Regional snow depth frequency curves for avalanche hazard mapping in central Italian Alps

The annual maximum 3-day snow fall depth H-72 is Often adopted as an input for avalanche hazard mapping procedure. The quantile of H-72 with a reference return period T, in years, is evaluated using standard statistical distributions. The choice of the proper distribution is a key step. This is particularly critical in those sites where short observed series of H-72 are available, because the required return periods are in the order of 300 years. Here, the distribution of H-72 is investigated for a network of 40 gauging stations in the Alpine part of Lombardia region, in the central Italian Alps, using a data set dating back to 1985. A regional approach based on the index value is developed to estimate the T-years return period quantiles of H-72 at each single site i, H-72i(T). The index value is the single site sample average, (mu H72i). For each site, the values of H-72i scaled, i.e. divided, by mu(H72i), H-72i* are calculated. These are dimensionless quantiles with given return period. Some standard statistical tests are carried out, showing that the distribution of H-72i* is homogeneous in space, i.e. it is similar for each site in the considered area, which can then be viewed as a homogeneous region. Therefore, it is possible to use a regional approach for the frequency estimation of H-72*. The values of H-72i* are then grouped together in only one sample and their frequency of occurrence is accommodated by a general extreme value, GEV, probability distribution. The proposed distribution, valid in each site of the study area, can be used to assess the T-years return period quantiles of H-72i*. Then, by multiplying H-72i* by the single site sample index value mu(H72i), an estimate of the T-years quantile of H-72i at the single site is obtained. A sampling strategy is also suggested to estimate the single site index value, i.e. the average mu(H72i), critical for the evaluation of the single site distribution of H-72i. The final result is an estimate of the 3-day snow fall depth with a given return period, H-72i(T) for each site. Also, the standard error of estimation of H-72i(T) is given. It is shown that the value of H-72i(T) estimated by the regional approach is far more reliable than that calculated by the distribution fitting carried out using the data at each single site, particularly for high return periods. The values of H-72i(T) can be used as inputs to dynamic models, for avalanche hazard mapping exercise. The sensitivity analysis of the avalanche hazard maps to the uncertainty in the estimation of H-72i(T) can be also carried out, using the reference methods available in the present literature. (c) 2006 Elsevier B.V. All rights reserved.