Improved Forest Fire Danger Rating Using Regression Kriging with the Canadian Precipitation Analysis (CaPA) System in Alberta

  • Author / Creator
  • Estimating precipitation is currently one of the key challenges of accurate Fire Danger Rating. New gridded precipitation analysis systems such as the Canadian Precipitation Analysis System (CaPA) may be superior to the current analytical interpolation strategies, thin plate spline (TPS), and inverse distance weighting (IDW). To compare the performance of CaPA and interpolation methods in the forested area of Alberta, I evaluated precipitation estimates from CaPA and 17 algorithms of five interpolation methods, including IDW, smoothed TPS, non-smoothed TPS, ordinary kriging, and regression kriging. Precipitation estimates were generated using station observations through leave-one-out cross-validation and were evaluated using a range of skill scores (MAE, Bias, ETS, and FBI). I also assessed impacts of these precipitation estimates on the Canadian Forest Fire Weather Index (FWI) System and examined the impacts of weather station density on model performance. Results show that CaPA was only a mid-tiered method (13th of 18), except in Doppler radar covered areas, where CaPA performed second best. Regression kriging (with CaPA as a covariate) was the best method and produced precipitation estimates with 19.6% lower MAE compared with IDW. I found that the best method shifted with station density; CaPA was the best method when fire weather station density fell below 0.6 stations per 10 000km2 while regression kriging was the best method above this threshold. Additionally, this study showed that the FWI System responded to precipitation estimates differently due to their varying drying time lag of the indexes. Quick drying indexes (FFMC, ISI, FWI) preferred methods with lower MAE (e.g., regression kriging with smoothing), while slow drying indexes (DMC, DC, BUI) preferred methods with lower Bias (e.g., regression kriging without smoothing). Overall, I recommend the use of regression kriging with CaPA as a covariate to estimate fire danger across landscapes.

  • Subjects / Keywords
  • Graduation date
    Fall 2017
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.
  • Language
  • Institution
    University of Alberta
  • Degree level
  • Department
  • Specialization
    • Forest Biology and Management
  • Supervisor / co-supervisor and their department(s)
  • Examining committee members and their departments
    • Xianli Wang (Canadian Forest Service)
    • Jen Beverly (Renewable Resources)
    • Nadir Erbilgin (Renewable Resources)