The diurnal cycle of rainfall in South Africa in the austral summer

Authors

  • Mathieu Rouault,

    Corresponding author
    1. Department of Oceanography, Mare Institute, University of Cape Town, Cape Town, South Africa
    2. Nansen-Tutu Center for Marine Environmental Research, James Building, University of Cape Town, Cape Town, South Africa
    • M. Rouault, Department of Oceanography, Mare Institute, University of Cape Town, Cape Town, South Africa.
    Search for more papers by this author
  • Shouraseni Sen Roy,

    1. Department of Geography and Regional Studies, University of Miami, Coral Gables, FL 33124, USA
    Search for more papers by this author
  • Robert C. Balling Jr

    1. School of Geographical Sciences & Urban Planning, Arizona State University, Tempe, AZ 85287, USA
    Search for more papers by this author

Abstract

The diurnal cycle of precipitation over South Africa during summer is analysed for the first time using hourly precipitation data from 103 stations for the period 1998–2007. Using harmonic analysis, we found the presence of a distinct diurnal cycle over most of South Africa regarding both the frequency and amount of precipitation events. The standardized amplitudes, indicative of the strength of the diurnal cycle across a region, are strongest over the interior and along the east coast of South Africa with up to 70% explained variance associated with the diurnal cycle. The time of maximum precipitation is late afternoon to early evening in the interior, and midnight to early morning along the Agulhas Current as well as inland in the northeast of the county. The proximity of the warm Agulhas Current plays a role in the diurnal cycle of rainfall at coastal stations. There is an early morning maximum in precipitation in the South West of the country with small amplitude in the diurnal cycle there. On average, the peak in precipitation amount leads the peak in frequencies of precipitation by 30 min to 1 h. Together with a high resolution climatological summer rain rate presented here and a detailed table for all stations, this study is a benchmark upon which model output or satellite estimate of the diurnal cycle can be compared within the region. Copyright © 2012 Royal Meteorological Society

1. Introduction

The diurnal cycle of rainfall is an important mode of variability within the global climate system. It is ultimately a manifestation of the atmosphere–ocean–land system's response to solar radiation, but diurnal precipitation patterns are compounded by any number of interacting physical and dynamic processes. The diurnal cycle is well documented over many of the world's landmasses and oceans, but significant gaps in the documentation remain for many regions. Maximum precipitation tends to occur in the late afternoon/early evening over land and in the early morning above the ocean. In both hemispheres, the amplitude of diurnal cycles is greater in summer than winter. Land surfaces experience much larger amplitude of diurnal cycle than the oceans, associated with intense surface heating in the interior of large landmasses, convergence of land and sea breezes in the coastal areas, and anabatic flows in valleys and highland areas (McGregor and Nieuwolt, 1998). However, variations from these general patterns occur in response to complex orography, local land–sea breeze circulations, mountain–valley breezes, and coastal–ocean–atmosphere interaction (Schwartz and Bosart, 1979; Kousky, 1980; Landin and Bosart, 1985; Mapes and Houze, 1993; Chen et al., 1996; Nesbitt and Zipser, 2003). Several studies have emphasized the presence of strong diurnal cycles related with localized convective processes in low latitude tropical areas (Houze et al. 1981; Albright et al. 1985; Machado et al. 1993). Furthermore, the interaction between topography and convection leads to more pronounced diurnal cycles in precipitation (Carter and Elsner, 1996; Sen Roy and Balling, 2004). Ramesh Kumar et al. (2006) examined the diurnal cycle of precipitation over the eastern equatorial Indian Ocean from October 2001 to 2003. Their results indicate a peak in active precipitation spells during the afternoon hours, with more than 80% of the variance of hourly rainfall explained for all seasons by the diurnal harmonics.

Much of the precipitation received in the summer rainfall region of South Africa (Figure 1) is of convective origin forced by large-scale dynamics (Tyson and Preston-White, 2000). South Africa has a subtropical climate and is affected by temperate and tropical weather systems. The country comes under the influence of the Southern Hemisphere high pressure system, but in summer, a heat low is found over the interior. Most of the interior lies on an elevated plateau and orography plays an important role (Tyson and Preston-White, 2000). A marked sea surface temperature gradient exists along the South African coast from the west, where a strong upwelling cools the coastal area, to the east where the Agulhas Current brings warm water southwards (Figure 2). In Figure 2, we used the 1984–2007 mean sea surface temperature in austral summer (November to March) estimated from daily nighttime NOAA satellite remote sensing available at 4-km resolution (Kilpatrick et al., 2001). The southwest region and the west coast get most of its rainfall in austral winter through temperate systems while the rest of the country gets most of its rainfall in austral summer. Summer rainfall is caused by a large-scale synoptic system leading to convection, but little is known about the diurnal cycle of rainfall. In fact, the last study of the diurnal cycle for South Africa was undertaken by Hastenrath (1970) who used only four rainfall stations. The aim of this article is to document and analyse the diurnal cycle of precipitation in South Africa in summertime using 103 stations and also together with a high resolution climatological summer rain rate chart (Figure 1), provide a benchmark upon which model output or satellite estimate of the diurnal cycle can be compared with. We primarily use harmonic analysis as a statistical technique employed widely to study diurnal rainfall patterns in other regions around the world. Section 2 covers techniques and methods, Section 3 reports the results, Section 4 discusses those results, and conclusions are drawn in Section 5 while a detailed table with relevant information for each station is found in the Appendix.

Figure 1.

The 1950–2000 mean austral summer (November to March) rainfall in South Africa in mm d−1 (modified from Hewitson and Crane, 2005)

Figure 2.

The 1984–2007 mean sea surface temperature in austral summer (November to March)

2. Data and methodology

We obtained hourly precipitation data from the South African Weather Service for 139 stations located across South Africa for the period 1998 to 2007 (Figure 3). Initial analyses of the precipitation data revealed 36 stations which had more than 15% of missing data and therefore were not included in the final analysis. The nearest neighbour statistic was calculated for the spatial distribution of the remaining 103 stations as the ratio between the observed mean distance among the sites and the expected mean distance given a random distribution. The ratio of 1.14 for the selected network of stations is random to dispersed at the 0.01 level of statistical significance. Many previous studies of diurnal variations in precipitation have utilized harmonic analysis to detect the diurnal cycle in precipitation patterns (Wallace, 1975; Sen Roy and Balling, 2007). The basic form of the harmonic equation used in the analysis is:

equation image(1)

where the estimated precipitation frequency is equation image; θ is derived as 2πX/N with X as the hour, and N is the number of observations, i.e. 24. Φ is the phase angle of the harmonic curve, also reinterpreted as the time of maximum precipitation. is the average hourly frequency over the N observations, and r is the frequency or number of times the harmonic curve is repeated in 24 h. The standardized amplitude that mainly reveals the relative importance of a diurnal cycle at each individual station was calculated by Ar/2. The portion of variance explained is computed as equation image, with σ as the standard deviation of the 24-h precipitation frequency values. A useful parameter related to hourly precipitation probabilities may be generated when the standardized amplitude is multiplied by 2 and added to 1. For example, standardized amplitude of 0.10 implies that the probability of precipitation in the peak period is 1.20 times the 24-h mean value.

Figure 3.

Distribution of weather stations used in that study across South Africa with orography

The results of the harmonic analysis, including the first harmonic standardized amplitudes and the portion of explained variance at each station, were plotted on maps and interpolated using an ordinary Kriging method which produced the lowest root mean square error for the interpolated surface. The basic formula used in this technique is:

equation image(2)

where the predicted values Z(s) are calculated with µ as the mean, s is the location, and ε is the spatially autocorrelated error. Ordinary Kriging is a stochastic interpolator which takes into account the statistical relationships between the different observations while generating the final surface interpolations. It uses data of known points and its performance depends on the sample size (Hughes and Lettenmaier, 1981). The spatial patterns of the time of maximum were plotted using rotated arrows to depict the time during the 24 h of a day, when the maximum occurred in the first harmonic fit.

3. Results

The raw hourly data were compiled within four scale matrices, one each for frequency and amount of rainfall for the summer months of November to March which is when most of summer rainfall occurs (Rouault and Richard, 2003). The matrices consisted of 103 rows, one for each station, and 24 columns for the hours in a day. Each column in the frequency matrix was a count of the number of times a precipitation event greater than 0.1 mm was recorded at a given station in a given hour over the time period. In case of the amount matrix, each column consisted of the amount of precipitation occurring during that hour. The total number of precipitation events recorded during the summer months of the entire study period for all 103 stations was 221 510, and the total precipitation associated with these events was 371 753.8 mm. In Table AI of Appendix, names of station, latitude, longitude, altitude, percentage of missing data (%), average number of hourly events for the summer months of November to March (N), average summer total amount (TA), average summer amount divided by average summer number of events (R), amplitude of frequency (AF), variance explained for frequency (VF), time of maximum frequency (TMF), amplitude of amount (AA), variance explained for amount (VA), and time of maximum amount (TMA) are listed.

The results of the first harmonic analysis for frequency of precipitation are shown in Figure 4. The standardized amplitudes associated with the first harmonic are mapped in Figure 4(b), which is also a measure of the strength of the diurnal cycle. The strength of the diurnal cycle can also be assessed by the variance explained by the first harmonic that is shown in Figure 4(a). Given the marked diurnal cycle in the frequency of precipitation, the times of maximum values are plotted in Figure 4(c). An arrow pointing north indicates midnight; east, 0600 local standard time (LST); south, 1200 LST; and west, 1800 LST. Amplitude, variance, and time of maximum for the summer rainfall hourly amounts are shown in Figure 5.

Figure 4.

Results of harmonic analysis for frequency of precipitation events during austral summer (November to March): (a) map of variance explained (values greater than 0.25 significant at 0.01 level of confidence), (b) standardized amplitude, and (c) time of maximum in the diurnal cycle for 103 stations located across South Africa. An arrow pointing to north indicate midnight, pointing to the east: 0600 LST, to the south: 1200 LST, to the west 1800 LST. Standardized amplitude of 0.10 implies that the probability of precipitation in the peak period is 1.20 times the 24-h mean value

Figure 5.

Results of harmonic analysis for hourly amount of precipitation during austral summer months (November to March): (a) map of standardized amplitudes, (b) variance explained (values greater than 0.25 significant at 0.01 level of confidence), and (c) time of maximum in the diurnal cycle for 103 stations located across South Africa. An arrow pointing to north indicate midnight, pointing to the east: 0600 LST, to the south: 1200 LST, to the west 1800 LST. Standardized amplitude of 0.10 implies that the probability of precipitation in the peak period is 1.20 times the 24-h mean value

The maximum timing of frequency of hourly rainfall reveals several coherent patterns (Figure 4): a morning maximum in the west, a late afternoon and early evening maximum in the interior, and a night maximum along the coast. There is also a night maximum in the northeast. Amplitude and variance of hourly amount and frequency are low for the west coast and more pronounced over the interior and along the east coast. The west coast receives most of its rainfall in winter from May to September (Rouault and Richard, 2003) mainly through cold fronts and cut-off lows that are large-scale systems landing randomly from the east. Cold fronts shift poleward during austral summer. In that respect, it is interesting to observe a coherent diurnal signal there in summer, where time of maximum for frequency of rainfall occurs on average at 0600 LST and time of maximum for rainfall amount occurs on average about half an hour earlier. The time of maximum for precipitation frequency increases roughly from southwest (0400 LST) to northwest (0800 LST) (Figure 4(c)). Average amplitude for frequency and amount of rainfall is the lowest on the west coast of the study area (0.18 and 0.15, respectively) (Figures 4(b) and 5(b)). An analysis of hourly winter rainfall (not shown) also reveals a similar early morning maximum for precipitation frequency and amount, with even lower values for the standardized amplitude. We note that in other parts of the world and during cold seasons, precipitation has a much weaker diurnal cycle than in summer, with a morning maximum in winter over most land areas that is at least partially enhanced by the morning maximum in lower tropospheric relative humidity (Dai et al., 1999; Dai, 2001; Sen Roy and Balling, 2005) because higher relative humidity increases the formation of condensates. This could well apply to the south west region of South Africa.

Along the southern coast the maximum timing of frequency of summer rainfall veers regularly clockwise from early morning at Struiss Bay (34.80°S and 20.06°E) near Cape Agulhas, Africa's southernmost point, to around midnight from Port Elizabeth (33.98°S and 25.61°E) to Durban (29.97°S and 30.95°E) and back to early morning at the far east near the Mozambique border where coastal water are the warmest (Figure 4(c)). Maximum timing of maximum amount is on average an hour earlier but in fact increases from west (half an hour) to east (2 h). The amplitudes for the frequency and amount of rainfall during summer are low from Struiss Bay to Port Elizabeth (an average of 0.13 for frequency and 0.14 for amount of rainfall), and begin increasing at East London (−33.03°S and 27.83°E) where they remain higher (an average of 0.25 for frequency and 0.4 for amount) all the way to the east of our study area (Figures 4(b) and 5(b)). The Agulhas Current is quite close to the coast at East London and from there starts moving towards the west and away from the coast. It is positioned a few hundred kilometers from the coast at the southern tip of South Africa (Figure 2). The nocturnal coastal maximum is typical of a region influenced by a warm ocean, and it is similar to that observed in India and the United States (Sen Roy and Balling, 2005, 2007). Temperatures of 24–30 °C are found along South Africa's east coast in summer. Measurements of the Agulhas Current show substantial transfers of water vapour in the marine boundary layer, and a deepening of the marine boundary layer due to intense mixing and unstable atmospheric stability created by the advection of colder, drier air above the current (Lee-Thorp et al., 1999; Rouault et al., 2000, 2003). The intensity of mixing in the local boundary layer is such that even during anticyclonic subsident conditions, cloud lines can be observed above the current (Lee-Thorp et al., 1998). At last, the nocturnal maximum amplitude for frequency and amount of rainfall during summer in the vicinity of Durban is well documented and is attributed to the combined effect of orographic effect of the Drakensberg range and of the warm Agulhas Current (Tyson and Preston-White, 2000).

Except for the northeast, the interior has a clear diurnal cycle with maximum amplitude and variance explained there during the summer season. Average amplitude for frequency and amount is 0.20 and 0.28, respectively (Figures 4(b) and 5(b)). Timing of maximum amount rainfall occurs on average an hour before the timing of maximum frequency (Figures 4(c) and 5(c)). Average explained variance for frequency is 76 and 62% for amount (Figures 4(a) and 5(a)). The region of maximum frequency and amplitude coincides with the occurrence of times of maximum mostly from 1900 to 2300 LST, except for isolated locations or in the northeast where the maximum frequency and amount overlapped with times of maximum occurring from around midnight to early morning. In the northeast, average amplitude for frequency and amount are lower (0.16 and 0.25, respectively) (Figures 4(b) and 5(b)). We note that the northeast is lower lying than the interior and the rainfall is also lower (except for the Soutpansberg Mountains) as seen in Figure 1. Lightning is also less frequent in that part of South Africa than in the interior or along the east coast (Collier et al., 2006).

4. Discussion

The results are generally in agreement with those of previous studies done in other parts of the world (Wallace, 1975; Dai, 2001; Sen Roy and Balling, 2005; Sen Roy, 2009) showing a midnight to early morning maximum in the frequency of rainfall events in coastal areas, a late afternoon to early evening maximum over the continental interior, and a nighttime maximum in mountainous areas. All of the above-mentioned spatial patterns are predominantly the result of convection, local land/sea breeze circulations, orography, convection, and mountain–valley processes. Kousky (1980) also found nocturnal maxima in rainfall activity in the interior highlands of Brazil and linked the pattern to a mountain/valley breeze circulation that increased rainfall at night. Similar findings for the interior of Sumatra in the form of late-evening convective showers have been reported. Furthermore, Winkler et al. (1988) found a stronger diurnal signal during the summer season, with a greater spatial extent of nocturnal maxima during winter for the central and eastern United States, which was re-emphasized in an updated study for winter rainfall by Sen Roy and Balling (2005). Night to early morning maxima is found in the central United States (Wallace, 1975), where daytime subsidence inhibits convection (Dai et al., 1999), and India (Sen Roy and Balling, 2007). The interaction of orography and convectional processes, leading to pronounced diurnal cycles in rainfall, was also indicated in smaller island landmasses including Puerto Rico and Hawaii (Carter and Elsner, 1996; Sen Roy and Balling, 2004). In the case of coastal areas, most studies indicate a nocturnal to early morning maximum in the occurrence of rainfall (Schwartz and Bosart, 1979; Chang et al., 1995; Dai, 2001) as we found in that study. One of the significant results of our study is the 30 min to an hour lead (on average) in the peak time of occurrence of amount of precipitation over time of maximum for frequency of precipitation. This can be a result of numerous less intense precipitation events following a major rain event.

5. Conclusions

This study represents a considerable improvement on the study done by (Hastenrath, 1970) and based on only four rainfall stations. It also provides metrics that can be used to compare the diurnal cycle in South Africa with the rest of the world. It is also amelioration on global study done from satellite estimate (Dai et al., 2007) or by extrapolating global dataset of hourly rainfall (Dai, 2001) that did not show the diversity of diurnal cycles experienced in South Africa. Together with values of mean summer precipitation rate displayed in Figure 1 and the Table AI shown in Appendix, our study presents a benchmark upon which model output or satellite remote-sensing estimate of the diurnal cycle of rainfall can be compared. For instance, in their modelling study of the diurnal cycle in South Africa Tadross et al. (2006) had to used a 3 hourly rainfall estimate from satellite remote sensing to compare with their model output which led to inconclusive results. Furthermore, the diurnal cycle of rainfall in South Africa presents a rich variety similar to what can be found elsewhere in the world but over a smaller area. The geographic distribution of South Africa and the presence of the warm Agulhas Current play an important role in the diurnal cycle of the rainfall over South Africa. Further work will be needed to chronologically extend our dataset to better understand the behaviour of diurnal cycles from the start to the end of summers or during drought or wet periods, most of those being related to El Niño Southern Oscillation (Rouault and Richard, 2005) and also to understand the drivers of those distinct diurnal cycles in South Africa.

Table Appendix. Table AI. Name of station, latitude, longitude, altitude, percentage of missing data (%), average number of hourly events for the summer months of November to March (N), average summer total amount (TA), average summer amount divided by average summer number of events (R), amplitude of frequency (AF), variance explained for frequency (VF), time of maximum frequency (TMF), amplitude of amount (AA), variance explained for amount (VA), and time of maximum amount (TMA).
        FrequencyAmount
StationLatitudeLongitudez%NTARAFVFTMFAAVATMA
Alexanderbaai− 28.5716.53250.2960.70.100.110.610.310.2223.92
Beaufort Wes− 32.3622.588990.8851571.80.220.7319.260.380.7218.79
Bethlehem Wo− 28.2528.3316870.62905211.80.200.8621.240.220.5618.33
Bisho− 32.927.285907.925325110.070.7623.190.170.5521.04
Bloemfontein Stad− 29.1226.1814061.12094051.90.190.9321.480.250.8019.78
Bloemfontein Wo− 29.126.3135312074312.10.220.9121.780.270.8220.41
Bloemhof− 27.3925.3712282.32073861.90.150.8221.210.210.6319.08
Brandvlei− 30.4720.489224.836521.50.430.8219.850.620.7018.83
Calvinia Wo− 31.4819.769750.542581.40.210.5819.480.310.3917.26
Cape Town Wo− 33.9718.6420.371650.90.200.765.850.160.424.12
Charters Creek− 28.232.4284.12946112.10.170.922.950.160.731.87
De Aar Wo− 30.672412860.81061681.60.220.7519.870.350.8418.93
Durban Wo− 29.9730.95141.53495941.70.270.8523.080.380.8022.36
East London Wo− 33.0327.831160.43554751.30.110.720.230.150.6523.38
Elliot− 31.3327.85146333654161.10.280.8821.610.410.5220.39
Ellisras− 23.6827.78392.61453092.10.150.700.210.220.5422.88
Ermelo Wo− 26.529.9817690.62795371.90.250.8522.050.410.8419.69
Excelsior Ceres− 32.9619.439581.149771.60.080.1717.230.190.2515.99
Fort Beaufort− 32.7926.634552.62222941.30.160.7822.570.350.7819.50
Gaints Castle AWS− 29.2729.5217621.15156381.20.220.8520.100.460.8718.22
Garies-Groenrivier− 30.8617.57172.321150.70.350.516.170.370.437.47
Geelbek− 33.218.1270.144360.80.240.826.210.180.455.62
George wo− 34.0222.381910.82723511.30.140.792.810.170.402.46
Grahamstown− 33.2926.56420.426826710.130.8523.260.240.7321.15
Graskop AWS− 24.9330.8514369.13796151.60.170.941.960.220.4923.15
Greytown− 29.0830.610290.24055871.50.250.8822.990.470.8020.59
Hermanus− 34.4319.22130.31321401.10.140.895.700.110.474.36
Hoedspruit− 24.3531.055243.21733932.30.210.923.300.210.622.09
Irene Wo− 25.9128.21152632604861.90.160.7822.610.240.7520.41
Jamestown− 31.1226.8160210.11772841.60.340.8619.700.500.7818.88
JHB Bot Tuine− 26.152816248.52414041.70.110.7120.700.210.6419.95
Johannesburg Int Wo− 26.1528.2316950.52625522.10.180.7920.410.310.7118.79
Joubertina AWS− 33.8423.8654601751961.10.130.740.880.100.3918.74
Kathu− 27.6723.0111865.21262311.80.210.8021.290.360.7920.58
Kimberley Wo− 28.824.7711965.115230320.240.8319.980.350.7918.59
Klerksdorp− 26.926.62132216.82024162.10.140.6219.810.270.5817.95
Knysna− 34.0623.09540.12834931.70.150.744.090.080.226.89
Koingnaas− 30.1817.28991.223160.70.390.707.840.420.599.58
Komatidraai− 25.5231.918312.19719220.160.742.620.290.620.72
Kroonstad− 27.6327.2314343.12223931.80.210.8721.660.270.7819.91
Kuruman− 27.4423.45131512.21262381.90.120.5423.730.200.5122.04
Ladysmith− 28.5729.7710691.227353820.310.8522.100.500.8120.63
Laingsburg− 33.220.876560387620.230.7318.040.370.7316.49
Lambertsbaai Nortier− 32.0318.33940.635270.80.320.746.830.180.337.91
Langebaanweg AWS− 32.9718.17313.639350.90.220.767.940.180.369.33
Lichtenburg− 26.1326.1714853.11914002.10.170.7120.640.280.6518.09
Lydenburg− 25.1130.48143361252361.90.210.7720.500.240.5418.96
Mafikeng Wo− 25.8125.5412812.31813992.20.130.6021.290.340.6319.05
Malmesbury− 33.4718.721020504810.160.666.950.140.255.74
Margate− 30.8530.331541.341584120.210.8622.950.300.7121.94
Marken− 23.628.389963.41643502.10.160.760.110.220.6823.66
Mbazwana Airfield− 27.4832.6824.92645852.20.200.933.350.300.892.18
Mount Edgecombe− 29.731.05940.53736001.60.270.8623.750.430.7522.47
Mtunzini− 28.9531.7383.53897301.90.220.911.230.220.7223.60
Nelspruit− 25.530.9288316.628156820.220.970.520.270.9023.07
Newcastle− 27.7729.9812388.322244020.240.8721.420.480.8519.69
Noupoort− 31.1924.9714950.61422551.80.290.8719.600.310.7318.56
Oribi Airport− 29.6530.472103384761.40.320.8422.050.530.8420.04
Paarl− 33.7218.971030.1741111.50.220.817.570.140.367.41
Paddock− 30.7530.275156.84866951.40.200.9023.040.320.8021.42
Phalaborwa− 23.9331.154073.41993841.90.190.924.790.280.682.74
Pietermaritzburg− 29.6330.46732.83996401.60.310.8822.260.530.8219.86
Pietersburg Wo− 23.8729.4512260.71743942.30.120.4920.960.270.7120.19
Pilanesberg− 25.2627.2310854.82074892.40.040.1722.860.410.0818.67
Plettenbergbaai− 34.0923.331390.32292881.30.150.863.950.130.603.67
Pongola− 27.4131.5931212.22503111.20.180.7623.260.330.5721.01
Port Edward− 31.0730.23114.14337241.70.220.8823.040.300.8722.73
Port Elizabeth Wo− 33.9825.61630.62032631.30.130.671.680.160.600.70
Porterville− 33.0118.98122057621.10.200.637.120.030.017.91
Postmasburg− 28.3523.0913218.21212201.80.180.5919.480.360.6919.54
Potchefstroom− 26.7327.0713493.62264041.80.190.7921.640.280.6219.46
Potgietersrus− 24.2129.01110710.71332762.10.150.6320.520.190.4020.72
Pretoria Eendracht− 25.7428.1813080.62514671.90.130.7222.170.190.6720.18
Pretoria Unisa− 25.7728.214395.621943120.110.7621.270.230.7019.44
Prieska− 29.6722.739490.6791541.90.230.8622.010.360.7020.12
Queenstown− 31.9226.8811045.31843211.70.300.8420.540.520.7519.12
Robbeneiland− 33.818.3720.159500.90.220.736.330.180.505.58
Rustenburg− 25.6527.23115022274421.90.100.7023.080.090.2422.53
Springbok Wo− 29.6717.8910060.735371.10.060.182.230.360.2617.42
Springs− 26.228.4315926.41913031.60.240.8721.010.320.7919.17
Standerton− 26.9329.2315628.91682581.50.260.8421.380.390.8419.96
Stilbaai− 34.3721.41030.21741951.10.200.884.530.190.644.39
Strand− 34.1418.85100.585971.10.110.734.660.160.523.30
Struisbaai− 34.820.0630.71251120.90.210.865.470.140.485.50
Taung− 27.5524.7711101.81904022.10.160.6720.060.300.5518.62
Thabazimbi− 24.5827.4297710.812325020.060.3822.880.070.0821.15
Thohoyandou Wo− 23.0930.386141.63156782.20.200.935.700.300.902.88
Tsitsikamma− 34.0323.9150.12734021.50.080.602.260.080.361.33
Tygerhoek− 34.1519.91511.61601951.20.110.636.230.020.013.78
Uitenhage− 33.7125.441570.92112471.20.120.691.160.130.5822.78
Ulundi− 28.331.425240.92484551.80.240.8723.250.420.8220.65
Umtata Wo− 31.5328.677471.53814571.20.210.8122.130.380.7719.87
Upington Wo− 28.4121.268350.87514920.320.9223.240.360.7622.26
Van Reenen− 28.3729.3816802.14667781.70.220.930.230.430.7620.96
Van Zylsrus− 26.8822.059383.9921401.50.210.872.560.290.881.16
Ventersdorp− 26.3226.82149414.21903461.80.190.7520.290.320.6918.28
Vereeniging− 26.5727.9514796.92454281.70.210.9220.920.320.7519.26
Virginia− 29.7731.051412984971.70.280.8523.380.440.7622.43
Vrede− 27.4329.1716703.62524221.70.240.8420.510.360.7619.22
Vryheid− 27.7830.811632.42232991.30.220.8221.270.480.8919.93
Welkom− 2826.6713430.41873101.70.200.7922.070.190.4720.53
Witbank− 25.8329.1815482.32585302.10.200.8021.530.370.7619.85
Worcester AWS− 33.6619.421990.143581.30.060.1614.560.240.4416.70

Acknowledgements

Mathieu Rouault thanks the South African Weather Service for providing the rainfall data and WRC, Nansen-Tutu Center for Marine Environmental Research, and ACCESS for funding.

Ancillary