Northeastern Naturalist B235 A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 Vol. 24, Special Issue 7 Distribution Data Support Warm Winter Temperatures as a Key Limit on the Range of a Goldenrod Gall Fly Host Race Amy V. Whipple1,2,*, Jason T. Irwin3, Paul L. Heinrich2, and Warren G. Abrahamson4 Abstract - Climate change has the potential to negatively impact organisms adapted to cold winters. Eurosta solidaginis (Goldenrod Gall Fly ) is a model organism for the study of cold tolerance in insects. The Goldenrod Gall Fly occurs as distinct host races on species of the goldenrod genus Solidago. The 2 well-studied host races exhibit high fidelity to their host-plant species on which they mate, lay their eggs, and overwinter as larvae. One host race, associated with Solidago altissima (Late Goldenrod) and referred to as the altissima host race, occurs virtually throughout its host-plant’s range across the eastern North America. Although S. gigantea (Giant Goldenrod) shares much the same range as S. altissima, the host race on S. gigantea referred to as the gigantea host race, is restricted to the northern portion of its host-plant’s range. We developed hypotheses to account for both the limited, northerly range of the gigantea host race as well as the widespread range of the altissima host race, which includes the southern US. Our first hypothesis is that the gigantea host race is limited by warm overwintering temperatures, which result in higher energetic losses over winter. This limitation would differentially affect the 2 host races given that altissima flies have greater mass and our data on egg production showed that smaller gigantea females produce fewer eggs than altissima females making it difficult to overcome such energetic losses. Our second hypothesis is that adult emergence times, which can be affected by spring temperatures during pupation, detrimentally alter the synchronization of host-plant and insect phenologies. Using presence–absence records and modeled proxies for winter energy loss and spring development time based on PRISM climate data, we conducted an auto-logistic regression on both proxies plus a spatial auto-covariate. This analysis supports the hypothesis that winter energy loss limits the geographic distribution of gall flies on S. gigantea. Climate and landscape changes, including cooling and deforestation during European settlement of northeastern US and southeastern Canada, were likely favorable for the S. gigantea host race. Ongoing climate and landscape change in the opposite direction will likely reduce the range and population size of the gigantea host race. Introduction With global climate change underway, it is increasingly important to understand the effects of warming on the winter ecology of organisms. Negative impacts of warming are often associated with factors such as summer heat stress. However, the importance of summer heat stress may be minimal in cooler, 1Department of Biological Sciences and 2Merriam-Powell Center for Environmental Research, Northern Arizona University, PO Box 5640, Flagstaff, AZ 86011. 3Department of Biological Sciences, Central Washington University, 400 E. University Way, Ellensburg, WA 98926. 4Department of Biology, Bucknell University, Lewisburg, PA 17837. *Corresponding author - amy.whipple@nau.edu. Manuscript Editor: Joshua Ness Winter Ecology: Insights from Biology and History 2017 Northeastern Naturalist 24(Special Issue 7):B235–B250 Northeastern Naturalist A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 B236 Vol. 24, Special Issue 7 northern latitudes where greater warming is expected to occur in the winter than summer (IPCC 2013). Winter warming is most often associated with putatively positive impacts, such as reduced frost damage, warmer temperatures during development, or a longer growing season (Altermatt 2010, Bale et al. 2002, Bradshaw et al. 2004, O’Connor et al. 2014). In particular, the northeastern US is expected to experience fewer frost days and longer growing seasons (Ahmed et al. 2013). On the other hand, negative impacts of winter warming could arise from disruption of organismal adaptations to continuous cold (Stuhldreher et al. 2014). Here we explore the role of temperature gradients at different times of year in setting a southern limit to an insect host-race’s range. Eurosta solidaginis Fitch (Goldenrod Gall Fly) is a model organism for the study of cold tolerance (Lee et al. 1995), insect ecology, and evolution (Abrahamson and Weis 1997, Abrahamson et al. 2001). The Goldenrod Gall Fly induces ball galls on the stems of host-plant species in the genus Solidago (Asteraceae; goldenrod). As temperatures increase in the spring, Goldenrod Gall Fly larvae pupate and emerge from their galls as non-feeding adults (Abrahamson and Weis 1997, Uhler 1951). During their short adult lifespan, the gall flies mate and females oviposit eggs into the stems of their goldenrod hosts. The larvae feed on tissue within the induced stem galls until they enter diapause in autumn. The larvae are freeze tolerant and are exposed during winter to wide temperature variation on the erect plant stems (Irwin and Lee 2003). In this study, we focused on gall flies attacking Solidago altissima L. (Late Goldenrod) and S. gigantea Ait. (Giant Goldenrod). We refer to these host races as “altissima flies” and “gigantea flies”, respectively. Solidago altissima and S. gigantea both have ranges that encompass all of the US east of the Mississippi River and beyond. Altissima flies are found throughout the range of S. altissima, but gigantea flies are restricted to the northern portion of their host-plant’s range from the east coast west into Minnesota at roughly the latitude of the Great Lakes (Abrahamson and Weis 1997). Thus, the range of gigantea flies is much more restricted than the range of their host plant or the range of altissima flies (Brown et al. 1995, 1996; Sumerford et al. 2000; Waring et al. 1990). The restricted range of gigantea gall flies raises the question of why its range is markedly reduced compared to altissima flies. Since southerly S. gigantea are susceptible to galling by gigantea flies in greenhouse tests (How et al. 1993), the restriction does not seem to be due to host-plant genetic variation. Hence, we examined the natural history of the Goldenrod Gall Fly to formulate biologically relevant hypotheses for climate-related range limits in the 2 host races. We posit 2 alternative hypotheses for the restriction of the gigantea host race to the northern US and southern Canada. First, energy losses due to warm winter temperatures limit the southern extent of gigantea flies. This hypothesis follows from findings that altissima fly larvae incur substantial mass losses when temperatures are above freezing during winter (Irwin and Lee 2000, Irwin et al. 2001). These freeze-tolerant larvae do best when they are not subjected to increased metabolic costs associated with thawing temperatures (Irwin and Lee 2003). Despite the metabolic costs of Northeastern Naturalist B237 A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 Vol. 24, Special Issue 7 warm winters, the altissima host race extends from northern regions to the Gulf of Mexico. Gigantea flies have lower average mass than altissima flies (Abrahamson and Weis 1997) and consequently have less energy reserves to compensate for increased metabolism during warm winter weather. Lower mass when coupled with energetic losses due to warm temperatures could translate into population declines if lower mass results in lower fecundity. Since the non-feeding adults have a full complement of eggs at emergence we were able to determine maximum potential fecundity for each host race from overlapping sites. We ascertained the egg loads of sympatric altissima and gigantea flies from the northeastern US to corroborate the potential for energetic costs having a larger influence on the ability of gigantea flies to sustain viable populations compared to altissima flies. Second, we hypothesize that warmer spring temperatures during pupation and adult development affect fly population persistence. This hypothesis is an obvious one to consider because spring emergence of flies is strongly influenced by temperature and is one of the major differences between the host races (How et al. 1993). Experiments with northeastern US populations of gall flies demonstrated that emergence time would have cascading influence on gall induction because the probability of gall induction on S. gigantea is influenced by host-plant age (Whipple et al. 2009). This influence could differentially affect the 2 host races for 2 reasons: (1) the host races have differing emergence times and (2) the host plants have different phenologies (Craig et al. 1993, Schmid et al. 1988, Whipple et al. 2009). Field-site Description Solidago altissima and S. gigantea are commonly found in old fields, road sides and disturbed areas (Abrahamson et al. 2005, Uhler 1951), but can also be found in less-disturbed meadow, prairie, and forest-edge habitats. For this study, the primary focal area was the northeastern US and upper Midwest where both host races are common (Appendix 1). Finding populations of Goldenrod Gall Flies is relatively easy because of the roadside habit of their host plants and the visibility of galls. Our site localities, presences, and absences are based on collections and field notes by the co-authors and a limited number of other former members of the Abrahamson Lab at Bucknell University. Our 120 localities include published gall locations as well as sites visited specifically for this study from the eastern and Midwestern US (latitude 40.320–46.763N; longitude 70.216–93.989W). Methods Egg counts of both host races We determined gall fly egg loads and egg lengths for female flies from VT and NH from 3 regions: near Lake Champlain, VT; in the foothills of mountains of northern VT; and in southern NH. We collected galls from both goldenrod species in the late fall and early winter of 2000 and stored them in freezers at -20 °C. In the spring of 2001, larvae were dissected from galls and reared in growth chambers at Northeastern Naturalist A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 B238 Vol. 24, Special Issue 7 24 °C, with 14 hours of light, and 80% relative humidity. We weighed 10 adult altissima females and 26 adult gigantea females upon emergence and dissected those specimens to obtain egg counts and the lengths of 10 individual eggs per female fly. Because adult flies do not feed and live only days, these counts represent their maximum potential reproductive output (Irwin and Lee 2000, Uhler 1951). Species presence and absence data We used records of Goldenrod Gall Fly presence and absence from our lab’s previously published research (Brown et al. 1995, 1996; Sumerford et al. 2000; Waring et al. 1990) plus additional records reported here (Appendix 1). All presences and absences were recorded by investigators who were studying both gall fly host races. Thus, detection of the galls of both host races when present would be very high, with only extremely low-density occurrences potentially being missed. Absences used in this study are sites where altissima galls and S. gigantea hosts were both found, but gigantea galls were not found. We restricted absences in this manner because our primary interest was testing for climatological constraints on gigantea versus altissima flies, not identifying factors related to host-plant range or factors that limit both host races. Our study includes records for 120 sites across the US range of gigantea flies as well as to the south of this range (Fig. 1). Additional occurrences reported for this study (rather than from literature) come from gallcollecting trips and presence/absence surveys during 1998–2003. The goal of these trips was the collection of galls for experimental studies on gall fly host-use (e.g., Whipple et al. 2009) or surveying localities for this project. Climatological data We used PRISM 2-km–scale, 1971–2000, “30-year normals”-modeled climate data for the continental US provided by the PRISM Climate Group (2016). These modeled data are based on 1971–2000 temperature observations which are then interpolated across the US using topography to estimate local climate parameters on a 2-km grid. Since most of our presence and absence data for gall flies was collected during the period 1990–2003, this climate data set is representative of the recent climate history experienced by gall-fly populations prior to and during our collections. We entered presence and absence records for gigantea flies in a GIS database and used these locations to extract the monthly mean, minimum, and maximum temperatures from the PRISM data for each location. These values were used to calculate the following 2 indices of the effects of temperature on biological rates: warming degree days related to emergence time based on April–May temperatures, and estimated energy consumed during overwintering based on Nov–Feb temperatures. The inclusion of New England and Midwestern localities helped to reduce the correlation between temperatures in these 2 distinct seasons (Fig. 1). Calculation of a degree-day estimate related to emergence time The single-triangulation method (Lindsey and Newman 1956, Roltsch et al. 1999, UC Statewide Integrated Pest Management Project 2002) uses daily Northeastern Naturalist B239 A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 Vol. 24, Special Issue 7 maximum temperature, daily minimum temperature, and a threshold temperature to estimate the number of degree-hours per day above that threshold. We modified the method to use the 30-year normal average monthly minimum (Tmin), average monthly maximum (Tmax), and days per month (Dmonth) to calculate the degree days per month above the threshold temperature (Tthreshold) and applied this formula to create a degree-day estimate for each locality as follows: Degree days above threshold per month = Dmonth([(12[Tmax – Tthreshold]2 / (Tmax - Tmin)]/24) We used a threshold of 10 °C and a simple sum of degree days over April and May (when flies pupate) to give an index of relative emergence times. Evidence from the literature suggests a linear response of pupal development time to temperature in other insect species, or a logistic response that is essentially linear over a wide range of temperatures (e.g., Duyck and Quilici 2002). Figure 1. Presence (white circles) and absence (black triangles) of gigantea flies mapped with: (A) estimated April and May degree days above 10 °C and (B) estimated energy consumption (grams carbohydrate and lipid) from November through February. Midwestern gigantea gall sites are a closer match for Northeastern gigantea gall sites in the winter temperature-driven energy consumption metric compared to the spring temperature-driven degree-days that relate to development time. Northeastern Naturalist A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 B240 Vol. 24, Special Issue 7 Calculation of energy consumed during overwintering To calculate the amount of lipid and carbohydrate consumed during overwintering, we measured the metabolic rate of diapausing gigantea fly larvae (following the methods of Irwin and Lee 2003) collected in the area of Canton, NY (44°37.46'N, 75°14.18'W), at 0, 5, 10, 15, and 20 ºC, for which the relationship between metabolic rate (expressed as μl CO2·g-1·h-1) and temperature was: Metabolic rate = 0.0255 x 1.0894Temp This equation closely describes the relationship with an r2 of >0.99. Based on a respiratory exchange ratio of 0.924 (Irwin and Lee, 2003), we then converted metabolic rate to grams of mass consumed (from both stored carbohydrates and lipids) per month: Mass of lipids and carbohydrates consumed per month = Dmonth(0.6305(0.0255 x 1.0894Monthly Mean Temp), where: Dmonth = days per month. Details justifying the use of this equation for the conversion are available in Irwin and Lee (2003). Using ArcGIS 10.5 (ESRI 2016) and 1971–2000 temperature-based modeled monthly data provided by the PRISM Climate Group (2016), we extracted the monthly means, minimums, and maximums associated with each gall-survey location. These values were used to calculate a summed November–February monthly energy-consumption index for each location, which represents grams of carbohydrate plus lipid. Spatial analysis of gall fly presence and covariates We tested our hypotheses via spatial auto-logistic regression to examine the association between gigantea fly distribution and indices of winter energy loss and spring development time. We analyzed the relationship between presence of gigantea galls and our 2 indices for the 120 localities listed in Appendix 1 using R (R Core Team 2016). There has been considerable debate on the best ways to handle the analysis of spatial data (Bardos et al. 2015, Beale et al. 2010, Dormann et al. 2007, Hawkins 2012, Legendre 1993). The consensus is that care needs to be taken in how spatial autocorrelation is dealt with since it can cause problems with the interpretation of covariates. One way to incorporate unaccounted for spatial covariance in binary presence/absence data is with auto-logistic regression, and Bardos et al. (2015) have recently clarified the settings necessary for valid use of this technique. An auto-covariate effect could arise from spatial process such as dispersal or from additional environmental variables such as gradients in soil properties. In the case of our analysis, we are not explicitly interested in the form or amount of autocorrelation so much as wanting to protect against it having a large influence on the estimates of coefficients for the covariates of interest. We followed Bardos et al. (2015) as implemented in the R package ‘spdep’ (Bivand et al. 2006). We calculated, and included in the analysis, a spatial auto-covariate (settings: second-order neighborhood with weights based on neighbor summation and independent of distance). We also included the covariates of interest: spring growing-degree-days estimate Northeastern Naturalist B241 A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 Vol. 24, Special Issue 7 and energy-consumption estimate. We ran models with varied combinations of predictors (1) as a check for high levels of collinearity that could cause coefficient estimates to change as factors change and (2) to be able to use insight from AIC values for models of varied complexity. Mapping of gigantea fly occurrence in relation to indices We created raster maps of the indices using ArcGIS 10.5 map algebra (ESRI 2016) and 1971–2000 temperature-based modeled monthly means, maximums, and minimums provided by the PRISM Climate Group (2016) to calculate the spring degree-day and energy-consumption estimates using the methods described above. We then mapped (Fig. 1) the indices and gigantea gall presence/absence data (ESRI 2016, PRISM 2016). Results Fewer eggs for gigantea flies Gigantea flies had eggs of the same lengths (mean = 575 μm, std. dev. = 22.8) as altissima flies (mean = 574 μm, std. dev. = 15.4). However, gigantea flies had fewer eggs and lower potential fecundity (mean = 121, std. dev. = 37.1) than altissima flies (mean = 165, std. dev. = 26.8). The linear model results for egg length and counts with species and site nested within species as factors are provided in Table 1. Gigantea fly distribution related to winter temperatures We ran the auto-logistic regression model with both the winter energy-consumption and spring growing-degree-days covariates and also with each covariate separately, for 2 reasons: first, because collinearity of climate-related covariates can cause dramatic changes in the results, and second, because we could compare Table 1. Linear-model results for egg length and egg count with host race and source site nested within host race as factors. Egg length is the same across host races, but egg counts differ between host races. Estimate SE t value Pr(>|t|) Egg length Intercept 572.66 14.11 40.58 less than 2.00e-16 *** Host race 2.35 8.36 0.28 0.781 Host race:site -0.65 1.48 -0.44 0.664 Residual standard error: 21.31 on 33 degrees of freedom; Multiple R-squared: 0.006452; Adjusted R-squared: -0.05376. F-statistic: 0.1071 on 2 and 33 DF; P-value: 0.8987. Egg count Intercept 208.49 22.87 9.11 1.56e-10 *** Host race -48.378 13.55 -3.57 0.00112 ** Host race:site 2.62 2.39 1.10 0.281 Residual standard error: 34.54 on 33 degrees of freedom; Multiple R-squared: 0.2786; Adjusted Rsquared: 0.2349. F-statistic: 6.374 on 2 and 33 DF; P-value: 0.00456. Northeastern Naturalist A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 B242 Vol. 24, Special Issue 7 model AIC statistics to determine which variables contribute substantially to improved model fit. The coefficients for the covariates did vary some among the models, but not to a degree that affects interpretation, indicating that we had enough independence of the temperature indices in different seasons to make meaningful comparisons. Running the model in 3 forms also allows comparison based on AIC values (Table 2). We found an AIC of 73.14 with the winter energy-consumption index, but adding spring degree days did not improve the model fit (AIC = 73.70). The model with spring degree days alone had an AIC of 109.51, and the null and residual deviance for the spring degree-day analyses indicate a lack of fit (Table 2). This finding corroborates statistical analysis based on coefficients and significance testing where the coefficient (standard error; P-value) was -18.05 (4.28; 0.0000244) for winter energy consumption and 0.01 (0.01; 0.248) for spring degree days. It also supports the finding that estimated winter energy consumption is strongly related to gigantea gall presence whereas the spring degree-day estimate is not. These relationships are illustrated using a color ramp of winter energy-loss estimates and Table 2. Gigantea gall presence and absence was modeled with an auto-logistic regression. The autocovariate is constructed to take into account spatial autocorrelation using the ‘spdep’ package (Bivand et al. 2006) in R with settings for valid auto-covariate construction as in Bardos et al. (2015). Here we show outputs for 3 general linear models: each of the 2 covariates separately and both covariates together. The covariates of interest are the estimated energy consumption due to temperatures above freezing in November to February and the estimated April and May degree-days (above 10 °C) as an index of pupation and emergence time. Coefficients Estimate SE z value Pr(>|z|) Model with both covariates Intercept 28.81 6.65 4.33 1.48e-05 *** Energy Nov:Feb -18.05 4.28 -4.22 2.44e-05 *** Growing DD Spring 0.01 0.01 1.15 0.248 Autocovariate 0.09 0.03 2.84 0.00446 ** Null deviance: 130.385 on 119 degrees of freedom; Residual deviance: 65.704 on 116 degrees of freedom AIC: 73.70 Model with energy consumption covariate Intercept 26.42 6.06 4.36 1.28e-05 *** Energy Nov:Feb -15.75 3.58 -4.40 1.06e-05 *** Autocovariate 0.09 0.03 2.79 0.00532 ** Null deviance: 130.385 on 119 degrees of freedom; Residual deviance: 67.144 on 117 degrees of freedom AIC: 73.14 Model with growing-degree-days covariate Intercept 1.51 0.85 1.78 0.0752 Growing DD Spring -0.01 0.005 -1.86 0.0625 Autocovariate 0.08 0.02 3.25 0.00116 ** Null deviance: 130.39 on 119 degrees of freedom; Residual deviance: 103.51 on 117 degrees of freedom AIC: 109.51 Northeastern Naturalist B243 A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 Vol. 24, Special Issue 7 spring developmental degree-day estimates along with our presence and absence data for gigantea flies (Fig. 1). Figure 1 shows the better match between gigantea fly locations in the east and west for energy stores based on winter temperatures (Fig. 1B) than for warming degree days in the spring (Fig. 1A). Discussion We tested 2 alternative hypotheses intended to explain the northerly distribution of the gigantea host race of the Goldenrod Gall Fly relative to the more widespread range of the altissima host race. We found substantial support for our first hypothesis, which addresses the potential negative impacts of warm winters. Irwin and Lee (2000) have previously noted that warmer winter temperatures could be detrimental to freeze-tolerant insects, such as the Goldenrod Gall Fly because it pupates in the spring and does not feed as an adult. Insects with this life history are particularly vulnerable to overwintering energy losses because they have no further opportunity to compensate for winter energy losses via adult feeding. Because altissima flies are much larger (Abrahamson and Weis 1997) than gigantea flies, they should have more resources and thus capacity to maintain positive population growth in warmer southern climates where altissima flies occur extensively in the complete absence of gigantea flies. This suggestion is supported by the losses of fecundity evident in altissima flies that experience warm overwintering temperatures (Irwin and Lee 2000) and by our demonstration of lower potential fecundity in gigantea flies compared to altissima flies under the same climatic conditions. Our results reinforce the hypothesis that warm overwintering temperatures may be an important constraint on the distribution of freeze-tolerant insects generally (Irwin and Lee 2000, Stuhldreher et al. 2014). The association between gigantea fly presence and steadier freezing temperature conditions, which leads to low stored-energy loss, is strong evidence in support of the first hypothesis. This strong relationship was not obscured by temperature conditions during spring or spatial auto-covariance, which represents other unstudied factors such as migration or additional environmental variation. The hypothesis of southern range limitations due to temperatures during spring pupation and adult development was not supported by our findings. There was no detectible relationship between spring growing-degrees-days that would drive emergence time and the southern range boundaries of gigantea flies. This hypothesis was based on a series of experiments demonstrating emergence-time differences between the host races (Craig et al. 1993), variation in gall induction rates through time and by goldenrod species (How et al. 1993, Horner et al. 1999, Whipple et al. 2009), and the effect of temperature on fly development time. All these factors suggest spring is a critical period for Goldenrod Gall Flies and their coordination with their host plant. A similar hypothesis, proposed by MacLean (1983) and Bale et al. (2002), suggests that at southern boundaries plants may develop too early for their insect herbivores. Gigantea flies emerge earlier when they develop at the same temperature as altissima flies (Craig et al. 1993), but S. gigantea is suitable for galling during an earlier Northeastern Naturalist A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 B244 Vol. 24, Special Issue 7 phenological window than S. altissima. Thus to some extent, warmer temperatures in spring could benefit rather than negatively impact gigantea flies. Cold spring temperatures are more likely to place a northern limit on gigantea fly distributions. Such a limit, if it exists, might be relaxed by a warming climate. Climate and land-use change at the time of European colonization of North America (Bonan 1999) may have promoted the expansion of the gigantea host race across the northeastern US due to both a cooling climate and land clearing for agriculture. Genetic data indicate that the gigantea flies are derived from the ancestral altissima host race and that the host shift may have occurred in the northeastern US and then spread to the Midwest (Brown et al. 1996). It is likely that this host shift occurred relatively recently (Stireman et al. 2005). Investigations of climate limitations on the ranges of the host races and host plants, combined with genetic studies of divergence in this group, have implications for speciation research in this system. For instance, such research will help determine the likely degree of sympatry of suitable hosts at the time of divergence. Our data support the contention that at high latitudes future climate-change impacts on insects are more likely to be influenced by winter temperatures and growing-season alteration (Bradshaw et al. 2004). Ongoing climate change, which is predicted in the northeastern US to involve warmer winters, fewer frost days, and longer growing seasons (Ahmed et al. 2013, IPCC 2007) has critically important implications for all gall flies given their sensitivity to warm winters. Gigantea flies in particular are vulnerable to this warming because their distribution range is already much narrower than that of altissima flies and they have less mass and fecundity to buffer against overwintering energy losses. The degree to which flies may be able to adapt to these warmer winters is likely limited by energetic constraints. Acknowledgments This work was supported by a National Science Foundation grants DEB-9981330 to W.G. Abrahamson and A.V. Whipple and DEB-0343633 to W.G. Abrahamson and J.T. Irwin, and by the David Burpee Chair of Bucknell University. We thank C. Abrahamson, L. Young, and S. Morehead for field support and an anonymous reviewer for comments that improved the manuscript. Any opinions, findings, and conclusions expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. Literature Cited Abrahamson, W.G., and A.E. Weis. 1997. Evolutionary Ecology across Three Trophic Levels: Goldenrods, Gallmakers, and Natural Enemies. Monographs in Population Biology 29. Princeton University Press. Princeton, NJ. Abrahamson, W.G., M.D. Eubanks, C.P. Blair, and A.V. Whipple. 2001. Gall flies, inquilines, and goldenrods: A model for host-race formation and sympatric speciation. American Zoologist 41:928–938. Abrahamson, W.G., K. Ball Dobley, H.R. Houseknecht, and C.A. Pecone. 2005. Ecological divergence among five co-occurring species of old-field goldenrods. Plant Ecology 177:43–56. Northeastern Naturalist B245 A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 Vol. 24, Special Issue 7 Ahmed, K.F., G. Wang, J. Silander, A.M. Wilson, J.M. Allen, R. Horton, and R. Anyah. 2013. Statistical downscaling and bias correction of climate-model outputs for climate- change impact assessment in the US Northeast. Global and Planetary Change 100:320–332. Altermatt, F. 2010. Climatic warming increases voltinism in European butterflies and moths. Proceedings of the Royal Society 277:1281–1287. Bale, J.S., G.L. Masters, J.D. Hodkinson, C. Awmak, T.M. Bezemer, V.K. Brown, J. Butterfield, A. Buse, J.C. Coulson, J. Farrar, J.E.G. Good, R. Harrington, S. Hartley, T.H. Jones, R.L. Lindroth, M.C. Press, I. Symrnioudis, A. Watt, and J.B. Whittaker. 2002. Herbivory in global climate change research: Direct effects of rising temperature on insect herbivores. Global Change Biology 8:1–16. Bardos, D.C., G. Guillera-Arroita, and B.A. Wintle. 2015. Valid auto-models for spatially autocorrelated occupancy and abundance data. Methods in Ecology and Evolution 6:1137–1149. Beale, C.M., J.J. Lennon, J.M. Yearsley, M.J. Brewer, and D.A. Elston. 2010. Regression analysis of spatial data. Ecology Letters 13:246–264. Bivand, R., L. Anselin, O. Berke, A. Bernat, M. Carvalho, Y. Chun, C. Dormann, S. Dray, R. Halbersma, N. Lewin-Koh, H. Ono, P. Peres-Neto, M. Tiefelsdorf, and D. Yu. 2006. The spdep package: spatial dependence, weighting schemes, statistics, and models. Available online at htSp://www.R-project.org/. Accessed 6 July 2017. Bonan, G.B. 1999. Frost followed the plow: Impacts of deforestation on the climate of the United States. Ecological Applications 9:1305–1315. Bradshaw, W.E., P.A. Zani, and C.M. Holzapfel. 2004. Adaptation to temperate climates. Evolution 58:1748–1762. Brown, J.M., W.G. Abrahamson, R.A. Packer, and P.A. Way. 1995. The role of naturalenemy escape in a gallmaker host-plant shift. Oecologia 104:52–60. Brown, J.M., W.G. Abrahamson, and P.A. Way. 1996. Mitochrondial DNA phylogeography of host races of the Goldenrod Ball Gallmaker, Eurosta solidaginis (Diptera: Tephritidae). Evolution 50:777–786. Craig, T.P., J.K. Itami, W.G. Abrahamson, and J.D. Horner. 1993. Behavioral evidence for host-race formation in Eurosta solidaginis. Evolution 47:1696–1710. Dormann, C.F., J.M. McPherson, M.B. Araújo, R. Bivand, J. Bolliger, G. Carl, R.G. Davies, A. Hirzel, W. Jetz, W.D. Kissling, I. Kühn, R. Ohlemüller, P.R. Peres-Neto, B. Reineking, B. Schröder, F.M. Schurr, and R. Wilson. 2007. Methods to account for spatial autocorrelation in the analysis of species distributional data: A review. Ecography 30:609–628. Duyuk, P.E., and S. Quilici. 2002. Survival and development of different life stages of three Ceratitis spp. (Diptera: Tephritidae) reared at five constant temperatures. Bulletin of Entomological Research 92:461–469. ESRI, Inc. 2016. ArcGIS 10 [Computer software]. Redlands, CA. Available online at http:// www.esri.com. Accessed 13 February 2016. Hawkins, B.A. 2012. Eight (and a half) deadly sins of spatial analysis. Journal of Biogeography 39:1–9. Horner, J.D., T.P. Craig, and J.K. Itami. 1999. The influence of oviposition phenology on survival in host races of Eurosta solidaginis. Entomologia Experimentalis et Applicata 93:121–129. How, S.T., W.G. Abrahamson, and T.P. Craig. 1993. Role of host plant phenology in host use by Eurosta solidaginis (Diptera: Tephritidae) on Solidago (Compositae). Environmental Entomology 22:388–396. Northeastern Naturalist A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 B246 Vol. 24, Special Issue 7 Intergovernmental Panel on Climate Change (IPCC). 2013: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [T.F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (Eds.).]. Cambridge University Press, Cambridge, UK. 1535 pp. Irwin, J.T., and R.E. Lee Jr. 2000. Mild weather temperatures reduce survival and potential fecundity of the Goldenrod Gall Fly, Eurosta solidaginis (Diptera: Tephritidae). Journal of Insect Physiology 46:655–661. Irwin, J.T and R.E. Lee Jr. 2003. Cold winter microenvironments conserve energy and improve overwintering survival and potential fecundity of the Goldenrod Gall Fly, Eurosta solidaginis. Oikos 100:71–78. Irwin, J.T., V.A. Bennett, and R.E. Lee Jr. 2001. Diapause development in frozen larvae of the Goldenrod Gall Fly, Eurosta solidaginis Fitch (Diptera: Tephritidae). Journal of Comparative Physiology B 171:181–188. Lee, R.E. Jr., R.A. Dommel, K.H. Joplin, and D.L. Denlinger. 1995. Cryobiology of the freeze-tolerant gall fly Eurosta solidaginis: Overwintering energetics and heat-shock proteins. Climate Research 5(1):61–67 Legendre, P. 1993. Spatial autocorrelation: Trouble or new paradigm? Ecology 74(6):1659–1673. Lindsey, A.A., and J.E. Newman. 1956. Use of official weather data in spring time: Temperature analysis of an Indiana phenological record. Ecology 37(4):812–823 MacLean, S.F. 1983. Life cycles and the distribution of pullids (Homoptera) in arctic and subarctic Alaska. Oikos 40:445–451 O’Connor, R.S., R.S. Hails, and J.A. Thomas. 2014. Accounting for habitat when considering climate: has the niche of the Adonis Blue Butterfly changed in the UK? Oecologia 174:1463. Available online at doi:10.1007/s00442-013-2850-1. PRISM Climate Group. 2016. PRISM climate data. Northwest Alliance for Computational Science and Engineering, Oregon State University, Corvallis, OR. Available online at http://prism.oregonstate.edu. Accessed 14 March 2017. R Core Team. 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available online at http://www.R-project. org/. Accessed 13 February 2016. Roltsch, W.J., F.G. Zalom, A.J. Strawn, J.E. Strand, and M.J. Pitcairn. 1999. Evaluation of several degree-day estimation methods in California climates. International Journal of Biometeorology 42:169–176. Schmid, B., G.M. Puttick, K.H. Burgess, and F.A. Bazzaz. 1988. Correlations between genet architecture and some life-history features in three species of Solidago. Oecologia 75:459–464 Stireman, J.O. III, J.D. Nason, and S.B. Herd. 2005. Host-associated genetic differentiation in phytophagous insect: General phenomenon or isolated exceptions? Evidence from a goldenrod–insect community. Evolution 59:2573–2587. Stuhldreher, G., G. Hermann, and T. Fartmann. 2014 Cold-adapted species in a warming world: An explorative study on the impact of high winter temperatures on a continental butterfly. Entomologia Experimentalis et Applicata 151:270–279. Sumerford, D.V., W.G. Abrahamson, and A.E. Weis. 2000. The effects of drought on the Solidago altissima–Eurosta solidaginis-natural enemy complex: Population dynamics, local extirpations, and measures of selection intensity on gall size. Oecologia 122:240–248 Northeastern Naturalist B247 A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 Vol. 24, Special Issue 7 Uhler, L.D. 1951. Biology and ecology of the Goldenrod Gall Fly, Eurosta solidaginis (Fitch). Cornell University Agricultural Station Memoir 300:1–51. University of California (UC) Statewide Integrated Pest Management Project. 2002. Degree days and phenology models. Available online at: http://www.ipm.ucdavis.edu/ WEATHER/ddconcepts.html. Accessed 13 February 2016. Waring, G.L., W.G. Abrahamson, and D.J. Howard. 1990. Genetic Differentiation among host-associated populations of the gallmaker Eurosta solidaginis (Diptera: Tephritidae). Evolution 44:1648–1655 Whipple, A.V., W.G. Abrahamson, M. Khamiss, P.L. Heinrich, A.G. Urian, and E.N. Northridge. 2009. Host-race formation: Promoted by phenology, constrained by heritability. Journal of Evolutionary Biology 22:793–804. Northeastern Naturalist A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 B248 Vol. 24, Special Issue 7 Appendix 1. Sites used for the spatial auto-logistic regression of presence and absence of gigantea flies. S.G. is the presence or absence of Solidago gigantea (Giant Goldenrod) coded at “p” or “a” respectively with n/a = plant not present; G.G. is the presence or absence of galls on S. gigantea coded as “p” or “a” respectively; A.G. is presence or absence of galls on S. altissima (Late Goldenrod) coded as “p” or “a” respectively; S-DD is the estimated degree days over 10 °C summed over April and May; and Energy is the estimated grams of carbohydrate plus lipid metabolized by a fly summed over November, December, January, and February. Lat (°N) Long (°W) State Souce S.G. G.G. A.G. S-DD Energy 40.320 87.062 IN This Paper p a p 244.54 1.880 40.961 76.879 PA Brown et al. 1996 p a p 203.43 2.017 40.968 75.122 PA Brown et al. 1996 p a p 208.26 1.955 41.021 77.738 PA This Paper p a p 182.46 1.896 41.643 87.067 IN This Paper p a p 212.76 1.911 41.724 75.648 PA This Paper p a p 131.93 1.683 41.775 75.591 PA This Paper p a p 103.59 1.582 41.786 75.506 PA This Paper p a p 77.56 1.468 41.790 75.536 PA This Paper p a p 89.72 1.521 41.805 75.563 PA This Paper p a p 96.15 1.545 41.812 75.571 PA This Paper p a p 96.15 1.545 42.025 84.350 MI This Paper p a p 162.27 1.678 42.186 85.644 MI This Paper p a p 205.08 1.805 42.262 84.996 MI This Paper p a p 174.41 1.715 42.372 72.635 MA This Paper p a a 178.84 1.844 42.423 72.623 MA This Paper p a p 173.82 1.820 42.546 85.908 MI This Paper p a p 157.96 1.755 42.611 72.583 MA This Paper p a p 146.92 1.701 42.711 85.464 MI This Paper p a p 157.10 1.651 42.733 85.482 MI This Paper p a p 157.10 1.667 42.794 85.516 MI This Paper p a p 154.14 1.669 42.843 75.551 NY This Paper p a p 113.03 1.568 43.016 85.936 MI This Paper p a p 132.14 1.745 43.163 85.706 MI This Paper p a p 144.86 1.630 43.413 86.351 MI This Paper p a p 114.32 1.736 43.548 85.767 MI This Paper p a p 120.77 1.553 43.565 86.348 MI This Paper p a p 114.79 1.682 43.603 85.811 MI This Paper p a p 124.50 1.531 41.373 89.798 IL This Paper p p p 241.77 1.654 41.708 92.779 IA This Paper p p p 196.03 1.496 41.930 88.750 IL Waring et al. 1990 p p p 178.57 1.526 41.978 89.362 IL Brown et al. 1996 p p p 200.27 1.553 42.035 93.621 IA Sumerford et al. 1991 p p p 228.03 1.520 42.639 85.455 MI This Paper p p 164.89 1.677 42.833 85.828 MI This Paper p p p 152.45 1.748 42.917 72.267 NH This Paper p p p 140.47 1.615 42.934 72.278 NH Waring et al. 1990 p p p 135.00 1.600 42.945 72.785 VT This Paper p p 71.02 1.401 Northeastern Naturalist B249 A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 Vol. 24, Special Issue 7 Lat (°N) Long (°W) State Souce S.G. G.G. A.G. S-DD Energy 42.953 72.316 NH This Paper p p p 136.44 1.587 43.031 71.450 NH This Paper p p a 128.96 1.667 43.034 72.664 VT Brown et al. 1995 p p p 94.50 1.437 43.049 86.109 MI This Paper p p p 131.62 1.826 43.068 72.458 VT Brown et al. 1995 p p p 126.90 1.566 43.084 72.428 NH Brown et al. 1995 p p p 123.84 1.554 43.098 71.465 NH Waring et al. 1990 p p a 122.74 1.579 43.126 71.441 NH This Paper p p a 122.00 1.563 43.131 71.453 NH Waring et al. 1990 p p a 123.30 1.563 43.133 71.417 NH This Paper p p a 122.00 1.560 43.168 72.457 VT Brown et al. 1995 p p p 120.40 1.535 43.263 85.812 MI This Paper p p p 135.34 1.612 43.263 72.596 VT Brown et al. 1996 p p a 107.13 1.435 43.265 72.577 VT Brown et al. 1995 p p p 109.09 1.447 43.351 72.359 NH This Paper p p p 121.42 1.490 43.361 72.311 NH This Paper p p p 115.44 1.499 43.381 72.416 VT Brown et al. 1995 p p p 120.26 1.473 43.422 85.790 MI This Paper p p 115.78 1.585 43.579 75.384 NY This Paper p p p 83.54 1.360 43.579 75.380 NY This Paper p p p 83.54 1.360 43.601 75.397 NY This Paper p p p 77.63 1.330 43.629 72.322 NH Brown et al. 1995 p p a 112.71 1.445 43.631 72.350 VT Brown et al. 1995 p p a 112.71 1.438 43.636 72.225 NH This Paper p p p 108.95 1.437 43.673 72.309 VT Brown et al. 1996 p p a 117.39 1.449 43.685 75.446 NY This Paper p p a 77.10 1.307 43.759 72.452 VT Brown et al. 1995 p p a 102.42 1.370 43.784 72.473 VT Brown et al. 1995 p p a 104.22 1.376 43.825 86.399 MI This Paper p p p 121.87 1.737 43.840 85.849 MI This Paper p p a 130.96 1.492 43.844 72.185 VT Brown et al. 1995 p p a 115.08 1.418 43.869 72.183 VT Brown et al. 1995 p p a 112.92 1.404 43.933 86.114 MI This Paper p p a 129.72 1.579 43.967 70.117 ME This Paper p p a 99.20 1.581 44.011 86.472 MI This Paper p p a 125.06 1.761 44.015 73.167 VT Brown et al. 1996 p p p 131.81 1.509 44.015 72.105 VT Brown et al. 1995 p p p 104.86 1.350 44.033 72.090 VT Brown et al. 1995 p p a 103.07 1.328 44.034 86.407 MI This Paper p p p 123.50 1.750 44.074 72.032 NH This Paper p p p 96.61 1.295 44.100 70.214 ME Waring et al. 1990 p p a 109.17 1.599 44.112 72.055 VT Brown et al. 1995 p p p 94.59 1.288 44.116 93.708 MN Brown et al. 1996 p p a 170.59 1.212 44.117 72.017 NH This Paper p p p 96.86 1.290 44.170 86.129 MI This Paper p p a 120.87 1.683 44.199 94.021 MN This Paper p p p 189.56 1.267 44.200 71.952 NH Brown et al. 1995 p p a 98.79 1.309 Northeastern Naturalist A.V. Whipple, J.T. Irwin, P.L. Heinrich, and W.G. Abrahamson 2017 B250 Vol. 24, Special Issue 7 Lat (°N) Long (°W) State Souce S.G. G.G. A.G. S-DD Energy 44.205 72.917 VT Brown et al. 1996 p p a 58.97 1.226 44.261 72.574 VT Waring et al. 1990 p p a 83.90 1.324 44.319 93.938 MN This Paper p p a 201.64 1.286 44.330 93.916 MN This Paper p p a 199.18 1.283 44.333 72.983 VT This Paper p p p 80.16 1.312 44.350 71.897 VT Brown et al. 1995 p p a 94.75 1.298 44.459 93.162 MN Brown et al. 1996 p p a 170.74 1.221 44.477 73.185 VT This Paper p p p 125.17 1.449 44.569 74.961 NY This Paper p p p 93.11 1.343 44.624 75.236 NY This Paper p p p 104.24 1.384 44.634 73.124 VT This Paper p p p 113.79 1.429 44.677 75.009 NY This Paper p p a 102.98 1.365 44.689 73.297 VT This Paper p p p 118.21 1.500 44.691 75.042 NY This Paper p p p 103.01 1.363 44.766 73.312 VT This Paper p p p 115.69 1.465 44.810 73.104 VT This Paper p p p 110.03 1.409 44.859 73.360 VT This Paper p p a 116.24 1.430 44.885 73.354 VT This Paper p p a 115.37 1.426 44.922 73.110 VT This Paper p p p 111.96 1.429 44.927 72.793 VT This Paper p p p 126.60 1.386 44.933 73.150 VT This Paper p p p 112.36 1.444 44.954 73.205 VT This Paper p p p 115.17 1.442 45.054 87.749 WI Brown et al. 1996 p p a 97.85 1.350 45.056 92.804 MN Sumerford et al. 2000 p p p 177.49 1.234 45.067 93.168 MN Sumerford et al. 2000 p p p 179.87 1.247 45.101 87.631 WI Brown et al. 1996 p p a 100.20 1.356 45.108 87.614 MI Brown et al. 1996 p p p 100.20 1.360 45.175 93.867 MN Sumerford et al. 2000 p p a 178.28 1.193 45.268 93.007 MN Sumerford et al. 2000 p p p 165.45 1.209 45.553 84.784 MI Brown et al. 1996 p p p 78.99 1.397 45.669 84.777 MI This Paper p p p 74.11 1.446 45.710 87.566 MI This Paper p p p 89.95 1.237 46.250 86.000 MI This Paper p p 68.58 1.262 46.559 91.599 WI This Paper p p p 89.65 1.045 46.763 92.171 MN This Paper p p p 40.72 1.073