The Changes in North American atmospheric circulation patterns indicated by wood cellulose

INTRODUCTION

Climate change is often manifested by changes in atmospheric circulation patterns and the global hydrologic cycle. Several general circulation model simulations have suggested that during the Last Glacial Maximum (LGM), the circumpolar vortex intensified and enlarged, a glacial anticyclone developed over the Laurentide Ice Sheet, and the position of the Northern Hemisphere jet stream was shifted southward relative to its current position (e.g., Bromwich et al., 2004; Kutzbach et al., 1993). Such a change in the general circulation pattern would have been reflected by changes in surface wind patterns, temperature, and moisture conditions in North America. Modeled changes in temperature and precipitation can be verified with paleoclimate data, such as pollen or lake-level records (Bartlein et al., 1998; Prentice et al., 1991; Webb et al., 1998). However, there are few reports on wind patterns (Sweeney et al., 2004). Here we report an investigation of changes in wind patterns and moisture transport pathways between the Holocene and the last glaciation (LG) in North America between 40° and 50°N using oxygen and hydrogen isotopic compositions in radiocarbon-dated wood cellulose.

BACKGROUND

Isotopic Distribution Across a Continent in Relation to Wind Patterns

It has been well established that oxygen and hydrogen isotopic compositions in precipitation contain information about atmospheric conditions along the moisture transport pathway. Precipitation near the moisture source has higher δ¹⁸O and δD values than farther from the source (e.g., Dansgaard, 1964). Thus, the spatial distribution of precipitation δ¹⁸O across a continent may reveal information about the moisture sources and transport pathways.

We first examine modern precipitation in North America between lat 35° and 55°N. Currently, the surface zonal winds of this area are dominated by the westerlies (Peixoto and Oort, 1992). Marine moisture is advected onshore either by the westerly winds on the west coast, or by the onshore wind sector of extratropical cyclones on both coasts. In addition, moisture is brought to North America from the Gulf of Mexico, although its influence decreases with latitude (Welker, 2000). Thus, as expected, the latitude-corrected precipitation δ¹⁸O is higher near the east and west coasts than inland (Fig. 1A). In contrast, in South America between lat 0° and 30°S, moisture is brought onshore mainly from the east coast by advection of easterlies in both the Trade Winds and the South American summer monsoon (Vuille and Werner, 2005), but not by extratropical cyclones. The latitude-corrected δ¹⁸O shows a general decrease from east to west along the moisture transport pathway (Fig. 1B), although there is a significant scatter due in part to the vagaries of the monsoons and the intertropical convergence (ITC).

Oxygen and Hydrogen Isotopes in Tree Cellulose

After three decades of research, we have learned that the isotopic compositions of oxygen (δ¹⁸O_C, with C representing cellulose) and nonexchangeable hydrogen (δD_CN, with CN representing cellulose nitrate) in wood cellulose are mainly determined by the isotopic composition of source water and the relative humidity of the air during the growing season (e.g., Edwards and Fritz, 1986; Epstein and Yapp, 1977; Roden et al., 2000), although many other factors, such as soil water evaporation, air temperature, and the degree of postphotosynthetic isotopic exchange introduce noise into the isotopic signal (Roden et al., 2000; Tang and Feng, 2001). Because the isotopic composition of source water can be related to the hydrologic cycle, δ¹⁸ O_C δD_CN and values have been used as paleo-climate proxies (e.g., Feng and Epstein, 1994; Yapp and Epstein, 1977). Shu et al. (2005) suggested that, by comparing the slope of the δD_CN versus δ¹⁸ O_C plot with that of the δD versus δ¹⁸O or δ¹⁸O plot of the corresponding source water, one can obtain the covariation between the δ¹⁸O (or δD) of source water and the relative humidity. For a given region, if the δ¹⁸O of precipitation is not correlated with the relative humidity, the slope of the δD_CN against δ¹⁸O_C plot should inherit the slope of the meteoric water line (close to 8). If, however, the precipitation δ¹⁸O is positively correlated with the relative humidity, the δ¹⁸D_CN versus δ¹⁸D_C plot should have a greater slope than does the δD versus δ¹⁸O plot of precipitation (if the precipitation δ¹⁸O is negatively correlated with the relative humidity, the δ¹⁸D_CN versus δO_C plot should have a smaller slope) (see Fig. 2).

Figure 3 contains two comparisons between the slope of δD versus δ¹⁸O in precipitation (the meteoric water line, or MWL) and that of δD_CN versus δ¹⁸O_C in wood cellulose (the cellulose line). The precipitation data in Figure 3A are the precipitation-weighted annual average meteoric δ¹⁸O from North America. The cellulose data are from wood samples widely distributed on the North American continent, from Florida to Alaska, modern or ancient, and thus there is no systematic covariation between the δD or δ¹⁸O in the source water and the local relative humidity. The slope of the cellulose line (8.4) is statistically indistinguishable from that of the MWL (8.2). Figure 3B is redrawn from Shu et al. (2005), who studied the relationship between the local MWL and the cellulose line along the west-to-east trajectory of Pacific moisture in the Olympic Mountains, Washington, USA. The relative humidity and the δ¹⁸O (and δD) of the orographic precipitation both decrease along this trajectory (they are positively correlated). As a result, the cellulose line has a slope of 17.4 ± 1.2, significantly greater than that of the local MWL (9.0 ± 0.3).

SAMPLES AND METHODS

The concepts of isotopic distribution in meteoric water and in wood cellulose are used in this study. We obtained wood samples from Samuel Epstein's laboratory at the California Institute of Technology. Samples from lat 40° to 50°N in North America were separated into two groups by age: Holocene samples younger than 10 ka, and LG samples from 14–36 ka. Sample locations and their radiocarbon dates are listed in Table 1 (see also GSA Data Repository1). The δD_CN data have been reported for a subset of samples (Tang et al., 2000; Yapp and Epstein, 1977) and the remaining data are from Epstein's laboratory database. The δ¹⁸O_C values were determined in the Stable Isotope Laboratory of Dartmouth College. The ground wood was extracted for α-cellulose using a standard technique (Epstein and Yapp, 1976). In case the extracted cellulose was stained with iron oxide, the latter was removed using the Waller (1980) method and the purity of cellulose was checked with X-ray diffraction and Raman spectroscopy. The δ¹⁸O_C values of pure and dry cellulose were determined using a temperature conversion elemental analyzer interfaced with an isotope ratio mass spectrometer. The analytical precision is 0.3‰.

RESULTS

We examine the data in two ways, corresponding to the two concepts discussed. One is to obtain the slope of the δD_CN versus δ¹⁸O_C relationship and infer the covariation between δD or δ¹⁸O in the source water and the relative humidity during the growing season. The other is to examine the spatial variation of δ¹⁸O_C and δD_CN. Both pieces of information are used to infer moisture transport patterns on a continental scale.

Plots of δD_CN against δ¹⁸O_C are shown in Figure 4. For the Holocene, regression yielded a slope of 14.2 ± 2.1. This slope is significantly greater than 8.2 (p <0.01), the slope of the modern MWL in North America (Fig. 3A). This result indicates that, in the region where samples were collected, there was a positive covariation between the δ¹⁸O (or δD) of meteoric water and the growing season relative humidity during the Holocene.

In contrast, the slope of δD_CN versus δ¹⁸O_C for the LG period is 7.6 ± 1.1. If the MWL during the glacial period had a slope of 8, similar to that in the Holocene (Joussaume and Jouzel, 1993), the slope of the cellulose line is not significantly different from the slope of the MWL (p ∼0.34). This suggests that during the glacial period, there was little covariation between precipitation δ¹⁸O (or δD) and relative humidity at lat 40°–50°N.

In order to infer the direction and pathways of moisture transport, we plotted δD_CN and δ¹⁸O_C against longitude in Figure 5. Because there is a limited range of latitude among our sample sites, and the expected latitude effect is not present in the cellulose data in either age group, we made no latitude correction.

For the Holocene samples, both the δD_CN and δ¹⁸O_C have high values near the east and west coasts and lower values toward the center of the continent. The second-order polynomial curves explain 68% of the variance in δD_CN (p = 0.006) and 61% in δ¹⁸O_C (p = 0.02), and both second-order coefficients are highly significant (p <0.002). This isotopic pattern suggests that during the Holocene, moisture was carried inland from both east and west coasts. For the LG data, the curvature is much less well defined. While the overall regressions remained highly significant (p <0.002), the coefficient of the second-order term is statistically insignificant for either δ¹⁸O_C (p = 0.86) orδD_CN (p = 0.22). In contrast to the Holocene, the LG isotopic pattern suggests that moisture came predominantly from the east coast and the influence of the west coast moisture was significantly weakened and restricted to the coastal region.

DISCUSSION

The isotopic data of North American wood cellulose from lat 40° to 50°N for the LG and Holocene time give two contrasting impressions. The first difference is that a positive covariation between meteoric δ¹⁸O (or δD) and relative humidity existed during the Holocene, but not the LG period. Second, during the Holocene, moisture advected onto the continent from both coasts and moved inland, while during the LG, moisture came predominantly from the east coast with a much smaller amount of moisture moving a short distance inland from the west coast.

The isotopic variation with longitude in the Holocene samples (solid curves in Fig. 5) is similar to the isotopic variation with longitude in modern meteoric water (Fig. 1A), suggesting that the longitudinal change in the δ¹⁸O (or δD) of wood cellulose is determined by zonal moisture transport consistent with the modern westerly wind zone of North America. This moisture transport pattern also explains the covariation between the precipitation δ¹⁸O and the relative humidity, because coastal sites having high precipitation δ¹⁸O also have relatively high precipitation amounts and relative humidities. This is confirmed by the fact that latitude-corrected precipitation δ¹⁸O is positively correlated with growing season (June–August) relative humidity (r² = 0.63; p <0.0001).

The δ¹⁸O_C in the glacial wood shows a more or less monotonic decrease from east to west. This suggests that moisture from the east coast dominated precipitation at all longitudes, except for the areas very close to the west coast, where the weak influence of Pacific moisture is suggested by the data. One possible cause of δD_CN such a change in the moisture source is that the prevailing easterlies expanded equatorward, into the study region, due to the southward shift of the jet stream and the development of the Laurentide glacial anticyclone under the powerful influence of continental glaciation. A southward shift of the storm tracks was associated with this equatorward displacement of the prevailing easterlies, leaving the study region north of the tracks affected by the synoptic-scale easterlies of the storm systems. Vegetation records from the Puget Lowland, west of the Cascade Range in Washington State, at a longitude comparable to our westernmost samples in the Pacific Northwest, show a drier climate during the LGM than the present day (Barnosky, 1981; Thompson et al., 1993). This record is thus consistent with the influence of easterly winds during glacial time. Sweeney et al. (2004) reported evidence for the influence of the glacial anticyclone that weakened prevailing south-southwesterly winds and reduced eolian deposition of the Columbia Plateau. Several modeling results with the LGM boundary conditions have shown prevailing easterly winds in the Pacific Northwest (Bartlein et al., 1998; Bromwich et al., 2004), in contrast to present-day westerlies (Bartlein et al., 1998; Kutzbach et al., 1993).

A good analog for the LG North America does not exist in the modern environment. However, northern South America is somewhat comparable because its main source of moisture is also from easterlies off the Atlantic. The longitudinal distribution of precipitation δ¹⁸O in tropical South America (Fig. 1B) shows a pattern similar to that of δ¹⁸O_C in our North American glacial wood, in that the δ¹⁸O is highest near the east coast and decreases with longitude to the west.

It is not clear what was responsible for the lack of covariation between the precipitation δ¹⁸O and relative humidity in North America during the LG, as implied by the slope of the cellulose data (Fig. 4B). One possibility is the presence of a cool littoral climate near the west coast during the LG, where offshore easterlies and coastal upwelling caused high relative humidity (and fogginess) coupled with low precipitation (e.g., the present climates of Lima and Casablanca). In such a scenario, the positive correlation between precipitation δ¹⁸O and relative humidity on the east side of the continent would have canceled the effect of the negative correlation on the west side. One way to test this hypothesis is to see if the δD_CN versus δ¹⁸O_C plot (Fig. 4B) exhibits concave curvature such that the right end of the curve with higher slope corresponds to the east coast, and the left end with lower slope corresponds to the west coast. Our data appear to be consistent with this pattern, although the number of points is too limited to yield sound statistics.

We have not discussed absolute changes in either isotopic composition (Yapp and Epstein, 1977) or relative humidity. While the δD_CN has a similar range between the LG and Holocene time, the δ¹⁸O_C seems to have higher values near the east coast during the LG (Fig. 5A). A quantitative explanation of this difference would require knowledge of changes in isotopic composition of ocean water and local precipitation, temperature, and relative humidity. Although such an inference would be very useful, it is much more difficult to accomplish and less certain than obtaining relative spatial trends, which we focused on in this contribution. In the future, full inversions of cellulose isotopic values to precipitation δ¹⁸O or δD, relative humidity, and temperature may be possible with data sets having better spatial coverage and when tied to other paleoclimate data and climate model simulations of isotopic tracers.