Flood Hazard Characteristics at the Global Scale: An Observation-based Perspective

Abstract

Understanding large-scale flood hazard characteristics using streamflow observations is important for designing effective mitigation and adaptation strategies that reduce the future impacts of floods. Barriers to understanding floods include fragmented (and even conflicting) scientific findings of regional studies, limited spatiotemporal coverage of streamflow observations, and the complexity of flood generating processes. This thesis aims to improve the observation-based understanding of flood hazard characteristics at the global scale, focusing on three key objectives: (1) collating streamflow databases to support global-scale hydrological research; (2) identifying global patterns of flooding characteristics using the most comprehensive available streamflow database(s); and (3) evaluating the ability of hydrological simulations to reproduce trends exhibited from streamflow observations. An important element of this research was the production of the Global Streamflow Indices and Metadata (GSIM) archive. GSIM was initiated to develop an unprecedented daily streamflow dataset containing more than 30,000 stations worldwide by compiling 12 free-to-access databases. Significant efforts were invested in developing GSIM to produce a comprehensive set of metadata (e.g. catchment identifiers, catchment boundary, and landscape attributes), to evaluate the quality of the streamflow records that are included, and to derive time series of indices capturing essential aspects of streamflow regimes. GSIM data products have all been made available through a public data repository to support hydrological research. The first empirical investigation of flood characteristics at the global scale in this research focused on changes in magnitude of annual maxima streamflow, using more than 6000 daily streamflow of the Global Runoff Data Base, the core database of GSIM. The investigation assessed the significance of trends using the Mann-Kendall test coupled with a bootstrapping field significance approach. Across most experiments, there were more stations with significant decreasing trends than significant increasing trends, indicating limited evidence to support the hypothesis of increasing trends in flood hazard globally. The detected trends were assessed in the context of upstream catchment attributes and the findings suggested a substantial influence of catchment size on changes in floods. The data arising from the GSIM project was used for a global-scale assessment of flood seasonality to identify homogeneous regions of flood producing mechanisms. The identified relationships of flood generation were then used to predict flood timing across the globe, including ungauged locations, using climate indices derived from atmospheric reanalysis dataproducts. GSIM was also used as an input to compare observed trends in streamflow extremes to trends identified from simulated discharge of six global hydrological models (GHMs) across more than 3,000 sites globally. The findings showed moderate capacity of GHMs in reproducing spatial pattern of trends, suggesting the usefulness of GHMs in assessing the widespread pattern of changes in flood hazard. The release of GSIM enables new opportunities for advances in hydrology research through better spatiotemporal coverage of observations. In addition, the observation-based investigations in this research have yielded important findings and represent a significant contribution to improved understanding of flood characteristics at the global scale.