Statistics and Meteorology of Cutoff Lows over South Africa 1970–2023

Abstract

The meteorology of cutoff lows over South Africa is characterized by statistical analysis of daily field data in the period 1970–2023. An index is formulated by subtracting 500 hPa geopotential height in the mid-latitudes from the subtropics. Cutoff lows (COL) are identified by positive values, mostly in autumn and spring. Statistics indicate that climate forcing is seasonal: La Nina/El Nino favors COL in March–May/September–November. Hemispheric regressions reveal anomalous highs across the southern mid-latitudes when COL are frequent over South Africa. A 14-case composite was formed from the most intense daily COL events in autumn and spring. The composite shows a NW- tilted Rossby wave and jet stream loop around the COL. Maritime easterlies induce a warm east—cool west SST pattern, but composite moist inflows are shallow, so stormy weather hugs the coastal plains. Overturning circulations meet in an upper-level “saddle” over South Africa. 500 hPa sinking motions to the Southwest are of similar strength to rising motions to the Northeast. A COL case study exhibited hourly rain rates >10 mm at Port Alfred 18–20 October 2012 fed by tropical inflow. New insights emerged from this study via composite height sections over South Africa.

1. Introduction

South Africa experiences a subtropical climate marked by a large annual cycle with a dry cold winter and a warm humid summer. Zonal winds alternate from westerlies in June–August to easterlies in December–February. During the transition seasons, convection can be stimulated by cutoff lows (COL) that occur when the upper-level jet stream splits and forms a ridge south of a trough [1,2,3]. The subtropical trough detaches from the circumpolar westerlies and lags behind the mid-latitude ridge. A cold-core vortex spins up, often folding the tropopause. The bifurcated jet and its associated anticyclone—cyclone pair drift slowly eastward, converging moisture and inducing rainfall over the southeast coastal plains of South Africa. Sometimes the convection deepens and is joined by a meridional cloud band [4], leading to heavy runoff from the Southeast-facing escarpment. COL may have harmful side-effects, but they also benefit South African water resources and crop production by extending the summer rainy season into spring and autumn.

There is some concern that global warming will destabilize the circumpolar westerlies, amplify Rossby waves, and shift storm tracks [5]. Rising temperatures have expanded the realm of subtropical anticyclones, but it is unclear whether surplus heat in tropical outflows will stimulate climate-weather interactions. One aspect is certain: Higher sea level and beach erosion make the coastal zone more vulnerable. Cutoff lows often generate large swells as they move over the warm Agulhas Current [6]. Their storm surges damage infrastructure and necessitate costly re-engineering.

Given these impacts and existing knowledge on COL [7], this paper seeks new insights on climate-weather links, seasonality and trends, the regional circulation, and air-sea interactions. In Section 2, statistical methods are outlined, which employ a simple dipole index from daily fields of 500 hPa geopotential height. Section 3 provides results for the COL index: temporal attributes, spatial regressions, composite structure, and a case study. Statistics are evaluated for meteorological features around South Africa in the period 1970–2023 underpinned by radiosonde and satellite measurements, but future model simulations were not evaluated. Conclusions are given in Section 4, which highlight expected results and new outcomes.

2. Data and Methods

To identify COL over South Africa 1970–2023, daily 25 km resolution European Reanalysis (ERA5) [8] 500 hPa geopotential height data were extracted over two areas: maximum over sea 37–45° S, 21–28° E and minimum over land 25–30° S, 22–27° E (Figure 1a). These were subtracted south minus north and when positive, the normal thickness pattern is reversed. Considering the probability distribution (Figure 1b), thresholds were applied: > −50 m for COL deemed to be present (21% occurrence), and > 100 m for intense COL (1% occurrence). More complex statistical cluster methods (e.g., empirical orthogonal functions) were explored but could not reveal a 500 hPa geopotential dipole pattern in South Africa longitudes. Persistence and seasonality in the daily COL record were determined by autocorrelation at 0–6 days and calculation of the mean annual cycle.

Regressions were performed between the COL index > −50 m and 500 hPa geopotential field, omitting winter (June–August), when subsidence from the Hadley circulation inhibits tropical inflows and convection over the highlands (Figure 1a). The regression map confirmed that the index and threshold capture the high-south/low-north COL pattern. The threshold and omission of winter reduces the sample size to 4240 days in the period 1970–2023.

Cases were ranked to identify days with ERA5-derived COL index > 100 m in the period 2000–2023. This era is underpinned by multi-satellite rainfall and may better represent climate—weather interactions in recent years. Using this group of 14 days (

Table 1. (a) List of days used in the composite analysis and their COL index values (500 hPa geopotential height south–north), * = case study. (b) Flow chart of statistical methods.

(a)
yr mm dd	COL
2000 11 01	185.8
2001 09 19	180.6
2012 10 20	* 152.0
2001 09 14	142.4
2003 05 11	140.2
2022 05 21	133.1
2017 05 13	130.0
2006 09 26	121.7
2012 09 06	118.8
2023 05 31	116.1
2019 04 23	114.8
2000 04 05	107.1
2003 03 25	101.5
2016 04 07	100.8
(b)

a), regional composites were analyzed for NCEP2 winds at 500 hPa and 925 hPa level [9], NOAA sea surface temperature (SST) anomalies [10] and sensible heat flux, CHIRP satellite rainfall [11], and NCEP2 Southern Hemisphere winds at 250 hPa.

To explore low-frequency teleconnections, daily COL index values > −50 m were summed to monthly intervals and 1-month lead correlations were explored with several known climate indices after detrending: Antarctic oscillation and Southern annular mode—describing the polar vortex in different ways, and tropical Pacific Nino3 SST and Indian Ocean Dipole (IOD)—indicating the amplitude/phase of the tropical ocean thermocline “see-saw” associated with the El Nino Southern Oscillations. Pair-wise correlations were done per month to understand the seasonality of climate-weather links. Coefficients > |0.26| achieve 95% confidence with ~54 degrees of freedom during the period 1970–2023. The monthly COL index (based on the ERA5 geopotential height difference between the mid-latitudes and subtropics) was filtered to retain oscillations above 1 year, and wavelet spectral power was calculated for periods below 8 years. The filtered temporal record was regressed onto Southern Hemisphere 500 hPa geopotential height fields, to determine whether COL over South Africa are shared across longitudes.

Lastly, a major COL event was analyzed for propagation 17–22 October 2012 from ERA5 hourly weather data, wherein Hovmoller plots of rainfall and zonal winds in the latitude band 25–33° S (Eastern Cape) were generated. Meteosat imagery, SA Weather Service Port Elizabeth (PE) radiosonde profile and East London (EL) weather station observations, and MERRA2 ozone maps [12] complete the case study evidence. Table 1b is a flow chart of methodology that indicates how sample size declines with each step. Figure 1a presents the topography, COL index areas, and long-term average SST > 25° C and 500 hPa winds > 25 m/s.

3. Results

3.1. Cutoff Low Space-Time Characteristics

Spatial regression of the South African COL index > −50 m with daily 500 hPa geopotential height fields (Figure 2a) shows the expected pattern of subtropical low of radius ~600 km centered 28° S, 25° E, and mid-latitude anticyclone of radius ~1800 km centered 47° S, 20° E. Lag-correlations indicate northeastward propagation at ~4 m/s for the high/low pair over a 4-day sequence (icons in Figure 2a). This direction suggests that cyclonic eddies are advected from the Western Cape to the Highveld and remain shallow due to declining planetary vorticity. The mid-latitude anticyclone had a WNW-tilted ridge axis (indicating Rossby wave-breaking), while the subtropical low remained circular.

Temporal statistics derive from the daily COL record 1970–2023 (Figure 2b). The time series is chaotic with positive dipole events throughout. The mean annual cycle of the upper quintile (Figure 2c) has low amplitude in summer, and peaks in March–April and September–November, as reported in [13,14]. Infrequent COL in December–February season appear to be linked to warming over the Kalahari (thermal high at 500 hPa). The subtropical jet from the South Atlantic is guided poleward around South Africa, inhibiting bifurcation. The winter season shows high index values, but COL impacts then (June–August) are usually mild. Autocorrelation of the daily COL index > −50 m (Figure 2d) exhibits persistence from 0–2 days, inferring that hazardous wind and rain are sustained up to 48 hours, followed by runoff and swell impacts. There is little seasonal difference in how persistence diminishes (log- shape). However, COL above the −50 m threshold are twice as frequent in March–May than in September–November. Inferences that emerged: During autumn, the atmosphere cools faster than the ocean, and upward heat fluxes stimulate marine cyclogenesis. In spring, the mid-latitude westerlies are faster, so COL have a shorter stay over South Africa.

One-month lead correlations of the monthly COL index > −50 m with climate indices demonstrate seasonality (Figure 3a). Coefficients with respect to Nino3 and IOD are negative in April–May (−0.31) and positive in September–October (+0.24), inferring that La Nina/El Nino favors COL over South Africa in autumn/spring, respectively. Similarly, correlations with the (detrended) Antarctic Oscillation and Southern annular mode exhibit seasonality that peaks +0.29 in April (polar vortex expansion) and dips −0.31 in October (polar vortex contraction). Overall, the weak coefficients indicate circumpolar westerly bifurcation is quite random, and that climate-weather links are conditional.

The 18-month filtered COL index > −50 m (Figure 3b,c) reveals inter-annual oscillations of ±3.8 and ±5.6 years. Faster oscillations cover the early record, slower oscillations come later. High values indicating greater frequency of COL occur in 1974, 1981, 1988, 1996, 2000, 2006, 2011, and 2022, while spells of less COL appear in 1992, 1998, 2010, and 2018–2019. Index values exceeded 140 m in September–October 2000. No significant trend is found, inferring that global warming has not destabilized the circumpolar westerlies. The frequency of jet bifurcation and COL incidence has remained steady, possibly due to a faster rate of warming over the Kalahari (0.04 C/yr) than the Southern coast (0.02 C/yr). Regression of the monthly COL index > −50 m onto Southern Hemisphere 500 hPa geopotential height fields (Figure 3d) reflects positive anomalies in the Southeast Pacific and south of Australia. Thus, sympathetic bifurcation spreads across the Southern mid-latitudes when COL are frequent over South Africa.

3.2. Composite Analysis of Cutoff Lows

Composite features of the most intense COL (based on 14 days listed in Table 1a) are given in Figure 4 and Figure 5. Regional 500 hPa winds (Figure 4a) exhibit the expected pattern of a NW- tilted Rossby wave and an associated jet stream looping around a COL centered on 29° S, 25° E. Low-level winds (Figure 4b) describe a marine anticyclone on 40° S, 5–45° E that feeds moist air onto the Southeastern coastal plains. Over the plateau, there seems to be a standing easterly wave on 23° S, 25° E. The composite circulation is sufficiently long-lived to generate anomalous air-sea interactions, producing a warm east/cool west SST pattern (Figure 4c). The Agulhas Current is +2 °C warmer, and the Benguela Current is −2 °C cooler than normal (with coastal upwelling). The upward-east/downward-west pattern of sensible heat flux (Figure 4c) reflects cold easterly winds swirling around the ridging anticyclone. A 2000 km band of marine values > 30 W/m² reaches the Southeast coastal plains and feeds convection there. Cold air temperatures (<14 °C) over the interior inhibit sea breezes, so runoff is synoptically pulsed and diurnal cycling is minimal.

The 14-case composite height section of the meridional circulation and vorticity (Figure 4d) reveals contrasting kinematic forcing in the 500–300 hPa layer. Over South Africa 33–25° S: rising motion, poleward airflow, and cyclonic vorticity (−6 × 10⁻⁵ s⁻¹) occur; over the mid-latitudes 47–39° S: sinking motion, equatorward airflow, and anticyclonic vorticity (4 × 10⁻⁵ s⁻¹) occur. The twin overturning circulations reflect jet stream bifurcation and represent new outcomes.

The composite sequence of 500 hPa vertical motion (Figure 5a) reflects a slowly propagating dipole comprised of sinking over the Benguela Current and rising over the interior plateau, in equal proportions. The twin zonal overturning cells (Figure 5b) suggest that subsidence to the west of the COL accentuates rising motion to the east [15]. The overturning rotors are separated by an upper-level “saddle” associated with slowing of the Rossby wave. The atmospheric moisture is shallow; hence, impacts tend to focus on the Southeastern coastal plains. As the ridging anticyclone skirts the coast of South Africa, its subsidence (Figure 5c) is responsible for constraining the depth of atmospheric convection. Composite rainfall is vigorous over the Agulhas Current (60 mm/day) and spreads onto the adjacent coastal plains at ~30 mm/day (Figure 5d). Rainfall in individual COL events can exceed 100 mm/day, as seen below.

A significant meteorological feature in composite 250 hPa winds is the poleward spiral and acceleration (50 m/s) of circumpolar upper-level westerlies across the South Atlantic (Figure 5e). A mid-latitude jet streak extends ~6000 km into the South Indian Ocean and supplies anticyclonic vorticity to the mid-latitude ridge (cf. Figure 4d). A shorter jet streak occurs northeast of the COL and supplies cyclonic vorticity. The two jet streaks are separated by 2000 km in South African longitudes (20° E) and converge in the South Indian Ocean (50° E). These features identify Rossby wave breaking that sustains COL over South Africa for ~10 days every autumn and spring [15].

3.3. Case Study of 16–22 October 2012

Here, an individual event is studied. The Meteosat visible image on 12 Z 20 October 2012 (Figure 6a) shows a huge comma-cloud extending along the southeast-facing coastal plains due to the COL centered on 28° S, 24° E. Tropical inflows could be traced to the Zambezi Valley and Mozambique Channel. Winds over the Benguela Current were upwelling-favorable and stimulated atmospheric subsidence west of the COL, as noted earlier (cf. Figure 4b and Figure 5b).

The 21 Z 19 October radiosonde profile at PE (34° S, 26° E) indicates neutral stability (CAPE = 0) with moisture below 500 hPa and dry air aloft (Figure 6b). Winds rotated counterclockwise from southerly (1000 hPa) through easterly (850 hPa) to northwesterly (500 hPa). Hovmoller plots along 0–50° E, averaged 26–33° S (Figure 6c,d) illustrate that transient rainfall (blue shaded) retrograded and intensified on 19–20 October between two westerly troughs on 17 and 22 October. The retrograding suggests upper-level Rossby wave breaking and production of a stationary vortex. Surface easterly winds from the ridging anticyclone formed a diagonal band U < −10 m/s (green shaded) that converged (∂U/∂x –10⁻⁴ s⁻¹) over the coastal plains. A sequence of 200 hPa ozone maps is presented in Figure 6e. Stratospheric injection and tropopause folding is suggested by the ozone maximum > 40 ppb of radius 500 km, which propagated across the Western Cape (32° S, 19° E on the 20th), tenuously linked to a mid-latitude trough (42° S, 40° E).

Cold onshore winds and shallow convection generated rainfall > 300 mm (Figure 7a–c) over the Eastern Cape. Hourly rain rates pulsed > 10 mm/hr on 18 and 20 October, causing sudden runoff. The SA Hydrology Service recorded peak discharge > 400 m³/s on the Kariega and Kowie Rivers, which led to ~$30 M damage around Port Alfred [16]. We take note of SAWS in situ observations pertinent to this event. Aircraft at PE measured low-level winds from 160° at 15 m/s on 17 October. The weather station at EL reported surface wind gusts from 200° at 23 m/s on the 18th as sea-level pressure dropped to 1000 hPa. Surface winds switched to easterly 070° at 14 m/s by the 21st, as the marine anticyclone ridged. Air temperatures at EL remained <15 °C from 19–21 October, activating latent + sensible heat flux > 400 W/m² over the Agulhas Current (Figure 7c,d). The waverider buoy off EL harbour recorded southeasterly swells of 3–4 m. Such air-sea interactions amplify the coastal impacts of shallow convection accompanying COL events.

4. Conclusions

This study has explored cutoff lows over South Africa using a simple dipole index from 500 hPa geopotential height over the region 16–52° S, 0–50° E in the period 1970–2023. COL weather systems occur ±10 days each year and are often well-predicted [17]. However convective cloud bands around the COL arrive randomly from north or east, and concentrate heavy rainfall in time and space, leaving communities vulnerable to hazardous runoff and service outages. A COL index was derived by subtracting mid-latitude maximum from subtropical minimum 500 hPa geopotential heights and applying a threshold > −50 m. COL exhibit seasonal amplification in March–May and September–November. Regression of the index onto fields of 500 hPa geopotential height revealed a bifurcated jet and Rossby wave looping around a subtropical low undercut by a mid-latitude ridge. More extreme cases (Appendix A, Figure A1) have a blocking high and longer duration. Temporal statistics suggest that tropical climate forcing is seasonally dependent: Pacific La Nina and Indian Ocean Dipole favor COL in March–May, while El Nino and +IOD favor COL in Sep–Nov. Southern Hemisphere regressions with the COL index 1970–2023 exhibit sympathetic anomalies in the mid-latitudes of the Southeast Pacific and Southern Australia, where jet bifurcation leaves the imprint of high pressures. These features (seen in Figure 3d and Figure 5e) suggest the influence of zonal wave-3 [18].

A composite was formed from daily COL index > 100 m in autumn and spring. These 14 cases represent the most intense COL with the greatest impact. Maritime easterlies induced a warm east–cool west SST pattern, but moist inflows were shallow, so stormy weather hugged the coastal plains. The pattern of composite rainfall (cf. Figure 5d) suggested the warm Agulhas Current as a source of thunderstorms that move ashore, consistent with model simulations in [17]. The regional-scale overturning during COL exhibit twin rotors that meet in an upper-level saddle over South Africa. 500 hPa sinking motions in the Southwest are of similar strength to rising motions in the Northeast. The height sections represent new outcomes that aid our understanding of COL meteorology.

A case study revealed hourly rain rates > 10 mm/hr at Port Alfred 18–20 October 2012, fed by a COL of radius > 500 km centered on 28° S, 24° E joined by a meridional cloud band. Convection stretched along the entire southeast-facing coastal plains. Short-term peak values in this event were ~10× above composite and highlight how mesoscale features within COL drive localized impacts. New insights from this analysis emerged from composite height sections. Cyclonic eddies can grow beneath a bifurcated jet and deepen over South Africa, causing coastal weather impacts in Mar–May and September–November. Further work will analyze the vertical structure of convection rolling in from the Agulhas Current (Appendix A, Figure A1).