Explore a statistical model designed for large spatio-temporal data sets exhibiting complex patterns, particularly those simulated by physics-based hydraulic models over high-resolution 2D meshes. Learn about an innovative approach that combines multi-resolution analysis with an extended lifting scheme for spatio-temporal data, addressing the challenge of long computation times in urban flood hazard assessment. Discover how this model leverages dominant spatial features identified through clustering to create an interpretable non-parametric representation. Examine the incorporation of low-resolution model information to develop a downscaling model, assuming representative high-resolution spatial patterns can be inferred from the training set. Analyze the model's application to a 2D dam break experiment with a synthetic urban configuration and a field-scale test case simulating dike break flood wave propagation in Sacramento. Compare the performance of this spatio-temporal lifting scheme-based model with spatial interpolation schemes and a principal component analysis variant, focusing on its superior ability to reproduce extreme events. Read more