PROBABILITY OF FLOODS WITH VERY LONG RETURN PERIODS

THIS PROPOSAL IS DIRECTED TO THE DEVELOPMENT OF A PROCEDURE FOR GENERATING LONG STREAMFLOW SEQUENCES. IT IS PROPOSED TO FIRST DEVELOP A LONG SEQUENCE OF HOURLY RAINFALL BY STOCHASTIC METHODS AND TO TRANSFORM THIS RAINFALL SEQUENCE INTO STREAMFLOW WITH A DETERMINISTIC HYDROLOGIC MODEL. PHASE I WILL FOCUS ON THE A STOCHASTIC RAINFALL MODEL CAPABLE OF DEVELOPING RAINFALL SEQUENCES AT MULTIPLE SITES. THE WORK OF FRANZ AND KELMAN WILL BE THE BASIS OF THIS MODEL. A 1000-YR FLOW SEQUENCE WILL BE DEVELOPED. THE ANNUAL FLOOD PEAKS WILL BE STUDIED TO DETERMINE THE BEST FIT FREQUENCY DISTRIBUTION. SUBSAMPLES OF 100, 200, 333, AND 500 YEARS WILL BE ANALYZED TO DETERMINE THE VARIATION OF UNCERTAINTY IN FLOOD FREQUENCY ESTIMATES WITH SAMPLE SIZE. THIS INFORMATION WILL BE USED IN PLANNING PHASE II OF THIS RESEARCH, WHICH WILL REQUIRE APPLICATION OF THE METHODS TO DATA FROM SEVERAL CATCHMENTS REPRESENTING A VARIETY OF HYDROLOGIC REGIMES.