Abstract
ABSTRACTDevelopment and dissemination of seasonal forecasts are integral components of the climate services provided by numerous meteorological services worldwide, offering estimates of meteorological variables on a seasonal time scale to aid local warning systems and decision‐making processes. The World Meteorological Organization (WMO) recommends that operational seasonal forecasts be objective and that the process be traceable and reproducible, including the selection and calibration of models. Following these guidelines, the Chilean Meteorological Service (Dirección Meteorológica de Chile, DMC) has implemented the next generation of seasonal forecasts, NextGen‐Chile. This new forecast system is based on a multi‐model ensemble using state‐of‐the‐art general circulation models (GCMs) from the calibrated North American Multi‐Model Ensemble (NMME) project. The forecasts from the GCMs are calibrated using a canonical correlation analysis‐based regression with a homogenised dataset of ground stations. The system is completed with two statistical models built using canonical correlation analysis on sea surface temperature (SST) in the ENSO and the Southwestern Pacific regions. Individually calibrated GCMs and statistical models are combined by weighing their hindcast skill to construct the final calibrated multi‐model ensemble (CMME) prediction. A verification analysis of probabilistic re‐forecasts during 2019–2021 has been performed, adding an average‐based ensemble forecast (CMME‐Mean). The CMME models outperformed the individual models in discrimination and showed less seasonal variability in performance than the individual models, adding consistency to the forecast. All metrics analysed during the verification process were maximised in the central region of Chile, which could be attributed to the high concentration of ground stations in the central region and the definition of a central region‐centred domain for the CCA calculation. Looking into the near future of NextGen‐Chile, a Flexible Seasonal Forecast is introduced as a more comprehensive approach for seasonal forecasts, allowing users and stakeholders to access information beyond the tercile seasonal forecast approach.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have