Using Classification Trees to Build Flexible and Intuitive Winter Weather Indices

In northern latitudes, road salt remains a key element of winter maintenance operations. At the same time, there is increased pressure to reduce salt usage without compromising level of service or safety. Road salt usage models provide a way of benchmarking and understanding spatial-temporal variations in maintenance operations, and therefore have value in working toward improved salt management practices. The current study outlines a new approach for modeling road salt usage that addresses many of the limitations of past models. This approach is developed and illustrated using automatic vehicle locator data for three seasons and one provincial highway patrol near Ottawa, Canada. Using categorical, hourly salt application rates for specific highway segments as the dependent variable, and various sources and types of forecast and observed weather conditions as the independent variables, five different treatment modes are modeled using classification trees. Results are promising in terms of both the accuracy of predictions and the ability of this inductive approach to identify key explanatory variables and related threshold values that affect the probability of different treatment options. The winter index that results from this approach can incorporate both the most likely treatment mode as well as its probability, and can be scaled such that the same data inputs and outputs can be used to characterize winter weather at various temporal scales, from individual storms to entire winter seasons.

• Corporate Authors:

  Transportation Research Board

  500 Fifth Street, NW
  Washington, DC  United States  20001
• Authors:
  • Andrey, Jean C
  • Brenning, Alexander
  • Mills, Brian
  • Kirchhoff, Denis
  • Abulibdeh, Ammar
  • Perchanok, Max S
• Conference:
  • Transportation Research Board 88th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2009-1-11 to 2009-1-15
• Date: 2009

Language

• English

Media Info

• Media Type: DVD
• Features: Figures; References; Tables;
• Pagination: 15p
• Monograph Title: TRB 88th Annual Meeting Compendium of Papers DVD

Subject/Index Terms

• TRT Terms: Automatic vehicle location; Deicing chemicals; Level of service; Maintenance practices; Spatial analysis; Winter maintenance
• Uncontrolled Terms: Temporal analysis
• Geographic Terms: Ottawa (Canada)
• Subject Areas: Highways; Maintenance and Preservation; I62: Winter Maintenance;

Filing Info

• Accession Number: 01124175
• Record Type: Publication
• Report/Paper Numbers: 09-1642
• Files: TRIS, TRB
• Created Date: Mar 16 2009 7:22AM