Next Article in Journal
Drought Hazards and Hydrological Variations in the South Hebei Plain of China over the Past 500 Years
Next Article in Special Issue
Reconstructed Phase Space of Tropical Cyclone Activity in the North Atlantic Basin for Determining the Predictability of the System
Previous Article in Journal
Greece 2023: Crazy Summer or New Normal—Lessons Not Learned
Previous Article in Special Issue
A Study on the Pre-Survey and Plan for the Establishment of the Korean Typhoon Impact-Based Forecast
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Atmosphere
Volume 15
Issue 10
10.3390/atmos15101242
atmosphere-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

WRF-ROMS-SWAN Coupled Model Simulation Study: Effect of Atmosphere–Ocean Coupling on Sea Level Predictions Under Tropical Cyclone and Northeast Monsoon Conditions in Hong Kong

by
Ngo-Ching Leung
1,
Chi-Kin Chow
2,
Dick-Shum Lau
2,
Ching-Chi Lam
2,* and
Pak-Wai Chan
2
1
Earth and Environmental Sciences Programme, Faculty of Science, The Chinese University of Hong Kong, Hong Kong, China
2
Hong Kong Observatory, Hong Kong, China
*
Author to whom correspondence should be addressed.
Atmosphere 2024, 15(10), 1242; https://doi.org/10.3390/atmos15101242
Submission received: 3 August 2024 / Revised: 24 September 2024 / Accepted: 14 October 2024 / Published: 17 October 2024
(This article belongs to the Special Issue Advances in Tropical Cyclone Prediction: Observation, Simulation, and Verification)
Browse Figures
Versions Notes

Abstract

:
The Hong Kong Observatory has been using a parametric storm surge model to forecast the rise of sea level due to the passage of tropical cyclones. This model includes an offset parameter to account for the rise in sea level due to other meteorological factors. By adding the sea level rise forecast to the astronomical tide prediction using the harmonic analysis method, coastal sea level prediction can be produced for the sites with tidal observations, which supports the high water level forecast operation and alert service for risk assessment of sea flooding in Hong Kong. The Coupled Ocean-Atmosphere-Wave-Sediment Transport (COAWST) Modelling System, which comprises the Weather Research and Forecasting (WRF) Model and Regional Ocean Modelling System (ROMS), which in itself is coupled with wave model WaveWatch III and nearshore wave model SWAN, was tested with tropical cyclone cases where there was significant water level rise in Hong Kong. This case study includes two super typhoons, namely Hato in 2017 and Mangkhut in 2018, three cases of the combined effect of tropical cyclone and northeast monsoon, including Typhoon Kompasu in 2021, Typhoon Nesat and Severe Tropical Storm Nalgae in 2022, as well as two cases of monsoon-induced sea level anomalies in February 2022 and February 2023. This study aims to evaluate the ability of the WRF-ROMS-SWAN model to downscale the meteorological fields and the performance of the coupled models in capturing the maximum sea levels under the influence of significant weather events. The results suggested that both configurations could reproduce the sea level variations with a high coefficient of determination (R2) of around 0.9. However, the WRF-ROMS-SWAN model gave better results with a reduced RMSE in the surface wind and sea level anomaly predictions. Except for some cases where the atmospheric model has introduced errors during the downscaling of the ERA5 dataset, bias in the peak sea levels could be reduced by the WRF-ROMS-SWAN coupled model. The study result serves as one of the bases for the implementation of the three-way coupled atmosphere–ocean–wave modelling system for producing an integrated forecast of storm surge or sea level anomalies due to meteorological factors, as well as meteorological and oceanographic parameters as an upgrade to the two-way coupled Operational Marine Forecasting System in the Hong Kong Observatory.

1. Introduction

Hong Kong is a metropolitan city lying over the South China coastal region, which is susceptible to the influence of tropical cyclones (TCs) and monsoons. On average, 5 to 6 TCs affect Hong Kong each year [1]. Against the backdrop of climate change, as air and sea temperatures increase, the intensity of TCs is also expected to increase, and the associated storm surges and waves will become more severe [2]. The combined effect of TC and northeast monsoon in the late season can further exacerbate the risk of coastal flooding in low-lying and vulnerable areas of Hong Kong. In winter, when a strong northeast monsoon occurs on the day of spring tides, high water levels may also cause coastal flooding in low-lying areas. Hong Kong has coastlines that are longer than 1000 km, including natural and artificial shorelines with complex morphology. The man-made coastlines are protected by coastal structures such as vertical or sloping seawalls, wave walls, piled deck structures, promenades, etc., especially those with infrastructure or transportation corridors behind them. The Hong Kong Observatory (HKO) provides high water level alerts or storm surge alert to relevant government departments to support their decision-making on emergency preparedness and response actions, such as deploying flood barriers or initiating evacuation when the situation warrants. Timely and accurate water level predictions are essential for taking precautionary measures to minimise loss of lives and property in severe flooding events. The total water level at the shoreline also includes a wave-driven component, which was found to have a significant contribution to shoreline total water levels in some studies [3,4,5] for the assessment of coastal change like dune erosion, overwash or inundation of beaches, and the water level forecast in this study only includes contributions from astronomical tide and wind- and pressure-driven surges like storm tides. The wave-driven component, i.e., wave runup or overtopping waves, is not considered in this study as there are no wave observations at tide gauge locations for model validation, and the effect of overtopping waves will depend on, among other factors, sloping of the seawall and the materials used which affect waves breaking and energy dissipation. The interactions between sea waves and coastal structures require separate techniques like empirical method, parameterisation technique or AI model to include these hydrodynamic processes [6,7,8] and take into account their effects on the total water level forecast, which are out of the scope of the current study.
Currently, storm surge forecasts are mostly generated from numerical models [9]. Yet, the accuracy of these forecasts highly depends on the predicted TC track, storm size, and intensity. Traditionally, surface wind and pressure fields derived from empirical parametric models such as Holland (1980) [10] and Jelesnianski et al. (1992) [11] are used to represent a nearly axisymmetric TC structure with an empirical formula for the radial distribution of wind or pressure. Once the distribution of either wind or pressure is determined, the other variable can be calculated from the gradient wind balance [12]. While these models are able to provide wind and pressure fields based on the user-specified TC track and intensity forecast as compared to the forecasts from individual numerical weather prediction (NWP) models, they are not able to describe details in the TC wind structure and are thus less accurate or representative when the TC wind profile deviates from the parametric model, for example TC wind field asymmetry due to interaction with other weather systems or uneven distribution of convection near the TC centre or other atypical TC structure.
With the advancement in NWP model parametrisation and increased resolution where TC intensity could be more accurately predicted, using model wind and pressure fields as input for storm surge forecasting could address the deficiencies of the TC parametric model. To account for the insufficient resolution of the global NWP model, downscaling by mesoscale atmospheric models has been adopted to generate wind and pressure fields for coupling with the ocean and wave models. These coupled atmosphere–ocean–wave models consider the atmosphere and the ocean as a unified system, which gives a more realistic picture of the Earth system. This approach incorporates the exchange of heat, momentum, and turbulent kinetic energy across the air–sea interface between the atmospheric boundary layer and the ocean upper layers into the forecasts [13]. Case studies [14,15,16] suggested that these fully coupled models significantly improved the intensity forecasts of TCs.
Dietrich et al. [17] conducted several numerical experiments with Hurricane Isaac that affected the Gulf of Mexico in 2012 to compare the performance of a parametric vortex model and a three-way coupled atmosphere–ocean–wave model in storm surge predictions. They found that while the parametric model performed better in hindcast scenarios, the fully coupled model demonstrated better forecast skills and improved early surge predictions with long lead times, which highlighted the value of coupled models in storm surge prediction despite their higher computational cost. Nevertheless, to our present knowledge, there seems to be a lack of studies that perform a systematic comparison between water level predictions from a two-way coupled ocean–wave model and a three-way coupled atmosphere–ocean–wave model with downscaling by an atmospheric model using identical meteorological inputs from global NWP products. Hence, this study was carried out to investigate the sensitivity of water height forecast to such differences in atmospheric forcing and evaluate the performance of the coupled models in water level forecasting, with a particular focus on TC and strong monsoon cases, which are the main causes of sea level anomalies in Hong Kong.
Version 3.6 of the Coupled Ocean-Atmosphere-Wave-Sediment Transport (COAWST) Modelling System [18] was adopted in this study. This open-source modelling framework comprises an ocean component using the Regional Ocean Modelling System (ROMS) [19], a wave component using the WaveWatch III (WWIII) [20] or the Simulating WAves Nearshore (SWAN) [21], and an atmospheric component using the Weather Research and Forecasting (WRF) model [22], where data fields are exchanged among these components through the Model Coupling Toolkit [23] and flexible combinations between them are allowed. This system has also been widely used in storm surge studies [24,25]. Currently, the Operational Marine Forecasting System (OMFS) in HKO has adopted a two-way coupling between the ocean model ROMS and wave model WWIII in the outer domains and SWAN in the innermost domain. Three-way coupling with the use of the WRF model was thus explored for the potential upgrade of the OMFS in the future in light of the study results.
This paper first provides a detailed explanation of the model configuration used, then introduces the selected TC and strong monsoon cases and the method of analysis, and subsequently presents, discusses, and concludes the results from the numerical experiments.

2. Data and Methods

A brief description of the experimental setup, selected cases for model performance evaluation and the method of analysis and evaluation are given in this section.

2.1. Model Configuration

In this study, there are four domains for the ocean and wave models, including SCS_PAC covering the South China Sea and western North Pacific, NSCS covering the northern part of the South China Sea, SHK covering the south of Hong Kong, and HKW covering Hong Kong waters, with their domain areas and resolutions detailed in Table 1. Bathymetric data are interpolated from the General Bathymetric Chart of the Oceans (GEBCO) [26] at 15-arc/second resolution. To save computational resources, COAWST was run independently on SCS_PAC with WWIII as the wave model, and the boundary condition was extracted and passed to the inner domain NSCS with SHK and HKW nested in it using a uniform child-to-parent ratio of five for further downscaling of the forecast with SWAN as the wave model. For the atmospheric model, another set of four domains was used, and the details are given in Table 1. The areal coverage of the atmospheric model is slightly larger than that of the ocean and wave models in order to minimise the noise induced from the boundaries of the atmospheric model. The coverage of model domains is illustrated in Figure 1.
The ocean model was based on ROMS, a free-surface model solving three-dimensional primitive equations under hydrostatic and Boussinesq approximations. The Chapman implicit, Flather and radiation scheme were used as the lateral open boundary conditions for free surface, vertically integrated momentum, and mixing turbulent kinetic energy, respectively, while the radiation boundary condition with nudging was used for three-dimensional momentum, temperature, and salinity. The initial conditions and boundary conditions for the outermost domain, SCS_PAC, were retrieved from the daily Global Ocean Reanalysis (GLORYS) product from the European Copernicus Marine Environment Monitoring Service (CMEMS). The boundary conditions for the inner domains were extracted from the run on SCS_PAC every 30 min, allowing the change in momentum transport to be better described by an increased temporal resolution. Astronomical tidal forcing was derived from the TPXO Global Tidal Solution [27] (TPXO9-atlas) of the Oregon State University (OSU).
WWIII was used as the wave model for SCS_PAC, while SWAN was used as the wave model for NSCS and its inner domains. Both WWIII and SWAN are third-generation spectral wave models. WWIII was designed for general-purpose wave modelling in deep waters, while SWAN was developed for wave modelling in nearshore water by the inclusion of shallow water effects, such as nonlinear triad wave–wave interactions and depth-induced wave breaking [28]. In SCS_PAC, the initial wave conditions were generated from the initial wind field of the atmospheric input, and a pre-run of an outermost WWIII model enclosing domain SCS_PAC provides the necessary boundary data for SCS_PAC. Output from WWIII was then applied to the boundary points of NSCS and was used as the boundary condition for the SWAN run on NSCS. The wave models interact with the ocean model by propagating information on wave conditions and receiving data of current and free surface height from it at a regular interval of 30 min. Vortex force formulation for wave effect on current and wave dissipation from wave model was adopted. The tight coupling between the ocean and wave component allows the modelling system to capture the interaction between current and waves.
For three-way atmosphere–ocean–wave coupled (hereafter referred to as “three-way coupling”) simulations, WRF with the Advanced Research solver (ARW) was used as the atmospheric model, which solves the compressible, non-hydrostatic Euler equations with a terrain-following vertical coordinate. The National Centre for Atmospheric Research (NCAR) tropical physics suite, which was shown to have satisfactory performance in TC simulations in other research [29], was adopted in this study, except that the Mellor–Yamada–Janjic (MYJ) Scheme [30,31] was used for the planetary boundary layers to enable the dependence of sea surface roughness on wave characteristics through the Taylor and Yelland formulation [32]. New Tiedtke [33,34] was adopted as the cumulus parameterisation scheme in the WRF model domains except in the finest domain, HKW, where convection was explicitly resolved and a microphysics scheme was used. The initial and boundary conditions for the atmospheric model were taken from the ERA5 global reanalysis [35] with 30 km resolution and 38 vertical levels. Under two-way ocean–wave coupled (hereafter referred to as “two-way coupling”), the WRF model was deactivated. Atmospheric forcing from the three-hourly ERA5 global reanalysis was applied to the ocean and wave models, as shown in Figure 2a. However, for three-way coupling, as illustrated in Figure 2b, meteorological data from the WRF model were passed to the ocean and wave model, while sea surface temperature and wave parameters describing the surface roughness were returned. Hence, interactions between the atmosphere and the ocean across the air–sea interface were enabled. A comparison of sea level predictions from these two configurations was made in this study.

2.2. Selected Cases

Seven weather cases were selected to illustrate the effect of three types of weather scenarios for the coupled model comparison study of sea-level anomaly prediction. Two of these were storm surge record-breaking super typhoon cases, three were significant storm surges due to the combined effect of TC and northeast monsoon, and two were solely monsoon-induced sea level anomalies. Details of the selected cases and the observed maximum sea level at five tide stations (QUB—Quarry Bay; SPW—Shek Pik; TBT—Tsim Bei Tsui; TMW—Tai Mui Wan; TPK—Tai Po Kau) managed by HKO in Hong Kong were tabulated in Table 2. The location of these tide stations is shown in Figure 3.
A comprehensive review of the two extreme storm surge cases, namely Super Typhoon Hato in August 2017 and Super Typhoon Mangkhut in September 2018, was presented in another study [36]. Coinciding with the astronomical high tide, Hato brought a maximum sea level of 3.57 mCD (metres above Chart Datum) at QUB, being the second highest since 1954 when instrumental records began [37] and ranked fifth if including non-instrumental records before 1954 [38]. Extensive flooding caused serious damage to coastal areas of Hong Kong, for example, as seawater washed into village houses in some places and into housing estates and some underground car parks. Hato developed over the western North Pacific on 20 August 2017. It propagated westward across the Luzon Strait into the northeastern part of the South China Sea. As it turned to the west–northwest and skirted at about 60 km south of the HKO on the morning of 23 August, it intensified into a super typhoon and reached a maximum intensity with 185 km/h estimated sustained wind near the centre. Its circulation remained compact. By the afternoon of the same day, it made landfall to the west of Hong Kong near Zhuhai.
Comparatively, Mangkhut was a fast-moving and rapidly intensifying storm with a large circulation. It also developed over the western North Pacific and strengthened into a Super Typhoon four days after its formation on 7 September 2018. It attained its peak intensity on 14 September, with an estimated maximum sustained wind of 250 km/h near its centre before crossing northern Luzon. It weakened but continued to traverse the northern part of the South China Sea quickly as a Super Typhoon. On 16 September, Mangkhut weakened into a Severe Typhoon while skirting past to the southwest of Hong Kong at about 100 km. It then made landfall to the west of Hong Kong near Tai Shan of Guangdong. Although convection over the eyewall has weakened after passing northern Luzon, Mangkhut maintained its structure with intense rainbands between 100 to 200 km from its centre. Due to the superposition of its extensive strong circulation and high moving speed when Mangkhut skirted past south of Hong Kong, where Hong Kong stayed in the dangerous semicircle of Mangkhut, ferocious wind and record-breaking storm surges were observed. Sea level at QUB rose by 2.35 m to 3.88 mCD, exceeding the record set by Hato in 2017. Severe inundation in coastal areas and destruction to coastal structures, including HKO’s tide station at Waglan Island (WGL in Figure 3), has made Mangkhut more devastating than Hato.
As the TC season in Hong Kong occurs mainly between May and October, TCs in autumn may coincide with the northeast monsoon. On the one hand, strong winds from the northeast monsoons generate high waves that push seawater ashore and cause high sea levels [39]. On the other hand, the monsoon may interact with TCs through its cooling and vertical shear effects, making TCs that hit Hong Kong in autumn highly variable in their tracks and intensities [40]. For this weather scenario, three recent cases, namely Typhoon Kompasu in October 2021, Typhoon Nesat in October 2022, and Severe Tropical Storm Nalgae in November 2022, were selected for analysis. Formed over the western North Pacific on 8 October 2021, Kompasu first moved northwards and gradually intensified. It turned westwards and crossed the Luzon Strait as a tropical storm on 11 October, then further strengthened into a typhoon with a maximum intensity of 120 km/h estimated sustained wind near its centre, skirted past about 360 km south of Hong Kong, subsequently weakened while making landfall over Hainan Island on 13 October. The Strong Monsoon Signal (SMS), alerting that strong northerly winds associated with the monsoon system have exceeded or were expected to exceed 40 km/h near sea level anywhere in Hong Kong, was issued by HKO on 08 UTC 11 October and was later replaced by TC warning signals. The maximum storm tide in Hong Kong was 3.53 mCD at TPK, and the maximum storm surge above astronomical tide level in Hong Kong was 1.36 m at SPW.
Nesat and Nalgae were TCs striking Hong Kong in 2022. Nesat developed over the western North Pacific on 15 October. It moved westwards across the Luzon Strait and intensified into a typhoon over the northern part of the South China Sea on 16 October. On the next day, Nesat moved past Hong Kong at about 380 km to its south–southeast, then turned west–southwestwards towards Xisha and bypassed Hainan Island from the south, eventually gradually weakened and dissipated on 20 October over the South China Sea to the southwest of Hainan Island under the influence of the northeast monsoon. The SMS was issued on the morning of 16 October till shortly after noon. TC warning signal was issued on the night of that day till the afternoon of 18 October. It was then replaced by SMS. A maximum sea level of 2.84 mCD at QUB and a maximum storm surge of 0.76 m above astronomical tide at TMW were recorded.
Nalgae was a rare TC that tracked very close to Hong Kong in late autumn and was documented in another study [38]. Forming on 26 October 2022 as a tropical depression over the western North Pacific, Nalgae gradually strengthened into a severe tropical storm on 29 October and propagated west–northwestwards across the Philippines. After entering the central part of the South China Sea, Nalgae weakened into a tropical storm briefly but later re-intensified into a severe tropical storm on 31 October while moving north–northwestwards, reaching a peak intensity with 110 km/h estimated sustained wind near the centre. On 3 November, Nalgae swept past Hong Kong at about 40 km southwest of HKO and made landfall to the west of Hong Kong over Zhuhai a few hours later. Under the influence of the combined effect of the northeast monsoon and TC Nalgae, strong winds with significant storm surges occurred. A maximum sea level of 3 mCD was registered at TBT, and a maximum storm surge of 0.73 m above astronomical tide was recorded at TMW.
Lastly, two recent cases of solely monsoon-induced sea level anomalies were selected for the coupled model comparison study. An SMS was issued early on the morning of 17 February 2022 to warn of strong easterly winds associated with the northeast monsoon till the early morning of 19 February. Strong winds were subsequently from the north and an SMS was issued again later on 19 February till the early morning of 21 February. A significant sea level rise occurred at TPK at midnight on 21 February, with an anomaly of 0.97 m above the astronomical tide. Another selected case of monsoon-induced sea level rise for the study was 21 February 2023, when an SMS was issued in the early afternoon of that day to warn of strong easterly winds associated with the monsoon. The SMS lasted till shortly before noon on the next day. The water height at TPK reached its maximum before midnight of 22 February with an anomaly of 0.77 m above the astronomical tide.

2.3. Method of Analysis

Two 96 h simulations were performed respectively, with two-way and three-way coupling in each weather case, initialising at 12 UTC about two days before the occurrence of the weather event in order to provide sufficient time for model spin-up. Outputs were generated every 30 min for the ocean and wave models and every 60 min for the atmospheric model. To assess the performance of sea level predictions from the ocean model, outputs from the innermost domain, HKW, were analysed through comparison with astronomical tide predictions and measurements at the five tide stations in Table 2 with locations marked in Figure 3, using sea level values interpolated from the nearest grid points. Excluding the model outputs from the first 24 forecast hours to eliminate the effect of model spin-up, observations at the model output times were extracted, and scatter plots of the model outputs against observations from all tide stations were drawn to evaluate the performance. Straight lines were fitted into the plots, and statistical parameters, including the coefficient of determination (R2), the bias, and the root-mean-square error (RMSE), were computed. In addition, peak values of sea level from the simulations and observations at the respective stations were extracted, and the associated biases were examined.
To analyse the difference in atmospheric forcing under the activation and deactivation of the atmospheric model component, the 10 m wind speed and wind direction from the models were also compared with the hourly mean wind speed and prevailing wind direction from the automatic weather station (AWS) at WGL. As WGL is an exposed offshore station (see Figure 3), wind conditions there are less affected by the local topography and thus are suitable for evaluation. Scatter plots were drawn for 10 m wind speed from the models against observations to examine their correlation, and statistical parameters, including R2, bias and RSME, were computed.
In addition, for the TC cases, the track and intensity of the TCs were analysed as these might significantly affect the ocean state. Parameters including latitude and longitude of the TC centre, minimum sea level pressure, and maximum 10 m wind speed near the centre, were extracted using the Geophysical Fluid Dynamics Laboratory (GFDL) Vortex Tracker Version 3.9a [41] and compared with the HKO’s TC best-track data. For two-way coupling model simulations, mean sea level pressure and surface wind speed from the ERA5 atmospheric forcing for NSCS were used for storm tracking. For the three-way coupling model run, outputs from the atmospheric model for NSCS were used, in which additional parameters describing the upper atmosphere, such as geopotential heights at 850 hPa, 700 hPa, 500 hPa and 200 hPa pressure levels, were also inspected. This assessed the performance of the atmospheric model in downscaling the meteorological conditions as input fields.

3. Results

This section gives a general overview of the results generated by the two-way and three-way coupling model simulations. The results of the three different weather scenarios under these two types of simulations will be discussed in the following sections.

3.1. TC Tracks and Intensity

Figure 4 shows the TC track forecasts from the two types of simulations as compared with the HKO’s TC best track for the TC cases under study. Table 3 shows the comparison of the forecast TC maximum intensity against the best-track data.
(i)
Hato: The tracks from the ERA5 dataset and the three-way coupling model simulations were consistent with the HKO’s TC best-track data, except that the storms tracked slightly more eastward after landfall at Zhuhai in the three-way coupling model run. The intensity was significantly underestimated for the two-way coupling model simulation. The minimum central pressure from the best-track data was 950 hPa, while that derived from the ERA5 atmospheric forcing was 984 hPa. The maximum sustained wind speed was 24.2 m/s, which was 53.0% lower than the best-track value of 51.4 m/s. On the other hand, the intensification of the storm was well simulated by the 3-way coupling model simulation, with a minimum central pressure of 965 hPa and a maximum sustained wind speed of 45.3 m/s;
(ii)
Mangkhut: The track from the ERA5 dataset aligned with the best track, but the intensity of the storm was underestimated, with the maximum sustained wind being only 32.9 m/s. For the three-way coupling model simulation, while the estimated maximum central wind speed of 59.7 m/s was closer to the best-track value of 69.4 m/s with around 14% underestimation, the storm track deviated significantly westward and made landfall at Zhanjian which was further away from Hong Kong;
(iii)
Kompasu: The track extracted from the ERA5 reanalysis generally agreed with the best-track data, while for three-way coupling, the storm tracked more southwards and passed Hainan Island to its south rather than making landfall over it as in the best-track data;
(iv)
Nesat: The storm tracked more northwards at a higher translational speed and made landfall on Hainan Island, leading to dissipation at a time earlier than the ERA5 reanalysis and the best-track data. This might be caused by a relatively poor description of the evolution of the ridge located near the storm, as the ridge intersected with the boundary of the domain NSCS. Constructing a larger regional model domain might improve the simulation results;
(v)
Nalgae: The track from the ERA5 dataset was consistent with the best-track data, while the TC in the three-way coupling model simulation turned westward with a much larger angle and dissipated at a location further away from Hong Kong.

3.2. Comparison with Observations at WGL Station

Figure 5 shows the comparison of surface meteorological conditions for different cases at weather station WGL. Figure 6 shows the scatter plots illustrating the correlation between model forecasts of 10 m wind speed against observations.
(i)
Hato: Although the two-way coupling model simulation significantly overestimated the minimum central pressure of the TC, it described the mean sea level pressure at WGL rather well, with a minimum of 988 hPa which was closer to the observed value of 982 hPa as compared with the prediction of 971 hPa from the three-way coupling model simulation. The discrepancy between the three-way coupling model and the observation may be due to the slightly eastward-deviated track from the atmospheric model. Consistent with the low bias on TC intensity, the two-way coupling simulation also significantly underestimated the maximum observed 10 m wind speed at WGL. While the wind speed from the two-way coupling simulation has an acceptable R2 with a value of 0.82 as compared with the observation, it has a negative bias of 1.7 m/s and an RSME of 5.1 m/s. Comparatively, as listed in Table 4, the R2, bias and RSME from the three-way coupling model were 0.92, −0.3 m/s and 3.3 m/s, respectively. The maximum wind speed in the two-way coupling of 17.7 m/s was much lower than the observed value of 32.5 m/s, while that in the three-way coupling was 33.9 m/s, which just differed slightly from the observed value. The results indicated that the three-way coupling simulation outperformed the two-way coupling simulation in capturing the surface meteorological conditions at WGL;
(ii)
Mangkhut: The minimum mean sea level pressure at WGL in the two-way and the three-way coupling simulation were 976 hPa and 994 hPa, respectively, being 1 hPa and 19 hPa higher than the observed value of 975 hPa. The minimum observed mean sea level pressure at WGL was well captured by the two-way simulation. However, given that both two-way and three-way coupling simulations gave a high R2 on wind speed, neither two-way nor three-way coupling simulations reproduced the maximum observed 10 m wind speed of 42.5 m/s at WGL. Their maximum values were 26.6 m/s and 29.3 m/s, corresponding to an underestimation of 37.5% and 31.2%. While the three-way coupling simulation gave an overall bias of +0.2 m/s on wind speed, which was smaller than −3.5 m/s from the two-way dataset, the RMSE of 5.2 m/s for two-way and 4.0 m/s for three-way were rather large in both simulations. Nevertheless, their wind directions generally agreed with the observation;
(iii)
Kompasu, Nesat, Nalgae: As the TC evolution in both simulations agreed with the best track data, the ERA5 dataset generally performed better in the two-way coupling for depicting changes in mean sea level pressure. It also has a higher correlation with the observations of 10 m wind speed. However, due to persistent underestimation in wind speed by the ERA5 dataset, likely due to coarser resolution, outputs from the three-way coupling model generally outperformed in describing the 10 m wind speed at WGL. However, neither the wind field in the two-way coupling nor the three-way coupling simulation could capture the sudden increase in wind speed around forecast hour 50 for Kompasu and around forecast hour 75 for Nalgae. Nevertheless, wind directions in both two-way and three-way coupling simulations were consistent with the observation, except for the last 24 h for TC Nalgae, in which the three-way coupling simulation failed to reproduce TC landfall near Hong Kong. Hence, the change in wind direction was not well captured;
(iv)
Monsoon cases in February 2022: While the ERA5 dataset could provide a reasonable description of the surface meteorological conditions at weather station WGL, the three-way coupling model failed to predict the increase in mean sea level pressure and the decrease in 10 m wind speed around forecast hour 30. The wind speed in the three-way coupling simulation at WGL has a coefficient of determination R2 of 0.32, which is much lower than that of the ERA5 dataset (R2 = 0.71). Its bias of +2.9 m/s and RMSE of 4.1 m/s was also larger than that of −0.9 m/s and 1.9 m/s, respectively, from the ERA5 dataset;
(v)
Monsoon case in February 2023: The surface meteorological fields predicted by the three-way coupling simulation were consistent with those interpolated from the ERA5 dataset and the observation. In particular, the three-way coupling model outperformed the ERA5 dataset in producing a wind speed closer to the observation. The bias and RSME from the three-way coupling simulation were −0.1 m/s and 1.8 m/s, respectively, which were smaller than those from the ERA5 dataset, −2.9 m/s and 3.6 m/s. The wind speed at WGL from both two-way and three-way coupling simulations has a high coefficient of determination of 0.90 and 0.87, respectively, as compared with the observation.

3.3. Comparison with Sea Level at Various Tide Stations

Figure 7 shows the scatter plots illustrating the correlation between model forecasts of sea level and observed at tide stations. The bias of the predicted maximum sea level in the three-way coupling simulations was plotted against that in the two-way coupling simulations in Figure 8 for all TC cases in this study. Table 4 summarises the results of the statistical parameters for 10 m wind at WGL and sea level at five tide stations for all weather cases.
(i)
Hato: The predicted sea levels from both runs were highly correlated with the observations at all tide stations, with R2 from the three-way coupling simulation of 0.95 higher than that from the two-way coupling simulation of 0.91. The bias and RSME were −0.21 m and 0.40 m, respectively, for the two-way coupling simulation and −0.14 m and 0.37 m, respectively, for the three-way coupling simulation, suggesting that the sea level predictions from the three-way coupling simulation were closer to the actual values. While the two-way coupling simulation could reasonably predict the occurrence of the storm surge, the peak values were significantly underestimated with a mean bias of −0.87 m, noting that the mean bias of the three-way coupling simulation was only +0.18 m. With the remarkable improvement in capturing the peak values of sea levels in the tide stations, the results indicated that coupling with an atmospheric model could enhance the ability of the modelling system to simulate extreme storm surges. A time series plot of sea level during the passage of Hato is given in Figure 9. The fluctuations in water level in the first 24 h were likely related to the model spin-up problem. There were incomplete data in the tide observations due to the malfunction of tide gauges;
(ii)
Mangkhut: R2, bias and RMSE for the two-way coupling simulation and the three-way coupling simulation were all comparable in magnitude. Excluding data from those tide stations with incomplete observations, the mean bias of two-way coupling and three-way coupling were −0.80 m and −0.88 m, respectively, suggesting that both simulations poorly captured the storm surge induced by Mangkhut. The underprediction of water height in the two-way coupling simulation was likely due to the underestimation of wind speed in the ERA5 dataset, and relatively weak winds were incapable of pushing enough seawater onshore. Although a higher wind speed was simulated by WRF in the three-way coupling simulation, the westward deviation of the TC track caused weaker winds over Hong Kong and, thus, weaker storm surge in Hong Kong waters. Considering non-satisfactory model performance, an additional set of model simulations was conducted using three-way coupling but replacing the ERA5 reanalysis dataset at 0.25-degree resolution with the European Centre for Medium-Range Weather Forecasts (ECMWF) model forecast at 0.125-degree resolution. Besides, another set of model experiments using the WRF model in the two-way coupling with the same resolution as in the three-way coupling was conducted. The comparison results of peak sea level forecasts are given in Table 5. The average peak sea level forecast from the two-way coupling using the WRF standalone model was about 3% higher than that from the three-way coupling using the ERA5 dataset as the initial and boundary conditions for the WRF model. However, it was about 13% lower than that from the three-way coupling using 0.125-degree resolution ECMWF model data as the initial and boundary conditions. The peak sea level forecast from three-way coupling using higher resolution ECMWF data as initial and boundary conditions for downscaling by the WRF model performed much better, especially for the TPK tide station.
(iii)
Kompasu, Nesat, Nalgae: As summarised in Table 4, the R2 for both two-way coupling and three-way coupling simulations for the three TCs were around 0.9 as compared with the observations. Bias and RMSE from the two-way and the three-way coupling model simulations indicated that the predicted sea levels from the three-way coupling model were generally closer to observations than those from the two-way coupling model. The results suggested that while the three-way coupling model might not be able to reproduce the evolution and track of the TCs well, a more realistic description of the surface wind conditions near Hong Kong might have compensated somewhat and thus achieved better results in sea level predictions.
In terms of the peak sea levels, the mean bias for the two-way coupling simulations of Kompasu, Nesat and Nalgae were −0.29 m, −0.26 m and −0.11 m, respectively, which was likely caused by the underestimation in surface wind speed by the ERA5 reanalysis dataset. For the three-way coupling simulations, the mean bias was −0.21 m, −0.08 m, and 0.05 m, respectively, and again performed better than two-way coupling. The time series for sea level during Kompasu, Nesat and Nalgae are shown in Figure 9.
(iv)
Monsoon cases: The time series plots of sea level for the monsoon events in February 2022 and February 2023 can be found in Figure 9. As there was an underestimation in mean sea level pressure and an overestimation in the 10 m wind speed for the event in 2022, reasonably, there was an overestimation in the monsoon-induced sea level anomaly, with a mean bias of −0.01 m and +0.34 m for two-way and three-way simulation, respectively. As for the monsoon event in 2023, the mean biases were +0.03 m and +0.22 m, respectively, indicating that both configurations could work well given reliable meteorological fields.

4. Discussion

Two-way coupling has been deployed in the HKO’s Operational Marine Forecasting System (OMFS) since December 2021. Model performance has been evaluated for significant wave height and current speed forecast for coastal waters and open seas against moored buoy observations and wave recorder measurements near the shores of Hong Kong and drifting buoy data over the South China Sea, as well as compared with the Mercator Ocean model reanalysis in 2022 [42]. In this study, the performance of storm surge and sea level forecasts for three types of weather scenarios, namely intense TCs, the combined effect of TC and northeast monsoon, and strong northeast monsoon coupled with spring tides was evaluated. A three-way coupling model configuration was set up for experimental testing of potential system upgrades of OMFS. The performance of storm surge and sea level forecasts for selected weather cases was evaluated and analysed in this section.

4.1. Cases of Extreme Storm Surge

TC intensity was significantly underestimated in the two-way coupling simulation driven by the ERA5 atmospheric forcing. These biases also agreed with the findings from the other studies. For example, Han et al. [43] conducted an evaluation of the ERA5 reanalysis dataset associated with TCs affecting Shanghai from 2007 to 2019 and found that the ERA5 dataset tended to underestimate the TC intensity, probably due to insufficient model resolution. The biases seemed to be increasing with TC intensity.
Figure 10 compares the model wind field in the two-way and three-way coupling runs when Super Typhoon Hato (2017) was in the vicinity of Hong Kong. While there still exists a bias of +15 hPa in the minimum central pressure and −6.1 m/s in the central pressure and maximum sustained wind in the three-way coupling, the results demonstrated the ability of the coupled atmosphere–ocean–wave model in downscaling the meteorological input for a better description of the TC wind structure and intensity.
For Super Typhoon Mangkhut, the TC forecast track in the two-way coupling deviated significantly westward after Mangkhut entered the South China Sea. The weakening of the TC during crossing Luzon (around forecast hour 10) was much faster than the observed. It might have led to different steering levels, which resulted in westward deviation from the actual track. This might be solved by initialising the model with meteorological fields of better quality and higher resolution.
To reproduce the track of Mangkhut, an additional simulation was performed using 0.125-degree resolution ECMWF model forecasts initialised at 12 UTC on 14 September 2018 as the model input and with the same model settings. Figure 11 shows that the track from this additional simulation was much closer to the best track with only a slight westward deviation. This suggested that the three-way coupling model could reproduce the TC tracks given appropriate initial and boundary conditions. Nevertheless, the preparation of such fields was beyond the scope of discussion in this study. When the TC track and intensity were well simulated, the peak sea level height could be basically reproduced. Although the two-way coupling using the WRF standalone model with the same resolution as in the three-way coupling produced a slightly higher peak sea level forecast than the latter, the three-way coupling using 0.125-degree resolution ECMWF model forecast as initial and boundary conditions for the WRF model still performed better as shown in Table 5, and this model configuration of COAWST could be set up for use in operational forecasting.
Apart from the drop in wind speed due to crossing landmass shortly after model initialisation, there was another sudden drop at around forecast hour 30 after the re-intensification of Mangkhut in the South China Sea. It was likely due to the boundary effects as the TC propagated from the coarser domain (NSCS) into the finer domain (SHK). This suggested that while downscaling the meteorological field by an atmospheric model might improve the simulation results, as in the case of Hato, it could also introduce additional errors into the simulation.
From Figure 8, it could be observed that the bias in storm surge height in cases of extreme storm surge cases, Hato and Mangkhut, was more dispersed and different from that in cases of the combined effect of northeast monsoon and TC as well as monsoon-induced surges. Since the TCs tracked very close to Hong Kong in these two cases, the wind and pressure gradients were steep. A slight difference in the location could cause a significant difference in the wind and pressure conditions. This poses a great challenge to extreme storm surge simulations by numerical models with finite resolution.

4.2. Cases of Storm Surge Under Combined Effect of TC and Northeast Monsoon

In Hong Kong, late-season TCs often encounter northeast monsoons in autumn. Their combined effect will bring more persistent higher winds and hence higher risk of significant storm surge, especially when the timing of peak storm surge height coincides with that of astronomical high tide. Interactions of TC with northeast monsoons may also lead to higher variability in intensity and track changes, which makes sea level forecasting more difficult. A comparison of Kompasu, Nesat, and Nalgae’s TC tracks showed that the tracks extracted from the ERA5 reanalysis generally agreed with the HKO’s TC best-track data. Yet, as more uncertainty was introduced after downscaling by the atmospheric model, the tracks forecast by the three-way coupling simulations could deviate from the input dataset, probably due to different physical parametrisation schemes used in the parent and child models.
The study results hinted that while the three-way coupling model might not be able to reproduce the evolution of the TCs well, a more realistic description of the surface wind conditions near Hong Kong might have compensated somewhat for that discrepancy and given better results in sea level predictions. Further system tuning of the WRF model setup would be needed to improve TC track and intensity forecasts to provide better forcing to sea level prediction. This might include enlarging the model domain covering the South China Sea (marked as NSCS in Figure 1) so as to minimise the boundary effect from the outer domain, reducing one layer of nesting (marked as SHK in Figure 1) and enlarging the model domain of the innermost (marked as HKW in Figure 1) at 1 km resolution for direct nesting with NSCS domain so as to avoid the treatment of convection representation across the “grey zone” of 3 km resolution. The role of convective parameterisation remains challenging in the grey zone [44,45], and the TC track forecast by the WRF model could also be sensitive to the cumulus parameterisation scheme [46]. Considering the asymmetry in the TC circulation due to interactions between the TC and northeast monsoon could only be simulated using numerical models as compared with that using a parametric model with a circularly symmetric TC wind profile such as Holland (1980) [10] and Jelesnianski et al. (1992) [11] in storm surge models, the use of numerical models is a way forward to improve sea level forecast under the combined effect of TC and northeast monsoon taking into account the current flow and wave interactions between the two systems.

4.3. Northeast Monsoon Cases

For the strong northeast monsoon event in February 2022, after comparing the surface pressure and wind fields of the two-way coupling and three-way coupling runs against the actual observations, it was found that the three-way coupling model over-predicted the cyclonic flow or troughing over the South China Sea as compared with the actual. Under the prevalence of the northeast monsoon over southern China, this cyclonic feature has unrealistically increased the pressure gradient over the South China coastal region and, hence, stronger than observed winds and higher than observed sea level anomalies were forecast for Hong Kong (Figure 5 and Figure 9). It also explained why the mean sea level pressure at WGL of three-way coupling simulations for the monsoon case of 22 February 2022 was lower than the observed, and the wind direction veered more to the east as compared with the two-way coupling run. The two-way coupling ran well-captured changes in mean sea level pressure and surface wind at WGL and gave better sea level rise anomaly predictions at various tide stations.
Both two-way coupling and three-way coupling captured well the mean sea level pressure at WGL in the simulation of the northeast monsoon case in February 2023; however, the surface wind speed at WGL was much lower in the two-way coupling as compared with the observed. Notwithstanding this, the sea level anomaly prediction in the two-way coupling simulation was similar to that of the three-way coupling with R2 of around 0.95 and RMSE of around 0.3 m. This hinted that there are factors other than wind speed and pressure that contribute to the rise of sea level, particularly in northeast monsoon situations. The current flow was believed to be making such a contribution which both three-way coupling and two-way coupling simulations were able to capture satisfactorily.

5. Conclusions

This study evaluated the Coupled Ocean-Atmosphere-Wave-Sediment Transport (COAWST) Modelling System comprising WRF, ROMS, WaveWatch III, and SWAN in predicting sea level variations associated with three types of weather scenarios, namely intense TCs, combined effect of TC and northeast monsoon, strong northeast monsoon coupled with spring tides, which may bring about significant sea level anomalies to Hong Kong. Model simulations for seven weather cases (two extreme storm surge cases of Super Typhoons Hato in 2017 and Mangkhut in 2018, three combined TC and northeast monsoon cases including Typhoon Kompasu in 2021, Typhoon Nesat and Severe Tropical Storm Nalgae in 2022, and two northeast monsoon cases respectively in February 2022 and February 2023) were performed by using a two-way coupled ROMS-SWAN model as in the current HKO’s OMFS and a three-way coupled WRF-ROMS-SWAN model both forced by the ERA5 reanalysis dataset or higher resolution ECMWF/WRF model forecasts. Sea level height forecasts were verified against observations at five tide gauge stations in different parts of Hong Kong. The study showcased the model’s ability to downscale meteorological fields and accurately capture maximum sea levels across various conditions. It was found that the WRF-ROMS-SWAN configuration delivered superior performance compared to other models, with reduced RMSE in surface wind and sea level anomaly predictions and a high R2 of approximately 0.9. The study could offer useful insights for improving the operational marine forecasting system and support the advancement of a more integrated model for predicting storm surges and sea level anomalies. As demonstrated in the case of Super Typhoon Mangkhut in 2018, the two-way coupling using the WRF standalone model with the same resolution as in the three-way coupling produced a slightly higher peak sea level forecast than the latter; however, the three-way coupling using 0.125-degree resolution ECMWF model forecast as initial and boundary conditions for the WRF model still performed the best. Considering that the accuracy of the TC track and intensity forecast highly affects the performance of the storm surge forecast, the configuration of the model domain and the choice of a convective parameterisation scheme in the WRF model will be further tuned and tested. A more systematic and comprehensive model validation would be conducted for a continuous period before upgrading OMFS with the three-way coupled atmosphere–ocean–wave modelling system for operational marine weather forecasting and warning services in Hong Kong. Techniques to estimate the contribution from wave runup to the total water level height and the effect of overtopping waves due to wave interactions with the coastal structures at different locations in Hong Kong will also be explored in the future.

Author Contributions

Conceptualization, C.-C.L. and P.-W.C.; methodology, N.-C.L., D.-S.L. and C.-C.L.; software, C.-K.C. and D.-S.L.; validation, N.-C.L., C.-K.C. and D.-S.L.; formal analysis, N.-C.L., C.-K.C. and D.-S.L.; investigation, N.-C.L., C.-K.C. and D.-S.L.; resources, C.-C.L. and P.-W.C.; data curation, N.-C.L., C.-K.C. and D.-S.L.; writing—original draft preparation, N.-C.L., D.-S.L. and C.-C.L.; writing—review and editing, C.-C.L. and C.-K.C.; visualization, N.-C.L., C.-K.C. and D.-S.L.; supervision, C.-C.L. and P.-W.C.; project administration, C.-C.L. and P.-W.C. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data are available upon reasonable request to the Hong Kong Observatory.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Lee, T.C.; Cheng, Y.Y. Typhoon 5–6. Hong Kong Observatory Blog on 3 April 2021. Available online: https://www.hko.gov.hk/en/blog/00000120.htm (accessed on 12 September 2024).
  2. Chen, J.L.; Wang, Z.Q.; Tam, C.Y.; Lau, N.C.; Lau, D.S.; Mok, H.Y. Impacts of climate change on tropical cyclones and induced storm surges in the Pearl River Delta region using pseudo-global-warming method. Sci. Rep. 2020, 10, 1965. [Google Scholar] [CrossRef]
  3. Stockdon, H.F.; Sallenger, A.H., Jr.; Holman, R.A.; Howd, P.A. A simple model for the spatially-variable coastal response to hurricanes. Marine Geol. 2007, 238, 1–20. [Google Scholar] [CrossRef]
  4. Stockdon, H.F.; Long, J.W.; Palmsten, M.L.; Westhuysen, A.V.D.; Doran, K.S.; Snell, R.J. Operational forecasts of wave-driven water levels and coastal hazards for US Gulf and Atlantic coasts. Commun. Earth Environ. 2023, 4, 169. [Google Scholar] [CrossRef]
  5. Birchler, J.J.; Palmsten, M.L.; Doran, K.S.; Karwandyar, S.; Pardun, J.M.; Oades, E.M.; Mulligan, R.P.; Whitehead-Zimmers, E.S. Skill assessment of a total water level and coastal change forecast during the landfall of a hurricane. Coast. Eng. 2024, 193, 104590. [Google Scholar] [CrossRef]
  6. Van der Meer, J.W.; Allsop, N.W.H.; Bruce, T.; De Rouck, J.; Kortenhaus, A.; Pullen, T.; Schüttrumpf, H.; Troch, P.; Zanuttigh, B.; EurOtop. Manual on Wave Overtopping of Sea Defences and Related Structures. An Overtopping Manual Largely Based on European Research, but for Worldwide Application, 2nd ed.; EurOtop: Frickenhausen, Germany, 2018; 304p, Available online: http://www.overtopping-manual.com/assets/downloads/EurOtop_II_2018_Final_version.pdf (accessed on 12 September 2024).
  7. Suzuki, T.; Altomare, C. Wave Interactions with Coastal Structures. J. Mar. Sci. Eng. 2022, 9, 1331. [Google Scholar] [CrossRef]
  8. Díaz-Carrasco, P.; Molines, J.; Gómez-Martín, M.E.; Medina, J.R. Neural Network calibration method for VARANS models to simulate wave-coastal structures interaction. Coast. Eng. 2024, 188, 104443. [Google Scholar] [CrossRef]
  9. Kohno, N.; Dube, S.K.; Entel, M.; Fakhruddin, S.H.M.; Greenslade, D.; Leroux, M.-D.; Rhome, J.; Thuy, N.B. Recent progress in storm surge forecasting. Trop. Cyclone Res. Rev. 2018, 7, 128–139. [Google Scholar] [CrossRef]
  10. Holland, G.J. An analytical model of the wind and pressure profiles in hurricanes. Mon. Weather Rev. 1980, 108, 1212–1218. [Google Scholar] [CrossRef]
  11. Jelesnianski, C.P.; Chen, J.; Shaffer, W.A. SLOSH: Sea, Lake, and Overland Surges from Hurricanes. In NOAA Technical Report; National Weather Service: Silver Spring, MD, USA, 1992; Volume 48, pp. 13–15. Available online: https://repository.library.noaa.gov/view/noaa/7235 (accessed on 12 September 2024).
  12. WMO. Guide to Storm Surge Forecasting. WMO-No.1076, 2011 Edition, 120p. Available online: https://library.wmo.int/viewer/28432/download?file=wmo_1076_en.pdf&type=pdf&navigator=1 (accessed on 12 September 2024).
  13. Katsafados, P.; Papadopoulos, A.; Korres, G.; Varlas, G. A fully coupled atmosphere–ocean wave modeling system for the Mediterranean Sea: Interactions and sensitivity to the resolved scales and mechanisms. Geosci. Model Dev. 2016, 9, 161–173. [Google Scholar] [CrossRef]
  14. Lok, C.C.F.; Chan, J.C.L.; Toumi, R. Importance of air-sea coupling in simulating tropical cyclone intensity at landfall. Adv. Atmos. Sci. 2022, 39, 1777–1786. [Google Scholar] [CrossRef]
  15. Sun, M.; Duan, Y.; Zhu, J.; Wu, H.; Zhang, J.; Huang, W. Simulation of Typhoon Muifa using a mesoscale coupled atmosphere-ocean model. Acta Oceanol. 2014, 33, 123–133. [Google Scholar] [CrossRef]
  16. Li, Z.; Tam, C.-Y.; Li, Y.; Lau, N.-C.; Chen, J.; Chan, S.T.; Lau, D.-S.D.; Huang, Y. How does air-sea wave interaction affect tropical cyclone intensity? An atmosphere-wave-ocean coupled model study based on Super Typhoon Mangkhut (2018). Earth Space Sci. 2022, 9, e2021EA002136. [Google Scholar] [CrossRef]
  17. Dietrich, J.C.; Muhammad, A.; Curcic, M.; Fathi, A.; Dawson, C.N.; Chen, S.S.; Luettich, R.A., Jr. Sensitivity of storm surge predictions to atmospheric forcing during Hurricane Isaac. J. Waterw. Port Coast. Ocean Eng. 2017, 144, 04017035. [Google Scholar] [CrossRef]
  18. Warner, J.C.; Armstrong, B.; He, R.; Zambon, J.B. Development of a Coupled Ocean-Atmosphere-Wave-Sediment Transport (COAWST) modeling system. Ocean Model. 2010, 35, 230–244. [Google Scholar] [CrossRef]
  19. Shchepetkin, A.F.; McWilliams, J.C. The regional oceanic modeling system (ROMS): A split-explicit, free-surface, topography-following-coordinate oceanic model. Ocean Model. 2005, 9, 347–404. [Google Scholar] [CrossRef]
  20. Tolman, H.L. A third-generation model for wind waves on slowly varying, unsteady, and inhomogeneous depths and currents. J. Phys. Oceanogr. 1991, 21, 782–797. [Google Scholar] [CrossRef]
  21. Booij, N.; Ris, R.C.; Holthuijsen, L.H. A third-generation wave model for coastal regions: 1. Model description and validation. J. Geophys. Res. Oceans 1999, 104, 7649–7666. [Google Scholar] [CrossRef]
  22. Skamarock, W.C.; Klemp, J.B.; Dudhia, J.; Gill, D.O.; Barker, D.M.; Duda, M.G.; Huang, X.-Y.; Wang, W.; Powers, H.G. A Description of the Advanced Research WRF Version 3; NCAR Technical Note 2008 NCAR/TN–475+STR; National Center for Atmospheric Research: Boulder, CO, USA, 2008; Available online: https://n2t.org/ark:/85065/d72n51q1 (accessed on 13 October 2024).
  23. Larson, J.; Jacob, R.; Ong, E. The Model Coupling Toolkit: A new Fortran90 toolkit for building Multiphysics parallel coupled models. Int. J. High Perform. Comput. App. 2005, 19, 277–292. [Google Scholar] [CrossRef]
  24. Amunugama, M.; Suzuyama, K.; Manawasekara, C.; Tanaka, Y.; Xia, Y. Typhoon-induced storm surge analysis with COAWST on different modelled forcing. J. Jpn. Soc. Civ. Eng. 2020, 76, I 210–I 215. [Google Scholar] [CrossRef]
  25. Drews, C.; Galarneau, T.J., Jr. Directional analysis of the storm surge from Hurricane Sandy 2012, with applications to Charleston, New Orleans, and the Philippines. PLoS ONE 2015, 10, e0122113. [Google Scholar] [CrossRef]
  26. Weatherall, P.; Tozer, B.; Arndt, J.E.; Bazhenova, E.; Bringensparr, C.; Castro, C.; Dorschel, B.; Drennon, H.; Ferrini, V.; Harper, H.; et al. The GEBCO_2021 Grid—A Continuous Terrain Model of the Global Oceans and Land; NERC EDS British Oceanographic Data Centre NOC: Liverpool, UK, 2021. [Google Scholar] [CrossRef]
  27. Egbert, G.D.; Svetlana, Y.E. Efficient inverse modeling of barotropic ocean tides. J. Atmos. Oceanic Technol. 2002, 19, 183–204. [Google Scholar] [CrossRef]
  28. de León, S.P.; Bettencourt, J.; Vledder, G.P.V.; Doohan, P.; Dias, F. Performance of WAVEWATCH-III and SWAN models in the North Sea. In Proceedings of the ASME 2018 37th International Conference on Ocean, Offshore and Arctic Engineering, Madrid, Spain, 17–22 June 2018. [Google Scholar] [CrossRef]
  29. Khaira, U.; Astitha, M. Exploring the real-time WRF forecast skill for four tropical storms, Isaias, Henri, Elsa and Irene, as they impacted the northeast United States. Remote Sens. 2023, 15, 3219. [Google Scholar] [CrossRef]
  30. Mesinger, F. Forecasting upper tropospheric turbulence within the framework of the Mellor-Yamada 2.5 closure. Res. Activ. Atmos. Oceanic Mod. 1993, 18, 4.28–4.29. [Google Scholar]
  31. Janjic, Z.I. The Step–Mountain Eta Coordinate Model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Weather Rev. 1994, 122, 927–945. [Google Scholar] [CrossRef]
  32. Taylor, P.K.; Yelland, M.J. The dependence of sea surface roughness on the height and steepness of the waves. J. Phys. Oceanogr. 2001, 31, 572–590. [Google Scholar] [CrossRef]
  33. Tiedtke, M. A comprehensive mass flux scheme for cumulus parametrization in large-scale models. Mon. Weather Rev. 1989, 117, 1779–1800. [Google Scholar] [CrossRef]
  34. Zhang, C.; Wang, Y.; Hamilton, K. Improved representation of boundary layer clouds over the Southeast Pacific in ARW-WRF using a modified Tiedtke cumulus parameterization scheme. Mon. Weather Rev. 2011, 139, 3489–3513. [Google Scholar] [CrossRef]
  35. Hersbach, H.; Bell, B.; Berrisford, P.; Hirahara, S.; Horányi, A.; Muñoz-Sabater, J.; Nicolas, J.; Peubey, C.; Radu, R.; Schepers, D.; et al. The ERA5 global reanalysis. Q. J. R. Meteorol. Soc. 2020, 146, 1999–2049. [Google Scholar] [CrossRef]
  36. Choy, C.W.; Lau, D.S.; He, Y. Super typhoons Hato and Mangkhut, part I: Analysis of maximum intensity and wind structure. Weather 2022, 77, 314–320. [Google Scholar] [CrossRef]
  37. Storm Surge Records in Hong Kong During the Passage of Tropical Cyclones. Available online: https://www.hko.gov.hk/en/wservice/tsheet/pms/stormsurgedb.htm?t=RANK&v=STORM_SURGE (accessed on 12 September 2024).
  38. Significant Storm Surge Events in Hong Kong Before 1954. Available online: https://www.hko.gov.hk/en/wservice/tsheet/pms/stormsurgedb_notes.htm (accessed on 12 September 2024).
  39. Thuy, N.B.; Tien, T.Q.; Wettre, C.; Hole, L.R. Monsoon-induced surge during high tides at the southeast coast of Vietnam: A numerical modeling study. Geosciences 2019, 9, 72. [Google Scholar] [CrossRef]
  40. Chan, P.-W.; Choy, C.-W.; Mak, B.; He, J. A rare tropical cyclone necessitating the issuance of gale or storm wind warning signal in Hong Kong in late autumn in 2022—Severe Tropical Storm Nalgae. Atmosphere 2023, 14, 170. [Google Scholar] [CrossRef]
  41. Biswas, M.K.; Stark, D.; Carson, L. GFDL Vortex Tracker Users’ Guide V3.9a. 2018. Available online: https://dtcenter.org/sites/default/files/community-code/gfdl/standalone_tracker_UG_v3.9a.pdf (accessed on 12 September 2024).
  42. Kong, W.; Lam, C.C.; Lau, D.S.; Chow, C.K.; Chong, S.N.; Chan, P.W. Model validation and applications of wave and current forecasts from the Hong Kong Observatory’s Operational Marine Forecasting System. Ocean Model. 2024, 190, 102393. [Google Scholar] [CrossRef]
  43. Han, Z.; Yue, C.; Liu, C.; Gu, W.; Tang, Y.; Li, Y. Evaluation on the applicability of ERA5 reanalysis dataset to tropical cyclones affecting Shanghai. Front. Earth Sci. 2022, 16, 1025–1039. [Google Scholar] [CrossRef]
  44. Hong, S.-Y.; Dudhia, J. Next-generation numerical weather prediction: Bridging parameterization, explicit clouds, and large eddies. Bull. Am. Meterol. Soc. 2012, 93, ES6–ES9. [Google Scholar] [CrossRef]
  45. Wang, R.; Qiao, F.; Liang, X.-Z.; Zhu, Y.; Zhang, H.; Li, Q.; Ding, Y. Role of convection representation across the gray zone in forecasting warm season extreme precipitation over Shanghai from two typical cases. Atmos. Res. 2021, 253, 105370. [Google Scholar] [CrossRef]
  46. Shepherd, T.J.; Walsh, K.J. Sensitivity of hurricane track to cumulus parameterization schemes in the WRF model for three intense tropical cyclones: Impact of convective asymmetry. Meteorol. Atmos. Phys. 2017, 129, 345–374. [Google Scholar] [CrossRef]
Atmosphere 15 01242 g001
Figure 1. Model domains used in the atmospheric, ocean and wave models, in which ‘SCS_PAC’ domain covers the South China Sea and western North Pacific, ‘NSCS’ the northern part of the South China Sea, ‘SHK’ the south of Hong Kong, and ‘HKW’ the Hong Kong waters.
Figure 1. Model domains used in the atmospheric, ocean and wave models, in which ‘SCS_PAC’ domain covers the South China Sea and western North Pacific, ‘NSCS’ the northern part of the South China Sea, ‘SHK’ the south of Hong Kong, and ‘HKW’ the Hong Kong waters.
Atmosphere 15 01242 g001
Atmosphere 15 01242 g002
Figure 2. A schematic diagram illustrating the coupling and data flow between different models. Two-way coupling refers to current–wave coupling, and three-way coupling is current–wave–atmosphere coupling.
Figure 2. A schematic diagram illustrating the coupling and data flow between different models. Two-way coupling refers to current–wave coupling, and three-way coupling is current–wave–atmosphere coupling.
Atmosphere 15 01242 g002
Atmosphere 15 01242 g003
Figure 3. Location of tide observations in the innermost domain ‘HKW’ of the ocean model used in this study.
Figure 3. Location of tide observations in the innermost domain ‘HKW’ of the ocean model used in this study.
Atmosphere 15 01242 g003
Atmosphere 15 01242 g004
Figure 4. Comparison of TC tracks from ERA5 input for two-way coupling runs (circle) and predicted by WRF in three-way coupling runs (triangle), with HKO TC best-track data (cross) for different TC cases in this study.
Figure 4. Comparison of TC tracks from ERA5 input for two-way coupling runs (circle) and predicted by WRF in three-way coupling runs (triangle), with HKO TC best-track data (cross) for different TC cases in this study.
Atmosphere 15 01242 g004
Atmosphere 15 01242 g005
Figure 5. Comparison of mean sea level pressure and 10 m wind speed from ERA5 input for two-way coupling model simulations (red) and predicted by WRF model in three-way coupling model simulations (blue) against observations (black) at weather station WGL for different weather cases.
Figure 5. Comparison of mean sea level pressure and 10 m wind speed from ERA5 input for two-way coupling model simulations (red) and predicted by WRF model in three-way coupling model simulations (blue) against observations (black) at weather station WGL for different weather cases.
Atmosphere 15 01242 g005
Atmosphere 15 01242 g006
Figure 6. Scatter plots of 10 m wind speed from ERA5 input for two-way coupling model simulations (red) and predicted by WRF in three-way coupling model simulations (blue) against observations at weather stations WGL for different weather cases.
Figure 6. Scatter plots of 10 m wind speed from ERA5 input for two-way coupling model simulations (red) and predicted by WRF in three-way coupling model simulations (blue) against observations at weather stations WGL for different weather cases.
Atmosphere 15 01242 g006
Atmosphere 15 01242 g007
Figure 7. Scatter plots of sea level predicted by two-way (red) and three-way coupling simulations (blue) against observations at various tide stations for different weather cases. Values before 24 h forecast were excluded to eliminate the effect of model spin-up.
Figure 7. Scatter plots of sea level predicted by two-way (red) and three-way coupling simulations (blue) against observations at various tide stations for different weather cases. Values before 24 h forecast were excluded to eliminate the effect of model spin-up.
Atmosphere 15 01242 g007
Atmosphere 15 01242 g008
Figure 8. Scatter plots of sea level bias (in metres) during maxima at different tide stations from the forecasts by three-way coupling model versus two-way coupling model for extreme storm surge cases (square) Hato (orange) and Mangkhut (red), combined effect of TC and monsoon cases (circle) Kompasu (deep blue), Nesat (light blue) and Nalgae (green), and monsoon-induced sea level anomalies cases (triangle) in February 2022 (pink) and February 2023 (violet).
Figure 8. Scatter plots of sea level bias (in metres) during maxima at different tide stations from the forecasts by three-way coupling model versus two-way coupling model for extreme storm surge cases (square) Hato (orange) and Mangkhut (red), combined effect of TC and monsoon cases (circle) Kompasu (deep blue), Nesat (light blue) and Nalgae (green), and monsoon-induced sea level anomalies cases (triangle) in February 2022 (pink) and February 2023 (violet).
Atmosphere 15 01242 g008
Atmosphere 15 01242 g009
Figure 9. Comparison of sea levels predicted by two-way (red) and three-way (blue) coupling simulations against observation (black) and predicted astronomical tide (green) at QUB, TBT and TPK tide stations. The times when SMS or TC warning signals were in force are shaded in light and dark grey, respectively.
Figure 9. Comparison of sea levels predicted by two-way (red) and three-way (blue) coupling simulations against observation (black) and predicted astronomical tide (green) at QUB, TBT and TPK tide stations. The times when SMS or TC warning signals were in force are shaded in light and dark grey, respectively.
Atmosphere 15 01242 g009
Atmosphere 15 01242 g010
Figure 10. Comparison of the surface wind field from the ERA5 reanalysis dataset in the two-way coupling simulation (left) and that predicted by the atmospheric model in the three-way coupling simulation (right) when Hato (2017) was near Hong Kong.
Figure 10. Comparison of the surface wind field from the ERA5 reanalysis dataset in the two-way coupling simulation (left) and that predicted by the atmospheric model in the three-way coupling simulation (right) when Hato (2017) was near Hong Kong.
Atmosphere 15 01242 g010
Atmosphere 15 01242 g011
Figure 11. Comparison of the track of Mangkhut from the three-way coupling simulations using the 0.25-degree resolution ERA5 reanalysis dataset (triangle) for the initial and boundary conditions and that using the 0.125-degree resolution ECMWF forecasts (circle) against the HKO’s TC best track.
Figure 11. Comparison of the track of Mangkhut from the three-way coupling simulations using the 0.25-degree resolution ERA5 reanalysis dataset (triangle) for the initial and boundary conditions and that using the 0.125-degree resolution ECMWF forecasts (circle) against the HKO’s TC best track.
Atmosphere 15 01242 g011
Table 1. Summary of model grids used in this study.
Table 1. Summary of model grids used in this study.
DomainSCS_PACNSCSSHK
(Nested in NSCS)		HKW
(Nested in SHK)
Models used
Atmospheric modelWRFWRFWRFWRF
Ocean modelROMSROMSROMSROMS
Wave modelWWIIISWANSWANSWAN
Atmospheric model
Latitude6.57° S–53.66° N12.98° N–26.72° N19.56° N–23.80° N21.11° N–23.08° N
Longitude94.62° E–155.38° E103.13° E–126.87° E110.47° E–118.09° E113.04° E–115.17° E
Number of grids216 × 250250 × 155241 × 145202 × 202
Vertical layers
(sigma coordinate)		48484848
Horizontal Resolution (km)28103.331.11
Nesting refinement ratio//33
Ocean and wave model
Latitude 5° S–50° N14° N–26° N20.95° N–23.05° N21.82° N–22.63° N
Longitude 100° E–150° E105° E–125° E112.95° E–115.05° E113.72° E–114.51° E
Number of grids201 × 221241 × 145127 × 127237 × 242
Vertical layers
(sigma coordinate)		30303030
Resolution (km)27.759.251.850.37
Nesting refinement ratio//55
Table 2. Selected cases and their associated water height records at different tide stations in Hong Kong.
Table 2. Selected cases and their associated water height records at different tide stations in Hong Kong.
NameIntensityDatesMaximum Sea Level in Metres Above Chart Datum (Maximum Storm Surge Above Astronomical Tide in Metres)
QUBSPWTBTTMWTPK
Cases of extreme storm surge
HatoSuper Typhoon20–24 August 20173.57 (1.18)3.91 * (1.54)4.56 (2.42)3.14 * (1.05)4.09 (2.42)
MangkhutSuper Typhoon7–17 September 20183.88 (2.35)3.89 (2.34)4.18 * (2.58)4.19 * (2.77)4.71 (3.40)
Cases of the combined effect of TC and northeast monsoon
KompasuTyphoon8–14 October 20213.38 (1.13)3.38 (1.36)3.29 (1.27)3.48 (1.21)3.53 (1.16)
NesatTyphoon15–20 October 20222.84 (0.70)2.83 (0.71)2.80 (0.72)2.83 (0.76)2.81 (0.69)
NalgaeSevere Tropical Storm26 October–3 November 20222.93 (0.65)2.99 (0.70)3.00 (0.70)2.95 (0.73)2.95 (0.69)
Cases of monsoon-induced sea level anomalies
//20–21 February 20222.72 (0.71)2.71 (0.66)2.88 (0.73)2.78 (0.83)2.80 (0.97)
//21 February 20233.03 (0.56)3.06 (0.55)3.34 (0.65)2.99 (0.58)3.25 (0.77)
* The Figure shown could not reflect the actual maximum due to incomplete data from the tide gauge.
Table 3. Comparison of the minimum central pressure and maximum sustained wind near the TC from ERA5 input for two-way coupling model runs, and predicted by WRF model in three-way coupling runs against HKO’s TC best-track data when the TC attained its peak intensity.
Table 3. Comparison of the minimum central pressure and maximum sustained wind near the TC from ERA5 input for two-way coupling model runs, and predicted by WRF model in three-way coupling runs against HKO’s TC best-track data when the TC attained its peak intensity.
TCMinimum Central Pressure (hPa)Maximum Sustained Wind (m/s)
2-Way3-WayBest-Track2-Way3-WayBest-Track
Hato98496595024.245.351.4
Mangkhut95594690032.959.769.4
Kompasu97798197024.733.433.4
Nesat98498096026.236.541.2
Nalgae97897798227.335.030.9
Table 4. Summary of statistical parameters describing the 10 m wind speed at weather station WGL and the sea level at five tide stations (after 24 h forecast) for the study cases. The unshaded and shaded parameters describe the two-way and the three-way coupling simulations, respectively.
Table 4. Summary of statistical parameters describing the 10 m wind speed at weather station WGL and the sea level at five tide stations (after 24 h forecast) for the study cases. The unshaded and shaded parameters describe the two-way and the three-way coupling simulations, respectively.
Event10 m Wind (m/s)Sea Level (m)Peak Sea Level Bias (m)
R2BiasRSMER2BiasRSMEQUBSPWTBTTMWTPKMean
Hato0.82−1.765.130.91−0.210.40−0.61/−1.28/−0.73−0.87
0.92−0.293.260.95−0.140.37+0.09/−0.25/+0.69+0.18
Mangkhut0.96−3.455.190.92−0.230.38−0.91−0.72//−0.77−0.80
0.91+0.183.950.91−0.160.35−0.69−0.53//−1.42−0.88
Kompasu0.95−3.854.470.88−0.380.49−0.36−0.29+0.09−0.46−0.42−0.29
0.85−1.132.930.85−0.320.47−0.28−0.21+0.17−0.40−0.33−0.21
Nesat0.82−3.473.820.92−0.340.38−0.38−0.31−0.03−0.41−0.17−0.26
0.82+1.342.280.90−0.120.23−0.16−0.05+0.11−0.18−0.14−0.08
Nalgae0.80−4.204.930.88−0.490.58−0.09−0.45+0.04−0.24+0.21−0.11
0.50+1.583.940.89−0.150.32+0.00−0.05+0.07−0.02+0.25+0.05
2022 monsoon0.71−0.941.940.95−0.170.32−0.06+0.01+0.21−0.19−0.04−0.01
0.32+2.914.120.94+0.090.31+0.30+0.38+0.56+0.23+0.22+0.34
2023 monsoon0.90−2.923.550.95−0.180.36−0.05+0.08+0.29−0.05−0.13+0.03
0.87−0.121.760.95−0.040.31+0.12+0.27+0.50+0.15+0.08+0.22
Table 5. Comparison of peak sea level forecasts (mCD) from ERA5 input and WRF standalone model input for two-way coupling simulations and three-way coupling simulations, respectively, using 0.25-degree resolution ERA5 reanalysis dataset and 0.125-degree resolution ECMWF forecast as the initial and boundary conditions against observations at various tide stations during the passage of Mangkhut in 2018.
Table 5. Comparison of peak sea level forecasts (mCD) from ERA5 input and WRF standalone model input for two-way coupling simulations and three-way coupling simulations, respectively, using 0.25-degree resolution ERA5 reanalysis dataset and 0.125-degree resolution ECMWF forecast as the initial and boundary conditions against observations at various tide stations during the passage of Mangkhut in 2018.
Tide StationPeak Sea Levels During TC Mangkhut (mCD)
2-Way (ERA5)2-Way (WRF Standalone)3-Way (ERA5—WRF)3-Way (ECMWF—WRF)Tide Gauge Observation
QUB2.973.233.193.623.88
SPW3.173.553.363.793.89
TBT3.503.933.914.083.38 *
TMW3.283.283.174.044.19 *
TPK3.943.443.294.474.71
* The figure shown could not reflect the actual maximum due to incomplete data from the tide gauge.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Leung, N.-C.; Chow, C.-K.; Lau, D.-S.; Lam, C.-C.; Chan, P.-W. WRF-ROMS-SWAN Coupled Model Simulation Study: Effect of Atmosphere–Ocean Coupling on Sea Level Predictions Under Tropical Cyclone and Northeast Monsoon Conditions in Hong Kong. Atmosphere 2024, 15, 1242. https://doi.org/10.3390/atmos15101242

AMA Style

Leung N-C, Chow C-K, Lau D-S, Lam C-C, Chan P-W. WRF-ROMS-SWAN Coupled Model Simulation Study: Effect of Atmosphere–Ocean Coupling on Sea Level Predictions Under Tropical Cyclone and Northeast Monsoon Conditions in Hong Kong. Atmosphere. 2024; 15(10):1242. https://doi.org/10.3390/atmos15101242

Chicago/Turabian Style

Leung, Ngo-Ching, Chi-Kin Chow, Dick-Shum Lau, Ching-Chi Lam, and Pak-Wai Chan. 2024. "WRF-ROMS-SWAN Coupled Model Simulation Study: Effect of Atmosphere–Ocean Coupling on Sea Level Predictions Under Tropical Cyclone and Northeast Monsoon Conditions in Hong Kong" Atmosphere 15, no. 10: 1242. https://doi.org/10.3390/atmos15101242

APA Style

Leung, N. -C., Chow, C. -K., Lau, D. -S., Lam, C. -C., & Chan, P. -W. (2024). WRF-ROMS-SWAN Coupled Model Simulation Study: Effect of Atmosphere–Ocean Coupling on Sea Level Predictions Under Tropical Cyclone and Northeast Monsoon Conditions in Hong Kong. Atmosphere, 15(10), 1242. https://doi.org/10.3390/atmos15101242

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop