Species increasingly face environmental extremes. While responses of morphological traits to changes in average environmental conditions are well-documented, responses to environmental extremes remain poorly understood. Bird bills contribute to thermoregulation, with considerable heat loss possible through the bill surface, and with bill morphology shaped by long-term thermal conditions. We used museum specimens to investigate the relationship of bill surface area (SA) in dark-eyed juncos Junco hyemalis to traditional measures of climate (temperature and precipitation) and to a novel measure of short-term relative temperature extremity, which quantifies the degree to which temperature maxima or minima have diverged from the recent five-year norm. We found that bill SA exhibits different patterns of association with relative extremity depending on the overall temperature regime and on precipitation. While thermoregulatory function predicts larger bill SA at higher relative temperature extremities, we found this to be the case only when the measure of temperature extremity existed in an environmental context that opposed it: relative minimum temperature in a warm climate, or relative maximum temperature in a cool climate. When, instead, environmental context amplified the relative temperature extremity, we found a negative relationship between bill SA and relative temperature extremity. We also found that the strength of associations between bill SA and relative temperature extremity increased as precipitation increased. Our results suggest that trait responses to environmental variation may qualitatively differ depending on the overall environmental context, and that environmental change that extremifies already-extreme environments should be of particular concern. Extreme-on-extreme environmental change may produce responses that cannot be predicted from observations in less extreme contexts, which should make it a priority for research on species' responses to climate change as well as trait evolution generally.Species increasingly face environmental extremes. Morphological responses to changes in average environmental conditions are well-documented, but responses to environmental extremes remain poorly understood. We used museum specimens to investigate relationships between a thermoregulatory morphological trait, bird bill surface area (SA), and a measure of short-term relative temperature extremity (RTE), which quantifies the degree that temperature maxima or minima diverge from the five-year norm. Using a widespread, generalist species, Junco hyemalis, we found that SA exhibited different patterns of association with RTE depending on the overall temperature regime and on precipitation. While thermoregulatory function predicts larger SA at higher RTE, we found this only when the RTE existed in an environmental context that opposed it: atypically cold minimum temperature in a warm climate, or atypically warm maximum temperature in a cool climate. When environmental context amplified the RTE, we found a negative relationship between SA and RTE. We also found that the strength of associations between SA and RTE increased with precipitation. Our results suggest that trait responses to environmental variation may qualitatively differ depending on the overall environmental context, and that environmental change that extremifies already-extreme environments may produce responses that cannot be predicted from observations in less extreme contexts.

Morphological data collection

We measured museum specimens of two subspecies of dark-eyed junco belonging to the "Oregon junco" group, Junco hyemalis pinosus and J. h. thurberi, collected between 15 March and 30 September from 1900 to 1950 within the state of California (Figure 1). We chose this 50-year timespan because it includes >90% of California junco specimens in the collections we accessed; after 1950, specimens are temporally and geographically sparse. We focused on the time period between 15 March and 30 September to ensure that the individuals studied were present to breed; juncos present in the winter may be long-distance migrants wintering in California. Specimens are held in the collections of the Museum of Vertebrate Zoology or the California Academy of Sciences. Metadata associated with the specimens were downloaded from VertNet (http://vertnet.org).

Bill length, width, and depth, as well as tarsus length and wing chord, were measured by KL with digital calipers, and bill surface area was calculated from the three linear bill measurements following Greenberg et al. "Heat loss may explain bill size differences between birds occupying different habitats.," PLoS ONE, vol. 7, no. 7, p. e40933, 2012. Further details can be found in LaBarbera et al. "Complex relationships among environmental conditions and bill morphology in a generalist songbird.," Evolutionary Ecology, vol. 31, pp. 707-724, 2017.

Obtaining environmental measures for each specimen

We obtained records for four monthly climate variables from the PRISM historical climate dataset (PRISM climate group 2015): precipitation and mean, minimum, and maximum temperature. Temperature and precipitation are standard measures of abiotic climate (Danner and Greenberg, "A critical season approach to Allen’s rule: bill size declines with winter temperature in a cold temperate environment," Journal of Biogeography, vol. 42, no. 1, pp. 114-120, 2015). The PRISM historical dataset provides GIS raster files containing the monthly means of the four climate variables as measured over 4 km-by-4 km grid cells across California. To determine which environmental conditions were associated with a given specimen, we used the latitude and longitude at which each specimen was collected to assign a raster cell. A buffer code was used to convert environmental values for each cell to those for a circle with a radius of 15 km centered on the collection locality for each specimen. For each calendar month, values for cells entirely within this circle were averaged to generate a single mean value per variable; raster cells that were only partially located within the 15 km radius were not included in these analyses. This procedure was performed for each of the five years prior to the collection date of a specimen.