Roles of Land and Topography on Propagating Convective Systems During the Heavy Rainfall Event of the 2002 Jakarta Flood, Indonesia

: The movement direction of propagating convective systems originating from both inland and offshore over the north coast of West Java in Indonesia is determined primarily by the prevailing wind. However, the role of a land-sea contrast and a rugged topography over southern West Java is also expected to affect propagating convective systems by increasing land-sea breezes and enhancing upward motion. These hypotheses are tested using a weather prediction model incorpo-rating convection (up to 3 km height) to simulate the heavy rainfall event during 26–29 January associated with the 2002 Jakarta flood. First, we addressed the influence of land-sea contrast and topography on the local circulation, particularly in the area surrounding Jakarta, by replacing the inland topography over western Indonesia (96°–119°E, 17°S–0°) with a water body with an altitude of 0 m. We then compared the results of model simulations with and without topography. The results show that the main role of the topography here is enhancing the upward motion and generating a deep convective cloud in response to the land-based convective system during 26–27 January 2002, which then continuously and rapidly propagates offshore due to the cold pool mechanism. Furthermore, the land-sea contrast has a significant role in increasing sea breeze under the rapidness of the landward propagation system during 28–29 January 2002, which was strengthened by the gravity waves and resulted in early morning convection over coastal regions.


Introduction
Heavy rainfall over the north coast of West Java has triggered severe floods in Jakarta, Indonesia [1][2][3][4][5]. These extreme events associated with landward-as well as offshore-propagating convective systems have previously been found to be controlled by prevailing background winds [6][7]. However, the effects of land and topography on this propagating convective system are not fully understood. On a regional scale, a previous study suggested that the sea breeze convergence is strengthened by mountain-valley winds, which enhance precipitation over inland Java [8]. [8]. However, no previous studies have investigated the effects of mountainous regions in southern West Java on coastal precipitation associated with extreme flooding events in Jakarta. On the other hand, several previous studies shared a clue about the role of land through Island rainfall enhancement mechanism that mainly creates significant diurnal rainfall [9][10][11][12] over the Maritime Continent [8,[13][14][15]. This mechanism shows that a localize buoyancy oscillation modulated by land-sea contrast can produce the mean circulation [12,[16][17][18], thus, may trigger extreme precipitation associated with several floods over the Maritime Continent (i.e., the Greater Jakarta).
In the past four decades, the capital city of Jakarta has experienced floods more frequently during the active period of the Asian monsoon (January-February). However, a severe flood occurred in Jakarta from 29 January to 10 February 2002, reported as the worst flood in the last three decades  in terms of its impact. Given that offshore (26-27 January 2002)-as well as landward (29-30 January 2002)-propagating convective systems may produce extreme precipitation (see: Fig. 1b, d), this event is an ideal case study to investigate the role of land-sea contrast and topography on offshore as well as landward propagating convective systems using a numerical model. In this case, satellite observational data could well-captured the propagation signal; however, it recorded lower precipitation amounts over Jakarta than station observations from 29 to 30 January 2002 (Fig. 1).
Of note, a cyclonic vortex that developed over the Indian Ocean near southwest Java created a predominantly meridional monsoonal flow that produced near-surface flows with a meridional orientation over West Java (Fig. 1). In this case, we hypothesize that the predominant northerly wind may interacts with a land by topographically elevated and complex area in southern West Java. To further investigate the role of land and topography in southern West Java on modulating precipitation over its north coast (including Jakarta), we conducted numerical modeling, based on a removing land and flattening topography approach using high-resolution numerical weather predictions from the Weather Research and Forecasting (WRF) model that incorporates convection up to 3 km height. In the following sections, we explain the model set-up and experimental design, and discuss the results of the modeling.

Model Set-up
The numerical experiments were simulated using WRF model version 4 To further investigate the heavy rainfall event, we validated the simulation results with ground-based observational data from the Agency for Meteorology, Climatology, and Geophysics (BMKG) station [21] and TRMM Multi-Satellite Precipitation Analysis (TMPA) Real Time 3B41RT (TMPA-RT). The TMPA-RT data are calibrated by rainfall gauge data and have been used in several extreme rainfall studies over tropical regions and the Maritime Continent [6,[22][23][24][25][26]. Furthermore, to identify the synoptic conditions related to background winds, we used high-resolution, near-surface ERA5 data from the European Center for Medium-Range Weather Forecasts (ECMWF), with a spatial resolution of 0.25˚ x 0.25˚ [27].

Experimental Design
To investigate the effect of land and topography on a propagating convective system, we used an experimental design (Fig. 2b) involving two scenarios: (1) the actual topographic conditions as the control simulation (CTL); and (2) a modified topography that is flattened (altitude = 0 m) and replaced with the sea (i.e., a water body), which is the scenario simulation (SCE). The CTL and SCE scenario simulations were applied to all domains, focusing on domain 3 (D03). Both experiments include the outermost domain, which captures the effects of regional processes (e.g., wind circulation and sea surface temperature) on the innermost domain circulation. We further analyze the role of land and topography through two selected episodes of propagation: offshore (26-27 January 2002) and landward (28-29 January 2002) propagations of convective systems.

Heavy Rainfall Observed and Background Conditions
Rainfall recorded by four meteorology stations in Jakarta and surrounding areas revealed a significant increase from 26 to 30 January 2002 (Fig. 1b). Moreover, heavy rainfall were associated with a cold anomaly after the Borneo Vortex had dissipated [2].
One interesting factor is that the offshore propagation of the convective system was in a direction against the prevailing wind. The interaction between the predominant nor-

Effect of Land and Topography on Propagating Convective Systems
The effects of land and topography on a propagating convective system are shown in Fig. 3. The presence of a land-based, as well as an oceanic convective system, produced maximum precipitation on 26 January 2002 during the afternoon and at night, respectively Interestingly, early morning precipitation over the coastal region also became persistent during 28-30 January 2002 in the CTL simulation, whereas the SCE simulation only captured the landward propagation of the convective system on 27 January (Fig. 3b-c).
Consequently, the difference (i.e., by subtracting) between the CTL and SCE simulations reveals large amounts of rainfall over the land and ocean on those days, apart from the afternoon rainfall on 28 January 2002 (Fig. 3d).
This vigorous rainfall depicted in the CTL is related to the general diurnal rainfall pattern over the Maritime Continent under the Island rainfall enhancement hypothesis [8,11,13]. Thus, previous studies also mentioned that removing islands leads to more impact in reducing rainfall than flattening orography over the western Maritime Continent (Sumatra and Kalimantan Islands) [15]. In this study case, both offshore (26-27 January 2002) as well as landward (29-30 January 2002) propagation of rainfall (Fig. 3b,d) suggested as an implication of removing land and flattening topography treatment in the SCE. We now consider the effect of topography on the maximum precipitation obtained from the CTL simulation on 26-28 January, given the qualitative agreement between the model results and observational data, which was also confirmed by the spatial rainfall pattern (Fig. 4a-c). The potential temperatures in the CTL and SCE simulations over West Java are similar (Fig. 4d-f).
However, there are slight differences in humidity between the northern and southern parts of West Java, which indicate the topographic effect of the mountain range might have formed dry (moist) areas in the northern (southern) regions (Fig. 4f). This result is consistent with a previous finding that the topography created upward and downward motions due to the large-scale circulation (i.e., Madden-Julian oscillation), and thus a quasi-stationary convective system near major topographic features over the Maritime Continent [28].  coastal regions, which occurred in the morning of 29 January 2002 (01:00-06:00 LT). Therefore, we now discuss the physical and dynamic processes related to the effect of topography on: (1) propagating convective systems; and (2) early morning convection. Figure 5a and d shows that strong offshore propagation occurred from 13:00 to 00:00 LT, over a distance of >110 km. The offshore propagation did not occur in the SCE simulation ( Fig. 5b and e). During 07:00-18:00 LT, the convective system was concentrated over the mountainous region, whereas during 19:00-00:00 SLT the system moved offshore (Fig.   5c and f). This strong offshore propagation is consistent with a previous study that found offshore-propagating convection from inland areas reaches a maximum in daytime [29].  To further understand the physical processes that caused the offshore propagation during 26-27 January 2002, we analyzed the hourly evolution of cloud water vapor, equivalent potential temperature, and meridional vertical wind vector. On 26 January, the landbased convective system formed from 13:00 LT over the southern part of the mountainous area and propagated northward to the coastal area (18:00 LT) (Fig. 7).
Notably, the cold pool created by the decaying convective cloud over the mountainous region travelled downward and triggered the formation of new convective cloud over the coastal region. In this case, the presence of the cold pool is evident from the lower equivalent potential temperature over the near-surface (<1 km) (Fig. 7c). This is consistent with a previous study that showed offshore-propagating systems are driven by a surface cold pool [7,29] and the squall line mechanism over the eastern Maritime Continent [30]. The development of this thunderstorm system associated with the leading edge of a propagating surface cold pool is consistent with the results of the present study. This distinctive offshore propagation also occurred on the following day on 27 January 2002 (Fig.   8). A land-based convective system developed over the mountainous area from 13:00 LT and strengthened until 18:00 LT (Fig. 8a).
However, the cold pool that formed from the decay of the convective cloud system appeared at 16:00 LT over the southern mountainous region. The cold pool created a new convective cloud system over the southern mountainous region and forced the existing convective cloud system to move northward. Interestingly, several convective cells connected the mountain and sea regions at 19:00 LT, and then the convective system over the mountainous region diminished at 20:00-00:00 LT (Fig. 8a).
A cold pool that developed from the decaying cloud over the coastal region created a new convective cell over the ocean at 22:00 LT, which propagated landward due to the strong northerly background winds ( Fig. 8a and c). In this case, we conclude that the interaction between the predominant prevailing wind and topography, in association with the cold pool mechanism, controlled the propagation direction of the convective system.
However, this was not the case for the SCE simulation (Fig. 8b). This rapid landward propagation of the oceanic convective system triggered a deep convective system over the mountainous region during the early morning on the following day (Fig. 9a). For the SCE simulation, early morning convection also occurred and appears to be strongly connected to the convective system over the ocean (Fig. 9b). In such a case, the early morning convection may also be affected by a gravity wave, which was further examined for the following day (Fig. 9).

Preprints
The SCE simulation results (Figs 7-9) suggest that the flattened topography does not significantly enhance northerly winds over the ocean or produce enhanced precipitation.
This result is consistent with a previous study that reported greater surface roughness on a flat island has an influence on a high-wind regime [31]. Thus, forcing caused by the thermal contrast between the land and sea is the main factor [31] controlling the diurnal cycle of precipitation over the Maritime Continent [32].

Effect of Gravity Wave on Early Morning Precipitation
Another interesting feature that can be explained by the simulation results is the early morning precipitation over the coastal region on 28-30 January 2002 (see Fig. 6). A previous study noted that the early morning precipitation peak over the north coast of West Java is strongly related to extreme precipitation [6]. We found that the early morning precipitation peak is clearly evident on 28-29 January 2002 (Figs 9-10). In the case of 29 January 2002, initial convection over the coastal region occurred suddenly at night (22:00 LT) and rapidly propagated landward (02:00 LT).
The initiation of this convection over the coastal region is difficult to explain solely in terms of the prevailing northerly wind. We found a wave-like structure expressed as a negative equivalent potential temperature anomaly at ~500-700 hPa which correspond to recently study [29]. The layered structure of the anomaly was disrupted when initial convection developed at 22:00 LT. This is consistent with a study [30] that suggested the strong prevailing wind could create strong asymmetry between the windward and leeward areas, thus generating gravity waves via a thermal contrast, as well as an upward deviation of the horizontal flow in response to the enhanced diurnal heating mechanism.
Our proposed mechanism for early morning convection and precipitation over the north coast of West Java involving meridionally propagating gravity waves is consistent with recent studies [29], and is also supported by previous studies of offshore propagation of convective systems due to gravity waves [30][31][32][33][34][35]. Topographically induced gravity waves have a larger amplitude in the windward direction and descend on leeward slopes [31]. The topography produces low-level circulation that is strengthened by the height of the topography. This possible mechanism occurred persistently during the three days of heavy rainfall over the coastal region during the severe 2002 Jakarta flood.

Conclusions
We investigated the role of land and topography on propagating convective systems, based on a case study of the heavy rainfall event during 26-29 January 2002 that produced a severe flood in Jakarta, using high-resolution (3 km system. This numerical experiment shows that the simulated rainfall is qualitatively consistent with satellite and station data. In this case, synoptic weather conditions were influenced by a cyclonic vortex off the south coast of West Java, which enhanced the nearsurface prevailing northerly wind.
We found that during 26-27 January 2002, the rugged topography over southern West Java significantly enhances upward convective motion over the mountainous region. Hence, the dominant role of the topography over western Java is producing offshore propagation. These results are consistent with a previous study [24] that showed diurnal and topographic forcing led to ~20% rainfall enhancement, as reflected by offshore-propagating modes of convection. The topography of southern West Java drives offshore propagation due to the cold pool mechanism, which is in a direction opposite to the prevailing northerly wind. The cold pool also triggers thunderstorms and allows convective cells to coalesce and separate into new cells. With the prevailing northerly winds strengthening, the new convective cells tend to propagate landward, leading to maximum precipitation over the coastal and mountainous regions. This cold pool has previously been identified from night-time land surface cooling and a low-level flow anomaly near the leading edge of the cold anomaly [25].
We also found that the land-sea contrast may have a role in a strong landward propagation during 28-29 January 2002, which is strengthened by wave-like activity associated with a meridionally propagating gravity wave. This study agrees with a previous study [26] that showed near-surface cooling induced by enhanced rainfall generated gravity waves, which then caused an offshore-propagating diurnal cycle of precipitation. Furthermore, our results demonstrate that disruption of the gravity wave coincides with initial convection over the coastal region in the early morning, which is consistent with the previous study [7,21]. This result also corresponds to the early morning precipitation peak related to extreme precipitation and strong landward and offshore propagation of a convective system. In addition, the landward propagation is consistent with a previous study that suggested this results in the timing of peak precipitation being more random [6,25].
However, the early morning precipitation peak identified in the present case study was also a feature of the 2007 Jakarta flood, which had a semi-diurnal pattern superimposed on the diurnal cycle, with the largest precipitation peak during the early morning Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 25 November 2021 doi:10.20944/preprints202111.0469.v1 [2]. Thus, our study indicates a slight chance that the same process will occur again in the future despite the interaction between the topography and the northern coastal area of Java needs to be associated with a similar combination of synoptic and mesoscale dynamics.