsnails was not a concern in this study. Empty shells were painted to resemble the gold sheen of live Pecos assiminea shells and then sterilized prior to the experiment in an autoclave at 121°C and 15 psi for 15 min. Standardized survey protocols developed by U.S. Fish and Wildlife Service for monitoring Pecos assiminea were implemented, which consisted of 9 × 20 cm metal quadrats (n = 12) placed along spring banks in habitat that appeared suitable for Pecos assiminea based on previous surveys. Within each quadrat, painted springsnail shells were placed in microhabitats where Pecos assiminea is typically found, e.g., within litter, on bare ground, or adhered to the base of plant stalks, and secured using a small amount cyanoacrylate by a researcher not involved in the ensuing surveys of the quadrats. One of four snail density treatment levels was randomly assigned to each quadrat: none (n = 0 shells), low (n = 3 shells), medium (n = 6 shells), and high (n = 9 shells).

A total of ten observers participated in the experiment, which was conducted on November 12, 2014. Four self-identified their experience level as being beginners, three as intermediates, and three as experts. Those categorized as beginners had never participated in a survey for Pecos assiminea or other gastropod species, those identified as intermediates had assisted an observer in at least one Pecos assiminea survey, and experts had conducted at least one Pecos assiminea survey as the primary observer. All observers were allowed to inspect a vial containing the surrogate shells to develop a search image prior to starting the experiment. The observers were not informed as to the exact number of shells in each quadrat, only that quadrats could contain some or no shells. The starting position was randomized among observers. All ten observers searched through each of the same 12 quadrats for painted phantom springsnail shells for 5 min. Observers recorded the time they began observing each quadrat, quadrat number, and total number of shells found in each quadrat. After surveys were completed, habitat variables within each quadrat were recorded. Variables were tested and met the assumptions of normality and homogeneity.

Detection and accuracy

A mixed-effects logistic model was used to evaluate the influence of treatment level, litter depth and composition, slope, and observer skill level (fixed effects) on the probability of detection. The probability of detection was the response variable and coded 0 for no detection and 1 for detection. The sequence of searches within a quadrat was used as a repeated effect and grouped by quadrat. This was done to account for the potential effect on detection of any disturbance to the vegetation and litter within a quadrat occurring as the quadrats were searched. Individual observers were treated as a random effect. The mixed models were constructed and evaluated using the PROC GLIMMIX procedure in SAS 9.4 (SAS Institute Inc., Cary, NC, USA). A value of α = 0.05 was used to judge the significance of all statistical tests.

A mixed-effects-model analysis of covariance (ANCOVA) was used to test the null hypothesis that the accuracy of shell counts did not vary as a function of observer skill level (fixed effect), litter depth (covariate), or litter composition (covariate). Accuracy was calculated as the difference between the number of shells observed and the number expected divided by the expected number of shells for each observation period. Similar to the evaluation of detection probability described above, the sequence of searches within a quadrat was used as a repeated effect with quadrat treated as the subject effect and observer as a random effect. The mixed-model ANCOVA evaluating factors influencing accuracy was constructed and evaluated using the PROC GLIMMIX procedure in SAS 9.4 (SAS Institute Inc., Cary, NC, USA). A value of α = 0.05 was used to judge the significance of all statistical tests.

Results

Detection

The naïve detection probability produced by experimental results was 72.2%, i.e., 65 detection events out of the 90 events where detection was possible. However, habitat conditions and the number of shells present in a quadrat influenced the ability of observers to detect snail shells in the quadrats (Figs. 2, 3), resulting in a predicted mean (± SE) detection probability of 67.4 ± 3.0%. The probability of an observer detecting a snail was negatively correlated to litter depth, while the number of snail shells present within a quadrat was positively correlated to the probability of an observer detecting snails. Other factors, such as the percent cover of vegetation or litter within the quadrat, or the number of times a quadrat was searched prior to a given observation or observer experience level had little to no influence on the probability of detection (Table 1).
Fig. 2

Relationship between the probability of detecting Pecos assiminea surrogates in a quadrat and the litter depth and number of surrogate snail shells present in that quadrat based on 120 searches of twleve 9 × 20-cm quadrats placed at Bitter Lake National Wildlife Refuge near Roswell, New Mexico in November 2014. The contours represent the modeled detection probability relative to litter height and number of shells present. The pie charts represent the raw detection probabilities (black for detection)

Fig. 3

Influence of litter height on probability of detection (a) and accuracy of counts (b) of Pecos assiminea surrogates in a quadrat based on 119 searches of 12 experimental quadrats containing 0, 3, 6, or 9 shells

Table 1

Parameter estimates (β; ± standard error) and test of fixed effects of mixed-model logistic regression evaluating the influence of observer and environmental effects on the detection probability of Pecos assiminea in quadrats surveyed at Bitter Lake National Wildlife Refuge near Roswell, New Mexico on November 12, 2014

Fixed effect

β estimate ± SE

F value

Df

P value

Number of shells present

0.67 ± 0.14

9.88

1, 6

0.02*

Quadrat order

1.30

9, 96

0.24

1st search

0.61 ± 1.21

2nd search

−1.46 ± 1.17

3rd search

−1.43 ± 1.16

4th search

−0.73 ± 1.15

5th search

0.46 ± 1.24

6th search

−1.31 ± 1.16

7th search

−3.26 ± 1.25

8th search

−1.29 ± 1.19

9th search

−2.00 ± 1.11

10th search

0.00

Vegetation (%)

−0.09 ± 0.13

0.45

1, 6

0.53

Litter (%)

−0.05 ± 0.11

0.22

1, 6

0.66

Litter depth

−0.51 ± 0.16

10.80

1, 6

0.02*

Bank slope

−0.03 ± 0.04

0.66

1, 6

0.45

Observer experience

1.15

2, 22

0.34

Beginner

0.00

Intermediate

1.01 ± 0.67

Expert

0.32 ± 0.66

Quadrat order was treated as a repeated effect with quadrat as a subject effect. Individual observers were treated as a random effect. P values considered significant are indicated with asterisk

Accuracy

While the probability of an observer detecting snail shells when present was relatively high, the chances of the same observers finding and reporting the correct number of snail shells were not. Observers accurately indicated the lack of shells in the zero-shell treatments in 93% of the opportunities. Two different observers recorded false positives in different quadrats. The mean (± SE) observer accuracy with the zero-shell treatments excluded, was 25.5 ± 4.3%, i.e., observers tended to underestimate the number of snail shells present by a factor of about four with a 95% CI of 3.3–5.0. Furthermore, accuracy varied according to both environmental and observer factors (Table 2). The number of times a particular quadrat had been sampled prior to a given observation and litter depth were both negatively correlated with accuracy. Observer fatigue, as measured by the sequence of observations, did not have a strong association with accuracy. Increasing observer skill level tended to be associated with greater accuracy, but this effect was not as pronounced as the other fixed effects in the model (Table 2).
Table 2

Parameter estimates (β; ± standard error) and test of fixed effects of mixed-model analysis of covariance evaluating the influence of observer and environmental effects on the accuracy of counts made of Pecos assiminea surrogates placed in quadrats surveyed at Bitter Lake National Wildlife Refuge near Roswell, New Mexico on November 12, 2014

Fixed effect

β estimate ± SE

F value

Df

P value

Litter depth

−1.92 ± 1.05

7.67

1, 68

0.01*

Proportion Phragmites litter

0.20 ± 0.11

3.31

1, 68

0.07

Quadrat order

3.39

9, 68

<0.01*

1st search

25.84 ± 9.83

2nd search

4.35 ± 9.76

3rd search

3.62 ± 9.75

4th search

15.46 ± 9.66

5th search

12.32 ± 9.59

6th search

−6.87 ± 9.55

7th search

−14.79 ± 9.41

8th search

−2.55 ± 9.30

9th search

−4.97 ± 9.10

10th search

0.00

Number of shells present

−0.65 ± 1.10

0.35

1,68

0.56

Observer experience

1.19

2, 68

0.31

Beginner

0.00

Intermediate

3.35 ± 5.98

Expert

9.14 ± 11.47

Quadrat order was treated as a repeated effect with quadrat as a subject effect. Individual observers were treated as a random effect. P values considered significant were indicated with asterisk

Discussion

Detection plays a large role in the accuracy of population estimates of abundance, distribution, and metapopulation dynamics. Using statistical methods that incorporate detection probabilities, such as occupancy modeling (Royle & Link, 2006), can more efficiently deal with zero-inflated data, i.e., survey data from difficult to detect species (Miller et al., 2012). Large bodies of research report the qualitative or theoretical effects of detection probabilities through modeling, but surprisingly few studies have empirically determined the probability of detection, especially in the context of estimating observer accuracy (Alldredge et al., 2007). However, empirical studies are essential when a quantitative theory of how detection effects population estimates is discussed. The results indicate the probability of detecting Pecos assiminea may be relatively high, particularly in locations where there is a relatively high density of snails and shallow layer of vegetative litter. While standing vegetation did not seem to affect detectability, the amount of litter in a quadrat was negatively correlated to snail detection. The vegetative species did not significantly impact detection which may allow this research to be applicable in other systems or for other species. Interestingly, the results indicate that prior experience had little effect on the probability of an observer detecting snails when present, likely because training has been shown to be unable to remove the effects of varying observer abilities on detection of a species in a survey (Frederick et al., 2003; Elphick, 2008). With these results on how environmental characteristics and observer variables affect the detection of small-bodied semiaquatic gastropods advances the understanding of the problem of imperfect detection.

Besides validating the accuracy of count data and evaluating influencing factors, it is also important to appropriately and accurately interpret the results of a monitoring survey. Observer accuracy was negatively influenced by the number of times a quadrat was searched prior to the observation. This relationship was consistent across all treatment levels (snail shells) except for the zero-shell treatment. This decrease in accuracy may have been due to the increasing amount of disturbance in the quadrat over time which influenced detection, but was detrimental to accuracy. However, it may also have been due to observer fatigue. Searching for small snail shells cryptically colored to allow concealment within a structurally complex matrix of litter and vegetation requires a high degree of focus and attention. Observer fatigue has been identified as a detrimental factor in detection probability in other studies (Habib et al., 2012; Ransom, 2012; Rodtka et al., 2015), but has not been noted as a factor influencing the accuracy of count data. Due to confoundment, the relative importance of these two factors could not be assessed. On average, observers underestimated the number of shells by a factor of about four, which is not uncommon in wildlife surveys (MacKenzie et al., 2002; Moilanen, 2002; Frederick et al., 2003; Bailey et al., 2004; Gu & Swihart, 2004; Royle et al., 2005; Elphick, 2008). This information can now be used to apply a calibration (4 * n) to obtain a relative abundance estimate from past and future Pecos assiminea surveys. Therefore, during any given survey, if 20 snails were found by an observer with perfect accuracy, it is likely representing a true abundance of 60–100 snails.

The results from the present study indicate that personnel surveying for Pecos assiminea and other similar species of semiaquatic snails require only minimal training to be effective. Further, the results suggest that allowing sufficient time for personnel to take breaks when designing or planning a survey may improve accuracy; however, additional research is warranted to confirm this finding. There are several other caveats that should be considered when applying the results of the current study to the design of monitoring surveys or interpretation of survey results for Pecos assiminea or similar semiaquatic gastropod species. Although litter depth and snail abundance were the primary covariates associated with detection, some forms of heterogeneity may be accounted for in other covariate information, such as site characteristics or environmental conditions (MacKenzie et al., 2002), which were either not measured at the time of sampling in this study or not examined as part of the current study. For example, the current study was conducted over a few hours during a single day. Changing light levels associated with sun angle on a diel basis or changing characteristics of the vegetation in the quadrats occurring seasonally could represent factors influential to detection or accuracy. Furthermore, behavior could also represent an influential factor on detection and accuracy that was unaccounted for in the current study. However, snail behavior is unlikely to have a significant impact on detection or accuracy within a survey compared to what was observed in the current study. Movement and shifts in habitat use related to season or weather conditions, such as recent rain, could potentially influence detection and accuracy between survey events.

A good understanding of how often and under what circumstances miscounting or making estimation errors during ecological surveys for many species is lacking (Elphick, 2008), particularly for freshwater gastropods. An estimated 60–75% of freshwater gastropods are imperiled (Lysne et al., 2008; Strong et al., 2008; Johnson et al., 2013). Understanding detection probabilities and accuracy allows for greater efficacy in surveying and population estimates, which is critical to the process of species conservation (Martin et al., 2007). In most field surveys, especially involving challenging species (i.e., rare, patchy, or elusive), low detection probabilities cause individuals to frequently go undetected when they are present, resulting in high nondetection errors from the survey (Gu & Swihart, 2004). Recognizing how to appropriately allocate limited resources for conservation, especially for organisms with low detection probabilities, is a major issue in the recovery of imperiled species (Campbell et al., 2002). These findings provide empirically derived detection probabilities to be compared to modeled detection probabilities and inform managers on conservation. It also illustrates that some factors that were assumed to be important for successful surveys, such as experience and training, may not greatly influence the effectiveness of survey efforts. Identifying detection rates will aid in the development of a successful monitoring program and thus better guide management decisions impacting conservation efforts of Pecos assiminea and similar species.

Notes

Acknowledgements

Funding for this research was provided by the U.S. Geological Survey (Cooperative Agreement No. G13AC00051). The authors thank F. Anaya, L. Clark, A. Godar, B. Johnson, K. Leuenberger, K. Metzger, J. Sanchez, F. Truetken, and B. Wadlington for their participation in this experiment. This manuscript benefited from the comments and suggestions provided by S. Fritts and M. Barnes. Phantom springsnail shells were provided by C. Funkhouser. The authors also thank Cooperating agencies for the Texas Cooperative Fish and Wildlife Research Unit and University of Hawaii system, Hawaii Department of Land and Natural Resources, the U.S. Geological Survey, Texas Tech University, Texas Parks and Wildlife, the Wildlife Management Institute, and the U.S. Fish and Wildlife Service. The use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

References

  1. Alldredge, M. W., K. H. Pollock, T. R. Simons & S. A. Shriner, 2007. Multiple-species analysis of point count data: a more parsimonious modelling framework. Journal of Applied Ecology 44: 281–290.CrossRefGoogle Scholar
  2. Anderson, D. R., 2001. The need to get the basics right in wildlife field studies. Wildlife Society Bulletin 29: 1294–1297.Google Scholar
  3. Bailey, L. L., T. R. Simons & K. H. Pollock, 2004. Estimating site occupancy and species detection probability parameters for terrestrial salamanders. Ecological Applications 14: 692–702.CrossRefGoogle Scholar
  4. Bouchet, P. J. & J. J. Meeuwig, 2015. Drifting baited stereo-videography: a novel sampling tool for surveying pelagic wildlife in offshore marine reserves. Ecosphere 6: 137.CrossRefGoogle Scholar
  5. Campbell, S. P., J. A. Clark, L. H. Crampton, A. D. Guerry, L. T. Hatch, P. R. Hosseini, J. J. Lawler & R. J. O’Conner, 2002. An assessment of monitoring efforts in endangered species recovery plans. Ecological Applications 12: 674–681.CrossRefGoogle Scholar
  6. Elphick, C. S., 2008. How you count counts: the importance of methods research in applied ecology. Journal of Applied Ecology 45: 1313–1320.CrossRefGoogle Scholar
  7. Frederick, P. C., B. Hylton, J. A. Heath & M. Ruane, 2003. Accuracy and variation in estimates of birds by individual observers using an aerial survey simulator. Journal of Field Ornithology 74: 281–287.CrossRefGoogle Scholar
  8. Gu, W. & R. Swihart, 2004. Absent or undetected? Effects of non-detection of species occurrence on wildlife–habitat models. Biological Conservation 116: 195–203.CrossRefGoogle Scholar
  9. Habib, T. J., D. A. Moore & E. H. Merrill, 2012. Detection and stratification approaches for aerial surveys of deer in prairie–parklands. Wildlife Research 39: 593–602.CrossRefGoogle Scholar
  10. Hershler, R., H. P. Liu & B. K. Lang, 2007. Genetic and morphologic variation of Pecos assiminea, an endangered mollusk of the Rio Grande region, United States and Mexico (Caenogastropoda: Rissooidea: Assimineidae). Hydrobiologia 579: 317–335.CrossRefGoogle Scholar
  11. Johnson, P. D., A. E. Bogan, K. M. Brown, N. M. Burkhead, J. R. Cordeiro, J. T. Garner, P. D. Hartfield, D. A. W. Lepitzki, G. L. Mackie, E. Pip, T. A. Tarpley, J. S. Tiemann, N. V. Whelan & E. E. Strong, 2013. Conservation status of freshwater gastropods of Canada and the United States. Fisheries 38: 247–282.CrossRefGoogle Scholar
  12. Kellner, K. F. & R. K. Swihart, 2014. Accounting for imperfect detection in ecology: a quantitative review. PLoS ONE 9: 1–8.Google Scholar
  13. Land, L. & G. F. Huff, 2009. Multi-tracer investigation of groundwater residence time in a karstic aquifer: Bitter Lakes National Wildlife Refuge, New Mexico, USA. Hydrology Journal 18: 455–472.Google Scholar
  14. Lysne, S. J., K. E. Perez, K. M. Brown, R. L. Minton & J. D. Sides, 2008. A review of freshwater gastropod conservation: challenges and opportunities. Journal of the North American Benthological Society 27: 463–470.CrossRefGoogle Scholar
  15. MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, A. J. Royle & C. A. Langtimm, 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83: 2248–2255.CrossRefGoogle Scholar
  16. Martin, J., W. M. Kitchens & J. E. Hines, 2007. Importance of well-designed monitoring programs for the conservation of endangered species: case study of the snail kite. Conservation Biology 21: 472–481.CrossRefPubMedGoogle Scholar
  17. Martin, T. G., B. A. Wintle, J. R. Rhodes, P. M. Kuhnert, S. J. Low-Choy, A. J. Tyre & H. P. Possingham, 2005. Zero tolerance ecology: improving ecological inference by modelling the source of zero observations. Ecology Letters 8: 1235–1246.CrossRefPubMedGoogle Scholar
  18. Miller, D. A. W., L. A. Wier, B. T. McClintock, E. H. C. Grant, L. L. Bailey & T. R. Simons, 2012. Experimental investigation of false positive errors in auditory species occurrence surveys. Ecological Applications 22: 1665–1674.CrossRefPubMedGoogle Scholar
  19. Moilanen, A., 2002. Implications of empirical data quality to metapopulation model parameter estimation and application. OIKOS 96: 516–530.CrossRefGoogle Scholar
  20. NMDGF. 2005. Recovery and conservation plan for four invertebrates: Noel’s amphipod (Gammarus desperatus), Pecos assiminea (Assiminea pecos), Koster’s springsnail (Juturnia kosteri), and Roswell springsnail (Pyrgulopsis roswellensis). Prepared by Blue Earth Ecological Consultants Inc. and New Mexico Department of Game and Fish, Santa Fe.Google Scholar
  21. Pilsbry, H. A., 1935. Western and southwestern Amnicolidae and a new Humboldtiana. The Nautilus 48: 91–94.Google Scholar
  22. Ransom, J. I., 2012. Detection probability in aerial surveys of feral horses. The Journal of Wildlife Management 76: 299–307.CrossRefGoogle Scholar
  23. Rodtka, M. C., C. S. Judd, P. K. Aku & K. M. Fitzsimmons, 2015. Estimating occupancy and detection probability of juvenile bull trout using backpack electrofishing gear in a west-central Alberta watershed. Canadian Journal of Fisheries and Aquatic Sciences 72: 742–750.CrossRefGoogle Scholar
  24. Royle, J. A. & W. A. Link, 2006. Generalized site occupancy models allowing for false positive and false negative errors. Ecology 87: 835–841.CrossRefPubMedGoogle Scholar
  25. Royle, J. A., J. D. Nichols & M. Kery, 2005. Modelling occurrence and abundance of species when detection is imperfect. OIKOS 110: 353–359.CrossRefGoogle Scholar
  26. Strong, E. E., O. Gargominy, W. F. Ponder & P. Bouchet, 2008. Global diversity of gastropods (Gastropods; Mollusca) in freshwater. Hydrobiologia 595: 149–166.CrossRefGoogle Scholar
  27. Taylor, D. W., 1987. Fresh-Water Mollusks from New Mexico and vicinity, Bulletin, Vol. 116. New Mexico Bureau of Mines and Mineral Resources, New Mexico.Google Scholar
  28. USFWS. 2005. Endangered and threatened wildlife and plants: listing Roswell springsnail, Koster’s springsnail, Noel’s amphipod, and Pecos assiminea as endangered with critical habitat; final rule. Federal Register, DOI, U.S. Fish and Wildlife Service, 50 CFR Part 17, RIN 1018–AI15. 9 August 2005.Google Scholar

Copyright information

© Springer International Publishing AG (outside the USA)  2017

Authors and Affiliations

  1. 1.Department of Natural Resources ManagementTexas Tech UniversityLubbockUSA
  2. 2.U.S. Geological Survey, Texas Cooperative Fish & Wildlife Research UnitTexas Tech UniversityLubbockUSA
  3. 3.U.S. Geological Survey, Hawaii Cooperative Fishery Research UnitUniversity of Hawaii at HiloHiloUSA

Personalised recommendations