Worried about heading out on the roads in heavy rain? A fascinating paper – delivered at the 44th Southern African Transport Conference (SATC) – reveals that heavy rain is not necessarily the biggest wet-weather danger. But rain can certainly turn already dangerous intersections and corridors into fatal hotspots…

A new paper presented at the 44th Southern African Transport Conference (SATC) has delivered a stark warning for Gauteng: rainfall is not simply an inconvenience for motorists. Under the wrong conditions, it becomes a powerful amplifier of fatalities on the roads.

The study, Precipitation Effects on Road Fatalities: A Machine-Learning Assessment Using Rainfall Records in Gauteng, was authored by Freeman Masuku and Karien Venter of the CSIR, together with Marion Sinclair of Stellenbosch University. It examined fatal road crashes in Gauteng between 2015 and 2019, combining fatality records with rainfall observations from 15 South African Weather Service stations.

Its central finding is both important and counter-intuitive: the most dangerous rainfall is not necessarily the heaviest. Light-to-moderate rain was found to significantly raise fatality risk, especially during the morning peak between 06:00 and 09:00, and the evening peak between 17:00 and 20:00.

Heavy rain, by contrast, can reduce exposure because drivers slow down, delay trips or avoid travel altogether.

The danger is concentrated

The research makes it clear that rain does not create risk evenly across the province. It sharpens existing danger.

Fatalities were found to cluster near major intersections, high-mobility arterial corridors and dense pedestrian environments. Township areas accounted for around 17% of fatalities, with hotspots identified in areas including Soweto, Tembisa, Katlehong and Sebokeng. However, the paper stresses that township status itself is not the main predictor of road fatalities. The real danger lies in the interaction between infrastructure design, pedestrian exposure, visibility and weather.

In plain terms, rain makes Gauteng’s already dangerous places more dangerous.

Intersections emerged as a particularly serious concern. The study found that proximity to intersections was consistently among the strongest predictors of fatality risk, suggesting that wet weather amplifies existing conflict points between vehicles, pedestrians, turning traffic and high traffic volumes.

Winter is a hidden danger season

While Gauteng’s rainfall is concentrated in summer, the study found that fatalities peak during the cooler winter period. That may seem surprising, but the explanation is practical.

Winter brings shorter days and more dark-hour travel. When occasional rain falls during this period, wet roads combine with poor visibility and reduced skid resistance. The result is a high-risk environment, particularly for pedestrians and road users moving through intersections and arterial routes.

These findings matter because road safety planning often treats the rainy season as the obvious period of concern. The paper suggests that authorities should also treat winter, especially dark peak periods, as a period of increased danger.

High-risk corridors identified

The research identified several road sections and corridors with disproportionate fatality burdens across the province. These include the Golden Highway, Main Reef Road, N12, N14, Heidelberg Road, R25, Randfontein Road, R21, Johannesburg Road, R80, Voortrekker Road, M1, N17, N1, N3, Albertina Sisulu Road, Old Pretoria Main Road, R24, William Nicol Drive, Solomon Mahlangu Drive and Chris Hani Road.

For transport authorities, this list is not merely descriptive. It provides a practical basis for deciding where limited safety budgets should be spent first.

Machine learning changes the picture

A key strength of the paper is its use of a Spatially Aware Poisson Random Forest Ensemble (SAPRF). The model combines traditional statistical analysis with machine learning and spatial data, allowing the researchers to identify nonlinear patterns that conventional models miss.

The SAPRF model substantially outperformed traditional regression. According to the paper, it achieved an R² value of 0.958, compared with 0.062 for Poisson regression and 0.872 for a standard Random Forest model.

This matters because road deaths are not caused by one factor in isolation. They emerge from combinations of weather, road design, pedestrian movement, traffic volumes, daylight conditions and location. A model that can capture those interactions gives authorities a sharper tool for prevention.

What needs to change

The paper’s recommendations are practical and targeted. Gauteng should integrate real-time rainfall data into traffic operations, using light-to-moderate rain during peak hours as a trigger for enforcement, variable message signs and temporary speed management. Critical intersections should receive better lighting, reflective markings and pedestrian visibility upgrades.

Drainage and skid-resistance treatments should be prioritised at known wet-weather crash locations. Stormwater inlets should be cleared before the rainy season, while high-risk corridors should be considered for resurfacing and pedestrian safety improvements.

The study also calls for spatial risk modelling to be embedded in provincial planning, allowing authorities to update hotspot maps, guide enforcement and direct infrastructure spending more intelligently.

The message is clear: Gauteng does not need to wait for more fatalities before acting. The data already shows where wet-weather risk is concentrated, when it is most likely to rise and which interventions could make the biggest difference. If authorities act on this data, many lives could be saved.