Author for correspondence: Joy K. Ward Tel: +1 785 864 5229 Email: firstname.lastname@example.org
•While studies of modern plants indicate negative responses to low [CO2] that occurred during the last glacial period, studies with glacial plant material that incorporate evolutionary responses are rare. In this study, physiological responses to changing [CO2] were compared between glacial (La Brea tar pits) and modern Juniperus trees from southern California.
•Carbon isotopes were measured on annual rings of glacial and modern Juniperus. The intercellular : atmospheric [CO2] ratio (ci/ca) and intercellular [CO2] (ci) were then calculated on an annual basis and compared through geologic time.
•Juniperus showed constant mean ci/ca between the last glacial period and modern times, spanning 50 000 yr. Interannual variation in physiology was greatly dampened during the last glacial period relative to the present, indicating constraints of low [CO2] that reduced responses to other climatic factors. Furthermore, glacial Juniperus exhibited low ci that rarely occurs in modern trees, further suggesting limiting [CO2] in glacial plants.
•This study provides some of the first direct evidence that glacial plants remained near their lower carbon limit until the beginning of the glacial–interglacial transition. Our results also suggest that environmental factors that dominate carbon-uptake physiology vary across geologic time, resulting in major alterations in physiological response patterns through time.
The last glacial period began c. 110 000 yr ago and reached a maximum for global ice volume at 18 000–20 000 yr ago. Glacial conditions persisted (except for brief interstadials) until the abrupt transition to the current interglacial period, beginning c. 14 000 yr ago. At the peak of the last glacial period, atmospheric CO2 concentrations ([CO2]) ranged between 180 and 200 ppm, which are among the lowest concentrations that occurred during the evolution of land plants (Berner, 2006; also see Pagani etal., 2009 for an account of similarly low concentrations c. 15 million yr ago). When grown at glacial vs modern [CO2], modern C3 plants show 40–70% reductions in photosynthesis and biomass production (Polley etal., 1993; Sage & Coleman, 2001), 20–30% lower survival (Ward & Kelly, 2004), and may even fail to reproduce (Dippery etal., 1995). This is a result of reduced CO2 substrate concentrations at carboxylation sites, as well as higher photorespiration rates. However, even at reduced paleo-temperatures where photorespiration is decreased, plants are still unable to overcome the severe, negative effects of low [CO2] (Ward etal., 2008). Such pronounced effects originating at the level of autotrophic physiology have been modeled at the ecosystem scale, and have been predicted to greatly reduce net primary production and carbon storage during glacial periods (Turcq etal., 2002; François etal., 2006). Admittedly, however, modern plants are often grown in glacial conditions for only a single generation, and therefore do not reflect evolutionary responses to low [CO2]. This realization prompted our recent studies of glacial Juniperus (juniper) trees that were fully preserved within the La Brea tar pits in southern California (Los Angeles), and that had tens of thousands of years to adapt to low [CO2]. Analysis of stable carbon isotope ratios of complete wood samples (that integrate all tree rings from a given individual) show severely reduced internal [CO2] (ci) during the last glacial period that are unprecedented in modern equivalents, strongly suggesting the existence of major carbon limitations on tree physiology (Ward etal., 2005). Thus, the low [CO2] of glacial periods likely produced a bottleneck on carbon exchange through reduced transfer of CO2 from the atmosphere to the biosphere. However, it is still unknown if low [CO2] presented an overriding limitation on plant physiology relative to other climatic factors (e.g. water, temperature).
The last glacial period represents an ideal time period for addressing this issue, since climate was more variable on an interannual basis and [CO2] was exceptionally low relative to modern times (Mayewski etal., 2004). More specifically, ice cores from Greenland indicate extreme stability of Holocene climate compared with that of the last glacial period (Dansgaard etal., 1993). In addition, Dansgaard–Oeschger (D–O) cycles, which are periods of rapid and abrupt changes in temperature, dust content, ice accumulation and greenhouse gas concentrations, were more prominent during the last glacial period relative to the Holocene (Broecker, 1994; Roy etal., 1996). These patterns recorded in Greenland ice are also documented in ocean sediment cores from the Santa Barbara Basin (Behl & Kennett, 1996; Heusser, 1998; Hendy & Kennett, 1999; Hendy etal., 2002), c. 100 km northwest of La Brea, our primary research site. Analyses of these cores show a strong teleconnection between atmospheric trends over Greenland and ocean dynamics off the California coast, identifying synchronous climatic events between the two records over the last 60 000 yr (Behl & Kennett, 1996; Heusser, 1998; Hendy & Kennett, 1999; Hendy etal., 2002).
In previous work, we did not have wood specimens that allowed for carbon isotope analysis of individual tree rings. Recent excavations at the La Brea tar pits have now yielded higher-quality Juniperus specimens, allowing for discernment of individual tree rings. Thus, these wood specimens make an excellent model system for testing the constraints of low [CO2] on tree physiology relative to the effects of other climatic factors during the last glacial period. Here we compare long-term responses of tree physiology, as well as interannual variation within individuals, between the last glacial period and modern times. In doing so, we find the first evidence that low [CO2] constrained the physiology of glacial trees, as evidenced by a dampened response to interannual climate variability.
Materials and Methods
For this study, glacial trees from the Rancho La Brea tar pits (Los Angeles) were sampled and 14C dated to 14.5–47.6 kyr before present (BP), with the majority of specimens dating to the last glacial period. Juniperus samples from Rancho La Brea cannot be identified to the species level, although analysis by a wood anatomy expert (Ward etal., 2005) and species distributions indicate these samples are either J.californica or J.occidentalis. Cores of modern trees were collected from three low elevation sites in the Angeles National Forest (J.californica, two trees per site, one core per tree) and three high elevation sites in the San Bernardino National Forest (J.occidentalis, three trees per site, one core per tree), which are close in proximity to La Brea. Only modern trees from natural areas with well-drained, nonirrigated soils were sampled. Low-elevation sites provided a same-site control for glacial La Brea (with the full suite of environmental changes through time), whereas high-elevation sites controlled for lower temperatures and higher precipitation of the last glacial period (see Table 1, Heusser, 1998; Daly etal., 2008), allowing for isolation of CO2 effects. Note that conditions at glacial La Brea were wetter than modern times, which differs from most regions that were drier during the last glacial period. While [CO2] does not vary with elevation, CO2 partial pressure decreases in proportion to total atmospheric pressure. Under modern conditions, partial pressures of CO2 at high-elevation sites are 10–30% lower than at low-elevation sites, producing an even more conservative comparison between glacial and modern conditions.
Table 1. Climate data for glacial and modern Juniperus sampling sites
Site name and coordinates
Mean annual precipitation (mm)
Mean annual temperature (°C)
Climate data for glacial La Brea (Heusser, 1998), and modern sampling sites (Daly etal., 2008). Modern samples were collected from San Bernardino National Forest (SBNF) and Angeles National Forest (ANF).
La Brea Tar Pits
+ 34°3′48″, − 118°21′22″
Modern high elevation (SBNF)
Big Bear Lake
+ 34°16′12″, − 116°55′29″
Hwy 38 Bend
+ 34°11′35″, − 116°47′6″
+ 34°9′52″, − 116°43′14″
Modern low elevation (ANF)
Mt. Emma Rd
+ 34°28′55″, − 118°4′2″
+ 34°29′42″, − 118°1′36″
+ 34°11′22″, − 117°26′11″
Stable isotope measurement
We measured stable carbon isotope ratios on alpha-cellulose from individual tree rings of glacial and modern Juniperus. Whole tree rings were analyzed in order to provide an integrated measure of the full annual response. Ring wood was separated under a dissecting microscope and alpha-cellulose was extracted from each ring using the method described by Ward etal. (2005). Previous work using this method has documented high purity levels of alpha-cellulose with no indication of asphalt contamination from the tar pits. Purity was based on theoretical O : H ratios (weight percent oxygen : weight percent hydrogen) of 7.79–8.08 for alpha-cellulose, with actual values falling well within this range (8.01 ± 0.02 and 7.97 ± 0.04 for modern and glacial samples, respectively, Ward etal., 2005). Because our specific compound reflected high purity levels, we do not believe that diagenetic processes would have influenced our results.
Of the five glacial wood specimens that were available with an adequate number of tree rings, three are trunk specimens while two may be either portions of the trunk or large branch sections. For modern trees, the 10 rings nearest the center were excluded, as is common on dendrochronological work, as the juvenile stage often exhibits altered physiological patterns. Apart from this exception, all available rings in all samples were analyzed. Isotope measurements were performed at the Keck Paleoenvironmental and Environmental Stable Isotope Laboratory (KPESIL) at the University of Kansas. δ13C values were calculated using the following formula:
where R is the ratio of 13C : 12C, using belemnite carbonate from the Pee Dee Formation, Hemingway, SC (PDB) as the standard. Data were converted to ‘per mil’ (‰) notation by multiplying δ values by 1000. δ13Ccell was converted to δ13Cleaf using a constant offset of –3.2‰ (Leavitt & Long, 1982; Ward etal., 2005). Carbon isotope discrimination was calculated as:
Conversion to carbon discrimination is necessary as it incorporates changes in δ13Cair through time. δ13Cair was −0.0066 (−6.6‰) during glacial times, but has decreased in modern times to −0.008 (−8.0‰; Leuenberger etal., 1992). From Δ, ci/ca was calculated as
where a is the fractionation against 13C as a result of slower diffusion across the stomata (4.4‰) and b is the fractionation against 13C as a result of Rubisco (27‰).
For each ring, ci was also calculated from the ci/ca ratio using ca values. For modern samples, ca values were obtained from direct atmospheric measurements (Keeling etal., 2009) and the Taylor Law Dome ice core (Etheridge etal., 1996). For glacial trees, ca values were obtained from the Vostok and EPICA Dome C ice cores (Lüthi etal., 2008). To obtain the appropriate ca values, 14C ages of glacial trees were first converted to calendar ages in order to coincide with ice core data (Beck etal., 2001). Since atmospheric [CO2] showed only minimal changes throughout the latter portion of the last glacial period that is encompassed in our study, we are confident that [CO2] values corresponding to converted ages are accurate to the actual conditions experienced by glacial trees.
Mean ci/ca values for high- and low-elevation modern Juniperus were not significantly different despite environmental differences between these locations (0.53 ± 0.05 and 0.53 ± 0.06, P = 0.1; ANOVA), and therefore, the two modern sets were grouped together for comparison to glacial values. Since the variance in ci/ca was significantly different between modern and glacial Juniperus (P < 0.0001), a Welch’s ANOVA was used to compare modern and glacial ci/ca values that account for lack of equivalence of variance.
The coefficient of variation (CV) was calculated for ci/ca in both modern and glacial samples. CV provides a measure of dispersion of data around the mean, allowing us to compare variation between groups. CV was calculated as:
where s is the standard deviation, and is the mean. Data are shown in percentage notation by multiplying CV by 100. In order to account for differences in chronology length between glacial (shorter) and modern (longer) samples, the following correction (Sokal & Rohlf, 1995) was applied to CV:
Correlation of modern ci/ca with climate
To determine correlations of modern ci/ca values with climate, monthly temperature and precipitation data were obtained for each site from PRISM (Daly etal., 2008). The PRISM model is ideal for this comparison as it accurately reflects climatic conditions in mountainous coastal regions with large elevational gradients and complex topography (Daly etal., 2008). Measures of temperature and precipitation alone provided only weak correlations with ci/ca, so vapor pressure deficit (VPD) was used for this correlation. VPD is a more integrative climatic parameter that combines water and temperature relationships and is closely linked to evapotranspiration, making this measure more directly related to plant physiology than temperature or precipitation alone. VPD was calculated from monthly average maximum (Tmax), minimum (Tmin) and dewpoint (Tdew) temperatures using:
where es(T) is the saturation vapor pressure at temperature T, calculated as:
In order to correlate ring isotopic composition with VPD, rings of modern trees were associated with specific calendar years. Ring width patterns from trees within the same site were correlated and aligned using marker years of high precipitation and growth.
Results and Discussion
The ci/ca ratio is driven by two fundamental processes: stomatal conductance, which controls the rate of CO2 diffusion from the atmosphere into the intercellular spaces of leaves; and chloroplast demand for CO2, which is determined by internal CO2 diffusion rates to carboxylation sites and photosynthetic biochemistry. Long-term trends in ci/ca over evolutionary timescales reflect the degree of coordination between processes affecting CO2 supply and demand within the leaf. In addition, shorter-term trends in ci/ca (e.g. annual rings) reflect integrated shifts in tree physiology in response to changing environmental conditions within the lifespan of a single individual.
We found that mean ci/ca of Juniperus was similar between glacial and modern trees (Fig. 1a; glacial average, 0.52 ± 0.02; modern average, 0.53 ± 0.05; P > 0.2). One possible explanation for this, although one not supported by the literature, is that both stomatal conductance and chloroplast demand for CO2 remained constant across this expansive period of [CO2] and climatic change. On the other hand, if only one of these factors predominantly changed through time, there would have been shifts in ci/ca, which were not observed here. It is therefore most likely that both stomatal conductance and chloroplast demand for CO2 were higher during the last glacial period, which would have enhanced CO2 uptake under limiting carbon conditions. When supply and internal demand for CO2 covary in the same direction, as has been observed even in highly disparate taxa (Franks & Beerling, 2009a), there are opposing effects on ci/ca, likely producing the stabilization effect observed here. When moving into the interglacial period, both stomatal conductance and chloroplast CO2 demand likely decreased, with the effect of saving water and nitrogen as CO2 became less limiting. In support of this idea, Ehleringer & Cerling (1995) hypothesized that ci/ca represents a metabolic set point that is maintained within species across time. In addition, increases in stomatal conductance are almost always observed in modern C3 plants grown at low [CO2] (Gerhart & Ward, 2010), and studies with glacial leaves show evidence for increased stomatal density and decreased stomatal size, which would have increased maximum stomatal conductance in the past (Beerling etal., 1993; Franks & Beerling, 2009b; but also see Malone etal., 1993 for responses of modern plants grown at low [CO2]). The wetter conditions of the last glacial period may have also provided increased nitrogen availability to support higher leaf nitrogen contents, which may have enhanced photosynthetic capacity.
Despite any physiological adjustments, ci values remained extremely low in glacial trees relative to modern trees as a result of consistently low ca throughout the last glacial period (Fig 1b; glacial average, 106 ± 6; modern average, 168 ± 20; P < 0.0001). Past studies have reported similarly low ci values in glacial needles of Pinus flexilis preserved in packrat middens (Van de Water etal., 1993; Beerling, 1994). When considering all available rings, the vast majority of glacial ci values fell outside the range of modern values. In fact, no modern trees experienced ci values below 114 ppm, and no glacial trees experienced values > 120 ppm, leaving only a narrow overlapping range. It is also interesting to note that ci values of glacial trees never fell below 90 ppm over an integrated annual period. This suggests that this may be a limiting concentration below which juniper trees may not maintain a positive carbon budget for basic physiological functions for survival (Campbell etal., 2005).
Interannual variation in ci/ca, represented by CVcorr, was significantly lower in glacial vs modern trees (P < 0.0002; Fig. 2). More specifically, low- and high-elevation modern trees showed CVcorr values of 8 ± 2% and 8 ± 3%, respectively. The only available Holocene specimen (14.5 kyr BP) showed an intermediate CVcorr value of 5%, while glacial specimens showed the lowest values, averaging 3 ± 1%. Furthermore, although modern trees show occasional, short-term periods of low interannual variation, these periods are rare. Glacial trees show consistently low variation in ci/ca in all cases. In fact, the two oldest glacial Juniperus samples (45.1 and 47.6 kyr old) correspond to the timing of D–O events recorded in Greenland glaciers (Blunier & Brook, 2001). The maintenance of low interannual variation in ci/ca, even during time periods of rapid and drastic environmental change that are characteristic of D–O cycles, suggests that the maintenance of low variation in glacial Juniperus physiology was consistent throughout the last glacial period.
Since glacial climate was much more variable than in modern times, one would expect glacial ci/ca to also show higher variation if trees were responding to similar climatic factors. To the contrary, we found reduced amounts of interannual variation in ci/ca during the last glacial period (Fig. 2), indicating that a stable environmental factor dominated tree physiology. During the last glacial period, [CO2] was extremely stable from year to year (EPICA, 2004), while water availability and temperature were predicted to have been highly variable (Behl & Kennett, 1996; Heusser, 1998; Hendy & Kennett, 1999; Hendy etal., 2002). In our study, extremely low ci values coupled with reduced variation in ci/ca, even under a highly fluctuating glacial climate, point strongly to low [CO2] constraints on tree physiology. While short-term studies with modern plants grown at glacial [CO2] show major carbon limitations on physiology, our findings highlight the strength and consistency of low CO2 constraints over evolutionary timescales.
In conclusion, this study has demonstrated that mean ci/ca has been maintained in Juniperus between the last glacial period and modern times, despite changes in temperature, precipitation and [CO2]; that glacial ci values were extremely low on an annual basis and occur only rarely in modern trees; that a limiting level for Juniperus physiology may exist at or near 90 ppm; and that interannual variation in ci/ca was greatly reduced in glacial Juniperus, likely as a result of the constraints of low [CO2] that overrode responses to other climatic factors. This is the first direct evidence from trees that actually lived and evolved under low [CO2] that carbon limitation persisted on an annual basis during the last glacial period. Moreover, our results suggest that the environmental factors that dominate carbon-uptake physiology can vary across geologic timescales, resulting in major alterations in physiological response patterns through time.
This work was supported by the National Science Foundation (NSF) CAREER (0746822) and PECASE awards, and the Wohlgemuth Faculty Scholar Award (endowed by D. Lynch) to J.K.W. The NSF C-CHANGE IGERT fellowship (0801522) and the Madison and Lila Self Graduate Fellowship supported L.M.G. at KU. The Department of Ecology and Evolutionary Biology (KU) also provided assistance for this work. The authors thank G. Cane (KPESIL, KU), E. Duffy (EEB, KU) and K. Colvin (EEB, KU) for assistance with isotope analysis, and J. Nickerman (Angeles National Forest) and M. Lardner (San Bernardino National Forest) for field permit assistance. The authors also thank three anonymous reviewers for their helpful comments.