Subseasonal-To-Seasonal (S2S) Prediction Skill of the Rainfall Diurnal Cycle Over the Maritime Continent and Its MJO Dependence

Abstract

The Maritime Continent (MC) is one of the rainiest regions globally, where the diurnal cycle contributes significantly to rainfall variability and can be modulated by the Madden-Julian oscillation (MJO). However, our understanding of diurnal cycle prediction skill and its relevance to the MJO in subseasonal forecast models remains limited. This study evaluates the subseasonal prediction skill of boreal wintertime (from November to January) rainfall diurnal cycle characteristics over the MC using ECMWF and NCEP S2S models and explores the influence of MJO activity on prediction skill. Evaluated diurnal cycle characteristics include the daily mean, diurnal range, and phase. Considerable errors appear in the first day forecast and slowly increase with lead time. Aggregation over longer lead windows can sustain or enhance prediction accuracy. Errors over land are attributed to earlier peaks and larger diurnal ranges than observations, while errors over the sea are typically found along coastlines and in some far ocean regions, such as the tropical northern West Pacific. Furthermore, we stratify the forecasted rainfall diurnal cycle by the MJO phase at forecast initialization to examine the MJO impact on prediction skill in S2S models. The predictions become less skillful when MJO convection reaches the MC because the models greatly underestimate the large diurnal range. This is related to insufficient low-level convergence at the grid scale and inaccuracies in representing the moisture-diurnal cycle relationship in the models. This study contributes to understanding the subseasonal rainfall prediction skill and MJO influence in rainfall prediction over the MC.