Recent years have witnessed a surge in mega-fires across the globe, impacting both human populations and ecosystems far beyond the immediate vicinity of the fires. These fires are increasingly occurring in regions and seasons previously considered fire-resistant. The inherent flammability of forests and woodlands, rich in live and dead plant biomass (fuel), makes them susceptible to fire. Fuel moisture content is a crucial factor determining fire risk. Major drying events can overcome natural firebreaks, leading to fire spreading even into typically fire-resistant ecosystems like rainforests. Quantifying the relationship between fire activity and fuel moisture content, especially with spatially explicit predictions of fuel moisture, is therefore crucial for understanding fire risk, particularly in the context of climate change. Vapor pressure deficit (VPD), derived from air temperature and humidity, serves as a reliable predictor of dead fuel moisture content across various biomes. VPD is also a key driver of plant mortality, further increasing flammability. Previous studies have shown VPD-based thresholds to be linked to increased fire activity in regions like southeast Australia and Mediterranean Europe. This study aims to identify VPD thresholds for the transition of global forest ecosystems from humid, non-flammable states to dry, flammable states. Using daily remotely sensed burned area data and hourly climate reanalysis data, the study will develop generalized linear models to predict fire occurrence probability as a function of daily maximum VPD, ultimately assessing climate change impacts on fire risk, focusing on carbon loss and human health.

Existing research highlights the increasing frequency and severity of global wildfires, often linked to climate change. Studies have explored the relationship between fire activity and fuel moisture content, with VPD emerging as a reliable predictor across various biomes. Previous work has established VPD-based thresholds indicating critical increases in burnt area in specific regions. The link between VPD and fire variability in temperate and tropical forests has also been documented. However, many studies focus on aggregate measures like total seasonal area burnt, lacking the daily-resolution insights needed to accurately assess the dynamic interplay between VPD and fire ignition probability. This study builds upon existing research by employing a high-temporal resolution approach (daily data), offering a refined understanding of the relationship between fuel moisture and fire activity on a global scale.

The study used a global dataset encompassing all forest biomes. Forest-dominant biomes were selected from a global classification of terrestrial ecosystems, further refined using a high-resolution global forest cover product. Data from February 2000 to 2020 was analyzed. Fire extent was derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) active fire product, using only high-reliability data. Daily VPD was calculated using daily maximum air temperature and dew point temperature from the ERA5 reanalysis dataset. For climate change analysis, three global climate models (ACCESS1.0, CNRM-CM5, and GFDL-CM3) from the CMIP5 dataset were selected based on their performance and ability to represent future climate scenarios (RCP4.5 and RCP8.5). Generalized linear models with a binomial error distribution and logit link function were used to assess the probability of fire incidence as a function of daily VPD for each forest biome and sub-continental window. Presence data consisted of VPD values on fire days, while absence data was randomly sampled from unburned grid cells. The receiver operating characteristic (ROC) curve was used to evaluate model performance. The annual frequency of days exceeding the VPD50 threshold (where the probability of fire exceeds 50%) was calculated using both ERA5 data (current frequency) and CMIP5 data (projected future frequency). To estimate the potential impact on human populations and forest carbon, population density data and aboveground biomass data were integrated with the projected changes in days exceeding the VPD50 threshold.

The study found a strong and predictable relationship between fire activity and VPD across all global forest biomes. Generalized linear models accurately predicted the probability of fire occurrence in most biomes (median true positive rate of 0.73). VPD thresholds above which the daily probability of fire exceeds 50% varied across biomes, reflecting latitudinal gradients, with higher thresholds in subtropical and tropical biomes (median 2.7 kPa) and lower thresholds in temperate and boreal biomes (median 1.3 kPa). The frequency of days exceeding VPD thresholds also varied significantly across biomes. Climate change projections under high emissions scenarios (RCP8.5) showed substantial increases in the frequency of days exceeding VPD thresholds in all forest biomes by 2081-2100, with the most significant increases in tropical South America (at least 150 additional days per year). Even under a lower emissions scenario (RCP4.5), widespread increases were projected. These increases were linked to significant exposure of carbon-rich forests (e.g., the Amazon) and human populations to increased fire risk. The study quantified this exposure using person-days and tonne-days of exposure to elevated fire risk conditions, highlighting the potential for substantial increases in both smoke-related health impacts and carbon loss.

The findings underscore the critical role of atmospheric water demand, as measured by VPD, in driving global forest fire activity. The strong correlation between VPD and fire probability demonstrates the potential for using VPD-based models to predict and manage fire risk under future climates. The projected increases in fire risk, particularly in carbon-rich tropical forests like the Amazon, highlight the significant threat to global carbon sinks and the potential for accelerating climate change through positive feedback loops. Similarly, the projections of increased exposure of human populations to wildfire smoke demonstrate substantial health risks in several regions. The study's findings emphasize the urgent need for mitigation strategies to reduce greenhouse gas emissions and adapt to the increasing fire risk. Improved fire management practices and ecosystem restoration efforts will also be crucial in mitigating the impacts of future fires.

This study provides compelling evidence of the strong link between VPD and global forest fire activity. The projected increases in fire risk under climate change highlight significant threats to global carbon sinks and human health. These findings underscore the urgency of climate change mitigation and adaptation strategies, including improved fire management and ecosystem restoration, to minimize future impacts.

The study acknowledges that other factors beyond VPD, such as human activity, fire weather, and long-term drying trends, influence fire activity. The models used did not explicitly account for these factors, which could influence the accuracy of predictions in specific regions. Furthermore, the spatial resolution of the data used might limit the accuracy of predictions at finer scales. The study focused on fire risk and did not explicitly model the resulting impacts, such as the actual area burned or economic losses.