Elevation-driven biases in seasonal weather forecasts: Insights from the Alpine region

Generating reliable weather forecasts over mountainous regions is challenging. Indeed, global seasonal weather forecasting systems inherit biases that affect hydrological applications. In this context, we investigate the dependence of seasonal weather forecast biases on terrain characteristics in the Trentino-South Tyrol region, northeastern Italian Alps, with reference to precipitation and 2-m temperature variables. To this end, the European Centre for Medium-Range Weather Forecasts (ECMWF) fifth-generation seasonal forecast system (SEAS5) dataset is used at a horizontal spatial resolution of 0 . 125 ° × 0 . 125 ° with 25 ensemble members in the hindcast period from 1981 to 2016. The observational reference dataset used to evaluate the biases is a regional product routinely adopted in the interested area for hydrological applications, characterized by daily gridded precipitation and mean temperature at a spatial scale of 250 m × 250 m during the time window 1980–2018. The spatio-temporal variation of the biases at the monthly timescale is investigated here, and its dependence on elevation is interpreted using linear regression models. A simple correction model is also devised to investigate and analyze the linear correlation between forecast biases and elevation bias and elucidate the role of ensemble members in bias generation. • Biases were computed using a high-resolution gridded observational dataset over the region. • Analyzed ECMWF SEAS5 monthly forecast bias dependence on elevation. • Observed elevation dependence differences using ensemble mean vs. mean of ensemble bias. • Linear correction models reduced MAE, confirming elevation-dependent forecast biases. • Inferred linear dependencies may not fully capture forecast biases in complex terrain.