Estimation of continuous daily streamflow at ungaged locations in southern New England.

摘要

The estimation of streamflow at ungaged locations has been an active area of research in hydrology for several decades. The focus of this dissertation is on the development of a new methodology, termed the streamflow-duration map-correlation transformation (SDMCT) method, to estimate continuous daily streamflow at ungaged locations. A new method--termed the streamflow-duration method--is proposed to estimate a daily, period-of-record flow-duration curve at an ungaged location by exploiting the strong structural relation among streamflow quantiles. A time series of daily streamflows is then created from the estimated flow-duration curve by transferring the timing of the daily streamflows at a selected index-streamflow station to the ungaged location at equivalent exceedance probabilities. It was found that the only reliable criterion for selecting the index-streamflow station was the the cross-correlation between the logarithms of the daily streamflows at the the index-streamflow station and the ungaged location. A geostatistical approach--termed the map-correlation method--is introduced to select the index-streamflow station whose daily streamflows are most-correlated with the daily streamflows at an ungaged location. Jack-knife cross-validation results from 19 streamflow-gaging stations show the daily streamflows estimated by the SDMCT method agree well with observed daily streamflows and are generally comparable to the agreement obtained from a calibrated rainfall-runoff model. The drainage-area-ratio method, a common method used to estimate daily streamflow at ungaged locations, when used with the map-correlation method for selecting an index-streamflow gage, was competitive with the SDMCT method. This research highlights the difficulties in applying standard techniques used in regional frequency analysis, such as distribution identification and fitting, to daily streamflows and suggests that traditional goodness-of-fit approaches, such L-moment diagrams and probability-plot-correlation coefficients are not useful in understanding the underlying processes that govern daily streamflow. The SDMCT method has been developed and applied to ungaged locations in southern New England but has applicability to other areas of the globe.