Improving the prediction of the Madden-Julian Oscillation of the ECMWF model by post-processing

Abstract

The Madden-Julian Oscillation (MJO) is a major source of predictability on the sub-seasonal (10 to 90 d) timescale. An improved forecast of the MJO may have important socioeconomic impacts due to the influence of MJO on both tropical and extratropical weather extremes. Although in the last decades state-of-the-art climate models have proved their capability for forecasting the MJO exceeding the 5-week prediction skill, there is still room for improving the prediction. In this study we use multiple linear regression (MLR) and a machine learning (ML) algorithm as post-processing methods to improve the forecast of the model that currently holds the best MJO forecasting performance, the European Centre for Medium-Range Weather Forecasts (ECMWF) model. We find that both MLR and ML improve the MJO prediction and that ML outperforms MLR. The largest improvement is in the prediction of the MJO geographical location and intensity.

Ensemble forecast of an index of the Madden–Julian Oscillation using a stochastic weather generator based on circulation analogs

“…For the region with coordinates (15 • S-15 • N, 90-150 • E), the forecast skill is significant with COR 0.5, at T = 30 d (orange line in Fig. B2), which remains the same results for this region compared to (Silini et al, 2022). The RMSE for the different regions is quite the same (Fig.…”

“…We found that the ECMWF forecast has the highest correlation until 20 d compared to the SWG forecast. The RMSE values of the ECMWF forecast are always small compared to the SWG forecast, which indicates a good reliability skill of the ECMWF forecast for lead times of 5 and 10 d. However, for lead times of 20 d the RMSE of the ECMWF forecast coincides with the RMSE of the SWG, which shows the improvement of the SWG forecast to lead times above 20 d. The skill scores of the ECMWF (COR and RMSE) (Silini et al, 2022) are computed for each lead time, which is different from our way of computing the skill score (considering the average lead time). Of course, this comparison was made to check the performance of our forecast and not to say that the SWG model can replace a numerical prediction.…”

“…12). The ECMWF reforecasts were taken from Silini et al (2022). We found that the ECMWF forecast has the highest correlation until 20 d compared to the SWG forecast.…”

Causal inference and forescasting methods for climate data nalysis

“…Although MJO predictions obtained using ML models do not outperform dynamical models (Martin, Barnes, and Maloney, 2021; yet, a hybrid approach (Silini et al, 2022a), combining dynamical models and ML techniques, manages to improve the dynamical models results. In this way, it is possible to use dynamical models that have been developed across decades, based on physical phenomena, in combination with data-driven ML techniques, an approach that has shown its ability to reduce the gap between observations and dynamical models' forecasts (Haupt et al, 2021;McGovern et al, 2019;Rasp and Lerch, 2018;Scheuerer et al, 2020;Vannitsem et al, 2021).…”

“…Recently, also machine learning approaches have been used to predict the MJO (Martin, Barnes, and Maloney, 2021;Silini, Barreiro, and Masoller, 2021), yet not exceeding the prediction skill of the best dynamical models. Nevertheless, a combination of machine learning and dynamical models exceeded the prediction skill of the latter alone (Kim et al, 2021;Silini et al, 2022a), suggesting a promising route to follow in the future to improve the prediction of the MJO.…”