Noncyclonic Winds
The UrbanFootprint Noncyclonic Wind Risk model focuses on noncyclonic winds, which may include winds from thunderstorms or derechos but doesn't include tropical cyclone winds (see Tropical Cyclone Winds) or tornadoes. This dataset details the probability at which winds will exceed a particular threshold annually across the United States.
Our model uses data from the NIST Standardized Extreme Wind Speed Database for the United States, which provides readings of extreme wind events from weather stations (often, but not always located at airports). As a result, some areas across the United States are located much further from one of these stations than others.
Because multiple factors can impact or indicate wind speed, we included these parameters as inputs to the model:
location (latitude and longitude)
elevation
the standard deviation of elevation
distance to the nearest mountain
distance to the coastline
distance to the nearest valley
total rated wind turbine capacity within 50 kilometers
To estimate values at unobserved locations, we used a Bayesian hierarchical model. In this hierarchical model, we have:
Spatial process – describes how the wind speed distributions change across the United States. Using a Gaussian Process, we modeled a parameter that controls the mean and standard deviation of the winds, and we used a Gamma distribution to model a shared parameter controlling the distribution shape.
Station process – describes the distribution of wind speeds at any location. Using a Gumbel distribution, we modeled how the winds are distributed, given the parameters at a location from the spatial process.
We present results as 1-minute sustained winds on an H3 grid at zoom level 5 for the contiguous United States.
Availability: Location Insights | Climate & Hazard Insights