Recent Challenges in the APCC Multi‐Model Ensemble Seasonal Prediction: Hindcast Period Issue.

Seasonal forecasts are commonly issued in the form of anomalies, which are departures from the average over a specified multiyear reference period (climatology). The model climatology is estimated as the average of the retrospective forecasts over the hindcast period. However, different operational centers that provide seasonal ensemble predictions use different hindcast periods based on their model climatology. Additionally, the hindcast periods of recently developed and upgraded newer models have shifted in the recent years. In this paper, we discuss the recent challenges faced by APCC multi‐model ensemble (MME) operations, especially changes in the hindcast period for individual models. Based on the results of various experiments for MME prediction, we propose changing the hindcast period, which is the most appropriate solution for APCC operation. This makes the newly developed models join the MME and increases the total number of participating models, which facilitates the skill improvement of the MME prediction. Plain Language Summary: In seasonal forecasting, it is well known that the MME, which combines different single‐model predictions from various operational and research centers, is a more effective way to improve forecast skill. Since 2005, the APCC has provided the MME seasonal forecasts, and the models participating in the APCC MME operations have been continuously changing. In particular, as the hindcast periods of newly developed models shift to the latest, they cannot participate in operational MME forecasts because of climatological discrepancies. However, over time, as the number of new models expected to provide skillful forecasts gradually increases, the APCC faces the challenge of continuously reducing the number of participating models or changing the hindcast period to more recent years. Considering various aspects such as the number of participating models, skills, and climatology period, we selected the most appropriate method for APCC operation. Thus, the MME prediction skill has improved over most of the globe and seasons because of the increase in the number of participating models, particularly the inclusion of newer models. Key Points: APCC, which combines all the information from different ensemble prediction systems, recently faced challenges in hindcast period issuesThe proposed solution leads to an increase in the number of models contributing to MME prediction, particularly recently developed modelsIt shows improved skills for both temperature and precipitation predictions over most of the globe and seasons [ABSTRACT FROM AUTHOR]