Paul D. Bates

Progress Toward Hyperresolution Models of Global Flood Hazard

Abstract Flood modeling at global scales represents a revolution in hydraulic science and has the potential to transform decision-making and risk management in a wide variety of fields. Such modeling draws on a rich heritage of algorithm and data set development in hydraulic modeling over the last 20 years, but conceptually the challenges of global flood modeling are the same as those faced at local and reach scales. This chapter therefore examines recent progress in the field of global flood hazard modeling and, in particular, looks at the development of so-called ‘hyperresolution’ models, defined here as those with a spatial resolution of 1 km or less. This chapter begins by examining how over the last 10 years the field of two-dimensional hydraulic modeling has moved rapidly from the study of single river reaches perhaps 10–60 km in length or restricted local areas of a few kilometers squared, to regional, continental, or even global scale models. This rapid advance has been enabled by three parallel developments: the discovery of new computationally efficient algorithms for solving 2D flow fields, advances in computing power and architectures, and the development of bespoke versions of global terrain data sets optimized for global flood modeling. To illustrate what is now possible this chapter uses the example of a recently developed hyperresolution model: SSBNflow. This is a 1/1200 arc second spatial resolution (∼90 m at the equator) global flood inundation model that solves the local inertial (i.e., a true hydrodynamic) form of the shallow water equations. The hydraulic engine is a clone of the well-known LISFLOOD-FP model. We describe the background to this model and its structure and report the results of a number of extensive validation tests. These show the model to perform well, within the best cases the performance, approaching that of a bespoke model built with local (rather than global) data. We describe how the hazard data can be used to produce global flood risk estimates, and, finally, we provide an extensive discussion of model limitations. Although this is still a young field, the progress, to date, has been rapid. Such models are already making a huge contribution to such fields as disaster risk reduction, disaster forecasting, risk finance, and humanitarian relief efforts, and the prospects for further improvements in model skill are very good.