Introduction

Climate change is predicted to increase the frequency of extreme single-day fire spread events, with major ecological and social implications. In contrast with well-documented spatio-temporal patterns of wildfire ignitions and perimeters, daily progression remains poorly understood across continental spatial scales, particularly for extreme single-day events (“blow ups”). Here, we characterize daily wildfire spread across North America, including occurrence of extreme single-day events, duration and seasonality of fire and extremes, and ecoregional climatic niches of fire in terms of Actual Evapotranspiration (AET) and Climatic Water Deficit (CWD) annual climate normals.

Methods

Remotely sensed daily progression of 9,636 wildfires ≥400 ha was used to characterize ecoregional patterns of fire growth, extreme single-day events, duration, and seasonality. To explore occurrence, extent, and impacts of single-day extremes among ecoregions, we considered complementary ecoregional and continental extreme thresholds (Ecoregional or Continental Mean Daily Area Burned + 2SD). Ecoregional spread rates were regressed against AET and CWD to explore climatic influence on spread.

Results

We found three-fold differences in mean daily area burned among 10 North American ecoregions, ranging from 260 ha day^-1 in the Marine West Coast Forests to 751 ha day^-1 in Mediterranean California. Ecoregional single-day extreme thresholds ranged from 3829 ha day^-1 to 16626 ha day^-1, relative to a continental threshold of 7173 ha day^-1. The ~3% of single-day events classified as extreme cumulatively account for 16-55% of ecoregional total area burned. There were four-fold differences in mean fire duration, ranging from 2.7 days in the Great Plains to 10.5 days in the Northwestern Forested Mountains. CWD had a weak positive relationship with ecoregional spread, and there was no pattern for AET.

Discussion

Regions with shorter fire durations had greater daily area burned, suggesting a paradigm of fast-growing short-duration fires in some regions and slow-growing long-duration fires elsewhere. Although climatic conditions can set the stage for ignition and influence vegetation and fuels, finer-scale mechanisms likely drive variation in daily spread. Daily fire progression offers valuable insights into the regional and seasonal distributions of extreme single-day spread events, and how these events shape net fire effects.

Daily Fire Progression and Identification of Extreme Events

We measured daily area burned (ha d^-1) for individual wildfires by interpolating spatially continuous daily progression maps following methods developed by Parks (2014; Fig. S1). This technique interpolates VIIRS and MODIS hotspot detections to map the most likely day of burning at 30-m resolution within final wildfire perimeters obtained from national repositories. Previous studies have successfully utilized this technique to study various aspects of fire activity, including daily area burned (Hart and Preston 2020), spread (Holsinger et al. 2016, Wang et al. 2017), and refugia (Meigs et al. 2020, Downing et al. 2021). We constrained all daily progression interpolations to final fire perimeters obtained from the Monitoring Trends in Burn Severity (MTBS, USA; USDA Forest Service and USGS (2023)) and National Burned Area Composite (NBAC, Canada; Hall et al. (2020)) national repositories. Centroids of final fire perimeters were used to identify within which ecoregion individual fires occurred. Fires < 400ha were removed from the NBAC dataset to be consistent with the MTBS dataset, and because the interpolation method is inappropriate for smaller fires due to the coarse spatial resolution of MODIS and VIIRS pixels. Similarly, fires with < 10 hotspot detections were excluded from our analyses due to reduced accuracies in daily progression maps interpolated from few detections. All detections that occurred between midnight and 0600 h local time were assigned to the previous day under the assumption that those locations likely burned during the previous day with lingering heat signatures. In total, we interpolated daily fire progression for 9,636 fires and 66,318 unique daily fire spread events, which ranged in size from 25 to 224,565 ha. Daily spread events followed a skewed log-normal distribution, so we performed all statistical analyses and calculations on log₁₀ transformed values and back-transformed to native values for presentation. One-way analysis of variance (ANOVA) and Tukey’s Honestly Significant Difference (HSD) were used to compare daily areas burned, fire duration, and fire season among ecoregions. All geospatial data processing and statistical analyses were completed in R 4.2.2 (R Core Team 2022).

Next, we developed two complementary definitions of extreme single-day fire spread thresholds to explore how occurrence, extent, and impacts of single day blow-ups vary among ecoregions. Specifically, following a distributional definition inspired by Coop at al. (2022), we identified the top 2SD of events across northern North America as “continental extreme spread events”, as well as by ecoregion, or “ecoregional extreme spread events”:

Similarly, we calculated continental and ecoregional “large” event thresholds (e.g., mean + 1SD) to identify single-day events that are substantially larger than common spread days, but not as massive as extreme spread days. By presenting both tiers of continental and ecoregion-specific spread events, we aim to provide 1) a common continental definition of extremes that facilitates intuitive comparison of occurrence, extent, and impacts among ecoregions and 2) regionally contextualized estimates that may more accurately characterize realized differences in extreme fire activity among ecoregions.

Duration of Fire Spread, Duration of Extreme Activity, Fire Season, and Interannual Trends in Extremes’ Occurrence

Day of burning information from daily fire progression maps provides insight towards several temporal aspects of fire activity. First, we calculated each fire’s duration as the difference between its maximum and minimum day of burning, plus one day so that the final day of burning was effectively counted (e.g., a fire that began and ended on the same day would have a calculated duration of 1 day). To determine how fire duration is related to average and maximum daily spread, we regressed fires’ duration against their average and maximum daily areas burned. We included a model term for ecoregion to allow different intercepts and trends among regions (e.g., linear regression: Average or Maximum Daily Area Burned ~ Duration*Ecoregion). Second, we characterized duration of single-day extreme fire activity by counting consecutively occurring days of ecoregional and continental extremes within each fire. Third, we examined differences in ecoregional fire seasons by modeling day-of-year of burning among ecoregions (e.g., ANOVA: day-of-year of burning ~ Ecoregion) and by plotting ecoregional frequency histograms of common, large, and extreme events’ occurrences across a year. Finally, we used Sen’s slopes calculated with the “trend” package (Pohlert 2023) to analyze temporal trends in ecoregional annual counts of both continental and ecoregion-specific large and extreme single-day spread events between 2002 and 2021.

Daily Fire Spread and the Climate Space of Fire

We characterized ecoregions’ average climatic conditions in terms of AET and CWD by sampling TerraClimate 2002-2021 annual climate normal reference climatology rasters within the cumulative area burned by all fires that occurred within each ecoregion. These climate normal values were summarized as the ecoregional average AET or CWD ± 1SD. To explore broad-scale patterns between daily fire spread and these key climatic gradients, we regressed ecoregional average daily spread rates as well as large and extreme fire spread thresholds against ecoregional burned areas’ annual average AET and CWD conditions (e.g., linear model; common, large, or extreme daily fire spread threshold ~ Ecoregion*[AET or CWD]).