Assessing thermal ecology of herpetofauna across a heterogeneous microhabitat mosaic in a changing aridland riparian system

Species–environmental relationships, including drivers of body temperature (T B), are important for understanding thermal ecology and physiological needs of species during climate change. This is especially important among ectotherms, including amphibians and reptiles (i.e., herpetofauna), in aridland riparian systems. Infrared thermography (IRT) can reliably and noninvasively estimate T B of ectothermic herpetofauna while simultaneously assessing thermal heterogeneity across a mosaic of microhabitats. We used IRT at a semi-arid riparian zone in Tucson, Arizona to assess herpetofauna–environmental relationships during early-season activity periods in 2020 and 2022. From mixed-effect modeling of desert riparian herpetofauna (n = 81), we found complex environmental relationships influencing T B. Microhabitat perch surface temperature (T S) best explained T B; many ambient conditions were inadequate at capturing T B. Herpetofauna were as warm by mid-morning than in the early afternoon, with T B approaching equilibrium with T S as mornings progressed. Less T B–T S variation was detected for anurans (e.g., canyon treefrogs, Hyla arenicolor) than with lizards, supporting that desert amphibians are sensitive to the thermal landscape. There was also complex environmental variation among thermally heterogeneous microhabitats used and available to herpetofauna. As perch T S increased, shade became more important, though shade sources may be less relevant. The methods and data obtained in this study can serve to establish baselines during seasonally and ecologically important periods and be used to monitor thermal ecology changes across time for herpetofauna or taxa with similar life history in aridland riparian systems. Although ectotherm thermoregulatory functions are undeniably complex, we recommend IRT as a rapid, noninvasive, and complementary tool to monitor cryptic thermal ecology in heterogeneous systems. Understanding species–environmental relationships and monitoring responses of wildlife across time can help guide more effective biodiversity conservation management strategies in a warming and drying world.


Introduction
Climate change in aridland ecosystems, such as the southwestern United States, is predicted to bring increased temperature and aridity while stressing water availability via reduced precipitation, elevated evapotranspiration rates, and anthropogenic demand (Archer and Predick 2008, Roby et al 2020, Warter et al 2023. Riparian zones are important corridors of biodiversity (Naiman et al 1993), especially in aridland ecosystems (Free et al 2015). Yet, riparian habitats and stream flow dynamics therein are vulnerable to 2. Material and methods 2.1. Study area Sabino Canyon Recreation Area (SCRA) resides in the southern foothills of the Santa Catalina Mountains in Tucson, Arizona, USA (Lazaroff 1993). The regional climate is semi-arid desert with bimodal seasonal precipitation peaks in winter/spring and summer, abutted by dry periods (McMahan et al 2021, Warter et al 2023. Our study area occurred in lower SCRA, adjacent Sabino Creek. At this location, stream flow is seasonally intermittent-receding and drying in late spring to early summer before recharging following summer monsoon precipitation (Lazaroff et al 2006, Blais et al 2023. Riparian habitat among the lower SCRA includes mesquite bosque and winter-deciduous woodlands among rocky, sandy substrates; Sonoran desertscrub (saguaro-palo verde communities) encompass the adjacent uplands (Lazaroff 1993). The heterogeneous microhabitat complexity in this system offers numerous resources to riparian fauna, including approximately 57 species of herpetofauna (Lazaroff et al 2006). Important in context for the region, contemporary ground water demands have negatively impacted Tucson-area watersheds (Bogan et al 2020, Bradley andColodner 2020). Tucson, including the Santa Catalina Mountains, has also experienced recent environmental extremes and disturbances, including near record heat and drought and the Bighorn Wildfire in 2020 (Wilder et al 2021); the latter negatively affected Sabino Creek (Blais et al 2023).

Surveys and sampling
Between February-April in 2020 and 2022, we performed visual encounter surveys (Doan 2016) along lower Sabino Creek's riparian habitat (32.3114 • N, −110.8105 • W, WGS84, elev = 822 masl). We began surveys at approximately 09:00 and worked sinuously upstream for ca. 1.8 km. We visually scanned surfaces of various microhabitats (e.g., detritus/debris piles, rocks, vegetation) adjacent (±50 m) the stream and with minimal disturbance, i.e., we maximized opportunity by covering ground and surface scanning rather than lifting rocks or rummaging through debris and litter (Catenazzi 2016). Surveys in 2020 were suspended after 8 April due to COVID-19 restrictions.
Our diel (morning) and seasonal (late winter-early spring) design aligns with early-season life history patterns of several species in the SCRA (Lazaroff et al 2006). For example, spring mornings are optimal for lizard activity (e.g., foraging, breeding behavior; Jones and Lovich 2009) and detecting perched canyon treefrogs (Hyla arenicolor; Preest et al 1992). In the summer, drastic environmental shifts (e.g., stream flow cessation, excessive temperature) can lead to behavioral or physiological changes by herpetofauna therein, such as aestivation strategies corresponding to hydric or thermal stress (Preest et al 1992, Jones and Lovich 2009, Griffis-Kyle et al 2018. Upon detection, we photographed herpetofauna with a DSLR digital camera (to aid in species identification) and a FLIR E8 infrared thermal camera (FLIR Systems, Wilsonville, Oregon) to simultaneously capture thermograms of individuals and the microhabitats that they perched upon, i.e., thermal niche (Goller et al 2014). The specifications of the E8 include 320 × 240 px ± 2 • C thermal resolution, FOV 45 • × 34 • , IFOV 2.6 mrad, and noise equivalent temperature difference sensitivity <0.06 • C. We set camera emissivity to 0.97 to match known ranges (0.95-0.98) for both amphibians and reptiles, respectively (Luna and Font 2013, Barroso et al 2016, Natchev et al 2022. To ensure adequate resolution of small-bodied animals against their microhabitat, we attempted to capture multiple IRT images at ca. 0.5-1 m distance at approximately 45 • angles to the flanks or dorsum of individuals (Playà-Montmany and Tattersall 2021; see figure 1). To neither induce stress or bias T B , we attempted to collect IRT data within 2 min, and track ⩽10 m distance if observer approach caused an individual to flee from its position; we captured thermograms of initial perch locations if fleeing occurred. By noninvasive design, we did not capture any herpetofauna nor measure body size. We noted (ad libitum) what the animal was doing upon first detection (e.g., basking, locomoting) but did not systematically sample behavior.

Microhabitat and environmental assessment
For each observation, we assessed microhabitat and environmental conditions (Mushinsky and McCoy 2016). We categorized microhabitat type (e.g., rock, woody debris), and vegetation height class within one meter; see table S1 for descriptions of qualitative variables. We used a Kestrel 5000 anemometer (Nielsen-Kellerman, Boothwyn, Pennsylvania) to record the following mean ambient conditions at ca. 1.5 m above surface: air temperature (T A , ±0.1 • C), barometric pressure (BP, ±0.1 millibars), relative humidity (rH, ±0.1%), and wind speed (±0.1 m s −1 ). We estimated percent overhead canopy density by averaging four directional readings from a spherical densiometer (Sprague and Bateman 2018); and estimated microhabitat shade to nearest 25% (Blais et al 2023). For every herpetofauna-used (i.e., observed) microhabitat, we subsequently (⩽10 min) acquired microhabitat and environmental data for a paired random (i.e., available) point (Row andBlouin-Demers 2006, Sprague andBateman 2018). Because of relatively small perceived space use and movement among commonly encountered SCRA herpetofauna, e.g., common side-blotched lizards (Uta stansburiana, Lazaroff et al 2006, Goller et al 2014, random points were ⩽10 m from paired used points; the distance and direction were determined using a random number generator. To capture environmental conditions of the broader area, we downloaded Daymet data (Thornton et al 2020) that encompassed a 1 km 2 grid centered on lower Sabino Creek (tile 11 015, https://daymet.ornl. gov/) for corresponding survey dates. These data included incident shortwave radiation flux density (srad, ±0.01 W m −2 ; Li et al 2021) and daily mean water vapor pressure (vpa, ±0.01 Pa; Weathers 1972).

Thermographic image data processing
Each IRT thermogram from the FLIR E8 contained a high resolution of per-pixel data (76 800 px ± 2% temperature uncertainty). We used line and polygon functions in FLIR Tools v.6.4 (FLIR Systems) to generate descriptive statistics (±0.1 • C) for individual T B and microhabitat perch surface (T S ), respectively (Goller et al 2014). For T B , we applied these functions across the head and central torso of individual's dorsal or side outlines to acquire mean, minimum, and maximum values. We cautiously excluded both potential radiation-biasing body edges and surrounding microhabitat surfaces (Goller et al 2014, Klein and Busby 2020, Schneider et al 2020 figure 1). To account for possible effects of regional heterothermy, we used mean values of T B for downstream analyses (Rowe et al 2020). This approach is due to working with small, skittish taxa under field conditions in which targeting certain body regions (e.g., eye, Barroso et al 2016) may not be easily photographed or adequately reflect T B (Playà-Montmany and Tattersall 2021). Because prior research (Luna and Font 2013, Goller et al 2014, Catenazzi 2016, Barroso et al 2020 found IRT to accurately capture internal body temperature, we infer that mean T B weighted across head and torso regions here proxies core T B of the individuals in this study. For microhabitat perch T S , we used polygon functions on surfaces immediately adjacent the specimen (⩽10 cm, see figure 1) but without including any part of the individual (Klein and Busby 2020). Mean perch T S were always estimated <1 m from individuals to reduce ambiguity from heterogeneous thermal mosaics among individual microhabitats (Goller et al 2014; see H. arenicolor thermogram in figure 1). We averaged multiple T S readings if surfaces were obscured or highly variable. We only retained observations with both T B and T S data for downstream modeling analyses.

Statistical analyses
Because it was not feasible to assess all types of microhabitats available (e.g., underground burrows, within complex refugia) in our noninvasive design, we followed Hyslop et al (2009) in considering patterns of surface-level microhabitat use by herpetofauna as a function of selected ecological factors. To test differences between used and randomly available plots, we performed mixed effect logistic regression modeling in the glmulti package (Calcagno 2020) via a 'glmer' function (Bates et al 2015). Predictor covariates included canopy density, vegetation height class, shade, microhabitat type, T S , and year; we allowed for covariate interaction and controlled for random effects of individual plots. We first tested covariates for multicollinearity (cutoff for omittance: r ⩾ 0.6). We used a corrected Akaike information criterion algorithm (∆AICc < 2, Venables andRipley 2013, Calcagno 2020) in glmulti to select optimal models. We followed Calcagno and de Mazancourt (2010) to weigh importance of individual covariates via inclusion in ⩾80% of models. We also explored linear relationships of mean T S to mean ambient T A per microhabitat type to assess thermal heterogeneity mosaics (Goller et al 2014).
We tested the difference in body temperature from perch temperature (∆T B − T S ) against mean T S per taxonomic family to inference the thermoregulatory processes of herpetofauna in this system (Goller et al 2014). Next, we used glmulti to perform multiple linear regression modeling to understand the extrinsic and intrinsic relationships between T B and the surrounding environment. Predictors included canopy density, BP, rH, srad, T A , T S , vegetation height, vpa, and wind; we added a quadratic term to T S as nonlinearity with temperature may exist (Woods et al 2015, Mushinsky andMcCoy 2016). We included species and activity level (i.e., active movement or inactive basking behavior, Taylor et al 2021) for taxon-specific and behavioral calibrations, respectively (Barroso et al 2016). We controlled for random effects of phylogeny (taxonomic Family) and year. We again used parameter weighting and ∆AICc < 2 to select optimal models. Prior to running T B models, we subset the data to include only species with >10 observations (i.e., Hyla arenicolor, Urosaurus ornatus, and U. stansburiana) to reduce bias from taxa with low sample size. Though body size should not impede the use of IRT to investigate thermal ecology (Barroso et al 2016), these three small taxa are similarly sized: snout-vent length ⩽64 mm and mass 2.6-8.1 g (Christian and Waldschmidt 1984, Preest et al 1992, Snyder and Hammerson 1993, Jones and Lovich 2009, Goller et al 2014. Although wetted skin surfaces may bias thermographic analyses (Schneider et al 2020), all individuals in this study appeared relatively dry, including H. arenicolor, which typically perch during the morning hours, often in direct sunlight (Preest et al 1992, Lazaroff et al 2006. Finally, we assessed how amphibian and reptile T B and ∆T B − T S trends changed across hourly intervals during spring mornings and early afternoons. We performed all statistical analyses in R v.4.1.1 (R Core Team 2021).
We accrued thermal data for 81 individuals (25 anurans, 55 lizards, and one snake, table 1). Three additional lizards escaped prior to obtaining thermal data. Sampling effort yielded 0.89 herpetofauna observations per person hour (3.2 ± SD 2.6 per person visit). We detected all individuals between 08:30 and 13:00. Behaviorally, herpetofauna most often exhibited little to no movement (i.e., basking/resting postures, 57.1%). Some phrynosomatid lizards displayed inter-and intraspecific territorially, but these were not quantified. Observer approach caused some lizards to flee whereas anurans mostly remained alert but immobile (i.e., crypsis posture but with eyes open).

Microhabitat associations
We most often encountered herpetofauna using rock microhabitats (n = 67, 81.7%), followed by debris (n = 8, 9.8%), open/low vegetation (n = 6, 7.3%), and one detection (1.2%) in woody microhabitats, respectively. More than half (53.9%) of used microhabitats were devoid of adjacent vegetation cover, such as unshaded rocks in the floodplain (see U. stansburiana in figure 1). Microhabitat surface temperatures were linked to ambient temperature, but overall thermal heterogeneity varied widely and by microhabitat type (linear model: F = 7.28, R 2 = 0.138, p < 0.001, figure 2(a), table S2). Three candidate models best characterized differences between used and available microhabitats (table 2(a)). After weighing covariates, the optimal representation included canopy, shade, and interactions of canopy-shade, canopy-vegetation height, and shade-T S (R 2 c = 0.28, ω = 0.245). After converting covariates to odds ratios, for each unit increase in shade-TS relationship, the odds of microhabitats being used were 1.26 times (95% confidence intervals: 1.02-1.55×) more likely than available, after accounting for effects of other terms ( figure 3(a)). There were trends that used microhabitats were under less dense canopy (p = 0.07), including when interacting with shade, albeit with some uncertainty as demonstrated by relatively wide confidence intervals (figures 3(b) and (d)). Also, there was some association that shadier microhabitats were more likely to be used when shade originated from high (>2 m) or low vegetation (<1 m, figure 3(c)). However, there were again wide confidence intervals across these interaction levels. We interpret these results as herpetofauna were somewhat more likely to use shadier microhabitats as surface temperature increased but the source of shade (e.g., vegetation versus sun angle in riparian canyon habitat) may be less relevant due to the variability in the data.
After excluding species with small sample size, three competing models best explained herpetofauna T B ; each included canopy, vegetation height, T S , and species but differed by wind and activity level (table 2(b)). After accounting for model weight and covariate importance threshold, the best representative model included T S and its quadratic term, canopy, height, wind, and species (R 2 c = 0.83). For each unit increase in T S and T S 2 , herpetofauna T B increased 1.51 • C (CI: 0.66-2.37) and −0.02 (CI: 0.00-0.03), respectively, after accounting for effects of other terms (figure 5). These asymptotic effects influenced the three focal taxa equivalently, but the two phrynosomatid lizards had T B greater than H. arenicolor. There were some trends that T B was associated with increasing wind (p = 0.10) and declining canopy density (p = 0.06) but with uncertainties, i.e., wide confidence intervals crossing zero (figure S1).

Discussion
As global biodiversity faces numerous challenges linked to Anthropogenic climate change (Newbold et al 2015, Trisos et al 2020, it is important to understand species-environment interactions and monitor how taxa, such as ectotherms, respond to environmental change (Huey 1991  Model selection based on change corrected Akaike information criterion (AICc < 2) and weights (ω) via the R package glmulti (Calcagno 2020   that microhabitats used by herpetofauna were linked to a surface temperature-shade interaction. Perch surface temperature was also the best predictor of body temperature, and this relationship approached equilibrium as microhabitats warmed. Our noninvasive field application of IRT acquired baseline proxies of relative thermal relationships between syntopic herpetofauna and their immediate surrounding environment during important springtime activity, a valuable step for assessing impacts of global climate change (Taylor et al 2021).

Thermal heterogeneity of microhabitats across an aridland riparian landscape
Microhabitat assessment can provide vital data that is potentially missed at macro scales (Mushinsky andMcCoy 2016, Campobello et al 2017). Thermal gradients can vary widely within and among landscape microhabitats (Goller et al 2014). Through IRT and environmental analyses, we addressed our first objective to discover a diverse thermal mosaic across various microhabitat types among the riparian landscape in the SCRA. Our top performing models of microhabitat association revealed an interacting relationship between shade and surface temperature. As perches warmed, those that were more shaded were more likely to be used. Used microhabitats tended to have less dense overstory canopy than those randomly available. Dense overstory may limit the amount of solar radiation reaching terrestrial surface microhabitats below or it could reflect behavioral attributes. For example, some lizards must balance thermoregulatory needs with predation risks when occupying thermally optimal perches (Broeckhoven andle Fras Nortier Mouton 2015, Sannolo et al 2019), and denser canopy overstory may obscure visual vantagepoints to detect predators, such as birds. There were also some trends of use towards high (>2 m, e.g., trees) and low (<1 m, e.g., forbaceous plants) vegetation classes (see figure 3(c)) but there was uncertainty in these data. This could be due to our sampling window prior to full leaf-out of riparian deciduous trees or that there remains unquantified variation in the system. A more plausible explanation could be that the existence of shade is more important than the source of it. Part of our sampling area occurred in a riparian canyon at times preceding the solar zenith angle and thus yielding natural shade, especially along canyon walls. The herpetofauna examined here may be selecting for perches where extent of shade offers certain thermal advantages, such as a refuge for regulating exposure to direct sunlight. Taken together, the various complex interactions in our top performing models allude that microhabitat associations by aridland riparian herpetofauna appear equally complex and drivers for such may extend beyond what we measured (e.g., behavioral, microhabitat fidelity, perch height, see Mahrt 1998, Goller et al 2014.

Influences on herpetofauna body temperature
Environmental temperatures can be strongly correlated with body temperatures in aridland lizards (Gadsden andEstrada-Rodríguez 2007, Woolrich-Piña et al 2012). For our remaining objectives, we found a relationship between herpetofauna T B and surface T S . Body temperature approached equilibrium with increasing T S . Though, lizards expressed more T B variability and range in relation to T S than anurans (see figure 2(b)). This may in part be due to aridland lizards possessing greater flexibility to warming conditions (Goller et al 2014, Griffis-Kyle et al 2018. Desert riparian amphibians, such as H. arenicolor, may be more sensitive to their environment (Preest et al 1992, Snyder andHammerson 1993), especially as the region warms and dries (Archer and Predick 2008, Roby et al 2020, Warter et al 2023. Temporally, herpetofauna in the SCRA were as warm at mid-morning (i.e., 09:00 onward) than they were in early-afternoon (e.g., 12:00-13:00). This differs from diel × seasonal warming patterns observed elsewhere. In northern Utah, for example, U. stansburiana were cooler in summer mornings and gradually warmed during afternoons (Goller et al 2014). We note that U. stansburiana body-perch temperature equilibrium during summer climate in Utah (ca. 39.0 • C, see figure 6(b) in Goller et al 2014) were proximal to what we observed during spring in Tucson (38.1 • C). Although full diel or seasonal T B limits were not captured by our study, we reiterate that we maximized opportunity by focusing our thermal ecology sampling temporally to when taxa are most detectable (Taylor et al 2021), such as spring mornings (Huey et al 1989, Chukwuka et al 2021. In the Sonoran Desert ecoregion, it becomes more challenging to detect herpetofauna diurnally during hotter parts of the year (Lazaroff et al 2006, Jones andLovich 2009). The surface activity, microhabitat use, and thermal relationships of herpetofauna during spring mornings in the SCRA therein proxy as a function of selected ecological factors (Hyslop et al 2009).
Through mixed effect modeling, we found surface temperature of perches (T S ) to be the most reliable predictor of body temperature for SCRA herpetofauna. Perch temperature may be a proxy of heat conduction with substrates, but we did not exclusively test the various heat sources that thermoregulate ectotherms (Ortega et al 2019). The complexity of parameters identified in our optimal T B model (see figure  S1, table 2(b)), however, likely relates to the thermoregulatory intricacies of ectotherms (Taylor et al 2021). It is noteworthy that ambient parameters (e.g., BP, rH, T A ) were not identified among top performing models. These results suggest that ambient parameters alone are likely suboptimal for inferring thermal ecology relationships and should be used in unison with thermal-sensing devices (Bakken and Angilletta 2014). Beyond T S , extrinsic factors that influence body temperature may be fluid and only partly explain herpetofauna thermoregulation in aridland riparian ecosystems.
Relationships between intrinsic and extrinsic factors can influence thermoregulatory behaviors throughout a day and season (Huey et al 1989, Caetano et al 2020. Herpetofauna often balance thermoregulation through active regulatory behaviors, such as posturing or shuttling within or between microhabitats (i.e., abiotic conditions) when they reach the bounds of their thermal tolerances (Huey 1991, Pianka et al 2017, Taylor et al 2021. Thermal avoidance behaviors may be deployed for reprieve from extremes (Huey et al 1989, Chukwuka et al 2021, especially during hot summers in aridland ecosystems Porter 2004, Jones andLovich 2009). Because the herpetofauna in this study are typically surface active or surface visible during spring mornings, and activity level (i.e., rest vs. locomote) appeared to have low weight among top performing T B models, it suggests that herpetofauna may not have reached critical thermal maxima levels within the environmental conditions assessed. This is supported by lizards frequently being observed in unshaded microhabitats, though we note canyon treefrog positions ranged from full sun exposure (e.g., exposed rocks) to full shade (e.g., canyon wall crevices). We sometimes observed U. ornatus or U. stansburiana occupying the top of the tallest structure (e.g., rock piles) in their immediate vicinity-structure characteristics are important for both species (Mahrt 1998, McElroy et al 2007. Could balancing thermoregulation with a certain degree of microhabitat site fidelity exist for some species in aridland riparian ecosystems? If so, to what extent will climate-induced increases in aridity and temperature stress this balance? For territorial lizards, it may be less costly to defend optimal thermal territory than to seek it out in complex, changing environments. Mature adults also may outcompete juveniles for optimal thermal refuges (Chukwuka et al 2021). Additional studies of thermal phenology patterns (e.g., daily, seasonally) and systematic behavior assessments (e.g., shuttling, time spent foraging), including influence from competition and territoriality, could elucidate deeper understanding of thermoregulatory patterns and species responses to environmental conditions (Goller et al 2014, Rowe et al 2020, Taylor et al 2021.

IRT applications and management recommendations
Advances in thermographic tools, such as IRT, have shown great promise in ecological studies (McCafferty et al 2021). Added benefits of IRT devices over single-read units or operative ground-penetrating devices include simultaneous real time captures of thermal mosaics, hundreds or thousands of data points per thermogram, and some relief of logistical or spatiotemporal resource demands. As IRT resolution and application increases, researchers can assess environmental relationships and microhabitat associations by animals in novel ways (Ganow et al 2015, Chukwuka et al 2021. Despite the practicality of handheld IRT devices, there are certain limitations in application and sources of uncertainty (see Chukwuka et al 2019, McCafferty et al 2021, Souza-Junior and de Queiroz 2022. For example, it may be challenging or impractical for wildlife managers and macroecologists to continuously adjust and calibrate equipment for various thermodynamic properties (e.g., per subject blackbody calibration, radiometry and spectrophotometry gear; Tattersall 2016) in field conditions, especially when working noninvasively with skittish or scurrying animals (Playà-Montmany and Tattersall 2021). A broader assessment of advantages and limitations of IRT use in thermal ecology studies have been discussed elsewhere (see Bakken and Angilletta 2014, McCafferty et al 2021, Playà-Montmany and Tattersall 2021. In this study, we exercised precautions to overcome regional heterothermy and other sources of error. We captured IRT at relatively short distances to animals (<1 m) to reduce errors related to temperature or atmospheric conditions (Playà-Montmany and Tattersall 2021, Taylor et al 2021, analyzed mean thermal data values from multi-pixel shape functions, across animal heads and torsos, rather than rely upon single-pixel 'spot' values (Rowe et al 2020), and controlled for other potential sources of variation, such as srad (Weathers 1972), vpa (Li et al 2021), and behavior (Mazzamuto et al 2023), during statistical modeling (Tattersall 2016). With such, we derived relative T B to T S relationships for multiple taxa, which may be more practical for wildlife management purposes than precise core temperature. The thermal heterogeneity across microhabitats detected in this study further supports that single-read instruments or ambient conditions alone may not adequately quantify the thermal mosaics available to species.
The application of IRT to investigate thermal ecology of herpetofauna comes at a key time. Climate change and its related effects (e.g., habitat loss, disease transmission) are directly linked to declines in biodiversity (Leigh andDatry 2017, Taylor et al 2021), which can have cascading effects in subsequent food webs (Zipkin et al 2020). Climate-linked disturbances can adversely affect herpetofauna populations in arid ecosystems (Griffis-Kyle et al 2018, Bateman and Riddle 2020), including the SCRA (Lazaroff et al 2006, Blais et al 2023. If climate change drives aridland riparian herpetofauna towards their thermal maxima earlier in the day or active season, would such shifts be physiologically or phenotypically disadvantageous? Although increased temperature may provide thermal opportunities for certain aridland lizards (Goller et al 2014, Griffis-Kyle et al 2018, climate-induced changes to riparian landscapes and heightened risks of exceeding thermal maxima may ultimately do more harm than good for aridland riparian herpetofauna (Kearney et al 2009, Bateman andRiddle 2020). This is also true for desert amphibians that are especially vulnerable to climate change (Griffis-Kyle 2016, Griffis-Kyle et al 2018), including specialist taxa or those that move less for thermoregulation, like H. arenicolor (Preest et al 1992, Snyder and Hammerson 1993, Davis et al 2015, Inman et al 2023. When thermal suitability declines due to altered environments, species either adjust through behavioral or physiological plasticity or experience demographic collapse (Bickford et al 2010, Sinervo et al 2010. The objective for herpetofauna may be less about getting warm and more about staying cool (Kearney et al 2009), for which their adaptive thermoregulatory capacity may be outpaced by rapidly changing systems (Griffis-Kyle et al 2018, Bodensteiner et al 2021). These concerns highlight the importance of gathering thermal ecology data and monitoring changes.
Effective conservation planning and management strategies to combat the effects of climate change can be bolstered by first identifying adaptive physiological potential and climate vulnerability of species (Griffis-Kyle 2016, Taylor et al 2021), and then linking that information to model important microhabitats and microclimates that could best serve as climate refuges (Scheffers et al 2014, Suggitt et al 2018. The methods and data obtained in this study can serve to establish baselines during seasonally and ecologically important periods and monitor thermal ecology changes across time for aridland riparian herpetofauna. The SCRA and greater Sonoran Desert ecoregion are in the midst of prolonged drought and extreme environmental disturbances (Bradley and Colodner 2020, McMahan et al 2021, Wilder et al 2021, Blais et al 2023. Considering how species such as aridland riparian herpetofauna respond to environmental changes and disturbances, maintenance of important riparian habitat heterogeneity and connectivity will be invaluable as thermal refuges for herpetofauna (Thompson et al 2018).

Conclusion
In this study, we examined the thermal ecology of aridland riparian herpetofauna. We used IRT and environmental assessments to address our objectives, including modeling the environmental compositions of microhabitats used or available to herpetofauna as well as assessing the relationships of thermal and environmental parameters on body temperature. We found that microhabitat perch surface temperature was most strongly linked to body temperature of herpetofauna and that a heterogenous thermal mosaic of microhabitats were available to species. In the SCRA, herpetofauna were as warm by mid-morning as they were in early afternoons. As the climate warms and dries in aridland systems such as the American Southwest, how will ectothermic taxa respond to changes? The extent to which environmental change will stress species and increase vulnerabilities is of importance for conservation practitioners and biodiversity managers (Griffis-Kyle et al 2018, Taylor et al 2021). Fortunately, both technological advances and reducing costs should enable IRT as a reliable tool for researchers and managers to simultaneously obtain body and perch surface temperature of ectothermic taxa in real time as well as monitor thermal heterogeneity across landscapes. Coupled with other behavioral and demographic assessments, long-term thermographic research may resolve our understanding of how aridland riparian herpetofauna-and other ectothermic taxa-may respond to altered and changing systems. Monitoring such species-environmental relationships across time will be paramount for managing and conserving biodiversity of aridland riparian ecosystems in a warming and drying world.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary information files).