Advancing sub-seasonal to seasonal multi-model ensemble precipitation prediction in east asia: Deep learning-based post-processing for improved accuracy.

The growing interest in Subseasonal to Seasonal (S2S) prediction data across different industries underscores its potential use in comprehending weather patterns, extreme conditions, and important sectors such as agriculture and energy management. However, concerns about its accuracy have been raised. Furthermore, enhancing the precision of rainfall predictions remains challenging in S2S forecasts. This study enhanced the sub-seasonal to seasonal (S2S) prediction skills for precipitation amount and occurrence over the East Asian region by employing deep learning-based post-processing techniques. We utilized a modified U-Net architecture that wraps all its convolutional layers with TimeDistributed layers as a deep learning model. For the training datasets, the precipitation prediction data of six S2S climate models and their multi-model ensemble (MME) were constructed, and the daily precipitation occurrence was obtained from the three thresholds values, 0 % of the daily precipitation for no-rain events, <33 % for light-rain, >67 % for heavy-rain. Based on the precipitation amount prediction skills of the six climate models, deep learning-based post-processing outperformed post-processing using multiple linear regression (MLR) in the lead times of weeks 2-4. The prediction accuracy of precipitation occurrence with MLR-based post-processing did not significantly improve, whereas deep learning-based post-processing enhanced the prediction accuracy in the total lead times, demonstrating superiority over MLR. We enhanced the prediction accuracy in forecasting the amount and occurrence of precipitation in individual climate models using deep learning-based post-processing.

Advances and challenges of operational seasonal prediction in Pacific Island Countries.

Seasonal climate forecasts play a critical role in building a climate-resilient society in the Pacific Island Countries (PICs) that are highly exposed to high-impact climate events. To assist the PICs National Meteorological and Hydrological Services in generating reliable national climate outlooks, we developed a hybrid seasonal prediction system, the Pacific Island Countries Advanced Seasonal Outlook (PICASO), which has the strengths of both statistical and dynamical systems. PICASO is based on the APEC Climate Center Multi-Model Ensemble (APCC-MME), tailored to generate station-level rainfall forecasts for 49 stations in 13 countries by applying predictor optimization and the large-scale relationship-based Bayesian regression approaches. Overall, performance is improved and further stabilized temporally and spatially relative to not only APCC-MME but also other existing operational prediction systems in the Pacific. Gaps and challenges in operationalization of the PICASO system and its incorporation into operational climate services in the PICs are discussed.

Toward a better multi-model ensemble prediction of East Asian and Australasian precipitation during non-mature ENSO seasons.

An effective and reliable way for better predicting the seasonal Australasian and East Asian precipitation variability in a multi-model ensemble (MME) prediction system is newly designed, in relation to the performance of predicting El Niño-Southern Oscillation (ENSO) and its impact. While ENSO is a major predictability source of global and regional precipitation variation, the prediction skill of precipitation is not solely due to typical ENSO alone, of which variability and predictability exhibit strong seasonality. The first mode of ENSO variability has large variance with high prediction skill for boreal winter and small variance with low skill for spring and summer, while the second mode shows the opposite phase. The regional prediction skills for Australasian and East Asian precipitation also show such seasonal dependence, with low skill and large spread of individual models' skills during the boreal spring to summer and high skill and small spread during winter. Using the individual models' reproducibility of the association between ENSO and regional precipitation, the prediction skills of the MME with selected models can improve at regional levels, compared to those for all-inclusive MME, during boreal spring to summer. While typical ENSO as a predictability source may still dominate during boreal winter, consideration of complex ENSO structure and its diverse impact can lead to a better prediction of regional precipitation variability during non-mature phase of ENSO seasons.

How do the strength and type of ENSO affect SST predictability in coupled models.

The effects of amplitude and type of the El Niño-Southern Oscillation (ENSO) on sea surface temperature (SST) predictability on a global scale were investigated, by examining historical climate forecasts for the period 1982-2006 from air-sea coupled seasonal prediction systems. Unlike in previous studies, SST predictability was evaluated in different phases of ENSO and for episodes with various strengths. Our results reveal that the seasonal mean Niño 3.4 index is well predicted in a multi-model ensemble (MME), even for four-month lead predictions. However, coupled models have particularly low skill in predicting the global SST pattern during weak ENSO events. During weak El Niño events, which are also El Niño Modoki in this period, a number of models fail to reproduce the associated tri-pole SST pattern over the tropical Pacific. During weak La Niña periods, SST signals in the MME tend to be less persistent than observations. Therefore, a good ENSO forecast does not guarantee a good SST prediction from a global perspective. The strength and type of ENSO need to be considered when inferring global SST and other climate impacts from model-predicted ENSO information.