1 2 3 4 5 6 Stable isotope analysis of diet confirms niche separation of two sympatric 7 8 of Namib Desert 9 10 10 11 11 1 1,2 1 1 1 12 Ian W. Murray, Hilary M. Lease, Robyn S. Hetem, Duncan Mitchell, Andrea Fuller 12 13 and Stephan Woodborne3,4 13 14 14 1 the 15 Brain Function Research Group, School of Physiology, Faculty of Health Sciences, University of 15 16 Witwatersrand, Johannesburg, South Africa, 16 17 17 2Biology Department, Whitman College, Walla Walla, Washington, USA, 18 18 3 19 iThemba Laboratories, Gauteng, South Africa and 19 20 4Stable Isotope Laboratory, Mammal Research Institute, Department of Zoology and Entomology, University of 20 21 21 Pretoria, Pretoria, South Africa 22 22 23 23 24 24 25 Abstract 25 26 26 27 We used stable isotopes of carbon and nitrogen to study the trophic niche of two species of insectivorous 27 28 , the Husab sand lizard Pedioplanis husabensis and Bradfield’s Namib day gecko living sympatrically 28 13 15 29 in the Namib Desert. We measured the δ C and δ N ratios in lizard blood tissues with different turnover times 29 30 (whole blood, red blood cells and plasma) to investigate lizard diet in different seasons. We also measured the 30 13 15 31 δ C and δ N ratios in available prey and plant tissues on the site, to identify the avenues of nutrient 31 32 movement between lizards and their prey. Through the use of stable isotope mixing models, we found that the 32 33 two lizard species relied on a largely non-overlapping but seasonally variable array of : P. husabensis 33 34 primarily fed on termites, and wasps, while R. bradfieldi fed mainly on , wasps and hemipterans. 34

35 Nutrients originating from C3 plants were proportionally higher for R. bradfieldi than for P. husabensis during 35 36 autumn and late autumn/early winter, although not summer. Contrary to the few available data estimating the 36

37 trophic transfer of nutrients in ectotherms in mixed C3 and C4/crassulacean acid metabolism (CAM) plant 37

38 landscapes, we found that our lizard species primarily acquired nutrients that originated from C4/CAM plants. 38 39 This work adds an important dimension to the general lack of studies using stable isotope analyses to estimate 39 40 lizard niche partitioning and resource use. 40 41 41 42 Key words: Namib Desert, niche partitioning, Pedioplanis, , stable isotopes 42 43 43 44 44 45 45 46 Introduction 46 47 47 For all species there exists a particular set of biotic 48 Correspondence: Ian W. Murray, Brain Function Research 48 and abiotic conditions that bound their existence, which 49 Group, School of Physiology, Faculty of Health Sciences, 49 may be thought of as the species’ niche (Hutchinson 50 University of the Witwatersrand, Johannesburg, South Africa. 50 1957). Species niches can be characterized by habitat 51 Email: [email protected] 51

1 1 requirements, geographical distribution, thermal niches δ15N may be used as an indication of an organism’s 1 2 or other dimensions, which may not be independent of trophic level (DeNiro & Epstein 1981; Peterson & 2 3 each other. Resource use is one niche dimension that is Fry 1987). Conversely, tissue δ13C changes very little 3 4 widely studied because the resources that organisms use (approximately 0–1.0‰), on average, across trophic 4 5 play an important role in determining species diversity, levels within a food web; consequently, δ13C may 5 6 and may allow different species within a similar feeding be used to trace carbon sources (DeNiro & Epstein 6 7 guild to coexist in the same habitat (Simberloff & 1978; Peterson & Fry 1987). During photosynthesis 7 8 Dayan 1991). Within the same feeding guild, a species plants discriminate against carbon dioxide molecules 8 9 may be a resource specialist or a resource generalist containing the 13C isotope. However, due to differences 9 10 (Futuyma & Moreno 1988). Habitats usually will be in the enzymes responsible for carboxylation, plants 10

11 able to support a greater number of specialist species which use the C3 photosynthetic pathway (e.g. trees 11 12 that consume non-overlapping resources than generalist and most forbs) have significantly lower 13C values 12

13 species that overlap in their resource consumption compared to plants that use either the C4 (e.g. many 13 14 (Roughgarden 1974). This is due to competitive grasses) or crassulacean acid metabolism (CAM; many 14 15 exclusion, the theory that two similar species are unable succulents) photosynthetic pathway. These differences 15 13 16 to coexist with one another unless there is some level lead to C3 plants having a lower δ C ratio compared 16

17 of divergence in how they use resources (Hardin 1960; to C4/CAM plants (Ehleringer et al. 1986, 1997). C3 17

18 Pianka 1974). Indeed, a basic premise of community and C4/CAM plants also have important structural 18

19 ecology is that the coexistence of otherwise similar differences (e.g. C4 plants are characterized by Kranz 19 20 species within a feeding guild may be accomplished by anatomy, which includes the presence of thick-walled 20 21 the use of distinct resources (MacArthur 1958; Bowers bundle sheath cells), which influence their nutritional 21 22 & Brown 1982). profitability to consumers, and may have distinct 22 23 Research on lizard community structure and function growth responses to seasonal patterns of precipitation 23 24 has been important for characterizing the concept of the and climate (Ode et al. 1980; Schulze et al. 1996; 24 25 species niche, as well as understanding how different Barbehenn et al. 2004a,b; Muldavin et al. 2008). 25 26 species coexist (e.g., Schoener 1977; Pianka 1986). Arid Consequently, ecosystem primary productivity can be 2 27 ecosystems in particular may be ideal places to examine divided into distinct resource compartments based on 27 28 species niche partitioning because in such environments plant photosynthetic pathways. 28 29 lizard diversity is often high, it can be relatively easy Importantly, the physiological differences between 29 30 30 to secure large samples of individual lizards (Pianka C3 and C4/CAM plants mean that they are likely to be 31 1986), and limited resources have the potential to affected differently under current projections of climate 31 32 32 intensify competition (MacArthur & Levins 1967). For change and enhanced atmospheric CO2 levels (IPCC 33 33 example, despite the low availability of plant resources, 2014). For example, higher CO2 levels may improve 34 34 the Namib Desert is home to a diverse lizard fauna with C3 plant nutrient and water use efficiency, and favor 35 high levels of endemism (Robinson & Cunningham plants with high demands for woody structural tissue, 35 36 1978; Murray & Schramm 1987; Herrmann & Branch such as trees, compared to herbaceous plants, such as 36 37 37 2013), and, as with other hot deserts, high lizard grasses (most of which are C4 in arid regions; Drake et 38 biomass may represent an important component of the al. 1997; Bond et al. 2003). However, warmer and drier 38 39 39 food web in this ecosystem (Pianka 1986). However, climatic conditions would tend to favor C4/CAM plants 40 few studies to date have examined resource partitioning (Bond et al. 2003). From a consumer’s perspective 40 41 and trophic dynamics within the Namib lizard fauna these differences matter because many 41 42 42 (Robinson & Cunningham 1978; Murray & Schramm selectively forage on either C3 or C4/CAM plants, and 43 1987; Murray et al. 2016). the nutritional quality of these plant groups is not the 43 44 13 44 The use of stable isotopes, particularly carbon (δ C) same (Ehleringer et al. 2002; Barbehenn et al. 2004a,b). 45 15 45 and nitrogen (δ N) stable isotopes, is an effective and Furthermore, enhanced CO2 levels may translate into 46 46 minimally invasive way to quantify the spatial and negative consumer effects due to lower plant tissue 47 47 temporal patterns of consumer resource partitioning nitrogen content and higher carbon to nitrogen ratios 48 48 (Gannes et al. 1997; Boecklen et al. 2011). Because (Ehleringer et al. 2002). 49 49 tissue δ15N increases by approximately 3.0‰, on Here we investigate and compare the resource 50 50 average, across each trophic level within a food web, partitioning of 2 sympatric and similarly-sized species 51 51

2 1 of insectivorous Namib lizards. The Husab sand lizard, Lizard tissue collection 1 2 Pedioplanis husabensis Berger-Dell’Mour & Mayer, 2 3 1989, is a 2.5–3.0-g lacertid lizard endemic to rocky We captured lizards during austral summer 3 4 substrates in the west-central Namib Desert between the (December 2012–January 2013) and austral autumn 4 5 ephemeral Swakop and Khan Rivers (Berger-Dell’Mour (May 2013) using noose poles. We took blood samples 5 6 & Mayer 1989). Bradfield’s Namib day gecko, (approximately 50 µL) from the infraorbital sinus with 6 7 Rhoptropus bradfieldi Hewitt, 1935, is a 3.0–4.0-g rock- heparinized capillary tubes before releasing the lizards 7 8 dwelling diurnal gecko endemic to the Namib Desert unharmed (Murray et al. 2014). All procedures were 8 9 (Branch 1998). We examine the trophic niches for each approved by the University of the Witwatersrand’s 9 10 of these lizard species by analyzing the carbon and Ethics Screening Committee (clearance 10 11 nitrogen stable isotope ratios in plant tissues, available certificate number 2012/50/03) and were in accordance 11 12 arthropod prey and lizard tissues. Because P. husabensis with the Namibian Ministry of Environment and 12 13 and R. bradfieldi differ in their foraging strategy and Tourism Research/Collecting Permit 1744/2012. 13 14 habitat use (Murray et al. 2014, 2015), we predict We collected whole blood from adult lizards of both 14 15 that there will be significant differences between their species between December 2012 and January 2013 15 16 trophic niches, evidenced by different tissue isotope (austral summer), and in May 2013 (austral autumn/ 16 17 values. early winter). Blood was sampled from 21 male and 17 18 5 female P. husabensis and 13 male and 8 female R. 18 bradfieldi during austral summer, and 17 male and 19 Materials and Methods 19 20 26 female P. husabensis and 7 male and 11 female R. 20 21 Study site bradfieldi in austral autumn. 21 22 We centrifuged the blood samples collected in 22 23 Our study site is along the dry Swakop River, autumn to separate out the plasma and red blood cells 23 24 Namibia, at Hildenhof, approximately 40 km east (RBC). Plasma was not available for the blood that we 24 25 of Swakopmund (22°42.049′S, 14°54.890′E; 210 m; collected during summer because we used the plasma 25 26 see Murray et al. [2014, 2015] for further details) water for the determination of field metabolic rates 26 27 in the Namib Desert. The dry riverbed vegetation is (Murray et al. 2014, 2015). We air-dried RBC and 27 28 characterized by a riparian woodland consisting of loaded approximately 0.4 mg of RBC and dried whole 28 29 scattered trees and shrubs including Vachellia erioloba blood into 4 × 6-mm tin cups (Costech Analytical 29 30 (camelthorn), Tamarix usneoides (tamarisk), Faidherbia Technologies, California, USA; #041070). In addition, 30 31 albida (ana tree), Euclea pseudebenus (wild ebony) and we pipetted approximately 15 μL of plasma into 4 31 32 Salvadora persica (mustard bush), growing in and along × 6-mm tin cups immediately after centrifuging and 32 33 the edges of the sandy riverbed (Cowlishaw & Davies air-dried the samples before folding the tin cups for 33 34 1997). Adjacent to the riverbed are bare rocky slopes analysis. We did not extract lipids from the blood 34 35 sparsely-covered with small shrubs such as Arthraerua samples because blood contains too little lipid to 35 36 leubnitziae (pencil bush) and Sesuvium sesuvoides confound analyses (Bearhop et al. 2000). Due to small 36 37 (desert pink). A narrow zone of more densely-spaced blood volumes, several of the R. bradfieldi samples did 37 38 shrubs such as Zygophyllum stapffii (dollar bush), not yield large enough nitrogen peaks to be analyzed 38 39 Lycium sp. and Salsola sp. (salt bush) is situated on the by isotope ratio mass spectrometry, resulting in fewer 39 40 silty substrates where the rocky slopes meet the river nitrogen isotope ratios being reported than carbon 40 41 channel. Perennial grasses make up a small proportion ratios. For several additional samples we did not have 41 42 of plant cover and generally are restricted to the edges sufficient RBC sample masses to run either carbon or 42 43 of the river channel (I. Murray, personal observation). nitrogen. We lost several plasma samples in the mass 43 44 The study site is in a hyper-arid system with mean spectrometer. 44 annual precipitation of approximately 25 mm, and 25– 45 In small insectivorous lizards, plasma has a carbon 45 50 fog days per year may be expected based on data 46 retention time (the average amount of time a carbon 46 from other similar sites (Olivier 1995; Haensler et al. 47 atom is retained in tissue and a means to estimate tissue- 47 2011; Eckardt et al. 2013). After sporadic precipitation 48 specific turnover times; Martínez del Rio & Anderson- 48 events, such as one during April 2013, annual grasses 49 Sprecher 2008) of 25 days while RBC have a carbon 49 such as Stipagrostis sp. were also evident. 50 retention time of 61 days (Warne et al. 2010b). Because 50 51 51

3 1 plasma and RBC have different biological turnover any diet switches in those arthropods could take several 1 2 rates, their isotope ratios reflect dietary history over months to be reflected in the arthropod tissue isotope 2 3 both short (plasma) and long (RBC) periods (Boecklen ratios (termites [Watson 1973]; ants [Mooney & Tillberg 3 4 et al. 2011). Consequently, plasma from blood collected 2005; Straka & Feldhaar 2007; Menke et al. 2010]). We 4 5 in May reflected diet in late autumn/early winter, and identified arthropods to the species level where possible, 5 6 RBC from blood collected in May reflected diet during and otherwise to the order, family or genus level, using 6 7 autumn. Although plasma was unavailable for blood references for southern African arthropods (Scholtz & 7 8 collected in summer, the use of whole blood is well Holm 1985; Marsh 1986; Uys 2002; Picker et al. 2004). 8 9 established in the published literature (e.g. Boecklen et As with plant samples, we dried arthropods in an oven 9 10 al. 2011) for estimating diet. The carbon retention time and homogenized individuals before loading them in 4 10 11 for whole blood is unknown for lizards, but because it × 6-mm tin capsules. 11 12 is likely to have a retention time between that of plasma 12 Stable isotope analyses 13 (25 days) and RBC (61 days), with RBC largely driving 13 14 whole blood carbon retention times (Flaherty et al. We analyzed all of our tissue samples for carbon 14 15 2010; Warne et al. 2010b), we can confidently make the (δ13C) and nitrogen (δ15N) stable isotope ratios using a 15 16 assumption that isotope ratios in whole blood collected continuous flow isotope ratio mass spectrometer (Delta 16 17 in summer reflect early summer diet. V Plus, ThermoFinnigan, Bremen, Germany) connected 17 18 to an Elemental Analyzer (Flash EA 1112 series, 18 Characterization of plant and arthropod 19 ThermoFinnigan, Bremen, Germany) in the University 19 20 resources of Pretoria Isotope Ratio Mass Spectrometry Laboratory. 20 21 21 We collected tissue from 30 plant species on the site The instrumental precision of these measurements 22 22 during May 2013, which represented a majority of the was ± 0.1‰ SD based on repeated measurements of 23 23 species growing during our lizard sampling activity. internal laboratory standards. All sample runs included 24 13 24 We sampled multiple leaves and stems from 3 to 5 a laboratory standard (Merck Gel δ C = −20.57‰; 25 15 25 randomly selected plants of each species and stored δ N = 6.8‰) and blank after each set of 12 unknowns. 26 2 them in paper envelopes. The plant tissues were dried Isotope concentrations are reported in delta notation 27 27 in an oven at 55°C and samples were homogenized (δ) in parts per thousand (‰): δX = (Rsample/Rstandard − 1) 28 28 with a clean mortar and pestle to create a homogenate * 1000. Rsample and Rstandard represent the ratio of heavy 29 13 12 15 14 29 for each species. We analyzed the carbon and nitrogen to light isotopes ( C/ C or N/ N) for the sample and 30 30 stable isotope ratios for each species using aliquots standard. The results are normalized to the international 31 15 31 (approximately 1 mg) of the dried homogenate. standards air for δ N and Vienna Pee Dee Belemnite for 32 δ13C. 32 We sampled arthropods from areas where 33 When carbon and nitrogen stable isotope ratios in 33 34 lizards were active and foraging during May 2013 34 by walking through the habitat and hand capturing consumer tissue are analyzed, there is often an offset 35 between the diet and the tissue termed the diet-to-tissue 35 36 arthropods (beetles, ants, termites and spiders) and 36 sweeping vegetation with a net (flies, bees, wasps, true discrimination factor (Δ). Diet-to-tissue discrimination 37 factors may significantly differ according to diet 37 38 bugs, as well as some beetles and spiders). We made a 38 concerted effort to sample ants and termites (identified as quality, growth rates, tissue or species (Caut et al. 2008; 39 Caut et al. 2009; Boecklen et al. 2011). Determining 39 40 key components of lizard diet; Murray et al. 2016) in the 40 same microhabitats where we saw lizards. Arthropods discrimination factors requires time and labor-intensive 41 feeding trials, which have not been carried out for 41 42 were kept cool (approximately 15°C) in vials for 1–3 42 days, a period in which we assumed that all gut contents all species (Gannes et al. 1997; Martínez del Rio & 43 Carleton 2012). Consequently, we used the mean Δδ13C 43 44 were metabolized, and then frozen (approximately 15 44 −4°C) for storage. We acknowledge the potential (0.4‰, 91 studies) and the mean Δδ N (2.3‰, 65 45 studies) determined for poikilotherm tissue diet-to-tissue 45 46 difficulties involved with inferring lizard consumption 46 of arthropods in summer based on the tissue isotope discrimination factors during controlled feeding trials to 47 adjust our lizard tissue δ13C and δ15N values (McCutchan 47 48 ratios of arthropods collected during late autumn/early 48 winter. However, for the primary prey items that lizards et al. 2003). We assumed that all lizard tissues analyzed 49 would have similar discrimination factors. 49 50 feed on, such as termites and ants, the long periods of 50 51 time required for growth and development means that Data analyses: Tissue stable isotope ratios 51

4 1 We tested for sex-related differences in tissue stable Data analyses: Mixing models 1 2 isotope ratios for both species using 2-sample t-tests. 2 We estimated the extent to which lizards used 3 Plasma and RBC samples from individual lizards are not 3 arthropods dependent on C /CAM versus C plant 4 independent, so we used repeated measures linear mixed 4 3 4 resources with a 2-end-point mixing model (Martínez 5 effects models to compare the carbon and nitrogen 5 del Rio & Wolf 2005): 6 isotope ratios between species across seasons. We used 6 ) 7 an unstructured repeated covariance type, and modeled 7 8 species, season and the species*season interaction as . 8 9 fixed effects and individual lizard as a random effect. 9 10 10 We conducted post hoc comparisons using a Bonferroni where “tissue” is either lizard plasma, whole blood 11 correction. We used 2-sample t-tests to compare the 11 or RBC; p is the fraction of C4/CAM plant resources 12 seasonal incorporation of arthropods feeding on C / 12 13 4 assimilated in lizard tissue; and D is the carbon 13 CAM plant resources by the two lizard species. 14 discrimination factor (0.4‰, McCutchan et al. 2003). 14 15 Data analyses: Isotopic niche metrics The subscripts “C3” and “C4/CAM” represent the carbon 15 16 isotope ratios of C3 and C4/CAM plant photosynthetic 16 To compare seasonal changes in lizard dietary niches 17 pathways, respectively. 17 (using different tissues to estimate seasonal dietary 18 We estimated the proportional contribution of 18 19 changes), we used the Stable Isotope Bayesian Ellipses arthropod prey groups to lizard tissues using the 19 20 in R (SIBER) package to calculate the standard ellipse Bayesian Stable Isotope Sourcing Using Sampling 20 21 area corrected for small sample sizes (SEAc) as well (SISUS; Erhardt & Bedrick 2013) software which 21 22 as the area of overlap for the summer, autumn and provides a significant advantage over other stable 22 23 late autumn/early winter dietary niches (Jackson et al. isotope mixing models because SISUS allows for the 23 24 2011). SEAc is a proxy for the trophic niche, and is the variability of stable isotope ratios in diet categories, 24 25 bivariate standard deviation of the stable isotope ratios as well as accounts for uncertainty in stable isotope 25 26 (e.g. carbon and nitrogen) characterizing a group of discrimination factors (Erhardt & Bedrick 2013). We 26 27 27 consumers; SEAc thus represents the core isotopic niche identified potential prey based on those groups that we 28 for each lizard species. We also describe the area of have found previously in lizard fecal pellets (Murray 28 29 the convex hull (TA), and the associated Layman niche et al. 2016). We used SigmaPlot 8.0 (Systat Software, 29 30 30 metrics estimating additional measurements of species San Jose, CA, USA), Microsoft Excel 2007 (Microsoft, 31 31 niche structure calculated using the package Stable Redmond, WA, USA), IBM SPSS 21.0 (SPSS, Chicago, 32 32 IL, USA) and R 2.15.2 (R Development Core Team 33 Isotope Analyses in R (SIAR), for comparative purposes. 33 2009) for all analyses. For all analyses, significance was 34 The TA (the smallest surface that encompasses all of the 34 accepted at P < 0.05 and values are reported as mean ± 35 carbon and nitrogen stable isotope ratios for individuals 35 SD. 36 of a species in a bivariate plot) is a geometric approach 36 37 that may be used to estimate consumer dietary niche 37 38 breadth, although TA is more sensitive to sample size Results 38 39 39 than SEAc, and fails to take into account uncertainty C plants and C /CAM plants growing on the site 40 3 4 40 within a dataset (Layman et al. 2007; Parnell et al. 2010; had non-overlapping carbon isotope ratios, a critical 41 41 Jackson et al. 2011). Layman niche metrics further observation allowing the sources of the nutrients 42 42 characterize diet spacing patterns between individuals in assimilated by insectivorous lizards to be traced back 43 43 a population, and include the mean distance to centroid to the plant functional groups consumed by their prey 44 44 (CD), the mean nearest neighbor distance (MNND) and (Fig. 1). Mean carbon isotope ratios were −26.2‰ ± 45 45 the standard deviation of the mean nearest neighbor 0.4‰ (range, −30.3‰ to −23.7‰; n = 16 species) in C3 46 46 plant tissues and −14.5‰ ± 0.3‰ (range, −16.4‰ to 47 distance (SDNND; Layman et al. 2007). For example, 47 −13.0‰; n = 14 species) in C /CAM plant tissues. Plant 48 high values of MNND would indicate a more diverse 4 48 tissue nitrogen ratios were 12.0‰ ± 1.0‰ in C plants 49 trophic niche, while high SDNND indicates a high 3 49 (range, 7.4‰ to 18.4‰) and 10.4‰ ± 0.9‰ in C /CAM 50 degree of unevenness in the spacing of the individual 4 50 plants (range, 6.8‰ to 19.3‰). 51 lizards in bivariate isotopic space (Layman et al. 2007). 51

5 Data analyses: Mixing models 1 1 2 2 We estimated the extent to which lizards used 3 3 arthropods dependent on C4/CAM versus C3 plant 4 15 13 4 Figure 1 Mean (± SD) δ N and δ C ratios resources with a 2-end-point mixing model (Martínez 5 5 del Rio & Wolf 2005): for seasonal diet as estimated from plasma 6 (late autumn/early winter), whole blood ) 6 7 (WB; summer) and RBC (autumn) from 7 . 8 the Husab sand lizard (Pehu, Pedioplanis 8 9 husabensis) and Bradfield’s Namib day 9 10 gecko (Rhbr, Rhoptropus bradfieldi). 10 where “tissue” is either lizard plasma, whole blood Blood tissue carbon and nitrogen isotope 11 15 11 or RBC; p is the fraction of C4/CAM plant resources 12 ratios are plotted relative to the δ N and 12 δ13C tissue values for individual species assimilated in lizard tissue; and D is the carbon 13 13 of plants belonging to different functional discrimination factor (0.4‰, McCutchan et al. 2003). 14 14 The subscripts “C ” and “C /CAM” represent the carbon groups (30 species; C3 shrubs/trees, C4 3 4 15 grasses, C shrubs and crassulacean acid 15 isotope ratios of C and C /CAM plant photosynthetic 4 3 4 16 metabolism [CAM] succulents) available 16 pathways, respectively. 17 on the site. Lizard blood tissue δ15N (2.3‰) 17 We estimated the proportional contribution of 18 and δ13C (0.4‰) ratios have been adjusted 18 arthropod prey groups to lizard tissues using the 19 by subtracting the appropriate diet-tissue- 19 Bayesian Stable Isotope Sourcing Using Sampling 20 discrimination factors determined for 20 (SISUS; Erhardt & Bedrick 2013) software which 21 poikilotherms (McCutchan et al. 2003). 21 provides a significant advantage over other stable 22 22 isotope mixing models because SISUS allows for the 23 23 variability of stable isotope ratios in diet categories, 24 24 as well as accounts for uncertainty in stable isotope 25 25 Table 1 Mean (± SD) δ15N and δ13C ratios of potential prey items collected during May 2013 for the Husab sand lizard (Pedioplanis discrimination factors (Erhardt & Bedrick 2013). We 26 2 27 husabensis) and Bradfield’s Namib day gecko (Rhoptropus bradfieldi) along the dry Swakop River bed in the Namib Desert, 27 identified potential prey based on those groups that we Namibia have found previously in lizard fecal pellets (Murray 28 28 13 15 et al. 2016). We used SigmaPlot 8.0 (Systat Software, 29 Prey category n Mean δ C (‰) Mean δ N (‰) 29 San Jose, CA, USA), Microsoft Excel 2007 (Microsoft, 30 Arachnida 30 31 31 Redmond, WA, USA), IBM SPSS 21.0 (SPSS, Chicago, Araneae 12 −17.9 ± 2.3 18.3 ± 2.4 32 32 IL, USA) and R 2.15.2 (R Development Core Team Insecta 33 33 2009) for all analyses. For all analyses, significance was Coleoptera 34 Psammodes/Physosterna/Zophosis/Scarabidae (beetles1) 14 −14.9 ± 1.9 17.2 ± 3.8 34 accepted at P < 0.05 and values are reported as mean ± 35 Somaticus/Gonocephalum/Stenocara (beetles2) 19 −20.1 ± 3.2 15.2 ± 3.4 35 SD. 36 Hemiptera 2 −21.9 ± 0.9 13.4 ± 3.3 36 37 Hymenoptera 37 RESULTS 38 Ants 38 39 Lepisiota capensis 4 −19.5 ± 2.2 16.7 ± 1.6 39 C plants and C /CAM plants growing on the site 3 4 40 Pheidole sp. 3 −15.3 ± 0.3 14.9 ± 0.1 40 had non-overlapping carbon isotope ratios, a critical 41 Camponotus sp. 2 −13.3 ± 0.3 14.5 ± 1.5 41 observation allowing the sources of the nutrients 42 Bees 4 −17.8 ± 4.6 14.8 ± 3.9 42 assimilated by insectivorous lizards to be traced back 43 Wasps 3 −18.7 ± 5.0 8.8 ± 3.6 43 to the plant functional groups consumed by their prey 44 Isoptera 44 (Fig. 1). Mean carbon isotope ratios were −26.2‰ ± 45 Trinervitermes sp. 6 −17.1 ± 0.3 5.8 ± 0.2 45 0.4‰ (range, −30.3‰ to −23.7‰; n = 16 species) in C 3 46 Hodotermes mossambicus 8 −18.6 ± 0.5 7.3 ± 0.6 46 plant tissues and −14.5‰ ± 0.3‰ (range, −16.4‰ to 47 Psammotermes allocerus/Amitermes sp. 16 −16.1 ± 0.8 11.8 ± 1.0 47 −13.0‰; n = 14 species) in C /CAM plant tissues. Plant 4 48 Based on similar tissue isotope ratios the genera Somaticus, Gonocephalum and Stenocara were combined into the category 48 tissue nitrogen ratios were 12.0‰ ± 1.0‰ in C plants 3 49 “beetles2,” and the genera Psammodes, Physosterna, Zophosis and Scarabidae were combined into the category “beetles1.” Sample 49 (range, 7.4‰ to 18.4‰) and 10.4‰ ± 0.9‰ in C /CAM 4 50 sizes (n) indicate the numbers of individuals sampled with the exception of the small Lepisiota capensis in which case each 50 plants (range, 6.8‰ to 19.3‰). sample was a homogenate of 4 individual ants from a single nest. 51 51

6 1 The potential arthropod prey groups of lizards autumn (RBC) but not summer (whole blood) dietary 1 2 occupied largely non-overlapping domains in carbon niches. The interaction between species and season also 2

3 and nitrogen isotope niche space (Table 1). For was significant (F2,67.392 = 6.554; P = 0.003; Table 2). 3 4 example, mean δ13C ranged from −21.9‰ ± 0.9‰ in Across seasons the tissue δ15N ratios increased similarly 4 5 hemipteran to −13.3‰ ± 0.3‰ in ants in the for P. husabensis, while for R. bradfieldi the δ15N ratios 5 6 genus Camponotus (Table 1). Arthropod prey groups were similar in summer and late autumn/early winter 6 7 also occupied a diversity of trophic levels, as evidenced but declined in the autumn dietary niche (Table 2). 7 8 by their tissue δ15N, which ranged from 18.3‰ ± 2.4‰ 8 9 in spiders to 5.8‰ ± 0.2‰ in termites of the genus 9 10 Trinervitermes. There also was significant variation in 10 11 tissue nitrogen and carbon isotope ratios for different 11 12 arthropod genera within the same order, apparent, for 12 13 example, in the distinct and non-overlapping δ13C and 13 14 δ15N ratios for genera of termites and ants (Table 1). 14 15 The distinct isotope ratios of the lizards’ potential prey 15 16 allowed unambiguous identification of their diets. 16 17 There were no sex-related differences in blood carbon 17 18 and nitrogen isotope ratios for both P. husabensis (RBC 18 19 15 13 19 δ N, t41 = −0.22; P = 0.829; RBC δ C, t41 = −0.55; 20 15 20 P = 0.582; plasma δ N, t21 = 0.32; P = 0.752; plasma 21 13 15 21 δ C, t23 = 0.49; P = 0.631; whole blood δ N, t5 = 1.98; 22 13 22 P = 0.105; whole blood δ C, t6 = 1.39; P = 0.215) and 23 15 23 R. bradfieldi (RBC δ N, t12 = −0.01; P = 0.996; RBC 24 13 15 24 δ C, t14 = −1.89; P = 0.08; plasma δ N, t7 = −1.63; 25 13 25 P = 0.148; plasma δ C, t15 = 0.10; P = 0.920; whole 26 blood δ15N, t = −0.3; P = 0.767; whole blood δ13C, t 26 18 10 Figure 2 Mean (± SD) proportional use of C /crassulacean 27 = −1.25; P = 0.238), so we combined male and female 4 27 28 acid metabolism [CAM] plant-derived resources, relative to 28 values for both species (Table 2). On average blood C resources, as estimated from the δ13C ratios for the summer 29 15 3 29 δ N did not differ between lizard species (F1,82.091 = (WB, whole blood), autumn (RBC, red blood cells) and late 30 30 0.009; P = 0.925), but there was a significant difference autumn/early winter (P, plasma) tissue δ13C ratios from the 31 31 between seasons (F2,67.392 = 21.170; P = 0.000) with the Husab sand lizard (Pedioplanis husabensis) and Bradfield’s 32 15 32 δ N ratio reflecting the late autumn/early winter dietary Namib day gecko (Rhoptropus bradfieldi). *P < 0.001; 33 niche (plasma) significantly higher than those reflecting 2-sample t-test for seasonal species differences in resource use. 33 34 34 35 35 36 36 37 37 Table 2 Mean (± SD) δ15N and δ13C ratios for austral summer dietary niches (November–January as estimated from whole blood), 38 38 autumn dietary niches (March–May as estimated from red blood cells), and late autumn/early winter dietary niches (April–May, 39 39 as estimated from plasma) for the Husab sand lizard (Pedioplanis husabensis) and Bradfield’s Namib day gecko (Rhoptropus 40 40 bradfieldi) 41 41 42 Isotope ratio Summer Autumn Late autumn/early winter 42 43 43 Pedioplanis Rhoptropus Pedioplanis Rhoptropus Pedioplanis Rhoptropus 44 44 husabensis bradfieldi husabensis bradfieldi husabensis bradfieldi 45 45 Mean δ13C (‰) −17.0 ± 1.2 −16.7 ±1.5 −16.0 ± 0.7 −16.9 ± 1.2* −17.3 ± 0.7 −19.4 ± 1.1* 46 46 (n = 27) (n = 21) (n = 43) (n = 16) (n = 39) (n = 17) 47 47 15 48 Mean δ N (‰) 10.8 ± 2.7 13.0 ± 1.0* 12.2 ± 2.7 10.9 ± 1.4 13.5 ± 2.7 12.6 ± 1.0 48 49 (n = 26) (n = 21) (n = 43) (n = 14) (n = 39) (n = 12) 49 50 Sample sizes (n) indicate the number of individual lizards from which samples were analyzed. *Significant inter-species difference 50 51 (95% confidence interval estimates;P < 0.05). 51

7 1 Blood δ13C ratios were on average significantly 18% beetles, 8% wasps, 6% bees and 5% spiders, while 1

2 lower in R. bradfieldi than in P. husabensis (F1,103.760 = the autumn diet for P. husabensis was made up of 52% 2 3 21.494; P < 0.001). In addition, δ13C ratios reflecting termites and 45% beetles, with only trace inclusions 3 4 late autumn/early winter diet (plasma) were significantly of other prey categories (Table 4). P. husabensis’s late 4 5 lower than the δ13C ratios reflecting both autumn (RBC) autumn/early winter diet included a higher diversity 5

6 and summer (whole blood) diet (F2,60.412 = 161.763; P of prey groups, and was made up of approximately 6 7 < 0.001). The interaction between species and season 31% termites, followed by beetles at 28%, spiders and 7

8 again was significant (F2,60.412 = 23.812; P < 0.001; Table bees at 16% each, and wasps at 9% of the diet (Table 8 9 2). For P. husabensis, tissue δ13C ratios reflecting both 4). In contrast, R. bradfieldi had a summer diet made 9 10 the summer and late autumn/early winter dietary niches up largely of hymenopteran insects (38% ants, 35% 10 11 were lower than those reflecting the autumn dietary wasps and 7% bees), with lower proportions of beetles, 11 12 niche (Table 2). In R. bradfieldi the δ13C ratios remained hemipterans and spiders. During autumn, R. bradfieldi’s 12 13 the same for both the summer and autumn dietary diet was almost completely made up of hymenopterans 13 14 niches, but were significantly lower for the late autumn/ (67% wasps and 30% ants), with only trace quantities 14 15 early winter dietary niche (Table 2). of other insects. The late autumn/early winter diet for R. 15 16 In addition to the species differences in seasonal bradfieldi was also primarily hymenopteran insects (29% 16 17 dietary niches, P. husabensis assimilated significantly 17 18 more nutrients from arthropods that fed primarily upon 18 19 19 C4 or CAM plants than R. bradfieldi during autumn (2 20 20 sample t-test; t57 = 3.91; P = 0.000) and late autumn/ 21 21 early winter (2 sample t-test; t54 = 8.78; P = 0.000) but 22 22 not summer (2 sample t-test; t46 = −1.30; P = 0.201; Fig. 23 2). During summer R. bradfieldi and P. husabensis both 23 24 derived approximately 75% of their diet from arthropods 24 25 25 that consumed C4/CAM plants. However, compared to P. 26 husabensis, during autumn and late autumn/early winter 2 27 R. bradfieldi sourced 10–20% fewer resources from 27 28 28 arthropods feeding on C4/CAM plants (Fig. 2). 29 29 More evidence for the significant differences in the 30 Figure 3 The distribution of the mean (± SD) δ15N and δ13C 30 isotopic niches between R. bradfieldi and P. husabensis 31 ratios for potential arthropod prey categories relative to the 31 was that both the SEA and TA of P. husabensis’s dietary 32 c mean (± SD) and the seasonal standard ellipse areas corrected 32 niche were larger than those of R. bradfieldi across all 33 for small sample sizes (SEAc) during summer (whole blood), 33 seasons (Table 3, Fig. 3). During the summer and late 34 autumn (red blood cells) and late autumn/early winter (plasma) 34 autumn/early winter, the dietary niche SEA was more 35 c for sympatric Bradfield’s Namib day geckos, Rhoptropus 35 than twice that for P. husabensis than for R. bradfieldi, 36 bradfieldi (upper panel) and Husab sand lizards, Pedioplanis 36 while during autumn P. husabensis’s dietary niche was husabensis (lower panel). Lizard blood tissue δ15N (2.3‰) 37 37 only slightly greater than that of R. bradfieldi (Table and δ13C (0.4‰) ratios have been adjusted by subtracting the 38 38 3). There was also considerable seasonal overlap in the appropriate diet–tissue–discrimination factors determined for 39 39 summer, autumn and late autumn/early winter SEA s for poikilotherms in a recent meta-analysis (McCutchan et al. 40 c 40 P. husabensis, while the dietary niche for R. bradfieldi 2003). 41 41 was spatially distinct across these seasons (Fig. 3). The 42 42 niches of the two species overlapped in summer such 43 43 that the area of that overlap occupied approximately half 44 44 of the total niche area in R. bradfieldi, but only one-fifth (Table 3). Furthermore, in all seasons the CD was 45 45 of P. husabensis’s summer niche area (Table 3). During greater for P. husabensis relative to R. bradfieldi, while 46 46 autumn, the overlap in the lizards’ dietary niches took the MNND and SDNND were either similar in size 47 47 up a similar proportion of the total area of the autumn or greater for R. bradfieldi compared to P. husabensis 48 48 dietary niche in both species (Table 3). However, during (Table 3). 49 49 late autumn/early winter there was almost no overlap 50 Analysis of the whole blood isotope ratios indicated 50 in the dietary niches of R. bradfieldi and P. husabensis 51 that the summer diet of P. husabensis was 63% termites, 51

8 1 Table 3 Calculated niche metrics based on the δ15N and δ13C ratios during austral summer as estimated from whole blood, austral 1 2 autumn as estimated from red blood cells and austral late autumn/early winter as estimated from plasma in the Husab sand lizard 2 3 (Pedioplanis husabensis) and Bradfield’s Namib day gecko (Rhoptropus bradfieldi) 3 4 4 Niche metric Summer Autumn Late autumn/Early winter 5 5 Pedioplanis Rhoptropus Pedioplanis Rhoptropus Pedioplanis Rhoptropus 6 6 husabensis bradfieldi husabensis bradfieldi husabensis bradfieldi 7 7 Standard ellipse area 8 8 SEAc 10.4 4.0 5.8 4.4 5.1 2.4 9 9 10 Area of overlap 2.3 2.0 0 10 11 Layman niche metrics 11 12 TA 27.0 12.1 21.7 7.9 15.9 5.1 12 13 CD 2.6 1.5 2.4 1.5 2.3 1.0 13 14 MNND 0.73 0.63 0.44 0.59 0.35 0.58 14 15 SDNND 0.43 0.54 0.29 0.38 0.18 0.38 15 16 16 Niche metrics are based on standard ellipse areas corrected for small sample sizes (SEAc) and the inter-species overlap between 17 17 seasonal SEAc, as well as the Layman niche metrics area of the convex hull (TA), distance to centroid (CD), mean nearest neighbor 18 distance (MNND), and the standard deviation of the mean nearest neighbor distance (SDNND). 18 19 19 20 20 21 21 Table 4 The relative contribution of arthropod prey groups to the diet of sympatric Bradfield’s Namib day geckos (Rhoptropus 22 22 bradfieldi) and Husab sand lizards (Pedioplanis husabensis), as calculated (mean ± SD) by the software Stable Isotope Sourcing 23 23 Using Sampling (SISUS) using carbon and nitrogen isotopes 24 24 25 Prey category Summer Autumn Late autumn/early 25 26 diet diet winter diet 26 27 Pedioplanis Rhoptropus Pedioplanis Rhoptropus Pedioplanis Rhoptropus 27 28 husabensis bradfieldi husabensis bradfieldi husabensis bradfieldi 28 29 Arachnida 29 30 Araneae 0.05 ± 0.04 0.04 ± 0.03 0.01 ± 0.01 0.01 ± 0.01 0.16 ± 0.10 0.05 ± 0.04 30 31 Insecta 31 32 Coleoptera 32 33 Psammodes/Physosterna/Zophosis/ 0.15 ± 0.08 0.07 ± 0.06 0.44 ± 0.07 0.01 ± 0.01 0.19 ± 0.08 0.04 ± 0.03 33 34 Scarabidae (beetles1) 34 35 Somaticus/Gonocephalum/Stenocara 0.08 ± 0.06 — 0.01 ± 0.01 — 0.09 ± 0.07 — 35 36 (beetles2) 36 37 Hemiptera — 0.05 ± 0.04 — — — 0.40 ± 0.07 37 38 Hymenoptera 38 39 Ants 39 40 Camponotus sp. — 0.21 ± 0.10 — 0.29 ± 0.01 — 0.04 ± 0.03 40 41 Pheidole sp. — 0.13 ± 0.10 — 0.01 ± 0.01 — 0.05 ± 0.04 41 42 42 Lepisiota capensis — 0.04 ± 0.03 — — — 0.08 ± 0.06 43 43 Bees 0.06 ± 0.05 0.07 ± 0.06 0.01 ± 0.01 0.01 ± 0.01 0.16 ± 0.13 0.07 ± 0.06 44 44 Wasps 0.08 ± 0.06 0.35 ± 0.04 0.01 ± 0.01 0.67 ± 0.01 0.08 ± 0.06 0.29 ± 0.04 45 45 Isoptera 46 46 — 47 Trinervitermes sp. 0.29 ± 0.10 0.32 ± 0.07 — 0.08 ± 0.06 — 47 48 Psammotermes allocerus/Amitermes 0.25 ± 0.16 — 0.18 ± 0.16 — 0.15 ± 0.10 — 48 49 sp. 49 50 Hodotermes mossambicus 0.09 ± 0.07 — 0.02 ± 0.01 — 0.08 ± 0.06 — 50 51 —, not included in the diet, based upon prior gut content analyses (Murray et al. 2016). 51

9 1 wasps, 17% ants and 7% bees) but with considerable higher degree of trophic diversity among individual P. 1 2 contributions from hemipterans (40%) and a low husabensis relative to R. bradfieldi, as evidenced by 2 3 proportion of spiders and beetles (Table 4). the higher CD (Table 3), but the higher SDNND and 3 4 MNND for R. bradfieldi in autumn and late autumn/ 4 5 Discussion early winter indicated that individuals of this species 5 6 had less redundancy in the trophic niche than did 6 7 The insectivorous lizards P. husabensis and R. individuals of P. husabensis (Layman et al. 2007; Table 7 8 bradfieldi occurred within an isotopically-diverse 3). These niche differences reflect the consumption of 8 9 landscape of C3 and C4/CAM plants (Fig. 1), and, distinct arthropod prey items and probably result from 9 10 consequently, had an isotopically-distinct prey base of differences in foraging strategies between the two lizard 10 11 arthropods available to them (Table 1). That isotopic species. 11 12 diversity allowed us to assess changes in lizard resource Pedioplanis husabensis uses an active foraging 12 13 use over time. There was considerable variation strategy and moves widely through its habitat (Murray 13 14 between the two species in arthropod resource use (Fig. et al. 2014). In contrast, R. bradfieldi uses a sit-and-wait 14 15 3, Table 4). Although both lizard species showed some foraging strategy in which it ambushes its prey from an 15 16 degree of seasonal variation in arthropod prey use, the immobile and exposed position (Murray et al. 2015). 16 17 dietary composition of the two lizard species did not Relative to sit-and-wait foraging lizards, actively- 17 18 overlap in its major constituents (Table 4). P. husabensis foraging lizards are likely to have larger territories 18 19 fed predominantly on termites and beetles, while and move over greater distances through a diversity 19 20 R. bradfieldi fed predominantly on ants, wasps and of habitats (Pianka 1986; Vitt et al. 2003). While we 20 21 hemipteran insects. Furthermore, despite the presence lack data on home range size and the spatial length of 21 22 of considerable C3 plant biomass in their immediate daily movements in P. husabensis and R. bradfieldi, 22 23 habitat, these two insectivorous lizard species showed a data from other communities of insectivorous desert 23 24 preference for arthropods dependent on C4/CAM plants lizards indicate that the hourly distances moved by 24 25 (Fig. 2). active foraging lizards are 4 to 4.5 times the distances 25 15 26 Overall, the δ N ratios for lizard tissues were similar, (Anderson & Karasov 1981; Huey & Pianka 1981) and 2 27 implying that R. bradfieldi and P. husabensis fed at the the home ranges 4 times larger (Anderson & Karasov 27 15 28 same trophic level, although the lower δ N ratios for P. 1981) than those of sympatric sit-and-wait foraging 28 29 husabensis in summer may reflect its high consumption lizards. As lizards forage over greater distances they 29 15 30 of termites with their relatively low tissue δ N ratios are likely to encounter a greater degree of habitat 30 15 31 at this time of year. However, when we examined δ N heterogeneity. Because landscape heterogeneity is 31 13 32 values in conjunction with δ C values seasonally, we positively correlated with arthropod diversity (Liu et al. 32 33 found notable differences between P. husabensis’s and 2013), it is likely that more widely foraging lizards may 33 34 R. bradfieldi’s dietary niche (Table 2). For example, come into contact with a more diverse assortment of 34 35 P. husabensis always had a larger dietary niche (TA prey and, thus, have a larger trophic niche, as we found 35 36 36 and SEAc) than did R. bradfieldi occupying the same for P. husabensis. Because actively-foraging lizards use 37 habitat (Fig. 3; Table 3). Depending on the season, the visual and chemosensory means to locate prey above 37 38 TA of P. husabensis was 2.2–3.1 times larger than the and below ground, they are also capable of feeding on 38 39 corresponding TA in R. bradfieldi. The TA incorporates a greater array of potential prey, such as subterranean 39 40 data from all individuals, including outliers that may be larvae and immobile insect pupae that are not 40 41 critical to capturing the population’s or species’ complete available to sit-and-wait foraging lizards (Pianka 41 42 trophic spectrum (Layman et al. 2012); however, it 1986; Vitt et al. 2003). Therefore, the consequences of 42 43 is a metric that is sensitive to sample size (Jackson et foraging actively may contribute to the larger trophic 43 44 al. 2011), and we sampled fewer R. bradfieldi than niche of P. husabensis. 44 45 45 P. husabensis (Table 2). The SEAc is a metric which Compared to P. husabensis, the variable and 46 46 characterizes the niche far more robustly given a limited non-overlapping seasonal SEA s for R. bradfieldi 47 c 47 sample size, and the SEAc results echoed those yielded may be related to a sit-and-wait predator foraging 48 48 from the TA analysis: SEAc for P. husabensis was 1.3– opportunistically during a particular time of year (Fig. 49 49 2.6 times larger than that of R. bradfieldi (Table 3). In 3). The reduced trophic redundancy and increased 50 50 addition to having a larger isotopic niche, there was a “unevenness” characterizing R. bradfieldi in isotopic 51 51

10 1 space during autumn and late autumn/early winter (high resources based on tissue stable isotope ratios. However, 1 2 MNND and SDNND; Table 3) implies that R. bradfieldi these plants were relatively minor components of the 2 3 show a less uniform pattern of resource use. This local flora (e.g. Mesembryanthemum guericheanum; 3 4 pattern could be due to individual geckos encountering represented by several widely scattered small 4 5 a relatively heterogeneous variety of arthropods during individuals). Consequently, an enriched carbon isotope 5 6 sit-and-wait foraging bouts. In contrast, individuals of ratio in insectivores here is likely to represent significant 6 7 7 the actively-foraging P. husabensis can target distinct use of C4 plants. 8 8 prey resources, specifically making its dietary niche Pedioplanis husabensis obtained more than 70% 9 9 more uniform. We acknowledge that we cannot be sure of its nutrients from arthropods that sourced most 10 10 that the isotopically-distinct SEAc found seasonally for of their carbon from C plants in all seasons that we 11 4 11 R. bradfieldi are the result of the inclusion of different studied, as did R. bradfieldi in summer and autumn. 12 types of arthropod prey in the diet; the diets of the 12 13 During late autumn/early winter, however, R. bradfieldi 13 arthropods themselves may have varied seasonally, and 14 preyed almost equally on arthropods dependent on C3 14 we did not collect arthropod samples during summer. 15 plants (Fig. 2). For P. husabensis, we believe that its 15 Further data collection would be required to better 16 consumption of termites brought about its tight linkage 16 address this question. 17 to C4 plants. The termite genera that it fed upon are 17 18 From the perspective of individual consumers, the known to feed largely on C4 grasses (e.g. Hodotermes, 18 19 relative importance of the C3 versus C4 components of Psammotermes and Trinervitermes; De Visser et al. 19 20 plant primary productivity varies by species, season 2008; Symes & Woodborne 2011), and our carbon 20 21 and habitat (Magnusson et al. 1999; Warne et al. isotope analyses of individual termites supported 21 22 2010a). In addition, C4 plant production represents an this assertion (Table 1). However, we are unable to 22

23 important component of food web nutrients, particularly distinguish isotopically between termites feeding on C4 23

24 in arid ecosystems (Ehleringer et al. 1997; Still et al. grasses and on woody C4 shrubs such Salsola sp. that 24 25 2003). There are relatively few studies estimating the occurred on the site, which means that the importance of 25

26 transfer of C3 versus C4-derived nutrients to higher- C4 grasses to the arthropods making up P. husabensis’s 26 27 level consumers, such as lizards in a nutritional dietary niche may be overestimated. In an entirely 27

28 landscape containing both C3 and C4 plants (Magnusson different system, consumption of termites also was 28

29 et al. 1999, 2001; Warne et al. 2010a). However, the considered to underlie flow of 4C grass-derived nutrients 29 30 available data indicate that lizards continue to acquire into lizard and frog tissues (Magnusson et al. 1999, 30 31 considerable amounts of nutrients from prey that feed on 2001). 31 32 32 C3 plant resources even when appreciable proportions Compared to P. husabensis, R. bradfieldi fed to 33 33 of total primary productivity stem from C4 plants a greater extent on arthropods that used C plant 34 3 34 (Magnusson et al. 1999, 2001; Smith et al. 2002; Warne biomass (Figs. 2 and 3). During late autumn/early 35 35 et al. 2010a). However, contrary to other published winter in particular, R. bradfieldi acquired nutrients 36 36 studies, our lizards included a large proportion of from arthropods that used significantly more C plant- 37 3 37 arthropods that consumed resources derived from C4/ derived resources than did P. husabensis, incorporating 38 38 CAM plants. up to 54% of its carbon from C plant resources (Fig. 39 3 39 40 While we cannot distinguish between the carbon 2). We surmise that the late autumn/early winter 40 41 isotope ratios of C4 and CAM plants on our study dietary niche of R. bradfieldi reflected incorporation 41 42 site, CAM plants were a minor component of the of arthropods using C3 plant production available after 42 43 landscape and were represented chiefly by scattered recent precipitation (Noy-Meir 1973, 1974; Polis 1997). 43 44 and isolated succulents, while perennial and annual C4 As a sit-and-wait forager R. bradfieldi is likely to feed 44 45 grasses and the C4 shrub Salsola sp. were conspicuous largely on more mobile arthropods that are active during 45 46 and regular components of the landscape (I Murray, its diurnal activity period (Pianka 1986). Indeed, the 46 47 personal observation). Some Namib Desert plants are SISUS mixing model results showed its diet to be made 47 48 capable of facultatively switching between C3 and CAM up of mobile and diurnally-active insects like ants, 48 49 photosynthesis depending on water stress (e.g. Winter wasps and hemipterans (Table 4). The ecology of these 49

50 et al. 1978), and insect use of these resources could arthropod groups also enables them to transfer this C3 50 51 lead to a misinterpretation of insectivore use of plant plant biomass to R. bradfieldi effectively. For example, 51

11 1 hemipterans make up a large part of the total available may increase slightly in the coastal Namib Desert, but 1 2 arthropod biomass after rare desert precipitation events decrease by 23–39% further inland (Haensler et al. 2

3 (Polis 1991), and many small wasps feed on C3 flower 2011). 3 4 4 pollen or are predators on insects that feed on C3 plant While we do not know how reliant on fog moisture 5 5 production (Scholtz & Holm 1985; Picker et al. 2004). versus precipitation the C4/CAM plants are that fed the 6 6 Our previous work documenting the diet of P. arthropods that the lizards preyed on, most of the C3 7 husabensis and R. bradfieldi during May of 2013 using plants in the dry riverbed are reliant upon ground water, 7 8 fecal pellet analyses generally supports our estimates and evidence exists suggesting that some riverbed trees 8 9 of diet composition based on the SISUS mixing model (e.g. Vachellia erioloba and Faidherbia albida) may 9 10 results (Murray et al. 2016). These fecal pellet analyses already be experiencing significant water stress from 10 11 showed that the diet of P. husabensis was dominated reduced ground water availability (Schachtschneider 11 12 numerically by termites (71%) and that of R. bradfieldi & February 2010). Furthermore, most of the primary 12 13 by ants (87%; Murray et al. 2016). However, we note that productivity on the gravel plains of the Namib Desert 13 14 14 the mixing model results and the fecal pellet analyses is from annual C4 grasses that grow in response to 15 do not align perfectly such that, in some seasons, rainfall (Henschel et al. 2005). While the effects of 15 16 P. husabensis and R. bradfieldi incorporated fewer reduced precipitation, warmer temperatures and higher 16 17 17 nutrients from termites and ants than the fecal pellet atmospheric CO2 levels could potentially lead to losses 18 18 analysis implied (Table 4). These contrasting results are of C4/CAM plant biomass and reductions in plant 19 perhaps not surprising given the very short periods over nutritional quality, lizards in this study occupy dry 19 20 20 which fecal pellet analyses survey diet (days) relative to riverbed habitat at the juxtaposition of a C3 riparian 21 21 the period over which the mixing model results based woodland plant community and a C4/CAM desert plant 22 on body tissues do (1–2 months), as well as the fact community. Their potential resource use flexibility, 22 23 that the fecal pellet diet analyses estimated prey items coupled with this habitat juxtaposition, may allow 23 24 and not proportional contribution to diet. We further enhanced consumer resilience despite negative climate 24 25 acknowledge that diet reconstructions estimated from change impacts to particular plant groups. 25 26 2 isotope mixing models may give false-positive results We describe and compare the movement of nutrients 27 27 even if the items are not included in the diet and it may from the C and C /CAM photosynthetic pathways of 28 3 4 28 be difficult to include coverage of all possible dietary primary productivity into two secondary consumers 29 29 items. In addition, here we have employed blood only, (lizards) in the Namib Desert. We show that two 30 30 and left unexplored the differential routing of prey sympatric species of insectivores consume isotopically 31 31 macronutrients and their associated stable isotope ratios distinct arthropod resources, and that despite the very 32 32 to different tissues (Podlesak et al. 2006; Voigt et al. high available biomass of C plants in the adjacent 33 3 33 2008). riparian plant community, these two lizard species both 34 34 Recent models imply that climate change in the rely heavily on a food web based on C4/CAM-based 35 35 Namib Desert could result in a mean annual increase of plant resources. Although the amount of flexibility 36 36 up to 3°C and a reduction in annual precipitation by up that these lizards and their arthropod prey have in their 37 37 to 22%, with coastal regions of the Namib Desert likely dietary ecology is unknown, we think it possible that 38 38 experiencing less pronounced change (Thuiller et al. any potential negative impacts that climate change 39 39 2006). In this xeric environment the potential benefits may have on the availability or nutritional quality of C3 40 40 that increased atmospheric CO levels may have for versus C4/CAM plants in this system may be partially 41 2 41 plant photosynthetic efficiency elsewhere are not likely buffered by the food web flexibility provided by the 42 42 to be capable of compensating for warmer and drier adjacent plant community types. Our findings highlight 43 43 conditions. C plant biomass is projected to decrease the importance of understanding how environmental 44 3 44 significantly in parts of the Namib Desert, while C change may impact different plant functional groups 45 4 45 plant biomass is likely to change to a much lesser extent when considering ecosystem-level implications of 46 46 in the Namib Desert (Thuiller et al. 2006). However, climate change for consumer populations. Expanding 47 47 modeling the impacts of climate change in the Namib the temporal, spatial and consumer scope of tissue stable 48 48 Desert is made more complex due to the significant role isotope analyses may be particularly useful for better 49 49 that fog-derived moisture plays in this system (Henschel understanding food web dynamics in the Namib Desert. 50 50 & Seely 2008). The number of days that fog occurs 51 51

12 1 Acknowledgments promoting the spread of grasslands and savannas. 1 2 Global Change Biology 9, 973–82. 2 We thank Mary Seely, Theo Wassenaar and 3 Bowers MA, Brown JH (1982). Body size and 3 Gillian Maggs-Kölling of the Gobabeb Research 4 coexistence in desert rodents: Chance or community 4 and Training Centre in Namibia for critical research 5 structure? Ecology 63, 391–400. 5 6 support and logistical help. We also thank Grant 6 Hall of the University of Pretoria for assistance in Branch WR (1998). Field Guide to Snakes and Other 7 of Southern Africa. Struik, Cape Town. 7 8 the analysis of stable isotope samples, and Jonathan 8 Caut S, Angulo E, Courchamp F (2008). Caution on 9 Clarke for assistance in the field. We are grateful to 9 isotopic model use for analyses of consumer diet. 10 Antoinette and Walter Skutsch for accommodation 10 Canadian Journal of Zoology 86, 438–45. 11 at Hildenhof. Research funding was provided by an 11 12 FRC individual grant to IWM from the University of Caut S, Angulo E, Courchamp F (2009). Variation in 12 15 13 13 the Witwatersrand’s Faculty of Health Sciences and a discrimination factors (δ N and δ C): The effect 13 14 National Research Foundation International Research of diet isotopic values and applications for diet 14 15 Grant (Namibia/South Africa Research Cooperation reconstruction. Journal of Applied Ecology 46, 443– 15 16 Programme grant # 89140) to DM and Gillian Maggs- 53. 16 17 Kölling. IWM acknowledges the support of the Claude Cowlishaw G, Davies JG (1997). Flora of the Pro- 17 18 Leon Foundation through postdoctoral fellowship Namib Desert Swakop River catchment, Namibia: 18 19 funding. Community classification and implications for desert 19 20 vegetation sampling. Journal of Arid Environments 20 21 References 36, 271–90. 21 22 DeNiro MJ, Epstein S (1978). Influence of diet on 22 Anderson RA, Karasov WH (1981). Contrasts in 23 the distribution of carbon isotopes in animals. 23 energy-intake and expenditure in sit-and-wait and 24 Geochimica et Cosmochimica Acta 42, 495–506. 24 widely foraging lizards. Oecologia 49, 67–72. 25 DeNiro MJ, Epstein S (1981). Influence of diet on 25 26 Barbehenn RV, Karowe DN, Chen Z (2004a). the distribution of nitrogen isotopes in animals. 26 27 Performance of a generalist grasshopper on a C-3 and Geochimica et Cosmochimica Acta 45, 341–51. 27 a C-4 grass: Compensation for the effects of elevated 28 De Visser SA, Freymann BP, Schnyder H (2008). 28 CO on plant nutritional quality. Oecologia 140, 96– 29 2 Trophic interactions among invertebrates in termitaria 29 103. 30 in the African savanna: A stable isotope approach. 30 31 Barbehenn RV, Karowe DN, Spickard A (2004b). Effects Ecological Entomology 33, 758–64. 31 32 of elevated atmospheric CO on the nutritional 32 2 Drake BG, Gonzàlez-Meler MA, Long SP (1997). More 33 ecology of C-3 and C-4 grass-feeding caterpillars. 33 efficient plants: A consequence of rising atmospheric 34 Oecologia 140, 86–95. 34 CO2? Annual Review of Plant Biology 48, 609–39. 35 Bearhop S, Teece MA, Waldron S, Furness RW (2000). 35 Eckardt FD, Soderberg K, Coop LJ et al. (2013). The 36 Influence of lipid and uric acid on δ13C and δ15N 36 nature of moisture at Gobabeb, in the central Namib 37 values of avian blood: Implications for trophic 37 Desert. Journal of Arid Environments 93, 7–19. 38 studies. The Auk 117, 504–7. 38 Ehleringer JR, Rundel PW, Nagy KA (1986). Stable 39 Berger-Dell’Mour HAE, Mayer W (1989). On the 39 isotopes in physiological ecology and food web 40 parapatric existence of two species of the Pedioplanis 40 research. Trends in Ecology and Evolution 1, 42–5. 41 undata group (Reptilia: Sauria: Lacertidae) in 41 42 the central Namib Desert (Southwest Africa) Ehleringer JR, Cerling TE, Helliker BR (1997). C4 42 43 with description of the new species Pedioplanis photosynthesis, atmospheric CO2 and climate. 43 Oecologia , 285–99. 44 husabensis. Herpetozoa 1, 83–95. 112 44 45 45 Boecklen WJ, Yarnes CT, Cook BA, James AC Ehleringer JR, Cerling TE, Dearing MD (2002). 46 Atmospheric CO as a global change driver 46 (2011). On the use of stable isotopes in trophic 2 47 influencing plant-animal interactions. Integrative and 47 ecology. Annual Review of Ecology, Evolution, and 48 Comparative Biology 42, 424–30. 48 Systematics 42, 411–40. 49 Erhardt EB, Bedrick EJ (2013). A Bayesian framework 49 Bond WJ, Midgley GF, Woodward FI (2003). The 50 for stable isotope mixing models. Environmental and 50 importance of low atmospheric CO and fire in 51 2 Ecological Statistics 20, 377–97. 51

13 1 Flaherty EA, Ben-David M, Smith WP (2010). Diet Layman CA, Araújo MS, Boucek R et al. (2012). 1 2 and food availability: Implications for foraging and Applying stable isotopes to examine food web 2 3 dispersal of Prince of Wales northern flying squirrels structure: an overview of analytical tools. Biological 3 4 across managed landscapes. Journal of Mammalogy Reviews 87, 542–62. 4 5 91, 79–91. Liu R, Zhu F, Song N, Yang X, Chai Y (2013). Seasonal 5 6 Futuyma DJ, Moreno G (1988). The evolution of distribution and diversity of ground arthropods 6 7 7 ecological specialization. Annual Review of Ecology, in microhabitats following a shrub plantation age 8 8 Evolution, and Systematics 19, 207–33. sequence in desertified steppe. PLoS ONE 8, e77962. 9 Gannes LZ, O’Brien DM, Martínez del Rio C (1997). 9 10 MacArthur RH (1958). Population ecology of some 10 Stable isotopes in animal ecology: Assumptions, warblers of northeastern coniferous forests. Ecology 11 caveats, and a call for more laboratory experiments. 11 39, 599–619. 12 Ecology 78, 1271–6. 12 13 MacArthur R, Levins R (1967). The limiting similarity, 13 Haensler A, Cermak J, Hagemann S, Jacob D (2011). 14 convergence, and divergence of coexisting species. 14 Will the southern African west coast fog be affected 15 American naturalist 101, 377–85. 15 by future climate change? Results of an initial fog 16 16 projection using a regional climate model. Erdkunde Magnusson WE, de Araújo MC, Cintra R et al. (1999). 17 17 2011, 261–75. Contributions of C3 and C4 plants to higher trophic 18 18 Hardin G (1960). The competitive exclusion principle. levels in an Amazonian savanna. Oecologia 119, 91– 19 6. 19 20 Science 131, 1292–7. 20 Magnusson WE, Lima AP, Faria AS, Victoria RL, 21 Henschel JR, Seely MK (2008). Ecophysiology 21 22 of atmospheric moisture in the Namib Desert. Martinelli LA (2001). Size and carbon acquisition 22 23 Atmospheric Research 87, 362–8. in lizards from Amazonian savanna: evidence from 23 24 Henschel JR, Burke A, Seely M (2005). Temporal and isotope analysis. Ecology 82, 1772–80. 24 25 spatial variability of grass productivity in the central Marsh AC (1986). Checklist, biological notes and 25 26 Namib Desert. African Study Monographs 30, 43–56. distribution of ants in the central Namib Desert. 2 27 Herrmann HW, Branch WR (2013). Fifty years of Madoqua 14, 333–44. 27 28 herpetological research in the Namib Desert and Martínez del Rio C, Anderson-Sprecher R (2008). 28 29 Namibia with an updated and annotated species Beyond the reaction progress variable: the meaning 29 30 checklist. Journal of Arid Environments 93, 94–115. and significance of isotopic incorporation data. 30 31 Huey RB, Pianka ER (1981). Ecological consequences Oecologia 156, 765–72. 31 32 32 of foraging mode. Ecology 62, 991–9. Martínez del Rio C, Carleton SA (2012). How fast and 33 33 Hutchinson GE (1957). Population studies – animal how faithful – The dynamics of isotopic incorporation 34 34 ecology and demography – Concluding remarks. into animal tissues. Journal of Mammalogy 93, 353– 35 35 Cold Spring Harbor Symposia on Quantitative 9. 36 36 Biology 22, 415–27. 37 Martínez del Rio C, Wolf BO (2005). Mass balance 37 38 IPCC (2014). Climate Change 2014: Synthesis Report. models for animal isotopic ecology. In: Starck 38 39 Contribution of Working Groups I, II and III to the JM, Wang T, eds. Physiological and Ecological 39 40 Fifth Assessment Report of the Intergovernmental Adaptations to Feeding in Vertebrates. Science 40 Panel on Climate Change. Pachauri RK, Meyer LA, 41 Publishers, Enfield, pp. 141–74. 41 eds. IPCC, Geneva, Switzerland, 151 pp. 42 McCutchan JH, Lewis WM, Kendall C, McGrath CC 42 Jackson AL, Inger R, Parnell AC, Bearhop S (2011). 43 (2003). Variation in trophic shift for stable isotope 43 Comparing isotopic niche widths among and within 44 ratios of carbon, nitrogen, and sulfur. Oikos 102, 44 communities: SIBER – Stable Isotope Bayesian 45 378–90. 45 46 Ellipses in R. Journal of Animal Ecology 80, 595– 46 Menke SB, Suarez AV, Tillberg CV, Chou CT, Holway 47 602. 47 DA (2010). Trophic ecology of the invasive 48 Layman CA, Arrington DA, Montana CG, Post 48 argentine ant: Spatio-temporal variation in resource 49 DM (2007). Can stable isotope ratios provide for 49 assimilation and isotopic enrichment. Oecologia 164, 50 community-wide measures of trophic structure? 50 51 Ecology 88, 42–8. 763–71. 51

14 1 Mooney KA, Tillberg CV (2005). Temporal and spatial Podlesak DW, McWilliams SR (2006). Metabolic 1 2 variation to ant omnivory in pine forests. Ecology 86, routing of dietary nutrients in birds: Effects of diet 2 3 1225–35. quality and macronutrient composition revealed 3 4 Muldavin EH, Moore DI, Collins SL, Wetherill KR, using stable isotopes. Physiological and Biochemical 4 5 Lightfoot DC (2008). Above ground net primary Zoology 79, 534–49. 5 6 production dynamics in a northern Chihuahuan Polis GA (1991). Complex trophic interactions in 6 7 Desert ecosystem. Oecologia 155, 123–32. deserts: An empirical critique of food-web theory. 7 8 8 Murray GW, Schramm D (1987). A comparative study American Naturalist 138, 123–55. 9 9 of the diet of the wedge-snouted sand lizard, Meroles Polis GA, Hurd SD, Jackson CT, Piñero FS (1997). El 10 10 cuneirostris (Strauch), and the sand diving lizard, Niño effects on the dynamics and control of an island 11 11 Aporosaura anchietae (Bocage), in the Namib ecosystem in the Gulf of California. Ecology 78, 12 12 Desert. Madoqua 15, 55–61. 1884–97. 13 13 Robinson MD, Cunningham AB (1978). Comparative 14 Murray IW, Fuller A, Lease HM, Mitchell D, Wolf BO, 14 diet of two Namib Desert sand lizards (Lacertidae). 15 Hetem RS (2014). The actively foraging desert lizard 15 Madoqua 11, 41–53. 16 Pedioplanis husabensis behaviorally optimizes its 16 17 energetic economy. Canadian Journal of Zoology 92, Roughgarden J (1974). Niche width: Biogeographic 17 18 905–13. patterns among Anolis lizard populations. American 18 19 Murray IW, Fuller A, Lease HM, Mitchell D, Wolf Naturalist 108, 429–42. 19 20 BO, Hetem RS (2015). Low field metabolic rates Schachtschneider K, February EC (2010). The 20 21 for geckos of the genus Rhoptropus may not be relationship between fog, floods, groundwater and 21 22 surprising. Journal of Arid Environments 113, 35–42. tree growth along the lower Kuiseb River in the 22 23 Murray IW, Fuller A, Lease HM, Mitchell D, Hetem RS hyperarid Namib. Journal of Arid Environments 74, 23 24 (2016). Ecological niche separation of two sympatric 1632–7. 24 25 insectivorous lizard species in the Namib Desert. Schoener TW (1977). Competition and the niche. In: 25 26 Journal of Arid Environments 124, 225–32. Tinkle DW, Gans C, eds. Biology of the Reptilia. 26 27 Noy-Meir I (1973). Desert ecosystems: Environment Academic Press, New York, pp. 35–136. 27 28 and producers. Annual Review of Ecology, Evolution, Scholtz CH, Holm E (1985). Insects of Southern Africa. 28 29 and Systematics 4, 25–51. Butterworths, Durban. 29 30 Noy-Meir I (1974). Desert ecosystems: Higher trophic Schulze ED, Ellis R, Schulze W, Trimborn P, Ziegler H 30 31 31 levels. Annual Review of Ecology, Evolution, and (1996). Diversity, metabolic types and δ13C carbon 32 32 Systematics 5, 195–214. isotope ratios in the grass flora of Namibia in relation 33 33 Ode DJ, Tieszen LL, Lerman JC (1980). The seasonal to growth form, precipitation and habitat conditions. 34 Oecologia 106, 352–69. 34 35 contribution of C3 and C4 plant species to primary 35 Simberloff D, Dayan T (1991). The guild concept and 36 production in a mixed prairie. Ecology 61, 1304–11. 36 the structure of ecological communities. Annual 37 Olivier J (1995). Spatial distribution of fog in the 37 Namib. Journal of Arid Environments 29, 129–38. Review of Ecology, Evolution, and Systematics 22, 38 115–43. 38 39 Parnell AC, Inger R, Bearhop S, Jackson AL (2010). 39 Smith KF, Sharp ZD, Brown JH (2002). Isotopic 40 Source partitioning using stable isotopes: Coping 40 composition of carbon and oxygen in desert fauna: 41 with too much variation. PLoS ONE 5, e9672. 41 investigations into the effects of diet, physiology, and 42 42 Peterson B, Fry B (1987). Stable isotopes in ecosystem seasonality. Journal of Arid Environments , 419– 43 52 43 studies. Annual Review of Ecology, Evolution, and 30. 44 Systematics 18, 293–320. 44 Still CJ, Berry JA, Collatz GJ, DeFries RS (2003). 45 Pianka ER (1974). Niche overlap and diffuse 45 Global distribution of C3 and C4 vegetation: Carbon 46 competition. PNAS 71, 2141–5. 46 47 cycle implications. Global Biogeochemical Cycles 47 Pianka ER (1986). Ecology and Natural History of 48 17, 1–14. 48 Desert Lizards. Princeton University Press, Princeton. 49 Straka J, Feldhaar H (2007). Development of a 49 50 Picker M, Griffiths C, Weaving A (2004). Field Guide to chemically defined diet for ants. Insectes sociaux 54, 50 51 Insects of South Africa. Struik, Cape Town. 100–4. 51

15 1 Symes CT, Woodborne S (2011). Estimation of food stable carbon isotopes in tissue and exhaled breath. 1 2 composition of Hodotermes mossambicus (Isoptera: Oecologia 157, 31–40. 2 3 Hodotermitidae) based on observations and stable Warne RW, Pershall AD, Wolf BO (2010a). Linking 3 4 4 carbon isotope ratios. Insect Science 18, 175–80. precipitation and C3–C4 plant production to resource 5 Thuiller W, Midgley GF, Hughes GO et al. (2006). dynamics in higher-trophic-level consumers. Ecology 5 6 Endemic species and ecosystem sensitivity to climate 91, 1628–38. 6 7 7 change in Namibia. Global Change Biology 12, 759– Warne RW, Gilman CA, Wolf BO (2010b). Tissue 8 8 76. carbon incorporation rates in lizards: Implications for 9 9 Uys V (2002). A Guide to the Termite Genera of studies using stable isotopes in terrestrial ectotherms. 10 10 Southern Africa. Plant Protection Research Institute Physiological and Biochemical Zoology 83, 608–17. 11 11 Handbook No. 15. Agricultural Research Council, 12 Watson JAL (1973). The worker caste of the 12 Pretoria. 13 hodotermitid harvester termites. Insectes Sociaux 20, 13 14 Vitt LJ, Pianka ER, Cooper WE Jr, Schwenk K (2003). 1–20. 14 15 History and the global ecology of squamate reptiles. Winter K, Lüttge U, Winter E, Troughton JH (1978). 15 American Naturalist 162, 44–60. 16 Seasonal shift from C3 photosynthesis to crassulacean 16 17 Voigt CC, Rex K, Michener RH, Speakman JR (2008). acid metabolism in Mesembryanthemum crystallinum 17 18 Nutrient routing in omnivorous animals tracked by growing in its natural environment. Oecologia 34, 18 19 225–37. 19 20 20 21 21 22 22 23 23 24 24 25 25 26 2 27 27 28 28 29 29 30 30 31 31 32 32 33 33 34 34 35 35 36 36 37 37 38 38 39 39 40 40 41 41 42 42 43 43 44 44 45 45 46 46 47 47 48 48 49 49 50 50 51 51

16