This document provides responses to public comments on the U.S. Environmental Protection Agency’s

[0] In the analysis of rainfall time series the assumption of stationarity is generally postulated. This key assumption undoubtedly facilitates mathematical tractability of stochastic models employed for either descriptive or predictive purposes. However, nature is complex enough so as to accommodate alternations between stationary and non-stationary modes, possibly co-existing in a time series of rainfall measurements. Formal statistical testing of the hypothesis of stationarity in data is very important, since the presence of non-stationarities can lead to seriously questionable results and misinterpretations of data analyses relying on stationarity ad hoc. The present work tests the hypothesis of weak stationarity on long rainfall time series records of high resolution (8 years with resolution of 1 hour at Marghera, Italy and 25 years with resolution 5 minutes at Firenze, Italy). This task is carried out with the implementation of non-parametric statistical procedures appropriate for testing homogeneity of mean, variance, and covariance structure, over time. The present work is also concerned with the effects of non-stationarity on the variance-time function of series of aggregates over a range of temporal scales of aggregation. Such effects are discussed with regard to observed rainfall data, and with regard to synthetic data obtained from simulations of certain non-stationary model processes constructed here for illustrative purposes.