Don R. Church, Larissa L. Bailey, Henry M. Wilbur, William L. Kendall, and James E. Hines. 2007. Iteroparity in the variable environment of the salamander Ambystoma tigrinum. Ecology 88:891–903.

Appendix B. Historic precipitation data and methods for assigning frequencies to survival and breeding probabilities in our stochastic simulations.

FIG. B1. Historic precipitation data and methods for assigning frequencies to survival and breeding probabilities in our stochastic simulations.

The graph represents annual precipitation totals for periods between the 1st of September through the 31st of August (the biological year for Ambystoma tigrinum beginning at the onset of the breeding season) from 1948 to 2003 with our four study years circled. We were fortunate to conduct the field study over a period that encompassed very different annual precipitation patterns. The graph represents precipitation data from two different sources. Twenty-five years of data (1948–1973) were obtained from Stuarts Draft, VA (Stuarts Draft, VA weather data archived at http://www.ncdc.noaa.gov) located approximately 2 km from the study site. Data for the subsequent 26 years (1973–2002) were obtained from Staunton, Virginia (Staunaton Sewage Plant, Staunton, Virginia, USA weather data archived at http://www.ncdc.noaa.gov) situated approximately 20 km due west of our study populations. These two sites have 23 years (1950–1973) of overlapping precipitation data that are positively correlated (r = 0.822, P < 0.001). The high variance in precipitation over our four-year study is evident by comparing the statistics for annual precipitation totals during the study period, 93.97 ± 35.75 cm (mean ± SD), with those for the annual rainfall totals of the 51 years previous to our study, 95.60 ± 18.53 cm.

Although the average precipitation totals for these two periods are nearly the same, the variance during the study period (1060.81) was greater than that of the preceding 51 years (343.33) by more than three-fold. The 1999–2000 season precipitation total, 105.44 cm, was 9.84 cm (0.53 SD) above the 51 year average and the following season (2000–2001) fell to 86.54 cm, a deficit of 9.06 cm below average. Annual precipitation for the 2001–2002 season was extremely low at 47.85 cm which was 47.75 cm below average. The 2002–2003 season received 37.4 cm above average precipitation with a total of 133 cm. We tested for serial independence (Sokal and Rohlf 1995) and found no autocorrelation in the annual precipitation totals (n = 51, η = 0.069, P > 0.1). An independently and identically distributed (iid) selection regime of survival and breeding probabilities was therefore justified for our stochastic simulations.

In order to assign frequency probabilities to our survival and breeding probabilities in our stochastic simulations, we used the Standardized Precipitation Index (SPI), rather than precipitation totals, as a surrogate for determining how often the conditions we observed have occurred historically. SPI was chosen because it is a running, weighted mean of precipitation that better indicates the hydrological state of our study ponds. It was important to use a measure of precipitation that reflected the hydrological state of the ponds because our CR analysis suggested that there is a positive association between survival and water level within the ponds. SPI data reflects the average of a large number of weather stations in the region and is therefore less susceptible to biases that may result from using precipitation data from a single weather station that is not located exactly at our study ponds. Furthermore, the hydrology of our study ponds is almost certainly a function of precipitation in the mountains as well as on the valley floor at the base of these mountains where the ponds are situated. Consequently, a regional precipitation index is more appropriate for assessing how the hydrology of the ponds has varied historically. We used the six-month SPI for February between 1895 and 2002 for Climate Region 5 of Virginia that includes the central mountains and Shenandoah Valley (source: www.ncdc.noaa.gov/oa/ncdc.html). The six-month SPI for February was used because it encompasses almost the entirety of the breeding season for tiger salamanders at our study site. The six-month February SPI’s for our study years were distributed within ± 3 standard deviations from the norm. To calculate the frequencies with which the year-types we observed occurred historically, we assigned each historical year to an observed year that was within the same standard deviation bracket. The resulting frequencies were 0.6542 for 1999–2000, 0.1215 for 2000–2001, 0.0374 for 2001–2002, and 0.1869 for 2002–2003. These frequencies were used for females in all three study populations as well as for males from Oak and Deep Ponds’ populations. However, the relatively long running average of February’s six-month SPI was not a good surrogate for male salamanders at Pond Two because Pond Two is very ephemeral and males enter early in the breeding season. Consequently we needed to place greater weight on precipitation early in the breeding season for Pond Two males. We compared historical years to our observed year types in two-dimensional space with the three-month SPI for October and the four-month SPI for February as axes. The resulting frequencies for Pond Two males were 0.3178 for 1999–2000, 0.0654 for 2000–2001, 0.0374 for 2001–2002, and 0.5794 for 2002–2003.