Climatologically, showery weather usually starts to affect the south China coastal areas in May due to the proximity of trough systems and the onset of southwest monsoon in the latter part of the month. However, in 2018, Hong Kong experienced an exceptionally hot and dry May under a spell of sunny and rain-free weather. A heat wave of 15 consecutive very hot days (maximum temperature at the Hong Kong Observatory reaching 33.0oC or above) prevailed over the territory in the latter half of the month from 17 to 31 May, by a large margin beating the previous record of 9 consecutive very hot days for May set in 1963. Rainfall in the month amounted to only 57.5 millimetres, less than 20% of the normal of 304.7 millimetres. The scorching weather and deficient rainfall in the month drew comparison to the disastrous drought requiring prolonged water rationing in Hong Kong in 1963.
Comparing rainfall in 1963 and 2018
Currently, the two main sources of fresh water supply in Hong Kong are rainfall collected from local catchment areas and water imported from Dongjiang, with the latter supplying about 70-80% of local demand. Table 1 compares the accumulated rainfall in Hong Kong and Heyuan, a station near the catchment areas of Dongjiang, from January to May in 1963 and 2018. As both Hong Kong and Heyuan are in the same climate zone, rainfall amounts between these two places are often highly correlated, experiencing drought / wet conditions at around the same time.
The driest year on record in Hong Kong was 1963 with an annual rainfall of only 901.1 millimetres. Rainfall in the first five months of the year was the lowest ever on record (Table 1). Similar level of rainfall deficit also occurred in Heyuan that year. In 2018, however, while the accumulated rainfall in Hong Kong from January to May was the second lowest on record, the rainfall deficit at Heyuan was apparently not as severe, a reflection of the high spatial variability in rainfall distribution that can exist even within a small region.
Drought monitoring using Standardized Precipitation Index
Since local rainfall and the definition of anomalous climate conditions such as drought may differ from places to places, the World Meteorological Organization (WMO) recommends adopting the Standardized Precipitation Index (SPI) to monitor the severity of drought conditions.
In gist, SPI is a normalized index representing the probability of occurrence of an observed rainfall amount when compared with the rainfall climatology at a certain geographical location over a long-term reference period. SPI is also designed to quantify rainfall deficit over different time scales (e.g. 3-, 6-, 12-, or 24-month rainfall), facilitating the analyses of drought impact on various water resource needs. For example, SPI for a period of 6 and 12 months (i.e. SPI-6 and SPI-12 respectively) or even longer reflects the longer term impact of drought on fresh water storage in reservoirs.
Negative SPI values represent rainfall deficit, whereas positive SPI values indicate rainfall surplus. Intensity of drought can be classified according to the magnitude of negative SPI values such that more negative SPI values represent more serious droughts. For example, negative SPI values greater than 2 are often classified as extremely dry conditions.
Never too early to save water
Figure 1 and Figure 2 show the variation of SPI-6 and SPI-24 in May for Hong Kong and Heyuan from 1953 to 2018 respectively. The extremely low SPI-6 values of less than -3.5 in 1963 and the less than -2.0 values of SPI-24 in 1964 in both Hong Kong and Heyuan clearly reflect the extremity of the prolonged and more widespread drought conditions in 1963 and 1964. The drought was eventually alleviated when Typhoon Viola affected Hong Kong and brought a total of about 300 millimetres of rainfall towards the end of May 1964.
Since about 80 percent of the annual rainfall of Hong Kong is recorded between May and September, whether 2018 will be as dry as 1963 of course very much depends on the local rainfall in the long summer yet to come. While the SPI-6 analysis for May 2018 suggests that the situation in Hong Kong borders on the extremely dry regime, the situation in Heyuan is graded as moderately dry and appears to be less critical. For longer term rainfall trend, the SPI-24 of both locations are still higher than the -1.0 level, basically within the near normal range.
Nevertheless, this is no reason for complacency as the rainfall situation in the context of drought avoidance can obviously change for better or for worse during the course of the year. The most sensible thing to do is to adopt good practice of water conservation irrespective of whether the year turns out to be wet or dry in the end. In the context of climate change, rainfall is expected to become more extreme and variable in the future. Therefore, we also need to stay vigilant and be prepared for the impacts of both extremes: heavy rain as well as severe drought.
T C Lee
 WMO No-1090, Standardized Precipitation Index User Guide
Figure 1 SPI-6 and SPI-24 in Hong Kong for May (1953 to 2018)
Figure 2 SPI-6 and SPI-24 in Heyuan for May (1953 to 2018)
Last time, we discussed the purpose of and the factors being considered in establishing the rainstorm warning system. We will look at some commonly asked questions this time.
Q: Why not implement district-level rainstorm warning?
The development and movement of a rainstorm are very quick at times and can change greatly in a short time. If district-level rainstorm warning were implemented, the warning levels for different districts at different time would likely be rather different, causing confusion to those who commute cross different districts. This will also pose great challenges to the media in conveying the rapidly changing warning messages to the public.
Taking the heavy rain episode on the morning of 25 May 2013 as an example (Figure 1), the hourly rainfall over Lantau Island, Lamma Island and Wong Chuk Hang once exceeded 40 millimetres (yellow region in Figure 1). The rain band subsequently moved northwards quickly. The rain over those districts previously under downpour subsided swiftly, while that over Tsing Yi, Tsuen Wan and Ma On Shan became heavy. If district-level rainstorm warning were implemented, the warning levels for the districts mentioned above would change within a short time, and the public would find it difficult to cope with: should they continue to go to work or to school? Should their decision be made based on the warning level of the departing district or that of destination? Should the district enroute to work or to school be also taken into consideration? The situation will then become highly complicated.
Figure 1 Rainfall distribution at 10:45 a.m. and 11:30 a.m. on 25 May 2013.
Q: Why were there occasions when the rain was not heavy or there was even no rain at my location while the Observatory had issued the rainstorm warning?
There are some degrees of randomness in where heavy rain falls and how its intensity changes. Taking the "Red Rainstorm" on the morning of 10 May 2016 as an example (Figure 2), most parts of the New Territories were affected by the torrential rain, yet rain was not heavy over Lantau Island. There was even no rain in many places of Hong Kong Island and Kowloon. To get hold of the changes in rainstorm, one can refer to the Observatory's radar imageries (Figure 3 - red, orange and yellow mean heavier rainfall) for the location, intensity and direction of movement of the rain area.
Figure 2 Distribution of past-hour rainfall at 7:30 a.m. on 10 May 2016.
Figure 3 Radar imagery at 7:30 a.m. on 10 May 2016.
Q: Why can't the Observatory predict the arrival of rainstorm a couple of hours in advance so as to avoid issuing the Red or Black Rainstorm Warning during "embarrassing time" (e.g. the time when travelling to work or to school) and minimize inconvenience?
Rainstorm can change vastly in a matter of an hour or even less. Issuing the Red or Black Rainstorm Warning accurately or providing early alert a couple of hours in advance is still a challenge at the current level of science and technology. If we often issue the rainstorm warning too early such that the rain band suddenly weakens or moves elsewhere, the adverse effect of "crying wolf" will come sooner or later. Taking the situation on the night of 6 May 2016 as an example (Figure 4), there was a very broad area of intense rain approaching Hong Kong from the northwest, but it weakened abruptly right before reaching Hong Kong and then skirted the north of the territory. This exemplifies the difficulty of location-based and quantitative forecast of rain. It is difficult to issue rainstorm warning a couple of hours in advance. If a warning has to be issued at the so-called "embarrassing time", while it will inevitably cause inconvenience to people going to work or to school, we hope the public can understand that such decision is made for the sake of safety. We advise every sector in the society to formulate contingency arrangements for rainstorm prior to the rain season so that they can be readily deployed when necessary.
Figure 4 Radar imageries at 7:30 p.m. (left) and 9:30 p.m. (right) on 6 May 2016.
L.S. Lee M.C. Chow
Origin of rainstorm warning
On the morning of 8 May 1992, the Observatory recorded some 110 millimetres of rainfall in an hour that set the highest hourly rainfall record at that time [Note 1]. There were serious flooding and landslips in places on that day while the traffic in various districts was paralysed. Tragically, even some people lost their lives. This rainstorm led to the establishment of the Rainstorm Warning Signal, which aims to reflect objectively the impact of heavy rain on our society. It also aims to caution the public about the major disruption that a rainstorm may bring, so as to enable them to take precautionary measures to avoid loss of life and property.
Time flies. The rainstorm warning system has been implemented for over a quarter-century. During 2013 to 2017, the Observatory had issued the Black Rainstorm Warnings five years in a row. As extreme weather will become more frequent due to global warming, a better understanding of the origin and meaning of the rainstorm warning system will enable us to respond timely under adverse weather.
Considerations of rainstorm warning
Although Hong Kong is small in area, its terrain is complex and the impact of heavy rain, flooding or landslip in different regions is not the same. The rainstorm warning should be operated with due consideration of the overall emergency response. We cannot consider solely the spatial coverage of heavy rain, but also need to take into consideration such risk factors as heavy-rain-induced flooding and landslip as well as the size of the affected population. Currently, data of over 100 rain gauges scattered around Hong Kong is being used by the Observatory to determine whether or not to issue a rainstorm warning signal and, if issued, the level of warning. These rain gauges are not distributed evenly across the territory. There are a number of factors being considered. The rain gauges are denser in areas of higher risk of flood and landslip and higher population density (Figure 1). If heavy rain falls over areas of denser rain gauges, the number of rain gauges recording high rainfall level will then be higher, and thereby higher chance of issuing rainstorm warning. In the event that the amount of rainfall reaches a certain level that the risk of landslip in Hong Kong becomes high and numerous landslips are likely, the Observatory will also issue a landslip warning [Note 2].
(a) Distribution of severe flooding cases from 2007 to 2016 (based on data provided by the Drainage Services Department). The blue line indicates the area applicable to the Special Announcement on Flooding in the northern New Territories;
(b) Distribution of landslip locations from 2007 to 2016. The redder the colour, the higher the number of landslip cases in the area. Red colour represents over 10 landslip cases per square kilometre in general (based on data provided by the Civil Engineering and Development Department);
(c) Population density in 2016 according to District Council districts (based on data provided by the Census and Statistics Department);
(d) Density distribution of reference rain gauges. Darker colour indicates a higher density of rain gauges, while lighter colour represents a lower density. Dark green means a density of over 3 rain gauges in 10 square kilometres in general.
Taking the heavy rain episode on the night of 8 June 2014 as an example (Figure 2), the Observatory issued the Amber and the Red Rainstorm Warning Signals in succession that night. While the rain was heavier on Hong Kong Island, Kowloon and Tseung Kwan O, there was generally not much rain over other parts of the New Territories. No rain was even recorded in some places. This episode of rain happened to fall over the areas of higher risk of impact brought by weather and of higher population density. The Observatory eventually issued the Red Rainstorm Warning and serious flooding affected many roads on Hong Kong Island and in Kowloon.
Figure 2 Rainfall distribution at 10:15 p.m. on 8 June 2014.
As for another heavy rain episode that affected mainly the northern part of the New Territories on the morning of 24 July 2015 (Figure 3), the Observatory did not issue a rainstorm warning, but issued the Special Announcement on Flooding in the northern New Territories that was tailored for that area. The rainfall distribution map shows that the hourly rainfall in most parts of Hong Kong did not reach 30 millimetres, except for the northern part of the New Territories. Hence, the rainstorm warning was not issued. To conclude, the overall impact of this heavy rain episode on Hong Kong was relatively low.
Figure 3 Rainfall distribution at 8:45 a.m. on 24 July 2015.
Localised heavy rain
Given that rainstorm warning is a territory-wide warning system directly affecting all people going to work and to school rather than a system catering for localised situation, the Observatory launched the Announcement on Localised Heavy Rain service in the summer of 2016. When rain is particularly heavy in individual districts of Hong Kong, and yet to reach the criteria of the Red or Black Rainstorm Warning Signal, the Observatory will indicate the affected districts and the recorded rainfall in the Announcement on Localised Heavy Rain to remind the public to take precautions against the risk of flooding. Since flooding may lag behind the rain, the risk will not abate immediately even when the rain starts to ease off.
"Flooding can happen even it is just Amber"
It is worth noting that when the Amber Rainstorm Warning is in force, flooding may occur in some low-lying or poorly drained areas, and rivers may also overflow. Therefore, the impact of the Amber Rainstorm Warning should never be underestimated. The Red and the Black Rainstorm Warnings indicate even worse situations. In particular, the Black Rainstorm Warning indicates an adverse situation with a high chance of serious flooding, river overflow, landslip, and traffic paralysis. The public should stay in safe places to avoid any danger.
Threshold of warning
Like the tropical cyclone warning signal, the rainstorm warning signal is a territory-wide warning system. However, the uncertainty of rainstorm is much greater than that of tropical cyclone. Heavy rain can develop in a short time, and the intensity of rain and the movement of rain band have some degrees of randomness. It is still a challenge to precisely predict them with the current science and technology.
One may wonder, what about lowering the threshold for the rainstorm warning, for example, issuing rainstorm warning as long as heavy rain occurs in certain individual areas (even though rain is not heavy in all other areas), or lowering the rainfall criteria for all levels of rainstorm warning (such as lowering the criterion of the Amber Rainstorm Warning from 30 millimetres per hour to 20 millimetres per hour, or even less)? If that is the case, the following situation may occur: the number of rainstorm warnings issued will increase substantially, yet the overall impact of rainfall on Hong Kong may not be serious every time. Over time, the public may feel accustomed to the warning, or even feel annoyed, eventually lowering their alertness to the rainstorm warning or even becoming negligent. When a destructive rainstorm comes then, the public may be caught off guard, causing unwanted loss of property or even human life.
Safeguarding public safety through science
The Observatory has always been serving the public with science and safeguarding public safety as its primary mandate. Through various weather warnings and information, the Observatory strives to remind the general public of the risks they may face. The Observatory also constantly performs reviews and seeks improvement, such as examining the distribution of the reference rain gauges of the rainstorm warning system, and providing more detailed weather information at the location of users through the MyObservatory mobile application. We advise members of the public not only to pay attention to the Observatory's information, but also assess the risks they may face based on their own experience and current situation. When necessary, they should stay away from dangerous places and seek assistance from emergency departments immediately.
Next time, we will talk about topics on rainstorm warning signal that you may be interested, such as the feasibility of implementing a district-level rainstorm warning, or the possibility of providing alert a couple of hours in advance before the rainstorm occurs, similar to that for tropical cyclone warning.
L.S. Lee M.C. Chow
Note 1: New records were made subsequently in 2006 and 2008.
Note 2: The rainstorm warning aims to alert the public to the impact brought by rainstorms, whereas the landslip warning aims to alert the public to potential landslip hazards and is issued by the Observatory in consultation with the Geotechnical Engineering Office.