20–60-day intraseasonal variation of summer rainfall in Thailand and its associated large-scale atmospheric moisture circulation

Variations of rainfall at sub-seasonal time scale can cause both wet and dry conditions that are commonly associated with anomalies of large-scale atmospheric moisture circulation. This study examines the characteristics of the sub-seasonal variations of summer rainfall in Thailand and their association with large-scale atmospheric moisture circulation. We analyzed long-term daily rainfall data from the Thai Meteorological Department (TMD) spanning from 1979 to 2019, along with data on atmospheric water vapor flux and transport obtained from the ERA-5 reanalysis. Here, we used a 20–60-day bandpass filter for both the TMD and ERA-5 datasets. The results showed that summer rainfall in Thailand is characterized by high intensity and variability of 20–60-day rainfall, particularly in the northeastern, eastern, and western coastal areas of upper Thailand, as well as in the western coastal area of lower Thailand. The characteristics of the 20–60-day atmospheric water vapor anomalies associated with wet and dry rainfall events varied both spatially and temporally, depending on the sub-region and month. Indeed, the wet conditions in upper Thailand during the summer season are strongly associated with southwesterlies (easterlies) of moisture convergence from the eastern Indian Ocean (South China Sea), which induced a total moisture budget of about 24 × 108 kg/s. In contrast, dry conditions are controlled by westerly anomalies of moisture divergence, with a moisture budget of −21.4 × 108 kg/s. For lower Thailand, wet conditions of the 20–60-day rainfall variation are associated with southwesterly anomalies and a cyclonic-like pattern of moisture convergence over the Andaman Sea, while an anti-cyclonic atmospheric moisture divergence is dominant, leading to below-normal rainfall (with a water budget of 2.7 × 108 kg/s and −2.4 × 108 kg/s for wet and dry events, respectively). The results presented here will benefit our understanding of regional intraseasonal rainfall variation, improving sub-seasonal to seasonal forecasting in Thailand which will help improve our water resource management and disaster risk reduction.


Introduction
The sub-seasonal variation of dominant atmospheric processes, including the Madden-Julian Oscillation (MJO) (Madden and Julian 1972) and the Boreal Summer Intraseasonal Oscillation (BSISO) (Yasunari 1979), modulates the complex atmospheric circulations that govern sub-seasonal rainfall variation over the Asian summer monsoon region (Annamalai and Sperber 2005;Hsu et al. 2004;Yasunari 1980).The associated zonal convection pattern propagates from the Indian Ocean to the Pacific Ocean, inducing landocean-atmosphere interaction in the tropical region (Li et al. 2013).Therefore, the sub-seasonal variability of rainfall and extreme weather events over Southeast Asia (SEA) is largely characterized by air-sea interactions originating from the tropical ocean (Lim et al. 2017;Xavier et al. 2014;Yang et al. 2021;Yokoi et al. 2007).Previous studies have indicated that atmospheric circulation originating from remote oceans is associated with variations in thermal heating of the local land surface temperature over Southeast Asia (SEA).These variations are regulated by soil moisture anomalies at sub-seasonal to seasonal (S2S) timescales (Amenu et al. 2005;Gao et al. 2020;Liu et al. 2009;Meng and Zheng 2006;You et al. 2018;Zhu et al. 2021).Moreover, it has been observed that soil moisture anomalies during the premonsoon months contribute to an enhancement of regional summer rainfall (Gao et al. 2020;You et al. 2018;Zhu et al. 2021).Changes in surface energy flux and evaporation processes modulate the land surface temperature, which subsequently influences atmospheric heating and air flows (Amenu et al. 2005;Meng and Zheng 2006).Additionally, the difference in thermal conditions between the South China Sea (SCS) and SEA sub-regions is associated with the onset of the summer monsoon (Liu et al. 2009).Thus, local land and ocean processes interact remotely and have the potential to modulate rainfall patterns in the region.Notably, the local topography, characterized by a diverse landscape comprising mountain ranges, plains, plateaus, and coastal areas, also plays a significant role in shaping local weather and climate at sub-seasonal timescales (Yokoi et al. 2007).
The sub-seasonal variability of summer rainfall in Thailand is a crucial aspect of the rainy season (Kikuchi et al. 2012).Notably, Kripalani et al. (1995) demonstrated the westward propagation of quasi-biweekly oscillation, which corresponds to 10-20-day variations in summer (June-September) rainfall across Thailand.Consequently, summer rainfall in the region exhibits a quasi-bimodal pattern, with distinct peaks occurring in June-July and August-September (Takahashi and Yasunari 2006).It is important to note that the first peak is associated with a 30-60-day southwest monsoon flow, while the second peak is linked to the 10-20-day westward propagation of tropical cyclones from the South China Sea (SCS) (Takahashi and Yasunari 2006;Yokoi and Satomura 2005;Yokoi et al. 2007).Furthermore, Takahashi and Yasunari (2006) revealed that the north-south oriented structure of strong monsoon westerlies induces a quasi-stationary ridge over the Indochina Mountain range and Dawna range in Myanmar, leading to a monsoon break during the mid-rainy season.Conversely, it triggers the propagation of stationary Rossby waves with strong westward moisture convergence across the SCS towards the Indochina Peninsula during the late-rainy season.Large-scale circulations also influence sub-seasonal rainfall variations in other regions.For instance, the westward propagation of anticyclone-cyclone waves over the western tropical Pacific plays a crucial role in the 10-20-day onset of rainfall over the Indochina Peninsula.Additionally, a trough-ridge seesaw pattern and subtropical ridge over the Indochina Peninsula can induce a 30-60-day summer monsoon onset, impacting rainfall patterns in the region (Mao and Chan 2005;Nguyen-Le et al. 2015;Zhang et al. 2002).Moreover, subseasonal rainfall anomalies in the Indochina region may contribute to the intraseasonal oscillation (7 to 25 days) of deep convective activities over the windward region, which is an elongated narrow mountain range located from east to west and adjacent to the South Himalaya (Fujinami et al. 2011;Fujinami et al. 2014).Moisture convergence induced by cyclonic circulation in the SCS is another source of subseasonal summer rainfall anomalies over Southeast Asia, particularly in northern Vietnam and its vicinity (Truong and Tuan 2019).In Southeast Asia, the northeastward propagation of moisture convection characterized by the boreal summer intraseasonal oscillation (BSISO) can influence rainfall patterns in certain areas (Chou and Hsueh 2010;DeMott et al. 2013;Kikuchi et al. 2012).The BSISO convection is controlled by the zonal advection of column-integrated moist static energy budgets over the Bay of Bengal (BoB) and Western North Pacific Ocean (Wang and Li 2020), subsequently enhancing rainfall over the Indochina Peninsula (Chou and Hsueh 2010;Shi et al. 2005;Wong et al. 2011).
In addition, key precursors related to temperature trends, sea surface temperature gradients, and sea level pressure patterns could have implications for the intraseasonal variation of summer rainfall in Thailand.This is because the onset of the South China Sea summer monsoon is closely linked to the initiation of the summer monsoon season, which in turn influences rainfall patterns in Thailand (Zhu and Li 2017b).Wu et al. (2023) also emphasized the importance of the SCS monsoon onset and the leading mode of the BSISO in influencing moisture processes, a factor that might be crucial in determining the intraseasonal variation of summer rainfall in Thailand.Moreover, the temporal and spatial characteristics of pre-summer extreme precipitation days in southern China could play a role in the intraseasonal variation of summer rainfall in Thailand, given their shared monsoon characteristics (Li et al. 2023).Recently, Zhu et al. (2023) explored the zonal displacements of the South Asian high (SAH) and the western Pacific subtropical high (WPH) associated with Meiyu.These displacements could modulate the atmospheric circulation patterns over Thailand during the summer monsoon season and contribute to intraseasonal variations in rainfall to a certain extent.Previous studies have highlighted that variations in intraseasonal climate conditions within a region are influenced by different factors or sources, all of which contribute to predictability.Therefore, understanding and accounting for these sources of predictability are important in subseasonal to seasonal (S2S) climate prediction (Zhu and Li 2017a;Zhu et al. 2017;Zhu and Li 2018).
While previous studies have contributed to enhancing our understanding of the characteristics and mechanisms of subseasonal variations in summer rainfall over the Southeast Asia (SEA) sub-region, there remains a lack of clear comprehension regarding intraseasonal rainfall anomalies specifically in Thailand, along with a focus on their associated large-scale atmospheric moisture circulation patterns.Thus, the primary objective of this study is to provide a characterization of the features of intraseasonal rainfall variations in Thailand, utilizing long-term daily rainfall data and analyzing the corresponding large-scale atmospheric circulations.
The structure of this paper is organized as follows:"Data and methods" provides a description of the data and methods employed in the study.In "Result and Discussion", we present the features of the 20-60-day summer rainfall variability and examine the behaviors of water vapor fluxes, transportation, and budgets associated with rainfall anomalies in different sub-regions of Thailand.Finally, "Conclusions" concludes the paper, summarizing the key findings and implications of the study.

Observation data
This study utilizes observed daily rainfall data acquired from surface monitoring stations operated by the Thai Meteorological Department (TMD) for a period spanning 41 years, from 1979 to 2019.This dataset has been widely employed in previous research studies (Khedari et al. 2002;Torsri et al. 2013;Torsri et al. 2022a;Torsri et al. 2022b).It comprises a total of 85 observation stations, which are distributed across two distinct climatic zones (Khedari et al. 2002;Torsri et al. 2013).Specifically, 67 stations are located in the North, Northeast, and Central-East regions, collectively referred to as upper Thailand (UT), while the remaining 18 stations are situated in lower Thailand (LT) or the South of Thailand (Fig. 1 provides a visual representation of the station distribution).
Furthermore, we incorporate 6-hourly reanalysis data from the European Center for Medium-Range Weather Forecasts (ECMWF) to augment our analysis.This dataset encompasses variables such as total column water vapor (TCW), the vertical integration of north-eastward water vapor flux, and moisture flux convergence.These data are derived from the ECMWF's ERA5 reanalysis, which has a horizontal resolution of 0.25° × 0.25° (Hersbach et al. 2020).
To examine the characteristics and potential influence of large-scale atmospheric circulations on intraseasonal rainfall variation in Thailand, the 6-hourly datasets are aggregated into daily averages.

Analysis methods
To facilitate statistical analysis, the summer season in this study is defined as the period from June to August (JJA), resulting in a total of 3772 days spanning the 41-year period.To isolate the sub-seasonal variation from the daily time series, we employ the bandpass Lanczos filter method (Duchon 1979) on both the daily summer rainfall and atmospheric circulation data.Previous studies by Yokoi and Satomura (2005) and Truong and Tuan (2019) have indicated that the intraseasonal variation of the rainy season in the Indochina Peninsula sub-regions is primarily influenced by atmospheric circulation with time scales ranging from 10 to 60 days.The 20-60-day period has been identified as the dominant mode of intraseasonal variation in this context.Therefore, we apply the bandpass filter method to derive a daily time series specifically capturing the 20-60-day period.The bandpass filter method is a technique used to isolate specific frequency components from a time series dataset.In our study, we employed a bandpass filter with a range of 20-60 days to eliminate interannual variations from the daily rainfall data.The choice of the number of weights in the filter determines the smoothness of the filtered time series, with a higher number of weights resulting in a smoother series.In this particular analysis, we utilized 121 weights to construct the filtered time series.The selection of the number of weights is based on the length of the filter, and for this study, we determined that a filter length of 121 provided an optimal balance between smoothness and preservation of relevant information.This choice ensures that the resulting series is adequately smooth without excessive loss of information or introduction of excessive noise.
In addition, we conducted a spectral analysis on the rainfall anomalies to quantify the spectral variance of summer rainfall in the upper Thailand (UT) and lower Thailand (LT) sub-regions separately.The analysis revealed that two frequency bands, namely 10-20 days and 20-60 days, play a significant role in the variation of rainfall in both It is important to note that, to capture regional statistics, the daily time series was initially calculated as area averages aggregated from all stations in the upper (UT) and lower (LT) Thailand.This aggregated time series was then subjected to the bandpass filter, specifically targeting the 20-60-day frequency range, for subsequent analysis.In order to identify the characteristics of wet and dry episodes in the summer rainfall events of Thailand, we employed a selection criterion based on the normalized 20-60-day filtered daily time series.Wet (dry) events were identified as those exceeding +1 (falling below −1) standard deviation from the mean for a given day of the month.
To quantify the moisture transport into the target region during strong and weak events, we employed the approach outlined by Schmitz and Mullen (1996) to compute the water vapor content at each boundary of the upper and lower regions.Additionally, we considered the moisture budget within each sub-region by summing the moisture transport across each boundary, following the methodology proposed by Li et al. (2012).To illustrate the atmospheric anomalous patterns associated with positive (wet) and negative (dry) summer rainfall events, we conducted a composite analysis.Statistical significance was determined using a two-tailed Student's t-test.The effective sample sizes for each event were estimated using the lag-1 autocorrelation coefficient, as described in Wilks (2011).

Dry and wet episode selection
Firstly, we examined the sub-seasonal variation of summer rainfall in Thailand and identified the number of wet and dry days within each summer month for both the upper and lower regions, using the ±1 standard deviation (std) threshold of the filtered rainfall anomalies (see Table 1).In the upper region, the frequency of wet summer rainfall events (i.e., rainfall anomalies > +1 std) varied from 166 to 197 days over the 41-year period, with the highest frequency observed in July.Conversely, the occurrence of dry events (i.e., rainfall anomalies < −1 std) ranged from 163 to 177 days, peaking in August.
Similarly, in the lower region (South), the number of above-normal (wet) and below-normal (dry) rainfall events varied on a month-by-month basis.The frequency of abovenormal events ranged from 155 to 188 days, with the lowest occurrence in June and the highest in August.Conversely, the frequency of below-normal events ranged from 145 to 194 days, also reaching its peak in August.
It is noteworthy that for composite analysis of atmospheric features within each sub-region, we used the average number of wet or dry events in a given month or season as representative of the respective episode.2b displays the variance of the 20-60-day summer rainfall in the upper region of Thailand.Overall, the variance of the 20-60-day rainfall in upper Thailand is pre-dominantly below 10.0 mm 2 /day 2 .However, notable areas with higher variance, exceeding 20.0 mm 2 /day 2 , are observed in the southeastern, eastern, and western parts of the upper Thailand sub-region.The lower Thailand (LT) region exhibits generally high summer rainfall intensity, surpassing 10.0 mm/day, particularly in its western part adjacent to the Andaman Sea (Fig. 2c).Moreover, this western region also experiences higher 20-60-day rainfall variation, exceeding 30.0 mm 2 / day 2 (Fig. 2d).These findings suggest that the contribution of low-frequency (20-60 days) rainfall variation in the UT sub-region is more significant in its southeastern and eastern parts, while in the LT sub-region, it is more pronounced in the western part.Notably, it should be emphasized that the 20-60-day variation of summer rainfall in Thailand exhibits greater prominence in coastal areas compared to inland areas, indicating the influence of land-sea thermal heating on atmospheric circulation patterns (Yokoi et al. 2007;Zhang et al. 2002).In the inland region, similar sub-seasonal rainfall patterns have been observed in northeastern Thailand, albeit on a shorter time scale of 10-20 days, which is associated with the westward propagation of cyclonic circulation over the South China Sea (SCS) (Takahashi and Yasunari 2006;Yokoi and Satomura 2005;Yokoi et al. 2007).Additionally, Fig. 2b demonstrates that 30-60-day rainfall variations dominate the coastal region of Thailand, which is related to the intensification of the monsoon trough and the onset of the southwesterly monsoon (Yokoi et al. 2007).

Characteristics of the 20-60-days water vapor transport and moisture budget associated with wet and dry events of summer rainfall anomalies over Thailand
In this section, we investigate various aspects of water vapor transport during the summer season in Thailand.Specifically, we examine the characteristics of 20-60 day atmospheric circulations, including total column water vapor (TCW), water vapor flux and its corresponding convergence anomalies (WF & Con), water vapor transport (WVT), and moisture budget.These characteristics are analyzed during two distinct extreme rainfall episodes, namely wet and dry Figure 3 presents the anomalous atmospheric circulations associated with wet and dry events in the UT region.During the wet episodes, a pronounced enhancement of TCW is observed, particularly in the southern and eastern parts of UT, extending to the western part (Fig. 3a).The TCW anomalies in these areas reach values of 3.0-4.0kg/ m 2 .This increase in TCW is attributed to the influence of strong southwesterly winds, which drive the transport of water vapor from the Andaman Sea into the region (Fig. 3c).Additionally, the northern part of UT exhibits significantly enhanced TCW, which may be associated with the influence of strong easterly winds that bring water vapor anomalies from the SCS to the region.It is hypothesized that these strong easterly winds may be triggered by convection occurring over the BoB (Liu et al. 2002).Therefore, the wet events in the UT region appear to be linked to the aforementioned anomalous atmospheric circulations.Furthermore, an a-like cyclonic circulation pattern is observed over the BoB in close proximity to the western coastal area of UT.This circulation pattern may contribute to additional water vapor flux convergence and the occurrence of heavy rainfall events in that particular area (Fig. 3c).It is important to note that previous studies, such as Ma et al. ( 2019 3a-d governing moisture convergence towards East Asia, thereby shaping the quantity and distribution of rainfall across Thailand.Conversely, during dry episodes, a significant moisture divergence is observed over the UT region.The prevailing anti-cyclonic circulation acts to transport moisture away from the region, leading to drier conditions (Fig. 3d).
Furthermore, we conducted a computation of the moisture budget associated with wet episodes in the UT region (Fig. 3e-f).The results reveal that water vapor is primarily transported into the region through its eastern boundary (25.2 × 10 8 kg/s) and southern boundary (18.8 × 10 8 kg/s).On the other hand, the outflow of moisture at the western and northern boundaries is significantly lower compared to the inflows, resulting in a total moisture budget of 23.8 × 10 8 kg/s (Fig. 3e).This observation aligns with the understanding that an intensified southwesterly monsoon flow over the Andaman Sea promotes moisture transport from the southern boundary to the UT region.Conversely, a weakened southwesterly monsoon over the SCS is frequently associated with the strengthening of the western North Pacific High (WNPH) (Zheng et al. 2013), which, in turn, impedes moisture transport through the eastern boundary of UT.These findings are in accordance with previous studies that have emphasized the significant role of large-scale circulations originating from these specific areas (Fukutomi and Yasunari 2002;Meehl et al. 2001;Takahashi and Yasunari 2006;Yokoi et al. 2007).In the case of the dry episode, a noticeable outflow of moisture occurs over the UT region, primarily from the eastern boundary at approximately 6.3 × 10 8 kg/s and the southern boundary at 18.2 × 10 8 kg/s.Consequently, this leads to a deficit in the total moisture budget, amounting to −21.4 × 10 8 kg/s (Fig. 3f).
Next, we examine the moisture divergence and budget specifically for the LT sub-region (Fig. 4).It is found that the TCW is significantly enhanced over the Andaman Sea due to the influence of a strong southwesterly monsoon and cyclonic circulation anomalies (Fig. 4a).This enhancement in TCW leads to moisture convergence primarily over the western part of the LT region (Fig. 4c).Conversely, the presence of anti-cyclonic circulation results in a deficit in rainfall and subsequently dry conditions within the LT region (Fig. 4d).The total moisture budget during wet and dry events in the LT region is estimated to be 2.7 × 10 7 kg/s and −2.4 × 10 7 kg/s, respectively.It is noteworthy that the total moisture budget in the LT sub-region is an order of magnitude smaller than that observed in the UT region.This distinction is important and may prove valuable in further investigations aimed at determining the sub-seasonal prediction skill of rainfall between these two sub-regions in Thailand.
This finding suggests a relative symmetry in sub-seasonal water vapor transport during both strong and weak events.Besides, Takahashi and Yasunari (2006) demonstrated that the rainy season in the Indochina Peninsula is closely linked to the development of strong monsoon westerlies, typically occurring from late May to early June.This highlights the significance of sub-seasonal variations in low-level large-scale circulations, which play a crucial role in the complex moisture convergence patterns that vary across different regions (Kanae et al. 2001;Zhang et al. 2002).Moreover, these sub-seasonal disturbances exert control over the sources of seasonal moisture and summer rainfall (Fujinami et al. 2011).Through detailed analysis, a notable lead-lag relationship has been identified between rainfall anomalies in the UT and LT sub-regions.Specifically, when considering the influence of the northward propagation of the boreal-summer intraseasonal oscillation on intraseasonal variability, it has been observed that a wet event occurring in the LT sub-region tends to precede a similar event in the UT region by a few days (see Fig. S2).

Characteristics of the 20-60-days water vapor
transport and moisture budget associated with wet and dry events in upper Thailand (UT) for each summer month Furthermore, to gain a comprehensive understanding of the sub-seasonal variation in the UT sub-region, we conducted a monthly analysis spanning from June to August.Our focus is to identify the dominant patterns that are significantly associated with wet and dry rainfall anomalies during this period.Specifically, we discuss the major patterns that contribute to the variability in summer rainfall over the sub-region.However, for clarity, we will include additional information regarding other patterns and aspects related to atmospheric water vapor, such as total column water vapor, water vapor flux and convergence, and water budget, in the supplementary materials.
While our analysis covers all 3 months, our findings highlight that the patterns and their corresponding atmospheric water vapor aspects in July and August (Figs.5 and 6) exhibit the most pronounced and noteworthy characteristics compared to those observed in June (Fig. S3).
In July, during the wet condition, a depreciation in the 20-60-day total column water (TCW) is observed over the UT sub-region compared to the previous month, with a magnitude ranging from 1.0 to 2.0 kg/m 2 (Fig. 5a).Additionally, the anomalous southwesterlies were more dominant over the Andaman Sea and the Gulf of Thailand, while the anomalous easterlies were suppressed (Fig. 5c).As a result, moisture was primarily transported into the UT sub-region through its western and southern boundaries, with a magnitude of 11.3 × 10 8 kg/s and 21.9 × 10 8 kg/s, respectively, while only a small portion originated from the eastern boundary (2.1 × 10 8 kg/s) (Fig. 5e).Conversely, there was relatively small moisture outflow (8.8 × 10 8 kg/s) at the northern boundary, resulting in a high positive net water budget (26.5 × 10 8 kg/s) within the sub-region (Fig. 5e).
During the wet condition in July, we observed that the magnitude of the 20-60-day TCW over the UT sub-region remained comparable to the previous month, ranging from 1.0 to 2.0 kg/ m 2 (Fig. 5a).However, there was a noticeable intensification of the anomalous southwesterlies over the Andaman Sea and the Gulf of Thailand, accompanied by a suppression of the anomalous easterlies (Fig. 5c).Subsequently, the moisture transport into the UT sub-region was predominantly through its western and southern boundaries, with magnitudes of 11.3 × 10 8 kg/s and 21.9 × 10 8 kg/s, respectively, while the contribution from the eastern boundary was relatively modest (2.1 × 10 8 kg/s) (Fig. 5e).Conversely, the outflow of moisture at the northern boundary was relatively minor (8.8 × 10 8 kg/s), leading to a significantly positive net water budget (26.5 × 10 8 kg/s) within the sub-region (Fig. 5e).
In the dry episode, a similar spatial pattern of TCW was observed in this month compared to the previous month (Fig. S3).The deficit in rainfall during this period was associated with pronounced anomalous moisture divergence over the sub-region (Fig. 5d), resulting in substantial outflow at its boundaries, particularly at the southern boundary (22.4 × 10 8 kg/s).As a result, a net negative moisture budget (−25.4 × 10 8 kg/s) was observed.
When considering the wet episode that occurred in August, it is observed that the 20-60-day TCW becomes weaker compared to June and July (Fig. 6a).The high rainfall events in this month are primarily influenced by moisture convergence from the anomalous easterlies, transporting moisture from the northern SCS across the uppermost part of the UT sub-region (Fig. 6c).Additionally, there is noticeable additional moisture movement from the BoB through anomalous southwesterly convergence.Consequently, the dominant moisture inflow is from the east, with the highest amount recorded (50.7 × 10 8 kg/s), along with surplus moisture inflow from the southern boundary of the UT sub-region (14.0 × 10 8 kg/s) (Fig. 6e).It is important to note that the total moisture inflow in August is lower than the amount recorded in June (Fig. S3), which may be attributed to the weakening of the southerlies in August and the high amount of moisture outflow at the western and northern boundaries of the UT sub-region, with magnitudes of 29.0 × 10 8 kg/s and 10.4 × 10 8 kg/s, respectively.These factors contribute to a net positive moisture budget remaining in the sub-region (25.3 × 10 8 kg/s).
In August, the rainfall deficit over the UT sub-region is also caused by the divergence of the 20-60-day TCW (negative value) (Fig. 6b).The spatial pattern of TCW in this month is similar to that of the previous months but with smaller magnitudes.Unlike the wet episode, moist air over the UT subregion is driven out by strong anomalous westerlies, accompanied by simultaneous northeastward moisture convergence (Fig. 6d).Consequently, there is an excess of moisture outflow at the eastern (45.1 × 10 8 kg/s) and southern (18.3 × 10 8 kg/s) boundaries, leading to a negative moisture budget (−21.2 × 10 8 kg/s) remaining in the UT sub-region (Fig. 6f).

Characteristics of the 20-60-day water vapor
transport and moisture budget associated with wet and dry events in lower Thailand (LT) for each summer month Additionally, the dominant anomalous atmospheric circulation observed over the lower region during June to August is examined (Fig. 7 and Figs.S4 and S5).Results show that among these months, August displays a particularly pronounced and noteworthy pattern compared to June and July.Consequently, our main discussion will center around the findings specific to August (Fig. 7), while the results pertaining to June and July are presented in the supplementary materials (Figs.S4 and S5). Figure 7a demonstrates that in August, the occurrence of a wet condition in the LT sub-region is associated with an extensive 20-60-day TCW, with magnitudes ranging from 2.0 to 5.0 kg/m 2 .However, the spatial pattern of the TCW in August differs significantly from that of the previous months, as the highest intensity of the TCW is displaced and extended further southward.Unlike the previous months, two dominant atmospheric circulations emerge due to the convergence of strong easterly moisture transport over the upper part of Thailand and the convergence of strong equatorial westerly moisture transport from the southern part of the BoB directly towards the tip of the LT sub-region (Fig. 7c).Consequently, a positive net moisture budget (12.7 × 10 7 kg/s) is induced by excessive moisture inflow through the western boundary (79.3 × 10 7 kg/s) and a smaller portion from the southern boundary (Fig. 7e).However, the moisture balance within the sub-region is diminished by a high moisture outflow (approximately 67.0 × 10 7 kg/s) at its eastern boundary.
During the dry episode of the LT sub-region, the TCW is diminished due to strong anomalous moisture divergence (Fig. 7b and d), resulting in abundant moisture flowing out Fig. 6 Same as Fig. 3, but for August from the western boundary with a magnitude of 70.8 × 10 7 kg/s (Fig. 7f).Although a significant amount of compensated moisture is transported into the sub-region at its eastern boundary (about 50.0 × 10 7 kg/s), it is relatively smaller than the moisture outflow, leading to a negative moisture budget (−17.0 × 10 7 kg/s).This indicates that the rainfall deficit in the sub-region is related to moisture outflow from the western and the southern boundaries.

Conclusions
The characteristics of the 20-60-day summer rainfall variation and its associated large-scale moisture circulation have been identified in this study.We analyzed data from rain-gauge stations and atmospheric moisture ERA-5 variables spanning a period of 41 years, from 1979 to 2019.Specifically, we examined the characteristics of the 20-60-day intraseasonal rainfall variation and its corresponding largescale moisture patterns during both wet conditions (above normal rainfall) and dry conditions (below normal rainfall) in the upper (UT) and lower (LT) Thailand sub-regions.
The composite analyses conducted in this study reveal key findings about the 20-60-day anomalies of summer rainfall in the UT sub-region during wet and dry conditions by which, during wet conditions, the above-normal rainfall in UT subregion is primarily induced by southwesterly moisture convergence.This convergence brings warm, moist air from the Indian Ocean and Gulf of Thailand towards the UT sub-region through its southern boundary.Additionally, easterly moisture Fig. 7 Same as Fig. 4, but for August convergence, located between 20° N and 30° N latitudes, plays a dominant role in causing above-normal rainfall during the season.Moisture transport occurs primarily through the southern and eastern boundaries of the sub-region.In July, the southwesterly moisture convergence brings significant moisture inflow from the Indian Ocean, particularly via the southern boundary, while additional moisture is also transported through the western boundary.In August, the dominant moisture inflow is observed at the eastern boundary, with easterly moisture convergence playing a major role in transporting warm, moist air from the South China Sea into the upper subregion.In contrast, during dry conditions, a reversal in largescale moisture circulation is observed in the upper sub-region.Divergence of westerly anomalies along the Tropic of Cancer, particularly northwesterly divergence anomalies, becomes dominant.As a result, moisture is excessively swept out from the sub-region through its eastern and southern boundaries.Additionally, due to the geographical characteristics of the uppermost and western parts of the UT sub-region, which include relatively high elevations and mountainous terrain, as air moves over land, specifically over northern South Asia and across the mountains, it loses moisture, creating dry air that is then transported into the UT sub-region.Consequently, abundant moisture outflow is typically observed at the southern and eastern boundaries, particularly in July and August.
In lower Thailand, above-normal 20-60-day rainfall variations during the summer season are typically attributed to excessive atmospheric moisture resulting from cyclonic moisture convergence.This convergence occurs as a result of southwesterly winds moving along the Equatorial Indian Ocean towards the western coast and southern part of the sub-region.This phenomenon is particularly evident in August, highlighting the importance of the southwest monsoon.In June, the moisture convergence pattern over the LT sub-region differs significantly from the following months by which easterly moisture inflow is more dominant compared to the subsequent months.During dry conditions in the summer season, an anti-cyclonic pattern of moisture divergence is observed over the LT sub-region.This pattern leads to an excessive outflow of moisture at the sub-region's western and southern boundaries, resulting in below-normal rainfall.Notably, the anti-cyclonic pattern is particularly strong and dominant in August.It is worth mentioning that the seasonal differences in intraseasonal variability are not as significant for the LT sub-region.
Overall, these findings enhance our understanding of the large-scale moisture circulation patterns and their impact on the 20-60-day summer rainfall variations in Thailand's sub-regions under both wet and dry conditions.We emphasize the significance of cyclonic and anti-cyclonic moisture patterns in driving these rainfall variations.However, it is important to acknowledge certain limitations and potential areas for future research.For instance, conducting a finer spatial resolution analysis at local scales, such as river basins or provinces, would provide more detailed insights into the variations and circulation patterns within these sub-regions.Additionally, exploring the underlying physical processes that drive the large-scale atmospheric patterns associated with intraseasonal rainfall variations would contribute to a deeper understanding of the mechanisms behind these observed variations and improve sub-seasonal to seasonal (S2S) climate prediction in Thailand.

Fig. 1
Fig. 1 Rainfall station locations (red dots) provided by the Thai Meteorological Department (TMD) in Thailand.The shading indicates the terrain elevation in meters.Two main sub-regions of Thailand, namely upper (UT) and lower (LT) Thailand, are defined

Fig. 2
Fig. 2 Spatial distribution of a climatology and b its variance of the 20-60 day bandpass filtered summer rainfall in units of mm 2 /day 2 for the UT subregion.c-d It is the same, but for LT sub-region Fig. 3 a-b The 20-60 day patterns of total column water vapor (TCW), c-d water vapor flux, convergence anomalies, and e-f water vapor transport anomalies at each boundary of the UT region during wet (left panels) and dry (right panels) summer rainfall events.Moisture balance is shown in the rectangular and units are indicated.Only values exceeding the 95% significant level are shown in shading areas and vectors in Fig.3a-d

Table 1
Frequency of occurrence of wet (dry) rainfall events occurred in a summer month over the period 1979-2019 based on above +1 (below −1) standard deviation of the rainfall anomalies for upper and lower Thailand