A basic problem in hydrology is the computation of confidence levels for the value of the T-year flood when it is obtained from a log Pearson III distribution using the estimated mean, estimated standard deviation, and estimated skew. Here we give a practical method for finding approximate one-sided or two-sided confidence intervals for the 100-year flood based on data from a single site. These confidence intervals are generally accurate to within a percent or two, as tested by simulations, and are obtained by use of a neural network. ¿ 1999 American Geophysical Union