The problem of determining a minimum cost design of a multilevel branching storm sewer system is formulated by using a serial approach to describing the system and proceeding through the computational algorithm. Discrete differential dynamic programing (DDDP) is the basic search technique employed. Results for an example using the serial approach are compared to those achieved by using an earlier nonserial DDDP approach. The several advantages of the serial approach become increasingly important as the size and level of branching of the system increase. These advantages are outlined and discussed, the superiority of the serial over the nonserial approach being demonstrated.

No abstract

The reliability of simulation results produced by watershed runoff models is a function of uncertainties in nature, data, model parameters, and model structure. A framework is presented here for using a reliability analysis method (such as first‐order second‐moment techniques or Monte Carlo simulation) to evaluate the combined effect of the uncertainties on the reliability of output hydrographs from hydrologic models. For a given event the prediction reliability can be expressed in terms of the probability distribution of the estimated hydrologic variable. The peak discharge probability for a watershed in Illinois using the HEC‐1 watershed model is given as an example. The study of the reliability of predictions from watershed models provides useful information on the stochastic nature of output from deterministic models subject to uncertainties and identifies the relative contribution of the various uncertainties to unreliability of model predictions.

This paper describes an evaluation of the assumption commonly employed in drainage design that the return period of the rainfall used to design a system is the same as the peak flow produced by that rainfall. Specifically, the sensitivity of the frequency response of three catchments to design storm parameters is examined. Parameters include, hyetograph shape, antecedent soil moisture and rainfall duration. A continuous simulation model is used to compute simulated historical frequency responses for three different long term rainfall records. Design storms are also developed from depth-duration-frequency analysis of the rainfall data. Comparisons are made on frequency graphs. It is concluded that significant parameter sensitivity exists but that an appronriate choice of design storm parameters can produce a design which yields peak flows of the desired return period.

No abstract