A non-homogeneous Markov model for the definition of climate change scenarios for coastal areas: the case of Agro-Pontina plain

This study addresses to the possible changes in Agro-Pontino rainfall under different global warming scenarios for the 21st century. The Agro-Pontino-plain is a reclamation region and presents the typical hydro-geological features of Mediterranean coastal environments. It is densely populated and productive, therefore, climate changes could adversely affect the socio-economic development of the area. Currently, due to the coarse resolution of Global-Circulation-Models, local climate variables simulations for limited size area are not accurate. Nonetheless, GCMs simulations of large-scale upper-air fields are generally considered reliable, therefore to bridge the gap between GCMs and local-scale processes different downscaling techniques are carried out. Here, a Hidden Markov Model and a Non-homogeneous Hidden Markov Model are developed using a 54-years record (1951-2004) of daily rainfall amount at 9 stations in Agro-Pontino-plain and re-analysis fields of atmospheric variables. In HMM and NHMM runs, we directly consider the entire year, rather than an a priori demarcation of seasons. The idea is to identify, directly using the HMM, the seasonal precipitation characteristics which may be related to the temporal sequence of ‘hidden states’ of atmosphere, subsequently modeled as dependent on appropriate fields of selected atmospheric variables. Daily rainfall variability is described in terms of occurrence of 5 ‘hidden weather states’ identified by the HMM and associated to variables representing the main characteristics of large-scale atmospheric circulation as obtained by re-analysis data, then, using NHMM, calibration and validation tests are made to identify the optimal predictors - GeoPotential Height and Temperature at 1000 hPa, Meridional & Zonal Wind at 850 hPa and Precipitable Water - to reproduce better the observed rainfall features on Agro-Pontino-plain.