Reconstruction of Extreme Rainfall Event on September 19-20, 2017, Using a Weather Radar in Bengkulu of Sumatra Island

Extreme rainfall accompanied by strong winds hit the province of Bengkulu in the western coastal area of Sumatera Island during September 19-20, 2017, causing floods and landslides in Seluma and Central Bengkulu district. This extreme rainfall was recorded by Bengkulu Meteorological Station about 257.0 mm day−1 using rain-gauge observation. The spatial distribution of extreme rainfall cannot be seen if only using a rain-gauge observation in this location. The spatial distribution of extreme rainfall is needed to identify the impact of rainfall on landslides in large areas. The study aims to (1) develop the reconstruction of the spatial distribution of extreme rainfall using weather radar and (2) investigate the trigger that caused extreme rainfall by analyzing the synoptic-scale tropical waves. Each weather radar datum is saved in a Constant Altitude Plan Position Indicator (CAPPI). To get rainfall information, the CAPPI must be derived from Quantitative Precipitation Estimation (QPE) values. In this paper, we derived CAPPI using a Marshall-Palmer reflectivity-rain rate relationship. The result shows that rainfall formed on September 20, 2017, 21.00 UTC with total daily rainfall ranged between 176 and 247 mm in both districts and the mean of total daily rainfall has exceeded the average of monthly rainfall. The analysis of tropical waves suggests that only Kelvin waves were active and served as a possible trigger factor while the Madden-Julian Oscillation (MJO) and Equatorial Rossby (ER) waves were inactive during this extreme rainfall.


Introduction
On 20 September 2017, an extreme weather event occurred in the province of Bengkulu on the western coast of Sumatra Island causing floods and landslides in several areas such as Kampung Bahari, Bengkulu, Padang Pelasan and Ngalam villages in Seluma district, and Tanjung Raman in Central Bengkulu district. Extreme weather is very likely to occur in the region of Bengkulu Province due to topographic contours and geographical location. e growth of convective clouds is one of the causes of extreme weather in the region which is related to regional and local factors such as tropical cyclone effects, eddy, shear line regions, and coastal shape [1,2].
During the extreme event, observation in the Bengkulu Meteorological Station recorded the maximum temperature of 26°C while the range of maximum temperature in September is about 29°-32°C. For rainfall measurement, it reached 257.0 mm per day at the Meteorological Station Fatmawati Bengkulu while at Pulau Baai Climatological Station it reached 230.2 mm per day. ese values are higher than the monthly average rainfall of September which is only about 220 mm/month for climatology of 1981-2010 so that these are classified as extreme rainfall [3,4]. e limited number of observations over the Bengkulu region makes extreme rainfall events difficult to be analyzed, and the existence of weather radar data offers a solution for this issue.
C-band weather radar is capable of detecting short-term weather conditions (near real-time) and has a high resolution and wide coverage area up to a radius of 240 km over the Bengkulu area. Weather radar measures electromagnetic radiation from rain clouds and has the potential to estimate rainfall intensity (R) by utilizing reflectivity data (Z). Weather radar can be used to provide early warning and analysis of extreme weather phenomena such as heavy rain events, tornadoes, gusty winds, and wind shear [5].
is study aims to reconstruct the extreme rainfall event in Bengkulu on 20 September 2017 using weather radar data. Furthermore, the possible trigger of this extreme rainfall was investigated by analyzing the synoptic-scale tropical waves.

Radar Data Processing.
In this study, the reconstruction of rainfall extreme uses C-band Doppler radar data at Bengkulu Meteorological Station (102.341280°E, 3.858720°S, 45.0 meters above ground level) with Gematronik type and maximum radius coverage of 240 km (Figure 1(a)). Data are recorded every 10 minutes in a volumetric format consisting of 10 Position Plans Indicator (PPI) scans (0.5°, 1.2°, 2.1°, 3.2°, 4.4°, 6.0°, 7.8°, 10.0°, 12.6°, 15.7°, and 19.5°) and each contains reflectivity decibels (dBZ with dBZ � 10 log Z) (Figure 1(b)). A python-based open-source software Wradlib was utilized to process weather radar data. Wradlib has been widely used in the processing of weather radar data and its application [6,7]. It has an important function in weather radar data processing, particularly in generating the Quantitative Precipitation Estimation (QPE) products. An example of its use is for the estimation of rainfall from radar data that shows good results in river flow estimation and simulation of flood events in the Philippines [8] and Bangka Island [9]. Wradlib has been used as a component to develop the Indonesia In-House Radar Integration System (Ina-RAISE) of BMKG [10,11]. e reconstruction of extreme rainfall event from 19 to 20 September 2017 utilized all of radar PPI layers and was performed in four stages: (1) Reading the data format, defining the coordinates on the map: Wradlib can be used to process weather radar data of volumetric formats from Rainbow Gematronik. e volumetric data from the radar was then processed and saved into a NetCDF format in Cartesian coordinates. is python programming library was developed by Postdam University and University of Stuttgart.
(2) Removal of clutter caused by nonmeteorological factors such as the presence of objects on the surface of the earth (mountains, hills, and tall buildings) or objects in the air (aircraft, birds, etc.) by clutter filters developed by Gabella and Notarpietro [12]. After that, attenuation correction commonly caused by radome (radar cover) is corrected using a method developed by Kraemer and Verworn [13]. (3) Gridding: e reflectivity data are displayed on each vertical slope angle of the radar automatically to calculate the value of Constant Altitude PPI (CAPPI), i.e., the horizontal radar reflectivity display at a certain fixed height, and also the maximum CAPPI (CAPPIMAX) value in altitude column. e CAPPI calculation specification has been designed to have a horizontal resolution of 0.5 km/pixel and a vertical resolution of 0.5 km from 0.5 to 10 km height column. (4) Conversion of reflectivity (dBZ) into rainfall intensity (mm/hr): estimated rainfall or QPE was derived from a common Z-R relationship of Marshall-Palmer with Z � 200R1.6 for general precipitation [11].

Tropical Wave Analysis.
Tropical wave analysis was conducted by investigating the intraseasonal synoptic-scale variation of atmospheric conditions during the event over Bengkulu Province. e investigated tropical waves include the Madden-Julian Oscillation (MJO), Equatorial Rossby (ER) wave, and Kelvin waves.

Event of Reconstruction.
Cumulonimbus clouds that caused extreme weather in the Bengkulu area were detected developing over the ocean. Rain clouds were moving from the west (Indian Ocean) to the east (Sumatra Islan) as shown in Figures 2(a) and 2(b). Based on the shape, extent, and life period, these rain clouds include the Mesoscale Convective System (MCS) phenomenon. e MCS phenomenon produces large clouds of Cumulonimbus (hundreds to thousands of km) with longer lifespans (more than three hours). At the time of this extreme rainfall, Cumulonimbus clouds reach more than 100 km with a lifespan of up to 18 hours in the area around Bengkulu. e greater the reflectivity value, the greater the intensity of the rainfall. Figure 2(a) showed some of the convective nucleus (Cumulonimbus cloud) stretching on the southwest coast of Bengkulu; the convective nucleus is characterized by high reflectivity values. A Cumulonimbus cloud is also capable of causing strong winds and the downburst process from cloud to surface so that the wave height increases in the Indian Ocean at West Bengkulu.
In Figure 2    circle on the radar image caused by the Cone of Silence region (a region not detected by radar imagery pans). Cone of Silence is located at the center of the radar due to the radar data emission that forms an angle that is not perpendicular to a radius of up to 3 km. e rain hit throughout the Bengkulu region with a total rainfall of at least 30 mm per day. In the northern coastal area of Bengkulu, the total rainfall reaches 60-90 mm per day. As for the central coastal area to the south of Bengkulu, the total rainfall reaches 120-210 mm per day.
Accumulated rainfall simulation results show that the phenomenon of MCS that caused the occurrence of extreme rain accompanied by strong winds hit the entire west coast of Bengkulu and concentrated in the central Bengkulu region,   From the graph in Figure 4, the village of Ngalam and Padang Pelasan in Seluma District was first washed down by the subsequent rain of the Kampung Bahari area in Bengkulu and Tanjung Raman in Central Bengkulu Regency. At Ngalam Village location within a 2-hour time span from 21 : 00 UTC to 23 : 00 UTC on September 19, 2017, accumulation of rainfall reaches 100 mm with intensity reaching approximately 50 mm per hour. Similarly, the intensity of rainfall in Kampung Bahari at 23 : 00 UTC to 00 UTC reached more than 50 mm per hour. As for the Padang Pelasan and Tanjung Raman areas, the average rainfall intensity is less than 25 mm per hour but rain falls continuously.

Trigger Factor.
On 19-20 September 2017, heavy rainfall triggered floods and landslides on the Indonesian island of Sumatra, particularly in Bengkulu Province (3.8°S; 102.26°E). e synoptic-scale climate variability was analyzed to determine the possible trigger factor, particularly the intraseasonal tropical waves such as Madden-Julian Oscillation (MJO), Equatorial Rossby Wave (ER), and Convectively Coupled Kelvin Wave (CCKW). During this period, the MJO was not active over the eastern Indian Ocean and Sumatra Island (red encircled in Figure 5). During an inactive or weak MJO, Bengkulu usually experiences less rainfall. In this case, MJO is not likely the main trigger for this extreme event.
e Equatorial Rossby wave was also not active over Bengkulu (black encircled in Figure 5) and did not contribute to the heavy rainfall event. e Scientific World Journal On shorter timescales, there was an active CCKW (Kelvin Wave) moving eastward faster than MJO and passed through the Bengkulu and the south Sumatra Island on 19-20 September 2017 (green encircled in Figure 5). is Kelvin wave had a very short period of about a few days over Bengkulu and enhanced the convection and generating MCSs. erefore, the Kelvin wave is likely the main factor triggered and contributed to the heavy rainfall over the Bengkulu during the period.

Conclusion
Limitations of surface rain intensity measuring devices led to measurement and analysis of extreme precipitation events causing floods and landslides in Bengkulu using QPE from C-Band Doppler weather radar. e results of atmospheric dynamic reconstruction on September 19-20, 2017, using CMAX radar products show that the MCS moving from the ocean to the mainland with the instability reaches 65 dBZ. QPE derived from Marsh-Palmer (MP) ZR relations calculated for one day from 19 September 2017 at 12 : 00 UTC to 20 September 2017 at 12 : 00 UTC shows the entire area of Bengkulu rain with varying intensity, where the highest total rainfall occurred in the waters of the western Indian Ocean Bengkulu to reach 390 mm per day which is 20-30 km from Bengkulu. Temporary QPE values using weather radar have been able to reconstruct the occurrence of extreme rainfall in the area of flood and landslide locations with daily rainfall accumulation of 236 mm per day in Kampung Bahari, 247 mm per day in Padang Pelasan, 217 mm per day in the Ngalam Village, and 176 mm per day in Tanjung Raman. e analysis of tropical waves suggests that only Kelvin waves were active and serve as a possible trigger factor, while the MJO and ER waves were inactive during the event over Bengkulu.
Data Availability e data supporting this article is provided within the article. e datasets generated and analyzed during the current study can be obtained from the corresponding author upon reasonable request.

Conflicts of Interest
e authors declare that they have no conflicts of interest.