Long-Term Trends in Pre-Summer Daytime and Nocturnal Extreme Hourly Rainfall in a Coastal City of South China

Extreme rainfall events are responsible for catastrophic disasters such as flooding and landslides, which can cause loss of lives and widespread social and economic destructions. During recent decades, extreme rainfall events have been reported to become more and more frequent and intensive over most regions of the world under the background of global warming (Allen and Soden ^[1]; Donat et al. ^[2]; Sugiyama et al. ^[3]; Sun et al. ^[4]; Zhang and Zhou ^[5]; Nguyen et al. ^[6]). More importantly, the increase and intensification in shortduration extreme rainfall events that happened within 3 hours are much more significant than those with longer durations, indicating that extreme rainfall tends to happen in shorter time scale (Fu et al. ^[7]; Chen et al. ^[8]; Fowler et al. ^[9]).

Located at the front edge directly influenced by the East Asian Summer Monsoon, coastal south China is one of the regions that have the most frequent and intensive extreme rainfall events in China. Rainy season in coastal south China is typically from May to August, and is divided into pre-summer (May-June) and midsummer (July-August) rainy seasons according to the migration of the East Asian Summer Monsoon and the shift of dominant synoptic systems (Ding ^[10]; Ding and Chan ^[11]; Yuan et al. ^[12]). Compared to the mid-summer rainfalls that are dominated by local thermal convections and tropical cyclones (Lee et al. ^[13]), pre-summer rainfall events are mostly associated with the activities of monsoonal flows during the onset of the East Asian Summer Monsoon (Luo et al. ^[14]; Xu et al. ^[15]). Vast amount of water vapor is transported from the tropical ocean to coastal south China during the onset of the East Asian Summer Monsoon, combined with the effect of land-sea contrast and complex terrain in coastal south China, leading to frequent severe convective rainfall events during pre-summer(Ding and Chan ^[11]).

Under the background of climate change, extreme rainfall, especially extreme hourly rainfall (EXHR) in south China has undergone the most drastic increase and enhancement all over the region during the past few decades. Moreover, the increasing trends show highly spatiotemporal heterogeneity. Spatially, higher increasing rates in EXHR are observed in urban areas and their downstream localities in south China (Fu et al. ^[7]; Su et al. ^[16]; Xiao et al. ^[17]; Wu et al. ^[18]). Temporally, the diurnal cycle of extreme rainfall over south China is modified, and the increase and intensification of nocturnal EXHRs is substantially more significant than those of daytime EXHRs over coastal south China (Su et al. ^[19]; Zhang and Zhai ^[20]). Compared to changes in large-scale circulations that act on the variation of extreme rainfall over the whole region, the distinct spatial and temporal patterns of trends in EXHRs are more likely related to changes in local factors and their interaction with large-scale circulations. Local conditions can directly trigger mesoscale convections, such as complex topography, land-sea contrast and induced local circulations, playing an important role in modulating spatiotemporal distribution of rainfall in south China (Luo et al. ^[14]; Chen et al. ^[21]). During the past three decades, urban area in south China has increased by over 10 times due to the rapid urbanization process, and this greatly modifies the land surface properties and local circulations that are closely related to the occurrence of rainfall events. It has been reported by a number of previous studies that the changes in local factors induced by urbanization have shown a significant impact on the rainfall pattern over the region (Su et al. ^[16]; Su et al. ^[19]; Wu et al ^[22]; Yan et al. ^[23]; Li et al. ^[24]). However, how the changes in local factors and their interaction with large-scale circulation modulate the long-term spatiotemporal variations in EXHRs over south China still remains unclear. Further studies on the contributions of and the change in local conditions to the trends in EXHRs and their spatiotemporal heterogeneity will provide an in-depth understanding on the physics and dynamics of extreme rainfall activities over coastal south China, which is also important for future projection of extreme rainfall events under the background of rapid urbanization and climate change over the region. Hong Kong is a coastal city located at the east coast of south China. The Hong Kong Observatory has established a comprehensive and dense observation network since the late 1980s to better monitor the long-term spatiotemporal meteorological variability all over the region (Shu et al. ^[25]). Furthermore, the drastic urbanization in south China since 1980s has greatly altered the land-surface properties, thermal conditions and local circulations all over the region (Lau and Ng ^[26]; Chao et al. ^[27]; Xia et al. ^[28]; Lu et al. ^[29]; Tse et al. ^[30]), which has pronounced effects on modifying the spatiotemporal rainfall pattern over urban and surrounding areas (Su et al. ^[16]; Li et al. ^[31]). Investigating the long-term trends in diurnal cycle of EXHRs in Hong Kong can provide an insight into the spatiotemporal variation in diurnal cycle of EXHRs over coastal areas of south China that are similar to Hong Kong but with rare available observations.

Based on the hourly record of rain-gauge data during 1988-2018 at 55 stations over Hong Kong, the long-term trends in pre-summer daytime and nocturnal EXHRs as well as their spatial pattern in coastal south China are examined and analyzed, and the possible local contributors to the trends and the underlying mechanisms are discussed in this work. The remainder of this paper is organized as follows. The datasets applied in this work is described in section 2. The methodology is presented in section 3. The results are given in section 4, along with the discussions. The conclusions are given in section 5.

Quality-controlled records of hourly rainfall rate from 1988 to 2018 at 55 stations over the entire region of Hong Kong were provided by the Hong Kong Observatory. Hong Kong is located to the east of the Pearl River Estuary (Fig. 1a): which is the front edge significantly affected by the East Asian Summer monsoon during rainy seasons. The red dots in Fig. 1b shows the location of the 55 stations. The qualitycontrolling procedures for the rainfall data include spatial and temporal consistency check and range check by inspecting the time-series data (WMO ^[32]; Zahumenský ^[33]): which help provide a reliable rainfall record at a high-temporal resolution over Hong Kong.

Figure 1.  Topography (grey shading) and urban areas over south China (a) and Hong Kong (b). The red rectangle in (a) denotes the location of Hong Kong, the blue grids denote urban areas in 1992, and the red grids denote urban areas in 2015 over south China. The green dots in (b) denote the location of the 55 rain-gauge stations over Hong Kong, and the red contours indicates terrain height in Hong Kong with an interval of 200 m.

The ERA5 dataset combines great amounts of historical observations into global estimates using advanced modelling and data assimilation systems (Hersbach et al. ^[34]). The ERA5 is based on the Integrated Forecasting System (IFS) Cy41r2 which was operational since 2016, and uses one of the most recent versions of the Earth system model and data assimilation methods (ECMWF ^[35]), benefiting from a decade of developments in model physics, core dynamics and data assimilation. As a result, the ERA5 performs markedly better compared to its coarser-resolution predecessors across most subregions in China in validations of comprehensive variables, such as temperature, wind, and precipitation (Jiang et al. ^[36]; Song and Wei ^[37]; Liu et al. ^[38]; Jiang et al. ^[39]; He et al. ^[40]; Liu et al. ^[41]). Furthermore, the ERA5 contains an ensemble component which provides estimates of analysis and short-range forecast uncertainty. This allows for the estimation of uncertainties in the reanalysis products. The 4-D assimilation system of the ERA5 makes use of 12-hourly windows in which observations are used. The resulting analysis fields follow the time evolution within the window and are stored hourly. Outputs of the ERA5 dated from 1950s provide hourly estimates of a large number of global atmospheric, land and oceanic climate variables at a ~30 km grid (0.28125°): and resolve the atmosphere vertically using 137 levels from the surface to the altitude of 80 km (Urraca et al. ^[42]; Wang et al. ^[43]).

A number of previous studies have been conducted to comprehensively examine the applicability of the ERA5 dataset in representing historical meteorological fields, including temperature (Chao et al. ^[27]; Zou et al. ^[44]; Tang et al. ^[45]), wind (Jiang et al. ^[36]; Liu et al. ^[38]; Chen et al. ^[46]), moisture (Zhang et al. ^[47]; Zhang et al. ^[48]), precipitation fields (Jiang et al. ^[39]; Jiao et al. ^[49]; Xie et al. ^[50]; Wu et al. ^[51]), as well as their long-term trends in China. These studies demonstrate that the ERA5 dataset shows a good performance in reproducing the spatiotemporal variations and long-term trends in meteorological fields over China, and outperforms satellite retrievals in some circumstances or regions.

In this study, 2-m temperature and 3-dimensional wind field over coastal south China during 1988-2018 from outputs of the ERA5 were used to investigate the contributors to the trends in daytime and nocturnal EXHRs over coastal south China and the underlying mechanisms.

In this study, an EXHR is defined as an event with hourly rainfall larger than the 95th percentile of all rainfall events (with hourly rainfall larger than 0.1 mm) at each station during 1988-2018. The annual occurrence frequency of EXHRs (R_fq95P) of a specific station is the count of EXHRs at that station in a year.

Linear trends of daytime and nocturnal R_fq95P at the 55 stations over Hong Kong during the period 1988-2018 are estimated, analyzed, and discussed. The MannKendall test (Kendall ^[52]) was conducted for trend detection. A trend was considered statistically significant if the probability of its occurrence by chance was less than 0.05. In addition, Sen's slope estimates (Sen ^[53]) were used to estimate the slopes of linear trends.

Spectral clustering is applied in this study for clustering the weather types associated to EXHRs. Spectral clustering is a modern clustering algorithm that outperforms traditional clustering methods such as k-means in many cases.

The clustering analysis was conducted upon the EXHR-associated hourly wind fields (u and v) at 925, 850, 700 hPa over south China (Fig. 1) from ERA5 outputs. To be specific, the EXHR-associated hourly wind field is defined as the wind field one hour prior to the hour when the EXHR occurs, to avoid taking account the feedback effect of the EXHRs on the wind field. For each EXHR, the associated 3-dimensional u and v fields are respectively standardized and flattened axis by axis. Finally, the two arrays of u and v are merged together (v follows u) to form a one-dimensional array (n arrays for n EXHRs) for clustering. The clustering algorithm is based on the k-means method.

To investigate the causes of the distinct trends in the occurrence frequency of EXHRs over Hong Kong, the trends in land-sea wind circulation over coastal south China is assessed and analyzed. Land-sea wind circulation is a secondary circulation imposed upon large-scale wind field at coastal regions induced by the thermodynamic contrast between land and sea. In this study, a method (Shen et al. ^[54]) is applied on hourly surface wind field of ERA5 outputs to separate land-sea wind circulation from large-scale wind field over coastal south China. The hourly 10-m x - (u) and y - (v) component from ERA5 outputs are used for land-sea wind calculation. Note that as the intersection angle between the coastline of south China and the latitude lines is ~23° (Fig. 1a), both the 10-m u and v are rotated counterclockwise by 23° before calculation to obtain new x- and y-components of wind, u' and v', which are perpendicular to the coastline of south China:

As land-sea wind is considered to be perpendicular to the coastline of south China, only v' is applied for the following calculation of land-sea wind. The full wind speed consists of large-scale background wind and secondary land-sea wind:

where v'_L denotes the y-components of large-scale wind that is perpendicular to the coastline of south China, and v'_LSW represents the y-components of land-sea wind. In this study, the large-scale wind is defined as the daily mean wind:

where v'_i denotes the hourly y-component of 10-m wind, and the land-sea wind component (v'_LSW) can be obtained by deducting the large-scale wind from full wind according to Eq. (3). v'_LSW with positive values are defined as sea wind (onshore) and those with negative values are defined as land wind (offshore).

As presented in Fig. 2, the occurrence frequency of pre-summer EXHR demonstrates a significantly increasing trend over the entire Hong Kong during 1988-2018. However, distinct spatial distributions of trends in daytime and nocturnal R_fq95P are observed. For daytime R_fq95P, the mean increasing rate of daytime R_fq95P is 0.062 h yr^-1 over the entire Hong Kong (Fig. 2a), indicating an increase by ~1.3% yr^-1 to the mean daytime R_fq95P over 1988-2018 (Fig. 2b). However, the linear trends of daytime R_fq95P illustrates highly spatial heterogeneity. It shows significantly increasing trends at most stations over northern part of Hong Kong, with rates ranging from 0.1 to over 0.2 h yr^-1 (Fig. 2a), indicating an increase by 2.5% yr^-1 to the mean occurrence of EXHR over 1988-2018 (Fig. 2b). On the contrary, no trends or insignificant trends are observed over southern part of Hong Kong (Fig. 2a and 2c).

Figure 2.  Linear trends in daytime (a-b) and nocturnal (c-d) R_fq95P by absolute values (left column) and percentage (right column) over Hong Kong during 1988-2018. The dots with black edges indicate trends with a significance level of 0.05.

The mean increasing rate of nocturnal R_fq95Pis 0.122 h yr^-1 (Fig. 2c), indicating an add-up by 2.2% yr^-1 to the mean nocturnal R_fq95P over 1988-2018 (Fig. 2d) and both are double of those for daytime R_fq95P. Different from daytime R_fq95P, the spatial distribution of the increasing trends in nocturnal R_fq95P is relatively homogeneous, with comparable increasing rates over each part of Hong Kong.

To better illustrate the distinct spatial distributions of daytime and nocturnal R_fq95P, the stations over Hong Kong are selected and classified into three groups according to the increasing rates of daytime and nocturnal R_fq95P: stations with significant trends in nocturnal R_fq95P and no significant trends in daytime R_fq95P (N + D0); stations with significant trends in both nocturnal R_fq95P and daytime R_fq95P (N+D+); stations with significant trends in daytime R_fq95P and no significant trends in nocturnal R_fq95P (N0D+). The locations and the increasing rates of daytime and nocturnal stations with of the stations in each group are presented in Fig. 3. The stations in N+D0 are mainly located over southern part of Hong Kong that is closer to the open ocean, with a mean increasing rate of 0.054 h yr^-1 and 0.138 h yr^-1 for daytime and nocturnal R_fq95P, respectively. The stations in N + D + are concentrated over northern part of Hong Kong, with a mean increasing rate of 0.122 h yr^-1 and 0.154 h yr^-1 for daytime and nocturnal R_fq95P, respectively. The stations in N0D+ are distributed at the eastern coastline of Hong Kong, with a mean increasing rate of 0.144 h yr^-1 and 0.115 h yr^-1 for daytime and nocturnal R_fq95P, respectively. The stations in the three groups are clearly distinguished from one another by locations, indicating the distinct trends in daytime and nocturnal R_fq95P over Hong Kong are highly related to local dynamics.

Figure 3.  Linear trends in daytime (left column) and nocturnal (right column) R_fq95P at stations in group N+D0 (a-b), N+D+ (c-d), and N0D+ (e-f) during 1988-2018. The dots with black edges indicate trends with a significance level of 0.05. * indicates the mean trend with a significance level of 0.05.

By applying spectral clustering method on wind fields at 700, 850 and 925 hPa over south China, the weather types of the EXHR at the stations in the three groups are clustered into three categories. As presented in Fig. 4, the three categories are distinguished by wind speed and direction over the region. The first category is featured by a southwesterly low-level jet (SW LLJ) at 700 hPa accompanied by a boundary low-level jet at 925 hPa, the second category is featured by southwesterlies (SW) at 700, 850 and 925 hPa, and the third category is marked by southwesterlies at 700 and 850 hPa, and southerlies (SS) at 925 hPa, with the lowest wind speed among the three categories. The mean linear trends for daytime and nocturnal R_fq95P related to different weather types over the stations in each group are presented in Fig. 5 and Table 1.

Figure 4.  Mean 700- (left column) and 925-hPa (right column) wind and geopotential fields of weather type SW LLJ (a-b), SW (c-d), and SS (e-f). The blue dashed contour indicates the area with wind speed higher than 12 m s^-1.

Figure 5.  Mean linear trends in daytime (red lines) and nocturnal (blue lines) R_fq95P for EXHRs associated with SW LLJ (a-c), SW (df) and SS (g-i) at stations in group N+D0 (left column), N+D+ (middle column), and N0D+ (right column). * indicates a trend with a significance level of 0.05.

For group N+D0, the mean daytime R_fq95P over all stations is 4.94 h yr^-1, which is evenly contributed by cases associated with SW LLJ, SW and SS; the mean nocturnal R_fq95P over all stations is 6.23 h yr^-1, and 75% of the EXHRs are related to SW LLJ and SS. For group N+D+ in which the stations show significant increasing trends in both daytime and nocturnal R_fq95P, the mean daytime R_fq95P over all stations is 4.87 h yr^-1, and over 70% of the EXHRs are related to SW and SS; the mean nocturnal R_fq95P over all stations is 5.48 h yr^-1, 34% and 37% of the EXHRs are associated with SW LLJ and SS, respectively. For group N0D +, the mean daytime and nocturnal R_fq95P over all stations is 5.32 h yr^-1 and 6.33 h yr^-1, and around 80% of the daytime and nocturnal EXHRs are related to SW LLJ and SW.

Overall, the statistics show that the occurrence frequency of EXHRs over Hong Kong is higher over areas closer to the sea, which is likely due to the convergence of low-level monsoonal flows along the coastline, especially when there are southwesterly lowlevel jets sitting over south China (Luo et al. ^[14]; Liu et al. ^[55]; Liu et al. ^[56]; Bai et al. ^[57]; Du and Chen ^[58]). Meanwhile, the occurrence frequency of EXHRs is higher during nighttime, especially in the southern part of Hong Kong, for which the mean nocturnal R_fq95P is 26% higher than the mean daytime R_fq95P. This is due to the fact that the thermal condition over areas closer to the sea is not as good as that over inland areas during daytime, however, during nighttime, the confrontation between land breeze and low-level southerly monsoonal flow results in convergence, which is the most enhanced along the coastline, leading to higher occurrence frequency of EXHRs over areas near the coast (Luo et al. ^[14]; Chen et al. ^[59]; Li et al. ^[60]; Chen et al. ^[61]; Li et al. ^[62]). In addition, most of the EXHRs at the eastern coastline of Hong Kong (N0D +) is associated with southwesterlies, especially those with relatively lower wind speed (SW). This might be caused by southwesterlies with lower wind speed, which results in weaker convergence at the upstream western coastline, which are not enough for triggering convection there. In this case, the convection tends to happen at downwind areas as the moist monsoonal flows cross over Hong Kong and gain enough energy from passing mountains and urban areas in Hong Kong. Moreover, the mean daytime R_fq95P over the stations in N0D+ is the highest among three groups. This is contributed by the convergence of southwesterly monsoonal flows and daytime sea breezes at the eastern coast of Hong Kong, where the sea breezes are easterlies (Lu et al. ^[29]).

For group N + D0, the general increasing trend of nocturnal R_fq95P is mainly contributed by the increase of EXHRs related to SW LLJ and SW with relatively higher wind speed, and the mean increasing rates of nocturnal R_fq95P associated to SW LLJ and SW are both 0.072 h yr^-1 (Fig. 5a and 5d). As presented in Table 1, this suggests a total add-up of 2.23 h over 1988-2018, which indicates an increase by 97% and 144% relative to the mean nocturnal R_fq95P (2.29 and 1.55 h yr^-1) associated to SW LLJ and SW over the stations in group N + D0 during the investigated period. Compared to nocturnal R_fq95P, daytime R_fq95P of all three weather types shows no significant trends during 1988-2018.

For group N+D+, the general increasing trend of daytime R_fq95P is jointly contributed by the increase of EXHR related to SW LLJ and SW, whereas the general increasing trend of nocturnal R_fq95P is mainly attributed to the increase of EXHR related to SW. The mean increasing rates of daytime R_fq95P related to SW LLJ and SW are 0.047 and 0.069 h yr^-1, indicating an add-up of 1.46 and 2.14 h to the mean daytime R_fq95P related to SW LLJ and SW, which doubles the mean daytime R_fq95P related to SW LLJ and SW during 1988-2018. On the other hand, the mean increasing rate of nocturnal R_fq95P related to SW is 0.107 h yr^-1, indicating an add-up of 3.32 h to the mean nocturnal R_fq95P related to SW, which suggests an increase by ~200% to the mean nocturnal R_fq95P related to SW during 1988-2018.

For group N0D +, the general increasing trend of daytime R_fq95P is equally contributed by the increase of EXHR related to the three weather types. The increasing rates of 0.057, 0.051, and 0.052 h yr^-1 suggest an increase of 1.77, 1.58, and 1.58 h over the period 1988-2018, which accounts for 100%, 66%, and 136% of the mean daytime R_fq95P during 1988-2018 associated with SW LLJ, SW, and SS, respectively. On the other hand, it shows a significant increasing trend in nocturnal R_fq95P associated with SW LLJ, with the increasing rate of 0.084 h yr^-1, indicating an add-up of 2.60 h to the mean over the period 1988-2018, which doubles the mean nocturnal R_fq95P related to SW LLJ during 1988-2018.

In general, the enhancements in both daytime and nocturnal R_fq95P are mostly contributed by the increase of EXHRs related to SW LLJ and SW with higher wind speed. This suggests that the trends in EXHRs over Hong Kong are highly dependent on the long-term variation of southwesterlies over south China during 1988-2018. Furthermore, for the stations to the south of Hong Kong, the variation of southwesterlies tends to lead to increase of nocturnal EXHRs only, which is different from the case for the stations at the north end and the eastern coastline of Hong Kong. This indicates that the impacts of the long-term variations in southwesterlies on daytime and nocturnal EXHRs are dependent on the locations of the stations.

The linear trends in low-level wind fields at 850 hPa, 900 hPa and 950 hPa over south China during 1988-2018 are presented in Fig. 6. Both daytime and nocturnal wind speed at 950 hPa shows a significant southward trend over coastal south China over the period 1988-2018, especially over the Guangdong-Hong Kong-Macau Greater Bay Area (to be referred to as "Greater Bay Area" hereafter), which is the most urbanized region in south China. This indicates a substantial weakening of low-level southerlies over coastal south China during 1988-2018 due to the drastic change of land surface properties induced by the urbanization process. The decreasing rate of wind speed at 950 hPa is as high as 0.05 m s^-1 yr^-1 over the Pearl River Delta, indicating a decrease of wind speed by ~1.6 m s^-1 over the period, which accounts for ~20% and ~15% of the mean daytime and nocturnal wind speed, respectively, in the region during 1988-2018. The decreasing trends are observed only at lower levels, and weaken with altitudes. There is no significant trend observed in wind fields at 850 hPa.

Figure 6.  Linear daytime (left column) and nocturnal (right column) trends in pre-summer for 850- (a-b), 900- (c-d), and 950-hPa (e-f) wind fields over south China during 1988-2018. Red dots represent trends with a significance level of 0.05. The black star denotes the location of Hong Kong.

The substantial weakening of low-level southerlies at the coastal area during 1988-2018 suggests an enhancement of low-level meridional convergence, as shown in Fig. 7a and 7b. On average, both daytime (Fig. 7a) and nocturnal (Fig. 7b) meridional divergence field at 950 hPa shows a negative trend during the period, indicating an enhancement of low-level convergence at the coastal region of the Greater Bay Area during 1988-2018, with much stronger enhancement during nighttime. The enhanced low-level convergence provides favorable dynamic conditions for convection initiation, leading to a general increase in EXHRs over coastal south China. Note that the enhancement of lowlevel convergence is stronger onshore during daytime, whereas offshore during nighttime, which is likely due to the diurnal variation of local circulation at coastal south China. During daytime, southerly sea breeze prevails at low levels during daytime at the coastline of south China, which moves the convergence center more inland by overlapping on southerly monsoonal flows. By contrary, northerly land-breeze prevails at low levels during nighttime at the coastline of south China, which tends to push the convergence center offshore (Luo et al. ^[14]; Chen et al. ^[59]; Li et al. ^[62]).

Figure 7.  Linear daytime (left column) and nocturnal (right column) trends in pre-summer for 950-hPa meridional (a-b) and zonal (cd) divergence over south China during 1988-2018. Black dots represent trends with a significance level of 0.05. The black star denotes the location of Hong Kong.

On the other hand, the blocking of low-level southerlies at the front edge of the Pearl River Delta also leads to an eastward detour of monsoonal flows over the ocean, as shown in Fig. 6. This results in an enhancement of low-level zonal divergence near the coastline. The linear trend in zonal divergence at 950 hPa during 1988-2018 over south China is presented in Fig. 7c and 7d. Both daytime (Fig. 7c) and nocturnal (Fig. 7d) zonal divergence shows a significant enhancement along the coastline and over the north part of South China Sea, and the enhancement is stronger during nighttime. Note that at the eastern coast of Hong Kong, there is a significant break in the enhancement of daytime meridional divergence along the coastline (Fig. 8c). This is probably due to the low-level easterly seabreeze at this location, where the coastline is in northsouth direction. This break tends to disappear in higherlevels above 950 hPa, where the easterly sea-breeze weakens gradually and turns into westerlies. The eastward-detour of the southwesterly monsoonal flows results in the increase in the EXHRs at the eastern coast of Hong Kong (Fig. 3e and 3f), especially during daytime (Fig. 3e) when the easterlies encounter with westerly sea winds at the eastern coast of Hong Kong.

Figure 8.  Mean daytime (left column) and nocturnal (right column) 700-, and 925-hPa wind and geopotential fields for EHXRs associated with weather types SW LLJ (a-d) and SW (e-h) at stations in group N+D0. The black star denotes the location of Hong Kong. The blue dashed contour indicates area with wind speed higher than 12 m s^-1.

The distinct trends in the occurrence frequency of daytime EXHRs over northern and southern part of Hong Kong indicates a northward or inland-ward shift of daytime EXHRs. Although the increase in low-level convergence at coastal south China induced by urbanization explains the general increase of EXHRs in Hong Kong, the reason for this northward shift of daytime EXHRs still remains unknown. According to Table 1, the increase in daytime EXHRs at the stations in group N+D+ is contributed by the cases related to SW LLJ and SW. Meanwhile, the daytime EXHRs at the stations in N+D0 related to SW LLJ and SW show no significant trends. Therefore, the discussion on the contributor to the northward shift of daytime EXHRs over Hong Kong will be focused on the cases associated with SW LLJ and SW.

Figures 8 and 9 show the mean daytime meteorological fields of EXHRs related to SW LLJ and SW in group N + D0 (Fig. 8) and N + D + (Fig. 9). In general, for those EXHRs at stations in group N + D0, which are located at the southern part of Hong Kong, the mean wind of daytime cases (Fig. 8a, 8c, 8e, and 8g) is slightly stronger than that of nocturnal cases (Fig. 8b, 8d, 8f, and 8h), especially at lower levels (Fig. 8c-8d and 8g-8h). The case is similar for those EXHRs at stations in group N+D+, which are concentrated at more inland northern area of Hong Kong (Fig. 9). The mean wind of daytime (Fig. 9a, 9c, 9e, and 9g) is lightly stronger than that of nocturnal cases (Fig. 9b, 9d, 9f, and 9h). Despite the subtle differences existing between the mean wind fields for different groups of EXHRs, those differences are relatively large-scale and thus unlikely the cause of the distinct trends in daytime EXHRs within such a small area in Hong Kong. The northward-shift of daytime EXHRs in Hong Kong is more likely caused by the changes in small-scale local factors, such as local thermal conditions and land-sea-breeze circulations, which may affect the location of rainfall in Hong Kong. Therefore, the changes in local daytime 2-m temperature and sea-wind prevailing during daytime over coastal south China in the period 1988-2018 are examined. Note that the definition of sea-wind is described in section 3.3.

Figure 9.  Mean daytime (left panels) and nocturnal (right panels) 700-, and 925-hPa wind and geopotential fields for EXHRs associated with weather types SW LLJ (a-d) and SW (e-h) at stations in group N+D+. The black star denotes the location of Hong Kong. The blue dashed contour indicates area with wind speed higher than 12 m s^-1.

Figure 10 shows the trends in mean daytime 2-m temperature and sea-wind speed over south China for all EXHRs associated with SW LLJ and SW at stations in group N+D0 and N+D+, respectively. The mean daytime 2-m temperature field of the EXHRs at stations in both groups are generally identical to each other, with relatively high temperature over the ocean and the western part of the Greater Bay Area (Fig. 10a and 10b). However, the mean sea-wind speed of the EXHRs at stations in group N+D+ (Fig. 10d) is significantly higher along the coastal region than that in group N+D0 (Fig. 10c), indicating that the EXHRs over northern part of Hong Kong are accompanied by stronger sea-wind compared with those that occurred over the areas more to the south - the convergence and convective systems that caused heavy rainfall are pushed northward by enhanced sea-wind front. The trends in mean 2-m temperature and sea-wind speed during pre-summer over 1988-2018 are further examined and presented in Fig. 11. The mean 2-m temperature shows significantly increasing trends during 1988-2018 over south China, especially at coastal region and over northern part of the South China Sea (Fig. 11a). More importantly, the seawind speed is significantly enhanced along the coastline of south China during 1988-2018, especially at the Pearl River Estuary (Fig. 11b). The rate of the enhancement of sea-wind can be up to 0.006 m s^-1 yr^-1, suggesting an increase of 0.18 m s^-1 in sea-wind speed during the period 1988-2018, accounting for ~30% of the mean sea-wind speed over this region. The substantial enhancement of sea-wind at the coastal south China leads to the sea-wind front and the convergence more to the north during daytime, and thus more daytime EXHRs at northern part of Hong Kong, which is the cause of the northward shift of daytime EXHRs over coastal south China.

Figure 10.  Mean 2-m temperature (upper panels) and sea wind speed (lower panels) for EXHRs associated with SW LLJ and SW at stations in N+D0 (a, c) and N+D+ (b, d). The black star denotes the location of Hong Kong.

Figure 11.  Linear trends in mean daytime 2-m temperature (a) and sea-wind speed (b) in pre-summer over the period 1988-2018. The black star denotes the location of Hong Kong. Black dots indicate trends with a significance level of 0.05.

Based on the hourly rain-gauge record during 1988-2018 at 55 stations in Hong Kong during 1988-2018, the long-term trends in the occurrence frequency of presummer daytime and nocturnal EXHRs in coastal south China and their spatial distributions are examined and analyzed. Despite a significant increase observed in the occurrence frequency of pre-summer EXHRs during the investigated period, the increase in daytime and nocturnal EXHRs show distinct spatial patterns. The increase in nocturnal EXHRs is observed over the entire Hong Kong, with relatively consistent increasing rates. However, the increase in daytime EXHRs is concentrated over the northern or eastern areas of Hong Kong-downstream areas of the southwesterly monsoonal flows, whereas no significant increasing trends in daytime EXHRs are observed over southern part of Hong Kong.

The clustering of weather types associated with daytime and nocturnal EXHRs further reveals that, the increase in nocturnal EXHRs over southern Hong Kong as well as the increase in daytime and nocturnal EXHRs over northern Hong Kong are mainly contributed by the increase of the cases associated with southwesterly monsoonal flows with relatively high wind speeds - wind speeds along the coastline are substantially lowered due to increased surface roughness induced by urbanization over the Greater Bay Area in south China during the past few decades, leading to enhanced convergence of low-level monsoonal flows and an increase in EXHRs at coastal south China. Meanwhile, the daytime sea-wind circulation at coastal south China is significantly enhanced during the investigated period. Enhanced sea-wind front during daytime is the main reason for the northward increase in the occurrence frequency of daytime EXHRs in Hong Kong.

In addition, the blocked low-level southwesterly monsoonal flows along the coastal south China are detoured eastward, leading to enhanced low-level convergence at eastern coast of Hong Kong, especially during daytime, when the easterly sea winds prevail, and resulting in an increase in EXHRs over eastern Hong Kong, with higher increasing rates during daytime.

This work provides insights into how the changes in local factors and their interaction with large-scale circulations contribute to the long-term spatiotemporal trends in extreme rainfall activities over coastal south China, which is important for future projection of extreme rainfall events under the background of rapid urbanization and climate change over the region. However, further quantitative investigations are required for a deeper understanding on the details and relative contribution of each physical and dynamic process, as well as the interaction in between.