Regional frequency analysis of rainfall extremes in Tuscany (Italy)

The first and the second level of a hierarchical approach to regional frequency analysis of annual maximum of rainfall height for storm durations of 1 hour, 3, 6, 12, and 24 hours in Tuscany, are here described. Over a territory of about 22000 km2, the data of 470 rain gauges, in the period from 1923 to 2002, have been collected and investigated. Particularly 295 time series larger than 10 years, characterized by a low spatial correlation and included in each of the five dataset (i.e. storm durations of 1, 3, 6, 12, 24 hours), have been selected. The statistical homogeneous zones and sub zones have been defined using the L-moments method. The regional frequency analysis has been carried out on the basis of a two components extreme value distribution and the parameters have been evaluated with the joint ML estimation method. The growth factors, of the second level of the hierarchical regional frequency analysis, have been estimated for each sub-zones and return period up to 200 years. A third level of the regional frequency analysis is also here proposed, using a multivariate regressive model to assess the index rainfall value for storm duration from 1 to 24 hours and a given return period, in any location of the studied territory. A preliminary investigation has been carried out using the Mean Annual Precipitation (MAP). Particularly the relation between MAP and the rainfall extreme value at the rain gauge sites for a given storm duration and an assigned time period has been investigated. Daily registrations from 1970 to 2004 of a dataset of 601 rain gauges have been analysed to obtain the MAP value at each measurement site. Inverse distance weighted method have been used to elaborate the spatial distribution of MAP. The results show that the MAP values are nearly sufficient for the estimation of the index value for one of the three sub zones defined at the second level of the hierarchical approach. The analysis has been carried out fitting a regression model on MAP values and on other geographic and climatic characteristics such as elevation, aspects and mean value of the annual maximum rainfall with given time duration.