Records of fluvial changes since the last glaciation have been most extensively studied by the analysis of sediments in terraces, floodplains, and valley fills. Where possible, such studies are combined with the results of recent advances in understanding the relationship of channel morphology to climatic and hydrological controls (Schumm and Brakenridge, 1987). However, preserved paleochannels are limited to certain special alluvial settings such as those studied on the riverine plain of eastern Australia (Schumm, 1968), the north Polish plain (Kozarski and Rotnicki, 1977), and the Gulf Coastal Plain of the United States (Baker and Penteado-Orellana, 1977). Paleochannel preservation tends to be best for meandering river environments and poorer for environments characterized by braided or straight channels. Most often the sediments must be analyzed without the morphological clues. Pitty (1971, p. 16) aptly summarizes the situation: ''Lending confidence to the geomorphologist during his hypothetical leaps between form and process is the flimsy safety net provided by the study of sediments."
Two broad classes of paleoflow estimation derive from the two classical divisions of fluvial hydraulics (Leliavsky, 1955): regime theory and sediment transport mechanics. The regime approach to fluvial paleohydrology involves the use of various empirical relationships that relate the driving variables of (1) relatively high-probability flow discharge and (2) sediment characteristics to various dependent variables, including paleochannel dimensions, river patterns, and gradients. The relationships apply only to alluvial rivers with beds and banks composed of the same types of sediment as in transport by the channel-forming flows. Paleohydrological work in this area was pioneered by Dury (1954, 1965) and by Schumm (1965, 1968).
Because of the interplay of sediment and water discharge in fluvial response, alluvial rivers may display a degree of complex response to changes in their drainage basins (Schumm, 1977). Lag times also occur between causative agents and responses. For example, the effect of glaciation on sediment yields continues as long as the unstable drift in proglacial or postglacial environments remains easily accessible to fluvial erosion and transport (Church and Ryder, 1972). The high sediment yields associated with glacier-related deposits may persist with long lag times from the emplacing processes. Church and Slaymaker (1989) show that Holocene sediment yields in British Columbia are dominated by these effects, and the influence of relict glacial sediments continues to the present day.
Most of the equations used in alluvial regime paleohydrology are listed by Williams (1984). The approach is subject to many limitations and provides relatively low accuracy of paleoflow retrodiction (Ethridge and Schumm, 1978; Rotniki, 1983; Dury, 1985). However, in a semiquantitative sense, the regime approach combined with detailed studies of floodplain sedimentology can show the pattern of fluvial responses to changing environmental conditions (Baker and Penteado-Orellana, 1977).
A variety of procedures from sediment transport theory have been used to relate sediment characteristics to shear stress, flow velocity, or stream power. Combined with information on paleochannel dimensions, these procedures can yield paleoflow estimates (Baker, 1974; Costa, 1983; Williams, 1983). Unfortunately, numerous problems may contribute to a relatively low accuracy level for these procedures (Church, 1978; Maizels, 1983).
Despite the problems with both the regime theory and the sediment transport mechanical approaches to fluvial paleohydrology, these methods have the most universal range of applicability with regard to ancient river deposits. Most of the literature on late Quaternary fluvial change is based on the study of alluvial valleys interpreted by classical stratigraphy combined with some regime or sediment transport theoretical analysis.
A relatively new development in Quaternary paleohydrology is the recognition that certain stable-boundary fluvial reaches may, under ideal circumstances, preserve remarkably complete and accurate records of river flood stages. This technique was used for cataclysmic Pleistocene glacial floods (Baker, 1973; Patton et al., 1979) and found also to apply to arid-region Holocene floods (Baker et al., 1979, 1983; Kochel and Baker, 1982). Paleodischarges are calculated using hydraulic flow models (O'Connor and Webb, 1988) that relate slackwater deposits and paleostage indicators (SWD-PSI) to paleowater-surface profiles. Modern SWD-PSI paleoflood hydrology results in remarkably complete catalogues of the number, timing, and magnitudes of the largest floods occurring over periods of centuries or millennia (Baker, 1987a). The data can be used directly in magnitude-frequency analysis (Stedinger and Baker, 1987; Baker, 1989), understanding regional patterns (Enzel et al., 1993) or in interpreting the effects of environmental change on flood time series (Baker, 1987b; Jarrett, 1991).
Although most SWD-PSI paleoflood hydrology studies have been done on relatively small rivers, the methodology is not limited in scale. If appropriate study reaches can be found, then large rivers can also be analyzed. Chatters and Hoover (1986) used the methodology to analyze a 1800-yr record of late Holocene floods on the Columbia