Role of updraft in dry‐season torrential rainfall in Greater Jakarta, Indonesia

Coastal urban areas in tropical regions, such as Greater Jakarta (GJ) in Indonesia, are susceptible to flood hazards from torrential rainfall. Efforts to understand the convective mechanisms leading to this type of rainfall have been carried out extensively, especially for events occurring in the wet season. However, hydrometeorological risks persist even in the dry season, when the rainfall is more closely correlated with the diurnal wind circulation. Among the aspects of convective activities and this diurnal circulation, the details of the updraft remain the least studied and open for discussion. We investigate the role of the updraft in urban torrential rainfall over GJ, using data from a C‐band Doppler radar and reanalysis dataset during the dry season of 2018–2019. Employing back‐trajectory calculations, we locate updraft origins corresponding to the peak of daytime inland rainfall. Most updraft origins are localized at an NW–SE‐oriented front nearly parallel to the coastline on the low‐lying plain, suggesting a region of atmospheric destabilization favorable for convective activity. The dry‐season torrential rainfall over GJ may involve convective activity from this front, that from earlier updrafts originating above the densely populated area, and further enhancement coming from orographic contributions. Our findings suggest the important role of updrafts in the rainfall generation, providing insights into the convective mechanism over GJ.


| INTRODUCTION
Understanding the convective mechanism in coastal urban areas is important to provide insights for mitigating the hydrometeorological risks of rain and flooding.One potential cause of urban flooding is torrential rainfall, or rainfall with a rate of ≥10Ámm/h (World Meteorological Organization, 1992).Torrential rainfallproducing convective activity in tropical regions is influenced by surface heating (Pielke Sr, 2001).Local diurnal wind circulation should also be considered when investigating urban rainfall, as the wind transports heat and moisture from the sea, urban areas, and land uses, which contributes to convective activity (Araki et al., 2006;Lestari et al., 2019).In urban areas, heat and moisture fluxes within the urban boundary layer are affected by changes in land uses, and the wind orients and transports them from their origin (Weaver & Avissar, 2001;Zhang et al., 2018).The updraft, or upward-moving air, holds an important role in bringing heat and moisture from lower to higher altitudes (Houze, 2014), especially during daytime when surface heating is the most intense.Therefore, in an urban area where the rainfall and the corresponding hazards are related to the diurnal wind, it is interesting to examine the relationship between wind, particularly updraft, and torrential rainfall occurrences.Such a relationship may be stable over time and can provide useful insights into the general rainfall characteristics on a diurnal scale.
The Indonesian Maritime Continent (IMC) has been known as the region where the most intense regional rainfall is situated, with convective activities that are sensitive to coupled ocean-atmosphere variability on spatiotemporal scales, especially along its coastlines (e.g., Arakawa & Kitoh, 2005;Yamanaka et al., 2018).Daytime coastal convective activity is a complex interaction involving sea breeze, surface heating, atmospheric moisture, and in-cloud vertical velocity (Bergemann & Jakob, 2016).Within the IMC, the Greater Jakarta area (GJ) fits all the importance and urgency of studies on urban rainfall characteristics.As a densely populated coastal urban agglomeration in the tropical region, which includes the capital megacity of Jakarta, its urban rainfall characteristics follow a diurnal pattern governed by land-sea breeze circulation (Katsumata et al., 2018;Mori et al., 2018).Therefore, GJ regularly faces the risk of flooding during both wet and dry seasons (Lestari, Protat, et al., 2022).Despite these persistent risks, studies on urban rainfall in GJ are still relatively limited, with the existing ones mainly focusing on the wet season, a larger regional area, and/or a shorter period of time (e.g., Andarini & Purwaningsih, 2021;Lestari, Sapan, et al., 2022;Supari et al., 2012).This study aims to examine the role of updraft in urban coastal torrential rainfall development by focusing on its distinct diurnal characteristics during the dry season on a local scale using multi-year weather radar data and wind trajectory analyses.Our results may provide more insight into the convective mechanism over GJ.

| THE GREATER JAKARTA AREA
The GJ is Indonesia's most densely populated area, a rapidly growing urban agglomeration consisting of the Special Capital Region of Jakarta, five satellite cities (Bogor, Depok, Tangerang, South Tangerang, and Bekasi), and three regencies of Bogor, Tangerang, and Bekasi (Figures 1a and S1).Jakarta is a megacity consisting of five municipalities (West, North, Central, South, and East Jakarta).GJ is situated on the northern coast of Java Island, between the Java Sea in the north and the mountainous area in the south.GJ is considered active in localized convective cloud formation as rainfall occurrences persist in both wet and dry seasons (e.g., Ferdiansyah et al., 2020).Prolonged torrential rainfall is a potential trigger for flooding produced by the development of mesoscale convective systems (MCS; Nuryanto et al., 2021).In particular, the higher-elevation Bogor area has been known for high rainfall occurrences throughout the year (Nakamura et al., 1994), increasing the risk of flooding in the more densely populated lowlying areas.
The Indonesian Agency for Meteorology, Climatology, and Geophysics (BMKG) defines the wet season during December-March and the dry season during June-September, with transitional periods from April to May and October to November (Aldrian & Susanto, 2003).Each season is characterized by different wind activity of the trans-equatorial monsoon flow (Hattori et al., 2011;Sofyan et al., 2004), while the wet season is heavily influenced by interannual variabilities such as El Niño-Southern Oscillation (ENSO), Indian Ocean Dipole (IOD), and Madden-Julian Oscillation (MJO), which significantly affect wind behavior and rainfall variability in Indonesia (Alsepan & Minobe, 2020;Wu et al., 2007).These abundant activities may obscure the diurnal pattern of winds that can be observed at altitudes of <1000 masl (meters above sea level), characterized by a northerly sea breeze during the daytime and a southerly land breeze during the nighttime (Hadi et al., 2002;Lestari et al., 2019).

| Spatiotemporal dominant wind field from reanalysis datasets
We examine the wind characteristics using the ERA5 reanalysis dataset from the European Centre for Medium-Range Weather Forecasts with a 1-h temporal resolution, a 0.25 horizontal resolution, and vertically provided at various pressure levels (Hersbach et al., 2020), from which we obtain the temperature values and threedimensional wind velocity components: the zonal wind u, the meridional wind v, and the vertical velocity w during 2018-2019.We convert the altitudes from hPa to meters according to the standard atmospheric model and the corresponding vertical velocities to m/s unit following Uma et al. (2021).
We transform the wind velocity vectors u, v, w ð Þ to spherical coordinate vectors s, θ, ϕ ð Þ consisting of wind speed s, the azimuth θ, and the vertical angle ϕ.These spherical wind vectors are used to examine the seasonal and diurnal wind characteristics and determine the dry season duration.We summarize the diurnal wind cycle by computing the average wind speed, the median azimuth, and the average vertical angle for each hour of the day at each grid point during the dry season.The result is a spatiotemporal "dominant" wind field representing the diurnal wind characteristics during the dry season of 2018-2019.
F I G U R E 2 Hourly maps of dry-season average cumulative rainfall in the GJ area.

| Rainfall estimation from radar reflectivity data
We obtained radar reflectivity data during 2018-2019 from a single-polarization C-band Doppler weather radar operated by BMKG at Tangerang City (Figure 1a).The radar provides conical scans every 8 min within a 200 km radius at nine elevation angles from 0.5 to 19.5 .The reflectivity data have been subjected to preprocessing steps using the Python-based wradlib library (Heistermann et al., 2013), corrected for attenuation following Jacobi et al. (2011), and converted to a geographical grid with a 500 m resolution (Permana et al., 2019).Estimating rainfall rate R from radar reflectivity Z involves the relationship Z ¼ aR b , where the constants a and b may vary locally (e.g., Marshall & Palmer, 1948;Rosenfeld et al., 1993).We compare the reflectivity values at the 2.4 elevation angle with measurements from six rain gauges located within GJ following Hutapea et al. (2020) and obtain an optimum relationship of Z ¼ 50R 1:5  (Figure 1a and S2).We exclude lower elevation angles due to the effects of topographic blocking (Morin et al., 2003) and electromagnetic interference (Figure S3).
To summarize the diurnal cycle of the rainfall, we estimate the rainfall rate from radar reflectivity, compute the hourly cumulative rainfall, and average them hourly throughout the dry season.The available number of radar observations may vary from the typical seven observations per hour, due to the removal of low-quality and erroneous observation data.Seasonal averaging introduces insensitivity to these variations while enhancing the spatial and diurnal pattern of the rainfall.In this study, we only focus on the dry-season averaged hourly cumulative rainfall over GJ (Figure 2).

| Wind back-trajectory analysis
From each location where the rainfall occurs at an initial altitude, we track the three-dimensional wind trajectory backward in time (back-trajectory) to the location when the updraft started.We adopt the advection calculation (Draxler & Hess, 1998;Stein et al., 2015) in a spatiotemporal dominant wind field.For each time instant t with a time interval of Δt, we estimate the previous location in time d t À Δt ð Þfrom a known point d t ð Þ as: where Þ are the wind velocity vector at d t ð Þ and the initial guess of upwind location d 0 , respectively, linearly interpolated from the dominant wind field.We iterate the above procedure to find the time and location when the trajectory is at its lowest altitude, representing the transition from downdraft to updraft before the rainfall occurrence, referred to as the "updraft origin."If no transition is found, we exclude the back-trajectory.Because the advection distance per time step should be <75% of the meteorological grid spacing (Draxler & Hess, 1998), which is 0.25 in our case, we use Δt ¼ 1 min, which provides detailed back-trajectories.The starting time for tracking will be determined from the rainfall analysis.Encounter with topography results in advection along the surface in time-forward tracking (Draxler & Hess, 1998), which is inaccurate in our timebackward tracking.Therefore, back-trajectories that encounter topography are excluded.We use topography data from the Indonesian Geospatial Information Agency (BIG) with an 8-m spatial resolution.
We estimate the back-trajectories from an initial altitude based on the cloud base height, considering that convective processes tend to occur in the lower portion of the clouds (Houze, 2014).We compute the dry-season average of the cloud base height from the ERA5 dataset during the starting time, which shows an altitude range of 1000-2500 masl over GJ (Figure S4) comparable to previous studies (Araki et al., 2006;Hashiguchi et al., 1996;Tjasyono et al., 2005).We linearly interpolate the initial altitudes for tracking from these seasonal average values.
Back-trajectories are calculated using UTM coordinates (zone 48, southern hemisphere).All time information is presented in Western Indonesian Time (UTC + 7).

| Diurnal cycle of wind and rainfall
Wind analysis at >1000 masl altitudes (Figure 1a) reveals monsoonal westerly winds during December-March (Figure 1b) and easterly winds during April-November (Figure 1c).This is consistent with the duration of wet and dry seasons from BMKG, respectively, but with easterly winds also present during both transitional periods, possibly due to the varying influence of interannual variabilities (Yamanaka, 2016).Therefore, we define the dry season as April-November.At <1000 masl altitudes, the monsoonal wind during the wet season heavily influences the sea breeze (Figure 1d), while the north-south orientation of the land-sea breeze system dominates during the dry season (Figure 1e).We focus on the dry season due to this stable and clear diurnal behavior, with the expected minimum influence of interannual variabilities on rainfall characteristics over GJ.
Our dominant wind field suggests the land breeze to sea breeze transition, as well as the earliest downdraft to updraft transition, occur at around 09:00 (Figure 1f,g), while the opposite occurs around 21:00 until midnight.These transition times vary due to different surface heating and cooling rates at different elevations.Lower wind speeds during land breeze (<1Ám/s) often cause difficulties in extracting their dominant direction (Figure 1f).At <1000 masl altitudes, the coastal and NE parts of GJ are the most affected by the monsoonal wind, as the easterly wind gradually changes to a north-south orientation when penetrating the GJ area.
The diurnal cycle of rainfall from radar observation confirms its relationship to the land-sea breeze cycle.Dryseason average rainfall distribution maps show the common traits of tropical coastal convective systems (Figure 2).Nighttime rainfall of <3Ámm/h over the sea and coastal areas occurs during the land breeze (21:00-08:00).Daytime rainfall formation and inland updraft start around 09:00 during the transition to the sea breeze.Widespread occurrence of >3Ámm/h rainfall starts around 11:00-13:00 as a sea-breeze front (Ferdiansyah et al., 2020), a west-east-oriented localized convective system near the coastal area over Tangerang and Jakarta.Meanwhile, orographic rainfall also starts to intensify during this time.The rainfall then rapidly intensifies over the area between the sea-breeze front and orographic rainfall.Torrential rainfall is observed over Bogor during 15:00-18:00, with its peak around 16:00 typical for inland rainfall in the IMC (Arakawa & Kitoh, 2005).Inland rainfall dissipates during 18:00-21:00 and gradually transitions toward the coastal area, repeating the diurnal cycle (e.g., Yamanaka et al., 2018).
The intense inland daytime convective processes and the torrential rainfall over higher elevations increase the flooding risk in densely populated areas along lower elevations.Next, we examine the updraft origin of the rainfall peak at 16:00 to explain the inland rainfall generation mechanism that may aid in reducing future flood risk.

| Relationship of updraft origin and rainfall
We locate the updraft origins from 42,831 radar data points with a dry-season average rainfall of >0.5Ámm/h during 16:00 over GJ (Figure 2).To ensure capturing the time of updraft start, we first calculate the back-trajectories for 12Áh, longer than the typical MCS lifetime (Nuryanto et al., 2021).From the updraft origin to the rainfall occurrence, the back-trajectories confirm that easterly winds from along the coastline east of GJ transition into northerly winds as they penetrate into GJ (the example of 0.5% of back-trajectories is shown in Figure 3a).
We examine the spatial distribution of the updraft origins using quartic kernel density estimation with a 5Ákm radius (Figure 3b), which reveals a strikingly linear feature of an NW-SE-oriented front of dense updraft origin, elongated from over Jakarta Bay to Indramayu regency.Vertically, the altitude of updraft origin increases with time (Figure 3c), with this front representing the horizontally stable updraft start during 10:00-13:00, only varying in altitude at 800-1700 masl.
We further characterize the updraft origin using the corresponding rainfall values (Figure 4a,b), time (Figure 4c), altitude (Figure 4d), and back-trajectory length from the origin to rainfall occurrence (Figure 4e).Before 10:00, the updrafts mostly originated at lower altitudes of 200-800 masl over areas between the NW-SE front and Bogor, particularly the megacity and peripheral densely populated areas, as well as the more inland parts of Karawang and Subang regencies.They are associated with >2Ámm/h rainfall over the southern half of GJ (mainly the Bogor area).
After 13:00, the updraft started at higher altitudes of 1700-3000 masl north of the NW-SE front in the coastal area over Bekasi and Karawang, heavily affected by easterly winds and spatiotemporally closer to the corresponding lower rainfall rates over the low-lying and coastal areas of GJ, including those over Jakarta Bay and northern Karawang.Updraft origins between 10:00 and 13:00 along the front seemingly correspond to the transition between coastal and inland rainfall over southern Jakarta, Tangerang, and Bekasi.
Despite the trend of increasing updraft origin altitude with time from the southern to the northern parts of GJ, the lengths of wind trajectories (Figure 4e) show 40-60Ákm-long back-trajectories from over the megacity, Tangerang, and Bekasi associated with more intense rainfall over Bogor.Shorter trajectories correspond to less intense rainfall over the megacity and coastal areas of GJ, while longer trajectories (>60Ákm) from further east over Subang and Indramayu correspond with low rainfall rates over Karawang and western Bekasi.
Finally, we summarize the relationships between updraft origin characteristics and the dry-season average rainfall at 16:00.The correlation between the starting time and altitude of the updraft is approximately linear and positive (Figure 5a).These altitudes uniformly increase with time before 11:00, after which different increase rates are observed north of the NW-SE front over the Subang-Indramayu area, Karawang, and Jakarta Bay.Torrential rainfall is associated with updrafts originating at 9:00-10:00 (Figure 5b) from 300 to 500 masl altitudes (Figure 5c) with 40-60 km-long trajectories (Figure 5d) found over densely populated areas in southern Jakarta, Depok, and around Bekasi City.These updraft altitudes for torrential rainfall are consistent with those for sea breeze circulation detected from the L-band boundary layer radar observation at South Tangerang (Hadi et al., 2000).We note that there are two peaks of torrential rainfall, which will be discussed in the following section.

| Convective mechanism and torrential rainfall
The positive correlation between the updraft starting time and altitude can be explained by differential surface heating and atmospheric pressure between land and sea, influenced by sea surface temperature over the Java Sea (Yamanaka et al., 2018).As the land heats faster, the increase in updraft starting altitude over the Jakarta Bay area is slower than that over Karawang and further east (Figure 5a), highlighting different heating characteristics for different areas.
The relationship between the updraft origin and the daytime torrential rainfall suggests the significant role of the NW-SE front in the rainfall generation over GJ.Surface heating may also control the horizontal stability of the updraft origin along the front, during which only its altitude continuously increases.We interpret the NW-SE front as a zone where colder, faster easterly winds from the Java Sea east of GJ collide with warmer, slower inland air parcels as surface heating progresses, forcing the warmer air to rise.This condition induces localized convection, promoting conditions for successive, rapid convective cloud growths (Bergemann & Jakob, 2016).Northeasterly winds along the coastal area then push these clouds southward, which may be responsible for the sea-breeze front around 11:00-13:00 and later rainfall over the northern part of GJ (Figure 2).
Updrafts before 10:00 from south of the NW-SE front during the beginning of the sea breeze and surface heating are associated with slower winds (e.g., Argüeso et al., 2016), transporting heat and moisture from lower altitudes over the urban area southward while picking up more moisture from the transpiration process of the abundant vegetation in the southern part of GJ.Longer trajectories provide opportunities for localized convective cell formation (Houze, 2014), which is also induced by morning low-level wind shear around the coastal and western parts of GJ (Nuryanto et al., 2018).Updrafts after 13:00 in the coastal area north of the NW-SE front occur under higher intensities of surface heating and land-sea temperature discrepancy, promoting more buoyancy that destabilizes the atmosphere, thus elevating the boundary layer height while promoting convective activity at higher altitudes, subsequently raising the cloud base height.This strong buoyancy produces heavy precipitation particles which plummet swiftly over a short distance (Houze, 2014) through the easterly winds.Therefore, the resulting rainfall cannot occur deeper within GJ.
As the day progresses, the coastal rainfall reduces the inland moisture supply (Ogino et al., 2017), while contributions from earlier updrafts become more dominant during the inland convective process.We suggest that the inland rainfall generation is a combination of the convective process from the long, slow transport of heat and moisture starting around 09:00 and the successive growth of clouds around the remnants of the sea-breeze front after 13:00 (e.g., Bergemann & Jakob, 2016), mainly involving contributions from the NW-SE front.However, the intense torrential rainfall occurrence may also involve other inland contributions.Orographic factor (Lestari, Protat, et al., 2022) may have caused the enhancement of the torrential rainfall peak over Bogor and another more enhanced peak closer to the mountain (Figure 4a), which was improperly associated with shorter wind trajectories (Figure 5d).As the precipitating inland clouds propagate southward, they are constrained by the mountains and interact with existing orographic clouds.The "seeder-feeder" mechanism of moisture gathering may occur, intensely enhancing the rainfall (Figure S5; Houze, 2014).
Meanwhile, longer wind trajectories from further east of GJ do not correlate with higher rainfall because they carry smaller quantities of moisture, or the moisture is insufficient to develop and sustain inland rainfall in GJ as the wind may have depleted its moisture into a more local coastal rainfall (e.g., Nuryanto et al., 2021;Ogino et al., 2017).
There is an apparent agreement between updrafts for the inland rainfall originating around the most densely populated area of GJ and previous studies that found a positive effect of urban land uses on rainfall through the convergence between sea breezes and urban heat islands (Kusaka, 2008;Kusaka et al., 2014;Shepherd et al., 2002).However, more dedicated investigations are necessary to study such an effect and will be left for future studies.

| CONCLUSIONS
We studied the role of updraft in the convective mechanism of dry-season daytime rainfall peak during 2018-2019 in Greater Jakarta (GJ), using weather radar data, reanalysis datasets, and back-trajectory analysis.We found a dense distribution of updraft origins along an NW-SE front separating the updrafts for inland and coastal rainfalls, where the altitude of updraft origin increases with time.The inland rainfall is generated by convective activities from updrafts over urban areas, collecting heat and moisture along their trajectory, and those following the sea-breeze front, driven by winds penetrating inland.Meanwhile, the torrential rainfall may involve additional enhancements from orographic factor.Lower rainfall rates over coastal and urban areas are caused by higher and nearby updrafts around the NW-SE front, where monsoonal winds meet the warmer air due to surface heating, promoting atmospheric instability and convective activity.An apparent relationship between the updraft origin over the densely populated area and the intense inland rainfall encourages further studies on the effect of urbanization on the convective mechanism.

F
I G U R E 1 (a) Topographical map of the study area showing regency/city divisions (black lines), the Greater Jakarta boundary (red line), and the location of weather radar and rain gauges.Red circle denotes an ERA5 grid point examined in (b)-(g).(b-e) Normalized distributions of horizontal wind direction versus wind speed at different seasons and altitudes.(f) Dry-season diurnal cycle of the horizontal direction of dominant wind field.Low wind speeds during land breezes hinder dominant direction calculations.(g) Diurnal cycle of dominant vertical wind velocity.

F
I G U R E 3 (a) Example of 0.5% of 42,831 wind back-trajectories from updraft origin to their corresponding rainfall observation point at 16:00.Altitude axes are exaggerated.(b) Density map of the obtained updraft origins.(c) Vertical distribution of updraft origins perpendicular to the magenta line in (b).

F
I G U R E 4 (a) Initial rainfall points (>0.5Ámm/h) for back-trajectory analysis and (b-e) the rainfall and analysis results shown at the obtained updraft origin points.

F
I G U R E 5 Comparison of values obtained from back-trajectory analysis and dry-season cumulative rainfall at 16:00.