The Italian Peninsula's climate is highly influenced by its complex topography and diverse regional weather systems, making high-resolution seasonal forecasting crucial for many societal sectors. Traditional seasonal prediction models, such as the CMCC SPSv3.5 (SPS), provide valuable insights but lack the spatial resolution necessary to capture local-scale climatic details. Thus, this study aims to provide a high-resolution seasonal forecast over Italy by enhancing SPS through statistical downscaling (SD) techniques tailored to the region's demand for finer-scale climate information. The SD method involves a three-step process that utilises observational datasets (ERA5 and CHIRPS) at 1/4° horizontal resolution and two machine-learning methods based on Empirical Quantile Mapping (EQM) and k -Nearest Neighbours ( k NN), translating 1° SPS forecasts into high-resolution fields by matching predicted conditions to observed patterns. Both SD methods were cross-validated over the 24-year hindcast period available (1993–2016), and the results indicate significantly enhanced seasonal predictions for the Italian Peninsula, with biases about 5–6 times smaller than those of the original SPS. The main component of this improvement is spatial accuracy, which allows the identification of domain characteristics that are unnoticed in SPS. The bias evaluated by lead time, key for seasonal forecasts, showed accuracy declining from lead month 1 onward. For instance, the 2 m temperature bias increased from −0.14/−0.31/−0.85 °C in lead month 1 to −0.68/−0.71/−1.41 °C in lead month 6 ( k NN/EQM/SPS), highlighting the challenge of maintaining predictive skill and the need for adaptive correction strategies to enhance lead-time reliability. Combining SD techniques with SPS outputs offers a solution for high-resolution seasonal predictions, supporting climate-sensitive applications by reducing forecast bias and improving spatial accuracy. • High-resolution seasonal forecasts over Italy via EQM and kNN downscaling. • Downscaling reduces SPS temperature biases by a factor of five to six. • kNN method enhances spatial detail, capturing orography and coastal contrasts. • kNN preserves seasonal cycles more consistently than EQM across lead months. • Refined seasonal forecasts supporting climate-sensitive applications.
Aragão et al. (Sun,) studied this question.