Rainfall Forecast and Real-Time Control: Balancing Accuracy and Performance in Urban Flood Mitigation

Abstract

Urban stormwater management is put under pressure by increasing urbanisation and climate change leading to more frequent urban flooding. Such flooding can introduce ecological and public health issues by releasing contaminated water into the environment. Upgrading traditional grey infrastructure and implementing blue-green infrastructure (BGI) can be limited in their applicability for flood mitigation making Real-Time Control (RTC) an increasingly popular alternative due to its cost-effectiveness and potential performance benefits.
This thesis investigated the accuracy of rainfall forecasts and examined how forecast-informed RTC procedures could reduce flooding frequency and overflow depth while minimising negative side effects. The rainfall forecast accuracy was assessed based on three key properties: rainfall depth, forecast horizon, and mean forecast intensity. Using the insights from this assessment, three forecast-informed RTC procedures were designed to enhance the flood mitigation performance of a reactive RTC.
With perfect forecast data, flooding was completely prevented for two of the procedures, which used the longest but distinct horizons. The procedure using the shortest horizon was not able to prevent every flooding event due to the restriction in pre-emptive water release ahead of the forecasted rainfall events. The forecast accuracy analysis showed that the accuracy declined with increasing rainfall depth and lengthening horizon. Furthermore, the results indicated a shift from underestimation to overestimation of the rainfall depth as the mean forecast intensity increased. Applying the real forecast data, all designed forecast-informed procedures demonstrated reduction in total overflow depth of up to 70%, with the reduction linked to pump operation. However, the flood mitigation performance did not align with the expected results based on forecast accuracy, indicating that the procedures’ logic and implementation played a significant role in determining their effectiveness.
The findings highlight the trade-offs inherent in using forecast-informed RTC procedures, particularly the balance between the uncertainties of longer forecast horizons and the need for sufficient lead time to take preventive action. By addressing these challenges, this thesis provides practical insights to inform the design and implementation of advanced RTC systems, marking a critical step toward more sustainable and resilient urban water management.