The valley floor of a 33.9 km2 watershed in western Colorado experienced gradual sedimentation from before ∼ 6765 to ∼ 500 cal yr BP followed by deep incision, renewed aggradation, and secondary incision. In contrast, at least four terraces and widespread cut-and-fill architecture in the valley floor downstream indicate multiple episodes of incision and deposition occurred during the same time interval. The upper valley fill history is atypical compared to other drainages in the Colorado Plateau.

One possible reason for these differences is that a bedrock canyon between the upper and lower valley prevented headward erosion from reaching the upper valley fill. Another possibility is that widespread, sand-rich, clay-poor lithologies in the upper drainage limited surface runoff and generally favored alluviation, whereas more clay-rich lithologies in the lower drainage resulted in increased surface runoff and more frequent incision. Twenty-two dates from valley fill charcoal indicate an approximate forest fire recurrence interval of several hundred years, similar to that from other studies in juniper–piñon woodlands. Results show that closely spaced vertical sampling of alluvium in headwater valleys where linkages between hillslope processes and fluvial activity are relatively direct can provide insight about the role of fires in alluvial chronologies of semi-arid watersheds.

Establishing the magnitude and frequency of floods within upland catchments on the basis of short-term gauged runoff records is crucially dependent upon the extent to which the record is truly representative. In the case of the River Dee, upstream of Crathie in Aberdeenshire, gauged discharge records are limited in length. Although the middle Dee has been gauged since 1929, the gauge within the upper catchment has only ten years of record. Thus, reliable estimates of the return intervals of extreme floods for this part of the Dee can only be obtained by using a variety of historical sources to extend the flood series.

Long-term rainfall records, where available, provide a valuable independent check on the reconstructed flood series. Such rainfall records, when analysed in terms of the magnitude, frequency and duration of major events, should, in general terms, correspond with the flood series. In this paper, the recurrence interval of extreme rainfalls of varying magnitude and duration in upper Deeside is estimated by extreme value analysis of the annual maximum series. The frequency of rainfall events above varying thresholds is also assessed. The existence of climatic fluctuations giving highly variable recurrence intervals for rainfall events of the same magnitude is demonstrated. Finally, the seasonality of frequent flood-producing storms is analysed. Patterns observed within the rainfall record are compared with those previously established within the historic flood series to substantiate and augment the flood record.