In this study, the extreme rainfall event on 2 June 2017 along the northern coast of Taiwan is studied from a modeling perspective. While a peak amount of 645 mm was observed, two 1 km experiments produced about 400 and 541 mm, respectively, using different initial and boundary conditions, and thus are compared to isolate the key reasons for a higher total amount in the second run. While the conditions in the frontal intensity and its slow movement are similar in both runs, the frontal rainband remains stationary for a long period in this second run due to a frontal disturbance that acts to enhance the prefrontal southwesterly flow and focuses its convergence with the postfrontal flow right across the coastline. Identified as the key difference, this low-pressure disturbance is supported by the observation, and without it in the first run, multiple slow-moving rainbands pass through the coastal region and produce more widely spread but less concentrated rainfall, resulting in the lower peak amount by comparison.
To explore and test the effects of Taiwan's topography in this event, two additional 1 km runs are also used. It is found that the removal of the terrain in northern Taiwan allowed the postfrontal cold air to move more inland and the rainfall became less concentrated, in agreement with a recent study. Also, when the entire island topography of Taiwan is removed, the result showed significant differences. In this case, the blocking and deflecting effects on the prefrontal flow are absent, and the heavy rainfall in northern Taiwan does not occur.
During the transition period from the northeastern to southwestern monsoon, there exists an early summer rainy season in many regions in East Asia, including China, Taiwan, Japan, and Korea (e.g., Lau et al., 1988; Ding, 1992; Chen, 2004; Ding and Chan, 2005). Known as the Mei-yu (plum rain) season in Taiwan, it is defined as May and June by the Central Weather Bureau (CWB). During this season, the Mei-yu front often forms repeatedly between the cold continental and warm maritime air masses (located in China and the subtropical western North Pacific, respectively) and moves in to affect the region and subsequently becomes stationary in order to bring about the continuous rainy conditions (Chen and Chi, 1980; Chen, 1992; Ding and Chan, 2005). When a Mei-yu front develops or approaches Taiwan, the horizontal pressure gradient can strengthen at times, and the wind speed south of the front increases to form a low-level jet (LLJ; e.g., Chen et al., 1994; Chen and Chen, 1995). Coming from the southwest, the LLJ can often transport the moist and unstable air toward Taiwan and the Mei-yu front from the upstream ocean, and thus has long been recognized as an important feature for causing heavy rainfall in Taiwan in the literature (e.g., Chen and Yu, 1988; Kuo and Chen, 1990; Chen et al., 2005, 2008). Under such conditions, organized mesoscale convective systems (MCSs) such as squall lines (Rotunno et al., 1988; Houze et al., 1989; Lin et al., 1990; Jou and Deng, 1992; Chen and Chou, 1993; Wang et al., 2014a) can develop near the front and make landfall in Taiwan (e.g., Kuo and Chen, 1990; Chen, 1992, 2004; Chen et al., 1998; Wang et al., 2011; Xu et al., 2012). The steep topography of Taiwan, with the highest peak reaching almost 4 km (Fig. 1a), also acts to enhance the convection or trigger new convection by forced uplift (e.g., Kuo and Chen, 1990; Nagata and Ogura, 1991; Lin, 1993; Lin et al., 2001; Wang et al., 2022b). Therefore, the long-term Mei-yu rainfall climatology shows two prominent centers over the mountain interiors in central and southern Taiwan (Chi, 2006; Wang et al., 2022b), and they even also appear in total rainfall in most individual seasons (e.g., Yeh and Chen, 1998; Chien and Jou, 2004; Wang et al., 2017, 2022a). A third but less pronounced center (Chi, 2006) appears in northern Taiwan (and in some seasons) and is typically associated with the Mei-yu front.
In addition to forced uplift, the topography of Taiwan also has thermodynamic effects and is an important contributor to the island circulation and diurnal cycle of rainfall during the Mei-yu season (e.g., Akaeda et al., 1995; Chen et al., 1999; Kerns et al., 2010; Johnson, 2011; Ruppert et al., 2013; Wang et al., 2014b, 2022b). It also exerts a significant blocking effect on the oncoming prevailing environmental flow (e.g., Pierrehumbert and Wyman, 1985; Banta, 1990; Yeh and Chen, 2002; Wang et al., 2005). In the latter situation, Yeh and Chen (2003) suggested that the deflection of southwesterly flow by the Central Mountain Range (CMR) of Taiwan can often produce a local barrier jet (BJ) off the northwestern coast of the island (also Li and Chen, 1998; Yeh and Chen, 2002). The low-level convergence induced by this BJ can lead to heavy rainfall in the area (Yeh and Chen, 2003) when a frontal rainband also arrives, thus contributing to the third rainfall center in northern Taiwan.
The CWB surface weather charts, overlaid with the NAVGEM
925 hPa flow field, surrounding Taiwan every 6 h from
The CWB upper-air charts surrounding Taiwan at
During the past two decades or so, only two events reached 500 mm in 24 h (defined as “extremely torrential rainfall” by the CWB) in northern Taiwan in the Mei-yu season, on 11–12 June 2012 and on 2 June 2017, respectively. With serious flooding in or close to the populous Taipei metropolitan area (Fig. 1b), each event caused severe property damage and economic loss, and thus demand particular attention from the research community. In the 11–12 June event in 2012, Taipei received a peak rainfall of 510 mm in 24 h, caused by two successive rainbands overnight: a prefrontal squall line and a stationary rainband that formed in the northern Taiwan Strait and extended into northern Taiwan (Wang et al., 2016). While each rainband lasted for about 6 h, the second one was studied in detail and found to form ahead (south) of the surface front over the northern Taiwan Strait, along the convergence zone between the southwesterly flow deflected by the topography and the unblocked west-southwesterly flow farther offshore in the environment (Wang et al., 2016). Contributing toward the vigor of the convection and thus total rainfall, the back-building process occurring inside the rainband without the presence of the cold pool was also studied. On the other hand, Chen et al. (2018) emphasized the high moisture content and moisture flux inside the marine boundary layer in this event. In their sensitivity test with the topography of Taiwan removed, the BJ did not form offshore of northwestern Taiwan (also Ke et al., 2019). Consequently, without the rainband between the BJ and the environmental flow, only a fraction of the observed rainfall was produced in northern Taiwan (Chen et al., 2018).
In the second event on 2 June 2017, the rainfall amount was even higher (645 mm in 24 h) and was maximized along the coast at the northern tip of Taiwan,
caused by a single intense, quasi-stationary and long-lasting rainband
along the Mei-yu front (more details in Sect. 3). This event was
responsible for much of the local economic loss of around USD 9 million in
the 2017 season (Huang et al., 2019). For this event, Wang et al. (2021,
hereafter referred to as WLC21) performed an ensemble-based sensitivity
analysis (ESA, Ancell and Hakim, 2007; Torn and Hakim, 2008; Bednarczyk and
Ancell, 2015) using 45 forecast members at grid sizes (
The sounding and horizontal wind profiles at Panchiao, Taipei (46692), launched at 12:00 UTC on 1 June 2017. Some relevant parameters are given at the top right.
A few questions remain regarding this event on 2 June 2017. First, it
appears quite challenging to reproduce a peak amount close to the
observation (645 mm) at the correct location at
The remaining part of this paper is arranged as follows. The data, numerical model and experiments are described in Sect. 2. In Sect. 3, an overview of the case on 2 June 2017 is given. The results of our 1 km tests on topographic effects are discussed in Sect. 4. In Sect. 5, the 1 km experiments are presented and contrasted to isolate the key differences in the model for a peak rainfall amount approaching the observed value along the northern coast. Further discussion is given in Sect. 6, followed by the conclusion and summary in Sect. 7.
Hourly rainfall (mm) from the Quantitative Precipitation
Estimation and Segregation using Multiple Sensors (QPESUMS), a radar-derived
product merged with rain-gauge observations (source: CWB and the National
Science and Technology Center for Disaster Reduction (NCDR) of Taiwan),
overlaid with surface horizontal winds (barbs, 1 full barb
Distribution of 24 h accumulated rainfall (mm, color) in
Taiwan
The observational data used in this study include weather maps, sounding data, rain-gauge data, merged rainfall estimates from radar, and gauge observations from the Central Weather Bureau (CWB) of Taiwan, as well as the gridded analyses from the National Centers for Environmental Prediction (NCEP), the Navy Global Environmental Model (NAVGEM), and the Naval Research Laboratory of the USA. The weather maps and sounding data at Panchiao (near Taipei) are used for the discussion in synoptic environments and thermodynamic conditions, and the hourly rainfall data (Hsu, 1998) and the Quantitative Precipitation Estimation and Segregation using Multiple Sensors (QPESUMS), a radar-derived estimates calibrated by rain gauges over land (Gourley et al., 2001), at 10 min intervals are used for the overview of the stationary rainband and extreme rainfall in the present event.
The analysis and discussion in this study are also aided by the use of
gridded global analyses during our case period. These datasets include the
final (FNL) analyses from the NCEP Global Forecast System (GFS) every 6 h,
at 0.25
The physical package used by all CReSS experiments (with references) in this study.
The domain configuration, initial and boundary conditions,
simulation period (UTC), and other relevant settings of four CReSS
experiments (F3, F1, S3, and F1) to investigate the heavy rainfall mechanism in
this study. For grid configuration, the numbers are in
The numerical model used in this study is the Cloud-Resolving Storm
Simulator (CReSS, version 3.4.2) developed at Nagoya University, Japan
(Tsuboki and Sakakibara, 2002, 2007). This is a single-domain cloud model
that employs a nonhydrostatic and compressible equation set and a
terrain-following vertical coordinate system. As shown in Table 1, all
clouds are explicitly simulated in CReSS using a double-moment bulk
cold rain microphysical scheme with six species of vapor, cloud water, cloud
ice, rain, snow, and graupel. While other more simple schemes (1.5- or
single-moment or warm rain only) are also available, this scheme is the
most complete and sophisticated one available and thus was chosen here as
in WLC21. Parameterized sub-grid scale processes include turbulent mixing in
the planetary boundary layer, radiation, and surface momentum and energy
fluxes with a substrate model (Table 1). In all our experiments, the above
physical options are all kept the same. Further details regarding the CReSS
model can be found in some earlier studies (e.g., Wang et al., 2012;
2014a, b, 2016, WLC21) or online
(
Design and brief description of the CReSS experiments included in this study to test the effects of topography. The control experiments of F3 and F1 are the same as those given in Table 2.
Comparison of areal-averaged 24 h rainfall (mm) inside the
three domains, denoted as large (L) domain (24.85–25.65
A total of eight experiments were performed and used in this study. Four of
them are at a horizontal grid size (
Based on the F3–F1 pair, two additional 3 and 1 km pairs of experiments were designed to test the impact of topography in the present event: the F3-NNT–F1-NNT pair where only the terrain in northern Taiwan is removed and the F3-NT–F1-NT pair in which the topography of the entire Taiwan is removed (Table 3). These runs were identical to the F3–F1 pair in all other aspects. In Fig. 1a and b, the respective regions of terrain removal are shown, and any topography exceeding 1 m is set to 1 m inside them. For each type of tests, the topography was removed in both the 3 and 1 km runs, so there is no lingering effects in the latter. In S3-NNT, both the Datun Mountain and Linkou Plateau are removed (Fig. 1b).
Model surface winds at 10 m height (m s
The extreme rainfall event in northern Taiwan during 1–2 June in the Mei-yu
season of 2017 is briefly reviewed in this section. First, the CWB surface
weather maps overlaid with the NAVGEM 925 hPa flow fields surrounding Taiwan
every 6 h from 12:00 UTC on 1 June to 06:00 UTC on 2 June 2017 are shown in Fig. 2. A
stationary surface Mei-yu front, with roughly an east–west alignment, was
located about 150 km north of Taiwan at 12:00 UTC (or 20:00 LST, where LST
Figure 3 shows the synoptic conditions aloft in the troposphere at 12:00 UTC on
1 June. Extending from the low pressure over the Sea of Japan, the front (or
trough) over the East China Sea and South China was at about 27.5
The sounding observation made at Panchiao (near Taipei) in northern Taiwan
at 12:00 UTC on 1 June (Fig. 4) showed a prefrontal environment that was well
mixed in the PBL below 900 hPa in the early evening (20:00 LST), consistent
with the strong vertical wind shear near the surface (Figs. 2a and 3).
Also with gradual veering, the flow south of the surface front increased to
40 kn in speed at 900–850 hPa and further to 50 kn at 500 hPa, clearly
reaching the criteria of the LLJ (Jou and Deng, 1992; Wang et al., 2014a).
Above the PBL, the temperature lapse rate suggested conditional instability of
up to about 540 hPa (Fig. 4). The convective available potential energy
(CAPE) of a surface air parcel was 576 J kg
In Fig. 5, hourly QPESUMS data (Gourley et al., 2001) near Taiwan, overlaid
with the NCEP FNL surface horizontal winds, are shown from 18:00 UTC on 1 June to
05:00 UTC on 2 June 2017 to depict the evolution of the intense rainband
associated with the Mei-yu front. The frontal rainband first reached the
northern tip of Taiwan around 18:00 UTC on 1 June (Fig. 5a) but remained
stationary until at least 02:00 UTC on 2 June 2017 (Fig. 5i). Only afterward did it
start to move slowly inland toward the south (Fig. 5j–l). Therefore, the
rainband stayed along the northern coast of Taiwan for some 9–10 h overall,
with roughly an east–west orientation throughout this period. The rain rate
estimates along the northern coast were often 50–90 mm h
Produced by the rainband seen in Fig. 5, the observed 24 h accumulated rainfall over Taiwan from 16:00 UTC on 1 June to 16:00 UTC on 2 June (00:00–24:00 LST) reached 645 mm right along the northern coast (Fig. 6a). Two other rainfall centers also appeared along the CMR in central and southern Taiwan, each exceeding 300 mm. In fact, out of the 645 mm in northern Taiwan, 635 mm of rainfall occurred within 12 h between 16:00 UTC on 1 June and 04:00 UTC on 2 June (see Fig. 5, also WLC21), causing serious inundation and economic loss along the northern coast.
Model-simulated surface frontal positions every 2 h
Distribution of 24 h accumulated rainfall (mm, color)
around northern Taiwan in experiment
Fractional distributions of different rain rate ranges
(mm h
As described in Sect. 2.3, three pairs of 3 km–1 km experiments driven by
the NCEP GFS real-time analyses and forecasts were performed to test the
effects of the topography of Taiwan, including the two control experiments
of F3 and F1. Using the real topography of Taiwan, the high-resolution 1 km
run (F1) was able to produce a maximum 24 h rainfall of 618 mm just offshore
of northern Taiwan, with a peak amount of 541 mm at the northern coast over
land (Fig. 6b) from the initial time (
Also downscaled from their respective 3 km runs but without the topography (or the northern part of it), results of two sensitivity experiments with otherwise identical settings, the F1-NNT and F1-NT (Table 3), are compared with F1. When the topography in northern Taiwan was removed in F1-NNT (see Fig. 1b), the peak 24 h rainfall was reduced to 422 mm and occurred near the coast of northwestern Taiwan, while significant rainfall also appeared along the northern slopes of the Snow Mountain Range (SMR) (Fig. 6c). In this case, the postfrontal cold air and the areas of heavy rainfall were able to move into the Taipei Basin, but there still existed a local maximum (about 350 mm) near the northern coast. Detailed comparison shows that the differences in F1 and F1-NNT are mostly minor, except in the precise location of the rainband near the northern coast of Taiwan (Fig. 7a–f). Thus, while the Linkou Plateau is also removed in F1-NNT, the topography in northern Taiwan did act to help concentrate the rainfall, and our 1 km test results are in general agreement with those of Tu et al. (2022) using 3 km models.
When the entire topography of Taiwan was removed, significant differences were obtained in F1-NT relative to F1 (Table 3). In this test, the peak 24 h rainfall near the northern coast of Taiwan is not even 100 mm (Fig. 6d), similar to the result of Chen et al. (2018) for the event on 11–12 June 2012. Not only is the rainfall surrounding northern Taiwan greatly reduced, but the rainfall centers in the mountains also disappear. This is because without the terrain, the near-surface (and low-level) southwesterly winds can blow across the flattened island without the blocking effect (Fig. 7g–i). Without deflection and convergence, the southwesterly flow over the Taiwan Strait during the event is weaker, thereby allowing the northerly flow to advance more rapidly. As a result, the surface front in F1-NT moves across northern Taiwan more rapidly by comparison. Also, the damming and southward intrusion of postfrontal cold air along the eastern coast of Taiwan in F1-NT does not occur (Fig. 7g–i), in contrast to both F1 and F1-NNT. The above result suggests that the local convergence at the front (and rainband) between the southwesterly and northeasterly flow, with the former being enhanced by the blocking effect of the island, was important to bring the rainfall along the northern coast up to a value over 350 mm, i.e., the amount attained in the F1-NNT experiment.
The two experiments at a grid size of 1 km with real topography (F1 and S1)
are contrasted in this section. The first one is the control experiment F1
that used the hourly outputs of F3 (i.e., M18 of WLC21) as IC/BCs and
started from 13:00 UTC on 1 June for 30 h (Table 2), and it produced 24 h
rainfall reaching 541 mm at the northern coast of Taiwan. As mentioned, M18
(
Model surface winds at 10 m height (m s
Model surface winds at 10 m height (m s
The model-simulated surface frontal positions at 2 h intervals are first shown in Fig. 8 to examine whether there are significant differences in the frontal moving speed between the two 1 km experiments. To illustrate the key differences, the less rainy case of S1 is chosen to be presented first. Linked to a faster frontal moving speed in S3 (by about 9 h too early), the surface front in S1 reached the northern tip of Taiwan at around 02:00 UTC and remained at the northernmost part of the island until about 15:00 UTC (Fig. 8a), and thus was stationary for around 13 h in total. On the other hand, the surface front in F1 moved through the same short distance near the northern coast in about 10 h, roughly from 16:00 UTC on 1 June to 02:00 UTC on 2 June (Fig. 8b), in close agreement with Fig. 5. Even though somewhat shorter in duration of frontal stagnation, F1 produced more rainfall along the northern coast of Taiwan than S1. Thus, while WLC21 identified the timing and speed of frontal movement as an important factor to an increased mean rainfall in northern Taiwan from their ESA, it does not appear to be as critical here when the peak rainfall reaches around 400 mm in both runs.
The modeled 24 h rainfall distributions in S1 and F1 are plotted and
compared in Fig. 9, with the accumulation period starting from 00:00 UTC on 1 June for S1 and 16:00 UTC on 1 June for F1, respectively. Immediately apparent
is that the rainfall in S1 is more widespread, with much larger areas
offshore and to the northwest of northern Taiwan receiving over 150–200 mm
(Fig. 9a) but a lower peak amount overland at 393 mm. On the contrary, the
rainfall in F1 is much more concentrated right around the northernmost part
of the island (Fig. 9b), with a much smaller area receiving over 200 mm but
higher peak amounts, reaching 541 mm overland as mentioned and 618 mm over
the ocean about 15 km offshore from the northern tip. As shown in Sect. 4,
the topography in northern Taiwan helped to concentrate the rainfall (also
Tu et al., 2022). If 12 h is used for accumulation, then the peak values are
576 mm (offshore) and 457 mm (on land) in F1 and 269 mm (on land) in S1,
respectively. As depicted in Fig. 9, three rectangular domains are chosen to
compute the areal-mean rainfall, with a size of
Next, the hourly rainfall at all model grid points in S1 and F1 inside each
of the three domains is classified, based on their intensity, into seven
groups of 0.01–1, 1–5, 5–10, 10–20, 20–50, 50–100, and
In order to examine the location and evolution of rainbands associated with the front near northern Taiwan in the two 1 km model experiments, hourly rainfall (ending at the indicated time) every 2 h in S1 and F1 is shown in Figs. 11 and 12, respectively. In Fig. 11 for S1, the three more intense rainfall periods over the northern coast of Taiwan in Fig. 10e can be identified: approximately during 05:00–09:00 UTC (Fig. 11a, b), 11:00–15:00 UTC (Fig. 11d, e), and around 19:00–20:00 UTC (Fig. 11h) on 1 June. While their moving speed may be slow, these rainbands indeed move continuously with time, across the northern coast of Taiwan in a successive manner (Fig. 11). At almost all the instances shown in Fig. 11, multiple rainbands near the front appear in S1 (see Fig. 8a), including the northern Taiwan Strait. In Fig. 12, on the other hand, a different scenario is seen in the F1 experiment: the northern coast of Taiwan receives heavy rainfall more or less continuously, roughly from 16:00 UTC on 1 June (Fig. 12b) to 04:00 UTC on 2 June (Fig. 12h) over a period of 12 h, consistent with Fig. 10f. This is because a local rainband in Fig. 12 forms between the prefrontal westerly or southwesterly winds (immediately offshore of northwestern Taiwan) and the cold northeasterly winds (north and northeast of Taiwan), right across the northern coast, and persists through much of this 12 h period in F1 (Fig. 12b–g).
Using plots like those in Figs. 11 and 12, hourly positions of rainbands around northern Taiwan in S1 and F1 were identified and contrasted in Fig. 13. Again, as old rainbands gradually move south after passing through the northern coast in S1, new bands form over the northern Taiwan Strait or north of Taiwan, and then approach and produce rainfall along the coastal region again (Fig. 13a–c). Between 06:00 and 22:00 UTC on 1 June, at least three different rainbands affect the northern coast in Fig. 13a–c with gaps in between (see Fig. 10e), thus producing widespread rainfall but a lower peak amount in S1 (see Fig. 9a). On the other hand, a single stationary rainband persists for a long time (of over 10 h) right across the northern tip of the island in F1 (Figs. 12 and 13d), roughly from 16:00 UTC on 1 June to 03:00 UTC on 2 June. Thus, the intense rainfall is more concentrated in a smaller area, and a considerably higher 24 h peak amount of 541 mm is achieved at the northern coast of Taiwan in F1. Note also that in Fig. 12, only a few other rainbands exist with comparable intensity nearby than the one responsible for the coastal rainfall in northern Taiwan.
Model-simulated surface rainband positions (at leading
edge) around northern Taiwan at 1 h intervals during
Model pressure (hPa, isobars every 2 hPa), horizontal
winds (m s
Model pressure (hPa, isobars every 2 hPa), horizontal
winds (m s
The CWB surface regional weather charts near Taiwan
every 3 h from
In Fig. 12 where the rainband is fixed in location for many hours, a slow-moving frontal disturbance is also visible to develop over the northern Taiwan Strait to the northwest of Taiwan since 14:00 UTC and until at least 22:00 UTC on 1 June (Fig. 12a–e). As the westerly flow to the south of its cyclonic center is enhanced, it appears to produce stronger near-surface convergence with the southwesterly flow off northwestern Taiwan and subsequently with the northeasterly flow off northern and northeastern Taiwan in F1. To further examine this linkage, the pressure, horizontal wind, and convergence fields in F1 at the height of 575 m are shown in Fig. 14 at the same times as in Fig. 12 for comparison.
In Fig. 14, the low-pressure disturbance along the Mei-yu front forms before
14:00 UTC on 1 June and is still identifiable at 00:00 UTC on 2 June (Fig. 14a–f).
From 14:00 to at least 18:00 UTC (Fig. 14a–c), narrow but intense convergence
lines (over 10
While the near-surface convergence associated with the rainbands in S1
(figure omitted) does not appear to be weaker than those in F1 (Fig. 14) by
comparison; the frontal intensity is the only factor yet to be discussed
among the five features identified in WLC21, and thus perhaps should be
addressed here. In Fig. 15, the equivalent potential temperature (
The near-surface frontal disturbance over the northern Taiwan Strait in F1
is identified to be the key feature that leads to the persistent convergence
zone and rainband fixed in location across the northern coast of Taiwan and
thus the considerably higher peak amount there (541 mm in 24 h), compared to
the S1 experiment (393 mm). One may ask if there is any observational
evidence to support the presence of this low. To address this point, the CWB
regional weather charts every 3 h during the heavy rainfall period are
presented in Fig. 16. While the frontal position (analyzed in this study) in
the regional chart often differs from the synoptic map (see Fig. 2) as
expected, a frontal low of about 999–1001 hPa in mean sea-level pressure
(MSLP) is seen to appear about 150 km north of Taiwan as early as 06:00 UTC
(Fig. 16a) and moves slowly eastward along the front as the latter gradually
approached Taiwan until 1 d later (Fig. 16b–i). Throughout this period,
the MSLP at its center was consistently about 1–3 hPa lower than the
surrounding, and the only time at least one enclosed isobar could not be
identified is 03:00 UTC on 2 June (Fig. 16h). Thus, the near-surface low along
the front, during the entire time when it is captured in F1 (see Fig. 14),
is confirmed to exist in the observation and quite persistent as well. In
Fig. 14, the spatial scale and amplitude of the low (
In this study, the extreme rainfall event on 2 June 2017 in northern Taiwan,
where the peak 24 h amount of 645 mm was observed over the coast, is studied
through numerical modeling. In an earlier study, WLC21 employed ensemble
sensitivity analysis to identify some factors important to differentiate
more rainy (around 150–350 mm) from less rainy (
In S1 where the peak 24 h rainfall is less (nearly 400 mm) at the northern
coast, the surface front has a stronger contrast in
For the topographic effect, our sensitivity tests indicate significant differences when the entire island topography of Taiwan is removed. Without the blocking and deflecting effects on the prefrontal flow, there is no heavy rainfall in northern Taiwan. However, the Datun Mountain and Linkou Plateau in northern Taiwan, when removed, produce heavy rainfall areas that are located more inland along the northern slopes of the SMR with a lowered peak amount (422 mm), and thus the response is in general agreement with Tu et al. (2022), where the reduction in peak rainfall is estimated to be around 25 %.
Although the peak amount in the S1 simulation is less, the F1 does produce 541 mm on land and only about 100 mm below the observation, driven by a 3 km forecast (M18 of WLC21) that was completed well before the occurrence of the actual event (roughly 60 h prior). Thus, the 1 km forecast offers some hope to successfully predict the event in advance in real time. Some related work is currently underway and will be reported in the future.
The CReSS model and the user's guide are available at
CCW contributed to conceptualization, methodology, investigation, formal analysis, resources, writing (original draft, review, and editing) supervision, project administration, and funding acquisition. TYY contributed to methodology, software, formal analysis, investigation, data curation, visualization, and writing (original draft, review, and editing). CSC contributed to data curation, formal analysis, visualization, and writing (review and editing). MSL contributed to software, data curation, and writing (review and editing). KT contributed to software and writing (review and editing). CHL contributed to data curation and writing (review and editing).
The contact author has declared that none of the authors has any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors thank the reviewers for their constructive comments that helped improve the paper. Useful discussions with Yu-Chieng Liou (National Central University) and Ben Jong-Dao Jou (National Taiwan University) are appreciated. Help from Ms. Yi-Wen Wang and Shin-Yi Huang is also acknowledged. The various data used in this study are provided by the CWB, the DBAHR, the National Science and Technology Center for Disaster Reduction (NCDR) of Taiwan, the NCEP, and the Center for Ocean–Atmospheric Prediction Studies (COAPS) of the USA.
This research has been supported by the National Science and Technology Council (NSTC), Taiwan (grant nos. MOST 108-2111-M-003-005-MY2, MOST 110-2111-M-003-004, MOST 111-2111-M-003-005, and MOST 111-2625-M-003-001).
This paper was edited by Heini Wernli and reviewed by two anonymous referees.