Six rainfall data time series, collected in various sites in Italy, are analyzed in order to determine whether a linear trend or long memory is present. Various techniques for estimating the intensity of long memory are applied, including the rescaled range statistic, the aggregated variance method, the periodogram method and maximum likelihood. A modified variance-type estimator is used in order to detect long memory even when some types of nonstationarity are potentially present. The results indicate that a decreasing trend, though not statistically significant, is present in every record, and that long memory is significant in only two series. ¿ American Geophysical Union 1996