quency of large floods has changed (Baker, 2000). Data on past hydrological conditions from the upper Mississippi River (Knox, 2000) and from sediments in the Gulf of Mexico (Brown et al., 1999) record large, abrupt shifts in flood regimes in the Holocene, which may have been linked to major jumps in the location of the lower Mississippi (delta-lobe switching). In the western United States, there is growing evidence that flood regimes distinctly different from today, and also episodic in time, were the norm rather than the exception. The frequency of large floods in the Lower Colorado River Basin, for example, appears to have varied widely over the last 5,000 years (Ely et al., 1993; Enzel et al., 1996), with increased frequency from about 5,000-4,000 years ago, then lower frequency until about 2,000 years ago, and some abrupt shifts up, down, and back up thereafter (Ely, 1997). Those flood-frequency fluctuations and substantial fluctuations elsewhere around the world (e.g., Gregory et al., 1995; Baker, 1998; Benito et al., 1998) appear to be linked to climate shifts but in poorly understood ways. Clearly, a predictive understanding of megadroughts and large floods must await further research.
This observation about droughts and floods applies at some level to all the abrupt climate changes recorded in proxy records. The data are clear. Ice-age events were especially large and widespread and involved changes in temperature, precipitation, windiness, and so on. Holocene events were more muted in polar regions, might have been more regionalized, and usually involved water availability, but often with important temperature changes as well. Multi-characteristic global-anomaly maps are not available for any of the abrupt changes, and additional records and proxy techniques will be required to provide such anomaly maps. Coverage gaps appear especially large in the oceans and southern latitudes, although broad gaps also exist elsewhere.
Instrumental records from scientific monitoring programs offer the possibility of capturing directly the relevant data on abrupt climate change with greater accuracy and spatial coverage than are possible from the necessarily limited proxy records. The relatively short period of instrumental records means that they have missed most of the abrupt changes discussed above, although some droughts and the warming from the Little Ice Age have been captured rather well. Instrumental records will become more valuable as their length increases, which argues for maintenance of key