Long-Term Rainfall Variability and Trends for Climate Risk Management in the Summer Monsoon Region of Southeast Asia

,


Introduction
Te Earth's climate has experienced remarkable transformations in recent decades, primarily attributed to global warming [1].Tese changes have led to a substantial escalation in the destructive impacts of foods, droughts, and other natural calamities across numerous nations [2].Consequently, climate change poses a severe threat to agricultural productivity and poses a signifcant challenge to food security at both local and global scales [3].
Rainfall plays a pivotal role as a fundamental environmental factor when examining the meteorological characteristics of a specifc area, enabling the monitoring of climate and weather pattern shifts [4].It serves as a crucial indicator for studying the impacts of climate change, as the spatial distribution of rainfall directly infuences global temperature fuctuations.Variations in rainfall intensity and magnitude are key indicators of changes in food occurrences, drought events, and exceptionally dry or wet seasons, which can result in signifcant harm to ecosystems, plant and animal life, property, and food security through declines in agricultural output [5][6][7].Moreover, investigating macrolevel variables and trends in rainfall is essential for researchers examining extensive geographic areas such as river basins, countries, or regions, enabling the exploration of sustainable solutions for food control and drought mitigation [8][9][10][11][12].
In addition, rainfall trend studies provide a solid foundation for micro-and macro-level resource management assessments based on the geographical context used for the analysis [13].Furthermore, studies related to rainfall trends have been conducted through several statistical models to provide a proper concept of agricultural activity, foods, droughts, and water systems.However, a signifcant number of studies using the spatial variability of rainfall are lacking [14][15][16][17][18][19].
However, studies have shown that worldwide rainfall has increased by 2% over the past ten decades [20], while other studies [21] have emphasized that the study of seasonal and annual precipitation, based on conventional climate center data, is a high priority.However, rainfall variations in river basins and climatic zones have not received signifcant attention in previously reported studies [22,23].It is expected that a signifcant increase in rainfall in Asia by 2099 will mainly be due to the summer monsoon [24].On the other hand, a substantial portion of the world's population is in the Asian monsoon region.Henceforth, drastic changes in rainfall may afect the socioeconomic activities in the region [24,25].On the other hand, the island kingdoms of Malaysia, Singapore, Indonesia, and Papua New Guinea, however, receive boreal winter monsoon rainfall, which extends from November to March [26][27][28][29][30].Moreover, Southeast Asian (SEA) countries such as Myanmar, Vietnam, Tailand, Lao People's Democratic Republic (Lao PDR), and Cambodia are largely afected by the South Asian summer monsoon system, bringing signifcant rainfall to those regions from May to September (list of abbreviations and acronyms are stated in Table 1) [31].
However, global climate projections show that foods will increase in many parts of the world over the next century due to increasing trends in precipitation [32].In order to manage climate risk in an area, it is most important to study historical weather conditions and how they have changed in the present.Also, the study of rainfall variability and trends at a particular place under consideration is very helpful in climate risk management, as regional climate behavior varies more or less compared to global climate change [33].On the other hand, precipitation variability and trend analyses are very important in formulating and implementing action plans to reduce exposure to extreme climatic conditions [34].
However, a few studies have been conducted to study the rainfall patterns and trends covering the summer monsoon region of SEA [35,36].Tese studies have shown changes in the SEA's monsoon system due to changes in temperature and monsoon rainfall in the late 21st century [37,38].Various studies have also found signifcant variations in summer monsoon rainfall in Southeast Asia [39,40].However, the SEA region has reported no proper climate change forecasting due to the lack of information on rainfall trends.
Henceforth, the objective of this study is to explore the rainfall variability and trends of summer monsoon countries of SEA using long-term rainfall data.Te study also focuses on the trends in rainfall over the year and monsoon, covering countries and major river basins in the study area.A long-term (1981-2021) grid-based (5 km × 5 km) precipitation dataset [41][42][43][44] has been used for rainfall variability and trend analysis in other parts of the world.We strongly believe that this study's fndings will help policymakers decide on sectors such as agriculture, urban planning, and disaster management.

Study Area.
Te Southeast Asian (SEA) region includes major land and oceanic islands and is characterized by tropical climatic conditions.However, the climatic setting of the SEA region is majorly controlled by the changes in the seasonal wind pattern.Te SEA region exhibits extremely complex climatic conditions and strong spatial rainfall patterns.Te root cause of these complex climatic conditions is the geographical changes caused by the proliferation of multiple islands near the sea. Figure 1 shows the summer monsoon region of SEA used as the study area covering Lao PDR, Tailand, Vietnam, Cambodia, and Myanmar, except for Malaysia in Southeast Asia.Although Malaysia belongs to the summer monsoon region of Southeast Asia, Malaysia is not used for this study because it has two land parcels separated by a wider ocean (more than 650 km).Te selected region receives rainfall mainly from the Summer Monsoon of South Asia [36].

Data. Climate Hazards Group InfraRed Precipitation with
Station Data (CHIRPS) used a spatial resolution of 5 km to generate the rainfall data, combined with satellite-estimated rainfall data with station-based rainfall measurements to provide reasonably accurate rainfall products.Te primary input data for this study is CHIRPS which is available in various time scales: daily, monthly, and annually.Monthly and annual CHIRPS data from 1981 to 2021 used for this study were downloaded from the Climate Hazard Center [45].
Boundaries of country, region, and major river basins were freely downloaded from the World Bank's ofcial boundary data catalog (https://datacatalog.worldbank.org/dataset/world-bank-ofcial-boundaries), and the FAO data  [46].A high degree of accuracy in major river basin boundaries is provided by the fact that the river basins are derived using hydrologically corrected elevation data (WWF HydroSHEDS and Hydro1K).

Methodology.
Annual data were used directly to study annual rainfall trends, and the cumulative sum of available monthly data for the relevant monsoon seasons (May to September) was used to compile data.Te CHIRPS data downloaded as NetCDF (Network Common Data Format) was converted to GeoTIFF through R-Studio software for easy reading and handling by the GIS system.Since the converted CHIRPS data are in raster format, the cumulative sum of monthly data to calculate monsoon rainfall was conducted using a geospatial data analysis technique called cell statistics.Te zonal statistical calculation of the spatial statistics tool kit of ArcGIS was then used to calculate the average values of monsoon and annual rainfall covering the geographical areas of each country and river basins.Te Mann-Kendall test and Sen's slope calculations were conducted by using the R-Studio statistical software package.

Mann-Kendall Test.
Te nonparametric Mann-Kendall (MK) test is mainly used in this study because it does not depend on the geographical settings of the particular region and is widely used to determine trends in precipitation, temperature, and river discharge [19,41,[47][48][49].Positive and negative values of the MK trend test indicate the increasing or decreasing trend of the considered parameters [50,51].Rainfall trend analysis is conducted using two approaches, i.e., the MK test and Sen's slope estimator.Whether there is a uniform linear increase or decrease trend in the considered parameter is analyzed through the MK test, and a quantitative value of that linear trend is given through Sen's slope.MK's trend test has been performed in a series of sequential data values x j and x i , where i � 1, 2, . .., n − 1, and j � 1, 2, . .., n.MK statistics "S" can be calculated by using the following equations: Te MK test requires more than ten records to analyze linear trends.Furthermore, the records used for the MK test are not considered that data are normally distributed or linear, but it needs to have autocorrelation.Te null hypothesis of the MK test is that there is no trend, and the alternative hypothesis is that there is a trend.If the number of records used for the MK test is 10 or greater than 10, the variance of S is calculated using the following equation: where t i is denoted by the number of ties of extent i.
Te Z value calculated through the MK test assesses whether or not there is a signifcant trend of increase or decrease.If the Z value is positive, it indicates an increase in the trend of the considered parameter, and the period for which the negative Z values are considered is a decreasing/negative trend of the relevant parameter.Because the MK test is a two-tailed test, |Z| > Zα/2 rejects the null hypotheses and is the level of Advances in Meteorology signifcance for the test.Te MK trend test was performed to detect trends with a 5% signifcance level and a 95% confdence level.
2.3.2.Sen's Slope Estimator.Whether or not the considered parameter has a signifcant increasing (positive) or decreasing (negative) trend is indicated through the MK test, and Sen's slope estimator is used in conjunction with the MK test to quantitatively determine the magnitude of the decrease or increase.Sen's slope estimation method uses a simple nonparametric and systematic procedure [52] to estimate the true slope without being limited to the calibration of a linear trend estimator.Using equations ( 5) and ( 6), Sen's slope can be estimated for a particular parameter.
where Qi is the slope at the ith time and x j and x k are the data values at time j and k (j > k), and N � n (n − 1)/2.Te median of the N values of Qi is determined as Sen's slope estimator (Qmed).

Results and Discussion
Tis section focuses on representing descriptive statistics of the country and the basin scale, followed by half-decadal rainfall variability analysis.

Descriptive Statistics Generated Using CHIRPS Data for
Country and Basins.Figure 2 represents annual rainfall variability over the entire study region and in the respective fve countries (Cambodia, Lao PDR, Myanmar, Tailand, and Vietnam) for 32 years (from 1989 to 2021).Although an apparent deviation in the annual rainfall can be observed in the region in 2015 and 2019 compared to other years, the overall rainfall represents an increasing trend in all countries and the entire study region.It is important to note that there was an apparent decrease in rainfall in 2015 and 2019 compared to other years.As Figure 2 indicates, among the countries in the region, the highest annual rainfall is in Myanmar, while the lowest rainfall is in Tailand.It is important to point out that there is no other country in the region that has received less rainfall than Tailand during the last 32 years.However, in 1986, 2000, 2009, 2014, 2016, and 2021 Cambodia received the highest annual rainfall in the region.
Te study of changes in annual rainfall at the river basin level is more practical and efective than at the country level because decision-making bodies such as the Basin Management Authority can easily manage their task.When considering the change in annual rainfall from the river basin level, Peninsula Malaysia receives the highest rainfall, and the Chao Phraya River basin (Figure 3) receives the lowest rainfall.
Table 2 shows the statistical parameters such as maximum, minimum, mean, standard deviation, and coefcient of variation of annual rainfall at the level of countries in SEA, while Table 3 shows the relevant annual rainfall statistics for the major river basins of the study area.Te highest average annual rainfall of 2137 mm was recorded in Myanmar, while the lowest average rainfall of 1641 mm was recorded in Tailand during the study period (between 1989 and 2021).Moreover, Myanmar and Tailand recorded their highest rainfall at 2444 mm and 2005 mm, respectively.It is important to note that the lowest rainfall between 1981 and 2021 was recorded in Tailand (1378 mm).
Considering the detailed rainfall statistics in the major river basins shown in Table 3, it is possible to identify essential diferences relative to the countries where they fall.Even though Tailand recorded the lowest average rainfall, the second-highest rainfall of 2318 mm was recorded in the "Gulf of Tailand Coast" river basin.On the other hand, it is important to detect signifcant deviations in rainfall when considering the river basins relative to the country level.In countries, the standard deviation is between 139 mm and 187 mm, but in the river basins, it is between 113 mm and 278 mm.
Te coefcient of variation (CV) is widely used to understand the interannual variability of the considered parameters.Te CV of country rainfall ranges between 7.34 and 9.56, while in the river basin, the range is between 7.30 and 12.55.Accordingly, it can be stated that a signifcant variability of annual rainfall is not identifed at the country and river basin levels.Te depiction of the 1 st and 3 rd quartiles for each country and river basin can be found in Figure 4.

Half-Decadal Average Rainfall
Variation.An important point to address in this study was the study of the diferences in the spatial and temporal distribution of the average fveyear rainfall.Te spatial and temporal variability of the fveyear average rainfall was studied for the target period (1981 to 2020) using eight intervals: 1981-1985, 1986-1990, 1991-1995, 1996-2000, 2001-2005, 2006-2010, 2011-2015, and 2016-2020.Figure 5 represents the spatial distribution of the fve-year average rainfall of the study region in fve diferent rainfall classes classifed based on the natural break statistical classifcation.
In addition to the spatial representation of that fve-year average rainfall, the change in the areas, where that rainfall was received over time was also investigated according to the rainfall classes considered.Te most signifcant fnding was   that a signifcant increase in the average rainfall above 3500 mm in the area was found from 1989 to 2020, and that value increased from 1.8% to 6.3% of the total land area studied.
Te rainfall of 2600-3500 mm also shows an increase in the areas receiving that rainfall, but the 2000-2600 mm and 1500-2000 mm classes do not observe a clear increasing or decreasing trend like other classes.However, the areas receiving less than 1500 mm of rainfall also saw a clear declining trend from 1981 to 2010, but a slight increase in the area over 2011-2015 and 2016-2020 can be observed.Te spatial distribution of the abovementioned changes can be better understood using Figure 6.Te coastal areas of Myanmar, Tailand, and parts of South Vietnam are identifed as the increased areas receiving more than 3500 mm of rainfall.

Annual Rainfall Trends in the Country.
Te annual rainfall trends were calculated through the Mann-Kendall (MK) trend test for fve countries, i.e., Cambodia, Lao PDR, Myanmar, Tailand, and Vietnam, including the entire region which is represented in Figure 7.However, the MK trend results show an increasing trend of rainfall in the entire study area and in all countries.However, Myanmar, Tailand, and Vietnam have signifcant rainfall trends.Conferring to Sen's slope estimator, the maximum annual rainfall increase recorded in Vietnam is about 5.63 mm/year, the highest among those fve countries.
Although it does not represent a statistically signifcant increase in rainfall, the minimum annual rainfall increase recorded in Lao PDR is 3.16 mm/year; furthermore, in countries that represent a statistically signifcant increase, an increase in annual rainfall of more than 5 mm/year is the key fnding of this analysis.

Annual Rainfall Trends in the River
Basins.Annual rainfall trends in the river basins also show an increasing trend while providing a positive Kendall's tau (calculated through the MK test using R-Studio statistical software) value for all the basins.However, except for the Mekong Peninsula of Malaysia, the Gulf of Tailand Coast, and the Salween, all the other river basins show a statistically signifcant rainfall trend (Figure 8).
In terms of river basins, the maximum annual rainfall trend of 11.21 mm/year has been recorded in the Chao Phraya river basin, followed by the Sittan with 9.12 mm/year.However, the lowest annual rainfall increase of 1.21 mm/ year was recorded in Peninsula Malaysia.Although a clear increase in annual rainfall can be detected in all countries and river basins, as described earlier, a detailed investigation of seasonal rainfall trends can provide important information needed to activate food security, such as crops and water management.
Tables 4 and 5 show the distribution of Kendall's tau, Z value, P value, and Sen's slope values at country and river basin levels for the monsoon seasons of dry intermonsoon (DIM-February to April), Southwest monsoon (SWM-May to October), and Northeast monsoon (NEM-November to January).An important fnding in this analysis is the presence of negative trends in both DIM and NEM monsoon rainfalls in Tailand and Myanmar, respectively.Tere is no statistically signifcant rainfall trend in the dry intermonsoon and Northeast monsoon in any of the countries except Vietnam, but positive rainfall trends (increasing) can be identifed for all Although there is a statistically signifcant increase in annual rainfall trends in Myanmar, Tailand, and Vietnam in the Southwest monsoon, the Southwest monsoon is limited to Myanmar and Tailand.Te main reason for this could be that Vietnam also receives signifcant rainfall during the Northeast monsoon.
Notably, no statistically signifcant increase or decrease was observed in any of the basins during the dry intermonsoon rainfall (Table 5).Te major river basins of the Hong (Red River), Irrawaddy, and Salween present a negative rainfall trend, while the other six basins represent a positive trend.In contrast, the Southwest monsoon shows a positive seasonal rainfall trend in all the basins, but only in the Bay of Bengal-Northeast Coast, Chao Phraya, Mekong, and Sittang basins showing a statistically signifcant increase in rainfall.Te Sittang basin indicates the highest trend (10.63 mm/year), and the minimum increase is in the Red River basin, which is 0.84 mm/year.
In the Northeast monsoon, only the Peninsula Malaysia and Vietnam-Coast river basins show statistically signifcant increases, with 5.05 mm/year and 3.37 mm/year, respectively.However, Chao Phraya, Irrawaddy, Salween, and Sittang indicate negative rainfall trends (decreasing) in the Northeast monsoon.

Conclusion
Tis study reports a long-term (1981-2021) rainfall variability and trend analysis carried out in Lao PDR, Tailand, Vietnam, Cambodia, and Myanmar, which are in the summer monsoon region of Southeast Asia together with its major river basins, using CHIRPS rainfall data.Accordingly, it was observed that for the study period (1981-2021), there is an increasing trend in rainfall for all fve countries; hence, the entire study region also shows a rising trend.Te average annual rainfall of these fve countries varies from 1641.38 mm to 2136.87 mm, whilst showing an increasing trend varying from 3.16 mm/year to 5.63 mm/year.Tis study also found that, for the study period, the average rainfall received above 3500 mm in the summer monsoon region of Southeast Asia has signifcantly increased from 1.8% to 6.3% of the total land area studied.Tis increase is mainly in the coastal areas of Myanmar, Tailand, and parts of South Vietnam.Tus, a signifcant increase in rainfall and the geographic area in which it occurs will increase the frequency of future food events.However, rainfall variability and trends provide valuable forecast information for formulating disaster risk management (DRM) policies and procedures to reduce severity and exposure to future foods.On the other hand, a clear declining trend of the areas receiving less than 1500 mm of rainfall has been observed from 1981 till 2010, and since 2010, it has slightly increased.Te observed seasonal rainfall trends during three monsoon seasons, dry intermonsoon (February-April), Southwest monsoon (May-October), and Northeast monsoon (November-January), also provide important information for efectively managing food security, such as crops and water management.Tese observations clearly show a climate change in the region that will afect the global climate in the future.Henceforth, it is recommended to consider the fndings of this study by the policymakers and other relevant stakeholders dealing with the sectors such as agriculture, urban planning, and disaster management, in the summer monsoon region of Southeast Asia.Te main problem associated, especially in developing countries, is the scarcity of weather stations; thus, it is also recommended that for similarlike study areas around the globe, researchers should investigate respective rainfall trends to better understand regional climate change by adopting the methodology followed in this study or with any other suitable approach.

Figure 3 :
Figure 3: Annual average rainfall distribution in nine river basins and the entire region.

Figure 4 :
Figure 4: Box plot of rainfall distribution of each country and river basin.

Figure 8 :
Figure 8: Annual rainfall trends in river basins.

Table 1 :
List of abbreviations and acronyms.
catalog (https://data.apps.fao.org/map/catalog/),respectively, considering their high accuracy Annual average rainfall distribution in fve countries and the entire study region.

Table 2 :
Country-wise descriptive annual rainfall (in mm) statistics.

Table 3 :
Descriptive annual rainfall (in mm) statistics in river basins.

Table 4 :
Rainfall trend at the country level in three monsoons.

Table 5 :
Rainfall trend at the river basin level in three monsoons.