KGS Home Geology Home Reports

Kansas Geological Survey, Open-file Report 96-35
Next page--Results

Variation in Opal Phytolith Assemblages as an Indicator of Late-Quaternary Environmental Change on Fort Riley, Kansas

by William C. Johnson and Steven R. Bozarth

Palynology Laboratory, University of Kansas
KGS Open File Report 96-35
July 1996

A Technical Report for
U.S. Army Construction Engineering Research Laboratory (USACERL)
in fulfillment of Contract No. DACA88-95-M-0422

The complete report is also available as an Adobe Acrobat PDF file.

Report (940 kB)

To read this file, you will need the Acrobat PDF Reader, available free from Adobe.


Environmental Change

The sensitivity of grassland community composition and its boundaries to short-term climatic variation during the historical period is well documented for the central Great Plains (Tomanek and Hulett, 1970; Küchler, 1972). Similarly, long-term prairie expansion and contraction, in response to climatic variation, is documented for the prehistoric time scale (e.g., Watts and Wright, 1966; Grüger, 1973; Bradbury, 1980; Webb et al., 1983). Despite the many published studies little is known, however, about the environmental conditions prevailing in the central Great Plains during the late Pleistocene and Holocene. Because the region has few bogs and natural lakes, there is a paucity of good sites for preservation of fossil pollen and botanical macrofossils. Therefore, climatic reconstruction in this region has traditionally relied heavily on the synthesis of other proxy sources such as vertebrate fauna, snails, and occasional botanical macrofossils. As a result, interpretations of late-Pleistocene vegetation of the region range, for example, from continuous taiga-like forest (e.g., Wells and Stewart, 1987) to grassland or steppe (e.g., Graham, 1987)

The quasi-continuously deposited loess of the central Great Plains represents some of the thickest and most complete loess deposits in North America and provides a largely untapped potential for reconstructing past climates. Ongoing research by this investigator and colleagues on magnetic records from these deposits indicates a tremendous potential for environmental reconstruction. For example, lower magnetic susceptibility of the loess associates with times of rapid accumulation, i.e., the cooler glacial intervals, whereas higher susceptibility associates with the intercalated paleosols and weathering zones, which represent times of landscape stability and/or reduced accumulation rates. In addition to correlation with long-term, regional cyclic climatic changes, the relatively rapid accumulation rates associated with loess deposits also permit the resolution of short-term changes in climate (e.g., Zhou et al., 1994; Heller et al., 1993; An et al., 1991, 1993).

Recent research using nonmagnetic parameters has also been particularly rewarding. Two approaches that are now being employed include stable isotope ratio analysis (SIRA), especially that of carbon, and biogenic opal analysis, namely that of phytoliths; these techniques have been very effective in yielding unique environmental information. The coincidental use of these three proxies of past environments provides a more comprehensive picture of past environments than use of only one parameter. Magnetic information, e.g., susceptibility, indicates times of weathering and soil development in the stratigraphic sequence; SIRA of carbon provides an overall impression of the type of climate to which the plants have adapted; and opal phytolith morphology permits identification of specific groups of plants.

Fort Riley

Geomorphological and geoarchaeological research on Fort Riley, Kansas for the U.S. Army Construction Engineering Research Laboratory (CERL) is being conducted to assist in the development of a dynamic paleoenvironmental model of late-Pleistocene and Holocene landscape evolution. Research began with an overview study by D.L. Johnson (1992). On-site work by D.L. Johnson, an assistant and the contractor in June, 1993 involved further reconnaissance, subsurface exploration with a vehicular-mounted coring rig, and documentation and sampling of natural exposures. Eighteen localities, numbered 21-38, were studied, with 16 yielding sediment and soil cores. These sites represented various landscape positions within Fort Riley. Of the sites which were cored, all except 21, 26, and 33 extended to bedrock or to the residual soil developed on the bedrock. Laboratory analyses were conducted on samples from cores 21, 24, and 25 by D.L. Johnson (1994) and subcontractors. Analyses included particle size determination (pipet, hydrometer, sieve, and elutriation), various wet chemical analyses (pH, cation exchange, organic matter, phosphorous, elemental ppm, base saturation), and 14C dating. Additional study has described the distribution of the loess mantle on the upland and refined the stratigraphy at the Sumner Hill locality, i.e., core site 21 (D.L. Johnson, 1996).

The first phase of the paleoenvironmental research was a test of the potential for SIRA (carbon) and opal phytolith analysis to provide a proxy time series of climate for the reservation (W.C. Johnson et al., 1994). Study was conducted on a core extracted from the Sumner Hill locality. Isotopic analysis of the carbon (δ13C) provided a time series that compares well with a composite of regional carbon isotope data, whereas the opal phytolith record was largely uninterpretable below about 3.5 m depth, i.e., for the lower 8.5 m of the core. On the basis of a recently derived chronostratigraphy of the site (D.L. Johnson, 1996; this study), the viable portion of the phytolith record does, however, include the time interval related to cultural occupation, i.e., the last 13,000 years or so.

A second phase of environmental reconstruction was designed to consider the record from several sites distributed over the reservation in upland and lowland environments. The research design was constructed as to make coincident use of magnetic analysis, SERA, and analysis of biogenic opal (opal phytoliths). This report presents the context, results and discussion of the opal phytolith analyses conducted.

Regional Late-Quaternary Loess Stratigraphy

An articulation of the regional loess stratigraphy is necessary for the appreciation of the late-Quaternary record preserved at Fort Riley. Considerable research has been conducted on these deposits in recent years, with an emphasis on development of the chronostratigraphy. Research results from Fort Riley are, however, making a major contribution to the data base and to the understanding of paleoenvironmental conditions for the central Great Plains.

A very good empirical relationship exists between cold stages in the marine oxygen isotope record and documented times of glaciation in the United States (Richmond and Fullerton, 1986b). Given the close correspondence between the glacial record and marine isotope record and the fragmentary nature of the former, it follows that the nearly continuous loessal record should be an excellent terrestrial cognate of the marine isotope record. Therefore, the climatic and chronological record assembled for the marine sequence should match the loessal record well. The generally accepted model relating climate to the loessal record indicates that periods of stability and pedogenesis are usually associated with warm interglacials, and periods of significant loess accumulation coincide with the colder glacial times (Kukla 1977, 1987). Morphologic and isotopic analyses of plant opal phytoliths from loess exposed at the Eustis ash pit in southwestern Nebraska support the model and the relationship with the marine isotope record for the Illinoian, Sangamon and early middle Wisconsin stages (Fredlund et al., 1985; Fredlund, 1993).

Pre-Illinoian Stages

Little is known of the pre-Illinoian loesses because far fewer exposures exist than of the Loveland (Illinoian) loess and certainly the Peoria (late-Wisconsinan) loess. Pedogenesis has been recognized in these early loesses, however. Zones of carbonate enrichment, occurring at about 410-360, 330-290, and 250-200 ka, within the Barton County sanitary landfill exposure were interpreted to be pedogenic in origin (Feng et al., 1994). These carbonate zones are likely analogous to the soils observed in the pre-Loveland loesses at the Eustis ash pit, Nebraska (Fredlund et al., 1985) and elsewhere in the region (Schultz and Martin, 1970; Frye and Leonard, 1951). The zones are temporally equivalent to the nonglacial, or warm marine isotope stages (Feng et al., 1994).

Fossil pollen evidence indicates that the grasslands of the Great Plains expanded during the interglacials and contracted, or perhaps disappeared, during the glacial periods (Kapp, 1965, 1970; Fredlund and Jaumann, 1987). This suggests that the carbonate enrichment was a product of grassland pedogenesis, not unlike that of today. Fredlund and others (1985) extracted the grass opal phytoliths contained within these multiple soil zones at the Eustis ash pit in south-central Nebraska and found that the soil-forming periods were warmer and periods of dust accumulation cooler.

Fort Riley. No pre-Illinoian Quaternary sediments have yet been recognized on Fort Riley as a result of this series of CERL-funded studies or are reported in the literature. It appears that major entrenchment and denudation preceded the Illinoian because exposures and cores to date have exposed only what is presumed to be Illinoian-age material overlying bedrock. For example, the Illinoian Sangamon soil, loess, and alluvium rest upon a strath, or bedrock-defended terrace at the Sumner Hill locality; conversely, middle Pleistocene loess rests on an erosion surface developed on limestone at the Bala Cemetery site.

Illinoian Stage

Loveland loess. The Loveland loess is the most widespread pre-Wisconsinan loess in the Midcontinent. Several investigators (e.g., Reed and Dreeszen, 1965; Ruhe, 1969; Willman and Frye, 1970; Ruhe and Olson, 1980) have described it throughout the Missouri, Mississippi and Ohio River basins. Further, it has been recognized south into Mississippi and Arkansas (McCraw and Autin, 1989). The Loveland has been far less studied (e.g., absolute chronology, geometry, mineralogical composition) than the Wisconsinan loesses, namely the Peoria.

The Loveland may be described as a yellowish-brown or reddish-brown eolian silt. Red hues increase toward the top of the formation due to development of the Sangamon soil within the uppermost Loveland. The thickest accumulations occur in the north-central part of the state: recorded thicknesses approach 15 m. A thining in the loess occurs both southward and westward such that the distribution becomes discontinuous to the southwest. In Kansas, the Loveland is typically less than 10 m thick, but produces a very distinctive mark on the landscape via its variation in stratigraphy. It occurs on uplands and valley side slopes. As a result, the Loveland and its capping Sangamon soil are well expressed in exposures, particularly freshly cultivated fields.

The absolute age of Loveland loess in Kansas is largely uncertain, but recent work at sections exposed in a Geary County quarry in northeastern Kansas, the Barton and Pratt County sanitary landfills of central Kansas, and the Eustis ash pit of southwestern Nebraska provided the first absolute-age information on the Loveland beyond that carried out at the paratype section. Oviatt and others (1988) reported TL ages of 136 ka and 130 ka for the upper part of the presumed Loveland loess exposed in an abandoned quarry near the town of Milford, immediately west of Fort Riley. TL age data from this and others sites in Kansas indicates that the Loveland loess began accumulating sometime before 130 ka (Feng et al., 1994; Johnson and Muhs, 1996). Recently, Maat and Johnson (1996) derived a TL age of approximately 160 ka immediately below the Sangamon soil in Loveland loess at the Eustis ash pit. Dating at the Loveland paratype section was the first attempt to employ the TL technique on loess in the Midcontinent, and results were consistent with the data obtained from loess at other localities in the North American Continent, including those in central Kansas. Four TL ages derived from the Loveland loess indicate the sediment was deposited approximately 140,000 ka. (Forman, 1990b).

Sangamon soil. This paleosol is strongly developed and occurs throughout the Midcontinent beneath deposits of the Wisconsinan glaciation and within deposits of the Illinoian glaciation or older deposits. The Sangamon soil has been recognized in Indiana (Hall, 1973; Ruhe et al., 1974; Ruhe and Olson, 1980), Illinois (Bushue et al., 1974; Follmer, 1979) where the type section is located (Follmer, 1978), Iowa (Simonson, 1941; Ruhe, 1956, 1969), Nebraska (Schultz and Stout, 1945; Thorpe et al., 1951) and Kansas (Frye and Leonard, 1952). In Kansas, the Sangamon soil is well expressed, occurring throughout the state. Although the soil has received considerable attention in northeastern Kansas (Frye and Leonard, 1949, 1952; Tien, 1968; Caspall, 1970; Bayne et al., 1971; Schaetzl, 1986), it has been recognized at many localities in the state (Bayne and O'Connor, 1968) and recently studied in central Kansas (Feng et al., 1994). Historically, it has been referred to as a "soil in the Sanborn formation" (Hibbard et al., 1944), the Loveland soil (Frye and Fent, 1947), and the Sangamon soil (Frye and Leonard, 1951). The color of the soil ranges from a vivid to pale reddish brown, with a loss in color occurring westward. Regionally, the soil character varies according to parent material, local drainage and climate which prevailed at the time of pedogenesis. The soil occasionally contains sufficient clay to create a subtle bench on cultivated slopes. Schaetzl (1986) noted that the soil appears to have been a very strongly developed Ultisol or Mollisol.

The Sangamon soil was first used in a time-stratigraphic context to differentiate deposits of the Illinoian and Wisconsinan glacial stages (Leverett, 1899). An appreciable time span for regional landscape stability and soil formation are suggested by oxidation, apparent deep leaching, and high clay accumulation. A major problem associated with the Sangamon soil is its diachronous upper and lower boundaries (Follmer, 1978, 1982, 1983). To further confuse the time element, the lower 1-2 m (3.3-6.6 ft) of the early Wisconsinan loess is typically weathered and forms a pedological continuum with the underlying Sangamon soil (Follmer, 1983), and early investigators mistakenly included the former in the Sangamon profile. The Sangamon should be considered a pedocomplex rather than a single soil which developed under a unique environmental condition (Schultz and Tanner, 1957; Fredlund et al., 1985; Morrison, 1987). It apparently represents two or more paleosols welded together to form a complex that reflects significant spatial and temporal variation in environmental conditions and an appreciable time span. Although laboratory data from exposures in central Kansas indicate the Sangamon soil was strongly weathered chemically, presumably under a warm, moist climate (Feng, 1991), recent data from the Eustis ash pit in south-central Nebraska indicate that the Sangamon soil is not very mature geochemically (Muhs and Johnson, 1996).

Because of apparent time transgressiveness, the age of the Sangamon soil is not precisely known. Follmer (1983) reported a radiocarbon age of 41,700±1100 yr B.P. on plant material from the top of the Sangamon in its type area in Illinois. Forman (1990a) reported TL ages of 140±20 and 70±10 ka from loess below the Sangamon soil at two separate sites in Iowa and Illinois, and concluded the Sangamon soil is diachronous and may consist of multiple soils. Feng (1991) and Feng and others (1994) reported a TL age of about 70 ka in the lowermost part of the Sangamon soil exposed in central Kansas and associate it with marine isotope stage 3. Although Richmond and Fullerton (1986b) assigned Sangamon time to 132-122 ka (isotope substage 5e), they acknowledged reported ages (relative and absolute) ranging from early Illinoian to middle Wisconsin. Basal ages on the overlying Gilman Canyon Formation from numerous locations in Kansas and Nebraska (Johnson, 1993a, b) provide a minimum age of about 45 ka for the Sangamon soil. Also, Forman (1990b) and Forman and others (1992) obtained TL and radiocarbon ages of 35-30 ka within the loess overlying the Sangamon soil at the Loveland paratype section in Iowa and the Pleasant Grove School section in Illinois.

Post-Sangamon time was one of extensive landscape instability including upland erosion, as evidenced by the partial or complete removal of the Sangamon soil. In a quarry near Woodruff in Phillips County, Kansas, the Loveland and Sangamon have been removed and the top of the Ogallala eroded. Similarly, the same units were apparently stripped and channels cut into the Smoky Hill Chalk prior to deposition of the Gilman Canyon Formation. Consequently, erosional truncation of the soil may be in part responsible for the apparent diachronous character of the soil. Deposition occurred at some locations: for example, a sandy zone overlying the Sangamon soil in the Phillips County sanitary landfill suggests a dry and windy transition to the Gilman Canyon Formation above. A similar unit has been observed by the author at the Eustis ash pit.

Although relatively little is known of the climate prevailing during Sangamon time, some recent research provides a first indication. In a regional examination of the stratigraphy representing Sangamon time, Dremanis (1992) noted that for the midwestern and eastern United States (the southeastern margin of the Laurentide ice sheet of North America) temperatures of the Sangamon climatic optimum were several degrees warmer and precipitation was less than at present. This time of maximum temperature and minimum precipitation, likely occurring during oxygen isotope stage 5e (c. 120ka), was followed by gradual cooling characterized by oscillations between cool and cold climates. Through global circulation model experiments, Harrison and others (1995) simulated regional climate and associated biome changes in North America for the two extremes in orbital parameters during isotope stage 5e. Maximum solar insolation, at about 125-126ka, produced midcontinental summer temperatures about 8°C warmer than today and a corresponding expansion of the warm season grasses. Conversely, minimum solar insolation, at about 115ka, depressed the summer temperatures 5°C below those of today, and expanded the cool season grasses and forests. Pedologic data obtained by Muhs and others (1996) support the notion of a climatic regime warmer and/or drier than at present.

Fort Riley. Thus far, investigations on the reservation have indicated that Illinoian age sediments are ubiquitous but relatively thin, and expressed primarily as the Sangamon soil developed within them. Best known expression of the Sangamon soil occurred at the Sumner Hill locality in a thin layer of Loveland loess overlying alluvium.

Wisconsinan Stage

Stratigraphy associated with the Wisconsinan glacial period in the central Great Plains consists of two loess units and associated soils: Gilman Canyon Formation and Peoria loess.

Gilman Canyon Formation. The Gilman Canyon Formation, first recognized in Nebraska (Reed and Dreeszen, 1965), is a middle to early-late Wisconsinan (cf. Farmdalian) loess. Equivalents of the formation have been recognized elsewhere: the Loveland loess is buried by the Roxana silt from Minnesota and Wisconsin to Arkansas and by the Pisgah Formation in western Iowa (Bettis, 1990). The Gilman Canyon of Nebraska and Kansas is typically dark in color, silty, leached of calcium carbonate, and heavily enriched in organic carbon via pedogenesis (melanization). As noted above, the formation was once considered to be the attenuated A horizon of the Sangamon soil (Thorpe et al., 1951; Reed and Dreeszen, 1965).

Reed and Dreeszen (1965) provide limited textural data and description of the Gilman Canyon Formation at the type section. Their description within the columnar section at the Buzzard's Roost exposures states (p. 62): "Upper 12 inches [31 cm] is medium dark gray, slightly humic, silt; middle 1 foot 1 inch [33 cm] is fight brownish-gray silt; basal 3 feet 8 inches [ 1. 12 m] is dark brownish-gray, humic, soil-like silt; entire thickness is noncalcareous ... 5 feet 9 inches [1.75 m]." Although all of these attributes described at the type section appear representative of the formation as observed in Nebraska and Kansas, the bimodal distribution of humus is curious: this suggests the existence of two periods of relative stability, or low accumulation rates, and an intervening period of accelerated accumulation rates. Consequently, the Gilman Canyon Formation often appears as one or more cumulic A horizons that are developed within a variably to noncalcareous loess, usually no more than 1.2 m thick. In a section revealing an expanded valley phase of the Gilman Canyon Formation, May and Souders (1988) recognized three distinct organic zones, each of which may represent a separate episode of pedogenesis. Two such zones have been recently observed by the author at the Eustis ash pit in south-central Nebraska. If two or more distinct periods of soil formation did indeed occur regionally, they are obscured at many localities, likely due to bioturbation. Overall, the formation reflects a sufficiently slow rate of loess fall (<.08 mm/yr) such that pedogenesis was operating more or less continuously, but with a decreased intensity at one or more times.

As expressed, the Gilman Canyon Formation is frequently overlain by .9-1.5 m of leached loess which is considered to be basal Peoria Formation. Correlative with the Gilman Canyon and overlying leached loess zone is the Citellus zone (a ground squirrel now recognized as the genus Spermophilus) of Nebraska (Condra and Reed, 1950). The leached zone is transitional between the well developed A horizon(s) in the Gilman Canyon and the calcareous Peoria loess above, and probably reflects a sufficiently slow accumulation of Peoria loess such that pedogenesis could keep pace only partially. A.B. Leonard (1951, 1952) supported the contention that the leached, or basal zone was slowly accumulating, early Peoria loess experiencing pedogenesis through inference that gastropods were originally present, but subsequently destroyed during weathering of the loess. Above the leached zone, the rate of accumulation of Peoria loess was sufficiently rapid (c. 0.6 mm/yr) as to preclude any soil development.

Radiocarbon ages from the Gilman Canyon Formation range from approximately 40 ka at the base to 20 ka at the top (May and Souders, 1988; Johnson et al., 1993a). The basal age of 40 ka agrees well with the time set by Richmond and Fullerton (1986a) for the beginning of the late Wisconsin. MWe Nebraska has many dated locations forming an arcuate pattern around the eastern and southern sides of the Sand Hills, data from Kansas are relatively limited. The ages in Kansas do show, however, good agreement from the south-central to the north-central part of the state and with those from Nebraska.

Given the radiocarbon time control and stratigraphic information currently available for the Gilman Canyon Formation within Kansas and Nebraska, it is clear that the associated soil(s) is a geosol, i.e. a laterally traceable, mappable, pedostratigraphic unit with a consistent time-stratigraphic position (Morrison, 1965; North American Commission on Stratigraphic Nomenclature, 1983, p. 865). The entire formation may be considered a geosol, but, because of the possibility for the existence of two or more identifiable cumulic A horizons merged or welded together, it may ultimately be considered a composite geosol.

Limited paleoenvironmental data are emerging for the Gilman Canyon Formation. δ13C values are a potential source of proxy data for vegetation type and hence climate (Krishnamurthy et al., 1982). When determinations are derived from the organic fractions of the soil, they reflect inputs by the plants, particularly grasses, growing on those surfaces. C3 (cool-season) species have an average δ13C composition of -27‰ and C4 (warm-season/arid) species -13‰, relative to the PDB standard (Deines, 1980). Terrestrial plant ecology of the Gilman Canyon Formation appears to have been characterized by primarily C4-type grasses, or a relatively warm, possibly dry climate. From plant opal phytolith morphology (Fredlund et al., 1985; Fredlund and Jaumann, 1987) and isotope data (Fredlund, 1993), it is evident that there existed a panicoid-dominated grassland, i.e., one of moist, temperate-adapted tall grasses. These data are not inconsistent with the δ13C values, since panicoid grasses are C4 types. Further, some of the C3-level values derived, specifically those from La Sena and Lime Creek sites, are reflecting former peaty or otherwise local, wet valley bottom environments (D.W. May, pers. comm.) which are characterized by C3 plants meso- or hygrophytic in habit.

Interpretation of a fossil pollen assemblage from a core extracted from Cheyenne Bottoms, a large marsh in central Kansas, indicates mesic conditions in the marsh and an upland vegetation of grass and sage with scattered trees in the valley and along escarpments during the period from approximately 30 to 25 ka (Fredlund, 1991). The Farmdalian-Woodfordian transition, approximately 25-24 ka, was characterized by increased aridity. The Muscotah Marsh fossil pollen record of northeastern Kansas reflects a mosaic of deciduous forest and prairie for the late Pleistocene (Grüger, 1973; Fredlund and Jaumann, 1987). Regionally, the Farmdalian grasslands were apparently found as far east as Iowa (Baker and Wain, 1985) and north to the Sand Hills region of Nebraska (Fredlund and Jaumann, 1987).

Peoria loess. Leverett (1899) first proposed the name Peoria for an interglacial period between the Iowan and Wisconsin glacial stages. When Alden and Leighton (1917) demonstrated the Peoria was younger than Iowan, usage shifted to that of a loess, rather than a weathering interval. Within the Midcontinent, several names have been used for post-Farmdalian loess. Ruhe (1983) prefers to use the term "late Wisconsin loess" because of the uncertainties in stratigraphic equivalency from one region to another. The Peoria Formation is typically an eolian, calcareous, massive, light yellowish-brown silt that typically overlies the Loveland Formation or an approximate equivalent of the Gilman Canyon Formation.

Ruhe (1983) notes three major features of late-Wisconsinan (Peoria) loess: it thins downwind from the source area, decreases in particle size systematically away from the source area, and is strongly time-transgressive at its base. The latter feature is unresolved and results in correlation problems. Ruhe (1969) realized a decrease in the age of the soil under the loess from 24,500 years B.P. near the Missouri River to about 19,000 years B.P. eastward across southwestern Iowa. A decrease from 25,000 to 21,000 years B.P. was noted for the base of the loess along a transect in Illinois (Kleiss and Fehrenbacher, 1973). The top of the loess also seems to be time-transgressive, ranging from about 12,500 years B.P. in Illinois (McKay, 1979b) to 14,000 years B.P. in central Iowa (Ruhe, 1969).

In Kansas, the Peoria is a reddish, yellowish, or tan buff color, homogeneous, massive, locally fossiliferous, variably calcareous, and ranges from coarse silt and very fine sand to medium to fine silt and clay (Frye and Leonard, 1952). Thicknesses vary from in excess of 30 m adjacent to the Missouri River valley to 0.6 m in discontinuous patches. Any accumulation less than 0.6 m is presumed unrecognizable in the field because it has become incorporated into the existing surface soil. Peoria loess typically rests conformably upon the Gilman Canyon Formation.

Despite the amount of attention given Peoria loess in Kansas, the source of the silt is not completely certain. Upon a review of the available data, Welch and Hale (1987) conclude that a single source was not likely for all loess deposits in Kansas, and that the loess was derived from a combination of three sources: glacial outwash river flood plains, present sand dune areas, and fluvial and eolian erosion of the Ogallala Formation. Research on trace element concentrations in loess (Johnson and Muhs, 1996) indicates, however, that the Platte River valley was the primary source, with secondary inputs from the major river valleys to the south (e.g., Republican, Smoky Hill, Solomon, Arkansas).

Although readily visible stratigraphic breaks such as the Jules soil recognized in Illinois (Frye and Willman, 1973; Frye et al., 1974; Ruhe, 1976; McKay, 1979a, b) and the soil zones in Iowa (Daniels et al., 1960; Ruhe et al., 1971) have not yet been identified in Kansas and adjacent Nebraska, evidence of one or more stable or vegetated surfaces is common. The only indication of soil development recognized is that of a Bt horizon in the Medicine Creek valley (May and Holen, 1993); interestingly, the soil has a probable Paleoindian association (May, 1991). The most common line of evidence for a discontinuity(ies) in Peoria loess deposition is that of plant remains, usually outcropping as lenses. Many of the age determinations were made from Picea remains, indicating a cool, moist environment. Although radiocarbon data document the burial of vegetative material throughout the Woodfordian, two temporal clusters or modes of ages appear from the limited data: one 18-17 ka and another 14-13 ka. The former time interval represents the last glacial maximum and the latter the time of major deglaciation (Ruddiman, 1987). Interpreting ice core data from Greenland, Paterson and Hammer (1987) record a dramatic decrease in atmospheric dust content from about 13,000; this period of reduced atmospheric dust may relate to the time of relative surface stability and tree establishment. Regional geomorphic data also support the existence of a hiatus at this time. May (1989), identifies deposition of the Todd Valley Formation in the South Loup River of central Nebraska at about 14 ka, which is subsequently buried by loess. Further, Martin (1990) identifies entrenchment in the Republican River of south-central Nebraska at about 13 ka, after which valleys were filled with late Peoria Loess.

Fort Riley. The Gilman Canyon Formation sediments and soil were present at all sites sampled for magnetic and isotopic analyses, indicating that the formation has both upland (loess) and valley (alluvium) phases preserved at Fort Riley, as elsewhere in the region. Radiocarbon ages on the formation range from about 19 ka to over 23 ka (table 1); these relatively young ages represent the most recent period of Gilman Canyon time pedogenesis; the relatively unaltered loess and alluvium was not sampled for dating.

Coring and backhoe trenching indicates that Peoria loess is typically relatively thin or undetectable. Only one of the magnetic and isotopic study localities exhibited clearly identifiable unaltered Peoria loess below the Brady soil.

Holocene Series

The beginning of the Holocene, about 10 ka (Hopkins, 1975), was a time of dramatic environmental change and attendant stratigraphic discontinuities. This boundary is generally considered only geochronometric, i.e., without specific stratigraphic reference, although a stratotype in Sweden has been proposed for the boundary (Marner, 1976) and has a reported age of 10,000±250 years B.P. (Fairbridge, 1983). Watson and Wright (1980) contended that major climatic and environmental change at 10 ka may be documented only on a local scale, i.e., all changes recorded in the stratigraphic record are diachronous. This notion now seems to be faulty on the regional and subcontinental scale in that research of the last decade has documented major pedogenesis at 10 ka in both alluvial and eolian/upland settings. This is the first major geosol to occur in the stratigraphic record of the region since the Gilman Canyon soil 1O,OOO years earlier.

Brady soil. The Brady soil was first named and described by Schultz and Stout (1948) at the Bignell Hill type locality, an eolian sequence exposed along a roadcut in the south valley wall of the Platte River of western Nebraska. The soil is developed within the Peoria loess and is overlain by the Bignell loess. The name was subsequently adopted by researchers in Kansas (Frye and Fent, 1947; Frye and Leonard, 1949, 1951). It is regionally extensive only in the northwestern and west central parts of Kansas, and even there it occurs discontinuously on the landscape. Frye and Leonard (1951) and Caspall (1970, 1972) recognized Brady development in northeastern and other parts of Kansas. Without the overlying Bignell loess, the Brady soil does not exist; the modern surface soil has incorporated post-Bradyan loess fall into its profile or may have developed in Peoria loess subsequent to the erosion of the Brady soil and Bignell loess. The Brady soil is typically dark gray to gray-brown and better developed than the overlying surface soil within the Bignell loess. Strong textural B horizon development and carbonate accumulation in the C horizon are typical, although it occasionally displays evidence of having formed under poorer drainage conditions than have associated surface soils (Frye and Leonard, 1951). Feng (1991) noted that the Brady soil, as expressed in Barton County, is strongly weathered both physically and chemically.

The age of the Brady soil has been uncertain, even at the type locality. Dreeszen (1970, p. 19) reported an age of 9160±250 (W-234) obtained in 1954 and another in 1965 of 9750±300 (W-1676), both from the type section but very likely contaminated by modern plant roots. Subsequently, Lutenegger (1985) reported an age of 8080±180 years B.P. but provided no specifics other than that the source was the A horizon of the Brady soil at the type section. Better age control for the type section has since been secured by this investigator: ages of 9,240±110 (Tx-7425) and 10,670±130 (Tx-7358) years B.P. were obtained on the upper and lower 5 cm, respectively, of the Brady A horizon.

The Brady soil has been recently dated at localities in Nebraska and Kansas. Souders and Kuzila (1990) obtained a radiocarbon age of 10,130±140 years B.P. on the Brady soil occurring within the Republican River valley of south-central Nebraska. Sites along Harlan County Lake upstream from Naponee have yielded a number of ages, ranging from 10,550±160 to 9,020±95 years B.P., on exposures of the Brady soil (Cornwell, 1987; Johnson, 1989; Martin, 1990; Martin and Johnson, 1995). Two radiocarbon ages of 9820±110 (TX-7045) and 10,550±150 (TX-7046) years B.P. have been derived from the upper and lower 5 cm, respectively, of the Brady A horizon exposed in Barton County, central Kansas (Feng, 1991).

Although it appears Brady pedogenesis occurred from about 10,500 to as recently as 8,500 years B.P., greater refinement of the Brady soil chronology is necessary, but present data clearly indicate it was a product of a major period of landscape stability at a time when widespread climatic shifts were occurring at the end of the Wisconsin. This was the first significant period of soil development since Gilman Canyon time, and represents the climate of the early Holocene. There is an isochronous alluvial soil found throughout the region which is particularly well expressed within the Kansas River basin (Johnson and Martin, 1987; Johnson and Logan, 1990). The two ages of 8,274±500 (C-108a) and 9,880±670 (C-471) years B.P. determined from alluvial fill (Fill 2A) at archaeological sites Ft-50 and Ft-41 on Harry Strunk Lake in southwestern Nebraska (Schultz et al., 1951; Libby, 1955) were the first radiocarbon determinations on the Brady soil. The soil, occurring in both eolian and alluvial contexts, appears to qualify, based upon present radiocarbon data, as a geosol, like the Gilman Canyon Formation soil. A typical exposure of the Brady soil would be that in Phillips County, located in west-facing roadcuts in the SW SW, Sec. 24, T4S, R19W. The locality was recognized by A.R Leonard (1952, p.42-3) and revisited by Johnson (1993b). It is the east face of a road cut about .8 km north of Speed, Kansas in which the Peoria loess, Brady soil, and Bignell Loess are visible. In the late 1940s and early 1950s the Loveland loess, Sangamon soil, and Gilman Canyon Formation were also exposed in the roadcut; they can yet be distinguished in a poor quality exposure around on the north face at the end of the roadcut. A profile within the road cut was excavated and sampled for radiocarbon dating in the uppermost and lowermost 5 cm of the A horizon: ages of 8,850±140 (Tx-6626) and 10,050±160 (Tx-6627) years B.P. were obtained, respectively (Johnson, 1993b). In sum, age data indicate a soil forming interval lasting 1,500-2,000 years.

Development of the Brady soil correlates well with indicators of regional climatic change. The fossil pollen record at Muscotah Marsh of northeastern Kansas indicates that spruce had essentially disappeared from the region by about 10,500 years B.P. As this decline occurred, deciduous tree species increased until about 9,000 years B.P., the time at which grassland expansion began (Grüger, 1973). On a hemispheric scale, the abrupt decrease in atmospheric dust noted in the Greenland ice core at 10,750 years B.P. (Paterson and Hammer, 1987) reflects decreased loess deposition and possibly Brady-age pedogenesis associated with relative terrestrial stability. Further, 180 levels within the same core suggest rapid warming about 10,750 years B.P., with the characteristic Holocene temperature regime being established about 9,000 years B.P.

Bignell loess. The Bignell loess was first described and named at the type locality in a bluff exposure on the south side of the Platte River valley southeast of North Platte, Nebraska (Schultz and Stout, 1945). It is typically a gray or yellow-tan, massive silt, calcareous and seldom more than 1.5 m (5 ft) thick. Although it is often somewhat less compact and more friable than the underlying Peoria loess, no certain identification can be made without the presence of the Brady soil. The Bignell loess does not form a continuous mantle on the Peoria; instead, it occurs as discontinuous deposits which are most prevalent and thickest adjacent to modern-day valleys, particularly the south side, and often within depressions on the Peoria surface. Feng (1991) speculates that the Bignell loess of central Kansas is relatively well weathered because it was derived from a pre-weathered source, the Brady soil surface, perhaps eolian and alluvial phases alike. This is consistent with the earlier interpretation derived in Nebraska that Bignell loess is at least partially comprised of re-worked Peoria loess (Condra et al., 1947, p. 33).

It appears from the radiocarbon ages obtained at the type section in Nebraska and the Speed roadcut in northwestern Kansas that the Bignell loess can be no older than 8,000 to 9,000 years B.P. Snails collected by A.B. Leonard from the lower part of the Bignell in Doniphan County, northeastern Kansas, produced ages of 12,500±400 (W-231) and 12,700±300 (W-233) years B.P. (Frye and Leonard, 1965). Because the shell material had absorbed an indeterminate amount of dead carbonate, Frye and others (1968) proposed an averaged age of approximately 11,000 years. Based upon the age data available for the Brady, the soil humate-derived ages are probably closest to reality.

A pronounced feature of the Holocene climate of the Great Plains was an extended warm, dry period (Wright, 1970; Benedict and Olson, 1978; Barry, 1983), identified as the Altithermal (Antevs, 1955) or, less commonly, as the Hypsithermal (Deevey and Flint, 1957). This dictates that the Bignell loess was a warm-climate loess, unlike the cold-climate loess of the Woodfordian. Reconstruction of the general circulation patterns for North America indicates that from the last glacial maximum about 18 to 15 ka there was no detectable change in atmospheric circulation: the westerly jet was split by the Laurentide ice sheet into a north and south flow around a strong glacial anticyclone (Kutzbach, 1985, 1987; COHMAP Members, 1988). By 9 ka, the ice had wasted appreciably, the jet was no longer split, orbital parameters were favoring increased temperatures, and zonal flow was dominating (Kutzbach, 1981,1985, 1987). Model results produced mean summer temperatures 2° to 4° C higher (COHMAP Members, 1988) and annual precipitation up to 25% less than at present in the region (Bartlein et al., 1984; Kutzbach, 1987).

Because of the increasing zonal flow and aridity of the Altithermal, species of the tall grass community migrated eastward to the present areas of mixed deciduous-prairie vegetation, i.e., the prairie-forest ecotone shifted eastward (VanZant, 1979; Semken, 1983; Webb et al., 1983). The fossil pollen record from Muscotah Marsh provides a disrupted but interpretable Holocene signal, indicating a middle Holocene prairie expansion (Grüger, 1973). Fossil pollen data from Cheyenne Bottoms suggest consistently lower water levels in the marsh during the middle Holocene (Fredlund, 1991, 1995). Molluscan fauna from the Bignell Loess of Kansas suggest that climate was somewhat drier than during Peoria time (Frye and Leonard, 1951). After a period of soil formation near the end of the Pleistocene, pedogenesis is not recognized until about 5,800 years B.P. in the sand sheet of the Great Bend Prairie, central Kansas (Johnson, 1991). Therefore, based upon various climatic proxies and a limited number of radiocarbon ages, it appears the Bignell loess was deposited, for the most part, from the end of Brady pedogenesis at about 8,500 years B.P. to about 5,500 years B.P.

Fort Riley. The Brady soil and Bignell loess are prominent elements of the late-Quaternary stratigraphy on the reservation. Together with the surface soil, they typically comprise the approximately upper 2 m of most sites documented (D.L. Johnson, 1996), including many of the sites analyzed for this report. Therefore, it appears that appreciable loess was deposited during the Holocene. This is not surprising given the proximal juncture of two large stream systems, which formed a significant loess source (dust from a relatively wide valley floor), and the situation of the sites on the north side of the river valley with southerly winds.

Paleoclimatic History during the Wisconsinan Glacial Stade

As the most recent glacial episode, the Wisconsin has the greatest chronostratigraphic resolution. However, existing knowledge of the climatic environment for this time interval is relatively limited and inconsistent for the central Great Plains. To date, either pollen records from peripheral areas or synthesis of various types of proxy data have been used to reconstruct climate history of the region.

The insolation record exhibits two relatively warm peaks (c. 50 and 30 ka) during marine isotope stage 3. Each peak is followed by a relatively minor and gradual decrease, culminating in the decrease to the glacial maximum (c. 18 ka). In contrast to the gradual insolation changes, most paleoclimatic records from this interval indicate rapid alternations of warm and cold events, which in frequency and timing appear to be unrelated to the Milankovitch forcing (Curry et al., 1992). Dansgaard and others (1985) showed that rapid and extreme fluctuations in stable isotopes (signifying air-temperature differences of 5°C) from two sites in Greenland appear to correlate over the past 50,000 years. These fluctuations are in phase with changes in CO2 and dust content. Rapid and substantial temperature fluctuations recorded in ice core segments during stage 3 also correspond with oscillations in the North Atlantic marine sediment record of species abundance and ice rafting (Heinrich, 1988). Using accelerator 14C ages on a planktonic polar species, Broecker and others (1988) identified four rapid climatic oscillation between 40 and 22 ka in the North Atlantic.

The regional upland vegetation, as inferred from the pollen record from Cheyenne Bottom in Kansas (Fredlund, 1995), appears to have been nearly treeless throughout the Farmdalian (ca. 30-24 ka). The pollen record suggests that, although regional tree and shrub populations were higher and more diverse than in Holocene, they were a secondary component of the overall vegetative structure. Grassland-sage steppe dominated the regional uplands surrounding the Cheyenne Bottoms basin throughout the Farmdalian period. The rise in Cheno-Am pollen percentages and an influx of sand beginning at about 25 ka probably mark the rapid onset of a cycle of aridity. Immediately after the onset of aridity, the most noticeable changes are declining percentages in the Cheno-Am types and rising percentages of Pinus. The increase in Picea and other arboreal pollen may signal a climatic shift toward cooler climatic conditions at ca. 24 ka. Unfortunately, the Woodfordian substage of the Wisconsin is missing from the Cheyenne Bottoms record.

Limiting 14C dates in North America indicate that glaciation commenced about 25-27 ka and thus allow less than 10,000 years for ice buildup prior to 18 ka (Andrews, 1987). The structure of deglaciation is uncertain. There is evidence supporting: (1) a smooth deglaciation model with fastest ice wastage centered on 11 ka; (2) a two-step deglaciation model with rapid ice wasting from 14 to 12 ka and 10 to 7 ka, and a mid-deglacial pause with little or no ice disintegration from 12 to 10 ka; and (3) a Younger Dryas deglaciation model with two rapid deglacial steps as in (2) above, interrupted by a mid-deglacial reversal with significant ice growth from 11 to 10 ka.

The data supporting the smooth deglaciation model are maps of Laurentide ice area based on 14C-dated glacial deposits (Andrews, 1987). Although there are subtle suggestions of more rapid retreat at or near the time of the two steps mentioned above, these curves indicate a steady progressive retreat of North America ice, with significant oscillations in retreat rate only at local spatial scales. Some marine δ18O curves also show a smooth progressive decrease toward Holocene values.

The step deglaciation model is also supported by some marine δ18O records (Mix, 1987). In addition, the distinctive patterns of change in sea-surface temperature of the North Atlantic Ocean and in Greenland ice-core δ18O values also show abrupt step-like warming at 10 ka and about 13 ka; these warming might be associated with step-like decreases in Laurentide ice volume. Regionally integrated rates of pollen change in eastern and central North America also show a rapid change in centered on 13.7 and 12.3 ka. (Ruddiman, 1987).

The Younger Dryas (e.g., Osborn et al., 1995: Bard et al., 1993; Peteet et al., 1992) deglaciation model is suggested by the strong signal of sea-surface temperature cooling between 11 and 10 ka in the North Atlantic Ocean. At least early and perhaps all of Brady pedogenesis coincides with an abrupt and brief cool interval correlative with the classic Younger Dryas cold interval of the North Atlantic region.

By the middle Holocene, drying had reached a maximum according to most studies. Northwest Texas was experiencing conditions of maximum temperatures, minimum precipitation, and eolian activity between 6000 and 4500 yrs B.P. (Holliday et al., 1983; Holliday, 1985; 1989; Johnson, 1987; Pierce, 1987). This episode coincides with δ13C values from soil organic matter from the same area revealing a shift from -23‰ in the early Holocene to -15‰ in the middle Holocene (Haas et al., 1986). These results were interpreted to represent a shift from cool-season C3 grasses to warm-season C4 grasses. Based on enriched δ13C values in soil carbonate from northwest Texas, Humphrey and Ferring (1994) also show a middle Holocene xeric episode, although the δ18O values from these same carbonates do not indicate a significant temperature change.

A noticeable shift back to cooler and/or wetter conditions was detected in many areas shortly after 5000 yr B.P. The Great Bend Sand Prairie transformed to conditions much like present (Arbogast, 1995). According to Humphrey and Ferring (1994), the return to mesic conditions after 5000 yrs B.P. was interrupted in north-central Texas by a brief warming and drying episode between 2000 and 1000 yrs B.P. Based on depositional environments, they concluded that cooler and wetter conditions returned after 1000 yrs B.P.

Study Localities

Sites were selected for opal phytolith study on the basis of spatial coverage of the reservation, landscape position, and absolute age control resulting from studies by D.L. Johnson (1994,1996). Sumner Hill (core site 21; D.L. Johnson, 1994) and Bala cemetery (core site 19; D.L. Johnson, 1994) were the upland sites investigated. The Sumner Hill site is located along the bluff line above Camp Funston and consists of approximately 12 m of Quaternary sediment over a strath cut into the Permian Wreford Limestone. A basal gravelly alluvium is overlain by over 11 m of loess, i.e., the Loveland loess and Sangamon soil up through Bignell loess capped by the surface soil. Due to the analytical constraints imposed by the sampling of cores, a series of six large trenches were excavated in the valley wall slope to form a stair-step exposure of the sediments in order that the entire sequence could be clearly viewed and thoroughly and accurately sampled. This site was investigated in the first phase of opal phytolith investigation (Johnson et al., 1994), but is included here to provide a comprehensive perspective. The other upland sample site is located adjacent to the Bala cemetery, a location nearly 40 km northwest of the Kansas River valley and about 1O km east of the Republican River valley. The site consists of approximately 2.75 m of loess overlying the Permian Winfield Limestone, a much thinner loess mantle than at the former site located adjacent to the river valley.

The Pump House Canyon site consists of a loess deposit situated on a strath located within a small tributary valley to the Kansas River valley. Sampling was conducted in a 3.3 m-deep backhoe trench. Although the trenching did not perforate the loess layer, its total thickness is likely less than 5 m because the upper part of the Gilman Canyon Formation was exposed. The Manhattan Airport site (core site 33), located approximately 1.5 km northwest of the airport, consists of clay-rich late-Pleistocene and Holocene alluvium. The site was backhoe trenched to expose the upper 3.8 m of fill, which included alluvial phases of the Sangamon soil(?), Gilman Canyon Formation, and Brady soil, which was overlain by Holocene alluvium.

Opal Phytolith Analysis

Due to a lack of old-growth trees suitable for dendroclimatology and bogs and natural lake for palynology and macrofossil analysis, the central Great Plains offers little in the way of climatic proxy sources. Opal phytolith analysis, however, has the potential to offer paleoenvironmental and paleoclimatological information that will provide a relatively detailed picture of the past.

The objective of this research was to determine the feasibility of recovering fossil opal phytoliths (siliceous plant cells) from sediment and soil samples collected at the study sites in order to reconstruct the vegetative history for the reservation. Phytoliths are the most common biosilicate in upland loess deposits and are the most useful for environmental reconstructions. Sponge spicules, another form of biogenic silica, may also be present in upland deposits but are much less common and of limited use when compared to the potential of phytoliths.

Grass opal phytoliths are the best studied and can be separated into morphologic categories related to the plant photosynthetic pathways and the major subfamilies of grasses. Twiss and his students (Twiss, 1980, 1983, 1987; Twiss et al., 1969; Kurmann, 1981, 1985) were the first to recognize the correlation between grass photosynthetic groups (adaptations) and phytolith morphology, i.e., the major subfamilies of grasses correspond to three morphologic classes of phytoliths.

Most Poaceae (grasses) employ the C3 pathway (Calvin) for the fixation of CO2 in the photosynthetic process. Commonly, these grasses belong to the Pooideae (festicoid) subfamily. Grasses included in this subfamily include the bromes (Bromus spp.), fescues (Festuca spp.), needlegrass (Stipa spp.), wheatgrass (Agropyron spp.), bluegrasses (Poa spp.), and many cereals such as rye (Secale cereale), oats (Avena spp.), barley (Hordeum vulgare), and wheat (Thriticum aestivum). The C3 grasses are widespread but are best adapted to the higher (cooler) latitudes and altitudes.

Conversely, the C4 grasses, employing the Hatch-Slack CO2 photosynthetic pathway, are most successful in the lower (warmer) latitudes and altitudes. This system is better adapted to high temperatures and low moisture conditions. In the Great Plains, two groups (subfamilies) of grasses typically utilize the C4 pathway, the Chloridoideae and Panicoideae subfamilies. Examples of grasses within the Chloridoideae subfamily are the three-awns (Aristida spp.), gramas (Bouteloua spp.), buffalo grass (Buchloe dactyloides), saltgrasses (Distichlis spp.), sandreed grass (Calomovilfa spp.), lovegrasses (Eragrostis spp.), muhly grasses (Muhlenbergia spp.), and dropseed grasses (Sporaholus spp.). Although the Panicoids are well adapted to high temperatures, they require more moisture than the Chloridoids, and are consequently better adapted to the eastern Great Plains, e.g., eastern Kansas. Included in the Panicoids are the bluestems (Andropogon spp.), panicums (Panicum spp.), indian grass (Sorgastrum nutans), gama grass (Tripsacum dactyloides), and the cereal grasses corn (Zea mays) and sorghum (Sorghum halepense).

Loess deposits of the central Great Plains have been found to contain large amounts of grass opal phytoliths, which produce an interpretable climatic record (e.g., Fredlund et al., 1985; Bozarth, 1991b, 1992b; Fredlund, 1993; Johnson, 1993a; Johnson et al., 1993b). Further, opal phytolith data from the Gilman Canyon Formation at the Eustis ash pit indicate the dominance of C4 grasses throughout most of the formation, and appear to correlate nicely with the δ13C values. Opal phytoliths, unlike δ13C values, can indicate the presence of an arboreal component (e.g., Rovner, 1971; Geis, 1973; Wilding et al., 1977) and differentiate to some extent between deciduous and coniferous trees (Bozarth, 1992a, 1993c).

Opal phytoliths are generally well preserved in most sediment and can be isolated from sediment samples and analyzed to reconstruct the paleoenvironment for a particular area. This has been successful on a number of sediment types, including loessal sites in China (Lu et al., 1991), Nebraska (e.g., Fredlund et al., 1985; Bozarth, 1991b, 1992b; Johnson et al., 1993a; Fredlund, 1993) and the Southern High Plains (Bozarth, 1995), as well as alluvium in Kansas (Kurmann, 1981, 1985; Bozarth, 1986) and the Southern High Plains (Bozarth, 1995), and swamp and upland sediment in Panama (Piperno, 1988).

Formation and Stability

Growing plants absorb water containing dissolved silica through their roots. Microscopic amorphous silica bodies are subsequently produced by the precipitation of hydrated silicon dioxide (SiO2·nH20) within the plant's cells, cell walls, and intercellular spaces. Silica bodies with characteristic shapes are called opal phytoliths. Phytolith is derived from the Greek words phyton, meaning plant, and lithos, meaning stone. Opal is the common name for hydrated silicon dioxide. Opaline bodies formed in plants without specific shapes are simply plant opal. Phytoliths form in most plants and are produced in many shapes and sizes. Many phytolith types are specific to particular groups of plants. Phytoliths are largely a "decay in place" fossil (Rovner, 1975) and represent the vegetation of a site at the time of deposition (Piperno, 1988).

The dissolution and stability of phytoliths in soil is not fully understood. Laboratory experiments demonstrate, however, that the solubility of silica is a function of temperature, particle size, pH, and the presence of a disrupted surface layer. Studies show that the solubility of amorphous silica increases linearly with temperature from 0°C. Particle size is another factor affecting stability as opal dissolution is greater with a decrease in size (Wilding et al., 1977, 1979). Pease (1967) experimentally determined that there appears to be a slight increase in phytolith solubility in the range of 5.0 to 8.5, an added increase between pH 8.5 and pH 9.0, and a large increase beginning at pH 9.0. Opal stability is also a function of the presence of certain metallic ions and sesquioxides. The adsorption of AI and Fe ions onto the surface of opal will decrease silica dissolution due to the formation of relatively insoluble silicate coatings. The presence of sesquioxides may increase dissolution of phytoliths due to the adsorption of monosilicic acid (Wilding et al., 1977).

Morphology and Taxonomy

Monocotyledons, particularly the Poaceae, produce a wide variety of morphologically distinctive opal phytolith forms. The most taxonomically useful types of grass phytoliths are silicified short cells. Several types of trapezoidal circular, rectangular, and elliptical short cells are diagnostic of the Pooideae (Brown, 1984; Twiss, 1987; Bozarth, 1992b), a C3 grass subfamily adapted to cool temperatures (Twiss, 1987). Saddle-shaped bodies occur most commonly in the Chloridoideae, a C4 grass subfamily (Brown, 1984; Twiss, 1987; Mulholland and Rapp, 1992) that flourishes in areas with warm temperatures and low available soil moisture. Saddle-shaped phytoliths are similar in appearance to double-edged battle axes formed by two opposite convex edges and two opposite concave edges. However, a few saddle-shaped phytoliths have only one concave side (Brown, 1984).

Bilobate and cross-shaped phytoliths are formed in the Panicoideae, the C4 grass subfamily (Brown, 1984; Twiss, 1987; Mulholland and Rapp, 1992) that thrives in warm temperatures and high available soil moisture (Twiss, 1987). Bilobates with indented, concave, or pointed lobes are formed only in grasses in the Panicoid subfamily. Bilobates with raised lobe edges and round or flat ends which are symmetrical in side view are also formed only in Panicoids (Bozarth, 1992b).

Bilobate phytoliths with raised lobe edges and round ends are also formed in Aristida (needlegrass, wiregrass), a genus in the Chloridoid subfamily (Gould and Shaw, 1968). However, bilobates formed in Aristida differ from Panicoid bilobates in that the raised edges on the top (the longer part) slope down at the ends. In addition, they are asymmetrical in side view as the top is more concave than the bottom (Bozarth, 1992b). Stipa, a genus in the Pooid subfamily (Gould and Shaw, 1968), also produces bilobates (Bozarth, 1992b). These bilobates differ from those produced in Panicoids and Aristida by not having raised lobe edges. Many have a small lobe on one side in the middle. Unlike most Pooids, Stipa species grow in dry areas (Pohl, 1968).

There are several other types of phytoliths produced in grass in addition to short cells. Long cells are relatively large, elongate bodies with smooth or wavy edges. Bulliform cells are large keystone shaped-cells. Dendriforms are cylindrical rods of varying length that have protrusions or spines radiating from a central core. Asteriforms are roughly spherical spiky phytoliths. Trichomes are silicified prickly-hairs composed of two parts, an outer sheath and an inner core. The outer sheath dissolves soon after being deposited on the soil, while the inner core remains well preserved. The silicified stomata are taxonomically useful at various levels but are typically not well preserved. The other types are not specific to any particular subfamily but are preserved in most sediment. Piperno (1988) reported that dendriforms and asteriforms are apparently formed only in grass floral bracts.

Non-grass monocots also produce numerous taxonomically valuable phytoliths. Cyperus (sedge) produce distinctive phytoliths in the form of cone shaped-bodies with round wavy margins. These phytoliths are occur both singly and in multiples. Truncated cones with multiple peaks and round wavy bases are formed in Scripus pallidus (bulrush). Both of these phytolith types appear to be diagnostic of the genera that produce them (Bozarth, 1993). Several types of phytoliths are produced in woody dicotyledons (deciduous shrubs and trees) and herbaceous dicotyledons (forbs and weeds). The two most common types of diagnostic dicot phytoliths are flat polyhedrons with 5-8 sides and anticlinal cells (Rovner, 1971; Wilding and Drees, 1971; Geis, 1973; Wilding et al., 1977; Bozarth, 1992a). Anticlinal cells have wavy, undulating walls with the appearance of jigsaw-puzzle pieces. Most of these polyhedral and anticlinal phytoliths consist only of silicified cell walls and are not well preserved in sediment (Wilding and Drees, 1974; Bozarth, 1992a). Other phytolith types formed only in dicots include branched elements with spiral thickening and honeycomb-shaped assemblages (Geis, 1973; Wilding and Drees, 1973, 1974; Bozarth, 1992a).

Several species of arboreal dicots produce opal spheres that range in size from 1 to 50 micrometers (Wilding and Drees, 1973, 1974). Opal spheres are also produced in conifers (Klein and Geis, 1978), but are much smaller (3 to 8 micrometers). Opaque opal spheres have been extracted from the A horizon of several forested soils in Ohio demonstrating that they are well preserved (Wilding and Drees, 1973, 1974).

Spiny spheres are formed in neotropical palms (Piperno, 1988) but have not been reported in temperate vegetation. However, the association of spiny spheres with deciduous tree phytoliths in a loessal site in Nebraska (Bozarth, 1992b) suggests that they are also formed in this, or an associated, group of plants.

Wilding and Drees (1973) reported opaque bladed forms (which appear to be opaque platelets), in white oak (Quercus alba). Similar particles were observed in isolates from a soil formed under deciduous forest.

Several families and genera of dicots produce phytoliths unique to those taxa. Opaque platelets with systematic perforations and certain types of segmented hairs are diagnostic of Asteraceae (the sunflower family). Platelets with irregular edges and echinate (spiny) sculpturing on one side are formed in the fruit of hackberry (Celtis occidentalis) and appear to be unique to that genus. These types are well preserved in sediment. Flat polyhedrons with 5-8 sides that are filled with coarse verrucae (bumps) appear to be unique to Ulmaceae (the elm family) (Bozarth, 1985, 1987c, 1992a).

Certain types of stalked verrucate phytoliths are specific to hackberry, mulberry (Morus), false nettle (Boehmaria), or nettle (Urtica). Elongate verrucate phytoliths with one or both ends tapering to a point are unique to Pilea (Bozarth, 1992a). Phytoliths with deeply scalloped surfaces of contiguous concavities are unique to Cucurbita (Bozarth, 1987a).

Several types of phytoliths are produced in the Pinaceae (pine family). Silicified, irregularly-shaped, polyhedral cells are the most common taxonomically useful Pinaceae phytolith. This type of phytolith is produced in Picea rubens (red spruce), P. mariana (black spruce), P. glauca (white spruce), P. engelmannii (Engleman spruce), and Pinus banksiana (jack pine) (Norgren, 1973; Klein and Geis, 1978; Bozarth, 1988, 1993c). Blackly polyhedra with smooth surfaces and at least eight non-parallel sides are characteristic but not diagnostic of Pinaceae, because they are also produced, although relatively infrequently, in grasses (Bozarth, 1993c).

In contrast to smooth polyhedrons, polyhedrons with bordered pit impressions on the surface are unique to the Pinaceae. This type of phytolith is abundant in Pinus (pine), Picea (spruce), Douglas-fir (Pseudotsuga), and less commonly in Larix (larch), Tsuga (hemlock), and Abies (fir) (Klein and Geis, 1978). Pseudotsuga menziesii (Douglas-fir) needles produce distinctive, branched, silicified particles (Brydon et al., 1963). This same type of phytolith was also reported in Douglas-fir by Garber (1966) as irregular shapes with spiny processes and by Norgren (1973) as amoeboid bodies with tapering, conical protrusions. Thin plates with wavy margins on all four sides are formed in needles of Picea glauca (white spruce) and appear to be unique to that species. Phytoliths with spiny irregular bodies are commonly formed in needles of Pinus banksiana (jack pine) and appear to be diagnostic of that species (Bozarth, 1993c).


Sampling was done either from a core extracted with a trailer-mounted Giddings drilling machine or from an exposure created by backhoe trenching. In both cases, samples were collected every 1O cm using a trowel cleaned thoroughly between samples. Samples were placed in sterile plastic bags for storage until extraction. Phytoliths were isolated from 5-gram subsamples using a procedure based on heavy-liquid (zinc bromide) flotation and centrifugation (Bozarth, 1991a). This procedure consists of five basic steps: 1) removal of carbonates with dilute hydrochloric acid; 2) removal of colloidal organics, clays, and very fine silts by deflocculation with sodium pyrophosphate, centrifugation, and decantation through a 7-micron filter; 3) oxidation of sample to remove organics; 4) heavy-liquid flotation of phytoliths from the heavier clastic mineral fraction using zinc bromide concentrated to a specific gravity of 2.3; 5) washing and dehydration of phytoliths with butanol; and 6) dry storage in 1-dram glass vials.

A representative portion of each phytolith isolate was mounted on a microscope slide in immersion oil under a 22x40 mm cover glass and sealed with clear nail lacquer. Each isolate was then studied at 400x with a research-grade Zeiss microscope. Each sample slide was first examined to determine the quality of preservation of the phytoliths. At least 200 phytoliths were counted in all of the samples with adequate preservation. A complete slide was scanned and all phytoliths classified in those samples with poor preservation.

Estimates of phytolith concentration were made using an indirect method reported by Piperno (1988). A known number of exotic spores (in this case Lycopodium) were added to each sample after the oxidation stage. The concentration of phytoliths (per gram) was computed as follows:

Phytolith conc. = no. of phytoliths counted x (total no. exotics added / no. exotics counted) / 5

Concentration permits an evaluation of the phytolith production, preservation, and sedimentation rate for a given sample interval.

Phytoliths were classified according to a convention that has been developed and used by other reports and publications. An extensive reference collection of plants native to the Great Plains has been developed in the palynology laboratory through field collection, research plots, solicited samples, and specimens supplied by the University of Kansas Herbarium. The phytolith reference collection consists of phytoliths extracted from complete or representative aerial portions of the following: 1) 25 species of 20 genera of 11 tribes of 6 subfamilies of the Poaceae (grass); 2) 11 species of 4 genera of 4 non-grass monocot families; 3) 65 species of 62 genera of 11 families of herbaceous dicots; 4) 20 species of 18 genera of 13 families of woody (mostly arboreal) dicots; 5) 14 species of 7 genera of 5 families of gymnosperms; and 6) 2 species of Equisetum. These reference materials include all the dominant species in the study area as reported by Kuchler (1974).

An unknown group consists primarily of phytoliths too poorly preserved to be classified any other way. There were a few other phytoliths included under this heading that were stuck under the cover glass and could not be rotated for three-dimensional viewing, thereby precluding positive taxonomic classification.

The biosilicate data are presented in computer-generated, percentage diagrams using Tiliagraph software (Grimm, 1992) designed for depicting fossil pollen data. For each site analyzed, two diagrams were constructed, one for all of the data and another for only the short cells diagnostic of grass subfamilies and the arboreal-type phytoliths. The latter diagram presents an unobscured perspective on the more diagnostic microfossils. Biosilicate data for each site have been subdivided on the diagrams into zones. Zonation is a common tool in botanical microfossil analysis, particularly palynology (fossil pollen analysis), and is rapidly being adopted in opal phytolith studies. This approach is generally used to identify periods which are similar with regard to the assemblages of the various microfossil types. Zonation has been used in this study to differentiate periods of concentration and preservation as well.

The phytolith data are also represented in this study as an aridity index, or measure of the moisture stress. The index may be computed using two or more of the three categories of phytoliths. Diester-Haass and others (1973) presented a climatic index consisting of the ratio of chloridoid phytoliths to the total number of chloridoid and panicoid phytoliths, based on a total count of 100, i.e., the higher the index, the more arid the climate and vise versa. If the pooid phytolith data are also used, an index of the ratio of chloridoid types to the sum of the three classes can be computed as an aridity index (Twiss, 1987). Diekmeyer (1994) created a third variation on the index consisting of the ratio of the chloridoid-panicoid sum to the sum of all three classes. The latter index has been adopted for use in this study because it appears to have the best intuitive basis.

Next page--Results

Kansas Geological Survey, Geology
Placed online Aug. 10, 2012
Comments to
The URL for this page is