Marshall E. Moss

THE SOUTHERN OSCILLATION INDEX AS A PREDICTOR OF THE PROBABILITY OF LOW STREAMFLOWS IN NEW ZEALAND

The Southern Oscillation index (SOI) can yield information about subsequent streamflows on the South Island of New Zealand; however, the relationship between streamflow and SOI may be nonlinear and heteroscedastic. To deal with such difficulties, the conditional probability of streamflows being below a critical magnitude, given a prior observation of SOI, can be considered a random variable, with its probability density function (pdf) estimated by Bayesian analysis of existing observations of streamflow and SOI. The conditional pdf can yield a probabilistic forecast of critically low streamflows given a precursor value of SOI. In the Clutha River basin of New Zealand, average austral summer inflow to the headwater lakes has an unconditional probability of nonexceedance of 360 m3/s of approximately 17%. However, during a moderate La Nina, average austral spring SOI = 12, the conditional probability of nonexceedance has an expected value of 55%. For an El Nino with a spring SOI of −12, the expected probability of nonexceedance is 18%; for a neutral SOI of 0, the expected probability is 10%. Thus, the probability of nonexceedance of seasonal streamflow can be seen to vary by a factor of more than 5 as a function of the SOI.

An intercomparison of hydrological network-design technologies

Abstract Two network-design technologies are compared by random sub-sampling of actual streamflow data. The technologies, Network Analysis for Regional Information (NARI) and Network Analysis Using Generalized Least Squares (NAUGLS), have a common objective, viz. to maximize regional information within a limited budget and time horizon. The data used for intercomparison are from a network of 146 streamgauges in the central part of the United States. In general, the results for the illustrative example indicate that the NAUGLS method conveys more information than the NARI method to the network designer interested in maximizing regional information about mean annual flows with a limited budget.

DESIGN OF SURFACE WATER DATA NETWORKS FOR REGIONAL INFORMATION / La conception des réseaux pour les données des eaux de surface qui fourniront des renseignements régionaux

Abstract The concept of equivalent years of record, a measure of parameter accuracy, in conjunction with regression analysis simulation leads to a methodology for specifying sets of numbers of gauging stations and lengths of record required to attain a prespecified level of regional streamflow-parameter accuracy. This methodology can be used in network design to identify the number of stations and length of record that is the least cost network that meets the accuracy criteria. The methodology also indicates situations in which the criteria cannot be met without improving the underlying regionalization model.

Supply and demand and the design of surface-water supplies / Offre et demande et l'établissement d'une alimentation en eau de surface

Abstract Procedures that traditionally have been used to size water-supply reservoirs and to design related treatment plants and distribution systems have ignored the powerful tool of economic marginal analysis. Marginal analysis, when combined with Bayesian techniques, can be used to specify optimum designs for water supplies that are dominated by the need for over-year storage of streamflow. The combination of these two tools also provides a means for evaluating the alternative of collecting additional streamflow data prior to designing the water-supply system. A simplified example of the Bayesian marginal analysis that has ten years of streamflow data as an input demonstrates the design technique and evaluation of an added year of streamflow data. The economic demand curve and the cost analyses that are contained in the technique are assumed to be known. This assumption limits the utility of the technique until it can be extended to include the economic uncertainties.