Printer Friendly

Use of probability of detection when conducting analyses of surveys of mesopredators: a case study from the Ozark Highlands of Missouri.

Management of mesopredators has become important because, in the absence of large carnivores, these smaller predators can become abundant and threaten migratory birds and small mammals (Crooks and Soule, 1999; Sinclair et al., 2005). Three common mesopredators, Virginia opossums (Didelphis virginiana), raccoons (Procyon lotor), and striped skunks (Mephitis mephitis), occur sympatrically throughout most of North America. Similar requirements for resources by these generalists have allowed them to become ubiquitous members of most natural and human-altered forested communities (Schwartz and Schwartz, 2001).

While studies of larger predators often aim to estimate true abundance (Kays et al., 2008), studies that examine ecology of mesopredators often use indices ofabundance for analysis: rate of capture (unique captures/100 trap nights; Disney et al., 2008) and rates of visitation to scent stations or relative abundance (Crooks and Soule, 1999; Dijack and Thompson, 2000; Sinclair et al., 2005). In several recent studies, researchers attempted to model relative abundance of raccoons, Virginia opossums, and striped skunks as functions of landscape and local habitats (Crooks and Soule, 1999; Dijack and Thompson, 2000; Sinclair et al., 2005; Disney et al., 2008) to predict risk of predation for forest-nesting birds. In their study in Missouri, Dijack and Thompson (2000) suggested that abundance of raccoons was related positively to latitude, density of streams, and mean size of patches on agricultural lands, whereas abundance of Virginia opossums was related positively to mean distance between patches of forest, latitude, density of streams, and related negatively to contagion. They determined that abundance of striped skunks did not relate to any characteristic of landscape examined. Other studies (crooks and Soule, 1999; Sinclair et al., 2005; Disney et al., 2008) have suggested that abundance of mesopredators increases as size of forested patches decreases.

Although variable detection may be accounted for to some degree in indices of abundance, the assumption is that probability of detection is constant and unaffected by habitat, surveying effort, or surveying method. Disney et al. (2008) attempted to compare indices of abundance between rates of visitation at scent stations and rates of capture for raccoons and Virginia opossums in a fragmented-forested landscape, but detected no association between the two methods. This suggests that indices of abundance, in general, are not appropriate for predicting true relationships of abundance or habitat. As an alternative to abundance or indices of abundance, MacKenzie et al. (2005, 2006) suggested using the state variable occupancy ([PSI]) when trying to elucidate relationships of habitat or distribution of species. This approach uses maximum-likelihood models to estimate occupancy by incorporating the additional parameter of probability of detection ([??]), which also can vary as a result of covariates of the model.

We suspect that surveying effort and covariates of habitat (i.e., heterogeneity of habitat and size of forested patch) will have effects on probability of detection and will bias indices of abundance if these are not accounted for prior to analysis. Our study aimed to examine how these variables affect probability of detection for three mesopredators in the Ozark Highlands of southern Missouri. In particular we attempted to model covariates that were revealed to be significant predictors of abundance from two studies in the central United States (Dijack and Thompson, 2000; Disney et al., 2008), as well as other predictors that we believed to be biologically significant.

Materials and Methods--Our study area included 10 counties in the Ozark Highlands of southern Missouri from Randolph county as the northern limit and the Arkansas border as the southern limit. Surveys were conducted on public lands managed by the Missouri Department of conservation, Missouri Department of Natural Resources, the United States Forest Service, and one private farm. Although trapping furbearers was legal and could be conducted on the lands in our study area, no reported trapping occurred at our sites prior to or during our surveys. We selected 14 sites with 5 sites at the northern edge of the Ozark Highlands and 9 sites in the southern portion of the region. The northern and southern sites were separated by >100 km and all sites were >4 km apart to ensure independence and allow analysis of landscape similar to Dijack and Thompson (2000). The use of 14 sites was similar to a study by O'Connell et al. (2006) in which they used 13 sites to estimate parameters of occupancy and detection for medium and large mammals.

Second-growth Quercus-Carya (oak-hickory) and mixed-hardwood forests dominate the ozark Highlands, which are interspersed with woodlands, savannas, prairies, and agricultural lands (Nigh and Schroeder, 2002). Our trapping sites primarily consisted of oak and mixed-hardwood forests; however, grasslands, croplands, and wildlife food plots also were common. Little was noted regarding composition of the understory, but most sites were characterized by nonnative shrubs, cool-season grasses, and forbs. Average temperature during our surveys was 8.4[degrees]C (range, 3.1-18.7[degrees]C).

During October 2008-April 2009, we set medium (106) and large (108) Tomahawk live traps (Tomahawk Live Trap Co., Tomahawk, Wisconsin) >100 m apart along established animal trails and natural funnels. Number of traps deployed at each site was 8-22 and was dependent on size of the area and availability of traps. This led to differences in trapping effort: 75 [+ or -] 8 SE trapnights/site (range, 28-137), but our analysis accounted for this variation, because we believed that inclement weather and low temperatures would affect detection negatively, we did not trap during precipitation events or when temperature was <0[degrees]C. We baited traps with mackerel or sardines and marshmallows and checked traps during 0700-1200 h. Initially, traps were set and run for 2-4 consecutive nights. After the last evening, traps were locked open, revisited after a 3-5-day rest period, and run again for an additional 2-4 evenings. Missing data from incomplete surveys at some sites (<8 full days) was accommodated by the occupancy-modeling approach (MacKenzie et al., 2006).

Once animals were captured, they were anesthetized with an intramuscular injection of Ketamine hydrochloride-Acepromazine (10 mg/kg and 1mg/kg, respectively). Standard measurements and gender were recorded for all animals that were handled. In some instances, animals were released without being handled or escaped before being handled. Each animal also received an individually numbered Monel ear tag (National Band and Tag Co., Newport, Kentucky). Following recovery, animals were released at the site of capture. All procedures were in accordance with guidelines of the American Society of Mammalogists (Gannon et al., 2007) and the University of Central Missouri Institutional Animal Care and Use Committee (permit 10-3209).

Using ArcGIS 9.3.1 (Environmental Systems Research Institute, Redland, California), we overlaid all trapping locations onto a digitized land-use-land-cover map. Following the analysis of landscape by Dijack and Thompson (2000), we created a 2-km-radius buffer at each site using a central point among the traps to measure covariates of landscape. We measured cumulative lengths of all roads and streams and used the Patch Analyst extension in ArcGIS 9.3.1 (R. Rempel, http://flash.lakeheadu.ca/~rrempel/patch/index.html) to measure total forested edge, mean size of forested patch, and total number of patches within each buffer. To test the hypothesis that latitude affects detection of mesopredators, we classified sites as northern and southern groups. We also included surveying effort per site (trapnights) as a covariate to be used in modeling. We standardized all continuous covariates to z-scores for analysis, but no other transformation was performed (Long et al., 2011). We did not include seasonal variation or covariates of temperature because our surveys were conducted over the course of one season. Additionally, because we did not trap in poor weather or low temperatures, we believe that any seasonal or temperature variation had little to no effect on our estimates.

Lack of recaptures of individual mesopredators precluded estimation of true abundance for our study sites, prior to analysis, we decided to group Virginia opossums and raccoons into one model due to their similar, generalist, ecological requirements, but maintained an additional covariate of species to model species-specific effects. We kept models of stripped skunks separate because we anticipated different effects of covariates based on previous studies (Dijack and Thompson, 2000; Disney et al., 2008). We then compiled all trapping records to create a binary history of detection (detected = 1, not detected = 0) for all three mesopredators. We developed 13 a priori models (Table 1) based on biologically plausible explanations of occurrence and detection and results of previous research (Dijack and Thompson, 2000; Disney et al., 2008). We included a null model with no effect of covariates and a global model that contained all eight possible covariates to ensure that there was no covariate interaction that lead to nonconvergence in the saturated model. For analysis, we used a single-season analysis in program PRESENCE 2.4 (J. E. Hines, http://www.mbr-pwrc.usgs.gov/software/presence.html). Because our naive estimates of occupancy were high, we believed raccoons and Virginia opossums occurred at all sites and we fixed the parameter for occupancy to 1.0 and only evaluated probabilities of detection as affected by covariates of surveying and habitat. This is similar to the first step of the two-step process employed by Yates and Muzika (2006) and Long et al. (2011) to first determine how variables affect probability of detection and then use those covariates as a constant set when deriving estimates of occupancy.

The best approximating models were selected based on the Akaike information Criterion corrected for small samples (AI[C.sub.c]) and Akaike weights ([w.sub.i]). We selected the 95% confidence set (summed [w.sub.i] = 0.95) of the supported models and removed the remaining models that were not contained in the confidence set to redistribute the Akaike weight among the top models. We then conducted model averaging (Burnham and Anderson, 2002) using spreadsheet software designed by B. Mitchell (www.uvm.edu/%7Ebmitchel/software.html) to estimate species-specific probabilities of detection and effects of covariates across multiple models.

RESULTS--We detected six species of mammals during the 1,052 trapnights from 73 sampling occasions. The most commonly detected mesopredator was the Virginia opossum (n = 83), followed by the raccoon (n = 47). Striped skunks were captured only on four occasions and were excluded from analyses. Oother incidental captures included the eastern woodrat (Neotoma floridana, n = 3), eastern fox squirrel (Sciurus niger, n = 2), and eastern cottontail (Sylvilagus floridanus, n = 1)

Of the 13 a priori models, only seven were contained in the 95% confidence set. A difference in probability of detection between species was supported by five of the seven models comprising the 95% confidence set (Table 1). Raccoons had a lower probability of detection than Virginia opossums, 0.46 [+ or -] 0.04 SE and 0.64 [+ or -] 0.05 SE, respectively. Mean size of forested patch appeared as a negative covariate in three of the top models (Tables 1 and 2), which agreed with the direction of our a priori hypothesis. Latitude, trapping effort, and total number of patches were all contained in the 95% confidence set of the model, but had little effect on parameters of detection with confidence intervals that strongly overlapped zero (Tables 1 and 2). Directions of effect of covariates, however, all agreed with our a priori hypotheses. Streams, roads, and total forested edge were not contained in the 95% confidence set and appeared to have no effect on probability of detection.

DISCUSSION--We determined that several covariates influenced ability to detect mesopredators in our surveys. All of these covariates have been suggested as drivers of abundance of mesopredators from other studies (Crooks and Soule, 1999; Dijack and Thompson, 2000; Sinclair et al., 2005; Disney et al., 2008), but these studies did not consider probability of detection in their analyses and we caution use of indices of abundance for drawing inferences about landscape. We suggest that modeling occupancy and the state variable of occupancy is a more appropriate approach to modeling relationships of habitat of mesopredators. This approach also is advantageous because multi-season and co-occurrence models of species have been developed and can be used to make strong inferences about factors that affect extinction and colonization of sites and dynamics of community interactions (MacKenzie et al., 2006).

Our surveys used varying numbers oftraps per site due to availability of traps (raccoons often damaged the medium-sized traps and rendered them unusable) and constraints of forested patches. Disney et al. (2008) also used varying numbers of traps (2-14) at different forested sites, leading to 80-560 trapnights/site. The occupancy-modeling approach allowed us to test how trapping effort affected our ability to detect presence of mesopredators. Although we believe that future surveying efforts should be standardized as much as possible, our results suggest that surveying effort (i.e., varying number of traps per site) explained only limited variation in our ability to detect mesopredators. We acknowledge that larger differences in trapnights per site may have more pronounced effects on probability of detection, but this can be modeled as a covariate of site to improve inferences in future surveys.

We were unable to estimate occurrence or probability of detection of striped skunks due to few detections. Neither the study by Dijack and Thompson (2000) nor that of Disney et al. (2008) was able to collect sufficient data for striped skunks to make inferences about their abundance. Our trapping results also demonstrated that striped skunks are detected infrequently and future studies should aim to improve surveying techniques for this species.

Dijack and Thompson (2000) suggested that abundance of raccoons and Virginia opossums in Missouri was influenced by latitude. They noted that agricultural lands in the northern portion of the state were dominated by croplands, whereas agricultural lands in the southern portion of the state were used more often for hay and pasture. Agricultural crops do provide an additional source of food for mesopredators, but there is a correlation between increasing agricultural lands and decreasing size of forested patches (Dijack and Thompson, 2000). Because we were unable to estimate true abundance from our surveys, we were unable to examine a numerical response for mesopredators to these alterations of landscape. However, interspecific and intraspecific interference competition often drive dynamics of communities and populations of carnivores (palomares and Caro, 1999) and, in this instance, dynamics of mesopredators. Smaller patches of forest surrounded by agricultural land should not exhibit a true increase in abundance of mesopredators due to increased interference competition and limited resources associated with the isolated patch.

Our results that probabilities of detection are higher at northern sites and that probability of detection is related negatively to size of forested patch corresponds with suggestions of Dijack and Thompson (2000) and Disney et al. (2008) because higher abundance might influence probability of detection. However, the exact relationship between abundance and probability of detection has not been examined. Furthermore, the use of scent stations in these studies (Dijack and Thompson, 2000; Disney et al., 2008) cannot distinguish between individual animals and the high rates of detection for this method are also a poor indicator of true abundance because a single individual can visit all stations at a single site in one evening (Disney et al., 2008). Our results suggest that activities of mesopredators are more restricted in small patches of forest within fragmented landscapes causing an increase in the probability of detection, which creates an artifact that abundance also is high.

We do not believe that any study has produced significant evidence that true abundance of mesopredators increases in smaller patches of forest. Disney et al. (2008) suggested that rates of visitation at scent stations were measures of activity by animals and that activity of raccoons and Virginia opossums were concentrated in smaller patches of forest, leading to increased risk of predation on forest-nesting birds. We believe that this hypothesis more accurately reflects the increased risk of predation on nests than does the suggestion that abundance of mesopredators is higher in smaller patches of forest.

Although our study was modest, we believe that the relationship between probability of detection and parameters in our analysis is evident and, with larger samples in future studies, the exact relationship and driving forces can be explored more robustly. Relative abundance and rates of capture are inaccurate indices of abundance, unless modeled as functions of sampling and covariates of habitat. We suggest that occupancy modeling is a more appropriate approach, but abundance of mesopredators also can be estimated directly by marking individuals and using capture-recapture analyses, or by using the repeated-count models of Royle and Nichols (2003). This is evidence that future surveys will benefit from inclusion of a parameter for probability of detection when objectives include estimating occupancy of sites (patches) or relationships of habitats for common mesopredators and exploring risk of predation on nests.

This research was funded by the Missouri Department of Conservation. We thank personnel of the Missouri Department of Conservation, Department of Natural Resources, and the United States Forest Service, as well as the Corson family, for allowing us access to their land for our surveys. We thank D. Fantz and anonymous reviewers for their help.

LITERATURE CITED

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

CROOKS, K. R., AND M. E. SOULE. 1999. Mesopredator release and avifaunal extinctions in a fragmented system. Nature 400:563-566.

DIJACK, W.D., AND F. R. THOMPSON, III. 2000. Landscape and edge effects on the distribution of mammalian predators in Missouri. Journal of Wildlife Management 64:20-216.

DISNEY, M. R., E. C. HELLGREN, C.A. DAVIS, D.M. LESLIE, JR., AND D. M. ENGLE. 2008. Relative abundance of mesopredators and size of oak patches in the cross-timbers ecoregion. Southwestern Naturalist 53:214-223.

GANNON, W. L., R. S. SIKES, and the Animal Care and Use Committee of the American Society of Mammalogists. 2007. Guidelines of the American Society of Mammalogists for the use of wild animals in research. Journal of Mammalogy 88:809-823.

KAYS, R. W., M. E. GOMPPER, AND J. C. RAY. 2008. The landscape ecology of coyotes based on large scale estimates of abundance. Ecological Applications 18:1014-1027.

LONG, R. A., T. M. DONOVAN, P. MACKAY, W.J. ZIELINSKI, AND J. S. BUZAS. 2011. Predicting carnivore occurrence with noninvasive surveys and occupancy modeling. Landscape Ecology 26:327-340.

MACKENZIE, D. I., J. D. NICHOLS, N. SUTTON, K. KAWANISHI, AND L. L. BAILEY. 2005. Improving inferences in population studies of rare species that are detected imperfectly. Ecology 86:11011113.

MACKENZIE, D. I., J. D. NICHOLS, J.A. ROYLE, K.H. POLLOCK, L.L. BAILEY, AND J. E. HINES. 2006. Occupancy estimation and modeling. Academic press, Burlington, Massachusetts.

NIGH, T. A., AND W. A. SCHROEDER. 2002. Atlas of Missouri ecoregions. Missouri Department of Conservation, Jefferson City.

O'CONNELL, A. F., JR., N. W. TALANCY,L. L.BAILEY,J.R.SAUER,R. COOK, AND A. T. GILBERT. 2006. Estimating site occupancy and detection probability parameters for meso- and large mammals in a coastal ecosystem. Journal of Wildlife Management 70:1625-1633.

PALOMARES, F., AND T. M. CARO. 1999. Interspecific killing among mammalian carnivores. American Naturalist 153:492-508.

ROYLE, J. A., AND J. D. NICHOLS. 2003. Estimating abundance from repeated presence/absence data or point counts. Ecology 84:777-790.

SCHWARTZ, C.W., AND E. R. SCHWARTZ. 2001. The wild mammals of Missouri. Second revised edition. University of Missouri Press, Columbia.

SINCLAIR, K. E., G. R. HESS, C.E. MOORMAN, AND J. H. MASON. 2005. Mammalian nest predators respond to greenway width, landscape context and habitat structure. Landscape and Urban Planning 71:277-293.

YATES, M. D., AND R. M. MUZIKA. 2006. Effect of forest structure and fragmentation on site occupancy of bat species in Missouri Ozark forests. Journal of Wildlife Management 70:1238-1248.

Submitted 5 April 2011. Accepted 10 May 2012. Associate Editor was Jennifer K. Frey.

Michael V. Cove, Liisa M. Niva, and Victoria L. Jackson *

Department of Biology and Earth Science, University of Central Missouri, Warrensburg, MO 64093

* Correspondent: vjackson@ucmo.edu
Table 1--Descriptions and expected direction of a priori models (except
the global model with all covariates) for detection of mesopredators
from surveys conducted in the Ozark Highlands of Southern Missouri,
October 2008-April 2009.

Hypothesis                            Model

No effects of habitat or survey on    p(.)
  detection
Increasing mean size of forested      p(mean size of forested patch)
  patch will affect detection
  negatively
Increasing trapping effort at sites   p(trapping effort)
  will affect detection positively
Increasing total number of patches    p(total number of patches)
  will affect detection positively
Species-specific detection and        p(species + latitude)
  positive influence from latitude
Species-specific detection and        p(species + road)
  negative influence from roads
Species-specific detection and        p(species + stream)
  positive influence from streams
Species-specific detection and        p(species + total edge of
  positive influence from             forest)
  increasing total edge of forest
Species-specific detection and        p(species + total number of
  positive influence from             patches)
  increasing total number of
  patches
Species-specific detection and        p(species + trapping effort)
  positive influence from
  increasing trapping effort
Species-specific detection and        p(species + mean size of
  negative influence from             forested patch)
  increasing size of forested
  patch
Species-specific detection,           p(species + mean size of
  negative influence from mean        forested patch + latitude +
  size of forested patch,             trapping effort)
  positive influence from
  increasing trapping effort

Hypothesis                            Structure of model

No effects of habitat or survey on    [[beta].sub.0]
  detection
Increasing mean size of forested      [[beta].sub.0] + [[beta].sub.1
  patch will affect detection         ](mean size of forested patch)
  negatively
Increasing trapping effort at sites   [[beta].sub.0] + [[beta].sub.1
  will affect detection positively    ](trapping effort)
Increasing total number of patches    [[beta].sub.0] + [[beta].sub.1
  will affect detection positively    ](total number of patches)
Species-specific detection and        [[beta].sub.0] + [[beta].sub.1
  positive influence from latitude    ](species) + [[beta].sub.2]
                                      (latitude)
Species-specific detection and        [[beta].sub.0] + [[beta].sub.1
  negative influence from roads       ](species) + [[beta].sub.2]
                                      (road)
Species-specific detection and        [[beta].sub.0] + [[beta].sub.1
  positive influence from streams     ](species) + [[beta].sub.2]
                                      (stream)
Species-specific detection and        [[beta].sub.0] + [[beta].sub.1
  positive influence from             ](species) + P2(total edge of
  increasing total edge of forest     forest)
Species-specific detection and        [[beta].sub.0] + [[beta].sub.1
  increasing positive influence       ](species) + [[beta].sub.2]
  from total number of patches        (total number of patches)
Species-specific detection and        ][[beta].sub.0] + [[beta].sub
  positive influence from             .1(species) + [[beta].sub.2]
  increasing trapping effort          (trapping effort)
Species-specific detection and        [[beta].sub.0] + [[beta].sub.1
  negative influence from             ](species) + [[beta].sub.2]
  increasing size of forested         (mean size of forested patch)
  patch
Species-specific detection,           [[beta].sub.0] + [[beta].sub.1
  negative influence from mean        ](species) + [[beta].sub.2]
  size of forested patch,             (mean size of forested patch)
  positive influence from             + [[beta].sub.3](latitude) +
  increasing trapping effort          [[beta].sub.4](trapping
                                      effort)

Hypothesis                            Expected result

No effects of habitat or survey on    --
  detection
Increasing mean size of forested      [[beta].sub.1] < 0
  patch will affect detection
  negatively
Increasing trapping effort at sites   [[beta].sub.1] > 0
  will affect detection positively
Increasing total number of patches    [[beta].sub.1] > 0
  will affect detection positively
Species-specific detection and        [[beta].sub.1] < 0,
  positive influence from latitude    [[beta].sub.2] > 0
Species-specific detection and        [[beta].sub.1] < 0,
  negative influence from roads       [[beta].sub.2] < 0
Species-specific detection and        [[beta].sub.1] < 0,
  positive influence from streams     [[beta].sub.2] > 0
Species-specific detection and        [[beta].sub.1] < 0,
  positive influence from             [[beta].sub.2] > 0
 increasing total edge of forest
Species-specific detection and        [[beta].sub.1] < 0,
  positive influence from             [[beta].sub.2] > 0
  increasing total number of
  patches
Species-specific detection and        [[beta].sub.1] < 0,
  positive influence from             [[beta].sub.2] > 0
  increasing trapping effort
Species-specific detection and        [[beta].sub.1] < 0,
  negative influence from             [[beta].sub.2] < 0
  increasing size of forested
  patch
Species-specific detection,           [[beta].sub.1] < 0,
  negative influence from mean        [[beta].sub.2] < 0,
  size of forested patch,             [[beta].sub.3] > 0,
  positive influence from             [[beta].sub.4] > 0
  increasing trapping effort

Table 2--Statistics for models of probability of detection derived from
surveys of mesopredators conducted in the Ozark Highlands of southern
Missouri. Included are models from the 95 % confidence set.

Model                           AI       [DELTA]AI    Akaike weight
                             [C.sub.c]   [C.sub.c]

[PSI](.),p(species + mean     120.48        0.00          0.351
  size of forested patch)
[PSI](.),p(mean size of       120.98        0.49          0.274
  forested patch)
[PSI](.),p(species +          121.18        0.70          0.247
  latitude)
[PSI](.),p(species +          123.88        3.40          0.064
  mean size of forested
  patch + latitude +
  trapping effort)
[PSI](.),p(species +          125.96        5.48          0.023
  trapping effort)
[PSI](.),p(species +          126.12        5.64          0.021
  total number of patches)
[PSI](.),p(trapping           126.27        5.79          0.019
  effort)

                               Number of
Model                          parameters      -2 log-
                                              likelihood

[PSI](.),p(species + mean          4            110.74
  size of forested patch)
[PSI](.),p(mean size of            3            113.98
  forested patch)
[PSI](.),p(species +               4            111.44
  latitude)
[PSI](.),p(species +               6            107.88
  mean size of forested
  patch + latitude +
  trapping effort)
[PSI](.),p(species +               4            116.22
  trapping effort)
[PSI](.),p(species +               4            116.38
  total number of patches)
[PSI](.),p(trapping                3            119.27
  effort)

TABLE 3--Model-averaged estimates of coefficients for covariates,
unconditional standard errors, and 95% CIs in models of
detection comprising the 95% confidence set from surveys of
mesopredators conducted in the Ozark Highlands of southern
Missouri.

Covariate           [beta] estimate      SE     Lower CI    Upper CI

Intercept                0.194         0.206     -0.210       0.598
Species                 -0.575         0.198     -0.963      -0.187
Mean size of
  forested patch        -0.548         0.164     -0.869      -0.227
Latitude                 0.431         0.266     -0.090       0.952
Trapping effort          0.001         0.001     -0.001       0.003
Total number of
  patches                0             0          0           0
COPYRIGHT 2012 Southwestern Association of Naturalists
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2012 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Cove, Michael V.; Niva, Liisa M.; Jackson, Victoria L.
Publication:Southwestern Naturalist
Article Type:Report
Geographic Code:1USA
Date:Sep 1, 2012
Words:4278
Previous Article:Spatial dynamics and growth of two native species of catostomids: are movements restricted?
Next Article:Temporal patterns of abundance of arthropods on sand dunes.
Topics:

Terms of use | Privacy policy | Copyright © 2022 Farlex, Inc. | Feedback | For webmasters |