Long-term homogeneity and trend analysis of seasonality and extreme rainfall under the influence of climate change in Johor River basin, Malaysia

There is a growing concern over the unprecedented shifts in seasonality and extreme rainfall over the last century across the globe due to climate change. However, there is still a lack of basin-scale study in documenting trend, changes, and variation across Peninsular Malaysia and its relation with flood occurrence. Therefore, Johor River basin (JRB) which plays a pivotal role in the country’s development was selected as a case study by evaluating 24 rainfall stations for homogeneity over the period 1970–2015 and followed by the analysis of rainfall seasonality and extreme rainfall trend. Mann–Kendall trend test was applied to show the area within JRB that is subjected to change at 95% significance level. Even though 75% of the stations had strong auto-correlation in the time series due to large-scale climate influence, the imputed rainfall stations showed no potential discontinuity with t-test statistics between 0.87 and 0.90 and R2 between 0.995 and 0.999 for the double mass curve of the yearly rainfall time series. The constructed PDF showed the restoration of the normal distribution of the rainfall data with higher peak as the missingness of the high-intensity rainfall has been restored. After that, a homogeneity test was conducted for annual and seasonal rainfall using a hybrid of Standard Normal Homogeneity test, Pettit test, Buishand Range test, and Von Neumann ratio test. The results showed that for the annual series, 43% of the total stations were found to be ‘useful’, but more ‘useful’ stations based on the seasonal data, with 67% and 52% of the total stations for NE and SW monsoons, respectively. As no anthropogenic effect can be deduced, the presence of inhomogeneity in some stations is associated with the influence of high climate variability. Although non-significant change was observed, the seasonality index (SI) showed that the annual rainfall regime in JRB is mainly classified as ‘rather seasonal with a short drier season’ (SI range from 0.39 to 0.47). Comparatively, the rainfall regime during the drier SW monsoon is more irregular (SI range from 0.67 to 0.72) than the wetter NE monsoon (SI range from 0.57 to 0.68). Spatially, increasing SI was observed in the downstream area indicating increasing occurrence of extreme rainfall over a shorter period. For trend analysis, RClimDex was utilised to compute eleven extreme rainfall indices as recommended by ETCCDI, consisting of frequency and intensity extreme rainfall indices. Generally, the results showed that increasing extreme rainfall in the form of frequency indices is more prominent throughout JRB particularly at the end of the NE monsoon. During the NE monsoon, frequency index of R10 showed a significant increasing trend at thirteen stations. Meanwhile, R20 and R25 showed increasing trend at five stations, mainly in the downstream and at the west of the basin. During the SW monsoon, R10 showed a significant increasing trend at six stations, but R20 and R25 showed a significant decreasing trend at two and one station, respectively. Based on the past flood record, increasing trend of R10, R20, R25, and CWD during the NE monsoon across flood prone area in Sayong river, Kota Tinggi town and downstream urbanised area is expected to worsen the flood conditions, which required improved flood mitigation and adaptation strategy.


Introduction
According to the IPCC report (O'Neill et al. 2016), climate change will increase the event of extreme rainfall throughout various parts of the world. Intra-and inter-annual rainfall patterns have shifted and are already becoming more intense, but high spatial variability was observed even at a basin scale, particularly in the tropical climate (Sharma et al. 2018). It has been documented that there have been increases in the frequency and duration of wet days, as well as more intense of the shorter-term rainfall events, all over the world (Tang 2019;Chen et al. 2020;Razmi et al. 2022). Increases in the intensity and frequency of extreme rainfall have resulted in devastating floods, and affecting agricultural activities, which have hampered the countries' socio-economic growth, especially the region of lowlying areas and low-income countries.
Due to increased frequency and intensity of extreme rainfall, the recent decade in Peninsular Malaysia (PM) has been marked by regular floods, particularly in the southern and eastern region (Alias et al. 2023). The major flooding seasons are mainly related to NE monsoon which comes with it, extreme rainfall. As a result, the PM is among the most vulnerable areas on the maritime continent, frequently experiencing the detrimental impacts of climate change. The state of Johor, in the southernmost part of PM, is one of the areas that has been affected by the excessive and erratic rainfall leading to the frequent occurrence of floods that cause the loss of lives, environmental damages, economic crisis and also drastically alter river morphology. In particular, the 2006/07 heavy rainfall event had caused major flooding in most parts of the state (Amin and Othman 2018). As a result, it is vital to look into the state's spatiotemporal distribution of extreme rainfall which can be a factor that causes worsening flood conditions due to climate change. Due to the lack of established scientific evidence on how basin-scale extreme rainfall has changed in the area, the mitigation action has been dawdling and any future works can be impeded due to the lack of such a basis. Therefore, information on the changes of spatiotemporal pattern relating to the frequency and intensity of extreme rainfall from the past records would help to apprehend and predict the magnitude of the extreme events like floods, severe storms, and landslides in various parts of the state. Besides, such information is required as supporting scientific documents to instil confidence for policymakers to invest in appropriate mitigation options in a number of key socio-economic sectors, notably infrastructure, agriculture, water supply, and transportation (Tang 2019;Rabiei et al. 2022).
To ensure robust assessment of the rainfall time series for extreme rainfall assessment, the accuracy of the long-term rainfall record will determine the reliability of a database for climate analysis. This is due to the several factors that often affected inhomogeneity in the time series, such as changes of operational routines, instrument errors and shifting of rainfall gauges. Also, the presence of natural catastrophes such as floods and landslides, as well as expanding urbanisation, may have an impact on rainfall data continuity. Furthermore, prior to the 1980s, rainfall monitoring systems were not properly distributed or equipped, resulting in substantial potential errors in Malaysia's long-term record of rainfall (Che Ros et al. 2016). Changes in the time series owing to the increased climate variability as a consequence of climate change must also be deliberated. Therefore, to produce a reliable analysis, it is essential to assess the time series homogeneity of the rainfall records before identifying the presence of shifts or trends and seasonality in the identified rainfall stations. Here, time series homogeneity tests are divided into two categories: 'relative method' and 'absolute method.' The relative method employs nearby stations in the testing procedure. On the other hand, the absolute method uses statistical analysis to apply the test to each station independently. The hybrid method, introduced by Wijngaard et al. (2003) which takes into consideration the method's sensitivity, has gained its popularity in recent years (Kang and Yusof 2012;Ahmad and Deni 2013;Lin et al. 2015;Chang et al. 2017). Therefore, the hybrid method will be employed in this study for the homogeneity assessment of the station rainfall time series and to assess its degree of confidence before further analysis.
For the purposes of defining some meaningful extreme indices which are consistent across a wide region, Frich et al. (2002) created the 'ETCCDI' indices to illustrate a broad range of climates and indicators. There are a total of 27 indices that have been established and used to investigate extreme climates in various places of the world (Keggenhoff et al. 2014;Razavi et al. 2016;Popov et al. 2017). Most often, extreme rainfalls were mainly investigated for frequency and intensity indices as these indicators could possibly cause stress to humans, the environment or biological conditions. Several works have been done over a larger scale of PM to understand the pattern of extreme rainfall (Paska et al. 2017), and its relationship with large-scale climate influence (Tangang et al. 2017;Tan et al. 2021). However, only one basin-scale study has been published (Muda river basin) by employing ETCCDI in Malaysia (Tan 2019;Tan et al. 2019b). Considering a high spatial variability of rainfall pattern in PM, the risk that comes from extreme rainfall can be different in terms of its severity, duration, and impact. Therefore, more assessment on basinscale extreme rainfall needs to be done across various main river basins in PM, to understand the relationship between extreme rainfall and the subsequent risk, such as flood (Tan et al. 2019b).
In addition to extreme rainfall, gaining insights into the long-term historical trend of rainfall seasonality would allow us to better understand the seasonal cycle of rainfall occurrence over the year in a particular region. The spatiotemporal information on rainfall seasonality at a certain location could aid in identifying areas of low-and high-intensity rainfall over a year. Any significant changes in rainfall seasonality will subsequently lead to temporal changes in runoff and evaporation. This has important implications in determining and designing water resiliency and conservation systems within the basin. To determine rainfall seasonality, SI is one of the powerful indicators which is calculated based on monthly rainfall concentration. Previous study by Guhathakurta and Saji (Guhathakurta and Saji 2013) on spatiotemporal rainfall variability and trend in the state of Maharashtra, India, found that SI helps in determining the distribution of the monthly rainfall and the information gained also enables a proper division between the states into different rainfall intensity regimes. This method has also been applied to other regions to gain a better insight into rainfall seasonality and its impact at a basin level (Livada and Asimakopoulos 2005;Mao et al. 2022;Swain et al. 2022). The information derived from SI can be used for a range of purposes, including studying hydrological cycles, monitoring water supplies, and as an early warning system for flood and erosion disaster monitoring (Coscarelli and Caloiero 2012;Zamani et al. 2018). The information also will facilitate water flows regulation from high rainfall intensity towards low rainfall intensity areas in the form of various water transfer projects. This is particularly important in south Johor as a raw water transfer project has been in operation by channelling water to Layang dam in Pasir Gudang and Lebam dam in Kota Tinggi to address water shortage. Another project, Projek Air Mentah RAPID (PAMER) involves a 77-km pipeline from Seluyut river in Kota Tinggi to Pengerang to serve the oil and gas industry. Thus, any further water transfer project needs to take rainfall seasonality into consideration to ensure optimum benefit.
Many works have been done on trend assessment for various hydro-meteorological and climate change parameters Jiang et al. 2021Jiang et al. , 2022Burgan 2022;Fahim et al. 2022;Salehie et al. 2023). For example, Burgan (2022) used Innovative Trend Analysis (ITA), Mann-Kendall (MK), Correlated Mann-Kendall and Seasonal Mann-Kendall and found that ITA is preferable in defining both the increasing and decreasing trend in sediment discharge in the coastal rivers in the Mediterranean and Black Sea regions of Turkey. Jiang et al. (2022) used MK test to analyse the temporal and spatial changes in vegetation coverage and drought during the growing season in the Yellow River Basin (YLRB) and Yangtze River Basin (YTRB) in China. Among other findings, temporally, they found in YTRB a significantly increasing trend (0.013/year), and spatially, 40% of the area showed a significant trend of wetness. Niu et al. (2020) also used MK test to analyse the trend in the spatiotemporal variations in temperature extremes in the YLRB and YTRB and found a high variability in the increasing trend of the investigated indices within the study area. In Bangladesh, Fahim et al. (2022) found poor sustainability of groundwater utilisation across the country due to a significant decreasing trend of groundwater using modified Mann-Kendall test. In another work, Salehie et al. (2023) used modified Mann-Kendall test to find trends in water availability in the Amu river basin and found that water availability was declining more rapidly (0.04 to -0.08 cm/year) in the tundra, warm, dry continental climate zones, and the basin's delta region. Among other methods, the nonparametric MK test has been widely used for extreme rainfall trend assessment (Zhang et al. 2022;Fu et al. 2023;Hu et al. 2023), because it can circumvent the issue raised by the skewness in the data, and is more robust against outliers, making it suitable for time series without requiring normality or linearity. Besides, the distribution of Kendall's τ can be estimated directly from a single time series, and as a result, no additional data or simulations are needed, and the computation is faster than with parametric tests (Fu et al. 2023). Therefore, the MK test will be employed for trend assessment in this study.
Regionally, several researches have been conducted on the trends in extreme climate in various parts of Malaysia (Varikoden et al. 2011;Dindang et al. 2013;Yip et al. 2015;Othman et al. 2016;Suhaila and Yusop 2018;Sa'adi et al. 2019;Khan et al. 2020). Based on the reported literature database of the Web of Science website, over 50 research articles were reported on the extreme rainfall studies over the PM, with 7 research articles in the State of Johor but none of it taking a particular interest in extreme rainfall trend and assessment at the basin level. The major keywords of the literature were presented in the thematic map as shown in Fig. 1. It is clear from the exhibited biblioshiny algorithm (Aria and Cuccurullo 2017) that extreme rainfall has an essential correlation with homogeneity, streamflow, variability, temperature, drought, and trend. However, it was also revealed that the study to correlate extreme rainfall with flood at the basin scale is lacking, despite the direct impact of flood at such level. Thus, establishing the extreme rainfall trend and time series assessment is pertinent for supplying sufficient information for river basin management and future flood mitigation plan. Although there is a growth of 5.65% in annual scientific publication in this topic across PM, there is a low annual scientific production on basin-scale cases, as the research only garnered traction in 2002 with five publications. Furthermore, there is still a lack of thorough investigation on the long-term homogeneity of rainfall station dataset and its reliability for extreme rainfall assessment in documenting trend, changes, and variation at basin-scale studies.
Until recently, most rainfall trend studies have been confined to macro-scale regions with a low number of rainfall stations and using selected climate indices to evaluate the climate trend and pattern. Only a few studies have attempted to understand basin-scale rainfall indices trends due to low number of available stations and an emphasis given to the conventional boundary system (administrative, political, development plan etc.) instead of natural boundary system (basin, ecosystem etc.). JRB is likely to face considerable water stress as a result of rapid growth in the south Johor region, including the Refinery and Petrochemical Integrated Development (RAPID) venture and several mega projects in Iskandar Malaysia. Besides, the JRB is critical for providing water to southern Johor and Singapore based on 1962 Water Agreement (Chuah et al. 2018). The river environment has Fig. 1 The biblioshiny algorithm visualisation for the occurrence of the major keyword on extreme rainfall studies in PM been reported to be stressed by vast plantation in the mid and upper regions of the river basin. (Heng et al. 2017). Excessive water abstraction, particularly during dry years, has frequently jeopardised the necessary low flow required to meet environmental standards (Pennan Chinnasamy et al. 2018). On the other hand, increase occurrence of extreme rainfall has cause frequent occurrences of flood event across the Johor river and many part of the basin. As a result, it's critical to comprehend the variations and patterns of extreme rainfall while creating long-term river management systems focused at preserving the river's health, water conservation and flood mitigation. To fill the gap, the novelty of the study found its relevance in terms of spatiotemporal mapping of the basin-scale, long-term homogeneity, and extreme rainfall trend assessment as well as classifying the rainfall seasonality across JRB. The trend results were then used to identify the area that is expected to be more vulnerable to the returning flood condition based on the official flood report data. JRB was selected as a case study due to its important role in the development of the country, and the basin is vulnerable to the worsening flood condition over the years due to climate change. The interest for basin-scale study has been growing in Malaysia, pursuant to the initiative to establish integrated river basin management (IRBM) across the main river, including JRB (Mokhtar et al. 2011;Nafchi et al. 2021;Babatunde et al. 2022). With the establishment of IRBM across the country, the findings from this study will have a far reaching effect in aiding the development of proposals and plans to lessen the basin-scale impact of climate change. Additionally, a 45 years of historical trend analysis from 1970 to 2015 will also help us to understand the extent of the current and future change in rainfall extremity in identifying areas with high risk of water-related hazards.
With a renewed interest in an integrated river basin management plan and policy to safeguard and conserve water resources by the main stakeholders in the State of Johor, it is in our interest to extend the analysis on basin-scale extreme rainfall and to classify the type of rainfall seasonality covering JRB. The high concentration of rainfall stations with long-term record in the JRB makes it possible for spatiotemporal variation analysis of the climate over the basin. The results of the study also will be important in revealing the area that is prone to flood, as well as identifying an area that is suitable for water conservation infrastructure. The findings can be used as scientific evidence for policy maker to manage the impact of climate change in the basin, as well as by farmers, water resource planners, and urban engineers to estimate water supply and plan for future demand, and flood mitigation strategy.

Study area
The Johor River is one of the primary rivers in Johor, Malaysia's southernmost state covering four districts, namely, Kluang in the upstream, Kota Tinggi and Kulai in the middle and Pasir Gudang in the downstream area (Fig. 2). The river spanning 122.7 km length with a catchment of 2636 km 2 is located at the latitudes of 1° 30ʹ-2° 10ʹ N and the longitudes of 103° 20ʹ-104° 10ʹ E. The river travels from Mount Gemuruh to the Strait of Johor, first in a north-south flow direction and then to the south-west. Linggui, Sayong, Lebam, and Tiram Rivers are its main tributaries. The total population in JRB recorded in 2015 was about 300 thousand with 70 thousand households (Department of Statistic Malaysia 2018). According to Peel et al. (2007)'s Köppen and Geiger climate classification, much of PM, including JRB, is classed as Af (Tropical rainforest climate). It has a stable temperature throughout the year, high humidity, and abundant rainfall, as well as a large spatial region of complicated topography with a wide range of local rainfall climatology. The average yearly rainfall is 2340 mm with the maximum recorded in 1995 at 3104 mm, while the minimum was in 2015 at 1826 mm. The mean maximum relative humidity, which is typical of the humid tropics, ranges between 70 and 80% (Irwan et al. 2019). Being located close to the equator, the yearly temperature is always high with an average of 26 °C. However, temperature fluctuates regionally due to local circulation and littoral coastal current, orography, geographical position, and atmospheric circulation features affecting the area.
The seasonal evolution of meteorological events in JRB is influenced by the NE and SW monsoons, which stimulate the regional and local atmospheric convection currents (Fig. 2a). Between November and March, the NE monsoon occurs, which is more apparent due to rapid increases in rainfall amounts. The intensity, frequency and duration of rain events during the NE monsoon are usually strengthened by the occurrence of cold surge and Borneo vortex at the end of the monsoon. However, due to the movement of the subpolar point crossing the equator and leavign the Southern Hemisphere during the northward equinox, the amount of rainfall falls significantly in February (Irwan et al. 2019). During this event, higher temperature is observed because the sun is located exactly perpendicular at noon across the equatorial region, as a result, less rainfall. On the contrary with the NE monsoon, the drier period of SW monsoon happened from May and September (Diong et al. 2015). Between the NE and SW monsoon, a transition period of inter-monsoon months of April and October generated locally convective events on a scale of 10 km or less, which is marked by a substantial rainfall variability (Joseph et al. 2008).

Rainfall stations
The performance of trend analysis is unable to accurately reflect the long-term climate patterns in case of a highly sensitive shorter time series, particularly at the start and end of the time series. To determine the rainfall trend, researchers used a statistical trend analysis for long-term observation data to reveal the magnitude of changes in climate patterns Fig. 2 The map of JRB within PM, its elevation and the locations of rainfall stations, and boxplot of mean monthly rainfall for 24 stations from 1970 to 2015. The blue and red box plots represent the NE and SW monsoon seasons, respectively over several decades. According to the WMO guidelines (Arguez and Vose 2011), when handling with inherent variability in climatological time series, a data period of at least 30 years is essential to establish independence in climate data. As a result, this analysis relied on daily rainfall data collected at 24 stations across JRB over a 45-year period (1970-2015) obtained from DID (Department of Irrigation and Drainage) Malaysia. The majority of the rainfall data were monitored and collected using rain gauges with automatic tipping buckets at 0.5 mm per tip, resulting in a rounding bias in the dataset (Fig. 2c). The monthly, seasonal (NE and SW monsoons) and annual totals of rainfall data were prepared to calculate the homogeneity, seasonality and extreme rainfall trend. Table 1 describes the selected rainfall stations and the annual rainfall average whereas their locations are shown in Fig. 1. It must be noted that one of the selection criteria of these stations was to get a good spatial rainfall distribution. The stations were selected within and close to the basin to enable a better spatial distribution of the rainfall assessment. There was no available observed data in the northernmost part of JRB. In addition, some stations do not have sufficiently long observation periods or suffer from a high percentage of missing data. As such, attempting to fill in missing data may not be productive. Therefore, out of the original 49 stations, only 24 stations were finally selected as listed in Table 1. Figure 3a, b showed the histogram percentage and temporal pattern of the missing data of the selected stations.

Extreme rainfall indices
In order to improve comparisons between studies on global climate extremes, the World Meteorological Organization (WMO) suggested using the ETCCDI extreme indices (Tan et al. 2019b). There are 27 core indices of climate extreme available in RClimDex, as recommended by ETCCDI with primary focuses on extreme events (Peterson et al. 2001).
Of which, eleven indices are related to extreme rainfall that can be further divided into two groups as shown in Table 2. The first group is related to frequency indices, namely, R10, R20, R25, CDD and CWD. The second group pertains to intensity, namely, Rx1day, Rx5day, PRCPTOT, SDII, R95P and R99P. The R10, R20, and R25 were used to define the number of days with daily rainfall exceeding 10 mm, 20 mm, and 25 mm, respectively. Meanwhile, the R95P and R99P were used to measure the total rainfall exceeding the threshold values of the 95th and 99th percentile of wet-day rainfall. The duration aspect of the extreme rainfall was assessed by using CWD and CDD, which represent the maximum number of consecutive wet days and consecutive dry days, respectively. In addition, SDII and PRCPTOT were also considered to investigate the annual intensity of the wet days and annual total rainfall. 1 mm was used as a threshold for CWD, CDD, SDII, and PRCPTOT. The time series of each index for daily, seasonal, and annual series were prepared for each station prior to trend analysis.

Procedure
The procedure used for the long-term assessment of the homogeneity and trend analysis of seasonality and extreme rainfall under the influence of climate change based on the  (R95p) Annual total rainfall when RR > 95th percentile mm Extremely wet days (R99p) Annual total rainfall when RR > 99th percentile mm Simple daily intensity index (SDII) Annual total rainfall divided by the number of wet days (defined as rain ≥ 1.0 mm) in the year mm Annual total wet-day rainfall (PRCPTOT) Annual total rainfall in wet days (RR ≥ 1 mm) mm/day selected 24 daily rainfall stations for the period of 1970 to 2015 in JRB are outlined below (Fig. 4). Details of the employed methods are provided in the following subsections.
1. The rainfall stations were initially selected based on the availability for long-term daily rainfall time series spanning more than 30 years, located within and close to JRB. The selected rainfall stations were then subjected to quality assessment being performed by using RClimTool. 2. After that, the missing rainfall data was imputed by using predictive mean matching provided within the MICE method from R programming packages, mice (van Buuren and Groothuis-Oudshoorn 2011). The consistency of the imputed time series was then assessed by constructing a double mass curve of the yearly time series. In addition, PDF was also constructed to assess the quality of the imputed datasets. Then, the sequential student's t test was used to evaluate homogeneity in the time series at each station, and possible outlier (or natural variability) was checked with the neighbouring stations, and evaluated based on the auto-correlation coefficient. 3. The data homogeneity test was then investigated using a hybrid of absolute method, namely, Pettit, Standard Normal Homogeneity test (SNHT), Buishand Range test (BRT) and Von Neumann ratio (VNR) tests to evaluate the reliability of the annual and seasonal (NE and SW monsoon) datasets. 4. The details on the descriptive statistics of the daily and monthly rainfall were then prepared to characterise the general rainfall pattern in JRB. 5. Finally, the trend for the annual and seasonal (NE and SW monsoon) rainfall seasonality based on SI, and extreme rainfall based on the selected ETCCDI indices was run using MK test at 95% significance level. The results were mapped for spatiotemporal analysis.

Data quality control and homogeneity test
The data quality assessment of the rainfall time series was initially performed using RClimTool to identify and replace irregular records in the database (Ferrari and Ozaki 2014). Here, a quality control (QC) log file was created for every station to document each change or acceptance of an outlier. The information derived from the QC log file was the percentage of daily rainfall data that fall within and outside the predefined limits and range of mean and standard deviation, identifying the cases of equal data in a period of longer than five consecutive days and the presence of outliers. A preliminary report consisting of graphical and descriptive analysis (Plot Charts, Graphs, Scatter plots or Boxplot) made by the application was also being assessed for each rainfall stations. The study takes advantage of the wide availability of the daily rainfall stations with long records located across JRB. However, the issue of missing value is always present in lengthy and continuous time series data, particularly in developing nations like Malaysia (Kamaruzaman et al. 2017). The relocation of the rainfall station, changes in the environment, instrument failures, and network reorganisations are the causes of the missing data and inhomogeneity issue. The employed rainfall data must be full, homogenous, and of high quality in order to produce analyses that are accurate. To address the issue of the data continuity for a practical study of the time series, the missing rainfall data at several stations were imputed by using predictive mean matching provided within the MICE method from R programming packages, mice (van Buuren and Groothuis-Oudshoorn 2011). When addressing with ambiguity in missing data, MICE created several imputations as opposed to a single imputation in the traditional method. A reasonable data value was taken from a distribution particularly built for every missing data point to replace the missing value. Many studies have proven the robustness of MICE in imputing missing rainfall data (Poyatos et al. 2018;Norazizi and Deni 2019). The selection of neighbouring stations of this study is based on a 20 km radius from the target station with a correlation value of 0.4, which is more than optimal for areas in PM based on the moderate effect size (Kamaruzaman et al. 2017). Any correlation below the threshold won't be taken into account. After the data imputation and quality control operations, the double mass curve for the yearly time series was created to see if there was a breakpoint. Possible outliers (or natural variability) were also checked by comparing values from neighbouring stations for consistency or potential breaks. The total rainfall series for the annual, seasonal, and monthly were then created and checked for any discontinuities. A widely used parametric sequential student's t test was also used to evaluate rainfall homogeneity by identifying whether or not a possible shifting point in the time series present at each station after imputation of missing value (Nashwan et al. 2019). The PDF was also constructed to assess the performance of the imputed datasets under different degrees of missingness. In addition, evaluation of the auto-correlation coefficient of a time series for yearly and monthly rainfall was performed to see if natural climate variability had an impact in record of the stations (Hyndman and Hyndman 2016). For the next subsequent assessment of homogeneity, rainfall seasonality, 1 3 and extreme rainfall trend analysis, only data series with sufficient temporal records that passed these quality control was included.
In this study, the hybridisation of several absolute methods as proposed by Wijngaard et al. (Wijngaard et al. 2003) is preferred for homogeneity tests because the locations of the rainfall stations are randomly scattered across JRB. The annual and seasonal rainfall series was investigated for homogeneity using Pettit, SNHT, BRT and VNR tests at 95% confidence level. When it comes to locate the significant change in a time series, the sensitivity of the SNHT, BRT, and Pettit tests varies. In detecting the changes at the start and end of the time series, SNHT test is more capable, but for the efficiency in identifying the changes in the middle of the time series, it is recommended to use BRT or Pettit tests (Martínez et al. 2010). Nonetheless, all of these three tests can be used to determine the break year that occurred in the time series. Meanwhile, VNR adopts a similar null hypothesis as the other three tests, but it also makes an assumption that the time series is not distributed randomly as an alternative hypothesis. Even though the time series randomness is well evaluated by VNR, unfortunately it does not identify the time series break's year. Taking into consideration all of the results that reject the null hypothesis presented by all of these four homogeneity tests, the results for homogeneity test are divided into three categories: "useful" (if one or none of the null hypothesis of the tests are rejected), "doubtful" (when two or all of the tests reject the null hypothesis), and "suspect" (when three or all of the tests reject the null hypothesis).

Seasonality index
Based on the amount of rainfall, the seasonality index (SI) classifies the climate of the basin into distinct classifications by using the equation (Abaje et al. 2010;Guhathakurta and Saji 2013;Patil 2015): where the average rainfall for the month n is define as x n and the average yearly rainfall is defined as R . The index varies from a value of 0 (all months have almost equal rainfall) to 1.83 (most rain occurs in a single month). There are seven rainfall regime classifications based on the value calculated by the SI, defined as very equable (≤ 0.19), equable but with a definite wetter season (0.20-0.39), rather seasonal with a short drier season (0.40-0.59), seasonal (0.60-0.79), markedly seasonal with a long drier season (0.80-0.99), most rain in 3 months or less (1.00-1.19) and extreme, almost all rain in 1-2 months (≥ 1.20). In order to investigate trends and changes of SI in time, the annual and seasonal SI indices were calculated for 45-year from 1970 to 2015 using the MK test as described in Sect. 4.3.1.

Trend analysis
The extreme rainfall indices were initially estimated using RClimDex, an R-based software tool created on behalf of the ETCCDMI by the Climate Research Branch of the Meteorological Service of Canada. RClimDex and its related manual and documents can be found at http:// etccdi. pacifi ccli mate. org/ softw are. shtml. Then, the presence and significance of the change was evaluated using the MK nonparametric test. Sen's slope approach was used to (1) SI = 1 determine the changing rate in the extreme at a 95% confidence level. The details of the methods are described below.

Mann-Kendall test
The nonparametric MK (Mann 1945;Kendall 1955) is a test for randomness against trend in which each data point is compared to all subsequent data points. The time series statistic (S) for x 1 , x 2, x 3 …, and x n of the MK is computed as, where The significance of the trend was statistically quantified using the normalised test statistic Z and the probability connected with S and the sample size, n, as follows: where Var (S) is the variance of S.
A positive or negative value for Z implies an upward or downward trend. If |Z| > 1.96, the null hypothesis of no trend is rejected at the 95% level of significance.

Sen's slope estimator
Kendall's tau-based slope estimator proposed by Sen (1968) was applied to calculate the trend magnitude of the equally spaced time series indices. This widely used estimator has been applied in hydro-meteorological studies because of its robustness against the effect of outliers in the time series (Chervenkov and Slavov 2019). The slope is calculated as a change in unit of measure per change in time using the equation: where at time t ∕ and t , Q ∕ is measure as the slope between data points x t ∕ and x t . Sen's slope estimator is then literally given by the median of all slope ( Q ∕ ) to show the rate of change over the whole period.
( Figure 4 provides an overview of the historical annual and seasonal rainfall patterns for the period of 1970 -2015 in JRB. Although there are differences in the boundary and amount of rainfall, a similar high variability of spatial pattern was observed between the annual (Fig. 5a) and seasonal rainfall (Fig. 5b, c). The highest mean annual rainfall was found in the north-eastern front of the basin, which is more exposed to the NE monsoon coming from the South China Sea. The rainfall pattern showed a decreasing trend transition to the north-western part of the basin, with the lowest rainfall in the westernmost area. The highest mean annual rainfall was recorded at Station 1740001 in the southeast (2565 mm) and the lowest at Station 1834001 in the West (2066 mm). About 45% (1032 mm) of the rainfall was recorded during the NE monsoon and the remaining 37% (856 mm) during the SW monsoon. The linear regression of mean annual rainfall shows a slightly decreasing trend over the study period. Meanwhile, the linear regressions of mean rainfall during NE and SW monsoons showed contrasting patterns of increasing and decreasing, respectively. This pattern may infer the increasing occurrence of extreme rainfall during the NE monsoon and drier condition during the SW monsoon.

Data quality and homogeneity test
It has been observed that the selected rainfall stations in JRB were subject to missing values of less than 20%, discontinuities and different starting and end dates, except for Station No. 1537113 (23.0%), 1636001 (28.4%), and 1834001 (43.5%). All missing data for each of the target stations have been imputed based on the selection of neighbouring stations (within 20 km radius) with good correlation (≥ 0.4) by predictive mean matching by using MICE. It was found that each of the target stations has at least two neighbouring stations that are suitable to be used for the imputation process. Application of imputation process enables more rainfall stations to be used across the basin that was previously deemed unusable due to high amounts of missing data. The imputation process also provides a longer and reliable period of rainfall time series analysis in the case of JRB compared to previous studies (Tan et al. 2014(Tan et al. , 2015(Tan et al. , 2019a. Previous works have shown a good capability of the MICE method for imputation of hydro-climatological data (Milo et al. 2019;Norazizi and Deni 2019;Farzandi and Rezaee-Pazhand 2021). RClimTool was used to construct a quality control (QC) log entry for each station to capture each difference or acceptance of an outlier. The information derived from the QC log file was the percentage of data within and outside the predefined limits and range of mean and standard deviation, identifying equal data in a period of longer than five consecutive days and the presence of outliers. A preliminary report consisting of graphical and descriptive analysis (Plot Charts, Graphs, Scatter plots or Boxplot) for the percentage of data within and outside the predefined limits and range of mean and standard deviation, identifying equal data in a period of longer than five consecutive days and the presence of outliers made by the application for each of the stations has been assessed and found to be reliable for further assessment (result not shown). With R 2 between 0.995 and 0.999, the double mass curve for the yearly rainfall time series reveals no discontinuity in the time series. This implies that the rainfall data following imputation is consistent. The sequential Student's t test was then used to evaluate homogeneity by identifying whether or not a possible shifting point in the time series is present at each station after imputation of missing value. The t test statistics for all stations were between 0.87 and 0.90. As the critical t values were much higher compared to the test statistics at 0.05 significance level, it can be concluded that potential discontinuity does not exist in the time series at any station. The constructed PDF for the selected imputed station (Fig. 6) showed that the characteristic of the rainfall has been retained with the restoration of the normal distribution of the rainfall data for the station even with the highest missingness (≥ 15%) (Kamaruzaman et al. 2017). The result of the imputed dataset also shows higher peaks, which indicate that the missing value for the high-intensity rainfall has been imputed which is important for a reliable assessment of extreme rainfall.
According to Sun et al. (2018), in order to properly estimate long-term rainfall change, it is crucial to adequately account for natural variability. By evaluating the data series' shift in time between the present value and its past values, and calculating the correlation with the original time series, stationarity may be evaluated using the autocorrelation function (Ridwan et al. 2021). In this study, the auto-correlation function of a monthly rainfall time series was estimated to have considerable auto-correlation in 18 of the 24 stations in JRB (Table 1), indicating the effect of natural climate variability (Hyndman and Hyndman 2016). A similar result was found by Nashwan et al. (2019) in the stations over the southern state of Johor, which displayed a heterogeneous result, demonstrating non-stationarity for various rainfall intensities that depend on the location of the station. The similar spatial output based on the homogeneous location of the stations with/without considerable auto-correlation gives a fair estimate of the area being influenced with the large-scale climate phenomena. Five of the six stations with no substantial auto-correlation were all in JRB's northwest region, which could be owing to the region's low exposure to large-scale climate phenomena. Che Ros et al. (2016) confirm that ENSO has a crucial role in causing unexpected increases in longterm rainfall variability in the Kelantan River basin. Suhaila and Yusop (2018) discovered discontinuity or break points in temperature time series that may be related to climatic conditions such as El Niño and La Niña occurrences. Furthermore, because 75% of the stations had strong auto-correlation in the time series, all of the stations evaluated in this study can be used for the subsequent analysis because the rainfall series shows no anthropogenic effect.
The data homogeneity test was further investigated using a hybrid of absolute method, namely, Pettit, SNHT, BRT and VNR tests to evaluate the reliability of the datasets. As shown in Table 3, for the annual series, 43% of the total stations were found to be 'useful', and 29% stations each under 'doubtful' and 'suspect' categories. The results are almost similar with the finding by Kamaruzaman et al. (2017) who found that 40% of the station is homogeneous after missing data has been imputed based on the selected 104 daily rainfall stations throughout PM. There are more 'useful' stations based on the seasonal data, with 67% and 52% of the total stations for NE and SW monsoons, respectively. 10% stations were 'doubtful', and 24% stations were 'suspect' for NE monsoon. Meanwhile, 19% stations were 'doubtful' and 29% stations were 'suspect' for SW monsoon. The results for VNR suggest that NE monsoon series are more randomly distributed compared to the SW monsoon. This indicates that rainfall temporal patterns are more consistent over the years during NE monsoon compared to SW monsoon. The latter might receive a sudden rainfall in different months over the SW monsoon. The study results showed more detailed information regarding homogeneity distribution of rainfall stations in JRB compared to other studies that cover a broad area of PM (Suhaila Syed Jamaludin et al. 2008;Kang and Yusof 2012;Ahmad and Deni 2013). The details of the results for the homogeneity test are presented in Table 4. Its spatial distribution can be observed in Fig. 5a.
Due to the low distribution density of the stations used in this study, the absolute method is considered to be more appropriate. However, the absolute test has difficulty to differentiate inhomogeneity from the influence of natural climate variability (Sahin and Cigizoglu 2010). To confirm the data inhomogeneity at each station, the year with the observable discontinuity was compared to those of neighbouring stations that had been deemed homogenous. All inhomogeneous sites with varying magnitude of rainfall variability near the break point showed a similar pattern. It's worth noting that the time series' breakpoint coincided with the ENSO event, suggesting that natural temporal variability might be a source of inhomogeneity. The correlation is similar with previous study by Suhaila and Yusop (2018). Therefore, we include all the 24 stations for the subsequent rainfall seasonality and trend analysis as no systematic errors can be inferred from the rainfall series. In addition, there are always reliable neighbouring stations for each inhomogeneous station that can cater for the need for spatial analysis except for Stations 1737001 and 1636001. The consequential results from the inhomogeneity of Table 3 The homogeneity test classification ('useful', 'doubtful' and 'suspect')  Class 2 'doubtful' 6 (28.57%) 2 (9.52%) 4 (19.05%) Class 3 'suspect' 6 (28.57%) 5 (23.81%) 6 (28.57%) rainfall records are rarely fully incorporated in the hydro-climatology analysis, but it can be a potentially large source of error. The current study is interested in applying the inhomogeneity results to ascertain the degree of reliability of the seasonality and trend analysis.

Rainfall descriptive statistic
The rainfall spatial distribution was highly variable considering the small size of the basin as shown in Fig. 7. The highest rainfall was in the East of the downstream area, while the lowest was in the North West of the upstream area. Spatially, only 3 stations along the West of the basin showed significant increasing trend under MK test and 1 station in the East showed significant decreasing trend. Although a similar spatial pattern with high correlation of 0.80 between annual mean daily and monthly rainfall was observed, daily rainfall showed higher variability than monthly rainfall with up to threefold deviation from the mean value. The comprehensive descriptive statistical information for the monthly rainfall at JRB is shown in Table 5 below, and it reveals a highly variable temporal pattern of rainfall. The standard deviation ranged from 38 to 132 mm, while the monthly averages ranged from 125 mm in February to more than twofold increase at 275 mm in December. The average    the equinox. All months showed right-skewed (except in May, July, and September) which indicates that more rainfall occurred at the lower bound, which means the tendency of rainfall above the average. The kurtosis showed a negative value mainly during the SW monsoon months of April, May, June, August, and September which means flatter than normal peak distribution. While mainly during the NE monsoon months of January, February, March, July, October, November, and December showed a positive value, which means sharper than normal peak. This indicates the distinct behaviour of rainfall intensity between the SW and NE monsoon. The Shapiro-Wilk test of normality showed that all months are normally distributed. The influence of the large-scale climate phenomena such as equinox during the month of February (Irwan et al. 2019), and the influence of cold surge and Borneo vortex during the peak of NE monsoon months of October, November and December can be a factor for sharper kurtosis peak and right-skewed rainfall distribution (Ng et al. 2022;Liang et al. 2023).

Seasonality
A long-term daily and monthly rainfall data could provide an important indicator to understand the rainfall regime that modulates the spatiotemporal distribution and occurrence of extreme rainfall. Therefore, the variability and concentration of annual and seasonal rainfalls in JRB were evaluated using SI based on the monthly rainfall. SI facilitates in detecting rainfall regimes based on monthly distributions in order to see if there have been any shifts over the last 45 years. Table 6 shows that the SI trend for both annual and monsoonal series is non-significant under MK test at 95% confidence level. However, it is worth revealing the trend direction of the SI to better understand the seasonal pattern across the basin. For the annual series, we can see that 3 stations (1734001, 1735125 and 1735142) which are located at the central West of the JRB show an equal but with a definite wetter season type of rainfall regime. The rest of the stations are rather seasonal with a short drier season, indicating that rainfall mainly occurred and distributed more than 6 months in a year across JRB.
However, the monsoon season shows relatively higher seasonality with SI values for SW monsoon range from 0.67 to 0.72 and for NE monsoon from 0.57 to 0.68. Although the SW monsoon brings lower rainfall in comparison with the NE monsoon, the rainfall regime during the SW monsoon was found to be more irregular. During the NE monsoon, seven stations mainly in the downstream of JRB show a rather seasonal rainfall regime with a shorter drier season. This shows that the high rainfall occurrences during NE monsoon period are temporally distributed in this area with three out of five months (Nov, Dec and Jan) being wetter. Meanwhile, the rest of the stations are seasonal with certain months (Dec and Jan) of the NE monsoon period being wetter than the other. Although no significant trend was observed, the SI results show more increasing trends as extreme rainfall over a shorter period may occur more in the downstream area.

Trend analysis
The MK test was employed to determine the presence of a trend in the extreme rainfall data series at 95% significance level. The results in Table 7 showed that both significant decreasing and increasing trends were found for annual extreme rainfall in JRB for 8 indices (R10, R20, R25, R95p, R99p, CDD, CWD and SDII). There are 7 indices with increasing trends (R10, R20, R25, R95p, R99p, CDD and CWD) and 3 indices with decreasing   Suhaila and Jemain (2009) also found a different class of rainfall being grouped homogeneously together across PM which might be due to the effects of orography and other geographical factors. For the seasonal trend, 6 indices were significantly changed for both NE and SW monsoons, namely R10, R20, R25, CDD, CWD and PRCPTOT. In general, a higher number of stations with a significantly increasing trend can be observed during the NE monsoon than SW monsoon for all indices except for CDD. On the contrary, more stations with a significant decreasing trend can be observed during the SW monsoon period. The distinct characteristic of the trend results between NE and SW monsoon suggests that monsoon plays a major influence that cause abrupt change in the frequency and intensity of the extreme rainfall (Tan 2018). It has been argued that the enhanced evapotranspiration and air moisture holding capacity resulting from climate change have altered the geographical and seasonal distribution of rainfall in many places of the world (Nashwan et al. 2019). Numerous studies also have documented an increase in the likelihood of extreme rainfall in PM's annual and seasonal rainfall mean and variability (Wong et al. 2016;Li et al. 2018;Cui et al. 2019).
Our results suggest that although annual trends can reveal the inter-annual changes of the extreme rainfall, the most important local features under the monsoon reveal more information regarding the trend of extreme rainfall for a particular period. Specific information can be determined at the temporal scale of the monsoon, such as the type and spatial distribution of the extreme rainfall trend, which was concealed from the annual trend. Various impacts of the changes in the extreme rainfall is of importance during the monsoon as the occurrence of floods and drier periods usually happen at certain times of the monsoon. Therefore, the subsequent analysis gives more focus on the spatial trend of the extreme rainfall during the monsoon.

Spatial distribution of monsoonal trend
The spatial distributions and trend of extreme rainfall indices show a distinct spatial trend characteristic between the NE and SW monsoons (Figs. 8,9). The spatial information on the flood occurrence and annual frequency of flood event based on the official annual flood report by DID is presented in Fig. 10. The relationship of extreme rainfall trend was discussed in relation with flood occurrence across JRB in the subsequent assessment. During the NE monsoon, 13 significant increasing trends were observed for R10 at the southern/ south-eastern (9 stations) and western (4 stations) parts of the basin. It was found that up to 33 days and 29 days per NE monsoon with more than 10 mm of rainfall on average was observed at the southern/south-eastern and western part of the basin, respectively. During the SW monsoon, only 6 significant increasing trends of R10 were observed at the south (2 stations) and western (4 stations) parts of the basin. It was found that up to 30 days per SW monsoon with more than 10 mm of rainfall on average was observed at both of these areas. Comparatively, more stations with significantly increasing CWD (4 stations) during the NE monsoon compared with 1 station during the SW monsoon indicate that the continuous wetter condition during the NE monsoon coupled with heavy rainfall up to 10 mm daily rainfall may potentially cause floods to happen. It was noticeable that the significant increasing trend for R10 is observed along the Sayong river, which has recorded flood events over several parts of the river in the past. Being in the upstream area, the river 1 3 channel was narrow and it is expected that increasing trend of R10 in the area may lead to more frequent flood events in the future. The middle part of the Johor river, in Kota Tinggi town which recorded a returning flood with one of the severe flood events in 2006/07, also  showed an increasing trend of R10. Therefore, flood protection and mitigation plans for this area are pertinent. The southern part of the basin, concentrated with urbanised and high population areas, will also be more vulnerable to flood events as several stations in this area showed an increasing trend of R10.
Significant increasing trends were observed at 5 stations for R20 and R25, respectively, mostly concentrated in the middle, South and south-eastern parts of the basin during the NE monsoon. For R20, it was found that up to 17 days, 18 days, and 19 days per NE monsoon with more than 20 mm of rainfall on average was observed at the middle, south and south-eastern parts of the basin, respectively. For R25, it was found that up to 13 days, 16 days, and 16 days per NE monsoon with more than 25 mm of rainfall on average was observed at the middle, south and south-eastern parts of the basin, respectively. One station showing an increasing trend at the junction between Linggiu and Sayong river showed both increasing trend for R20 and R25. The rapid migration of the converged water upstream is particularly a concern, as the immediate area downstream, in Kota Tinggi town, is prone to flood. Although the increasing trend was observed in the south and south-eastern parts of the basin rising the fear of flood, such an event is expected to increase water availability in the Seluyut and Layang reservoir which frequently recorded low levels of water in recent years. On the contrary, during the SW monsoon, two significant decreasing trends for R20 at the West and South while one significant decreasing trend for R25 was detected at one station in the South. The results suggest that rainfall frequency has significantly increased during the NE monsoon, and on the contrary has decreased during the SW monsoon. As for SW monsoon, the rainfall frequency for R20 and R25 showed significant reduction indicating that rainfall is shifting and contributing more R10 across the basin.
Significant increasing trends in CDD were observed at four stations during the NE monsoon; three stations in the south and one station in the western parts of the basin. On the other hand, a significant decreasing trend was observed at one station in the central-north. A similar spatial pattern during the SW monsoon with significant increasing trends were observed at three stations in the south and one station in the west. Significant decreasing trends were also observed at one station each in the central-north and in the west. The significant increases in extreme rainfall as represented by indices R10, R20 and R25 in some areas across JRB doesn't necessarily mean that it is temporally uniform during the monsoon period. The significant increase in CDD during the NE monsoon, particularly in the south, may suggest that the extreme rainfalls (R10, R20 and R25) happen consecutively over a longer duration but within similar months of Dec and Jan, and followed by longer and more frequent dry day within the same monsoon period in Feb and Mar.
For CWD, four stations with significant increasing trends were found across JRB during NE monsoon, while only one station with significant increasing trend was found during the SW monsoon. The stations with significant increase in CWD based on the R10, R20 and R25 indicate a trend of longer duration of high rainfall events in these areas towards the end of monsoon season. Therefore, it can be inferred that the longer duration and increasing trend of extreme rainfall at the end of the NE monsoon are expected to cause more severe flood events in the future in Kota Tinggi town and Sayong river. For SDII, only one station with a significant decreasing trend was observed in the south during NE monsoon. This indicates that although rainfall frequency (R10, R20, R25, including CWD) has increased in the other part of the basin, its intensity has decreased further downstream. This also means that shorter duration storms are becoming more often at the lower end of the basin. One station with a significant decreasing trend was also observed during the SW monsoon but located in the north west of the basin. The low intensity rainfall in this area might be due to the decreasing rainfall frequency (R10) as this area is shielded by high elevation terrain.
More rainfall has been recorded in the urbanised area downstream of the basin. This could lead to an increased occurrence of flooding events. In the downstream area, excess runoff could be quickly discharged into the sea which presents its own challenge in terms of water conservation. Coupled with increasing sea level due to climate change, it is expected that the communities living in the downstream area will be more vulnerable to floods caused by increased rainfall extremity. Therefore, it would be in the interest of water conservation to consider giving more attention to the north-western part of the basin which receives higher frequency (R10) of extreme rainfall during both monsoons. Generally, the results showed that increasing extreme rainfall in the form of frequency indices is more prominent throughout the basin. Improved design and infrastructure to handle sudden increase of rainfall over a shorter duration during the end of NE monsoon should be considered to mitigate or reduce the impact of flood and, at the same time, cater the needs for water conservation in preparation for the drier period of SW monsoon that follows.

Conclusion
In this work, statistical tests are used to determine the long-term homogeneity, rainfall seasonality, and trend of extreme rainfall in JRB. The homogeneity test found that seasonal datasets during the NE and SW monsoons resulted in higher degree of reliability compared to the annual datasets. However, the presence of inhomogeneity in the time series was found in some of the stations, which might have occurred due to the influence of climate variability affecting the local climate in the study area. Careful notations about the changes in the nearby station should be considered to confirm any inhomogeneity detected in the rainfall time series. Coupled with the imputation of the missing value, the homogeneity test should be applied as a precursor for any advanced statistical assessment of rainfall to ensure a good reference site can be obtained as many as possible.
The SI defined over JRB was found to be 'rather seasonal with a short drier season' with certain areas located at the central West of the basin show an 'equal but with a definite wetter season'. Monsoon-based SI assessment indicates that the relatively drier period of SW monsoon receives sporadic monthly rainfall compared with more consistent high monthly rainfall during the NE monsoon. Although rainfall seasonality has not significantly changed, an increasing pattern was observed especially in the downstream area, which may indicate an increasing tendency of intense rainfall during the peak of the NE monsoon. To gain more information on how the different parts of the PM may be different in terms of rainfall seasonality, it is recommended to apply SI for the whole of the PM to understand how the monsoon influence may be affecting the different parts of the region.
Extreme rainfall characteristics must be understood since they can have a significant influence on society, particularly in the light of climate change. The results showed that both significant decreasing and increasing trends were found based on the applied ETCCDI indices, with some indices showing both directions of the trend, indicating a variability of non-uniform trend at the local spatial scale. A further monsoon-based trend assessment suggests that global warming is linked with the distinct influence of the NE and SW monsoons in JRB. It was found that frequency-based indices, namely R10 (13 stations), R20 (5 stations), and R25 (4 stations), showed more stations with significant increasing trend during the NE monsoon compared with the SW monsoon. Therefore, flood vulnerability of the area along the Sayong river, the middle part of the Johor river, in Kota Tinggi town, and the urbanised downstream area are expected to increase due to the returning flood with the worsening condition during the peak of the NE monsoon months. Thus, strategies for flood protection and mitigation should be focused on this area. On the other hand, increased water availability due to extreme rainfall can be a valuable resource that needs to be captured for maintaining water level in the Seluyut and Layang reservoirs which frequently recorded low levels of water in recent years.
However, the research is also subject to several limitations that must be borne in mind. For example, the inhomogeneity test used in this work, related to the start and end of climatological periods and their assessment can be open to misinterpretation even though consideration of the usage of several method's sensitivity was considered. Meanwhile, the MK test is subject to the effect of auto-correlation that persists within the data, which needs to be removed as it will inadvertently increase the chance for significance in trend (Da et al. 2022). Besides auto-correlation, the significance of the trends over time is sensitive to the assumptions about whether the underlying data have a short-term or long-term persistence (LTP) due to the influence of natural variability that found to be exist in some of the rainfall station, as shown in the prior homogeneity assessment (Amirthanathan et al. 2023). Therefore, even though the MK test is widely accepted for trend assessment, other methods capable of reducing or removing auto-correlation and the influence of LTP should be considered to confirm the significance of the trend.
As extensively discussed by Alexander et al. (2019), there are several limitations on the usage of ETCCDI for extreme rainfall assessment. For example, if changes are unevenly distributed over the distribution, the outcome for the percentile-based indices may depend on the specific threshold choice. Besides, the sensitivity of the sample used to determine the percentiles is also an issue. In order to prevent biases from being created, the reference period chosen must also be standardised with sufficient length. Other drawbacks include dry regions in which the majority of the days are dry and distributions where 0 mm values can exceed the 90th percentile. The definition of a wet day itself may also be difficult because small amounts of rainfall are sometimes undetected at manual sites or, conversely, spurious small reports are made at automatic sites that may be caused by dew accumulation. Additionally, indicators don't always tell the whole story and occasionally need to be examined in conjunction with other indicators and variables that take into account shortages in soil moisture, streamflow, and groundwater, all of which are impacted by temperature changes and consequently affected by evapotranspiration. Another example is the fact that Rx1day shows the annual maximums of daily rainfall but provides no information on when these annual extremes will occur.
Nonetheless, this research provides significant contributions to regional climate change research and will enhance our understanding of the impacts of climate change at the basin level. Extension of this approach to other river basins, where little is currently known regarding trends in hydrological extremes, would be valuable for highly variable climate in the tropic. Further study also should be made to investigate and confirm the possible causes of homogeneity and the observed trends in extreme events. This study also can be a reference for other works which utilise the gridded-based rainfall datasets for comparative assessment. Importantly, this study can be used for scientific evidence to guide policymakers in pursuant of integrated river basin management in case of JRB, and for the subsequent future works relating to future projection of extreme rainfall under climate change scenario.

Availability of data and materials Not applicable.
Code availability Not applicable.

Declarations
Ethical approval Not applicable.

Consent to participate Not applicable.
Consent for publication We, (Zulfaqar Sa'adi, Zulkifli Yusop, Nor Eliza Alias) hereby declare that we participated in the study in the development of the manuscript titled (Long-term homogeneity and trend analysis of seasonality and extreme rainfall under the influence of climate change in Johor River basin, Malaysia). We have read the final version and give our consent for the article to be published in the journal of Natural Hazards.