Abstract
Accurate Madden–Julian oscillation (MJO) prediction is important for estimating the magnitude of and preparing for upcoming extreme climate and weather events. Current methods for MJO prediction exhibit limited predictive skills due to inevitable bias. In this work, we devise an ensemble meritocracy strategy for deep learning-based bias correction for MJO prediction, namely MJO-Net. MJO-Net consists of two modules: a stochastic MJO prediction (SMP) module, which can generate thousands of stochastic predictions, and a skill evaluation (SEM) module, which selects the elite prediction members from the SMP outputs for the final prediction results. MJO-Net demonstrates competitive performances in 46-day MJO prediction compared with the long short-term memory (LSTM) and the original European Centre for Medium-Range Weather Forecasts (ECMWF) models. Our ensemble meritocracy strategy achieves higher skills than the traditional ensemble strategy, which indicates the great potential of our method in improving other ensemble-based weather and climate predictions.
© 2025 American Meteorological Society. This is an Author Accepted Manuscript distributed under the terms of the default AMS reuse license. For information regarding reuse and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).
