Can the use of NAO index improve the stochastic forecasting of drought?

Drought monitoring and forecasting play a very important role for an effective drought preparedness and management. Within this context, the use of large scale climatic patterns that supposedly exerts an influence on the climatic variability in a region, such as ENSO, NAO or EB, can potentially improve the early warning of severe drought impacts and the forecasting of its evolution in time and space. In the paper, a model able to estimate transition probabilities of Standardized Precipitation Index classes corresponding to drought of different severities is extended in order to include information provided by an exhogenous variable such as a large scale climatic index. The analytical structure of the model enables to overcome the difficulties related to the relatively limited number of droughts generally observed in historical records. The model has been applied and tested with reference to SPI series computed from areal precipitation in Sicily, making use of NAO as exhogenous variable. More specifically, transition probabilities between SPI-based drought classes have been evaluated in terms of conditional probability of SPI drought class at the future month t+M, given the SPI drought class at the current month t, for different starting months and time horizon M. Then the analysis has been repeated by also taking into account the present NAO value. Results seem to indicate that drought transition probabilities are affected by the NAO index, especially by its negative values and for transitions from Extreme drought to Extreme drought and from Extreme drought to Non-drought condition. The statistical significance of such variation has been tested by means of a Montecarlo analysis under the null hypothesis of no correlation between SPI and NAO, and reveals that the effect of NAO on drought transition in Sicily should be considered significant.