Modelling wildfire occurrences and their present and future patterns of variability over the contiguous United States

Keeping, T. R. (2025) Modelling wildfire occurrences and their present and future patterns of variability over the contiguous United States. PhD thesis, University of Reading. doi: 10.48683/1926.00127660

Abstract/Summary

Wildfires are an intrinsic part of the Earth system, affected by people, vegetation and climate. However, extreme fire years and their impacts on people are of increasing concern. Understanding wildfire frequency is challenging across timescales, both due to stochasticity in individual wildfire occurrences and because significant climate variability in fire weather conditions is poorly sampled by the short observational record. This thesis characterises patterns in wildfire occurrence over the contiguous United States of America (USA) accounting for these two sources of uncertainty. First, developing a probabilistic model to account for the randomness of individual wildfire events. And second, using a large ensemble (LE) to account for the wider distribution of possible driving conditions in a reference (2000-2009) and future (+2°C) climate. Mean annual wildfire occurrences are projected to increase with climate change, though variability between fire years increases at a faster rate. An ensemble of model training runs and the LE application of the wildfire occurrence model enabled key drivers of daily and annual wildfire frequency to be found. Vegetation productivity was a key effect at both timescales, and an increasing control on variability in annual wildfire occurrences due to the emerging limitation of fuel availability on wildfire likelihood in dry regions of the USA. Climate variability is partially predictable from climate modes, which have been linked to increased wildfire in some regions. Multiple modes were found to have widespread, statistically significant associations with annual USA wildfire patterns, with these associations mostly strengthening in response to future climate change. This thesis advances wildfire occurrence modelling and quantification of interannual wildfire variability using LEs, both useful to global wildfire models. It is the first study to quantify the spatially varying effect of climate modes on USA wildfire likelihood, of potentially major utility in projecting seasonal fire danger.

Item Type	Thesis (PhD)
URI	https://centaur.reading.ac.uk/id/eprint/127660
Identification Number/DOI	10.48683/1926.00127660
Divisions	Science > School of Archaeology, Geography and Environmental Science
Download/View statistics	View download statistics for this item