USGS visual identity mark and link to main Web site

USGS Open-File Report 94-023

Pliocene Pollen Data Set Dynamics: Tulelake, California, And Lost Chicken Mine, Alaska

David P. Adam
U.S. Geological Survey, Menlo Park, CA 94025
Pollen data sets of Pliocene age from continental sites are relatively rare. Pollen work with a paleoecological slant has focused primarily on deposits of Holocene and Upper Quaternary age, which are common and readily sampled in many regions. Stratigraphic palynologists, in contrast, have focused on pre-Pliocene deposits; most of the pollen grains encountered in Pliocene deposits cannot be distinguished from modern forms, and the simple presence or absence of particular forms is not as informative as it is for older deposits.

Although many or most of the pollen types important in Pliocene deposits represent plants still extant, the associations formed by those plants in the past may well have been different than those of today. Even if Pliocene pollen assemblages have modern analogs, those analogs often represent geographical ranges far removed from the Pliocene sites. For example, poorly-preserved pollen samples from the Pliocene Santa Clara Formation near San Francisco contain up to 50% spruce (Picea) pollen; spruce does not occur in the area today (Adam et al., 1983).

Pollen records from the Pliocene often span rather long time intervals; the longer the interval under study, the more work is required to achieve fine time resolution. Time series are thus often severely under sampled compared to Holocene studies, with sampling intervals measured in millennia rather than centuries or decades. Pollen zones then tend to be characterized not only by the frequencies of the various pollen types within them, but also by the nature of the variability within and between types.

Because pollen assemblages from Pliocene deposits often suggest climatic conditions significantly different from those of today, it is important to understand not only the nature of the assemblages, but their underlying dynamic structure. Assuming that the sequence of pollen samples through time is known, the dynamics can be investigated by plotting the samples in stratigraphic order in some appropriate phase space, chosen so as to clarify the patterns present in the data. This approach to past vegetation dynamics is illustrated below using data from Tulelake, California, and Lost Chicken mine, Alaska.


The Tulelake record is from a 334-m core that spans the past 3 Myr (Adam et al., 1989, 1990). Initial inspection of the pollen diagram suggested that the behavior of pine pollen vs. TCT (Taxodiaceae, Cupressaceae, and Taxaceae) pollen varied through time in a significant way. A plot of pine vs. TCT pollen for the entire section (1A) was not helpful, but a plot of the Pliocene part of the core (1B) indicated a well-developed pattern of variability between the two types. When stratigraphically adjacent samples are connected with lines (Figure 1C), the nature of the underlying dynamics of the data set emerge. Nearly all of the connecting lines have a negative slope, indicating that when pine increases, TCT decreases and vice versa. Only rarely do both types increase or decrease together.

Figure 1. Plots of pine pollen percentage vs. TCT (Taxodiaceae, Cupressaceae, and Taxaceae) pollen from Tulelake, Siskiyou County, California.
This figure is available as a GIF, PICT, or TIFF (line-art) image. The three parts are:
Comparison of the pattern of pine vs. TCT pollen discussed above with modern data from central California suggested that the late Pliocene climate at Tulelake had much in common with the present climate at middle elevations along the western slope of the Sierra Nevada near Yosemite, and with the climate that prevailed at lower elevations in the northern California Coast Ranges during the cooler parts of deep-sea oxygen isotope Stage 5 (Adam et al., 1990). Comparison of the various sites was greatly facilitated by the phase-plot technique illustrated here.

Lost Chicken Mine

The Lost Chicken Mine is a working placer mine along the Taylor Highway in east-central Alaska (see also Ager, this volume). It includes exposures of plant- and mammal-bearing fluvial sediments and peats of upper Pliocene age, as indicated by the Lost Chicken tephra layer, with a glass isothermal fission-track age of 2.9 0.2 Ma (John Westgate, oral communication). Thirty-seven pollen samples were analyzed from various stratigraphic units; age relationships between samples are sometimes clear but sometimes inferred. The Pliocene samples apparently represent a warm interval at or near the end of the Gauss Normal paleomagnetic chron; the Holocene samples all predate the White River Ash, which has an age of 1200 years (Pw, 1975).

Figure 2. Plot of Lost Chicken, Alaska, pollen samples against Detrended Correspondence Analysis axis 1 and 2.
This figure is available as a GIF, PICT, or TIFF (line-art) image.
Plot symbols represent various stratigraphic units, numbers are sample numbers (some numbers are "missing"), and arrows indicate direction of stratigraphic succession (older to younger).
The pollen counts were reduced to a condensed data set and subjected to a detrended correspondence analysis (DCA; Gauch, 1982). The DCA reduced the data set to four primary orthogonal axes that summarize the variability within the data. Because each axis reflects patterns of variability that apply to the entire data set rather than just to individual variables, shifts in the behavior of the data with respect to these axes are likely to reflect changes in the dynamics of the underlying climate system than simple plots of one taxon versus another.

The pollen samples are plotted against the first two DCA axes in 1, with samples from the same stratigraphic unit in known stratigraphic order connected by lines. The Holocene samples (29-33) are clearly set apart from the other samples, which are all Pliocene in age. In addition, inspection of the sequence of samples through time suggests that the Pliocene samples represent two separate dynamic regimes: (1) Axis 1 and Axis 2 scores positively correlated, and (2) Axis 1 and Axis 2 scores negatively correlated.


Changes in the vegetational dynamics recorded in the Tulelake and Lost Chicken data sets are reflected in the phase plots shown in figures 1 and 2 by changes in the slopes of the lines connecting stratigraphically adjacent points. Particular dynamic regimes are represented by elongate clouds of points in the phase space. These clouds are roughly linear in the examples selected, but curvilinear clouds may also occur. The character of a cloud representing a single regime should be independent of the time intervals between samples; this property is useful in dealing with Pliocene data sets that usually lack detailed time control. Shifts in the location or orientation of the cloud are best identified through inspection of the sequence of lines connecting the points in stratigraphic order; the shifts show up as abrupt changes in slopes of the lines.

The method illustrated here is intended to augment rather than supplant standard methods of interpretation. It provides a way to characterize groups of samples that covary, and to contrast them with adjacent groups. When samples are spaced relatively far apart compared to the frequencies of the underlying climate signals, such characterizations may provide useful insights into the behavior of the record.


This page is <>
Maintained by Eastern Publications Group Web Team
Last modified 28-Feb-2001