Open Collections

Rodell, Christopher

University of British Columbia

Fire Danger Rating Systems (FDRS) are critical for wildfire risk assessment, resource allocation, and estimating wildfire smoke emissions, but their reliance on once daily weather inputs limits their ability to capture variations in fire behavior throughout the day. This dissertation introduces a new approach that integrates Numerical Weather Prediction (NWP) with machine learning (ML) to develop an hourly fire danger rating system, improving both fire danger assessments and wildfire emissions modeling. The first part of this research develops an Hourly Fire Weather Index (HFWI) system, enhancing the traditional daily Fire Weather Index (FWI) by incorporating hourly NWP forecasts. Validated against surface fire weather stations and satellite-based Fire Radiative Power (FRP) observations from wildfire case studies in Canada and the United States, the HFWI better captures dynamic fire behavior. For example, HFWI captures shifts in fire danger timing during late evening and early morning, and allows for the possibility of multiple periods of increased fire danger within a day. These improvements enable HFWI to provide more accurate fire danger assessments, particularly during atypical fire behavior. In the second research component, a neural network (ML) model is introduced to predict hourly FRP, a key metric for understanding wildfire intensity and smoke emissions. By combining NWP-derived HFWI and fuel load estimates, the ML model predicts potential FRP for each hour across large spatial areas. Trained on over 3,600 wildfire cases and tested on additional cases, the model outperforms persistence-based approaches by responding more effectively to changing weather conditions. This novel FRP forecasting method offers operational potential for predicting hourly fire danger and smoke emissions across North America. Overall, this work advances fire danger assessments and lays the foundation for a new, dynamic wildfire emissions modeling system, providing high-resolution, hourly predictions of fire intensity that can be used to improve smoke production forecasts.

Text

2024-12-09

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/89803

Atmospheric Science

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/