Kernel Regression of Directional Data with Application to Wind and Wildfire Data in Los Angeles County, California
Abstract:This article describes a method of kernel regression that can be used to investigate the relationship between a directional explanatory variable and a real-valued response variable. Cross-validation and bootstrap methods for obtaining sensible bandwidths and standard error estimates are also described. The proposed method is applied to wildfire and meteorological data from Los Angeles County, California, with the goal of summarizing and quantifying the impact of wind direction on the total area burned in wildfires ignited on a particular day. The results confirm that winds blowing from the northeast and east are associated with the ignition of wildfires of significantly larger burn areas than winds from other directions; the mean burn area of wildfires ignited on days with winds from the northeast is approximately 4.7 times that associated with the winds from the southwest. A major reason for this increase in burn area is confounding of wind direction with meteorological variables such as temperature, wind speed, and humidity, but even when generalized additive modeling is used to control for these meteorological variables, winds from the northeast are still associated with the ignition of fires of substantially larger burn areas.
Document Type: Research Article
Publication date: August 1, 2011
More about this publication?
- Membership Information
- ingentaconnect is not responsible for the content or availability of external websites