You are here

Harvard Forest Data Archive

HF323

Northern Range Limit of Aphaenogaster picea in Maine 2015

Related Publications

Data

Overview

  • Lead: Aaron Ellison, Amy Arnett, Sara Helms Cahan, Nicholas Gotelli
  • Investigators: Andrew Nguyen, Megan Brown, Jordan Zitnay
  • Contact: Aaron Ellison
  • Start date: 2015
  • End date: 2015
  • Status: completed
  • Location: Maine
  • Latitude: 44.60 to 45.17
  • Longitude: -70.44 to -68.33
  • Elevation: 147 meter
  • Taxa: Formicidae, Aphaenogaster picea
  • Release date: 2019
  • Revisions:
  • EML file: knb-lter-hfr.323.2
  • DOI: digital object identifier
  • Related links:
  • Study type: short-term measurement, modeling
  • Research topic: ecological informatics and modelling; physiological ecology, population dynamics and species interactions; regional studies
  • LTER core area: populations
  • Keywords: ants, distribution, physiology, temperature
  • Abstract:

    Low temperatures at poleward range margins of terrestrial species tend to match cold tolerance limits, suggesting that range boundaries may be set by evolutionary constraints on cold physiology. The northeastern woodland ant, Aphaenogaster picea, occurs up to approximately 45 °N in central Maine. We combined presence-absence surveys with regression-tree analysis to characterize its northern range limit, and assayed two measures of cold tolerance operating on different time-scales to determine whether and how marginal populations adapt to environmental extremes. The boundary was predicted primarily by temperature, but low winter temperatures did not emerge as the primary correlate of species occurrence. Low summer temperatures and high seasonal variability predicted absence above the boundary, whereas high mean annual temperature (MAT) predicted presence in southern Maine. Locations between these zones formed an east-west band where presence was conditional on precipitation. In contrast, assays of cold tolerance across multiple sites indicated substantial local adaptation of cold tolerance at the range edge, with a 4-minute reduction in chill-coma recovery time across a 2-degree reduction in MAT. Baseline tolerance and capacity for additional plastic cold-hardening shifted in opposite directions, with hardening capacity approaching zero at the coldest sites. This trade-off suggests that populations at range edges may adapt to colder temperatures through genetic assimilation of plastic responses, potentially constraining further adaptation and range expansion.

  • Methods:

    Field surveys and niche modeling

    Aphaenogaster picea is a common forest ant species that ranges from the high elevations of Virginia into northern Minnesota and Maine. Its northern range boundary---estimated from geo-referenced museum specimens---occurs in Maine near 45 °N. To further characterize the northern range boundary of A. picea, we combined presence-absence data from previous field surveys at 27 sites (see HF147) with new data collected for this study in July and August 2015 from 75 additional sites. These 75 sites were sampled randomly along a 65-km East-West belt transect centered on 45 °N latitude and running across all of central Maine. For the July 2015 survey of 32 sites, two researchers searched each site haphazardly for A. picea colonies for 20 minutes in deciduous and mixed hardwood forests. For the August 2015 survey of 43 additional sites, two researchers established 50 × 50 m plots and searched them for colonies for 20 minutes each (40 person-minutes per plot).

    We used classification and regression tree (CART) analysis, implemented in the rpart package (version 4.1-10) in R version 3.4.2, to determine which of the 19 bioclimatic variables (at 2.5-min resolution) downloaded from WorldClim (http://www.worldclim.org/bioclim) best predicted the occurrence of A. picea. To obtain the optimal regression tree and to avoid over-fitting the data, we pruned the tree so that it had the lowest complexity parameter and smallest cross-validated error. For each cross-validation, CART models were fitted to a training set and then used to predict presence or absence in the testing set. In total, we analyzed 10 independent cross-validations.

    Field collecting and rearing conditions in a common garden

    To test for local adaptation in cold performance, we collected 16 colonies from 16 unique sites along the range boundary in July, 2015. At each site, we collected whole colonies, including workers, larvae, pupae, and queens (when possible). Collected colonies were housed in 22 × 16 cm plastic containers and maintained in a 12:12 light:dark cycle at ≈50% humidity and 25 °C (i.e., within the range of optimal development). To minimize the contribution of the source environment, colonies were lab-acclimated in these conditions for at least one month before any physiological measurements were taken. Ant workers typically live between a few weeks to months. Colonies readily nested within glass test-tubes that were plugged with water-saturated cotton to maintain humidity. Each colony was fed 100 μL of 20% honey in water and one bisected meal worm three times a week.

    Constructing cold-performance curves

    We exposed ants to a series of cold pre-treatments, recovery treatments, and subsequent temperature treatments to construct cold-performance curves for adult workers from each lab-acclimated colony. Cold-resistance was measured as the Chill Coma Recovery (CCRT): the time in seconds needed for an individual ant worker to orient itself in an upright position and take one step after a 1-hour exposure to −5 °C.

    To assess basal cold-resistance and cold-hardening ability, we measured the CCRT of ants that were pre-treated for 1 hour at one of four temperature treatments: 25, 5, 0, or −5 °C (fig. 1). The 25 °C pre-treatment is the control that represents basal cold-resistance, and the −5, 0, and 5 °C pre-treatments represent cold-hardening at different temperature levels. All ants were temperature-treated in a circulating water bath (Polyscience, USA) that contained a 50% ethylene glycol solution to prevent freezing. For each colony, four ants per pre-treatment were placed in a sealed 15 × 160 mm glass test tube for one hour at the pre-treatment temperature. After one hour, the ants were removed and placed into a test tube and allowed to recover at 25 °C for one hour, then placed in the −5 °C treatment for another hour. After this second hour, ants were placed in glass tubes with water-soaked cotton plugs. CCRT was measured by an observer without prior knowledge of pre-treatment groups. We excluded 24 out of 272 ants tested that were lost or did not survive handling.

    Evaluating trade-offs in cold performance

    We adapted multivariate methods from quantitative genetics to estimate the correlation structure between basal cold-resistance and cold hardening in A. picea. In Kingsolver's original analyses, the data consist of performance traits that were measured for replicated genotypes For a set of n traits, the resulting n × n G matrix measures the additive genetic variance in the traits along the diagonal, and the additive covariances between pairs of traits in the off-diagonal elements. In our analyses, the data consist of chill coma recovery time (CCRT) that was measured at four different pre-treatment temperatures for 16 different colonies. We equate CCRT measured at the 4 treatments as 4 "traits" of the populations. We estimated the colony level, 4 × 4 variance-covariance matrix by first fitting a mixed-effect model and then extracting variance and covariance components using the lme4 package in R.

    Y_ijk = μ + C_jk + ϵ_ijk

    where Y_ijk represents the value of CCRT for ant worker i of colony j measured for each pre-treatment temperature k, μ is the fixed effect of the intercept, C_jk is the random effect of colony j at pre-treatment temperature k, and ϵ_ijk represents the residual error. Colony and residual within colony-level effects were treated as random effects with an unconstrained covariance structure. We extracted variance and covariance components that make up the 4 × 4 matrix with the VarCorr() function from the mixed effects model.

    Estimates of the variance-covariance matrix were calculated from untransformed data because each trait (CCRT under each pre-treatment temperature) have the same units. We then decomposed G using principal components analysis (PCA) to produce orthogonal eigenvectors, which represent independent axes of genetic correlations between traits. The first principal component, gmax, is the eigenvector that explains most of the variation in the G matrix. The pattern of loadings for gmax can be biologically interpreted as falling within one of three broad scenarios. In the first scenario, if colonies exhibited a constant difference in performance across all pre-treatment temperatures (additive variation), then the gmax loadings would be all negative or all positive. In the second scenario, if colonies varied in optimal cold performance between low and high pre-treatment temperatures (cooler-warmer variation), then the gmax loadings would be positive and negative or vice versa from low to high pre-treatment temperatures. In the third scenario, if colonies exhibited generalist-specialist variation, then the gmax loadings would be opposite between intermediate and extreme pre-treatment temperatures.

    Determining the relationship between cold-resistance and climate

    To explore how variation in cold resistance traits is related to the local thermal environment at the range margin, we tested for an effect of local temperature MAT on baseline cold resistance and cold hardening capacity. To calculate baseline cold resistance, we transformed CCRT so that higher values indicate greater cold resistance. Specifically, the maximum CCRT value in the entire dataset CCRTmax was treated as a reference level, and each CCRT measurement (CCRTobs) was subtracted from CCRTmax (CCRTmax – CCRTobs) to measure relative cold resistance. To calculate cold hardening capacity, we used the same data transformation and for each colony subtracted the average CCRT at each pre-treatment temperature (−5,0, 5 °C) from the average CCRT at the 25 °C pre-treatment temperature (−5, 0, and 5 °C; CCRT25 °C − CCRTpre-treatment temperature). To detect simple linear and non-linear relationships between local temperature and cold resistance or cold-hardening, we fit a regression model with cold resistance or cold-hardening as the response variable, and a linear and a quadratic term for temperature as the predictor variable. We used AIC model selection to determine whether the quadratic term should be retained or dropped from the final model.

  • Use:

    This dataset is released to the public under Creative Commons license CC BY (Attribution). Please keep the designated contact person informed of any plans to use the dataset. Consultation or collaboration with the original investigators is strongly encouraged. Publications and data products that make use of the dataset must include proper acknowledgement.

  • Citation:

    Ellison A, Arnett A, Helms Cahan S, Gotelli N. 2019. Northern Range Limit of Aphaenogaster picea in Maine 2015. Harvard Forest Data Archive: HF323.

Detailed Metadata

hf323-01: cold physiology

  1. colony: unique colony name
  2. date: date collected
  3. state: Vermont or Maine, USA
  4. county: county in which the colony was collected
  5. town: town in which the colony was collected
  6. lat: latitude (unit: degree / missing value: NA)
  7. lon: longitude (unit: degree / missing value: NA)
  8. altitude.ft: altitude (unit: foot / missing value: NA)
  9. temp: ground temperature (unit: celsius / missing value: NA)
  10. wind.speed: recorded wind speed at time of collection (unit: metersPerSecond / missing value: NA)
  11. humidity: recorded humidity at time of collection (unit: dimensionless / missing value: NA)
  12. wind.chill: recorded wind chill at time of collection (unit: fahrenheit / missing value: NA)
  13. heat.stress.index: heat stress index at time of collection (unit: dimensionless / missing value: NA)
  14. barometric.pressure: barometric pressure at time of collection (unit: atmosphere / missing value: NA)
  15. soil.temp: soil temperature at time of collection (unit: celsius / missing value: NA)
  16. canopy.photo: reference number for the picture of the canopy at the site of collection
  17. tree.species: types of trees at collection site
  18. understory: type of understory at collection site
  19. habitat.photo.number: reference for habitat photo at collection point
  20. nest.substrate: type of substrate collected from
  21. ct.max: upper thermal limit, time at which ants lose motor function (mean) when ramped at 0.1 degree C per minute (unit: celsius / missing value: NA)
  22. mat: from bioclim database: mean annual temperature (unit: celsius / missing value: NA)
  23. mdr: from bioclim database: mean diurnal temperature (unit: celsius / missing value: NA)
  24. iso: from bioclim database: isothermality (unit: dimensionless / missing value: NA)
  25. sd: from bioclim database: standard deviation (standard deviation * 10) (unit: dimensionless / missing value: NA)
  26. t.max: from bioclim database: thermal maximum (unit: celsius / missing value: NA)
  27. t.min: from bioclim database: thermal minimum (unit: celsius / missing value: NA)
  28. tar: from bioclim database: temperature annual range (unit: celsius / missing value: NA)
  29. twq: from bioclim database: mean temperature at wettest quarter (unit: celsius / missing value: NA)
  30. tdq: from bioclim database: mean temperature at driest quarter (unit: celsius / missing value: NA)
  31. t.warm.q: from bioclim database: mean temperature at warmest quarter (unit: celsius / missing value: NA)
  32. t.min.q: from bioclim database: mean temperature at coldest quarter (unit: celsius / missing value: NA)
  33. ap: from bioclim database: annual precipitation, mean (unit: millimeter / missing value: NA)
  34. pwm: from bioclim database: precipitation at wettest month (unit: millimeter / missing value: NA)
  35. pdm: from bioclim database: precipitation at driest month (unit: millimeter / missing value: NA)
  36. psd: from bioclim database: precipitation standard deviation (coefficient of variation) (unit: dimensionless / missing value: NA)
  37. pwq: from bioclim database: precipitation at wettest quarter (unit: millimeter / missing value: NA)
  38. pdq: from bioclim database: precipitation at driest quarter (unit: millimeter / missing value: NA)
  39. p.warm.q: from bioclim database: precipitation at warmest quarter (unit: millimeter / missing value: NA)
  40. p.min.q: from bioclim database: precipitation at coldest quarter (unit: millimeter / missing value: NA)
  41. pretreat.temp: pre-treatment temperature (unit: celsius / missing value: NA)
  42. pre: chill coma recovery time after the pre-treatment temperature in seconds (unit: second / missing value: NA)
  43. treatment.recovery.s: chill comma recovery time in seconds (unit: second / missing value: NA)
  44. diff: the difference in chill coma recovery time between treatment.recovery.s and pre (unit: second / missing value: NA)

hf323-02: Aphaenogaster picea distribution

  1. n: reference number for sample
  2. date: date
  3. state: Vermont or Maine, USA
  4. county: county
  5. locality: micro site identification
  6. habitat: type of area
  7. lat: latitude (unit: degree / missing value: NA)
  8. lon: longitude (unit: degree / missing value: NA)
  9. masl: elevation (unit: meter / missing value: NA)
  10. subfamily: subfamily of ant species - Myrmicinae
  11. ant.genus: genus of ant species – Aphaenogaster
  12. ant.species: species name of ant; picea
  13. code: HF code for genus species
  14. collection: who collected the colony
  15. collector: person who collected the colony
  16. found.notfound: presence-absence data
    • 0: absent
    • 1: present

hf323-03: R markdown script

  • Compression: none
  • Format: R code
  • Type: R code

hf323-04: R markdown script output

  • Compression: none
  • Format: pdf
  • Type: pdf