Introduction

Seminal works have documented global changes in flora and fauna through the Earth’s history (Cerling et al. 1997, 1998; Iacumin et al. 2000; Burton et al. 2001). Recently, many studies have used stable isotope ratios to explore the effects of global warming, increased CO2 concentrations, and invasions of exotic species on plant and animal physiology, community composition, and ecosystem function (Vander Zanden et al. 1999; Arens et al. 2000; Barber et al. 2000; Bergengren et al. 2001; Hibbard et al. 2001; Kritzberg et al. 2004; Pace et al. 2004).

The ratio 13C/12C plays a prominent role in studies of global warming and increased CO2 concentrations. Because of fractionation during photosynthesis, CO2 from biogenic sources is depleted in 13C relative to CO2 from inorganic sources (O’Leary 1981), and recent large-scale influxes of CO2 from organic sources (e.g., fossil fuel combustion) have decreased atmospheric CO2 δ13C (13C/12C, Friedli et al. 1986; Francey et al. 1999). Such a decrease provides a marker that can be used to construct atmospheric budgets and calculate input of biogenic, including anthropogenic, carbon relative to input of inorganic carbon (Keeling et al. 1989). Since plants assimilate CO2 from the atmosphere, plant tissue δ13C relates to atmospheric CO2 δ13C (O’Leary 1981). Consequently, atmospheric geoscientists and paleoclimatologists have used plants to infer past atmospheric conditions (Arens et al. 2000), for example, using tree ring cellulose δ13C as an indicator of annual atmospheric CO2 δ13C (Freyer 1979; Mazany et al. 1980; Stuiver et al. 1984; Stuiver and Braziunas 1987; Lipp et al. 1991). Further, time-specific measures of atmospheric CO2 δ13C have been inferred from kernels of maize grown during known years (Marino and McElroy 1991), and long-term records of δ13C have been reconstructed from sediment cores of peat (White et al. 1994). Additionally, δ13C of carbonates in aquatic invertebrates such as sponges (Druffel and Benavides 1986; Böhm et al. 1996), coral (Nozaki et al. 1978), and mollusks (McConnaughey et al. 1997) have been shown to relate to oceanic dissolved inorganic carbon (DIC), which relates to atmospheric CO2 δ13C; and long-term records of these organisms could be used to infer historic atmospheric conditions.

Since plants form the foundation of most food chains, and consumer δ13C correlates with δ13C of diet (DeNiro and Epstein 1978; Peterson and Fry 1987; Kelly 2000), animal tissue δ13C can indirectly relate to atmospheric CO2 δ13C. Long-term trends of δ13C of fossilized faunal tissue have been correlated with δ13C of atmospheric CO2 over the past 40,000 years (Iacumin et al. 1997; Iacumin et al. 2000; Richards and Hedges 2003). Further, Cerling and Harris (1999) observed a decline in mammalian tooth enamel δ13C values consistent with declines observed in atmospheric CO2 δ13C.

In ecosystem studies, stable isotope analysis has been used to study food-web dynamics, migration, transfer of nutrients between ecosystems and animal behavioral ecology (Kling et al. 1992; Ben-David et al. 1997, 1998a, 2004; Hobson 1999; Kelly 2000; Post 2003; Pace et al. 2004). For example, in trophic studies, δ15N is useful for determining trophic position of a consumer because δ15N becomes enriched (i.e., ratio of 15N : 14 N increases) as nitrogen passes through food webs; hence, consumers with greater δ15N values typically occupy higher trophic positions (DeNiro and Epstein 1981; Peterson and Fry 1987; Kelly 2000). Unlike δ15N, δ13C is not appreciably changed as carbon moves through food webs (Peterson and Fry 1987; Post 2002), however, in terrestrial ecosystems, because 13C is enriched in C4 plants relative to C3 plants, δ13C is useful for determining photosynthetic pathway of plants that form the foundation of terrestrial trophic systems (Peterson and Fry 1987). Likewise in lakes, δ13C of algae and detritus in littoral regions is enriched relative to δ13C of phytoplankton from pelagic regions, and δ13C of consumers can be used to discriminate between these two energy sources (France 1995). Further, isotope analyses have demonstrated links between terrestrial and aquatic ecosystems, with aquatic systems being supplemented by carbon of terrestrial origin (Pace et al. 2004) and terrestrial systems being supplemented with carbon of aquatic origin (Ben-David et al. 1998b).

With the example set by previous work (Cerling et al. 1997, 1998; Iacumin et al. 2000; Burton et al. 2001) and the current increase in use of stable isotope analysis, it is likely that a rapid expansion of studies of recent human-induced global changes will follow. Though changes in atmospheric CO2 δ13C provide a useful tool for estimating biogenic contributions to atmospheric CO2 concentrations, large-scale depletion of atmospheric CO2 δ13C produces a substantial change in isotopic baseline for stable carbon isotope studies and represents a potentially serious confounding factor for ecosystem studies. Atmospheric CO2 provides, directly or indirectly, the fundamental carbon source for most terrestrial and aquatic food chains, and comparison of tissues samples collected over a temporal span can yield biased results unless atmospheric change in CO2 δ13C is controlled. Establishing an appropriate isotopic baseline is a critical component of stable isotope investigations (Cabana and Rasmussen 1994; Vander Zanden et al. 1997; Post 2002), and although potential bias from atmospheric depletion of CO2 δ13C has been identified by other researchers (Cerling and Harris 1999), temporal changes in atmospheric CO2 δ13C have not been fully addressed in most of the literature on isotopic ecology.

Herein, we construct a correction factor, based on a recently published, high-precision record of atmospheric CO2 (Francey et al. 1999), to facilitate tissue sample comparison while controlling for atmospheric change. We then demonstrate its use by applying this correction factor to collagen δ13C data from a 100-year series of mountain lion (Puma concolor) bone samples collected in California, and we address potential effects of precision of this correction factor on comparison of samples collected through time. Finally, we identify limitations in the applicability of the correction factor and indicate areas in which additional research is needed.

Materials and methods

Atmospheric model

To correct for changes in δ13C of atmospheric CO2, we modeled a high-precision record of atmospheric CO2 δ13C from direct atmospheric measurements and from air trapped in Antarctic ice (Francey et al. 1999). The Francey et al. (1999) data were used because (1) they represent a complete, long-term record of postindustrial atmospheric conditions; (2) precision estimates for each point, useful for variance estimates in our analyses, are reported; (3) measurement errors present in earlier datasets (e.g., Friedli et al. 1986) have been corrected; (4) trends correspond closely with data represented elsewhere (e.g., Keeling et al. 1989); and (5) high-latitude CO2 δ13C values do not differ from annual global means (Keeling et al. 1989) and therefore serve as a reliable proxy for generalized global trends.

We modeled the Francey et al. data (1999) from 1880 to 1996 using a negative exponential model:

$$\delta ^{13}C = k - e^{at^2}, $$
(1)

where k and a were estimated parameters and t was an index to year in which 1 represented 1880.

Example data set

We used a directory of museum collections of mammals (Hafner et al. 1997) to identify and query 33 museums from the USA regarding their holdings of mountain lions from California. From responses to these inquiries we compiled a database of 280 mountain lion specimens collected in California (Long and Sweitzer 2001) and analyzed bone samples of 194 individual mountain lions collected between 1893 and 1995. Bone collagen was used for isotopic analysis because of its availability in museum collections; further, slow turnover rate of bone collagen makes this tissue advantageous for providing an integrated lifetime average of dietary input (Tieszen and Boutton 1989). For most mammals, fractionation enriches tissue δ13C much less than other stable isotopes such as 15N, and tissue δ13C is often enriched only 1–6% over dietary δ13C (Kelly 2000).

Museum curators often value skulls more highly than postcranial skeletal material, and as isotope values of collagen show very little to no variation between different bones within a single organism (DeNiro and Schoeninger 1983), we preferentially sampled the latter when available. Using a rotary tool equipped with a small cutting blade, approximately 1 g of postcranial skeletal material was sampled from the axis of a long bone or a portion of rib. For specimens represented by skulls only, a tile-cutting bit was used to section bone from the tentorium inside the cranium.

Extraction, purification, and lipid removal of bone collagen followed Matheus (1997). One milligram of recrystallized collagen was loaded into an 8×5 mm tin capsule (Elemental Microanalysis Limited) for isotopic analysis, which was performed on a Europa Scientific continuous flow isotope ratio mass spectrophotometer (University of California Davis, Stable Isotope Facility and University of Alaska Fairbanks, Stable Isotope Laboratory). Isotope data are expressed as parts per thousand deviation from a PDB limestone standard (Kelly 2000).

To reduce spatial variability and to control for potential differences in climate and community structure, which may affect δ13C of animal tissue, we used collection location data to categorize each mountain lion sample into one of ten physiognomically similar geographic subdivisions of California (Hickman 1993), hereafter referred to as ecoregions (Fig. 1). Analyzed specimens were distributed among the ecoregions, including Northwest California (n= 36), Cascade Ranges (n = 9), Sierra Nevada Mountains (n = 61), Central Western California (n = 48), and Southwest California (n = 33). Because the Cascade Range region was relatively underrepresented with museum specimens, tissue samples from this ecoregion were grouped with geographically contiguous Northwest California, and these combined data are hereafter referred to as Northern California (n = 45). Additional specimens (n = 7) were distributed among three other ecoregions (Great Central Valley, Mojave Desert, and Great Basin), but because of small sample sizes, we excluded these samples from analyses.

Fig. 1
figure 1

Specimens of mountain lions (Puma concolor) collected in California, USA, between 1893 and 1995 characterized by five ecoregions. Completely overlapping collection locations are not registered separately. Data from the Cascade Ranges and Northwestern California were grouped for analyses

Full size image

Before application of the correction factor, we assessed temporal variation in collagen δ13C within each ecoregion using Spearman rank correlation (R s ), due to non-normal distribution of collection dates. Further, by using a nonparametric test, we did not assume a linear relationship between collection date and bone collagen δ13C. Statistically outlying δ13C values were checked using studentized residuals. Two samples were determined to be outliers (t = 3.53, P < 0.05) and were excluded from subsequent analysis.

Application of the correction factor

To correct δ13C values of bone collagen in our sample based on changes in atmospheric CO2 δ13C, we calculated change in atmospheric CO2 δ13C between 1880 and the year in which a tissue sample was obtained and added this value to tissue δ13C values. To estimate precision of adjusted δ13C values for bone tissue, we added the variance of measurement error from the analysis of bone samples to the variance of the estimate of the predicted δ13C value from the negative exponential model using a Taylor series approximation (Seber 1982):

$$\widehat{\operatorname{var}}\left({\widehat\delta^{{\text{13}}}{\text{C}}_{{\text{adjusted}}}} \right) = \widehat{\operatorname{var}}\left({\widehat\delta ^{{\text{13}}}{\text{C}}_{{\text{bone}}}} \right) + D(\widehat\theta)I'(\widehat\theta)D(\theta),$$
(2)

where \(D(\widehat\theta)\) is the vector of partial derivatives of the negative exponential model with respect to parameters k and a, and \(\widehat I(\widehat\theta)\) is the estimated variance–covariance matrix coefficients in the negative exponential model. To reassess temporal variation after controlling for atmospheric variation, we conducted Spearman rank correlations on corrected bone collagen δ13C.

Results

Atmospheric model

The negative exponential model of atmospheric CO2 δ13C explained a significant amount of variation in the Francey et al. (1999) data (F 2,88 = 4800.8, P < 0.001, Fig. 2). Parameter estimates and associated standard errors were found to be k = −5.5656 (1.2606×10−2) and a = 6.0932×10−5 (5.8652×10−7). Residuals exhibited homoscedasticity, and 95% confidence intervals of predicted values encompassed the actual values for all but two data points.

Fig. 2
figure 2

Change in atmospheric CO2 δ13C since 1880 (data from Francey et al. 1999), with Eq. 1 fit to the data (F 2,88 = 4,800.8, P < 0.001)

Full size image

Example data set

Consistent with observations that temporal changes in δ13C of atmospheric CO2 affect δ13C of consumer tissue, mountain lion bone collagen δ13C was negatively correlated with collection date in three of four ecoregions (n ≥ 32, P≤0.01, Fig. 3). Mountain lion bone collagen δ13C in the Sierra Nevada ecoregion demonstrated no significant trend (n = 61, P = 0.14, Fig. 3).

Fig. 3
figure 3

Correlation of mountain lion bone collagen δ13C with collection date for specimens from four ecoregions of California: (a) Northern CA, (b) Sierra Nevada, (c) Central Western CA, and (d) Southwestern CA

Full size image

Correction factor application

Application of the correction factor to the mountain lion bone collagen δ13C data removed the trend from three ecoregions that had shown significant declines (n ≥ 32, P ≥ 0.18, Fig. 4), and adjusted bone collagen δ13C values of mountain lions from the Sierra Nevada demonstrated a significant temporal increase (n = 61, P < 0.001, Fig. 4). Adjusted δ13C estimates from bone tissues had CVs < 2%.

Fig. 4
figure 4

Correlation of corrected mountain lion bone collagen δ13C with collection date. Specimens are categorized by four ecoregions of California, and Eq. 1 (see Fig. 2) has been applied to bone collagen δ13C values to control for atmospheric variation

Full size image

Discussion

Changes in δ13C of atmospheric CO2 have been shown to alter δ13C of plant (Marino and McElroy 1991; Arens et al. 2000) and animal tissue (Richards and Hedges 2003), but implications of recent declines in atmospheric CO2 δ13C for application of stable isotope analysis in exploration of recent global changes have largely been overlooked (but see Cerling and Harris 1999; Barber et al. 2000). Using mountain lion bone collagen as an example, we documented significant temporal declines in collagen δ13C values throughout much of California, and these declines were consistent with patterns associated with depletion of atmospheric CO2 δ13C (Cerling and Harris 1999).

Comparison of results from our analyses before and after correction was applied illustrates the need to account for atmospheric variation in long-term stable isotope studies. Application of the correction factor changed the significance of results in all four ecoregions tested. After correction, three ecoregions demonstrated no significant trend. In the Sierra Nevada ecoregion, we observed a significant temporal increase in mountain lion bone collagen δ13C, a pattern likely relating to prey-switching behavior in response to changing prey availability in this region (Long 2001). Without correction, significant ecological changes may have been assumed in three ecoregions that demonstrated significant change in tissue δ13C. On the other hand, potentially important ecological changes may have been overlooked in the Sierra Nevada ecoregion, where no trend was detected before correction.

That atmospheric depletion in CO2 δ13C was detectable in a generalist, apex predator indicates that results of atmospheric change are transmitted up trophic chains and are pervasive even in systems with large environmental variation in multiple processes. Effects of depleted atmospheric CO2 δ13C have been clearly documented in a variety of plant tissues (Marino and McElroy 1991; Arens 2000), and because plants assimilate carbon directly from the atmosphere, the depletion pattern is most accurately recorded in plant tissue. At higher levels in trophic chains, additional sources of isotopic variation are introduced (e.g., mixed diet), but long-term changes in atmospheric CO2 δ13C are indirectly recorded in herbivore tissue (Cerling and Harris 1999; Iacumin et al. 2000; Richards and Hedges 2003), and as suggested here, in carnivore tissue. Therefore, correcting for atmospheric change in stable isotope studies is important regardless of trophic position.

Cerling and Harris (1999) measured δ13C of grasses and ungulate tooth enamel from samples collected in the 1960s and in the 1990s. To compare these samples, they accounted for atmospheric variation by correcting for an approximate 1‰ change in δ13C of atmospheric CO2 between 1956 and 1997, based on data presented by Keeling et al. (1979) and Friedli et al. (1986). Our correction technique is similar, but uses a model based on long-term data from 1880 until modern times and incorporates the precision of this model in the variance estimate for the corrected δ13C ratio.

Application of a correction factor introduces additional variability in ratio estimates and, consequently, reduces precision. However, source data for the correction factor were precise (typically error <0.5% of point estimates, Francey et al. 1999), and our negative exponential model fit the data well (Fig. 2). In our study, measurement error associated with the correction factor was much less than variation exhibited in bone tissue δ13C values. Therefore, measurement error, determined from Eq. 2, did not substantially reduce the precision of the estimates.

Although data analyzed here are exclusively terrestrial, similar patterns appear in aquatic systems. For example, long-term depletion of atmospheric CO2 δ13C has lowered δ13C of dissolved inorganic carbon, which in turn has lowered δ13C of precipitated carbonate in aquatic invertebrates (Nozaki et al. 1978; Druffel and Benavides 1986; Böhm et al. 1996). Further, aquatic phototroph tissue δ13C relates largely to DIC δ13C (Osmond et al. 1981), and because, as in terrestrial systems, aquatic consumer tissue δ13C is determined largely by δ13C of their food (Post 2003), effects similar to those observed in terrestrial systems should be apparent in aquatic systems. Consequently, correction for atmospheric change in stable carbon isotope investigations is also important for aquatic investigations.

In both aquatic and terrestrial systems, however, time lags between atmospheric change and ecological assimilation of atmospheric CO2 may represent a potentially serious confounding complication. For instance, if isotopic changes in atmospheric CO2 require a long time to equilibrate between systems (Libby et al. 1964), direct application of our correction factor may not be appropriate. In terrestrial systems, though tissue δ13C of plants reflect current atmospheric CO2 δ13C, lags may exist (1) if there are delays in assimilation of plant material into trophic webs, and (2) if there are delays in assimilation of dietary material into tissue, particularly tissues with low-turnover rates in especially long-lived species. For instance, as bone collagen has low metabolic turnover relative to other tissues such as hair, blood, and muscle, δ13C of bone collagen is generally considered to represent an integrated lifetime average of dietary input (Tieszen and Boutton 1989). Though specific estimates in animals are rare, Hobson and Clark (1992) showed isotopic turnover of bone collagen δ13C in Japanese Quail (Coturnix japonica) to be 173.3 days. However, this rate is likely longer in adults of long-lived species such as Asian elephants (Elephas maximus, Sukumar and Ramesh 1992) and humans (Libby et al. 1964).

In addition to metabolic lag in tissues, long residence time of carbon in low trophic levels may delay reflection of atmospheric trends at higher trophic levels. In many systems, annual plant growth, which relates isotopically to current atmospheric conditions, forms the bulk of diet for primary consumers, and in these cases, time lags would be minimal. Exceptions, such as deep-water lentic systems with detritus-based trophic webs and long detrital residence time, would delay atmospheric trends. For example, time lags in incorporation of algal δ13C values were recorded in nematodes under experimental conditions, and these time lags were consistent with the mixing rates of the sediments (Moens et al. 2002). Once carbon is sequestered by primary consumers, further substantial delays are unlikely, as soft tissues with quick metabolic turnover rates typically provide much of the carbon that is assimilated by higher order consumers (Tieszen and Boutton 1989). However, researchers need to be aware of potential violations of these assumptions, and correct for long-turnover rates when appropriate.

In our analysis of mountain lion tissue, we corrected date of all samples based on collection date, though this surely introduced variation associated with time lags. However, mountain lions are not a particularly long-lived species; Beier (1993) estimated that most mountain lions do not live past 5–6 years of age, with maximum longevity of approximately 12 years. Hence, when analyzing bone collagen, we expect lag from atmospheric CO2 to tissue δ13C to be, on average, 5 years or less. Correcting for a lag of 5 years, even in our most recent sample from 1995, where atmospheric change is most extreme, changes δ13C by less than 0.15‰ and, over the range of our investigation, introduces negligible variation. However, in studies of longer-lived species, with significantly greater life expectancies, or in studies examining tissues such as tooth dentine, which has turnover rates much slower than bone collagen (Richards et al. 2002), correction for time lags may be appropriate.

In aquatic systems, a third source of lag may be present, as turnover rate between atmospheric CO2 δ13C and DIC δ13C can be quite variable, especially in large bodies of water. For example, Böhm et al. (1996) found no isotopic equilibrium between oceanic surface water DIC and atmospheric CO2, and because equatorial North Atlantic Ocean surface water DIC has only approximately 10±4 years to equilibrate with the atmosphere before being recycled into deeper waters, DIC δ13C consistently reflected only 50–65% of the atmospheric decline (Böhm et al. 1996). Therefore, depletion of tissue δ13C in oceanic organisms, especially deep-water organisms, will likely be less than depletion of terrestrial tissue δ13C. In smaller bodies of water, equilibration between atmospheric CO2 δ13C and DIC δ13C can occur more rapidly, and where dissolved CO2 quickly equilibrates with atmospheric CO2 our correction should be appropriate.

Further factors complicating temporal δ13C correction in aquatic systems relate to variable sources of carbon input. For example, using δ13C Post (2002) identifies three sources of DIC that may be incorporated into trophic webs in lakes: dissolution of atmospheric CO2, remineralization of allochthonous or autochthonous organic carbon, and weathering of carbonates. Dissolution of atmospheric CO2, directly, and most remineralization of organic carbon, indirectly, relate to atmospheric CO2. In systems where these sources of carbon dominate, such as small and medium-sized lakes (Post 2002), temporal correction for atmospheric change is appropriate. Pace et al. (2004) suggested that 22–50% of dissolved organic carbon in lakes is derived from terrestrial origins, and if differences exist in assimilation rates of atmospheric CO2 between terrestrial inputs and autochthonous production, temporal δ13C correction may be complicated in aquatic systems. Further, carbon derived from weathering of carbonates (e.g., limestone), does not relate to current atmospheric conditions, and in large lake systems where substantial amounts of assimilated DIC may originate from this carbon source (Post 2002), direct application of our correction factor is not appropriate. Related to these potential complications, we recommend additional research on applicability of our proposed isotopic correction for aquatic systems, specifically in relation to turnover time, equilibration rates, and multiple sources of carbon input.

The ability to use stable isotope analysis to explore human-induced changes in food-web dynamics, predator–prey interactions, patterns of migration, transfer of nutrients between ecosystems, and animal behavioral ecology as well as alterations of structure and function of ecosystems (Bergengren et al. 2001; Hibbard et al. 2001) has the potential to provide insights into ecological processes that alter regional patterns of biodiversity (Petchy et al. 1999; Post et al. 1999). Ability to correct tissue samples for changes in isotopic baseline resulting from long-term alteration of atmospheric CO2 δ13C will facilitate analyses that span a large temporal range. Although effects of increased variability from the correction factor may be problematic in some cases, observation of negligible variation in our case is encouraging. Potentially confounding effects related to carbon turnover rate and, in aquatic systems, variable input of carbon sources, warrant additional research relating to appropriate temporal correction. In general, because of the pervasive nature of change in atmospheric CO2 δ13C, we recommend controlling for this source of variation for any 13C study in which tissue samples span even a moderate time frame.