For centuries humanity has been engineering the second largest river system in the Netherlands, the river Meuse system. This resulted in improvements for the shipping and water safety functions, but simultaneously has had a negative impact on other functions of the river Meuse system. Nature and water quality, for example, degraded due to the construction of dikes, weirs, locks, bend cut-offs, summer bed mining and floodplain reduction. Additionally, climate change is affecting the fresh water availability due to low river discharges in combination with little precipitation in dry periods. On top of that, weir leakage and levelling of ships lead to water loss, which amplifies the fresh water availability problems even more. What the future might bring is deeply uncertain, however, the current problems are expected to grow larger due to influence of climate change and socio-economic developments.With the upcoming task to replace the hydraulic structures (weirs and locks) in the Meuse river, a unique opportunity arises to improve the problems that the river Meuse system experiences and prepare it for future developments. Replacement strategies need to be developed and tested on the system. Traditionally separate models will be constructed to simulate shipping, nature, water quality and hydraulics. They tend to be complex and computationally heavy and more often than not, these models make predictions based on a small set of the most probable scenarios. However, in the rapidly changing times of today, these scenarios are often not sufficient enough to account for the highly uncertain future. Therefore, in this thesis, one model is constructed containing the multi-functionality of the river Meuse system which is tested on a broad range of future scenarios to test replacement strategies for the hydraulic structures. MethodologyRobust Decision Making (RDM) is used to explore replacement strategies from a multifunctional perspective and evaluate them based on their robustness. RDM is a method which helps making decisions that lead to the construction of a robust system without making predictions and a system is robust if the applied strategy performs satisfactory under a wide variety of future conditions. RDM works with four steps, first the river Meuse system is modelled in the system dynamics software Vensim. This Meuse Model consists of submodels that simulate: the flow of water, the flow of shipping and the performance of the functions water quality, fresh water availability and the shipping load on the locks. All functions are therefore assessed in one model. The second step explores the behaviour of the current system by analysing the performance of the functions under future scenarios (i.e., combinations of future uncertainties). To cope with the deep uncertainty of the future, the probability of occurrence for each scenario is assumed to be uniformly distributed. Thirdly, design options are developed based on the vulnerabilities of the current system and combinations of design options form replacement strategies. In the fourth and last step, the replacement strategies are applied to the system in a policy analysis, after which the design options are evaluated based on their robustness. Exploration of the current systemExploring the current system behaviour over a variability of plausible futures demonstrates that in some scenarios problems arise for the functioning of the system. The performance of the fresh water availability is in general robust. The river Meuse system is designed to maintain a specific water level to provide for shipping. However, especially in weir section Grave, which contains outflow to the river Waal and a relatively large leakage of the weir at Grave, fresh water availability problems occurred during extreme drought events. In approximately 18% of the runs, the fresh water availability was smaller than full capacity. A scenario discovery revealed that several uncertainties have crucial impact on the undesired performance of the fresh water availability. It is found that leakage of the weirs and certain dry discharge years are influential uncertain input variables that cause problems for the fresh water availability.The water quality in the weir sections of the river Meuse reaches for most of the runs risky or undesired behaviour. The weir sections Linne, Grave and Lith are the most vulnerable because the weirs in these weir sections maintain relatively large water depths. This results in low flow velocities and therefore high risk for algal and cyanobacterial blooms and low diffusivity of toxic substances. For the intensity of shipping, two future scenarios are composed under socio-economic developments. The first leads to a slight increase of ship intensity and the second one leads to a slight decrease of ship intensity in the coming decades. Most of the locks seem to be able to cope with this change. Only the lock at Grave might experience issues as it contains the smallest and least amount of lock chambers in the corridor. For approximately 30% of the runs, the intensity to capacity ratio of the lock at Grave took either risky or undesired levels (>0.5). Formulating replacement strategies Based on these findings, a number of design options are formulated with the main goal to improve the robustness of the fresh water availability. Robustness performance of other functions, however, should not deteriorate, and preferably improve by applying the design options. Weir section Grave is used as main focus for improvement as this weir section is most vulnerable for future scenarios, and improvements will therefore have the largest impact. The design options are: •Installation of pumps (all weir sections): limit the outflow of water in dry periods•Heightening crest levels of the weirs at Grave or Lith: increase buffer capacity•Relocating the weir at Grave in upstream direction: increase the average water depth •Dynamic crest level (all weir sections): temporarily increase buffer capacity in drought periods•Efficient locking (all locks): limit the daily outflow of water in dry periods by imposing a minimum amount of ships that need to enter the lock chamber before starting the locking cycle Analysing the replacement strategies Combinations of design options are made to form replacement strategies, which are now applied to the system. All design options lead to improvement of the fresh water availability for weir section Grave. By introducing efficient locking, however, the intensity/capacity value of the lock at Grave increases compared to the current system for more than 80% of the runs. Raising the crest level of the weir at Grave leads to a decrease of flow velocity, and therefore water quality, for approximately 80% of the runs. Changing this to a dynamic crest level results in a decrease in flow velocity for approximately 25% of the runs, which is significantly less. By relocating the weir at Grave, the robustness of the fresh water availability in weir section Lith is slightly reduced and the ship draught downstream of the weir at Grave is at risk. To increase the water depth and remove the risk of insufficient ship draught, the crest level of the weir at Lith is raised. However, an increase in water depth leads to a decrease in flow velocity and therefore water quality. Installing pumps creates the largest robustness for freshwater availability, does not induce any deterioration for the other functions and is therefore seen as best performing replacement strategy. ConclusionThe research implemented a general robust decision making method to analyse complex systems and applied it to the replacement of hydraulic structures in rivers. Thereby it has shown how various design options can be evaluated for multiple functions and under uncertain future scenarios. This approach can be used to select design strategies that perform well under a broad scope of future uncertainties and thus make the system more robust. However, the model and tools used in this thesis also know many limitations. Costs, for example, are left outside of the scope of this thesis but is expected to have a large impact on the decision making process. Furthermore, potentially better fitting methods such as Many Objective Robust Decision Making or Dynamic Pathways are available to analyse replacement strategies. Because of these limitations, the results from this study can be seen as a first interpretation and setup for further research.