Drought identification and probabilistic characterization

An effective drought monitoring system, able to provide a timely warning about the possible onset of a drought event, as well as to describe its evolution in time and space, is a necessary premise to adequately mitigate drought impacts. Therefore, an accurate selection of methods and tools for drought identification and characterization to be implemented within the drought monitoring system is required. In the first part of the paper, an analysis of the applicability of some of the most known drought indices, as valuable tools to monitor drought phenomena through the different components of the natural hydrological cycle, is presented. In particular, the Palmer index and the Standardized Precipitation Index (SPI) are described and applied to the Simeto river basin in Sicily, revealing that the former is generally more severe in classifying hydrometeorological conditions. With reference to the SPI, an analysis is carried out oriented to select the most appropriate probability distribution for precipitation, as well as to evaluate the uncertainty related to its computation due to limited sample size. Also, a stochastic methodology for drought forecasting through the SPI index is proposed. The investigation on drought indices is performed on monthly precipitation series, covering the period 1921-2003, and on monthly temperature series from 1926 to 2003, observed in Sicily. The second part of the paper deals with the probabilistic study of drought characteristics, namely duration and accumulated deficit, and in particular with the assessment of return periods corresponding to severe drought events, which can provide useful information for water resources systems planning and management. In particular, an analytical model to evaluate the return period of drought events, based on a bivariate probability distribution of duration and accumulated deficit is derived, that enables to compute the probabilistic features of droughts on the basis of the parameters of the underlying hydrometeorological variable (e.g. annual precipitation) and on the fixed threshold considered for drought identification. The proposed model is applied to study spatial distribution of drought return period with predefined accumulated deficit and duration over the Simeto river basin.