Determination and analyses of rainfall thresholds for landslide initiation in the Greater Durban Metropolitan Region, Eastern South Africa.

Abstract This study investigates rainfall thresholds for landslide initiation in the greater Durban Metropolitan Region, eastern South Africa, a region highly susceptible to rainfall-induced landslide hazards. The rainfall threshold has been established using quantile regression analysis applied to landslide and rainfall datasets spanning 1970 to 2022. The quantile regression results based on the 70th percentile provided a cumulative event rainfall (E)-duration (D) threshold of E = 108D 0.67 , consistent with previous regional studies. Further analysis of landslide and rainfall data highlighted the critical influence of antecedent rainfall, particularly over 10–30-day periods, in predisposing slopes to failure, with shorter antecedent periods (≤ 5 days) correlating strongly with event-day rainfall. Temporal probability analysis reveals a recurrence interval of approximately three years for significant landslide-triggering events, emphasising the vulnerability of the region. Additionally, the observed increase in landslide frequency and severity across the Metropolitan region in recent years highlights the potential impact of climate change and anthropogenic activities. These findings provide crucial insights for developing effective landslide risk management strategies and enhancing early warning capabilities. By integrating the proposed threshold into early warning systems, timely alerts can be provided, minimising casualties and property losses. Understanding the role of antecedent rainfall conditions enhances prediction accuracy and informs targeted interventions during prolonged rainfall events. Evolving precipitation patterns due to climate dynamics and rapidly evolving urban land use mean that periodic reassessment of rainfall thresholds is needed, incorporating data from real-time moisture monitoring systems.