Knowledge-Enhanced Deep Learning for Simulation of Extratropical Cyclone Wind Risk

Boundary-layer wind associated with extratropical cyclones (ETCs) is an essential element for posing serious threats to the urban centers of eastern North America. Using a similar methodology tropical cyclone (TC) risk (i.e., hurricane tracking approach), ETC can be accordingly simulated. However, accurate and efficient assessment field inside currently not available. To this end, knowledge-enhanced deep learning (KEDL) developed in study estimate boundary-layer winds over Both physics-based equations semi-empirical formulas are integrated as part system loss function regularize neural network. More specifically, scale-analysis-based reduced-order Navier–Stokes that govern European Centre Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA) ERA-interim data-based two-dimensional (2D) parametric formula (with respect radial azimuthal coordinates) prescribes asymmetric pressure respectively employed rationalism-based empiricism-based knowledge enhance The KEDL, using standard storm parameters spatial coordinates, central difference, translational speed, approach angle, latitude center, surface roughness) network inputs, provide three-dimensional (3D) arbitrary high computational efficiency accuracy. Finally, KEDL-based model coupled large synthetic track database (SynthETC), where 6-hourly center location deficit included effectively assess along US northeast coast terms annual exceedance probability.