Minimum variance streamflow record augmentation procedures

Traditional approaches for augmentation of short‐streamflow records have exploited the cross‐correlation among flows at two or more gages to obtain maximum likelihood estimates of the mean and variance of the flows at the short‐record gage. While such estimators asymptotically have minimum variance among all asymptotically unbiased estimators, they are not necessarily optimal for the small samples of interest in hydrology. Improved estimators of the mean and variance of the flows at the short‐record gage are developed. The gains in information transfer associated with the use of both the traditional and improved estimators of the mean and variance are documented. Monte‐Carlo results illustrate the performance of these estimators when the true cross‐correlation of the flows must be estimated. The information transfer gains when the true cross‐correlation must be estimated are comparable to the gains achieved when the cross‐correlation is known. The potential advantages and limitations of these new estimators are discussed within the framework of augmentation and/or extension of both peak annual floods and also monthly streamflow records.