RESEARCH PAPER

 

Forecasting extreme hourly rainfall in South Africa for disaster risk reduction: thresholds and return periods

 

 

Jan H Vermeulen^{I, II}; David W Hedding^II; Nthabiseng Letsatsi^III

^IDisaster Risk Reduction, South African Weather Service, Centurion, South Africa
^IIDepartment of Geography, University of South Africa, Johannesburg 1710, South Africa
^IIIClimate Service, South African Weather Service, Centurion, South Africa

Correspondence

 

 

---

ABSTRACT

Highest and percentile values determined for daily, hourly and 5-min rainfall data (July 1994 to June 2021) from 64 automatic weather stations across South Africa were used to define extreme hourly and 5-min rainfall intensity. Internationally, 99.9^th and 99.99^th percentiles are typically considered as thresholds for hourly and sub-hourly extreme rainfall when forecasting for disaster risk reduction assessments. In South Africa (SA), the average of the 99.9^th percentile for hourly rainfall values is 29.9 mm/h. This represents a good indicator of extreme hourly rainfall in SA and is a useful threshold for forecasting flash floods. The average highest of the hourly rainfalls for SA, 53.9 mm/h, should be a good indicator of more extreme hourly rainfalls for the country. The average of the 99.99^th percentile for 5-min rainfall values is 12.8 mm/5 min, which equates to 2.6 mm/min. Significantly, the 5-min rainfall data is used to establish South African categories based on rainfall intensity and total rainfall, whereby an event can be classified as a cloudburst, downpour or shower. Using the newly established local categories, the severe thunderstorm of 4 April 2000 at Hoedspruit that produced 132.2 mm in 25 min from an intensifying upper air trough system was classified as a cloudburst. Interestingly, the 66.2 mm recorded in 5 min during this event makes it the world record holder for all-time highest 5-min rainfall, passing the previous world 5-min rainfall record of 63.0 mm in 5 min recorded at Porto Bello, Panama, on 29 November 1911. Return periods of expected maximum daily, hourly and 5-min rainfall, based on yearly highest values, were also calculated for South Africa. This study presents expected maximums for 5-min rainfall in return periods of 10, 25, 50 and 100 years, a first for South Africa, which can inform strategies for disaster risk reduction.

Keywords: rainfall, hourly rainfalls, rainfall intensity, return periods, cloudburst, South Africa

---

 

 

INTRODUCTION

Extreme very short-term or hourly rainfall events mostly cause damage to infrastructure, property, the loss of livestock and even the loss of lives from the resulting flash floods and mass wasting (Wu and Lin, 2017; Thayyen et al., 2013; Zeng and Wang, 2022; Woolley et al., 1946). Extreme rainfall events are categorised in decreasing magnitude from cloudbursts to downpours to showers. Woolley et al. (1946, p. ii) introduced the term 'cloudburst' as "...a torrential downpour of rain which by its spottiness and relatively high intensity suggests the bursting and discharge of the whole cloud at once". Although, most of the events studied by Woolley et al. (1946) are now considered as downpours, this study led to the recognition that such extreme rainfall events, together with specific hydrological conditions, can lead to flash floods (Doswell et al., 1996; Knos et al., 2022). For example, in South Africa (SA), a downpour from a severe thunderstorm at Linksfield, Johannesburg, on 9 November 2016 caused extensive damage to roads and hundreds of vehicles with the tragic loss of 7 lives (SAWS, 2017). The City of Johannesburg Metropolitan Municipality reports that this event also severely impacted 862 people from 373 households in the Setswetla informal settlement, next to the Jukskei River and downstream from Linksfield (Mvulane, 2020). Measured at the OR Tambo International Airport (ORTIA), which is 11 km to the east of the Linksfield disruptive rainfall event, the downpour produced 83 mm in the preceding hour ending at 18:10, with the highest 5-min rainfall amount of 17.2 mm recorded at 17:35.

The Glossary of Meteorology of the American Meteorological Society (AMS, 2015) uses a threshold of 100 mm/h rainfall intensity to identify a cloudburst event. Importantly, this definition is based on the rainfall total for the period of 1 hour. Very few rainfall events in SA exceed the threshold set by the AMS but some local events nevertheless cause significant amounts of damage, which makes it difficult to implement disaster risk reduction (DRR) strategies. Thus, the aim of this study was to define local rainfall intensity categories and to calculate percentiles and return periods of daily, hourly and 5-min rainfall storms to forecast them more accurately in the context of DRR for SA. The 99.9^th and 99.99^th percentiles are proposed as thresholds for forecasting extreme hourly and 5-min rainfall, respectively, for local municipalities, provinces and SA. In addition, the study addresses extreme rainfall events in SA in relation to current predictions for regional climate change, which suggest that extreme events will increase in both frequency and intensity (Seneviratne et al., 2021; Lehman et al., 2015).

 

DATA AND METHODS

In SA, the measurement of 5-min rainfall intervals became widespread by the South African Weather Service (SAWS) from 1994, as autographic gauges were replaced with electronic gauges. Currently, SAWS have 246 automatic weather stations (AWS) and 170 automatic rainfall stations (ARS) across the country. The Campbell-Scientific CR1000X data logger of the AWS is connected to a 0.2 mm resolution TB4MM tipping bucket gauge made in the USA that has the following specifications (Campbell Scientific Inc., 2024):

• Measurement range: Max. rainfall intensity = 700 mm/h (58.3 mm/5 min)

• Accuracy: ± 2% inaccurate at 250 mm/h (20.8 mm/5 min) and ± 3% inaccurate at 500 mm/h (41.7 mm/5 min)

In other words, the 5-min rainfall of 66.2 mm measured at 23:20 on 4 April 2000 at Hoedspruit, Limpopo (see Table A2, Appendix) could have been anything between 64.2 and 68.2 mm using a 3% inaccuracy. This illustrates the uncertainties in the measurement of extreme or highest 5-min rainfall events.

For this study, recorded 5-min, hourly and daily rainfall data from AWSs (see Table A1, Appendix) were quality controlled to find those stations with > 21 years continuous data. Since weather forecasting in SAWS is done provincially by 5 regional offices (ROs); the combining of the stations was done provincially and not according to climatic regions. In addition, to obtain a good spatial representation across SA, a weighted approach based on the area of the provinces was adopted. Based on these criteria, 64 AWSs were selected across SA with the following numbers per province: Northern Cape (NC) - 15, Eastern Cape (EC) - 8, Free State (FS) - 7, Western Cape (WC) - 7, Limpopo (LP) - 7, North West (NW) - 7, KwaZulu-Natal (KZN) - 6, Mpumalanga (MP) -5 and Gauteng (GP) - 2 (Table 1). The provinces of Mpumalanga and Gauteng, the two smallest provinces and conveniently co-located, are combined for averaging purposes (Fig. 1). Combining the provinces of Mpumalanga and Gauteng as a spatial unit facilitates the comparison of thresholds with the other provinces, because the average of just two stations for Gauteng does not compare well with the approximate average of seven stations for the other provinces. The combined spatial unit termed 'MG' places it in the same size category as the other provinces, apart from the Northern Cape, which covers approximately 30% of SA.

The 90^th, 99^th, 99.9^th and 99.99^th percentiles and highest rainfall values for each of the 64 selected stations from 1 July 1994 to 30 June 2021 were determined for the daily, hourly and 5-min datasets as follows: The rainfall datasets were sorted in descending order to link the date and time with the rainfall event. Highest daily, hourly and 5-min rainfalls were then checked for consistency, and faulty records were removed for each station. The highest values as well as 90^th, 99^th, 99.9^th and 99.99^th percentiles for the 64 stations were grouped by province and for MG (see Table A2, Appendix). The stations presented in Table 2 exclude the station at Rustenburg because its 5-min data were missing for its highest hourly rainfall storm. Table 2 summarises the maximum 5-min accumulated rainfall over an hour (M) for 63 stations and includes the following variables: date and time, M, highest 5-min rain rate, cloudburst classification, downpour minutes, daily rainfall total and percentage rainfall contributions of (i) hourly to daily, (ii) 5-min to hourly and (iii) 5-min to daily.

The American Meteorological Society's Glossary of Meteorology (AMS, 2015) uses a threshold of 100 mm/h rainfall or 8.33 mm/ 5 min rainfall intensity to identify a cloudburst. For this study, if the highest 5-min rainfall is less than or equal to 8.33 mm, it is regarded as a shower and when the highest 5-min rainfall is greater than 8.33 mm but is less than or equal to 25.0 mm, it is considered a downpour. When the highest 5-min rainfall is greater than 25.0 mm, it is considered a SA-cloudburst. Downpour minutes of a storm are the period for which the rainfall exceeds the downpour rainfall rate (8.33 mm/5 min). In addition, the expected maximum rainfall amounts in return periods of 10, 25, 50 and 100 years for 5-min, hourly and daily rainfalls were calculated based on annual maximums from 1995 to 2021 for the 64 stations (Table A3, Appendix). The Fisher-Tippett Type I statistical or Gumbel distribution for extreme values was used following similar short-duration rainfall analysis studies (see WB, 1983; Reich, 1961; Gumbel, 1967). During the quality control process, the years 2000 and 2010 were removed for Komatidraai's return period calculations because of missing data. Estimates for parameters of the Gumbel distribution are calculated using the method of moments approach (Kruger, 2022). The formulae use the average (x) and population standard deviation (s) of the rainfall extremes (r) for each of the daily, hourly and 5-min datasets for each station. The cumulative density function (Gumbel) is:

The Gumbel function F when calculated with an Excel spreadsheet gives the return period in years for rainfall in mm. The cumulative density function can be written such that estimated extreme rainfall (r) is calculated for a given return period (F), namely: r(F) = z - b ln. The return periods presented in Table A3 (Appendix) are the first published for 5-min rainfall observations in SA.

 

RESULTS

The discussion of the highest rainfall and percentile values presented in Table 1 follows for daily, hourly and 5-min rainfall and the investigation of rainfall forecasting thresholds.

Daily data

Normal annual rainfall across SA varies from below 200 mm in the west to above 1 000 mm along the east coast (Kruger, 2007). This increasing trend in rainfall amount and intensity from west to east is confirmed by the trend in the average highest and average 99^th percentile of daily rainfall amounts for provinces presented in Table 1. The weighted average for the 99^th percentile of daily rainfall is 46.8 mm which is close to the 50 mm criterion that is conventionally used to issue a warning for disruptive rainfall in SA, although, the 50 mm threshold for heavy or extreme daily rainfall is an over-estimation for the Cape provinces (Northern, Western and Eastern) and the Free State, but an under-estimation for the remaining provinces in SA. Nevertheless, the 50 mm or slightly lower 47 mm is a reasonable indicator of heavy or extreme daily rainfall for SA. The 99^th percentile values for daily rainfall amounts (Table 1) indicate that more extreme rainfall events occur in the east and north-east of SA while less extreme rainfall events occur in the southern interior and western parts of the country (see Fig. 2). The average of the 99.9^th percentile of daily rainfall totals for SA is 92.6 mm. It is important to note that the stations at Vioolsdrif and Augrabies Falls in the NC did not receive enough rainfall to calculate the 99.9^th percentile of daily rainfall for these stations (see Table A2, Appendix) and were thus excluded from this calculation for SA. Fortunately, the highest recorded rainfall totals for storms at Vioolsdrif and Augrabies Falls are available (see Table A2, Appendix). These highest rainfall totals are regarded as extreme rainfall indicators.

Hourly data

The largest value for the average highest hourly rainfall of 69.5 mm for Mpumalanga and Gauteng (MG) demonstrates that these provinces experience the most convective rainfall (see Fig. 3). The 40.2 mm calculated for the Western Cape, representing the smallest average highest, demonstrates that it experiences the least convective and more stratiform rainfall (see Fig. 3). There is a 29.3 mm increase in average highest hourly rainfall from the southwest (WC) to north-east (MG) over SA, indicating the contrast between summer convective and winter stratiform rainfall. The average highest hourly rainfall of 53.9 mm for all provinces indicates that hourly rainfall greater than 53.9 mm should be considered an exceptional event, especially for drier provinces.

The weighted average for the 99^th percentiles of the data is 13.9 mm/h, which is close to the criterion of > 12.6 mm/h for a heavy rainfall rate or intensity used by WB (1982). A rainfall intensity between 2.5 and 12.5 mm/h is regarded as moderate and less than 2.5 mm/h is a slight rain intensity (WB, 1982). Figure 4 indicates the scarcity of heavy (> 12.6 mm/h) rainfall events over the western, southern and south-eastern parts of SA. The 29.9 mm/h value, which is the weighted average for the 99.9^th percentile, is a useful indicator of extreme hourly rainfall intensity, but it is important to recognise the spatial variation between the provinces, i.e. 20.7 mm/h in the WC to 41.2 mm/h in LP. The 99.9^th percentile values indicate that rainfall of above 30 mm/h occurs for most of the north-eastern part of SA, with below 30 mm/h typically occurring over most of the central, western and southern parts (Fig. 5).

Five-minute data

The value of 22.0 mm/5 min recorded in the LP is the largest average highest value and indicates that the LP experiences the most intense convective rainfall. Conversely, the WC is shown to experience the least intense rainfall with the smallest average highest value of only 11.3 mm/5 min, which again demonstrates that it experiences mostly stratiform rainfall. The weighted average of the highest values for all provinces is 15.5 mm/5 min. Figure 6 indicates a trend with the largest values in the northern and eastern provinces and the smallest in the southern and western provinces of SA. This same trend is observed for highest hourly rainfall values over SA (see Fig. 3 for comparison).

The 99.9^th percentiles for the 5-min data indicate that 8.9 mm and 5.1 mm are the largest and smallest average highest values for MG and WC, respectively. The 99.9^th percentiles of 5-min data that are > 8.0 mm occur in LP, NW, KZN and MG and are comparable to > 8.33 mm/5 min, which is equivalent to the 100 mm/h definition of a cloudburst by the American Meteorological Society's Glossary of Meteorology (AMS, 2015). Thus, the rates of 8.0 mm/5 min or 96 mm/h are equivalent to the 1.6 mm/min threshold which was used to identify an extreme precipitation event over Southern China on 7 May 2017 (see Zeng et al., 2021). Figure 7 indicates that 5-min values > 8 mm occur over most of the central, northern and eastern parts of SA, with < 5 mm in places over the extreme western and southern parts of the country. This observation illustrates the scarceness of these intense events in the stratiform rainfall region of SA.

The 99.99^th percentile values are an indicator of the most extreme or highest rainfall events. The calculated 18.8 and 8.5 mm represent the largest and smallest averages for the LP and WC provinces, respectively. The weighted average of 13.0 mm for the 99.99^th percentiles is observed for all provinces, apart for the NC. The NC is not included in Table 1 because it has an average of only 12.1 mm for the 12 stations with sufficient rainfall data to calculate the 99.99^th percentile. For instance, Alexander Bay, Vioolsdrif and Augrabies Falls, all bordering Namibia, have insufficient rainfall to calculate the 99.99^th percentile for these locations. Thus, the average 99.99^th percentile for rainfall over 5 min for the remaining 61 stations across SA is 12.8 mm, which is rounded up to 13 mm. The 99.99^th percentiles' weighted average of 13.0 mm/5 min is equivalent to the extrapolated threshold of 2.6 mm/min for extreme precipitation presented by Zeng et al. (2021). The 99.99^th percentile values that are > 13.0 mm for LP, NW, KZN and MG provinces indicate that these provinces experience the most extreme 5-min rainfall events.

Comparison of M with clock hourly rainfalls and classification of M's storms

For 34.9% of the 63 stations (see Table 2), the date of the maximum 5-min accumulated hourly rainfall (M) amount is different to the date of the highest hourly rainfall amount that is listed in Table A2 (Appendix). These 22 differently dated events or storms are indicated in bold in Table 2. In other words, 65.1% of the clock hourly events are also the M events for a station. The M value is on average 8.2 mm larger than the highest hourly rainfall for the 63 stations considered. The difference between the maximum accumulated and highest hourly rainfall totals varies from 0 to 36.8 mm. The 36.8 mm difference for Port Edward is for two different events, but it can be higher for an individual event; for example, the storm on 3 January 1997 at Port Edward had an hourly rainfall amount of 77.2 mm at 22:00, resulting in a difference of 37.8 mm. The M value is on average 10.8 mm larger than the highest hourly rainfall for each of the storm events for the 63 stations. The highest hourly rainfall for a station is only equal to M for 6.4% of the time, namely for the four underlined stations in Table 2. The highest hourly rainfall totals, dates and hours for the remaining 59 stations are shown in Table A2 (Appendix).

Table 2 shows that two storm events are classified as SA-cloudbursts, namely:

• The severe thunderstorm of 4 April 2000 at Hoedspruit that produced 132.2 mm in 25 min from an intense upper-air trough system (NOAA PSL, 2000). It follows from the cloudburst definition (> 25 mm/5 min) that the storm is a 25-min cloudburst because 132 > 125 mm over 25 min. The storm could also be regarded as a 50-min downpour of 157.8 mm. According to WMO-cloudburst definition (> 100 mm/h), this storm, as well as the Pretoria Unisa, Port Edward and Bloemhof storms, are classified as WMO-cloudbursts, although in the context of SA, the latter three events are classified as downpours because their highest 5-min rainfalls < 25 mm (Table 2). The intense upper trough system that resulted in the Hoedspruit cloudburst also caused an M of 67.6 mm at Nelspruit 6 h earlier, from a thunderstorm (see Table 2). The Nelspruit 20-min downpour of 33.6 mm was confirmed by the Ermelo radar system.

• The severe thunderstorm of 12 November 2001 at Kuruman, NC, that produced 30.2 mm in 5 min, due to an upper air perturbation system (NOAA PSL, 2001), is the other event which can be considered a SA-cloudburst. The storm is a 5-min cloudburst because 30 > 25 mm, and a 35-min downpour of 62.0 mm that caused flooding in the town.

The maximum hourly rainfall amount of 133.8 mm recorded at the Unisa weather station in Pretoria overnight on 18 to 19 January 1996 is the largest storm classified as a downpour (see Table 2). This severe storm, which produced 185.8 mm in 110 min, caused serious flooding damage in Pretoria (SAWS, 1996). The storm was caused by a high-pressure system ridging to the south of SA with an associated upper air trough (DEAT, 1996). The afternoon radiosonde ascent at Irene, 15 km south of Pretoria, on 18 January 1996 indicated a precipitable water (PW) content of 27.4 mm. A PW > 25 mm means high energy air availability and downpours or cloudbursts can be expected. High energy air is moist warm air usually from a tropical origin and these conditions typically arise over SA in summer (Tyson, 1986). A map of the maximum 5-min accumulated hourly rainfall amounts (see Fig. 8) depicts downpours > 80 mm in places over the northern and eastern provinces of SA. The map further shows downpours with < 30 mm in places over the western and south-western parts of the country. Figure 8 compares well with the map of highest hourly rainfalls (see Fig. 3).

The Hoedspruit cloudburst of 66.2 mm in 5 min on 4 April 2000 is the all-time highest 5-min rainfall recorded until 30 June 2023 for SA. Associated with this event, the 10-, 15- and 30-min rainfall totals of 121.0, 128.8 and 132.2 mm, respectively, are also the all-time highest records for SA. Furthermore, the recorded 66.2 mm in 5 min is greater than the 63.0 mm recorded at Porto Bello, Panama, on 29 November 1911 - the current all-time global 5-min rainfall record according to literature (Burt, 2007; Cornthwaite, 1919). Although, the 66.2 mm recorded is less than the estimated 129.5 mm for the global maximum 5-min rainfall total given by a regression equation developed by Ramage (1995).

Return periods

Table A3 (Appendix) indicates that for GP, the expected maximum rainfall for a 10-year return period is 12.7 and 50.9 mm for 5-min and hourly rainfall events, respectively. These values compare well with the highest 5-min and hourly rainfalls of 11.0 and 50.6 mm, respectively, at Bolepi House in Erasmusrand, Tshwane, GP, that has a 15-year climate record (Vermeulen, 2022). Table A3 (Appendix) further indicates that every 100 years one can expect hourly rainfall storms for Pretoria (GP) and George (WC) that produce 79.6 and 39.6 mm, respectively. The more than double calculated value for Pretoria versus George illustrates the difference between the summer convective rainfall regime of the GP and the winter stratiform rainfall regime of the WC. The return periods calculated from the Gumbel distribution for the Hoedspruit and Kuruman SA-cloudbursts as well as for the Pretoria, Port Edward and Bloemhof downpours are presented in Table 3.

The SA-cloudbursts at Hoedspruit and Kuruman are 1-in-900-and 1-in-300-year events, respectively, based on their F rainfall amounts. Hoedspruit became a WMO-cloudburst in 10 min with its 121.0 mm for the highest 10-min rainfall total. Kuruman never became a WMO-cloudburst because its M is < 100 mm/h. However, the three downpours are all WMO-cloudbursts. The rarity of the Pretoria downpour lies in its hourly rain that is a 1-in-1 000-year event, while its highest 5-min rainfall is only a 1-in-14-year event. The downpours at Port Edward and Bloemhof are 1-in-200- and 1-in-600-year events, respectively, based on their M values. The average return periods for the two SA-cloudbursts based on 5-min and hourly data are 647 and 1 092 years, respectively. The average return periods for the three downpours based on 5-min and hourly data are 65 and 2 923 years, respectively. The hourly events are the rarest for both SA-cloudburst and downpour groups. The average return period of the 5-min events of the SA-cloudbursts is an order of magnitude larger than the value of the downpours. Thus, the rarity of a SA-cloudburst lies in its extremely high rainfall intensity, namely F > 25.0 mm. The average F of 48.2 mm for the two SA- cloudbursts is much larger than the 16.3 mm average F for the three downpours.

 

DISCUSSION

Currently, rainfall above 50 mm/day is regarded as the 99^th percentile threshold to identify heavy or extreme daily rainfall in SA and the world (Goswami et al., 2010; Hitchens et al., 2012). However, the weighted average of the 99^th percentile of values for daily rainfall across SA is 46.8 mm and is a more appropriate threshold for extreme daily rainfall in the South African context, although it must be noted that there is significant spatial variation of daily rainfall across SA. For example, the average of 99^th percentile values for KZN is 61.7 mm/day, which is significantly greater than the suggested value of 46.8 mm/day to delimit extreme rainfall in SA. Consequently, the suggested extreme rainfall criteria for SA underestimates heavy rainfall in the eastern parts of South Africa such as KZN, and 62 mm would be a more appropriate criterion to classify extreme rainfall for this province. Similarly, the average of 35.0 mm/day for the 99^th percentile of recorded values for the NC is considerably less than 46.8 mm/day. Therefore, rainfall from a storm in NC should only exceed 35 mm/day to be regarded as extreme. Consequently, regional statistics should be adopted to classify extreme rainfall events across SA to better prepare strategies for DRR.

Some countries, such as China, use the 99.9^th percentile value of hourly rainfall as the threshold to identify extreme hourly rainfall (Luo et al., 2016). Even though there is spatial variation in the average of 99.9^th percentile values for hourly rainfall across SA, from 20.7 mm in the WC to 41.2 mm in the LP, the average value of 29.9 mm/h (rounded to 30 mm/h) is an appropriate criterion for extreme hourly rainfall and a useful threshold for forecasting flash floods in SA. The calculated averages of the 99.9^th percentile values and 99^th percentile values for 5-min rainfall across SA are 7.3 mm and 3.5 mm, respectively. The spatial variation of average 99.9^th percentile values ranges from 5.1 mm/5 min in the WC to 8.9 mm/5 min in the provinces of Mpumalanga and Gauteng. The average 99.99^th percentile value for 5-min rainfall is 12.8 mm (rounded to 13 mm/5 min). The average values of the 99.99^th percentile range from 8.5 mm in the WC to 18.8 mm in LP. Thus, if the average rate of 13 mm/5 min of rainfall is sustained over an hour, it would result in a rainfall event of156 mm/h. The maximum hourly rainfall amount of 133.8 mm recorded at Pretoria (see Table 2) is the largest downpour and has a highest 5-min rainfall rate of 13.4 mm. The 13 mm/5 min indicator of extreme 5-min rainfall events is also the extrapolated equivalent of 2.6 mm/ min - the threshold for extreme sub-hourly rainfall used by Zeng et al. (2021). Thus, there is a 1 in 10 000 probability (0.01% chance) that a daily rainfall event will be a cloudburst (Bharti et al., 2016), meaning a 99.99^th percentile event. Therefore, the calculated value of 12.8 mm, representing the 99.99^th percentile of 5-min rainfall, gives an indication of a possible cloudburst event in SA.

The classification in Table 2 indicates that the Hoedspruit and Kuruman storms of 4 April 2000 and 12 November 2001, respectively, should be regarded as cloudbursts because their highest 5-min rainfall amounts exceed 25 mm. According to the WMO cloudburst definition (> 100 mm/h) there are 4 cloudbursts in Table 2, namely: Bloemhof, Port Edward, Pretoria and Hoedspruit. The Bloemhof, Port Edward and Pretoria storms are re-classified as downpours because their highest 5-min rainfalls < 25 mm. According to the WMO definition, the Kuruman event is not a cloudburst, but is a SA-cloudburst because the highest 5-min rainfall amount > 25 mm. Dimri (2017) proposes > 100 mm/15 min for a Himalayan cloudburst. Therefore, only the event recorded in Hoedspruit with 128.8 mm/15-min qualifies as a cloudburst according to the criterion used by Dimri (2017). The 132.2 mm for the 25 min duration of the Hoedspruit storm is interpolated to 5.3 mm/min which exceeds the 2.6 mm/min threshold for the extreme sub-hourly rainfall rate used by Zeng et al. (2021) to indicate possible flash flooding.

Using the data for the 64 stations across SA from July 1994 to June 2021, the median value of the return periods for WMO-defined cloudbursts, SA-defined cloudbursts and downpours are 38 000, 16 000 and 2 years, respectively. Meaning that WMO- and SA-defined cloudbursts are extremely rare, while a downpour event can be expected once every 2 years across the whole of SA. The minimum value for a return period is 0.5 years at Pietermaritzburg in KZN while Tshwane in GP, experienced 14 downpour events over a 15-year period at Bolepi House in Erasmusrand (Vermeulen, 2022).

According to Dimri (2017), an hourly rainfall > 70 mm is indicative of a possible WMO-cloudburst. The highest hourly rainfalls of the 64 AWSs in Table A2 (Appendix) was compared with 57 autographic rainfall stations of WB36 (WB, 1983) for the average period from 1954 to 1972. Subsequently, the frequency of stations which met the following criteria was calculated:

• Highest rainfall > 70 mm/h: 14 and 12 stations for the WB36 and this study, respectively. The two differences between the frequencies of the two climate periods compared is statistically insignificant according to Wilcoxon's sum of ranks test (Langley, 1968). Importantly, the Wilcoxon's test shows a less than 90% chance of a significant difference.

• Highest rainfall > 100 mm/h: 1 and 2 stations for the WB36 and this study, respectively. The 1 September 1968 COL-storm (Hayward and Van den Berg, 1968) at Gqeberha, formerly known as Port Elizabeth, had a highest intensity of 112 mm/h, while the Hoedspruit and Pretoria Unisa storms have highest intensities of 132 and 102 mm/h, respectively.

The average highest intensity of 117.2 mm/h for the last two mentioned events compares well with Gqeberha at 112.4 mm/h. The WMO-cloudburst index cannot be considered for the Wilcoxon's sum of ranks test because there are not enough cases. Therefore, this study shows no statistically significant change from 1954 to 2021 in the intensity and frequency of cloudburst storms. It needs to be noted that the change in the rainfall recording instrumentation from autographic to electronic instruments during 1994 may affect the homogeneity of the datasets being compared.

The percentage rainfall contributions of M to daily (PCM2D), 5-min to M (PCF2M) and 5-min to daily (PCF2D) are presented in Table 2. On average, 76% and 16% of the daily rainfall occurs over 1 h and 5 min, respectively, for the 63 stations considered. Therefore, for a 50 mm/day forecast, maximums of 38 mm/h and 8 mm/5 min, respectively, can be expected for the storm events. On average 20% of the hourly rainfall in SA occurs in just 5 min. Therefore, for a 50 mm/h forecast, a maximum of 10 mm/5 min can be expected for the storm event. These calculations have implications for (i) planning DRR strategies, and (ii) evaluating severe storms. For example, in the early morning hours of 9 December 2022 a severe thunderstorm produced 176.7 mm rainfall in 5 hours at Florida in Johannesburg, GP. The resulting flash floods from this storm caused damage to roads and houses and cars were washed away. The averages of both PCM2D and PCF2D, when applied, give estimates of 134 mm/h and 28.3 mm/5 min for M and F, respectively.

Comparison with international downpours

The maximum 5-min accumulated hourly values for SA compare well with international events (see Table 4). The M values for Port Elizabeth and India indicated in bold italics are estimates calculated by adding 10.8 mm to their clock hourly values. The average diflerence between the maximum 5-min accumulated and clock hourly rainfall is 10.8 mm for the 63 stations across SA. The difference of 32.9 mm between maximum minute accumulated and clock hourly rainfall for the South China station compares well with the 37.8 largest difference of SA's Port Edward. The 'daily' and 'highest 5-min' values, when available for international stations, compare well with SA stations (see Table 4), except for the 66.2 mm/5 min of Hoedspruit that is classified as a once-in-900-years storm (see Table 3). Application of the 20.2% for the average PCF2M of the 63 stations (see Table 2) to the 30 October 2015 storm in Texas, USA, gives estimates of F (see bold in Table 4).

The F values of 35.8 and 29.6 mm for LCRA-Gauge and Austin-Bergstrom, respectively, show that this WMO-cloudburst storm is possibly a SA-cloudburst storm.

 

CONCLUSION

The rainfall rates of 46.8 mm/day and 13.9 mm/h, based on the 99^th percentile values, are suggested as more appropriate values to classify heavy or extreme rainfall across SA. In addition, the 99.9^th percentile values of 30 mm/h and 7.3 mm/5 min, which are representative of a 1-in-1 000 event provide useful thresholds for the forecasting of flash floods, and to develop appropriate DDR strategies in SA. Furthermore, the 99.99^th percentile value of 13 mm/5-min, which is representative of a 1-in-10 000 event is an indicator of a possible cloudburst storm in SA. The study also demonstrates the vast spatial variability observed in extreme rainfall rates across various temporal scales, specifically between the eastern and western parts of SA.

Importantly, a SA-cloudburst is defined as an hourly rainfall storm with a sub-hourly rainfall rate of > 25 mm/5 min. It is noteworthy that the 66.2 mm recorded in 5 min during a cloudburst at Hoedspruit on 4 April 2000 is the new all-time 5-min rainfall recordholder of the world. The calculated expected maximums of 5 min, hourly and daily rainfall totals in SA for return periods of 10, 25, 50 and 100 years presented here are a first for SA and will be useful for local hydrological modelling in the context of DRR, and specifically for flooding. The study also demonstrates the usefulness of recording 5-min data across SA to assess the maximum 5-min accumulated hourly rainfall (M) which can be used to evaluate hourly rainfall derived from numerical weather prediction (NWP) models. Lastly, evaluation of extreme rainfall events across SA is crucial to provide a better understanding of regional climate, particularly within the context of climate change which predicts increases in the intensity and frequency of extreme events (Seneviratne et al., 2021). However, comparison between the periods of 1995 to 2021 and 1954 to 1972 (WB36: WB, 1983) does not indicate a statistically significant change in the frequency and intensity of extreme rainfall events across SA.

 

AUTHOR CONTRIBUTIONS

Jan H Vermeulen - conceptualisation and methodology of the study, data collection and fieldwork, sample/data analysis, interpretation of results, writing of the initial draft, revision after review ; David W. Hedding - methodology of the study, sample/ data analysis, interpretation of results, revision of the initial draft, revision after review; Nthabiseng Letsatsi - generating the rainfall maps and field work.

 

ACKNOWLEDGEMENTS

The following people are gratefully acknowledged: Mr Ishaam Abader, CEO of SAWS, Dr Jonas Mphepya, Executive: Weather and Climate, Ms Vanetia Phakula, Acting Senior Manager Disaster Risk Reduction and Mr Ezekiel Sebego for the sabbatical leave to do the research work, Karin Oxley and Anastasia Demertzis, SAWS Library, Lucky Dlamini, SAWS Climate Service, for supplying the 5-min, hourly and daily rainfall data, Dr Andries Kruger for return period formulas and quality control of storm events, Charlotte McBride and the team within Climate Service who discovered the world record holder, at 66.2 mm, for highest 5-min rainfall.

 

REFERENCES

AMS (American Meteorological Society) (2015) AMS Glossary of Meteorology. URL: http://glossary.ametsoc.org/wiki/Main_Page (Accessed November 2022).         [ Links ]

BURT CC (2007) Extreme Weather - A Guide and Record Book. W. W. Norton & Company, New York.         [ Links ]

BHARTI V, SINGH C, ETTEMA J and TURKINGTON TAR (2016) Spatiotemporal characteristics of extreme rainfall events over the Northwest Himalaya using satellite data. Int. J. Climatol. 36 3949-3962. https://doi.org/10.1002/joc.4605        [ Links ]

CAMPBELL SCIENTIFIC INC. (2024) Product manual. Tipping bucket rain gages. Revision 2/2024, Campbell Scientific Inc, USA, Logan UT. 7 pp.         [ Links ]

CORNTHWAITE HG (1919) Panama rainfall. Mon. Weather Rev. 47 (5) 298-302. https://doi.org/10.1175/1520-0493(1919)47<298:PR>2.0.CO;2        [ Links ]

DEAT (Department of Environmental Affairs and Tourism, South Africa) (1996) Weather Bulletin January 1996. Department of Environmental Affairs and Tourism, Pretoria.         [ Links ]

DIMRI AP, CHEVUTURI A, NIYOGI D, THAYYEN RG, RAY K, TRIPATHI SN, PANDEY AK and MOHANTY UC (2017) Cloudbursts in Indian Himalayas: A review, Earth-Sci. Rev. 168 1-23. https://doi.org/10.1016/j.earscirev.2017.03.006        [ Links ]

DOSWELL CA III, BROOKS HE and MADDOX RA (1996) Flash flood forecasting: An ingredients-based methodology. Weather Forecast. 11 560-581. https://doi.org/10.1175/1520-0434(1996)011<0560:FFFAIB>2.0.CO;2        [ Links ]

GOSWAMI B, MUKHOPADHYAY P, MAHANTA R and GOSWAMI BN (2010) Multiscale interaction with topography and extreme rainfall events in the northeast India region. J. Geophys. Res. Atmos. 115 D12114. https://doi.org/10.1029/2009JD012275        [ Links ]

GUMBEL EJ (1967) Statistics of Extremes. Columbia University Press, USA. 375 pp.         [ Links ]

HAYWARD QL and VAN DEN BERG HJC (1968) Die Port Elizabeth-stortreëns van 1 September 1968, Weather Bureau News Letter No. 234. Weather Bureau, Pretoria.         [ Links ]

HITCHENS NM, BALDWIN ME and TRAPP RJ (2012) An object-oriented characterization of extreme precipitation-producing convective systems in the Midwestern United States. Mon. Weather Rev. 4 1356-1366. https://doi.org/10.1175/MWR-D-11-00153.1        [ Links ]

KRUGER AC (2007) Climate of South Africa - Precipitation, WS47. SAWS, Pretoria. p. 4, p. 29.         [ Links ]

KRUGER AC (2022) Statistics in Climatology WKD711. BSc. (Hons) Meteorology, University of Pretoria, p. 47 pp.         [ Links ]

LANGLEY R (1968) Practical Statistics. Pan Books Ltd., London, 399 pp.         [ Links ]

LEHMAN J, COUMON D and FRIELER K (2015) Increased record-breaking precipitation events under global warming. Clim. Change 132 501-515. https://doi.org/10.1007/s10584-015-1434-y        [ Links ]

LUO Y, WU M, REN F, LI J and WONG W (2016) Synoptic situations of extreme hourly precipitation over China. J. Clim. 29 (24) 8703 -8719. https://doi.org/10.1175/JCLI-D-16-0057.1        [ Links ]

MVULANE P (2020) Settling on water pathways: A case study of Setswetla vulnerability to flash floods. Unpublished MSc. dissertation, University of the Witwatersrand. 155 pp.         [ Links ]

NIELSEN ER and SCHUMACHER RS (2018) Dynamical insights into extreme short-term precipitation associated with supercells and mesovortices. J. Atmos. Sci. 75 2983-3009. https://doi.org/10.1175/JAS-D-17-0385.1        [ Links ]

NOAA PSL (National Oceanic and Atmospheric Administration -Physical Sciences Laboratory, USA) (2000) NCEP/DOE Reanalysis II. URL: https://psl.noaa.gov/data/gridded/data.ncep.reanalysis2.html (Accessed August 2022).         [ Links ]

NOAA PSL (National Oceanic and Atmospheric Administration - Physical Sciences Laboratory, USA) NCEP (2001) NCEP/DOE Reanalysis II. URL: https://psl.noaa.gov/data/gridded/data.ncep.reanalysis2.html (Accessed August 2022).         [ Links ]

RAMAGE CS (1995) Forecasters guide to tropical meteorology (AWS TR 240 updated), AWS/TR-95/001. Air Weather Service, U.S. Air Force, AWSTL. 392 pp.         [ Links ]

REICH BM (1961) Short duration rainfall intensity in South Africa. S. Afr. J. Agric. Sci. 4 (4) 589-614.         [ Links ]

SAWS (South African Weather Service) (1996) Climate summaries of South Africa, January. South African Weather Service, Centurion. SAWS (South African Weather Service) (2016) Climate summaries of South Africa, November. South African Weather Service, Centurion. SAWS (South African Weather Service) (2017) Caelum - 1991 to 2016 (unpublished). Climate Service, South African Weather Service, Pretoria.         [ Links ]

SELLICK NP (1961) Intensity duration curves for rainfall in Rhodesia. Bibliography of Regional Meteorological Literature, Vol. II Southern Africa, 1948-1960. Rhodesia Nyasaland Meteorological Service, Salisbury. 1 224 pp.         [ Links ]

SENEVIRATNE SI, ZHANG X, ADNAN M, BADI W, DERECZYNSKI C, DI LUCA A, GHOSH S, ISKANDAR I, KOSSIN J, LEWIS S and co-authors (2021) Weather and climate extreme events in a changing climate. In: Masson-Delmotte V, Zhai P, Pirani A, Connors SL, Péan C, Berger S, Caud N, Chen Y, Goldfarb L, Gomis MI and co-authors (eds)Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change J. Cambridge University Press, Cambridge. 1513-1766.         [ Links ]

SMITH JA, BAECK ML, ZHANG Y and DOSWELL CA III (2001) Extreme rainfall and flooding from supercell thunderstorms. J. Hydrometeorol. 2 469-489. https://doi.org/10.1175/1525-7541(2001)002<0469:ERAFFS>2.0.CO;2        [ Links ]

THAYYEN RJ, DIMRI AP, KUMAR P and AGNIHOTRI G (2013) Study of downpour and flash floods around Leh, India during August 4-6, 2010. Nat. Hazards 65 (3) 2175-2220. https://doi.org/10.1007/s11069-012-0464-2        [ Links ]

TYSON PD (1986) Climate Change and Variability in Southern Africa. Oxford University Press, Cape Town. 220 pp.         [ Links ]

VERMEULEN JH (2022) An investigation of extreme very short-term rainfall at Bolepi House in Erasmusrand, Tshwane from May 2003 to April 2018: Implications for disaster risk reduction assessments in South Africa. SAWS Weather-Smart Newsletter, October. 21-26.         [ Links ]

WB (Weather Bureau) (1982) Weather codes for land stations (surface observations). Weather Bureau, Department of Transport, South Africa.         [ Links ]

WB (Weather Bureau) (1983) WB36. Part II. Extremes of rainfall, temperature and wind. Weather Bureau, Department of Transport, South Africa.         [ Links ]

WU M and LIN G (2017) The very short-term rainfall forecast for a mountainous watershed by means of an ensemble weather prediction system in Taiwan. J. Hydrol. 546 60-70. https://doi.org/10.1016/j.jhydrol.2017.01.012        [ Links ]

WOOLLEY RR, MARSELL RE and GROVER NC (1946) Cloudburst floods in Utah, 1850-1938, Geological Survey, Water Supply Paper 994. Government Printing Office, Washington DC. 128 pp.         [ Links ]

ZENG Z and WANG D (2022) On the local rain-rate extreme associated with a mesovortex over South China: Observational structures, characteristics and evolution. Mon. Weather Rev. 150 1075-1096. https://doi.org/10.1175/MWR-D-21-0033.1        [ Links ]

ZENG Z, WANG D and CHEN Y (2021) An investigation of convective features and Z-R relationships for a local extreme precipitation event. Atmos. Res. 250 105372. https://doi.org/10.1016/j.atmosres.2020.105372        [ Links ]

 

 

Correspondence:
Jan H Vermeulen
Email: janhverm@yahoo.co.uk

Received: 8 December 2023
Accepted: 15 October 2024

 

 

APPENDIX

 

Table A1 - Click to enlarge

 

 

Table A2 - Click to enlarge

 

 

Table A3 - Click to enlarge