Thermal Traits of Anurans Database for the Southeastern United States (TRAD): A Database of Thermal Trait Values for 40 Anuran Species

We identified nine traits that are related to an animal’s response to temperature (referred to as thermal traits; Table 1). We chose traits that are commonly used to parameterize the biophysical model NicheMapR, which uses mass–balance equations to estimate how solar radiation, air temperature, and soil temperature interact to affect the body temperature of a model ectotherm (Peterman and Gade, 2017; Kearney and Porter, 2020). Although mass, emergence temperature, and activity period are not strictly thermal traits (Taylor et al., 2021), they are directly related to an animal’s response to temperature, and these traits are expected to change due to direct and indirect effects of climate change (Sheridan et al., 2018; Levy et al., 2019).

Physiological traits included a species’ preferred temperature, critical thermal maximum, and critical thermal minimum. These traits are related to physical and metabolic processes that alter an animal’s thermoregulation. Preferred temperature is the body temperature to which the species regulates and is often closely associated with optimal performance temperatures (Angilletta et al., 2002). Both critical thermal maximum and critical thermal minimum describe physiological limits and estimate the temperature at which an individual loses coordinated movement ability (i.e., biological death; Lutterschmidt and Hutchison, 1997).

Mass estimates the size of the animal, which can alter an animal’s heat conservation through their surface area to volume ratio and microclimatic experience (Kearney et al., 2021). Although mass is typically considered an emergent property between species’ trophic traits and the environment, macroecological patterns in adult anuran body size suggest that anurans leverage their thermoregulatory abilities to reach larger sizes in low energy areas, allowing them to conserve heat in low temperatures (Olalla-Tárraga and Rodríguez, 2007). As a result, anurans of different sizes could have different thermal traits and needs. In this database, we recorded any available mass estimates for different life stages; mean mass was reported across all life stages with the exception of three maximum adult mass values (Table 2).

Behavioral traits included activity period, retreat emergence temperature, basking temperature, and foraging temperature limits. These traits aid thermoregulation by allowing animals to select times and habitats that meet their physiological needs. The activity period describes when the animal prefers to be active on the landscape each day. Activity is phylogenetically conserved (Anderson and Wiens, 2017) and can potentially reduce an animal’s exposure to extreme temperatures (Hutchison and Dupré, 1992). Anurans generally cease surface activity in the winter and withstand extreme winter temperatures by seeking hibernacula, overwintering within deep water, or using freeze tolerance (Pinder et al., 1992). A species’ emergence temperature describes the environmental temperature at which an animal annually initiates surface activity after winter dormancy (Green et al., 2016). Basking temperature is the environmental temperature at which an animal begins basking to quickly increase its temperature and improve metabolic performance; however, basking may also increase the rate of water loss and risk of dehydration (Freed, 1980; Muths and Corn, 1997). Although few studies have investigated the benefits of the behavior, basking behavior has been documented in both toads (Bufonidae) and treefrogs (Hylidae; Wells, 2010). Foraging maximum and minimum temperature describes the range of environmental temperatures within which an animal is foraging on the landscape. Species foraging temperature limits likely differ based on species’ predation risk, energy loss to thermoregulation, and energy gained through foraging (Putnam and Bennett, 1981).

We took a species-centric approach to search the literature for thermal trait data. We used a combination of US state agency species lists and International Union for the Conservation of Nature (IUCN) range maps to identify all anurans native to Alabama, Florida, Georgia, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia comprising the southeastern United States (Fig. 1; International Union for Conservation of Nature, 2017). We excluded Collinses’ Mountain Chorus Frog (Pseudacris collinsorum) because it was recently described as distinct from the Mountain Chorus Frog (Pseudacris brachyphona; Ospina et al., 2020). Species names were standardized using the Global Biodiversity Information Facility (GBIF) Backbone Taxonomy (GBIF Secretariat, 2019). For the trilling chorus frog complex (Pseudacris spp.) and North American cricket frogs (Acris spp.), we updated species names according to the study location using distributions outlined in Lemmon et al. (2007) and Gamble et al. (2008), respectively. As a result, the traits identified as Western Chorus Frog (Pseudacris triseriata) were reassigned to other species, and the Western Chorus Frog was removed from the focal species list. For each species, we used the Integrated Taxonomic Information System (ITIS) to ensure taxonomic synonyms were included in our search for trait values (ITIS, 2021).

We collected thermal traits described for each taxon found within Hutchison and Dupré (1992), Rome et al. (1992), Wells (2010), and Herpetological Review (2017). We also built species keyword pairs (e.g., “Pseudacris fouquettei” critical thermal maximum) and searched Web of Science’s (2022) Core Collection using species keyword searches for all available sources; species’ literature searches began 1 September 2020 and concluded 5 July 2022 (Table 1). If the Web of Science search returned no records, we searched Google Scholar (2022) with the species keyword pair. Once sources were identified, we screened sources’ abstracts for any mention of thermal traits. This method identified peer-reviewed articles, dissertations, and theses as sources from which we transcribed reported trait information.

From each source, we recorded the species, trait value reported (e.g., mean mass), life stage, sex, study location, trait value variation (e.g., standard deviation of mass), and acclimation temperature, if available. Because there are many ways to measure and describe animals’ thermal traits (Lutterschmidt and Hutchison, 1997; Navas et al., 2021; Taylor et al., 2021), we categorized the methodology used to identify a trait value as the “type of evidence” (Table 2). Type of evidence could be the physiological endpoint used to quantify a trait, the type of experiment used to describe a trait, or the type of observation. Evidence type categorization reflects precision ranked A–C for most to least precise and aims to help database users make informed decisions about trait value inclusion based on how the values were originally measured. For trait data that were reported graphically without a numerical description, we used WebPlotDigitizer to digitize any plots (Rohatgi, 2021). When trait values were reported from a different source (e.g., for comparison among species within a discussion), we incorporated trait values from the original source following the same methodology described here. Overall, our process retains intraspecific trait data when available, allowing multiple trait values recorded for a single trait for a given species.

We conducted quality control measures to reduce transcription errors when recording traits found in the literature. Each trait value was checked by two separate authors. After data compilation, we checked the database for formatting errors (e.g., typing errors such as taxon name misspellings). For numeric traits (all except Activity), we identified trait values that were more than 2.5 standard deviations away from the mean at the family, genus, and species level to identify potential errors at levels of organization that might have different means and standard deviations. These flagged trait values were double checked; if the value was correctly reported, we made no change. Overall, 32 continuous trait values (4.2%) were flagged as outliers, and one value was corrected. No traits were imputed for this database.

We then summarized intraspecific trait data for interspecific comparisons. For categorical activity data, we summed the number of sources that indicate a species’ diel activity type. For continuous traits, if the trait is a minimum trait (i.e., critical thermal minimum, emergence temperature, minimum foraging temperature), we retained the minimum trait data point across all references. If the trait is a maximum trait (i.e., critical thermal maximum, maximum foraging temperature), we retained the maximum trait data point across all references. For all other traits (i.e., mass, thermal preference), we calculated the mean across all references. The TRAD database can be found at DuBose et al. (2023) and the Open Traits Network, a decentralized network of trait databases that follow open science principles (Gallagher et al., 2020).

We reported the number of unique trait values (intraspecific trait values), the number of sources, and the journal with the most sources for the TRAD database. We investigated whether the publication of physiological trait data has increased or decreased through time using a Mann-Kendall test, which tests for a monotonic trend in the number of unique species and trait combinations (R Core Team, 2020; McLeod, 2022). We quantify the matrix completion of the database for adult traits as a percent of the 360 trait values possible (9 traits for 40 species). To further quantify database completion, we calculated trait coverage as the number of species (maximum = 40) that had at least one value for the trait in question and trait completeness as the number of traits (maximum = 9) that had at least one value for the species in question. Total trait coverage and trait completeness are reported as a percentage, but trait-specific trait coverage and species-specific trait completeness are reported as either n species or n trait values, respectively. We report median and interquartile range (IQR) trait coverage and trait completeness for the TRAD database for adult or unknown life stages. We also report median trait coverage and trait completeness for all five life stages: eggs/embryos, tadpoles, metamorphs (intermediate life stages between tadpole and juvenile), juveniles, and adults.

We also investigated taxonomic and trait-based data gaps in thermal trait availability. We evaluated differences in rank mean trait completeness among genera and families using a Kruskal-Wallis rank sum test and Dunn’s test to evaluate significant differences among groups (alpha = 0.05; Dunn, 1964; Dinno, 2017). To investigate whether species’ range size is correlated with trait availability, we tested whether area of occurrence was correlated with number of recorded adult trait values and trait values for all life stages for each species using a Spearman’s rho correlation. We used previously calculated estimates of area of occurrence for most species (DuBose et al., 2022a, 2022b). Briefly, species’ point occurrences from data repositories of museum records, survey results, and community/citizen science efforts (HerpMapper, 2019; GBIF.org, 2021a, 2021b) were filtered to remove geographically, climatically, and taxonomically uncertain points and used to identify U.S. Geological Survey (USGS) 12-digit hydrologic unit code (HUC12) watershed sub-basin boundaries that contained at least one point occurrence (U.S. Geological Survey, 2020; DuBose et al., 2022a). The areas of occupied USGS HUC12 watershed sub-basin boundaries were then summed to produce an estimate of the area of occupancy (as a proxy for range size) for each species. When no point occurrences were available for the species (i.e., Dusky Gopher Frog Lithobates sevosus), we substituted the IUCN extant range for point occurrences, identified USGS HUC12 sub-basin boundaries that contain the range, and summed the areas of those boundaries (International Union for Conservation of Nature, 2017). All data summarization and analyses were completed in R with the aid of R packages googlesheets4 and googledrive (D’Agostino McGowan and Bryan, 2020; R Core Team, 2020; Bryan, 2021), and plots were created with ggplot2 (Wickham, 2016).

In total, we found 858 reported trait values for thermal traits from 267 sources, with most sources published in Herpetological Review (n = 47). The publication of thermal trait data has been increasing since 1945 (Mann-Kendall tau = 0.4, P < 0.001; Fig. 2), with the highest amount of species trait data published in 1988 (n = 29). Mass estimates accounted for 40.5% of all trait values, physiological traits were 37.9% of all trait values, and behavioral trait values were 21.7% of all trait values. Methodology consistent with the top evidence category (A) accounted for 47.8% of trait values; however, most trait values for critical thermal maximum (88.9%), emergence temperature (67.5%), and basking temperature (50%) were associated with a lower evidence category (B; Table 2).

Our database had a 43.9% matrix completion, 47.5% median trait coverage, and 44% median trait completeness. For the adult life stage, traits had a median trait coverage of 19 of 40 species (IQR = 24 species). Although mass was the most reported trait value (348 trait values for 31 species), activity had the highest trait coverage, with activity estimates available for 34 species. Trait values for basking temperature were available for only three species (Green Treefrog [Dryophytes cinereus]; Spring Peeper [Pseudacris crucifer]; and Blanchard’s Cricket Frog [Acris blanchardi]), making it the trait with the lowest trait coverage (Table 1).

The availability of trait data varied greatly between species (Fig. 1), with Pseudacris crucifer having the highest number of values reported in the database (n = 37 traits from unique sources). Across all species, the median trait completeness was four of nine traits (IQR = 2 traits). Cope’s Gray Treefrog had the highest trait completeness for adults, with eight of nine traits having at least one recorded trait value. We found no trait values for three species of chorus frogs (Brimley’s Chorus Frog [Pseudacris brimleyi]; Southern Chorus Frog [Pseudacris nigrita]; and Little Grass Frog [Pseudacris ocularis]), and we found values for only one trait for the Mountain Chorus Frog (P. brachyphona; emergence temperature; Fig. 1).

Adult trait values were most often reported within the literature (n = 455), although the life stage was not clearly stated for 74 trait values. The remaining trait values were reported for tadpoles (n = 223), metamorphs (n = 52), juveniles (n = 31), and eggs (n = 23). Sixteen species had reported trait values for the tadpole life stage, with critical thermal maximum and mass trait values being the most common (94 and 71 trait values, respectively). When considering trait completeness across all life stages, American Toad (Anaxyrus americanus) had the highest number (n = 15) of unique trait and life stage combinations.

We found both taxonomic and trait-based differences in trait availability within TRAD. Anuran genera varied in their species’ trait completeness (Kruskal-Wallis H₇ = 14.8, P = 0.038; Fig. 3). The genus Scaphiopus, of which only one species is within our study region (Scaphiopus holbrookii), had the highest trait completeness. The genus Pseudacris had the lowest median trait completeness (2 traits, n = 9 species), despite having the species with the most trait values in the database (P. crucifer; Fig. 2). Differences between genera’s rank mean trait completion were not statistically significant (P > 0.10) after Bonferroni correction. Among families, Scaphiopodidae had the highest median trait completeness (7 traits, n = 1 species) and Hylidae had the lowest median trait completeness per species (3 traits, n = 20 species), but rank mean trait completeness was not statistically different among families (H₄ = 3.8, P = 0.44). Species that had higher areas of occurrence often had higher trait completeness (Fig. 4; rho = 0.74, P < 0.001 for adults, rho = 0.79, P < 0.001 for all life stages).

The TRAD database provides a data reservoir for thermal trait data that extends the anuran ecological trait data stored in ATraiU (Moore et al., 2021) and complements other physiological trait databases by providing information about behavioral thermoregulation traits (e.g., emergence temperature; Anderson and Wiens, 2017; Bennett et al., 2018). By compiling TRAD, we identified gaps in anuran thermal trait data, which allows researchers to optimize future data collection by targeting species and traits that are understudied. For example, basking and foraging traits could be used to investigate how temperature shifts might impact frog foraging behavior but are not available for most species. Critical thermal maximum values could be a potentially better predictor of species’ mortality from increased temperature than local climatic values that also exhibit intraspecific variation (Cicchino et al., 2023). Studying the thermal traits of species of Pseudacris could fill in data gaps and improve estimates of the genus’s thermal sensitivity. By identifying gaps in this trait database, we hope to enable researchers to balance uncertainties in scientific knowledge with costs associated with further trait collection when investigating species thermal ecology and climate change vulnerability.

Data gaps for some species and traits might be due to different permitting needs. Field and laboratory studies require collecting permits and animal care and use (e.g., American Association for Laboratory Animal Science [IACUC]) approvals, and measurement of some thermal traits can be physiologically stressful to test animals, which is not ideal for species of conservation concern (Camacho and Rusch, 2017; Taylor et al., 2021). Some thermal traits are more difficult to collect than others: some can be estimated using agar models (Ortega-Chinchilla et al., 2022), some can be measured in the field, and others ideally are measured on laboratory-acclimated animals (Taylor et al., 2021). Therefore, alternative techniques or improvements to current techniques to reduce harm to test animals may help improve availability of trait data (Taylor et al., 2021).

When it is infeasible or unethical to measure trait values (i.e., species of conservation concern), researchers can identify surrogate species that are ecologically similar (i.e., have similar trait values) using trait dissimilarity matrices calculated from existing trait databases (Pracheil et al., 2016). Precision of trait imputation, which can also complete trait matrices with missing trait data by incorporating trait covariance among a hierarchical structure (i.e., a phylogeny or taxonomy), increases with trait data availability (Penone et al., 2014; Johnson et al., 2021). By compiling available trait data, TRAD aids future traits-based research and eases the incorporation of rare, newly described, or understudied species into comparative analyses (like species of Pseudacris, many of which were described in 2007 after 73% of values were already recorded [Lemmon et al., 2007]).

Trait-related and taxonomic biases are common in trait databases (González-Suárez et al., 2012; Violle et al., 2015; Etard et al., 2020) and affect the results of comparative analyses (Freckleton, 2009). For example, correlations among range size and trait data availability have potentially skewed our understanding of the relationship between species’ traits and extinction risk by missing part of the trait gradient; incomplete trait data for mammals that have small ranges have potentially caused us to overestimate the negative relationship between range size and extinction risk (González-Suárez et al., 2012; Fig. 4). For anurans, species’ range size and climate niche breadth have been used to quantify extinction risk and climate change sensitivity (Foden et al., 2013; Griffis-Kyle et al., 2018; DuBose et al., 2022a). Multi-species comparisons of climate change sensitivity based on thermal traits need to carefully consider the species inclusion criteria or how to deal with missing trait data, because species with smaller ranges or within the genera Psuedacris are less likely to have the trait data necessary for their inclusion in the comparison, despite a propensity for higher sensitivity to climate change (DuBose et al., 2022a).

Given climate change is altering water availability (Reid et al., 2019) and hydration affects an animal’s thermoregulation (Anderson and Andrade, 2017; Bovo et al., 2018), understanding how species vary in their water conservation traits will aid in the identification of at-risk species. Although our database does not fill the critical knowledge gap in anuran water conservation trait values, TRAD can be used to identify study species for investigations on how water availability and temperature interact to impact animal fitness.

Interspecific comparisons of thermal traits can inform conservation, ecological, and evolutionary research by revealing differences among species’ sensitivity to abiotic change (Hoffmann et al., 2021), illuminating drivers of community assembly (Wong et al., 2022) and quantifying the evolutionary importance of thermal tolerance (Bennett et al., 2021). Trait databases remove a barrier to trait-based research by collating trait values from the literature and providing transparency regarding the original source and categorical methodology. By collating trait values for 9 thermal traits and 40 species, the TRAD database improves the availability of existing thermal trait data for species native to the southeastern United States. The TRAD database can be leveraged to conduct interspecific comparisons of vulnerability to temperature change, physiological niche space and limitations, and drivers of geographic range limits, among other ecological and evolutionary patterns and processes, for anurans native to the southeastern US.

Associated U.S. Geological Survey (USGS) data release (DuBose et al., 2023) is available at https://doi.org/10.5066/P9HZFHSR. Code used to compile the database files and for this manuscript is archived at https://github.com/TraciPopejoy/AnuransMNMs/tree/main/ATraiU%202.0/code. Unless an alternative copyright or statement noting that a figure is reprinted from a previous source is noted in a figure caption, the published images and illustrations in this article are licensed by the American Society of Ichthyologists and Herpetologists for use if the use includes a citation to the original source (American Society of Ichthyologists and Herpetologists, the DOI of the Ichthyology & Herpetology article, and any individual image credits listed in the figure caption) in accordance with the Creative Commons Attribution CC BY License.

We thank Nicole Beard, Michelle Cox, Emma White, and Emma Hutin for completing early trait compilation tasks. We thank Dr. Martha Muñoz for their advice on database compilation. Support for this research was provided by G19AS00060 of the U.S. Geological Survey (USGS). Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.