Climate Change Impacts on River Floods: Uncertainty and Adaptation

Abstract

The modelling frameworks, which include greenhouse gas emission scenarios, climate models, downscaling methods and hydrological models, are generally used to assess climate change impacts on river floods. In this research, the uncertainty associated with each component of the modelling framework is analysed with particular reference to climate change impacts on flood frequency. A method of risk-averse economic optimisation has been proposed for adapting river dikes to climate change under uncertainty. The Huai River Basin in China has been selected as a case study. The outputs of climate models, i.e., General Circulation Models (GCMs), under greenhouse gas emission scenarios have been commonly used as fundamental inputs of the climate change impact assessments. The analysis in this thesis employed the climate model projections of the WCRP CMIP3 and CMIP5 datasets. In Chapter 2, a brief introduction of emission scenarios, as well as a preliminary analysis of the simulative ability and future projections of the participating climate models, is provided. The results confirm the necessity to bias-correct and downscale the climate model outputs before being used in impact-related studies. The annual mean temperature in the study area is suggested to increase up to 8oC at the end of this century under a high greenhouse gas emission scenario without mitigation measures. The standard deviation of precipitation intensity is suggested to increase, especially in summer, which may in the future lead to high-magnitude floods. Empirical statistical downscaling methods are becoming increasingly popular in climate change impact assessments that require downscaling multi-GCM projections. In Chapter 3 empirical statistical downscaling methods are classified based on calibration strategies and statistical transformations. Ten combinations of calibration strategies and transformation methods were used to represent a range of empirical statistical downscaling methods. To test the performance of these methods in downscaling daily precipitation and temperature, an inter-model cross validation was carried out using an ensemble of 16 GCMs. These downscaling methods were further applied to downscale the climate for the future period to assess the associated uncertainties. The results show that the change factor based methods outperform the bias correction based methods in projecting the probability distribution of downscaled daily temperature. With the change factor calibration strategy, simply adding (for temperature) or multiplying (for precipitation) the mean change factor is sufficient to represent most of the relative changes projected by GCMs. The use of quantile based methods appear to be advantageous only at the tails of the distribution. More sophisticated bias correction based methods are needed to remove the biases in the higher-order statistics of the GCM outputs. The two calibration strategies led to fundamentally different temporal structures and spatial variability of the downscaled climatic variables. Bias correction based methods produced larger uncertainty bounds of inter-annual variability than the change factor methods. For downscaled precipitation, the uncertainty arising from the downscaling methods is comparable to the uncertainty arising from GCMs, while more uncertainty is introduced by calibration strategies than statistical transformation methods. There is a growing consensus that the performance of hydrological models should be routinely evaluated before being used in impact-related studies. The uncertainty, which stems from transferring calibrated models to a changing future climate, is receiving increasing attention. Chapter 4 assesses the uncertainties associated with the parameter calibration of the lumped Xinanjiang hydrological model when assessing the climate change impacts on river flow. The transferability of model parameters was tested in the context of historical climate variability using the differential split-sample test. The parameters calibrated from the periods representing differing climatic conditions were used to project future river flow in a changing climate. The uncertainties in projected future river flows stemming from the choice of calibration periods and parameter equifinality were compared. The results show that the transferability of the parameters calibrated from a wet period to a dry period is poorer than the other way around. The model error as well as the variability in the simulations due to equifinality increase with the increase of the difference in rainfall between the calibration and validation periods. The uncertainty due to the choice of calibration periods takes the majority of the total parameter uncertainty in the projected future mean discharge. When the calibration period contains enough information on climate variability, the equifinality effect and the choice of calibration periods contribute comparable magnitudes of uncertainty in terms of extreme discharge. Five sources of uncertainty mentioned above were compared in Chapter 5, i.e. GCM structure, greenhouse gas emission scenario, downscaling method, choice of period for calibrating the hydrological model, and non-uniqueness of hydrological parameters. Multiple samples of flood frequency curves were generated through the combinations of different emission scenarios, GCMs, downscaling methods and hydrological model settings. All samples were given equal weights in the analysis. The results show that the future flood magnitude is expected to increase, not only due to the increase in mean precipitation, but also due to the increase in variation of precipitation. Nonetheless, there is still a small likelihood that the flood quantiles with a high return period (above 20 years) will decrease in the future. The results of uncertainty comparison suggest that the GCM structure is the dominant source of uncertainty, emission scenarios and empirical statistical downscaling methods also result in considerable uncertainty, and the uncertainties related to hydrological model are less than those related to other uncertainty sources. To guarantee a safe flood defence in a changing environment, the adaptation to climate change needs to be considered in the design of river dikes. However, the large uncertainty in the projections of the future climate leads to varied estimations of future flood probability. How to cope with the uncertainties in future flood probability under climate change is an inevitable question in adaptation decision-makings. In Chapter 6, the uncertainty introduced by climate projections was integrated into the ‘expected predictive flood probability’, and the risk-aversion attitude was introduced in the adaptation of river dikes. The uncertainty in the climate change projections on flood probability was represented by the uncertainty in the parameters of the probabilistic model. This parameter uncertainty was estimated based on the outputs from the GCMs participating in IPCC AR4. The parameter uncertainty, estimated from the selected GCMs under different scenarios, was integrated into the expected predictive probability of flooding, which was then used in the risk-averse economic optimization. Different optimal results were obtained based on varied values of the risk-aversion index which represents the risk-averse altitude of decision makers. The case of a dike ring area in the Bengbu City in the Huai River Basin is studied as an example using the proposed approach. The results show that the uncertainty of climate change decreases the optimal safety level and increases the optimal dike heightening up to 8.23 m (with the risk-aversion index of 1.5) in a gradually changing climate. The value would be even larger if the climate will change sooner. Integrated adaptive measures rather than only dike heightening are needed to respond to the uncertain impacts in the future. The proposed approach enables decision makers to cope with climate change and the associated uncertainty by adjusting the level of risk aversion.