Bayesian models of thermal and pluviometric time series in the Fucino plateau

This work was developed within the Project Metodologie e sistemi integrati per la qualificazione di produzioni orticole del Fucino(Methodologies and integrated systems for the quality improvement of horticultural products in the Fucino plateau), sponsored by the ItalianMinistry of Education, University and Research, Strategic Projects,Law 448/97. Agro-system managing, especially if necessary to achievehigh quality in speciality crops, requires knowledge of main featuresand intrinsic variability of climate. Statistical models may properly summarize the structure existing behind the observed variability, furthermore they may support the agronomic manager by providing the probability that meteorological events happen in a time window of interest. More than 30 years of daily values collected in four sites located on the Fucino plateau, Abruzzo region, Italy, were studied by fitting Bayesian generalized linear models to air temperature maximum/minimum and rainfall time series. Bayesian predictive distributions of climate variables supporting decision-making processes were calculated at different time scales, 5-days for temperatures and 10-days for rainfall, both to reduce computational efforts and to simplify statistical model assumptions. Technicians and field operators, even with limit-ed statistical training, may exploit the model output by inspecting graphs and climatic profiles of the cultivated areas during decision-making processes. Realizations taken from predictive distributions may also be used as input for agro-ecological models (e.g. models ofcrop growth, water balance). Fitted models may be exploited to moni-tor climatic changes and to revise climatic profiles of interest areas, periodically updating the probability distributions of target climatic variables. For the sake of brevity, the description of results is limited to just one of the four sites, and results for all other sites are availableas supplementary information.