Factors affecting Southern Leopard Frog, Rana sphenocephala, distribution and detection in five coastal plain refuges.

William J. Barichivich, USGS Florida Integrated Science Center,
7920 N.W. 71^st Street, Gainesville, FL 32653 http://cars.er.usgs.gov/armi/

Presented at the 2005 Joint Meeting of Ichthyologists and Herpetologists, Tampa, FL, 6-11 July 2005.

Background

The southeastern region of the DOI Amphibian Research and Monitoring Initiative, SEARMI, encompasses Alabama, Florida, Georgia, North Carolina, South Carolina, and Tennessee.  Since October 2000, 7153 observations of 36 species of amphibians have been documented from five southeastern coastal plain refuges.  A single species, the Southern Leopard Frog (Rana sphenocephala), is represented by 10.5% of these observations.  Due to the large number of observations and the occurrence of this species at all our study sites, the ARMI metric, Proportion of Area Occupied (PAO) was used to test the importance of four environmental factors on the distribution and detection probability of R. sphenocephala.

Methods

• Study sites: Harris Neck, Lower Suwannee, Okefenokee, St. Marks and Savannah National Wildlife Refuges (Figure 1).
   
• Repeated site visits using a combination of crayfish traps, dip-nets, frogloggers, and visual and aural surveys.
   
• Single season model, program Presence (v. 2.0 <050526.1129>).
   
• Tested effects of invertebrates, fish, pH, and specific conductance (SpC) on site occupancy and detection probability of Southern Leopard Frogs.

Figure 1.  Locations of USFWS National Wildlife Refuges sampled within the southeast ARMI region. (click image to enlarge)

Results

• Naïve estimates of seasonal occupancy range from 0.1875 to 1.0 and estimated occupancy (Ψ) from 0.1875 to 1.0 (Table 1).
   
• Detection probabilities range from 0.3333 to 0.7778 (Table 1).
   
• Akaike Information Criterion (AIC) values do not suggest any one of the four covariates modeled explains variation in occupancy or detection probability better than any other.
   
• There were no trends in occupancy or detection probability within or between refuges or seasons.

Table 1. Naïve and estimated site occupancy (Ψ) and detection probabilities (p) for each USFWS National Wildlife Refuge sampled by SEARMI.  Lower Suwannee NWR not included due to lack of site revisits. Zero (0) indicates no observations and (–) indicates missing surveys.  (click image to enlarge)

Figure 2.  Effects of two biotic covariates on R. sphenocephala observations.  HN=Harris Neck, SV= Savannah, LS=Lower Suwannee, SM=St. Marks, RSPH=R. sphenocephala.  (click image to enlarge)

Figure 3.  Effects of two abiotic covariates on R. sphenocephala observations.  HN=Harris Neck, SV= Savannah, LS=Lower Suwannee, SM=St. Marks; (+)=R. sphenocephala present, (-)=R. sphenocephala absent. (click image to enlarge)

Table 2. Site revisit matrix based on detection probability (p) and proportion of area occupied (Ψ).  (click image to enlarge)

Discussion

• A multi-season model may be a better choice if you have the data.
   
• Revisits may not be possible between seasons, e.g. seasonal access and pond drying.
   
• The biotic (invertebrate presence, fish presence) and abiotic (pH, SpC) covariates modeled may be autocorrelated.
   
• The biotic covariates may be too coarsely defined (e.g. Gambusia ≠? Lepomis).
• No combination effects were tested (Figures 2 & 3).
   
• A minimum of three site visits is recommended by USGS biometricians, but in cases of high occupancy or low detection probability more visits may be required (Table 2). May not know if more visits are necessary until after surveys are conducted.