Josh Van Buskirk. 2005. Local and landscape influence on amphibian occurrence and abundance. Ecology 86:1936–1947.

Appendix A. Methodological information and a map of the study area.

The study area included large parts of Kanton Zürich and eastern Kanton Thurgau, between Zürich, Switzerland, and the German border near Schaffhausen (Fig. A1). Within this region, ponds were chosen for study on the basis of their geographic locations, accessibility, availability of permits, and semi-natural condition. Ponds in this area were discrete water bodies; most were relatively small and shallow, but permanent (Fig. A2). Large lakes were specifically avoided. The average surface area was 1270 m² (geometric mean was 318 m²), average depth was 52 cm, and the average pond dried out in 20% of the years. Distances from the study ponds to the nearest neighboring pond (not necessarily included in the survey) ranged from 10 m to 1.4 km.

Estimates of density for amphibians, fish, and aquatic insects were obtained by pipe-sampling and quantitative dip-netting. For pipe-sampling, I rapidly dropped a hollow pipe through the water column, pressed it into the substrate, and removed the captured animals. The pipe (1 m long, 35 cm diameter, 0.096 m²) was thrown 20 times if the pond was large enough; more heterogeneous ponds received as many as 40 samples. The water and substrate captured within the interior of the pipe was repeatedly scooped with a net, and all animals were identified and measured. After I judged that all animals had been removed, I performed five additional scoops and moved to the next pipe-throw if nothing further was caught. I also spent 10–15 minutes dipnetting each pond in search of uncommon taxa that were not detected in the pipes; densities of such taxa were set to one-half the minimum possible pipe-sampling density.

The decision to make 20 samples per pond involved a compromise between accuracy and effort. Accuracy increased with the number of pipe throws, but after 20 samples the density estimates for all major taxa were within 20% of the asymptotic estimate (Fig. A3). In all, 82% of the 379 sampling occasions had 20 pipe throws, and 12% had 30 pipe throws. Ponds that received <20 pipe throws were too small to accommodate more samples.

I employed a second, more rapid survey method for a smaller number of ponds beginning in the year 2000. An area of 1.02 m² was sampled by sweeping a dip-net (0.51-m wide) a length of 2 m through the pond. This was repeated on average 34 times in each pond (range 4–73 sweeps), and all captured animals were identified and measured. Here again, the number of samples depended on the size and heterogeneity of the pond. Dipnetting underestimated the densities of all taxa. I used results from 114 pond records for which I had both pipe-samples and dipnet data to produce regression models that estimated "true" densities (judged from pipe-sampling) from the dipnetting data. I then used these models to estimate densities in ponds that were not pipe-sampled.

Not all ponds were surveyed in any single season or year. On average, I visited 46 ponds/year (range 32–64) during May samples and 35 ponds (24–45) during July samples. Of the 79 ponds included in the May samples, 54% were surveyed on 4 or more years (Fig. A1). The corresponding value for July samples was 45% of ponds. There were several reasons for adding or dropping a pond from the survey. Ponds could not be sampled when dry, and during dry years I included new ponds to compensate for those that were not sampled. Ponds were added to the study if they contributed animals to experiments. Some ponds surveyed during the first year were dropped as I expanded geographic coverage in the second year.

FIG. A1. Map of ponds included within the May survey period. Each circle represents a pond, and the diameter of the circle represents the number of years the pond was sampled (between 1 and 7 years). Most of the same ponds were included in the July survey.

FIG. A2. Frequency distributions of area and mean depth for the 83 ponds included within the survey.

FIG. A3. The accuracy of density estimates improved with the number of pipe samples. Black lines were calculated by randomly selecting increasing numbers of pipe throws from ponds with densities above the median, including records from the May survey during 2000–2003. Gray lines represent sampling occasions with densities in the lower 50% of all records. Accuracy was better in ponds with high density.