Sea level variations and storm surges are expected to increase as a result of climate change. 570 cities and some 800 million people are by 2050 estimated to be exposed to these phenomena when emissions do not decrease (UCCRN, 2018). It is, however, deeply uncertain if and to what extent emissions will decrease. Additionally, the effects of climate change are not fully understood due to its complexity, resulting in a wide range of uncertainty. Flood risk measures can be implemented to reduce flood risk. The economic evaluation of such measures is affected by other factors of uncertain nature such as economic growth. The Dynamic Adaptive Policy Pathways (DAPP) approach has been identified as an approach capable of identifying and implementing effective flood risk strategies under uncertain conditions (Haasnoot et al., 2013). This approach aims to provide decision-makers with insight into what action to take when. Simultaneously, the approach focuses on ensuring flexibility which facilitates adjustments to unforeseen conditions. Within this approach, so-called adaptation pathways need to be developed. These pathways describe a sequence of actions over time required to ensure a minimum level of flood safety. The DAPP approach has been applied to simplified cases in previous research, but not yet to more detailed master-planning for which the creation and economic evaluation of pathways is a not straightforward process. This results in the main research question of this thesis:

How can the Dynamic Adaptive Policy Pathways (DAPP) approach be used to create and select effective flood risk strategies under highly uncertain conditions?

This research question has been addressed by developing a framework capable of creating adaptation pathways and evaluating them with an incorporated scenario-based economic evaluation. This framework is supplemented with a probabilistic assessment that evaluates the performance of pathways in the full range of possible but uncertain futures. The use of these tools has afterwards been validated by applying them to a fictive case and a case study along the South East Coast of Singapore. The framework can be used to assess the sensitivity of uncertainties. Additionally, it can be used to obtain the conditions for which a different pathway turns out to be more effective. Damages corresponding to certain water levels are required as input and are used to build the damage function of a specific project area. These damages can be modelled with damage modules like the Global Flood Risk Tool (GFRT). The damage function and other basic information like dimensions and characteristics of the project area are used to perform an economic optimisation of individual measures. A selection of adaptation pathways (consisting of combinations of measures over time) is made based on prerequisites and a scenario-based evaluation. The scenario-based evaluation is supplemented with a probabilistic assessment as the use of scenarios might lead to cognitive biases and does not cover the full range of future possibilities (Hoffmann, 2017). This probabilistic assessment is conducted by means of a Monte Carlo simulation and can be used to evaluate the robustness of pathways in the full range of possible futures. A pathway can afterwards be selected and trigger values that should initiate the implementation of subsequent measures can be obtained via the framework.

The developed method has first been applied to a fictive case. A small area with a standard damage function was assumed for this. After the framework had been validated, the input-conditions were altered to assess the individual influence of each uncertain input value. The outcome turned out to be dependent on the characteristics of the project area and the assumed conditions. An increased area resulted in a pathway with a flood wall instead of a landfill as first measure being most effective. Out of the uncertainties, the discount rate and socio-economic growth rate turned out to most significantly affect the Net Present Value (NPV) and Cost-Benefit Ratio (CBR) of flood risk strategies. Especially, low discount rates and high socio-economic growth rates resulted in pathways built up of individual measures with lifetimes exceeding the technical lifetime of measures. To prevent this, and to ensure a flexible approach, the lifetimes of individual measures were restricted. A flexible approach enables one to adjust to conditions other than those that have been assumed. Measures with a long lifetime ahead lead to bigger investment costs and these are irreversible when the conditions turn out to be less severe than anticipated. The contrary also turned out to be possible. These short lifetimes corresponding to measures with practically infeasible low heights were prevented by setting a minimum height at which the linear relationship between the height and costs of measures starts. The probabilistic assessment was conducted after a pathway was selected. Increasing the range of uncertainty for this pathway logically led to a wider range of possible outcomes. This wider range of outcomes can be a reason not to select a certain pathway as the probability of a low NPV is higher. The fact that this difference could not be observed in the outcome of the framework underlines the need for this probabilistic assessment. Finally, trigger values were set for the selected pathway of the fictive case. The trigger values were initially set for the planned subsequent measures. However, it could be observed that the obtained trigger value for the planned subsequent flood wall increment did not result in sufficient time to implement a storm surge barrier without dropping below the required minimum level of flood safety. Therefore, trigger values corresponding to the next planned subsequent measure could potentially lead to the exclusion of other possible subsequent measures. Instead of setting trigger values for the planned subsequent measures, trigger values should provide enough time for all possible subsequent measures to forestall the exclusion of possible measures and ensure flexibility.

The findings from the fictive case led to adjustments of the framework to ensure practical feasibility and flexibility. The adjusted framework was afterwards applied to a real-life case along the South East Coast of Singapore characterised by a narrow water level distribution, i.e. a relatively small difference (∼15 centimeters) between water levels that differ a factor of 10 in return period. A storm surge barrier turned out to be a factor of 10 more costly than alternative solutions, irrespective of the assumed sea level rise scenario. Therefore, it was concluded that the storm surge barrier was not the desired flood risk reduction measure for the project area and the area was subdivided into smaller areas to further optimise the flood risk reduction strategy. The framework was used to create flood risk reduction strategies. No measures turned out to be required until 2092 for the assumed conditions. The sensitivity analysis of this case study showed that assuming a more severe sea level rise scenario (an additional 58 centimeters in 2100) and accounting for additional storm surge caused by climate change, could result in measures being required over 50 years earlier. Without taking this into account, two pathways turned out to satisfy the safety standards until 2200 for the assumed conditions. One pathway solely consists of a flood wall and subsequently flood wall increments (AP10) while dryproofing is the first measure of the other pathway and subsequently a flood wall and a flood wall increment are implemented (AP33). The costs of AP10 turned out to be slightly higher in the full range of futures than that of AP33 but the probability of obtaining a higher NPV than obtained from the framework was also higher (58% vs. 51% of the outcomes higher than the NPV of AP10 obtained in the framework). As the differences between the probabilistic assessments of the pathways are not substantial, the preferences of local stakeholders are even more important. Dryproofing can lead to inundation of the land for high water events with a return period lower than the safety standard while Singapore has set the goal to protect its coastlines and prevent inundation of the land. Therefore, the pathway consisting of a flood wall and flood wall increments turns out to be the most suitable solution for the project area.

This research contributes to the existing knowledge related to the DAPP approach as it smooths the not straightforward process of creating and evaluating adaptation pathways under uncertain conditions. It showed how to economically optimise individual flood risk reduction measures and build adaptation pathways out of those measures. A framework has been developed to automate this process and instantly evaluate the effectiveness of such pathways for set conditions. This study also showed how a probabilistic assessment of these adaptation pathways could be used to select a robust flood risk strategy. This developed method can further be fine-tuned by including transfer costs that reflect the costs of maintaining flexibility in the face of deep uncertainty. Additionally, flexibility can be incorporated within the probabilistic assessment to enable alteration of type and/or height of subsequent measures for conditions different than assumed. This would result in a more accurate assessment. Automating the integration of the probabilistic assessment within the framework can eventually lead to one complete tool which enhances the applicability and allows for the probabilistic assessment of all pathways instead of just a selection of pathways.

To reduce the flood risk in areas prone to flooding, possible adaptation measures must be investigated. In this thesis a Python-based framework has been developed that can screen flood adaptation measures as an alternative to the existing analytical method of Royal HaskoningDHV. The framework is limited to the costs and effects on the economic flood risk (the benefits) of eleven different measures such as levees, landfills, and flood proofing. First, the existing methods for screening measures were investigated. The resulting strong and missing elements were combined to form the functional requirements of the framework. The underlying script for the framework was elaborated with these requirements. Subsequently, the framework was applied to two case locations, namely Phu Loc (Vietnam) and Waal Eemhaven (Rotterdam, The Netherlands). Because these locations had already been developed by Royal HaskoningDHV, the results could be compared. The existing methods showed that clear results and the possibility for the user to select the measures and location are the main strengths. In addition, an uncertainty analysis and well-organized input table were found to be useful for the framework. The results of the simulations at the case locations were comparable to the results of the existing method of Royal HaskoningDHV for the measures considered. The framework is faster than the existing method and determines the optimal measure(s) and protection level. Therefore, it is a good alternative for screening the eleven considered measures. The framework can be supported with a hydraulic model to include more measures, for example river widening. It is recommended to add an extensive cost database to the framework, so that cost calculations and uncertainty analysis can be performed more easily in future applications.