An eco-evolutionary approach to defining wildfire regimes

This paper proposes an eco-evolutionary framework that defines distinct global wildfire regimes based on the seasonal timing and magnitude of fuel availability (GPP) and drying (VPD), demonstrating that while vegetation and human activities modify fire properties, these environmental drivers fundamentally constrain fire behavior and offer a basis for simplified Earth System modeling.

The Gist

Imagine the Earth as a giant, living kitchen. In this kitchen, wildfires are like the occasional, necessary cleaning of the stove or the burning off of old food to make room for new growth. But just like a kitchen fire, you can't have a blaze without two specific ingredients: fuel (the food) and dryness (the heat to make it catch).

This paper by Sandy Harrison and her team is like a master chef trying to understand exactly when and where these kitchen fires will happen, not by looking at the fire itself, but by looking at the ingredients and the weather.

Here is the breakdown of their discovery, served up with some simple analogies:

1. The Two Ingredients: Fuel and Dryness

The authors argue that to understand wildfires, you only need to track two main things:

• Fuel Availability (The "Green Stuff"): They measure this using something called GPP (Gross Primary Production). Think of this as the "growth meter." It tells us how much grass, leaves, and wood the plants are growing. If the plants are growing fast, there is a lot of fuel waiting to burn.
• Fuel Dryness (The "Thirst Meter"): They measure this using VPD (Vapour Pressure Deficit). This is a fancy way of saying "how thirsty the air is." If the air is very thirsty, it sucks the water out of the plants, turning them into dry kindling ready to ignite.

2. The Timing Game: The "Perfect Storm"

The big insight of this paper is that it's not just about having fuel or having dry air; it's about when they happen.

Imagine a baker making bread:

• Step 1 (Growth): The baker needs to let the dough rise (plants grow). This happens when it's wet and sunny.
• Step 2 (Drying): The baker needs to put the dough in a hot oven to bake it (the fuel dries out). This happens when it's hot and dry.

If the dough rises and the oven turns on at the exact same time, you get a mess. But if the dough rises first, and then the oven turns on, you get perfect bread.

In nature, wildfires happen best when there is a time lag:

1. First, the plants grow big and lush (Fuel builds up).
2. Then, the weather turns hot and dry (Fuel dries out).
3. Boom! A spark (from lightning or a human) creates a fire.

The authors looked at the whole world and found that different regions have different "baking schedules." Some places grow fuel in spring and dry out in summer. Others grow fuel in winter and dry out in spring.

3. Sorting the World into 18 "Fire Neighborhoods"

Using a computer program that acts like a smart sorting machine, the team looked at these "growth vs. drying" schedules for every corner of the globe. They didn't just look at where fires are now (which can be random); they looked at the potential for fire based on the climate.

They ended up dividing the world into 18 distinct "Fire Neighborhoods" (or clusters).

• Neighborhood A: Might be a place where trees grow all year but rarely get dry enough to burn (like a rainforest).
• Neighborhood B: Might be a place where grass grows fast in the wet season and dries out perfectly in the dry season (like a savanna).
• Neighborhood C: Might be a cold forest where the fire season is very short and intense.

The cool part? These 18 neighborhoods perfectly match the actual fire behavior we see. Some have huge, fast fires; others have many small, slow fires. The "baking schedule" predicts the fire type.

4. Humans and Vegetation: The "Kitchen Tweaks"

Once they sorted the world into these 18 neighborhoods, they asked: "Do humans or different types of plants change the rules?"

• Humans: Humans are like people walking into the kitchen.
  • If we build roads or plant crops, we break up the "fuel." It's like putting a firebreak in the kitchen. This usually stops big fires from spreading, but it might increase the number of small fires (like people burning trash or clearing land).
  • However, humans can't override the weather. If the air isn't thirsty enough (not dry enough), even a human with a match can't start a massive wildfire. The "kitchen" just won't let it happen.
• Plants: Different plants act like different types of fuel.
  • Grass is like dry paper—it burns fast and spreads quickly.
  • Trees are like wet logs—they burn slower but can get very hot.
  • The type of plant changes how the fire behaves, but the timing of the fire is still dictated by the climate (the growth and drying cycle).

The Big Takeaway

The authors are saying: "Stop trying to predict fires just by looking at the fire history. Look at the ingredients."

By understanding the rhythm of plant growth and air dryness, we can predict where and when fires are likely to happen, even before they start. This is a huge step forward for Earth System models (computer simulations of our planet). It's like moving from trying to predict a storm by looking at the clouds after it rains, to predicting it by understanding the temperature and humidity before the storm forms.

In a nutshell: Wildfires are nature's way of balancing the books. This paper gives us a simple, elegant rulebook: Fuel grows first, then the air dries it out, and that's when the fire happens. If you know the schedule, you know the fire.

Technical Summary

1. Problem Statement

Current methods for defining wildfire regimes (often termed "pyromes") rely on clustering observed fire properties (e.g., frequency, size, intensity) derived from satellite data. This "top-down" approach has several critical limitations:

• Stochasticity: Wildfire occurrence is highly stochastic; short observational windows (e.g., recent decades) may not capture the full range of fire behavior, especially in regions with long fire return intervals (e.g., boreal forests).
• Decoupling: Observed fire properties are not always coupled. For instance, high burnt area does not necessarily imply high intensity or specific ignition sources.
• Lack of Mechanistic Drivers: Classifications based on emergent properties do not explain why regimes differ. They fail to distinguish between regions with no fire due to fuel limitation (dry deserts) versus those with no fire due to dryness limitation (wet rainforests).
• Climate Change Sensitivity: Emergent properties may change under future climate states, whereas the fundamental biophysical drivers remain constant.

The authors propose that wildfire regimes are fundamentally shaped by eco-evolutionary optimality (EEO), where vegetation and fire properties are adaptations to environmental conditions. Therefore, regimes should be defined by their fundamental drivers rather than their observed outcomes.

2. Methodology

The study employs a "bottom-up" approach, clustering the globe based on the biophysical drivers of fire: fuel availability and fuel dryness.

• Key Drivers:
  • Fuel Availability: Represented by Gross Primary Production (GPP). The authors use the P-model (an EEO-based light-use efficiency model) to simulate GPP, as global observed GPP is not available.
  • Fuel Dryness: Represented by Vapour Pressure Deficit (VPD), a measure of atmospheric dryness that drives fuel desiccation.
• Data Sources (2003–2020):
  • GPP: Simulated using the P-model forced by ERA5-Land climate data, SoilGrids soil data, and fAPAR data.
  • VPD: Derived from ERA5-Land.
  • Fire Properties: Aggregated from the Global Fire Atlas (ignitions, size, speed, duration), MODIS MCD64A1 (burnt area), and GFED5 (carbon emissions).
  • Human/Vegetation Modifiers: Cropland/pasture cover (HYDE 3.3), road density (GRIP), population density, and fractional vegetation cover (ESA CCI).
• Clustering Algorithm:
  • The study uses X-means clustering (a variant of k-means that automatically determines the optimal number of clusters using the Bayesian Information Criterion) on the seasonal phasing and magnitude of GPP and VPD.
  • The analysis was performed at 10 km and 50 km resolutions to ensure robustness.
• Statistical Analysis:
  • Hotelling's $T^2$ statistic: Used to test the multivariate difference between cluster medians across seven fire properties.
  • Pairwise Wilcoxon tests: To identify which specific fire properties distinguish clusters.
  • Generalized Linear Models (GLMs): Used to quantify the influence of GPP, VPD, vegetation cover, and human activity on the within-cluster variability of fire properties.

3. Key Contributions

• EEO-Based Pyroclimate Definition: The paper introduces a novel framework defining wildfire regimes ("pyroclimates") based on the phase difference and magnitude of GPP and VPD seasonal cycles, rather than observed fire statistics.
• Identification of 18 Distinct Regimes: The authors identified 18 statistically distinct global clusters that align intuitively with major biomes (e.g., transitions from tundra to boreal forest, or tropical rainforest to savanna).
• Decoupling of Fire Properties: The study demonstrates that fire properties (size, speed, duration, intensity) can vary independently within and between regimes, challenging models that assume fixed relationships between these metrics.
• Quantification of Human vs. Environmental Control: The study disentangles the effects of environmental drivers (GPP/VPD) from human activities (fragmentation, ignition sources) and vegetation traits.

4. Key Results

• Cluster Characteristics:
  • The 18 clusters cover the globe, with Cluster 17 (largest area) and Cluster 18 (smallest area) representing distinct environmental niches.
  • Optimal Burning Conditions: Clusters with the greatest temporal separation between peak GPP (fuel build-up) and peak VPD (fuel drying) exhibit the highest burnt area and fire size. This confirms the hypothesis that fire requires a sequence of growth followed by drying.
  • Intermediate Productivity: Carbon emissions peak at intermediate GPP levels, consistent with the intermediate productivity hypothesis.
• Driver Analysis (GLM Results):
  • Environmental Drivers: GPP and VPD are the dominant drivers of within-cluster variability in burnt area and carbon emissions. VPD has a consistently positive effect on burnt area; GPP generally has a positive effect, except in wet tropical forests (Cluster 9) where high GPP implies wet fuel.
  • Human Activities: While humans modify fire regimes, they do not create entirely new regimes.
    • Crop Cover: Increases the number of ignitions (agricultural burning) but reduces fire size, speed, and duration due to landscape fragmentation and reduced fuel continuity.
    • Road Density: Generally reduces fire speed and duration (fragmentation) but can increase ignitions in boreal regions (edge effects, increased access).
    • Population Density: Rarely a dominant factor compared to land cover and climate.
  • Vegetation Traits: Shrub cover is a significant positive predictor for fire speed and burnt area. Tree cover generally suppresses fire spread, while grass cover promotes it.
• Regime Distinctiveness: 17 of the 18 clusters are statistically distinct from at least one other cluster regarding all seven fire properties. Only one pair (Clusters 8 and 18) was statistically indistinguishable.

5. Significance and Implications

• Earth System Modeling (ESM): The study provides a robust basis for developing simpler, more mechanistic fire models for ESMs. Instead of complex, computationally expensive process-based fire models, an EEO approach driven by simulated GPP and VPD can effectively parameterize fire regimes.
• Future Climate Projections: By defining regimes based on fundamental drivers (GPP/VPD) rather than emergent properties, the framework is more robust for projecting fire regimes under future climate change and varying CO2 concentrations, which alter GPP independently of temperature.
• Management and Policy: The results suggest that land management (suppression, fragmentation) operates within strict environmental constraints. While humans can suppress fires or alter ignition sources, the fundamental "pyroclimate" (the potential for fire based on fuel and dryness timing) dictates the upper limits of fire behavior.
• Theoretical Advancement: The work bridges eco-evolutionary theory with fire science, positing that vegetation traits and fire regimes co-evolve to optimize ecosystem functioning under specific climatic constraints.

In summary, Harrison et al. successfully demonstrate that wildfire regimes are best understood as environmental niches defined by the seasonal interplay of productivity and dryness, offering a predictive, mechanistic alternative to descriptive, observation-based fire classifications.