2018/07/09 by Jeremy Diaz, Diaz, Jeremy, Maxwell B. Joseph +1
Earth and Planetary Sciences · Environmental Science · #Applications (stat.AP) #Climate variability and models #FOS: Computer and information sciences #Flood Risk Assessment and Management #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Meteorological Phenomena and Simulations
paper · pdf · doi:10.48550/arxiv.1807.03456
openalex publication_date 2018/07/09 · openalex created_date 2022/08/04 · openalex updated_date 2026/07/28
Tornadoes are the most violent of all atmospheric storms. In a typical year,\nthe United States experiences hundreds of tornadoes with associated damages on\nthe order of one billion dollars. Community preparation and resilience would\nbenefit from accurate predictions of these economic losses, particularly as\npopulations in tornado-prone areas increase in density and extent. Here, we use\na zero-inflated modeling approach and artificial neural networks to predict\ntornado-induced property damage using publicly available data. We developed a\nneural network that predicts whether a tornado will cause property damage\n(out-of-sample accuracy = 0.821 and area under the receiver operating\ncharacteristic curve, AUROC, = 0.872). Conditional on a tornado causing damage,\nanother neural network predicts the amount of damage (out-of-sample mean\nsquared error = 0.0918 and R2 = 0.432). When used together, these two models\nfunction as a zero-inflated log-normal regression with hidden layers. From the\nbest-performing models, we provide static and interactive gridded maps of\nmonthly predicted probabilities of damage and property damages for the year\n2019. Two primary weaknesses include (1) model fitting requires log-scale data\nwhich leads to large natural-scale residuals and (2) beginning tornado\ncoordinates were utilized rather than tornado paths. Ultimately, this is the\nfirst known study to directly model tornado-induced property damages, and all\ndata, code, and tools are publicly available. The predictive capacity of this\nmodel along with an interactive interface may provide an opportunity for\nscience-informed tornado disaster planning.\n