The research domain area was Greater Jakarta, commonly referred to as Jabodetabek, which consists of Daerah Khusus Ibukota Jakarta, Bogor, Depok, Tangerang and Bekasi. It is considered a functional region, a classification based on its high concentrations of economic activity and population. It lies within an area of complex coastal–continental interaction. The province of Banten lies to the west, while that of West Java extends across the eastern and southern regions. The northern region is directly adjacent to the Java Sea. In addition, Sumatra Island is located to the northwest and contributes to the broader regional-scale atmospheric configuration.

The selection of cors and meteorological stations was based on the spatial extent of the study domain across Greater Jakarta, and data availability during the MCS period. The meteorological station of Serang (96737), as well as the BAKO and CJKU cors stations, were excluded from the analysis. That of Serang did not record observational data during the MCS period, thereby preventing data processing and subsequent analysis. In addition, data from the BAKO and CJKU cors stations were not accessible due to data request restrictions imposed by Badan Informasi Geospasial (BIG), the official data provider.

Tabular data of extreme weather and flood disasters based on (B.N.P.B., 2025) are accessible at https://dibi.bnpb.go.id/. The dataset encompasses categories such as location, time, casualties, damage and losses. These data points were utilized as benchmarks to quantify the impact of intense precipitation events specifically attributed to MCS. By integrating the disaster data, our study evaluates the vulnerability and real-world consequences of MCS-driven hydrometeorological hazards.

Rainfall observation data 3h from synoptic observations of Meteorological Stations Tanjung Priok (WMO ID 96741), Cengkareng (WMO ID 96749), Kemayoran (WMO ID 96745), Curug (WMO ID 96739), and Citeko (WMO ID 96751). Databased on (B.M.K.G., 2025), accessible via https://dataonline.bmkg.go.id. These data are utilized for a temporal analysis of rainfall and cloudiness at specific meteorological station points during MCS event.

Brightness temperature (Tb) data from the Global Geostationary Satellite based on (Janowiak et al., 2017) are accessible at https://disc.gsfc.nasa.gov/. The data category selected was NCEP/CPC L3 Half Hourly 4km Global Merged IR V1 (GPM_MERGIR). Due to the 30-minute temporal resolution, there are two data points per hour. The spectrum of the top-of-atmosphere emission is related to an equivalent temperature (Minnett & Barton, 2010), while Tb is used to identify and track convective clouds associated with events (Feng et al., 2023). In addition, the precipitation feature (PF) of the Global Precipitation Measurement (GPM) mission satellite are based on (Huffman et al., 2023). The data category selected was GPM IMERG Final Precipitation L3 Half Hourly 0.1 degree V07 (GPM_3IMERGHH). This provides rainfall estimates derived from the passive microwave (PMW) sensor using water vapor motion vectors ((Jin et al., 2021); (Joyce & Xie, 2011)). The 30-minute IMERG precipitation data are averaged to represent hourly rainfall total. According to (Feng et al., 2023), PF is defined as contiguous regions characterized by measurable precipitation rates to enhance convective cloud system (CCS) segmentation,with the use of hourly data considered sufficient for MCS identification.

GNSS RINEX observation data based on (B.I.G., 2025) are accessible at https://srgi.big.go.id/. The data were acquired from the following Continuously Operating Reference Stations (CORS): CBTU (Cibitung), CJKT (Tanjung Priok) and CTGR (Tangerang). The data were used to conduct a temporal analysis of PWV at CORS station points during the MCS. Model ERA5 data of specific humidity and air pressure based on (E.C.M.W.F., 2025) are accessible at https://cds.climate.copernicus.eu/ in NetCDF format.

The data have an hourly temporal resolution, a spatial resolution of 31 km, and 37 vertical pressure levels. The dataset was employed to analyze the PWV over the area and grid points during the MCS. Brightness temperature infrared-3 (IR-3) data from Himawari-9 satellite band 8 (wavelength bandwidth 6.2 µm) was based on (B.M.K.G., 2025) via Himawari-Cast in NetCDF format. The data have a temporal resolution of 10 minutes and spatial resolution of 2 km. The dataset is able to indicate upper-level moisture atmospheric water vapor and was thus used to analyze the water vapor parameter during the occurrence of the MCS.

In line with (Feng et al., 2023), the study employed the PyFLEXTRKR algorithm for MCS identification and tracking. This can determine the distribution and movement of convective clouds based on the temperature fields derived from satellite imagery. Fundamentally, the algorithm executes five main steps: reading input data, identifying cloud features, tracking, computing statistics, and final tracking. It monitors all convective clouds that exceed a specified minimum area threshold, which is defined based on the modified cloud-identification parameters outlined in Table 5. MCSs are then identified based on the tracked clouds.

Therefore, MCS tracking within PyFLEXTRKR encompasses the initial stage of convective cloud growth (before reaching mesoscale dimensions) through to the dissipation stage. MCS cloud trajectory is marked with a track number corresponding to the trajectory index in the track statistics file. Each track number generates a variety of final statistical variables. Various MCS types are separated using predefined criteria based on the statistical values of each track. In this study, algorithm categorization is based on MERGIR brightness temperature data and IMERG precipitation data from the GPM satellite.

The modified physical characteristics presented in Table 5, which are used as threshold inputs for PyFLEXTRKR and adjusted for the Indonesian Maritime Continent, were developed from MCS criteria proposed by (Maddox, 1980) and (Nuryanto et al., 2019). According to (Maddox, 1980), MCSs were identified using Tb-enhanced infrared (IR) satellite imagery from the Middle Latitude, focusing on single-cell system to as mesoscale convective complexes (MCC). Physical criteria included cloud shield with Tb ≤ 241 K (−32 °C) and areas exceeding 100,000 km², as well as interior cold cloud region with Tb ≤ 221 K (−52 °C) and areas greater than 50,000 km².

Furthermore, according to (Nuryanto et al., 2019), MCS identification used the GTG method (Whitehall et al., 2015) to Tb IR Multi-functional Transport Satellite (MTSAT) on Tropical Area, using physical thresholds of mean Tb ≤ 221 K (−32 °C) and a system size greater than 10,000 km². Despite methodological differences, both studies adopted common constraints of eccentricity greater than 0.7 and a minimum duration of 6 hours. Determination of PWV from the GNSS observation data is based on an algorithm developed by (Geng et al., 2019) using the PRIDE-PPPAR software package. PWV is derived from zenith tropospheric delay (ZTD), specifically the component categorized as zenith non-hydrostatic delay (ZNHD), which is commonly referred to as zenith wet delay (ZWD).

ZTD represents the total excess path length, or time delay, imposed on GNSS signals as they propagate through a neutral atmosphere in the vertical direction 199220252023((Bevis et al., 1992);(Isnaini et al., 2025);(Lian et al., 2023)). This delay results from changes in the atmospheric refractive index driven by both dry gases (zenith dry delay/zenith hydrostatic delay) and water vaporZWD/ZNHD). Although ZTD introduces substantial error in high-precision GNSS positioning, it remains a fundamental parameter in meteorology for assessing the troposphere’s dynamic conditions using Equation 1 and 2.

(1)

(2)

where ZWDini = zenith wet delay initial (mm); ZWDcorr = zenith wet delay correction (mm); N_nh(z) = non-hydrostatic refractivity at height z.

Calculation of PWV requires a component weighted mean temperature (Tm (K)). This is a quantity defined by (Davis et al., 1985), derived by integrating a function of temperature and water vapor pressure profiles along the atmospheric column. In essence, Tm represents the water vapor-weighted mean temperature of the atmosphere. The specific Tm algorithm used in this study follows the Equation 3 and 4 detailed by (Landskron & Böhm, 2018).

(3)

(4)

ρ_v = water vapor density (Kg m-3); T =temperature (K); R_v= specific gas constant for water vapor (461,51 J Kg-1 K-1); ρ_w = water density (0,997 kg m-3); k₂'= refractivity constant (16,48 K mb-1); k₃ = refractivity constant (3.776 K2 mb-1).

PWV ERA5 was calculated based on specific humidity and air pressure parameters across 37 levels. These values were then integrated from the surface layer up to the tropospheric boundary layer (or tropopause). The calculation algorithm utilized followed Equation 5 methodology established in research by (Jiang et al., 2016) and (Zhang et al., 2019).

(5)

q = specific humidity (Kg Kg-1); p_s = surface pressure (Pa); ρ_v = water vapor density (1000 kg m-3); g = gravity (9.780325 m s2).

PWV data analysis on GNSS, ERA5 and BT Upper-level Moisture using Equation 6, 7, 8, 9, and 10 statistical method bias, mean absolute error (MAE), RMSE (root mean square error), SD (standard deviation), corr (Correlation), and R2 (Coefficient Determination).

(6)

(7)

(8)

(9)

(10)

N = total data; xi = PWV GNSS (mm); y_i = PWV ERA5 (mm) or BT upper-level moisture (K); = mean of PWV GNSS; = mean of PWV ERA5 or BT upper-level moisture.

Figure 3 illustrates the observed rainfall data collected by rain gauges at the meteorological stations. Measurements were taken every three hours, adhering to the standard synoptic observation procedure. A total of five stations were utilized, each identified by its WMO ID number. They are distributed across the Greater Jakarta area (Jabodetabek), with their locations detailed in Figure 1 and Table 4.

Table 6 presents the statistical values. The maximum rainfall peak across all stations occurred between 15:00 and 18:00 UTC, with the absolute highest maximum rainfall recorded of 42.0 mm at both Tanjung Priok and Citeko stations. The overall average maximum rainfall across all stations was 30.6 mm, which is classified as heavy rainfall. The observed rainfall fluctuations demonstrated significant increases and decreases, forming a unimodal peak pattern across all stations. This pattern is clearly reflected in the trend line, which represents the moving average.

During heavy rainfall over Jabodetabek, analysis of the convective cloud system was conducted using the PyFLEXTRKR method with the Tb Global Geostationary satellite and Pf GPM satellite. The results are presented in Table 7 and Figures 4-5. Table 7 compiles the physical MCS characteristics, including phase, period, time, area, eccentricity, Tb, latitude,and longitude. Figure 4 shows further processing related to the spatiotemporal characteristics of MCS, while in Figure 5 the spatial distribution of MCS clusters during the initiation, mature and termination phases can be seen.

Table 7 shows the MCS characteristics based on Tb and Pf. The table presents phase, period, time, maximum area of Tb, maximum area of Pf, Pf eccentricity, mean Tb, latitude and longi-tude. The MCS demonstrated a total lifetime of 15 h, comprising a 12 h development and ini-tiation phase, 1 hour mature phase, and 2 h dissipation and termination phase. At the beginning of the development and initiation phase, MCS exhibited a maximum Tb area of 0.161 x 1011 m2 (16100 km2) and a maximum Pf area of 0.026 x 1011 m2 (2600 km2). The system then developed into the mature phase, with the MCS reaching a maximum Tb area of 2.063 x 1011 m2 (206300 km2) and maximum Pf area of dan 0.659 x 1011 m2 (65900 km2). At the end of the termination phase, the MCS maximum Tb area was recorded as 1.643 x 1011 m2 (164300 km2), and maximum Pf area as 0.469 x 1011 m2 (46900 km2). This evolution is consistent with the temporal distribution pattern of the maximum Tb and Pf areas shown in Figure 4a, where a significant increase in area was observed during the development and initiation phase, fol-lowed by the highest peak during the mature phase, and a subsequent reduction in area during the dissipation and termination phase. Furthermore, a strong relation between the maximum Tb and Pf areas is evident, as illustrated in Figures 4a and 4b.

The spatial output of PyFLEXTRKR for MCS identification is presented in Figure 5. The MCS labeled index number 2 is the focus of this study, as it was responsible for the heavy rainfall over Jakarta. At the beginning of the initiation phase, the MCS with Tb ≤ 221 K (−52°C) only covered the Bogor and Tangerang areas, with an estimated rainfall of up to 10 mm h⁻¹. In the mature phase, the MCS covered the entire Jabodetabek area, with an estimated rainfall reaching 15 mm h⁻¹. Finally, during the termination phase, the MCS only covered the Bekasi area, with an estimated rainfall of 3 mm h⁻¹. The distribution of MCS eccentricity is shown in Figure 4c. Statistically, the eccentricity has an average of 0.878, with a maximum value of 0.970 and a minimum of 0.734. Therefore, the resulting data range is 0.236. This indicates that the MCS clouds generally had a rounded and circular shape. The tight spread of eccentricity values in the boxplot indicates low variability. The fluctuation pattern of mean Tb is presented in Figure 4d. The pattern exhibits an inverse relation compared to the data for the MCS maximum area (Figure 4a). Mean Tb reaches its peak faster than the maximum area value. Statistically, mean Tb has an average of 209.436 K (-63.564 °C), with a maximum of 212.213 K (-60.787 °C). The minimum mean Tb reached was 206.146 K (-66.854 °C). The MCS trajectory from the initiation phase to the termination phase is shown in Figure 4e. The observed trajectory moved from the southeast of West Java at 09.30 UTC and ended in Lampung, Sumatra at 00.30 UTC. It was fairly long because of its relatively long lifetime of 15 h. During the end of the termination phase shown in Figure 5c, the MCS is still observed in the brightness temperature image to have a large area. This occurs even though the cloud cells have already split when viewed from their precipitation features.

In the following observations, the MCS cloud cell was not found to dissipate immediately, but to undergo redevelopment, as shown in Table 8 and Figures 6-7. Table 8 indicates the physical characteristics of the regenerated MCS, including phase, period, time, area, eccentricity, Tb, latitude and longitude. Figure 6 depicts the tracking trajectory of the extended MCS, while Figure 7 presents the spatial distribution of the MCS clusters during the initiation phase, representing a continued stage of cloud development from the previous MCS event. The subsequent MCS cloud cluster is identified by index number 5.

The MCS entered its initiation phase at 02.30 UTC. The cloud cell that had already split during the dissipation stage at 00.30 UTC developed again into an MCS. At the beginning of the ini-tiation phase, the MCS had a maximum Tb area of 0.0097 × 10¹¹ m² (970 km²), a maximum Pf area of 0.047 × 10¹¹ m² (470 km²), an eccentricity of 0.911, and a mean Tb of 211.407 K (−61.593°C). The formation location was at 2.011°S, 104.705°E. In the mature phase it reached a maximum Tb area of 1.905 x 1011 m2 (190500 km2) and maximum Pf area of 0.615 x 1011 m2 (61500 km2). Although at that time the MCS had not yet reached its maximum area, the mean Tb had the lowest temperature of 204.985 K (−68.165°C). At the end of the termination phase, the maximum MCS Tb area was recorded as 0.143 x 1011 m2 (14300 km2) and maxi-mum Pf area as 0.935 x 1011 m2 (93,500 km2). The movement of the MCS can be observed in Figure 6 in the form of a track trajectory. It had a very long lifetime of 29 h.

During the MCS period, statistical analysis was conducted at each cors station using time-series and scatterplot diagrams, as shown in Figures 8 and 9. The time series diagrams in Figure 8 indicate the temporal variability of the PWV derived GNSS and ERA5. The plots facilitate identification of PWV evolution in response to MCS development and assessment of fluctuation patterns, trend consistency and their temporal correspondence. Figure 9 presents scatterplot diagrams to examine the distribution and relationship between PWV GNSS and ERA5. Each plot is complemented by statistical uncertainty, including bias, MAE, RMSE, and SD. The framework provides a comprehensive and detailed statistical evaluation of the relation between PWV GNSS and ERA5 during the MCS period.

In the diagrams, the timescale follows the temporal resolution of the ERA5 data, available only hourly. Figure 8 shows that the GNSS and ERA5-derived PWV values gradually increased from 09.00 UTC. They then reached their highest (peak) PWV values at 15.00–16.00 UTC for GNSS and 13.00–16.00 UTC for ERA5. This corresponds to the time when the MCS center was closest to the station, at 15.30 UTC, as shown in Figure 4e. For the CBTU station (Figure 8a), the highest (peak) PWV values for GNSS and ERA5 occurred at 15.00 UTC, at 66.442 mm and 64.756 mm respectively. For the CJKT station (Figure 8b), peak PWV for GNSS occurred at 16.00 UTC at 67.289 mm, while peak PWV for ERA5 occurred at 14.00 UTC at 64.426 mm. For the CTGR station (Figure 8c), peak PWV for GNSS occurred at 15.00 UTC at 66.999 mm, while the peak PWV value for ERA5 occurred at 13.00 UTC at 64.474 mm. GNSS PWV then decreased gradually, before increasing again at 21.00 UTC, although not significantly. ERA5 PWV shows irregular increases and decreases. The PWV values increased significantly at 22.00 UTC, before decreasing significantly at 23.00 UTC.

Figure 9 shows that the relation between GNSS PWV and ERA5 PWV varied greatly during the MCS period. Correlation values were 0.041 (very low) at the CBTU station, 0.378 (low) at the CJKT station, 0.560 (moderate) at the CTGR station, and 0.280 (low) when all the stations were combined. The trend line resulting from this relation has a positive slope.

Figure 10 presents a scatterplot that illustrates the relationship between GNSS PWV and the distance between the MCS center and stations, as well as mean Tb during the evolution of the MCS. This aims to identify the factors that influence PWV changes measured at the CORS sta-tions. Distance refers to the position (latitude and longitude) of the MCS center detected by PyFLEXTRKR, as shown in Table 7, relative to the position (latitude and longitude) of the sta-tions shown in Table 4.

Figures 10a, 10b, 10c and 10d indicate strong to very strong correlations at each CORS station, together with a composite analysis. The correlation ranges from −0.798 to −0.862, confirming a consistently strong inverse relation between PWV GNSS and the distance from the station to the MCS track. The results demonstrate that distance is a key controlling parameter influen-cing PWV variability during the MCS period. A negative correlation implies that shorter dis-tances between CORS stations and the MCS trajectory are associated with higher PWV values, whereas greater distances correspond to lower ones. This pattern reflects the enhanced mois-ture accumulation in proximity to the active convective system. The tendency is further sup-ported by the time series analysis presented in Figure 11, which shows increasing PWV values as the MCS approaches the station, and decreasing ones as it moves away.

Mean Tb refers to the average brightness temperature detected by PyFLEXTRKR, as shown in Table 7. Figures 10e, 10f, 10g and 10h indicate that correlation is very weak to weak. There-fore, mean Tb has a small influence on PWV. ERA5 cannot be compared with the distance of stations to the MCS track and mean Tb because of the differences in observation times. ERA5 is available at full hour intervals, whereas PyFLEXTRKR outputs are not.

Figure 11 illustrates the timeseries relationship between PWV and distance. The results show that PWV changes follow the variations in distance that occur. Maximum PWV values are reached when the distance between the MCS and station is small. Therefore, the PWV value closely depends on the distance of the MCS from the observation location.

Figures 12, 13 and 14 present a comparison of PWV GNSS and Tb upper-level moisture. The timescale follows the temporal resolution of the Himawari satellite data, which is available every 10 minutes. The time series plots in Figure 12 illustrate the temporal variability of PWV and Tb, enabling identification of fluctuation patterns and temporal relations at each CORS station and in composite. Figure 13 shows scatterplots indicating the relation between PWV and Tb based on the composite distribution of CORS stations. Furthermore, Figure 14 shows scatterplots of each station and the composite with statistical uncertainty, including bias, MAE, RMSE and SD.

Figure 12 shows that fluctuations in GNSS PWV and BT upper-level moisture form opposite curves during the MCS period. When GNSS PWV increases as the MCS approaches the station location, Tb upper-level moisture decreases. This is indicated by negative correlation values at all CORS stations. Figure 13 presents a point distribution of PWV GNSS and Tb Upper-level moisture. The polynomial curve represents the best trend line with the highest coefficient of determination (R²) of 0.256. This is because the R² values for the linear (0.077), logarithmic (0.071), power (0.070) and exponential trend lines are lower, with the exponential value being negative.

In addition, Figure 13 shows that the highest PWV values and lowest Tb upper-level moisture values occur at different times, but still within a short time interval. GNSS PWV reaches its highest (peak) value first, followed by the lowest Tb upper-level moisture value. For the CBTU station (Figure 10a), highest (peak) PWV occurs at 15:30 UTC, with a value of 66.470 mm, and the lowest upper-level moisture Tb occurs at 16:30 UTC with a value of 191.661 K (−81.489°C). For the CJKT station (Figure 10b), highest (peak) PWV occurs at 15:50 UTC with a value of 67.383 mm, and lowest BT upper-level moisture occurs at 17:00 UTC, with a value of 191.661 K (−81.489°C). For the CTGR station (Figure 10c), highest (peak) PWV occurs at 14:40 UTC with a value of 67.051 mm, and lowest Tb upper-level moisture occurs at 15:50 UTC, with a value of 189.539 K (−83.611°C).

Figure 14 presents scatterplots illustrating the relation between Tb upper-level moisture and the distance of stations to the MCS track, including each CORS station and the composite analysis. The CORS stations have weak to moderate negative correlations, with correlation coefficients of −0.232 for CBTU, −0.142 for CJKT and −0.470 for CTGR, while the combined dataset produces a correlation of −0.264. Negative slopes of the regression lines indicate that increasing station–MCS distance tends to be associated with slightly lower Tb, although the strength of this relationship remains limited. The statistical indicators also confirm this pattern, as reflected by relatively high MAE and RMSE values of approximately 145 to 149 km, which demonstrate significant dispersion of data around the regression line. Among the stations, CTGR exhibits the strongest moderate negative association, whereas CBTU and CJKT show very weak relationships. Overall, the findings indicate that variations in MCS proximity exert only a minor to moderate influence on upper-level brightness temperature at the observation locations, suggesting that other dynamical and thermodynamical processes during MCS evolution play a more dominant role in controlling BT variability.

The rainfall observations show a significant increase in precipitation, with heavy‐rain catego-ries between 09:00 and 21:00 UTC at five meteorological stations, Tanjung Priok, Kemayoran, Cengkareng, Curug and Citeko. The rainfall fluctuations in Figure 3 represent a typical rainfall pattern associated with an MCS, namely high rainfall with heavy intensity over a long period. Maximum rainfall occurred at 15:00–18:00 UTC, during which time the MCS entered the ma-ture phase at 17:30 UTC, with its track was closest to the Greater Jakarta area at 15:30–17:30 UTC, as shown in Figure 4e. Therefore, the MCS caused maximum impact in the Jabodetabek region. These findings align with earlier studies which have reported MCS events that gene-rated heavy to extreme rainfall over Jakarta during other periods ((Nuryanto et al., 2019), (Nuryanto et al., 2021); (Nuryanto et al., 2018); (Nuryanto et al., 2018)); over the western coast of Sumatra (Rais et al., 2021); and in Papua (Febrizky et al., 2023). In addition, MCSs have been shown to contribute significantly to rainfall enhancement across Indonesia ((Azka & Trilaksono, 2024); (Ismanto, 2011); (Trismidianto & Satyawardhana, 2018)).

The MCS characteristics show a long lifetime of 15 h. The PyFLEXTRKR method monitors MCS evolution very well, including its lifetime, phase, area, eccentricity, Tb, latitude and lon-gitude. The maximum Tb area and maximum Pf area have a strong correlation, indicating that the method is highly effective in observing affected areas. This aligns with earlier applications of the method in Indonesia ((Azka & Trilaksono, 2024); (Rajagopal et al., 2023)). The eccentricity of MCS clouds shows consistency and persistence in their elliptical–circular shape, with va-lues of ≥ 7. The boxplot diagram in Figure 4c shows that the eccentricity data are tightly clus-tered, with Q1 of 0.846 and Q3 of 0.924.

The minimum mean Tb is 206.146 K (−67.004 °C), which occurs during the mature phase at 17:30 UTC. The MCS tracking results in Figure 4e show the distance between the MCS center and observation stations, allowing for further analysis in Figures 10 and 11. The mean Tb ex-hibits a small difference compared to (Nuryanto et al., 2019), who obtained mean Tb 208 K (-65.150 oC). In addition, the movement of the MCS tends to originate in the southeast, which is associated with the Australian monsoon during that period. This finding is consistent with ((Chen et al., 2022); (Ismanto, 2011); (Ramos-Pérez et al., 2022); (Rustiana et al., 2019)). After an MCS passes its termination phase, clouds form a new system in the form of an extension of the previous MCS. The observation results in Table 8 show that the subsequent MCS had stronger characteristics and intensity, with a lifetime of 29 hours, a maximum Tb area of 2.531 × 10¹¹ m² (253,100 km²); maximum Pf area of 1.152 × 10¹¹ m² (115,200 km²); maximum eccentricity of 0.984; and minimum mean Tb of 204.985 K (−68.165°C). The direction of movement tended to be from the southeast, moving across the Sumatra region.

PWV observed by GNSS and ERA5 experienced fluctuations during the MCS period, as shown in Table 8. However, the correlation is weak to moderate. Maximum correlation of PWV fluctuations between GNSS and ERA5 during the period occurs at CTGR station, with a value of 0.560. PWV fluctuations produced by GNSS have a smoother pattern. PWV shows regular increases and decreases at each time step, following the movement and evolution of the MCS. IN addition, ERA5 shows significant anomalous PWV changes at 22:00–01:00 UTC.

PWV-GNSS has a stronger relation with the distance between stations and the MCS track than with mean Tb, as shown in Figure 10. The correlation is strong to very strong. Therefore, it can be observed that fluctuations in PWV follow the distance of the MCS from the observer, as shown in Figure 11. The closer the MCS is to the observation point, the higher the PWV value; the further away the MCS is, the lower the PWV value. PWV reached its maximum value when the distance between station and MCS was shortest, at 15:30 UTC. This can be seen in the MCS tracking trajectory shown in Figure 4e. Distance therefore has a significant influence on PWV.

GNSS PWV has a moderate relation with Tb upper-level moisture, corresponding to water vapor channel 8 on the Himawari-9 satellite. A maximum correlation of 0.49 was observed at the CTGR station, as shown in Figure 12. In addition, distribution of relation value can be represented as a downward-opening curve with an R² value of 0.256. Therefore, BT tends to follow fluctuations in PWV that occur during the MCS period.

The correlation results relate to temporal differences between the parameters and instruments, including GNSS, ERA5, the Himawari satellite, and PyFLEXTRKR outputs. All correlations and data comparisons refer to the same time and point locations, ensuring that the datasets were equivalent and comparable.