Updating the existing intensity–duration–frequency (IDF) curves (focusing on 1-24 h) to reduce the global warming risks are typically based on regional climate models (RCMs). However, uncertainties are known to be present in such updating and need quantifying, especially whether bias correction (BC) methods should be used or not is less understood. Furthermore, the updating is still challenging for flood-prone cities in Southeast Asia due to a lack of sub-daily gauge-based observations. Therefore, in this study, firstly, we investigate the uncertainties related to the use of the BC, and the choices of different RCMs, RCP scenarios, parameter estimation methods by using the ‘analysis of variance’ method. Secondly, we propose a framework for developing and updating IDF curves in the data-scarce city (with only daily gauge observations) in Cambodia (Phnom Penh City (PPC)) by utilizing the rainfall information from the nearby cities (Ho Chi Minh City (HCMC) and Can Tho City (CTC) in Vietnam), based on the scaling property of rainfall duration versus intensity. A comparison between the RCMs with and without BC reveals that BC can efficiently remove the model bias, and the changes of rainfall intensity with BC are consistent with the underlying physical mechanisms (i.e., where atmospheric circulation tends to drive a larger increase in rarer rainfall events than less rare ones), supporting the application of BC before developing IDF curves. Further quantification of different uncertainty sources reveals that RCMs account for the largest share of the total variance (41.0% and 47.9% in HCMC and CTC, respectively). It is followed by the BC, with 39.4% and 29.7% in the two cities, respectively. RCP scenarios and parameter estimation methods provide the comparable and lowest uncertainty. The projected IDF curves in PPC, developed based on the integrated rainfall information (i.e., daily rainfall in PPC and scaling properties in nearby cities), reveal an apparent upshift (e.g., 33.6–43.7% increase in 20–year rainfall intensity in the near future (2026–2045) under RCP 4.5) of the IDF curves compared with the curves in the historical period (1986–2005). However, the deep uncertainty in such projection requires us to adopt a dynamic adaption strategy that can be iteratively updated based on new information. The current framework on quantifying different uncertainty sources and updating the IDF curves for data-scarce cities can provide suggestions for similar studies in other regions globally.
Read full abstract