Assessment of seasonal forecasts of meteorological variables and irrigation requirements of maize in Pianura Padana

The impact of climate change on agriculture represents a crucial challenge for food and water security in the coming decades. In this context, seasonal forecasts are an essential tool for adaptation in the agricultural sector, offering valuable information for medium-term decision-making. This study, therefore, aims to analyze the uncertainty and to assess the consistency of seasonal forecasts of meteorological variables such as maximum temperature, minimum temperature, and precipitation in the Po River Basin, with a particular focus on the Pianura Padana area, where the agricultural production is concentrated. Moreover, the seasonal forecasts of irrigation requirements of maize in the area of interest will be estimated to assess the propagation of uncertainty to dependent variables. Medium-term climate forecasts rely on slow, complex changes in Earth's components. Climate models, initialized with data on the system's state, predict climate evolution over weeks or months. "Ensemble Members" represent possible evolutions, indicating the probability of different future scenarios. In this thesis, products from 2018 to 2022 from the ECMWF forecast center (SEAS5 system) and those generated by two multi-model ensembles (MME and MMES) only in 2022 will be examined. The goal is to assess the variability of the Ensemble Members over time and space by comparing different initialization times (including those from March to June), various lead times (from March to September), and the differences associated with the spatial grid cells at a 1°x1° horizontal resolution. Subsequently, the errors generated by ECMWF, MME, and MMES forecasts for the entire Po Basin area will be analyzed, comparing them with data from the ERA-5 climate reanalysis dataset at a 0.25°x0.25° resolution. A more detailed analysis will also be conducted on two grid cells in the Cuneo province, comparing ECMWF data both with ERA-5 data and with data downloaded from ARPA Piemonte, in order to highlight differences between forecasts in grid cells at elevations below 400m and those at elevations above 400m. Finally, using a model for estimating irrigation requirements, these needs can be calculated based on both observed and forecasted data, then compared to assess their quality. The results show that during the summer period, forecasts for precipitation and temperature have narrower ranges of variability compared to the spring period, with a corresponding lower error range. This pattern is particularly evident for temperatures, much less so for precipitation. In general, the errors associated with precipitation are significantly higher than those related to temperatures. Specifically, errors related to ten-day cumulative precipitation show magnitudes similar to the forecasts themselves, while the average errors over three intervals within a month can be very small. Moreover, minimum temperatures appeared to have larger errors than maximum temperatures, despite having narrower confidence intervals. However, through the estimation of bias using hindcast data and the subsequent bias correction of real-time data, it emerged that minimum temperatures showed a significant increase in accuracy, with very low error values. From the detailed analysis of the two grid cells, it was observed that the ECMWF products are more accurate when compared to ERA5 products than to the observed data from ARPA Piemonte. Comparing the single ECMWF model with the two multi-model ensembles, it was found that the multi-model ensemble with the largest number