The problem addressed in this project if the estimation of the probability or return period for floods when the available information is for more than one site. Since floods events are non-deterministic, the prediction of such events is usually accomplished by using probabilistic modeling, usually based on data available at the site of interest. But often data exists at regional sites as well, and this is the information that will be used. The project will explore the joint modeling of flows, including parameter estimation, and determining the precision and robustness of the modeling/estimation procedures, testing them against actual flood data. Selected joint distribution functions that have marginals used in single-site flood prediction will be studied together with the joint models for which some relationship among their parameters are assumed. Procedures will be developed to estimate parameters and quantile points as well as deriving the standard errors of such estimators. Criteria will be developed to indicate under what conditions better estimators are obtained using multiple-site data, and the robustness of these procedures will be assessed. Software and corresponding manuals for the developed flood prediction procedures will be produced for personal computers whenever possible. The P.I.s are highly capable to carry out this work, thus a Standard Grant is recommended in the amount of $154,310 with a first increment of $79,526.
