The Madden-Julian Oscillation (MJO) is a planetary-scale weather disturbance circling the tropical atmosphere every 30-60 days. The MJO has significant impact on global weather prediction beyond two weeks. The project attempts to answer how long and how well the MJO can be predicted by the global numerical weather prediction models. We use the World Meteorological Organization Subseasonal-to-Seasonal (S2S) datasets produced by the leading numerical weather prediction centers around the world. We have developed a new approach to quantify the prediction and predictability of the MJO convection using the OLR (Outgoing longwave radiation)-based MJO Index, which we found is capable of tracking the intraseasonal oscillations across all the seasons. Our quantitative analysis shows that the MJO convection can be predicted up to five weeks ahead by the best model (Figure 1), exceeding the conventional estimate by more than one week. The prediction and predictability of the intraseasonal convection depends on seasons: it is more predictable in the northern winter seasons than in summer in the S2S models.

The MJO cycles are closely related to the conditions at the earth’s surface and the top of the MJO. At the surface, the Maritime Continent is the key pathway for the MJO propagating eastward from the Indian Ocean to Pacific Ocean. Above the MJO towering clouds, the stratospheric quasi-biennial oscillations (QBO) significantly modulate the MJO amplitude, and propagation, which is known as the MJO-QBO connection. We have investigated how these two conditions can impact the MJO prediction in the S2S models. We found that all the S2S models display large bias in predicting the MJO convection near the Maritime Continent, which is considered as a barrier for the MJO prediction. The MJO-QBO connection in the S2S models is qualitatively consistent with the observation. Further analysis indicates that this connection comes from two sources: persistence of initial MJO and QBO states in the model, and dynamically predicated MJO/QBO by the model. We have tested these two using a forecast numerical model by switching stratospheric conditions to different QBO states, and found that tropospheric MJO conditions still have a dominant effect in explaining the apparent QBO influence in the forecast model. We expect that these findings will help improve the MJO simulation and prediction.    

The MJO-QBO connection is important for understanding and predicting the long term changes and trend in the MJO activities under climate change scenarios. Toward this goal, it is important to test whether or not the numerical models can faithfully simulate the observed MJO-QBO connection. For this, we have designed and conducted experiments using a hierarchy of state-of-the-art numerical models, from a cloud-resolving model, to a global numerical weather prediction model, and to a climate model. The cloud-resolving model with idealized large-scale dynamics simulates time variations of two MJO events. We varied the characteristics QBO wind and temperature anomalies in this model, and found that the QBO influence on MJO convection depends on both the amplitude and the height of the QBO temperature anomaly: lower-altitude and larger-amplitude temperature anomalies have more pronounced effects on the MJO convection. The MJO-QBO connection in the climate model was tested by nudging the stratosphere winds toward the observation so that the model bias in the QBO can be mostly eliminated. Nevertheless, the ensemble simulations still lack an MJO–QBO connection, which we attributed to the model bias in the MJO convection. These experiments lead us to conclude that the climate models with parameterized convection have difficulties capturing the observed MJO-QBO connection.  To correct this bias, however, is challenging, because of lack of theoretical understanding of the MJO-QBO interaction, and because parametrized convection in the climate models is highly uncertain.

The most uncertain physical process in the MJO is moist convection. To tackle this issue, we have examined how convection develops over the Maritime Continent region, focusing on idealized small islands. The key finding is that prevailing low-level wind together with diurnal variations in the surface temperature essentially control rain and local circulation around the island. The result implies that when the MJO passes densely populated large tropical islands, it is important to look at rain from island and the wind together, and understand their relationships.

The project servers the scientific community and society by generating new knowledge of the MJO, and by developing a new approach for the MJO prediction. The broad impact also lies in the training that the project offered to a PhD student, an undergraduate, an under-represented minority research associate.

Last Modified: 05/30/2021
Modified by: Shuguang Wang