Prediction Skill of the MJO Teleconnection Signals in the NASA GEOS Subseasonal Reforecasts

Tropical-extratropical teleconnections are considered key to advancing subseasonal prediction. The Madden Julian oscillation (MJO), characterized by large scale convective envelopes propagating along the tropical Indo-Pacific sector, is known to modulate midlatitude circulation and associated weather patterns. Although there is a general consensus on the MJO's influence on the midlatitude circulation, which is thought to be due to modulations of the North Atlantic Oscillation (NAO) and the Pacific North America (PNA) pattern, relatively less is known about the predictability of these teleconnection signals in dynamical forecast models. The composite evolution of the midlatitude circulation anomalies and associated wave train structure as the delayed response to tropical heating are reported in many studies that have examined reanalyses and long model simulations. However, it is yet to be determined whether they lend any beneficial subseasonal forecast skill, especially to weekly mean surface temperature and precipitation over North America. Investigating useful predictable signals from the MJO teleconnections is also complicated by the fact that the MJO is a moving heat source with an approximate periodicity of 30-60 days, and that the structure and amplitude of the midlatitude response can be sensitive to the longitudinal positioning of the heating anomaly as well as the propagation speed of the MJO. The objective of this study is to investigate the impact of MJO teleconnections on forecast accuracy at 2-3 week lead over North America, with an emphasis on the above-mentioned lesser known aspects of these teleconnections. To this end, we utilize a suite of subseasonal reforecasts performed with the latest NASA GEOS-5 seasonal-to-subseasonal (S2S) system. These reforecasts were performed as part of the NOAA SubX project, wherein the NASA GEOS-5 atmosphere-ocean coupled model was run at degree horizontal resolution, initialized every 5 days for the period 1999-2016. The GEOS-5 model shows skillful predictions of the MJO, with the correlation coefficient based on the real-time multivariate MJO (RMM) index staying at or above 0.5 up to forecast lead 26-36 days. The system is thus a useful tool for investigating MJO teleconnection processes.