METHODS

Both sediments and plant fossils were systematically collected for paleoenvironmental analysis. Paleosols were logged and described in the field following previously established methods (Tabor et al. 2002; Tabor and Montaņez 2004). Pedogenic horizons within the Chilga strata were recognized using the criteria of Kraus and Aslan (1993). Paleosol tops were identified on the basis of a marked change in lithology or sedimentary features, whereas profile bases were delineated at the highest occurrence of unaltered parent material. Field descriptions of paleosols (e.g., thickness, color, type, and distribution of mottling; soil structure and mineralogy; size, morphology, and distribution of authigenic minerals) follow the methods of Retallack (1988). Paleosol and lithologic colors were identified in the field with dry samples using Munsell color charts (Munsell Color 1975). In order to avoid recent weathering, outcrop faces were dug back minimally 60 cm to provide a fresh surface for describing and sampling the paleosols. Approximately 500 g of paleosol matrix was collected from each paleosol horizon and stored in either canvas or plastic bags.

The mineralogical composition of sand-size and coarser grains, as well as soil microfabrics, were determined petrographically from doubly-polished thin sections following the methods of Moorhouse (1959) and Brewer (1976). Powdered bulk samples were analyzed by X-ray diffractometry to determine the mineralogical composition of several different paleosol horizons. X-ray diffraction analyses were performed on a Rigaku Ultima III X-ray diffraction system configured with a vertical Theta:Theta wide angle goniometer , using CuK at 40Kv and 30mA. Measurements were performed with step-scan increments of 0.01° 2 , counting times of 2s per increment, 0.5/1.0 mm for the primary slits and 0.2/0.3 mm for the receiving slits. Mineralogical composition of the samples was determined following the methods of Moore and Reynolds (1997).

Plant fossils were collected at 50 localities across the basin in the course of surveying, but systematic excavations at three localities have produced the majority of >1100 specimens. One locality, exposed laterally along 60 m of one stratigraphic level at the Guang River section, was subdivided and excavated at four sublocalities. To minimize collecting bias, even fragmentary excavated plant fossils were collected if there was any chance of identification. When necessary, fossils were prepared with a fine-point airscribe. Cuticle was collected from fossil specimens by chipping off an approximately 1 cm2 sample, which was then prepared using hydrofluoric acid to dislodge inorganic matrix and a weak solution of bleach or hydrogen peroxide to oxidize the organic matter to transparency. Specimens were digitally imaged using one or all of the following: a high resolution single lens digital reflex camera, a digital camera mounted on a reflected or transmitted light microscope, and a LEO 1450 variable pressure scanning electron microscope.

Plant fossil specimens were compared with modern representatives from herbarium and live collections at Missouri Botanical Garden, Royal Botanic Gardens, Kew, or Fairchild Tropical Gardens, and in consultation with botanists who specialize in the relevant plant groups (see Acknowledgments).

Large vertebrate fossils were located by surface prospecting, and all bone and dental fragments were collected. Many localities were suitable for excavation, and some of these sites produced associated postcranial remains. Specimens were subsequently prepared with an airscribe as needed. In an effort to locate small vertebrate fossils, fine-grained sediments were screen-washed at several promising localities. However, the deposits did not yield small bones or teeth.