WIXLIB

36V. J. Alaasam, J. E. Keehn, A. M. Durso, S. S. French, and C. R. Feldmannoise from generators and blade movements, habitat alteration(e.g., disturbance and fragmentation), and increased human activity (Orloff and Flannery 1992; Lovich and Daniels 2000; Rabinet al. 2006; Arana et al. 2015; Perrow 2017). Wind farms have alsobeen shown to alter community composition and structure, decrease native species richness and diversity, increase exotic speciesrichness, and potentially homogenize ecological communities(Pearce-Higgins et al. 2009; Keehn and Feldman 2018a). Thesetypes of biotic and abiotic changes at wind farms have the potential to increase physiological stress for wildlife (Crino et al.2011; Janin et al. 2011; Blickley et al. 2012; Tennessen et al. 2014;Agnew et al. 2016; French et al. 2017; Injaian et al. 2018; Kleist et al.2018; Łopucki et al. 2018; Thaker et al. 2018), yet few studies havemeasured bioindicators of physiological stress in the terrestrialspecies that inhabit wind farms (Mikolajczak et al. 2013; Agnewet al. 2016; Łopucki et al. 2018; Thaker et al. 2018; Klich et al. 2020).This is important because physiological stress has fitness andsurvival consequences, including immunosuppression and impaired reproduction (Calow and Forbes 1998; Hofer and East1998; Möstl and Palme 2002). Thus, environmental disturbancesthat increase stress may also increase susceptibility to disease orparasitism and affect survival or recruitment (Fitze et al. 2004;Lafferty and Kuris 2005; Martin 2009). Alternatively, wind farmsmay actually decrease physiological stress for some animals ifpublic access to developed areas is restricted (Lovich et al. 2011;Agha et al. 2015), if wind farms harbor fewer predators (Winderet al. 2014; Keehn and Feldman 2018b; Thaker et al. 2018), or ifscrap building materials and storage areas provide additional refuges (Lovich and Daniels 2000). Given the tremendous carbon,water, and economic advantages of wind energy (US Departmentof Energy 2015), understanding physiological responses of localtaxa to wind farms is essential for developing strategies to mitigate the potential ecological costs of wind farms.Oxidative balance is a well-studied physiological biomarkerof both the capacity of organisms to cope with environmentaldisruption and the deleterious effects of disruption (Cookeet al. 2013; Beaulieu and Costantini 2014). Oxidative balancecan be quantified as the blood concentration of reactive oxygenmetabolites (dROMs) relative to the neutralizing capacity of bloodantioxidants (Costantini 2014). Overproduction of dROMs exceeding the body’s natural antioxidant capacity (OXY; e.g., as aresult of increased energetic demands) causes oxidative stress andeventual damage to tissues and immune cells (Costantini et al.2011). Oxidative balance is a particularly useful metric for understanding the effects of chronic physiological stress, as it oftenincreases with chronic glucocorticoid elevation and immune response activation (Costantini et al. 2011) and can sometimespredict survival and reproduction (Bize et al. 2008; Monaghan et al.2009; Freeman-Gallant et al. 2011; Saino et al. 2011; Losdat et al.2013; Costantini et al. 2016). Additionally, oxidative balance is ageneral physiological response that can reflect a variety of disturbances (Isaksson 2010, 2015), including air pollution (Lodoviciand Bigagli 2011), noise pollution (Münzel et al. 2017, 2018),urbanization (Lucas and French 2012), predation (Guerra et al.2013), and infection rates (Isaksson et al. 2013), making it a valuable tool for assessing the compounding effects of multiple stressorsthat are often associated with human-induced habitat change(Beaulieu and Costantini 2014). Thus, oxidative balance is useful for investigating the complex relationships between anthropogenic disturbance, physiological stress, and disease. For example,disturbance may increase physiological demands and consequentlyreduce immunocompetency or directly increase prevalence andtransmission of pathogens (Patz et al. 2000; Daszak et al. 2001;McCallum and Dobson 2002), activating immune responses, whichcan, in turn, increase physiological stress.We examined the influence of wind farms on oxidative stressand an associated aspect of disease resistance (parasite load) in acommon desert reptile. To date, relatively few studies have investigated the effects of wind farms on reptiles (Santos et al. 2010;Lovich et al. 2011; Ennen et al. 2012; Agha et al. 2015; Keehn andFeldman 2018a, 2018b; Thaker et al. 2018; Keehn et al. 2019).Nevertheless, reptiles are essential components of many ecologicalcommunities, serving diverse roles in nutrient cycling and trophicinteractions and sometimes serving as pollinators, seed dispersers,and ecosystem engineers (Spiller and Schoener 1998; Bouchardand Bjorndal 2000; Godínez-Álvarez 2004; Dorcas et al. 2012;Kinlaw and Grasmueck 2012; Cortes-Gomez et al. 2015; de Miranda 2017; Reiserer et al. 2018). We focused our efforts on thecommon side-blotched lizard (Uta stansburiana), an abundant andwidespread inhabitant of western North American deserts (Tinkle1967; Wilson 1991; Stebbins 2003), where wind farms have recently been established. Side-blotched lizards are ideal for studying the effects of wind farms because they are relatively easy tocapture in high numbers, and they have limited dispersal and highsite fidelity (Tinkle 1967; Wilson 1991; Doughty and Sinervo 1994;Keehn et al. 2019), such that individuals experience narrow localenvironments over their lifetime and are strongly influenced bylocalized disturbance (Tinkle 1967; Tanner and Hopkin 1972; Lucasand French 2012).Here, we investigated oxidative balance and external parasiteload in side-blotched lizards at three wind farms and three reference sites in the San Gorgonio Pass wind resource area (SGWRA)in the Mojave Desert in Southern California. Previous work on thissystem has shown that wind farms in the SGWRA have higherlevels of noise pollution and human disturbance, as well as alteredbiotic communities that include reduced native plant richness anddiversity and increased nonnative plant richness compared withundeveloped reference sites (Keehn and Feldman 2018a). Suchchanges in plant community composition may be associated withshifts in the richness and abundance of the arthropods these lizardsrely on and might also impact behavior, foraging, thermoregulation, physiology, and other aspects of lizard life history (Dursoet al. 2020). Indeed, side-blotched lizards display slightly lowersurvival at these wind farm sites compared with nearby referencesites (Keehn et al. 2019). However, wind turbine sites also possessfewer avian predators of lizards (Keehn and Feldman 2018a), andpredator attack rates on side-blotched lizards appear lower onwind farms, resulting in less wary side-blotched lizards at thesesites (Keehn and Feldman 2018b). Therefore, predicting the impacts of wind farms on the health and physiology of local lizardsis not straightforward. On one hand, increases in noise, disturbance, and disruptions to native communities (and possible resource

Lizard Stress and Ectoparasitism at Wind Farmsavailability and use) may increase signs of oxidative stress andparasite loads in lizards on wind farms. On the other hand, reductions in predation pressure (and perceived threats) may dampenstress levels (Clinchy et al. 2012; Graham et al. 2017; Thaker et al.2018) and attendant parasitism rates in lizards at wind farms. Byincreasing our understanding of the multifaceted consequencesof wind energy development on the health of local species, weanticipate that this study will inform management decisions tosupport wildlife conservation goals in renewable energy development. Such work is vital as renewable energy expands as a globalindustry.MethodsStudy SpeciesUta stansburiana is a short-lived, oviparous lizard that inhabitsarid parts of the western United States in high abundances (Tinkle1967; Stebbins 2003). They can be found across a wide range ofterrain types, including hillsides, rocky outcrops, sand dunes,grasslands, and sagebrush flats (Tinkle 1967), but they are mostcommon in open landscapes with proximity to rocks and shrubsfor basking and shelter (Tinkle 1967; Stebbins 2003). Mean lifeexpectancy of side-blotched lizards is 1–1.5 yr at southern latitudes,so population turnover is roughly annual in southern deserts(Tinkle 1967; Parker and Pianka 1975; Wilson 1991; Sinervo et al.2000b; Zani 2005; Keehn et al. 2019). Mating occurs in the spring,and most eggs hatch in late summer (Parker and Pianka 1975;Doughty and Sinervo 1994; Stebbins 2003).The trombiculid mite, Neotrombicula, is the most prevalentexternal parasite of U. stansburiana (Goldberg and Bursey1991b) and infects a diverse array of other western vertebrates(Walters et al. 2011). These chiggers occur mostly on the upper37eyelids, gular folds, and ventral surfaces and cause necrosis of theepidermis at the point of attachment. Infestations last approximately 1 wk and result in integumental lesions and a diffuse inflammatory response (Goldberg and Bursey 1991a). Infectionsoccur on both male and female hosts at similar densities and arehighest between February and June but persist in lower densitiesthrough November (Goldberg and Bursey 1991b).Study SitesWe examined lizards and their external parasites from three windfarm sites and three undeveloped references sites in the SGWRA ofthe Mojave Desert in Southern California (fig. 1). The SGWRA wasfirst developed for wind energy production in the mid-1980s and iscurrently one of the largest wind production regions in the UnitedStates (Pasqualetti 2001), covering more than 18,000 ha. We selected three wind farm sites (Painted Hills, Mesa, and MountainView) and three reference sites (Mission Preserve, Mesa Control,Mountain View Control) with comparable abiotic characteristics(e.g., elevation, slope, substrates) and biotic communities (table 1).All sites are also relatively disturbed, with trails, dirt roads, andhuman visitation by maintenance workers at wind farms andrecreators at reference sites (located on public lands). Nevertheless,wind farms have greater levels of disturbance overall, with a higherdensity of roads and human traffic, as well as introduced plantspecies (Keehn and Feldman 2018a).Field and Live Animal MethodsWe collected lizards (n p 180) from our six study sites during a6-d period (June 27–July 2, 2015), visiting two sites per day andsampling each site twice on different days (table S1; tables S1,Figure 1. A, Male common side-blotched lizard (Uta stansburiana) displaying breeding colors. B, Close-up view of tail base of the same lizard,showing trombiculid mite (Neotrombicula) infestation; mites (bright red) are embedded at the base of the scales that encircle the tail. C, Six field siteslocated in the San Gorgonio Pass wind resource area in the Mojave Desert in Southern California. Wind farm sites: Painted Hills (PH), Mesa (ME), andMountain View (MV). Reference sites without wind turbines: Mission Preserve (MI), Mesa Control (MC), and Mountain View Control (SV). Weselected wind farm sites and reference sites that were generally similar in abiotic and biotic characteristics (table 1). Photo credits: R. W. Hansen.

22.840.224.229.1 5 15.159.926.549.945.7 5 24.233.9385, 2116.620333.9441, 2116.675733.9126, 2116.6277Disturbanceindexa34.0012, 2116.604733.9614, 2116.676633.8924, 2116.6788Latitude, longitude66.447.960.558.3 5 9.545.540.743.343.2 5 2.4Avg noise levels(dBA)466.3715.1648.0609.8 5 75.5649.2468.8500.9539.6 5 70.2Elevation(m)34.146.838.839.9 5 12.735.235.527.532.7 5 6.0Avg annual rainfall(cm)22,23,21,66,26,22,25,73,021021112518n 555.37.84.40.85.63.86.88.83Avg mass 5 5552.282.271.532.642.262.643.192.77Avg SVL 5 SD(mm)Note. Sample size and important physical measures (mass and body size reported as snout-vent length [SVL]) of the lizards (Uta stansburiana) examined from these sites are also given (additional information and rawdata reported in table S1, available online). MI p Mission Preserve; MC p Mesa Control; SV p Mountain View Control; PH p Painted Hills; ME p Mesa; MV p Mountain View; avg p average.aTotal length of road (weighted by 0.5 for footpath, 1.0 for two track, 3.0 for residential, 5.0 for secondary, and 10.0 for highway) within a 1,000-m radius from plant richness study plots; digitized in ArcMap from 1-mresolution imagery from the California National Agricultural Imagery Program. See Keehn and Feldman (2018a) for further details.Reference:MIMCSVAvg 5 SDWind farm:PHMEMVAvg 5 SDStudy siteTable 1: Study sites in the San Gorgonio Pass wind resource area in the Mojave Desert in Southern California along with location, disturbance levels, and some important abioticand climate characteristics

Lizard Stress and Ectoparasitism at Wind FarmsS2 are available online). Using a handheld “lizard loop,” wecaptured adult lizards, recorded exact capture locations for eachindividual with a GPS receiver (Garmin), quickly labeled animalswith a unique code, then placed them in individual cloth bags fortransport back to a field station (Boyd Deep Canyon DesertResearch Center, University of California Reserve). At the fieldstation, we measured body size (snout-vent length [SVL]), taillength, and mass. We noted whether individuals had a recent tailloss because tail loss may decrease immunocompetence (Kuo et al.2013; Argaez et al. 2018). We also recorded sex, and if female, wepalped the abdominal region by hand to record the number ofyolked follicles (1–4), because reproductive investment can influence stress and immunocompetence. Finally, we counted thenumber of external parasites for each individual using a handheldlens (#5, #10, and #15).Following these field efforts, we transported animals toUtah State University to obtain measures of oxidative balance.We transported lizards on June 29, 2015 (after sampling all sixsites once), and again on July 2, 2015 (again after visiting allsites). Within an hour of arrival, we collected blood samplesfrom all individuals via retro-orbital sinus using a heparinizedcapillary tube. We immediately placed blood samples on ice,separated blood plasma from cells via centrifugation, and storedsamples at 2207C. Note that lizards were held in captivity (in a coolroom at the field station) for 1–3 d before transport and bloodsampling. However, acute stress of collection and short-termcaptivity should have no bearing on our measures of oxidativebalance (described below), which are markers of long-term oxidative damage (Costantini et al. 2011). Research demonstrates significant increases in oxidative damage following only prolongedglucocorticoid treatment or stress exposure (i.e., multiple weeks)in most vertebrates, including lizards (Costantini et al. 2011;Voituron et al. 2017).On completion of the study, we humanely euthanized all animals and preserved them as voucher specimens in the herpetologycollection of the University of Nevada, Reno, Museum of NaturalHistory (table S1). We followed Institutional Animal Care andUse Committee protocols (UNR 00586; USU 2068) for the safecare, handling, and use of live animals in this study.Oxidative StressTo measure dROMs, we used a kit (Diacron, Grosseto, Italy) thatdetects the level of hydroperoxides, which are dROMs that signalboth protein and lipid oxidative damage (Costantini and Dell’Omo2006; Lucas and French 2012). We diluted 5 µL of blood plasmainto 100 µL of the provided acidic buffered solution and followedthe end point mode protocol, with modifications for use on a 96well microplate. To measure OXY, we used an OXY-adsorbent test(Diacron, Grosseto, Italy), which measures blood antioxidantbarrier by quantifying the oxidant action of hypochlorous acid(HClO; Vassalle et al. 2008). Briefly, we diluted 2 µL of bloodplasma in 100 µL of distilled water and followed manufacturerprotocol but modified for microplate use. We then mixed a 5-µLsubsample of diluted plasma with 100 µL of the HClO solutionprovided. All samples were run in singlicate. Interassay variation39was calculated using six replicate standards as the standard deviation of the standards divided by the average. Interassay variationwas 4.6% and 2.3% for dROMs and OXY, respectively.AnalysesA biologically relevant assessment of oxidative stress includesboth the levels of circulating dROMs and the absorbancecapacity of the blood (OXY) within individuals, so we used anintegrative index of oxidative stress to combine these complementary measures. We standardized dROM and OXY usingthe formula xs p (x 2 m) SD, where xs represents a standardized value, x represents the corresponding original value,and m and SD represent the mean and standard deviation,respectively. We then consider the oxidative stress index as thedifference between the standardized dROM and OXY values foreach individual, such that higher values of our index correspondto a greater differential between dROM and OXY (Vassalle et al.2008).We calculated body condition for each individual using a scaledmass index (Mi) and the equation M i p M # (SVL0 SVL)b ,where M is body mass, SVL0 is the mean SVL of the population (allsamples), and b is calculated as the slope from an ordinary leastsquares regression of body mass on SVL divided by Pearson’scorrelation coefficient (Peig and Green 2009).We conducted all statistical analyses in R version 3.5.3 (RDevelopment Core Team 2019) and assessed for multicollinearityof all variables using Pearson’s product-moment correlation (table S2). To test whether parasite loads varied between wind farmand reference sites, we used a generalized linear mixed model witha Poisson distribution, employing the lme4 package (Bates et al.2015). We included fixed effects of sex, body condition, SVL, sitetype (wind farm or reference), and tail loss (binomial yes/no),because recent tail loss and active tail regeneration may causephysiological stress or decrease immunocompetence (Kuo et al.2013; Argaez et al. 2018). We also tested for an effect of oxidativestress because environmental conditions that increase oxidativestress may simultaneously compromise immune function and result in higher levels of parasites. We compared models of increasing complexity by including interaction terms and selected a finalmodel using Akaike information criterion (AIC) comparison anda maximum likelihood approach (table 2). Unique site ID was included as a random effect to control for site effects in all models.The lsmeans package in R (Lenth 2016) was employed to obtainTukey-adjusted pairwise comparisons of significant predictors.To test whether oxidative stress varied between wind farm andreference sites, we used a generalized linear mixed model employing the lme4 package (Bates et al. 2015). As in the previousmodel, we included fixed effects of sex, body condition, SVL, sitetype (wind farm or reference), and tail loss. We also include parasiteload as a predictor because parasitic infection is likely to result inupregulation of an immune response and could explain variance inoxidative stress (Costantini and Møller 2009; Spence et al. 2017).In a second model, we limited our samples to females and includedthe number of yolked follicles (an estimate of clutch size) as apredictor because reproductive investment may also impose a

40V. J. Alaasam, J. E. Keehn, A. M. Durso, S. S. French, and C. R. FeldmanTable 2: Comparison of generalized linear mixed models to predict ectoparasite load in Uta stansburianausing maximum likelihoodModel, factorParasites site type 1 sex 1 body cond 1 SVL 1 tail lossSite typeSexSVLBody condTail lossParasites site type 1 sex 1 body cond 1 SVL 1 tail loss 1 stressSite typeSexSVLBody condTail lossStressParasites site type 1 sex 1 body cond 1 SVL 1 tail loss 1 stress 1site type # sexSite typeSexSVLBody condTail lossStressSite type # 001**.011*.631AICc3,082.1592,668.6412,670.674Note. All models include site as a random effect. All models include fixed effects of site type (wind or reference), sex (male or female), snout-vent length (SVL),body condition (cond), and tail loss. Stress was calculated as a combined metric of reactive oxygen metabolites concentration and neutralizing capacity ofantioxidants in blood plasma. AICc p corrected Akaike information criterion.*P ! 0.05.**P ! 0.01.***P ! 0.001.trade-off with self-maintenance and affect immunocompetenceor susceptibility to physiological stress (Smith and French 2017).We also used generalized linear models to investigate whethervariation in oxidative stress (i.e., stress index, as described above)was driven specifically by dROMs or OXY by testing raw values ofdROMs (mg H2O2/dL) and OXY (HClO/mL). We compared models, keeping fixed effects constant but increasing complexity byincluding interaction terms, and selected a final model using AICcomparison and a maximum likelihood approach (table 3). Uniquesite ID was included as a random effect to control for site effectsin all models. The lsmeans package in R (Lenth 2016) was employed to obtain Tukey-adjusted pairwise comparisons of significant predictors.ResultsAcross all sites, we caught more females than males, but theproportions of males and females caught at reference sites andwind farms were similar; at wind farm sites, we caught 66 females and 21 males, and at reference sites, we caught 73 femalesand 18 males. Males were 1:47 5 0:16 g heavier (mean difference 5SE; t p 9:208, P 0:001) and 4:50 5 0:25 mm longer (meandifference SVL 5 SE; t p 8:055, P 0:001) than females acrosssites. There was no difference in body condition between sexes(t p 0:022, P p 0:983). There was no difference in individualmass (t p 20:835, P p 0:405), SVL (t p 20:740, P p 0:460),or body condition (t p 20:226, P p 0:821) between wind farmsand reference sites. Summary statistics of mass and length of specimens at each site are included in table 1 (additional raw dataprovided in table S1). Collinearity of fixed effect variables is reported in table S2.Our best-performing model of parasite load included no interaction terms (table 2). We found a significant effect of site typeon parasite load (table 4; fig. 2A). Lizards at wind farm sites hadlower parasite loads (mean 5 SE p 8:18 5 2:69) compared withreference sites (mean 5 SE p 47:89 5 15:58). There was a positive correlation between SVL and parasite load (b p 0:062,SE p 0:009, 95% confidence interval [CI] p 0.045–0.079, P 0:001). There were also significant effects of oxidative stress andtail loss on parasite load but no effects of sex or body condition(table 4).Our best-performing model of oxidative stress included aninteraction of site type and parasite load (table 3). There wereno effects of site type (wind farm or reference), tail loss, bodycondition, or parasite load on oxidative stress in either sex(table 5; fig. 2B). There was a significant effect of sex (table 5),

Lizard Stress and Ectoparasitism at Wind Farms41Table 3: Comparison of generalized linear mixed models to predict oxidative stress levels in Uta stansburiana usingmaximum likelihoodModel, factorStress site type 1 parasites 1 sex 1 body cond 1 SVL 1 tail lossSite typeParasitesSexSVLBody condTail lossStress site type 1 parasites 1 sex 1 body cond 1 SVL 1 tail loss 1site type # parasitesSite typeParasitesSexSVLBody condTail lossSite type # parasitesStress site type 1 parasites 1 sex 1 body cond 1 SVL 1 tail loss 1site type # parasites # sexSite typeParasitesSexSVLBody condTail lossSite type # parasitesSite type # sexParasites # sexSite type # parasites # . All models include site as a random effect. All models include fixed effects of site type (wind or reference), parasite load, sex (male or female), snout-ventlength (SVL), body condition (cond), and tail loss. Stress was calculated as a combined metric of reactive oxygen metabolites concentration and neutralizing capacityof antioxidants in blood plasma. AICc p corrected Akaike information criterion.*P ! 0.05.**P ! 0.01.where females had higher oxidative stress than males (meandifference p 0:356, SE p 0:169, P p 0:037), but this effectemerged in only two of our three top model

erence sites in the San Gorgonio Pass wind resource area in the Mojave Desert, California. We quanti fied individual external par- . Research characterizing the effects of wind farm develop-ment on terrestrial wildlife has been slowly accumulating over the past two decades (Menzel and Pohlmeyer 1999; de Lucas