First, we compare the evolutions of precipitation anomalies averaged over SC from Day -15 to Day +15 among the four clusters (Fig. 2). The precipitation anomalies in all events turn from negative to positive on Day -4 and their magnitudes and horizontal patterns are of little difference (figures not shown).
Then, the evolutions of meridional advection of mean moisture by intraseasonal wind anomaly averaged over SC for each cluster is examined in Fig. 3. As mean moisture is maximum near the equator, positive (negative) moisture advection anomaly is induced by southerly (northerly) wind anomaly. All the clusters show significantly positive moisture advection anomaly over SC prior to Day -4, suggesting that the intraseasonal southerly anomaly plays a crucial role in premoistening SC.
The correspondence of tropical convection anomalies to each cluster is examined in Fig. 4, in which the left panels display horizontal patterns of the tropical OLR anomalies on Day 0 and the right panels show their zonal evolutions from Day -25 to Day 25 along the equator (20°S-20°N average). In Cluster 1, a zonal dipole pattern of OLR anomalies is prominent on Day 0, with an enhanced convection anomaly over the equatorial eastern Indian Ocean and a suppressed one over the western Pacific Ocean (Fig. 4a). Such pattern resembles that observed in the MJO phase 2-3, as defined by Wheeler and Hendon (2004). In Cluster 2, an enhanced convection anomaly is seen over the Maritime Continent, while suppressed ones are over the Indian Ocean and near the date line, respectively (Fig. 4c). Such distribution is similar to that found in the MJO phase 4-5. Furthermore, the enhanced convection anomalies in these two clusters both display a continuous eastward propagation along the equator, which initiate from the Indian Ocean and dissipate near the date line (Figs. 4b and 4d). These characteristics are identical to those in a typical MJO, although some phase lags can be found. In Cluster 3, a large area of suppressed convection anomaly is found over the eastern Indian Ocean and western Maritime Continent while a small area of enhanced convection anomaly is east of the Philippine Sea on Day 0 (Fig. 4e), but the convection anomalies do not propagate eastward continuously (Fig. 4f). The tropical convection anomalies are weak and scattered in Cluster 4 (Fig. 4g), and no significant signals could be found in the longitude-lagged time diagram (Fig. 4h). Thus, the first two clusters are accompanied by the MJO while the other ones are not.
Next, the correspondence of the mid-latitude ISO to each cluster is explored. Since Yao et al. (2020) revealed that the southward propagation of surface low pressure anomaly from the higher latitudes towards SC is responsible for the intraseasonal precipitation variation, Fig. 5 investigates the meridional propagation of 850-hPa geopotential height anomalies averaged over 80-120°E in each cluster. It is of interest to observe that all events are accompanied by significant negative geopotential height anomalies propagating southward from the higher latitudes, which reach SC on Day 0, although some detailed difference may exist. Such finding is consistent with Yao et al. (2020).
Figure 6 further presents evolutions of the horizontal patterns of 850-hPa geopotential height and wind anomalies from Day -8 to Day 0 with an interval of 2 days in each cluster. There are significant differences between Clusters 1-2 and Clusters 3-4. For Clusters 3-4, no significant geopotential height and wind anomalies can be found near the equator, but a large scale low-pressure anomaly is clearly seen over Euro-Asia continent in middle and high latitudes. The low pressure anomaly is near 60°N, 100°E on Day -8 and moves southward with time. The southwesterly wind anomaly across the eastern China on Day -6 which is crucial for SC premoistening as shown in Fig. 3 is clearly in association with the southward propagating low-pressure anomaly as the geostrophic wind response. This is consistent with the mechanism proposed by Yao et al. (2020).
In contrast, significant geopotential height and wind anomalies could be observed in the tropics in Clusters 1 and 2, showing the existence of the MJO. Furthermore, the southwesterly wind anomaly across the eastern China on Day -6 in Cluster 1 looks rather related to the anomalous anticyclone over the WNP than to the low-pressure anomaly over East Asia, for the continental low-pressure anomaly and the associated geostrophic wind is located too north.
The above results suggest that the precipitation events in Clusters 3 and 4 are primarily triggered by the mid-latitude ISO, while those in Clusters 1 and 2 are closely related to the tropical MJO as well. Therefore, we will further explore how the MJO trigger the precipitation events in the following section. Since Cluster 1 corresponds to an enhanced convection anomaly over the Indian Ocean and Cluster 2 over the Maritime Continent, hereafter they are referred to as the IO group and MC group, respectively.