Abstract/Summary

The aim of this research is to investigate how large-scale climate variability affects flooding in the Amazon basin, using this assessment to demonstrate the potential predictability that these modes can provide to enable earlier warning of impactful floods. To address this a multi-stage approach is adopted; first to understand the gaps and confidence in the state of current knowledge on how climate variability affects both rainfall and river discharge in the Amazon basin, secondly, to understand the skill of global hydrological models for undertaking further assessment, and thirdly to undertake a robust assessment of the impact of climate variability on different flood characteristics while considering different methodological approaches in more detail. An assessment of the robustness in the results of previous studies suggests the need to explore in detail the physical mechanisms leading to flood events on an individual basis. While composite analysis of several floods identified a particular response associated with La Niña conditions, investigation into individual events show it is unknown if the same response would be identified for all events individually. The performance of eight large-scale hydrological models are evaluated for their ability to capture previous peak river flows. The choice of precipitation input is found to be the dominant component of the hydrometeorological modelling chain, with improvement found when ERA5 is the chosen meteorological forcing. Calibration of the Lisflood routing model is identified to have no impact on the ability to capture flood peaks, stressing the need to use an objective function that fits the purpose of the model. Examination of how climate variability impacts flood characteristics in the Amazon basin identified significant changes for both flood magnitude and duration during the negative ENSO phase, particularly in the north-eastern Amazon. This response was not identified for eastern Pacific ENSO events, highlighting how results can differ between ENSO types, while no notable impact or pattern is observed for flood timing. This thesis has provided important information on how climate variability impacts less studied flood characteristics (flood timing and duration) which are associated with important flood types (e.g. early or long floods). Future work should focus on the improvement of climate reanalysis to produce a longer-term dataset consistent with observations to extend climate analysis. This would allow the examination on the impact of climate phases at a more granular scale (e.g. analysing the strength or combination of climate phases)