Intraseasonal Variability of Apparent Heat Source over the Arabian Sea and Its Relationship with East Asian Summer Precipitation

Abstract

Boreal summer intraseasonal oscillation (BSISO) is highly related to summer monsoon activities, tropical cyclones, flood disasters, and other extreme weather events in the Northern Hemisphere. The propagation of BSISO has considerable complexity. The apparent heat source (Q₁) is the heat generated by radiation, heat conduction, and latent heat release, and their anomalies significantly affect the atmospheric circulation and relevant precipitation. We selected 27 significant events from 30- to 60-day Q₁ anomalies in the northeast Arabian Sea (12–22° N, 66–76° E). K-means cluster analysis was used to further divide significant events into 19 Type-I events and 8 Type-II events. In the equatorial region, the Type-I events have continuous eastward propagation, while the Type-II events have no significant eastward propagation features before −10 days. In East Asia, the northward propagation of the Type-I events is significant and continuous, while there is no northward propagation of the Type-II events. The moisture analyses show that the horizontal advection term plays the most important role in the propagation of convection in most regions. The evaporation term of the Type-I events also plays a significant role in East Asia, and may be related to the difference between the two types of events there.

1. Introduction

Intraseasonal oscillation (ISO) is a periodic oscillation with a time scale of 10–90 days, which is significantly correlated with the evolution of the monsoon system [1]. The land–sea thermal configuration is one of the main reasons for the formation of monsoons. Diabatic heating is the leading thermal forcing factor promoting atmospheric motion, and the feedback of cumulus convective heating is an important dynamic mechanism stimulating ISO of the tropical atmosphere [2]. The climatology shows that the equatorial Indian Ocean–South Asia–equatorial Western Pacific region is the most extensive atmospheric heat source region in the world and has a strong influence on precipitation anomalies [3].

Boreal summer intraseasonal oscillation (BSISO) is a kind of propagating convection–circulation coupling system with a significant baroclinic structure [4,5,6], and is an important factor affecting the short-term climate in Asia [7]. Based on an EEOF analysis, Kikuchi et al. proposed the bimodal ISO index to represent MJO in boreal winter and BSISO in boreal summer [8]. Lee et al. defined a BSISO index based on the MV-EOF analysis of the daily OLR field and the 850 hPa wind field in the Asian summer monsoon region. BSISO1, composed of the first two modes, represents BSISO with a 30–60-day period that typically propagates northward and eastward. BSISO2, which consists of the third and fourth modes, mainly represents the oscillation with a period of 10–20 days and propagates northward and westward. The 30–60-day BSISO convection first appears in the equatorial Indian Ocean and then spreads eastward and northward, with the northward component reaching the Arabian Sea and Bay of Bengal, and the eastward component reaching the Maritime Continent and then spreading northward to the South China Sea and the Western Pacific Ocean.

Chen and Wang divided the 30–60-day boreal summer intraseasonal oscillation (BSISO) into three types using cluster analysis: the canonical modes of eastward and northward propagation, the dipole mode of northward propagation, and the eastward propagation mode [9]. During the Indian summer monsoon (ISM) season, the northward propagation of BSISO results in the intraseasonal anomaly of SST and precipitation over the East Arabian Sea and the Bay of Bengal. The BSISO that propagates eastward from the equatorial Indian Ocean is an important source of 30–60-day BSISO in East Asia [10]. It propagates eastward along the equator in the form of a coupled Kelvin–Rossby wave, reaches the Maritime Continent and then propagates northward, and finally dissipates in the south of China.

Kikuchi outlined proposed mechanisms that may affect the northward propagation of BSISO, including wind shear, water vapor convection feedback, and air–sea interactions [11]. Wang and Xie suggested that the northward propagation is caused by Rossby wave radiation. Chen and Wang proposed that northward propagation is mainly driven by horizontal humidity advection related to the Rossby wave response, while eastward propagation along the equator is mainly driven by vertical moisture transport related to the Kelvin wave. In terms of air–sea interactions, BSISO induces significant SST changes in the Arabian Sea, the Bay of Bengal, the eastern equatorial Indian Ocean Basin, and the South China Sea [12,13]. A positive sea surface temperature (SST) anomaly before BSISO convection supports its northward propagation. The positive SSTA leads BSISO convection by a quarter period, and SST affects the convergence and divergence of the boundary layer and changes the sensible and latent heat fluxes at the air–sea interface [14]. In tropical monsoon areas, the increasing downward shortwave radiation and easterly winds anomaly increase the sea surface heat and reduce evaporation. These cause an increase in SST, which in turn increases atmospheric instability and promotes the development of convection. The active convection weakens the downward shortwave radiation and the west wind anomaly, which cools the sea surface, enhances evaporation, reduces SST, increases atmospheric stability, and weakens convection, forming a cycle [15].

Atmospheric intraseasonal oscillation can induce intraseasonal changes in SST [16] and mixed layer [17] and sea surface height [18,19] through thermal and dynamic effects. The intraseasonal variation in SST can also have a significant effect on the atmosphere, further influencing precipitation by altering the instability and divergence of the lower atmosphere. In addition, BSISO can also influence atmospheric circulation and weather at mid-latitudes through the excited Rossby wave train [20,21,22,23].

Atmospheric apparent heat source (Q₁) is the heat generated by radiation, heat conduction, and latent heat release. The Arabian Sea, located on the southwest side of the Tibetan Plateau, and the Bay of Bengal are the regions with the strongest 30–60-day intraseasonal oscillation of Q₁. In the 30–60-day intraseasonal oscillation of the East Asian monsoon system, Q₁ shows two oscillation bands in the north and south, which are associated with the tropical and subtropical monsoon. The seasonal difference in the atmospheric heat source in the tropical Indian Ocean is one of the reasons for monsoons, which mainly affects the precipitation in East Asia through upper- and lower-level circulations [24]. The heat fluxes in the Northwestern Pacific Ocean, the Arabian Sea, and the Bay of Bengal are closely related to the Asian monsoon. Compared with sensible heat and longwave radiation, latent heat is the largest heat loss term in the tropical ocean surface, which is affected by the vertical gradient of wind speed and sea surface specific humidity. With the onset of the Asian summer monsoon, the latent heat flux in the Arabian Sea, the Bay of Bengal, and the South China Sea increases significantly, which provides moisture sources for the Asian summer precipitation [25].

Previous studies have shown that the BSISO has a significant impact on the summer climate variation in both South Asia and East Asia, and the distribution of heat anomalies will affect atmospheric circulation and climate change in the Asian monsoon region. In the past 30 years, the intraseasonal precipitation and the variability of 30–60-day BSISO in the northeast Arabian Sea have increased. This phenomenon only occurs in the Arabian Sea because its warming is faster than any other region [26]. There are few previous studies that focus on the relationship of BSISO and precipitation between the Arabian Sea and East Asia. This research discusses its relationship with summer precipitation in East Asia based on the intraseasonal anomaly of Q₁ in the Arabian Sea.

2. Data and Methods

2.1. Data

We utilized u-wind, v-wind, air pressure, potential temperature, vertical velocity, precipitation, and specific humidity data from the ERA-Interim daily dataset at the 1° × 1° horizontal resolution during the period of 1997–2018. The anomalous intraseasonal fields were acquired by removing the time mean and climatological annual cycle and then processing with a 30–60-day band-pass filter. Lastly, the data from 1 June to 30 September each year were chosen for subsequent analysis [27].

2.2. Calculation of Apparent Heat Source (Q₁) and Apparent Moisture Sink (Q₂)

The apparent heat source and apparent moisture sink were calculated based on the following method proposed by Yanai et al. [28,29,30]:

$${\mathrm{Q}}_1 ={ \mathrm{C}}_{\mathrm{p}}{\left({\displaystyle \frac{\mathrm{p}}{{\mathrm{p}}_0}}\right)}^{{\displaystyle \frac{\mathrm{R}}{{\mathrm{C}}_{\mathrm{p}}}}}\left({\displaystyle \frac{\partial \mathsf{\theta }}{\partial \mathrm{t}}} + \mathrm{V}\cdot \nabla \mathsf{\theta } + \mathsf{\omega }{\displaystyle \frac{\partial \mathsf{\theta }}{\partial \mathrm{p}}}\right)$$

(1)

$${ \mathrm{Q}}_2= -\mathrm{L}\left({\displaystyle \frac{\partial \mathrm{q}}{\partial \mathrm{t}}}+\mathrm{V}\cdot \nabla \mathrm{q}+\mathsf{\omega }{\displaystyle \frac{\partial \mathrm{q}}{\partial \mathrm{p}}}\right)$$

(2)

where C_p is the specific heat capacity at constant pressure, p is the air pressure, p₀ = 1000 hPa, R is the gas constant, θ is the potential temperature, V is the horizontal wind (calculated using u-wind and v-wind), ω is the vertical velocity, L represents the condensation latent heat coefficient, q is the specific humidity, and ∇ is the isobaric gradient operator.

The column-integrated apparent heat source and apparent moisture sink can also be characterized as:

$$⟨{\mathrm{Q}}_1⟩ = ⟨{\mathrm{Q}}_{\mathrm{R}}⟩ + \mathrm{LP} + \mathrm{S}$$

(3)

$$\begin{array}{c}⟨{\mathrm{Q}}_2⟩=\mathrm{L}\left(\mathrm{P}- \mathrm{E}\right)\end{array}$$

(4)

$$\begin{array}{c}〈〉={\displaystyle \frac{1}{\mathrm{g}}}{\int }_{{\mathrm{p}}_{\mathrm{T}}}^{{\mathrm{p}}_{\mathrm{S}}}\left(\right)\end{array}\mathrm{d}\mathrm{p}$$

(5)

where p_T represents the pressure at the cloud top, p_S is the surface pressure, Q_R is the atmospheric radiation, P and E are precipitation and evaporation, respectively, and S is the sensible heat flux. It can be seen that Q₁ is mainly composed of heating due to atmospheric radiation, the release of latent heat (L) by precipitation (P), and the sensible heat flux (S). Q₂ is mainly accounted for by the net latent heat flux due to precipitation (P) and evaporation (E), as we can see from Equation (4).

2.3. Selection of Significant Events

We identified a key region A (12–22° N, 66–76° E) around the Arabian Sea with significant intraseasonal variability. Significant events were selected according to the regional averaged Q₁ anomaly time series in region A. Significant events are those where the peak is greater than one standard deviation and there is at least one value (trough) less than one standard deviation below the mean. Finally, 27 significant events were selected. One event was divided into 8 phases and each phase lasts 5 days long (a pentad). The phase 0 is defined as the pentad mean of intraseasonally filtered daily data, with the central date being day 0. Pentad 1 (−1) is defined as the pentad mean succeeding (preceding) pentad 0, and so on [31]. Each event was defined from phase −4 to +3, and composite analysis was then conducted for phases −2 to +1 of the selected events. Figure 1 shows an example time series (2010) showing the selection criteria, with the highest point of a large event indicated.

2.4. K-Means Cluster Analysis

The significant events were analyzed by K-means cluster analysis following the methods of Wang et al. [31] and Chen and Wang [9]. Cluster analysis is an analysis method that classifies research objects based on their characteristics, so that the individual differences between the same groups are relatively small, and the differences between different groups are relatively large. K-means cluster analysis first randomly selects K objects as the initial cluster center and assigns them to the nearest cluster center by calculating the Euclidean distance between each object and each cluster center. The cluster center is then recalculated after assigning each sample, and this step is repeated until the cluster centers no longer change and the cluster sum of squares reaches its minimum. The selection of K can be judged by SSE (within-cluster sum of squared errors) and S (silhouette coefficient). When the SSE reaches the inflection point or the S reaches its maximum, the corresponding K value is the most suitable. The SSE and the S were both 2, indicating that it is reasonable to divide significant events into two categories (K = 2). The temporal domain for cluster analysis is from pentad −1 to pentad 3.

2.5. Moisture Budget

The intraseasonal column-integrated moisture equation is:

$${\displaystyle \frac{\partial ⟨\mathrm{q}{⟩}^{\prime }}{\partial \mathrm{t}}} = -⟨\mathrm{V}\cdot \nabla \mathrm{q}{⟩}^{\prime } -{ ⟨\mathsf{\omega }{\displaystyle \frac{\partial \mathrm{q}}{\partial \mathrm{p}}}⟩}^{\prime } -{ \mathrm{P}}^{\prime } +{ \mathrm{E}}^{\prime }$$

(6)

where q is the specific humidity, V is the horizontal wind, ω is the vertical velocity, p is the air pressure, P is the precipitation, E is the evaporation, ’ represents the intraseasonal component, and $\langle \rangle$ represents column-integration from surface to 100 hPa. The left term represents moisture tendency, the first term on the right represents intraseasonal horizontal advection, the second term represents the intraseasonal vertical moistening due to convection, P is the precipitation term, and E is the evaporation term [9].

3. Results

3.1. EOF Analysis

Power spectrum analysis (Figure 2) of the unfiltered Q₁ in region A shows a clear intraseasonal period of around 42 days. We then describe the 30–60-day band. The results of the EOF analysis on the Arabian Sea show that the Q₁ from June to September has a significant intraseasonal oscillation over the region A (Figure 3). Therefore, the composite analysis was performed based on averaged Q₁ time series from 1 June to 30 September of each year in region A.

During significant events (Figure 4), Q₁ and Q₂ exhibit typical BSISO distribution and propagation patterns. Q₁ (Q₂) anomalies generated in the equatorial Indian Ocean then spread northward and eastward, eventually forming a northwest–southeast banded distribution, while the opposite anomaly appeared in the Indian Ocean, forming a cycle. Compared with Q₁, the distribution of Q₂ in the region south of 20° N is basically the same as Q₁, but the trend of Q₁ and Q₂ anomalies drastically diversified past 20° N in East Asia. South of 20° N, the spatial patterns of Q₁ and Q₂ anomalies are very similar, while north of 20° N, there are significant differences. Patterns of precipitation anomalies in East Asia are closer to Q₁ anomalies than to Q₂ anomalies.

In significant event phase 0, there are anomalous cyclones in the eastern Arabian Sea to the Bay of Bengal, the South China Sea, and the western equatorial Pacific, corresponding to the positive anomalies of Q₁ and precipitation at this phase. There is an anomalous anticyclonic from the East China Sea to southern Japan, which corresponds to a positive divergence anomaly of moisture flux and a negative precipitation anomaly. Hao et al. mentioned that, during the eastward movement of MJO, anticyclonic circulation was stimulated on the northern side of the convective area [32]. Although MJO storm clouds could not move directly north to high latitude to directly affect precipitation in northern China, anticyclonic clouds formed on the northern side of the cyclone could strengthen the southerly winds in northern China. It provides favorable conditions for precipitation, which is consistent with the distribution of moisture flux divergence, moisture flux, and precipitation in northern China during phase 0.

3.2. Composite Analysis of Two Types of Significant Events

The significant events clearly show the distribution and change features of BSISO, so significant events are further classified in this section. All events were divided into two classes (Type-I and Type-II) according to the spatiotemporal evolution of Q₁ anomalies within the domain of 10° S–25° N, 50–150° E (yellow box of Figure 5) using K-means clustering. Before clustering, Q₁ was 5° × 5° averaged to reduce the factors considered in clustering. We ended up with 19 Type-I events and 8 Type-II events.

The distribution of the Type-I events is similar to that of significant events, but the moisture flux of the Type-I events is significantly larger than that of the overall significant events. In phase 0 to 2 of Type-II events (Figure 6), the negative Q₁ anomaly originating in the central Indian Ocean does not significantly propagate to the east. There is also northward propagation in Type-II events, but it is weakened during the propagation. In Type-II events, the northwest–southeast extended anomaly is not as strong and continuous as the banded anomaly of the Type-I events, and there will be breakage in the middle. This kind of distribution is similar to the northward dipole model presented in Chen and Wang [9], but the difference is that the north dipole model does not include eastward propagation but propagates northward in the Indian Ocean and the Western Pacific Ocean. However, for the Type-II events in this paper, eastward propagation still exists in the equatorial region, although it is much weaker than the Type-I events, as is the northward propagation (Figure 7).

The difference between the two types of events is greater in the East Asia area. In Type-I events (Figure 5), there are strong cyclonic water vapor flux anomalies and negative water vapor flux divergence anomalies over the South China Sea at phase −0, which are replaced by positive vapor flux divergence anomalies until phase 3. In Type-II events (Figure 6), there is no negative moisture flux divergence anomaly over the South China Sea during phase 0, and there is a partial negative water vapor flux divergence anomaly over the South China Sea during phase 2 which dissipates during phase 3. Compared with Type-I events, the anomalies in the South China Sea are much weaker. In Type-I events, there is an anticyclonic water vapor flux anomaly between eastern China and southern Japan in phase 0, which corresponds to the positive water vapor flux divergence and negative precipitation anomaly in southern Japan, and the negative water vapor flux divergence and positive precipitation anomaly in the Bohai Sea and surrounding area. This anticyclonic moisture anomaly is weakened during phase 1 and disappears during phase 2, while the region does not exhibit this change during Type-II events. In the Type-II events, an anomalous southwestward moisture flux appears in eastern Japan at phase 0, and a cyclonic moisture anomaly and a negative water vapor flux divergence anomaly are formed at phase 1, disappearing at phase 3.

In phases 0–1 of Type-I events, negative–positive–negative water vapor flux divergence anomaly distribution is formed in the South China Sea–Philippines, Japan, and the Sea of Okhotsk. However, this kind of distribution does not appear in Type-II events, which indicates that the Type-I events may also be related to EAP (East Asia-Pacific) teleconnection. When convection intensifies over the Northwest Pacific monsoon region, there is a pronounced circular Rossby wave train from WNP (Western North Pacific) to the West Coast of the United States [33,34,35]. EAP teleconnection on the intraseasonal scale can be observed; the water vapor transportation anomaly first appeared near the Philippines, and then anomaly centers successively appeared in the eastern part of China, Japan, and the Sea of Okhotsk, showing a tripole pattern. The three centers are weakened at the same time and disappear after 3–4 days, which is related to the Rossby waves stimulated near the Philippines [36].

To analyze the dominant features of the meridional and zonal propagation of the two types of events, the Hovmöller diagram of the zonal mean of 10° S–10° N and the meridional mean of 50–100°E were plotted (Figure 7). The results show that in Type-I events there is a significant eastward propagation in the equatorial region, and this eastward propagation is rapid, continuous, and almost uniform. In Type-II events, however, the positive anomaly in the central equatorial Indian Ocean almost does not spread to the east during −20 to −10 days, while it spreads rapidly to the east of 150° E during −10 to 0 days. In both types of events, northward propagation occurs in the Indian Ocean (50–100° E), but in Type-II events, the positive anomaly is weakened in the process of propagation. In the Western Pacific, there is a significant and continuous northward propagation of the Type-I events, while no significant northward propagation is observed in the Type-II events.

To explore the reasons for the differences in equatorial propagation between the two types of events, vertical structures of specific humidity, vertical velocity, and zonal velocity were demonstrated (Figure 8 and Figure 9). In equatorial, the eastward BSISOs are mainly manifested as a baroclinic pattern in the vertical direction, and the upper and lower troposphere show opposite features and the zonal winds tilt westward with height [37]. The eastward propagating vertical structure is prominent in Type-I events but markedly weak in Type-II events.

3.3. Moisture Budget Analysis of Two Types of Significant Events

According to Equation 2–6 the moisture budget analysis was carried out for the two types of events. Figure 10 shows the effect of the right three terms of the moisture equation on the moisture tendency in phase 0 of the two types of events. In both types of events, the horizontal moisture advection term is basically consistent with the moisture tendency, indicating that the horizontal moisture advection term plays the most important role in convection propagation. The distribution of the in-column term in both types of events is almost opposite to the moisture tendency, indicating that this term prevents propagation in most of the phases and areas. In Type-I events, spatiotemporal coherence of the evaporation anomaly term and the moisture tendency anomaly is seen in East Asia. The spatiotemporal pattern of the evaporation term anomaly in Type-II events in East Asia does not coincide with the pattern of Type-I events, and does not have apparent spatial consistency with the direction of the anomaly of moisture.

The moisture budget of the two types of events is compared in the Pacific Ocean (120°–160° E). For northward propagation over the Pacific Ocean (Figure 11 and Figure 12) in the Type-I events, the influence of the horizontal advection term is greater south of 20° N, and the influence of the evaporation term is significantly increased north of 20° N and even exceeds the horizontal advection term in some phases. This phenomenon does not appear in the Type-II events. Both north and south of 20°N, the most significant term is horizontal advection.

When comparing the impact of the horizontal advection term, in-column term, and evaporation term on moisture tendency, the horizontal advection term is the most important in most phases and regions.

4. Conclusions and Discussions

In this research, significant events were selected and analyzed according to region A (12–22° N, 66–76° E) time series. Using K-means cluster analysis, 19 Type-I events and 8 Type-II events were further divided according to Q₁ anomalies in the range of 10° S–25° N and 50–150° E (yellow box in Figure 5), and a moisture budget analysis was carried out for the two types of events. The conclusions are as follows:

The reason why the distribution of Q₁ and Q₂ in the East Asia region is quite different might be because Q₁ includes sensible heat flux but excludes the effect of evaporation, whereas Q₂ takes into account the effect of evaporation. The similarity of Q₁ and Q₂ south of 20° N indicates that the heating in this region is mainly supplied by condensation latent heat release, while the difference in East Asia north of 20° N indicates strong sensible heat flux or evaporation. The similarity of Q₁ and precipitation anomalies in East Asia indicates that the difference between Q₁ and Q₂ in East Asia is caused by evaporation.

The eastward propagation of the Type-I events in the equator is continuous and rapid, while the eastward propagation of the Type-II events hardly occurs within −20 to −10 days, but spreads rapidly to the east after −10 days, accompanied by an intensity reduction during propagation. The reason might be that the eastward-propagating vertical structure (baroclinic pattern, westward tilt) is observed in the Type-I events, while it is obviously weaker in the Type-II events. Over the Indian Ocean (50–100° E), both types of events spread northward, but the Type-II events are smaller and weakened halfway through. In East Asia, the northward propagation of the Type-I events remains continuous and could spread north of 30° N from the equatorial region, while the Type-II events have no significant northward propagation. The results of moisture analysis show that the horizontal advection term plays the most important role in the propagation of convection. In East Asia, the analysis of the moisture budget indicates that evaporation also plays a crucial role in the northward propagation of Type-I events. Conversely, the evaporation term plays little effect on Type-II events in East Asia, which may be one of the reasons for the weak northward propagation.

In the Type-I events, there is a wave train in the middle latitude, but not in the Type-II events; the Type-I events also have anomalies on the land around the Arabian Sea, while the anomalies of the Type-II events exist almost exclusively over the ocean. Ding and Wang pointed out that, in the mid-latitude wave train, the height over Central Asia can enhance convection over northern India, Pakistan, and the Arabian Sea [38]. Therefore, the difference between the two types of events could be decided by whether the Q₁ anomaly in the northeast Arabian Sea in the Type-I events is influenced by the joint action of the equator and mid-latitude while that in the Type-II events is not. This should be further studied and discussed.

Konda et al. proposed that the Indian Ocean dipole (IOD) can modulate BSISO. BSISO shows an incoherent spatial pattern in positive IOD (pIOD) years, while during negative IOD (nIOD) years the development of BSISO is coherent and more evident [39]. In nIOD years, the strong SST meridional gradient contributes to the northward propagation, which is related to zonal wind anomalies. Easterly wind anomalies inhibit evaporation and increase SST, while westerly wind anomalies lead to cold SST anomalies. After comparison, 7 of the 19 Type-I events are in the nIOD years and 2 are in the pIOD years, while only 1 Type-II event is in the nIOD years. Whether the two different types of events are affected by the IOD should be further discussed. The appearance of Type-I events in nIOD years (7 of 19) in comparison to Type-II events (1 of 8) hints at potential phase modulation by IOD, but due to a small dataset there is no specific statistical significance (p = 0.236 from Fisher’s exact test). Larger dataset studies in the future could render such a correlation more apparent.

This research classifies Q₁ anomaly events in the Arabian Sea, but the reasons for the difference between the different types of events are not discussed. This issue may be further studied in the future. The significant events in this article mainly occur in July and August, which may be affected by the selection standard of significant events and could be further analyzed by changing the selection methods of significant events. In this paper, 19 Type-I events and only 8 Type-II events were identified. The small number of Type-II instances contributes further to the uncertainty of the composite results and the moisture budget analyses of this type. Characteristics defined as representative of Type-II occurrences are to be interpreted carefully and may be less important than those of Type-I (n = 19). A small sample size will have an impact on significance testing. Long-period BSISO events exhibit significant asymmetry, which has been neglected by most previous studies due to the EOF or band-pass filtering methods [40]. Its convection activity has slow progression and rapid decay in Eastern IO [41], and is opposite in WNP. This asymmetry of convection anomaly is caused by SST anomaly amplitude asymmetry [40]. EOF and bandpass filtering were used in this paper, so the results might be affected. In addition, numerical simulation sensitivity experiments can further study the influence processes of baroclinic structures and evaporation terms on the two types of events. These could be further improved in the future.