Prospects for Future Range Expansions The Red Imported Fire

Abundant evidence of northward temperature limitations on ant distributions led us to begin by examining relationships between January mean minimum temperatures and the occurrence of RIFA.

We selected winter temperature minima rather than other temperature-related variables such as growing season length because experimental evidence indicates that absolute temperature minima, not summer or active season minima, drive mortality and brood failure in RIFA. In addition, growing season length (frost-free period) incorporates a threshold of 0°C, while the red imported fire ant appears to encounter temperature limits only well below freezing (see earlier). We selected January mean minimum temperatures because annual lows occur more often in January than in any other month. Absolute annual temperature minima rather than minima at any particular time of the year appear most relevant to RIFA's biology.19 We also chose to limit our analysis to the portion of the United States east of the Rocky Mountains. This helped to avoid imputing climatic factors for the absence of RIFA in areas whose climates may be suitable (e.g., central California), but which likely remain uninvaded simply because the ant has not been introduced or had the opportunity to spread to them.

As expected, RIFA occurrence probabilities20 very closely reflected January mean minimum temperatures. Within the range of minimum January temperatures that occur in the United States, RIFA's occurrence probability peaks at the highest temperatures with a value of 1 (100% of counties infested) (Fig. 7.1). Below

Figure 7.1. The relationship between January mean minimum temperature and the probability of red imported fire ant occurrence in the eastern United States. A logistic curve fitted through the data yields an adjusted R2 value of 0.992. Model and parameter values are cited in the text.

+ 1.5°C, occurrence of RIFA declines rapidly to values near 0 (0% of counties infested) by -5.0°C.

We used data from -9 to +15°C to calculate a logistic model using ordinary least squares:

Occurrence

where C and B are constants and T is January mean minimum temperature.21 We selected a logistic model because it provides a simple fit to data falling entirely on one side of the peak of a bell-shaped curve. The range of January mean minimum temperatures encountered in the continental United States is confined to relatively cold ones. RIFA occurrence peaks at the warmest available winter temperatures in the United States. This suggests that the only portion of the theoretical bell-shaped relationship between temperature and occurrence encountered in the United States is to the left (low temperature side) of RIFA's optimum January minimum temperature (Gleason 1926, Saetersdal and Birks 1997). The scaling constant B reflects the sensitivity of occurrence to January minimum temperature and the direction in which occurrence changes with temperature. The constant C sets the level of occurrence associated with a January minimum mean temperature of 0°C. For January mean minimum temperature, the modeled constant values were C = -2.586 and B= -1.203. The model provided an extremely strong fit to the data, with a raw R2 value of 0.997 and a corrected R2 of 0.992. For comparison, a C constant-only logistic model yielded a raw R2 of 0.691 and a corrected R2 of zero.

We also examined the relationship between growing season length22 and the probability of RIFA occurrence to test the hypothesis that temperature constraints on foraging and development prevent RIFA colonies from surviving in areas with long freeze periods. Once again, RIFA occurrence peaked at 100% in analysis areas with long growing seasons, declining rapidly below a growing season length of 225 days and reaching 0% at a growing season length of 135 days (Fig. 7.2).We used data from 135 to 365 days to calculate a logistic model using ordinary least squares. For growing season length, modeled constant values were C = 18.111 and P = -.088. The model provided nearly as good a fit as that of January mean minimum temperatures, with a raw R2 value

Figure 7.2. The relationship between growing season length and the probability of red imported fire ant occurrence in the eastern United States. A logistic curve fitted through the data yields an adjusted R2 value of 0.972. Model and parameter values are cited in the text.

of 0.992 and a corrected R2 of 0.972. Although some have argued that moisture also limits RIFA distributions, we chose not to analyze the relationship between occurrence and precipitation because available evidence counters the assumption that RIFA is already nearing the western (dry) limit of its distribution (Pimm and Bartell 1980, CAST 1976, Francke et al. 1986, Vinson and Sorenson 1986, Francke 1988, Allen et al. 1993, Stoker et al. 1994, Killion and Grant 1995).

We used occurrence values predicted by minimum January temperature model to estimate the potential effects of warming alone on RIFA distributions in the eastern United States. We chose to use minimum January temperatures rather than growing season length because (1) stronger experimental evidence supports an absolute minimum temperature limit to RIFA survival, (2) January minimum temperature allows more direct application of the findings of global climate models (GCMs), which predict mean annual temperature changes directly, and (3) model fit of January minimum temperatures was similar to that of growing season length.23 We raised January minimum temperatures by 1,2, 3, and 4°C and applied model occurrence probabilities to each 1°C U.S. isotherm based on its new January temperature minimum. We then added infested counties to each region by hand to reflect differences in temperature minima within each 1°C region and large differences in precipitation. Within 1°C regions, warmer counties received priority over cooler ones; in the western extreme of the range, wetter counties received priority over drier ones. We

Table 7.4. Projected Impacts of Warming on the Red Imported Fire Ant's Distribution in the Eastern United States.

Current

At +1°C

At+2°C

At+3°C

At +4°C

Number of partially infested counties

25

86

73

96

102

Number of fully infested counties

696

813

981

1096

1268

Total number of infested counties

721

899

1054

1192

1370

Total area of infestation (x 1000 km2)

1261

1583

1846

2038

2271

also made an effort to include counties contiguous to invaded areas before including more distant counties, though long-distance transport by vehicles or even water make noncontiguous distributions possible. The number of counties infested by RIFA increases by approximately 150 with each additional 1°C of warming, and the total area infested by RIFA increases by nearly 200,000 to over 320,000 km2 with each additional 1°C of warming (Table 7.4). At 1°C of warming,24 states such as Virginia and Tennessee, which currently suffer little or no RIFA infestation, become susceptible to invasion over substantial areas. With 2°C of warming, additional advances into the moister parts of western Texas, Oklahoma, much of Virginia, and coastal areas of states as far north as Maryland occur. With 3°C of warming, almost all of Tennessee and Virginia become susceptible, as well as additional areas of Kentucky, Oklahoma, Texas, and the southernmost portions of Illinois and New Jersey. By 4°C of warming, our model suggests that RIFA will be able to spread solidly into an area nearly double that of its current extent, affecting nearly twice a many as the 721 counties that it currently plagues (Table 7.4). While these figures represent only a climate-based first approximation of the potential for RIFA spread under warming, they suggest that substantial increases in the extent of RIFA's effects will occur.

0 0

Post a comment