Emissions in Gauteng

Abstract

Emission reduction interventions assist decision-makers in setting targets for environmental regulations and policies. These are necessary to address the growing concern of air pollution. In the UK alone, £800m have been invested in an air quality programme to meet their net-zero emissions target by 2050 (World Economic Forum, 2020a). We recognise that informed decision-making is vital for capital investment into transport interventions, especially in a developing country like South Africa.

We focus on emissions generation in the Gauteng province to understand how the actual traffic emissions vary from our estimations with the tools at our disposal. The tool we utilise is the Multi-Agent Transport Simulation (MATSim) emissions model based on the Handbook Emission Factors for Road Transport (HBEFA). MATSim is a powerful modelling framework that can produce transport simulations of an entire city with a high level of detail (Fourie, 2009; Van Velden, 2012; Zhuge et al., 2014; Ziemke et al., 2019).

The problem we face is that the European-based emissions model does not account for the driving conditions and vehicle types affecting real-world driving emissions on South African road networks. We address the diversity of our local driver population by creating a synthetic population representing the Gauteng vehicle population. MATSim’s Agent-Based Model (ABM) enables us to model emission profiles for each vehicle represented as an agent. In the synthetic population, we include passenger cars and heavy vehicle types. We estimate the aggregate CO2, CO and NOx emitted on a provincial level and the individual emissions per vehicle type.

We use PEMS equipment to conduct Real Driving Emissions (RDE) tests with which we validate our MATSim emissions model for Gauteng. We conduct these tests for both vehicle types represented in our synthetic population: a passenger car and a heavy vehicle. By comparing the PEMS data to MATSim’s estimations on a predetermined test route in Pretoria, we find that the emissions model accounts for ±80% of the CO2 emissions from these vehicle types. Furthermore, the observed CO emissions are 2.3–2.9 times higher than the simulation. MATSim also underestimates NOx emissions for the heavy vehicle type and overestimates these pollutant emissions for the light vehicle.

Our investigation of the emissions on the test route reveals that different road types and driving conditions factor into the variance we observe in our local emissions model. MATSim struggles more to estimate the emissions on steep suburban roads than on urban or freeway sections. Regarding driver behaviour, aggressive drivers might cause more carbon and NOx emissions than conservative drivers. Weather conditions also influence this behaviour, and we heed the notable difference between our warm South African and wet European weather.

We accomplish our research goals of building a representative Gauteng emissions model in MATSim, investigating how this model performs “out-of-the-box” and quantifying the gap between our local simulation and the reality of traffic emissions in South Africa.