Radar-Based Rainfall Estimation of Landfalling Tropical Storm “PABUK” 2019 over Southern Thailand

,


Introduction
A large number of studies on rainstorm analysis obtained from capability for spatial and temporal storm profiles measurement of weather radar are widely used to detect the convective storms and study of convective storm structure [1][2][3] and also the process of the rainfall system itself by providing real-time regional information, and with the existence of long radar data sets, these data could be also applied for climatological applications.In addition, weather radar is also tool that combines meteorology and hydrology [4][5][6]; the meteorological information measured by radar are used for hydrological analysis as referred to Peng et al. [7], and they explained that the advantage of using radar for precipitation measurement is the coverage of a large area in real-time, and radars also experience difficulty in achieving an accurate estimation for hydrological applications.e single-polarization [3] and dual-polarization [8,9] weather radar was used to study the relationship between radar reflectivity and rainfall rate which is developed for rainfall measurement by using the Z-R relationship.e uncertainty of radar rainfall estimation by using Z-R relationship was proposed by Chen et al. [10] and Gou et al. [11].e developed Z-R relationships are needed in ailand to provide a more systematic and comprehensive approach to achieve water management.
Applications of weather radar in ailand are still limited mostly for meteorology and monitoring the weather routines.Not much work has been done in the field of hydrological and heavy rainfall cases.Finding rainfall intensity is one of the essential applications for weather radar in the process of hydrology, flood management, and early winning system in case of severe weather situation.For the purpose in radar rainfall estimation, the relationship between radar reflectivity and rainfall rate is developed for rainfall measurement by using the Z-R relationship.e Z-R relationship is highly dependent on the precipitation types and wind conditions such as convective, stratiform or mixed types, and deep convection [12][13][14][15].Event type is one of the major influences of Z-R relationship that must be studied accordingly.Moreover, the location of areas and seasonal also plays an important factor in applying Z-R relationships to radar rainfall measurements [1,16].Most weather radars in ailand are not calibrated for the Z-R relationship.As the results, the developed radar rainfall estimation in case of severe weather situation from tropical storm is needed in ailand providing a more systematic and comprehensive approach to achieve in water management and also additionally to implement in flood warning purposes.Because the southern part of ailand is a major economic tree plantation zone, especially fruits, oil palm, and rubber tree, it is almost in transition from water richness to water scarcity because of the increasing demands on this limited resource as well as there is no universal Z-R relationship that can be applied to all cases of rainfall events.erefore, the focus of this paper will be on the optimization of Z-R relationships during tropical storm and heavy precipitation which were tuned to fit the rain gauge measurements that turn into inaccuracies over the central region of ailand.
is article is structured as follows.e data and methodology section presents the overview of PABUK tropical cyclone, the technical characteristics of the radar used, radar data analyses, and the statistics for the analysis.
is is followed with presenting the results of rainstorm properties and radar rainfall estimation during PABUK event, and the article closes with is a brief discussion about the conclusion reached.e influence of tropical storm PABUK caused strong wind shear, heavy rain, and flash floods from D1 to D3 where the accumulated rainfall reached 150-300 mm a day across 14 provinces of southern ailand, especially in Nakhon Si ammarat, Surat ani, Chumphon, Ranong, Phatthalung, Songkhla, Pattani, Yala, and Narathiwat, the maximum rainfall in 24 hours reached 309.3 mm, and maximum wind speed was 89 kmhr −1 at Nakhon Si ammarat province on D2.In addition, PABUK also results in rising sea levels and blowing into the shore as storm surge in the coastal region of upper southern region; the images of the damage caused by a tropical storm PABUK are as shown in Figure 1. e study area has been southern ailand, characterized by a complex topography and directly influenced by the South China Sea and Indian Ocean (Figure 2).e C-Band Doppler Radar, which represents a good compromise between range and reflectivity that can provide rain detection up to a range of 240 km, from ai Meteorological Department (TMD) was used in this study.

Data and Methodology
e reflectivity data from Songkhla's radar located in the eastern coastline near the landfall of PABUK tropical storm were appropriately used to investigate the rainstorm properties and estimated radar rainfall that affected the southern region of ailand.Radar is installed at Sathingpra District, Songkhla, at the elevation of 33 m MSL in southern ailand as shown in Figure 2.
e radar with EDGE ™ software collected the reflectivity data as volume scan to the highest altitude up to 5 km provided in the universal flies (UF) format [17].e files were obtained every 15 minutes up to the effective range of 240 km to the highest altitude up to 5 km provided in the volume format files for 4 elevation angles: 0.5 °, 1.5 °, 2.4 °, and 3.4 °, and a Doppler filter is applied to remove ground clutter and fixed echoes.
In order to characterize rainstorm properties, radar reflectivity data in horizontal polarization were run through TITAN and used the 30 dBZ reflectivity threshold to identify a convective storm cell before tracking their movement as referred to Dixon and Wiener [18]; Johnson et al. [19], and Potts et al. [20].All of rainstorms were selected and analyzed by the dataset from the criteria as suggestion by Chantraket et al., [1] including their 5 properties as exhibited in Table 1.However, it may be mentioned that in this study, only those rainstorms are considered whose rain centers are located in the effective range of 240 km of Songkhla's radar.e event numbers differed in each day; these data were then analyzed to express properties of individual rainstorms.e total of 2,557 rainstorm events were chosen from D1 (1,014 rainstorms), D2 (962 rainstorms), and D3 (581 rainstorms), respectively, during occurred PABUK tropical cyclone.e example case of rainstorm events from TITAN analysis is presented in Figure 3, and the preliminary of statistical analysis of all properties is illustrated as Table 2.
In order to estimate radar rainfall, the relationship between radar reflectivity and rainfall rate which is developed for rainfall measurement was used as explained in the following equation: where a and b are the relationship parameters, Z is the radar reflectivity in mm 6 m −3 , and R is the intensity of precipitation mmh −1 .e rainfall events were used to obtain the appropriated Z-R relationship for tropical storm in southern ailand as well as to test an accuracy of the proposed radar rainfall estimation based on different Z-R relationships as referred to Kirtsaeng and Chantraket [16] of Songkhla's radar of Z � 104R 1.5 , Z � 162R 1.5 , and Z � 184R 1.5 and the operational Z-R relationships of TMD Z � 300R 1.4 as referred for all rain types especially for deep convective [21,22].In order to evaluate the suitable Z-R relationship for tropical storm of PABUK, their measurement wascompared with the precipitation recorded by the rain gauges from automatic meteorological stations of Hydro Informatics Institute (HII) and TMD. e study was performed by using 24-hour accumulations of 156 rain gauges in and around southern regions in the coverage of radar effective range (see Figure 2) having been scrutinized during D1 to D3. ese data were procured from the (i) 129 stations of HII and (ii) 27 stations of TMD.After procurement of 24-hour accumulations from different sources, these were subjected to extensive quality control tests to remove gross errors, archival errors, and reformatting problem; however, the suspected data were checked for validation from different reliable sources.e study was carried out using 24-hour accumulations, comparing with estimated rain accumulation from Z-R relationship that occupied the position of rain gauges.Rain gauge measurement and radar estimate of 24-hour accumulations greater than 1.0 mm and less than 300 mm were considered to be valid.Daily rainfall distribution in southern ailand caused by tropical storm PABUK during D1, D2, and D3 is shown in Figure 4.
e statistical indexes used to evaluate the different of estimated rain accumulation from Z-R relationship related with the rain gauges as recommended by [23] are as follows.
Root mean square error (RMSE) is as follows: Mean error (ME) is as follows:     Mean absolute error (MAE) is as follows: Bias is as follows: where G is 24-hour accumulations of rain gauges at station i in mm, R is 24-hour accumulations of radar rainfall computed using Z-R relationship at the point with coincided to rain gauge station i in mm, and N is the number of precipitation records.Several Z-R relationships would be specified by the calculation of equations ( 2) to (5).Whichever relation provides the minimum of the four statistical measures will be selected as the most suitable relations for the study.

Rainstorm Duration (SDUR)
. SDUR is the time elapsed from the first radar reflectivity of 30 dBZ until the disappearance of precipitation. is study revealed that the average individual SDUR during D1 to D3 was around 0.9 to 1 hour, and almost all SDURs in D1 were less than 2 hours as in D2 and D3 tended to be longer duration than D1 as 1 to 3 hours.It is implied that the longer SDUR in D2 and D3 would extend the potential of rainfall intensity near the center of tropical storm PABUK in accordance with extremely heavy rainfall in D2 and D3 of rain gauge measurement when the rainstorm made landfall over southern ailand.e time series and frequency distribution of SDUR on D1 to D3 are illustrated in Figure 5.  Mathematical Problems in Engineering 7

Rainstorm Bases (SBAS).
e SBAS show the minimum height of radar reflectivity, as the minimum reflectivity threshold and altitude are determined as 30 dBZ and 0.6 km, respectively.e results of this study showed that an average of SBAS is quite similar in D1, D2, and D3; between 1.8 and 2.0 km MSL, all events of occurred rainstorms are lower than  4 km during PABUK occurrence.e time series and frequency distribution of SBAS on D1 to D3 are illustrated in Figure 6.

Rainstorm Reflectivity (SREF).
e result from this study shows that on average maximum value, storm's reflectivity peaks that occurred during PABUK are distinguished among three days and tend to be higher SREF from D1, D2, and D3 as 35.0 dBZ, 38.6 dBZ, and 43.5 dBZ, respectively.Furthermore, the maximum SREF was found as a stronger reflectivity more than 60 dBz which corresponds to the precipitation intensity and the development of PABUK tropical cyclone during D1 to D3 as well.e variation of SREF along three days in PABUK period is presented in Figure 7.

Rainstorm Area (SARE).
e average SARE during PABUK period shows difference among D1, D2, and D3. e largest SARE from individual rainstorm is shown in D2 that is approximately 55.8 km 2 .It is seen that the area of individual rainstorms of D2 when PABUK made landfall over Nakhon Si ammarat province was found to be more potential rainstorms than D1 and D3 according to the extremely heavy rainfall in D2 occurred near the landfall point and inland of southern part.e D2 rainstorm composes of several large cells as well as they can also lead to larger areas of precipitation.e variation of these properties along three days in PABUK period is presented in Figure 8.

Rainstorm Speed (SVEL).
e TITAN algorithm can provide the information of storm tracking and its movement.e results obtained from this study show that the average SVEL of D1, D2, and D3 was 23.1 kmhr −1 , 22.5 kmhr −1 , and 18.2 kmhr −1 , respectively.Most of SVEL has tend to be at lower speed when landfalling and passing through the land of southern ailand.Investigating the maximum speed in D1, D2, and D3 as illustrated in Table 2, it is found that maximum SVEL of individual rainstorms was observed in D2 consistently with the report of maximum sustained wind during PABUK occurred period as well.e time series and frequency distribution of SVEL on D1 to D3 are illustrated in Figure 9.

Radar Rainfall Estimation.
e radar estimated rainfall during PABUK landfall period was analyzed by the step as explained in the previous section.e result of estimated rainfall accumulation using Z-R relationship in four trials of (1) Z � 184R 1.5 , (2) Z � 104R 1.5 , (3) Z � 162R 1.5 , and (4) Z � 300R 1.4 is compared.e comparison of the 24 hr accumulated radar rainfall and the 24 hr accumulated gauge rainfall using the four trials of Z-R relationship is presented in Figure 10, and the statistical measures comparing these two sets of data are also calculated and summarized in Table 3.
Figure 10 shows the images of estimated daily radar rainfall attained from four Z-R relationships in D1, D2, and D3 and the scatter plot of the 24 hr accumulation of estimated radar rainfall attained from the different Z-R relationships and 24 hr accumulated gauge rainfall during D1 to D3. e estimated daily radar rainfall using the four trails of relation in D1, D2, and D3 was plotted as shown in the left side of radar images.From the scatter plot, it can be noted that the estimated radar rainfall accumulation is mostly higher than accumulated rain gauges except for Z � 300R 1.4  and also shows that Z � 104R 1.5 can provide the closest compared with the scatter plot of the other relations.
An agreement between estimated radar and gauge rainfall was examined using the statistical measures resulting from the four trials of Z-R relationships.e results show that the Z � 104R 1.5 is acceptable for overall statistical measures, with minimum of the four statistical measures, RMSE, ME, MAE, and BIAS, between the estimated radar and calculated rain gauge rainfall for the data sets in D1, D2, and D3. e calibrated Z-R relationship of Z � 104R 1.5 is therefore appropriate to be used for an estimation of accumulated radar rainfall in the tropical storm of PABUK.

Conclusions
e study presented the physical properties of rainstorm and radar-based rainfall estimation during tropical storm PABUK moving into the lower Gulf of ailand and making landfall over southern ailand which affected the southern regions with widespread heavy rainfall and flash floods.Derived from the data set of radar reflectivity and rain gauges during three days of PABUK, all storm properties were analyzed with TITAN, and estimated radar rainfall specified the appropriated Z-R relationship by the selected statistical measures.e results are shown as follows: (1) ree days (D1, D2, and D3) during the tropical storm of PABUK in order to investigate rainstorm of southern ailand obtained the 5 properties of rainstorms by using TITAN, which provided the important analysis tool to identify rainstorms and their movement in this study.It is revealed that rainstorms were found to be the most effective clouds over southern region.In accordance with the Table 3: Comparisons of the statistical measures gained from the different Z-R relationships of Z � 184R 1.5 , Z � 104R 1.5 , Z � 162R 1.5 , and Z � 300R 1.4  Mathematical Problems in Engineering characteristics of rainstorms, it can be seen that rainstorms in D2 and D3 when PABUK made landfall over Nakhon Si ammarat province, southern ailand, show a longer lifetime, higher reflectivity, and larger rain-cells as well as it was found efficient in terms of rainfall amount than in D1 consistently with the records of high precipitation depth in southern in that periods.
(2) e appropriated Z-R relationship acceptable for estimated radar rainfall during the tropical storm of PABUK in southern ailand is Z � 104R 1.5 , which provided the minimum of the four statistical measures (RMSE, ME, MAE, and BIAS) so far as the southern basin is concerned.e results should be especially useful in urban design problems as well as in hydrologic design problems during unusual cases such as extremely heavy rainfall from tropical storm on the southern basin.(3) ese results are provided to assess the planning of water resources on a probability in a particular region or a basin and made to provide improvement of hydrometeorological relations that are pertinent to hydrological applications in the southern region and also capable of adapting to the other parts of ailand.Relationships presented are subject to modification as additional data are collected in the heavy windy rainstorms, and further research is conducted.

Figure 3 :
Figure 3: Some case of rainstorm events obtained from TITAN analysis of Songkhla's radar on D2 at 14 : 03 UTC: (a) TITAN analysis image and (b) cross section of selected rainstorm.
January 2019 over Phanang, Nakhon Si ammarat province, at latitude of 8.2 °N and longitude of 100.2 °E.Maximum sustained wind is 75 kmhr −1 , and the storm was moving northwest at a speed of 18 kmhr −1 .PABUK became the first tropical storm to make landfall over southern ailand since Linda in 1997.
2.1.Overview of PABUK Tropical Cyclone.Tropical storm PABUK, which has originated from the low-pressure zone in the South China Sea, developed from tropical depression first defined on 31 December 2018. is storm moved westward into the lower Gulf of ailand, and it made landfall on 4

Table 1 :
Radar-obtained storm characteristics from TITAN and their units.