This is an opinion piece authored by Shawkat Sohel, a Research Fellow at the Forest Research Institute (FRI), University of the Sunshine Coast (UniSC), where he contributes to an ARC Future Fellowship funded project. 

Wildfires are escalating in intensity and frequency around the world, from the blistering heat of Australia’s Black Summer to the smouldering landscapes of California and the Mediterranean. These events are no longer isolated disasters — they are symptoms of a climate system in crisis. But despite our growing fund of firefighting technologies, one critical gap remains: the ability to predict where and when fires will likely ignite, well before they start. Today, most wildfire detection systems —focus on real-time monitoring of active fires using thermal or infrared data. While these tools are crucial for emergency response, they do not offer much time for preparedness. By the time a flame is spotted, the damage has often begun. 

What if we could forecast wildfire risk the same way we predict storms — days or even weeks in advance? Is there a ‘tipping point’ in vegetation moisture beyond which probability of fire occurrence is high? That’s where an innovative approach integrating plant functional traits (PFTs) and microwave remote sensing comes in — and why the world urgently needs to embrace it. 

Understanding what fuels the wildfire

At the heart of every wildfire is a simple equation: heat, oxygen, and fuel. But not all fuel burns the same way. A eucalyptus forest behaves differently from a pine plantation or a grassland, and even within those ecosystems, the fire risk depends on the traits of the plants themselves such as—vegetation moisture, successional stage (i.e., pioneers, early secondary, late secondary, climax phase), plant hydraulic trait, root trait, wood density, biochemical and phenological trait of plants and how they respond to water stress1. These traits, collectively known as PFTs, influence how much water a plant holds and at what point it dries out and in turn determines how flammable it becomes. Most importantly, incorporating PFTs with plant water use strategies can significantly enhance the interpretation of plant water use pattern2 and how tolerant they are in extreme water stress environment before ignition.  

Unfortunately, current fire prediction models often treat vegetation as a homogeneous layer — failing to capture this functional diversity. That’s a dangerous oversimplification in a world where climate change is drying landscapes unevenly and creating highly flammable conditions in once-resilient ecosystems.

From satellites to vegetation moisture: A new way to see wildfire risk

Wildfires are increasing in intensity and frequency worldwide, exacerbated by climate change, which drives severe droughts and reduced vegetation water content (VWC). This growing threat highlights the urgent need for an early detection system to enable proactive wildfire management. While millions of dollars have been invested in post-ignition detection technologies, the ability to forecast wildfire risk before ignition remains a critical gap. For example, Australia’s flammability monitoring system relies on manual VWC measurements and coarse-resolution MODIS data3. Furthermore, MODIS-derived vegetation indices only infer VWC indirectly through proxies such as “greenness”, limiting their precision for capturing fine-scale moisture variation across ecosystems and plant functional traits (PFTs)2. 

Fine-scale variation in VWC is crucial for wildfire risk assessment because low moisture levels directly increase vegetation flammability and the likelihood of fire ignition. Ground-based VWC measurements, while accurate, are time-consuming, expensive, and spatially limited. Thus, advanced remote sensing techniques offer a promising alternative for monitoring vegetation moisture. Among these, Vegetation Optical Depth (VOD), derived from microwave remote sensing, has shown strong potential as a proxy for VWC4. For example, the Vegetation Optical Depth (VOD) index can be used as a proxy for VWC, representing the attenuation effects of microwave radiation by vegetation cover, on a scale of 0 (dry) to 3 (saturated)5 where higher frequencies detect leaves and branches and lower frequencies detecting trunks6,7. Microwave remote sensing such as Sentinel 1a and 1b (a C-band radar appropriate for crops and forest) and ALOS Phased Array type L-band Synthetic Aperture Radar (PALSER; appropriate for forests) together have the ability to detect VOD in the top few millimeters of the vegetation, through to cloud cover8-10. There are only a few microwave remote sensing coarse resolution products capable of measuring VOD and even fewer high resolution VOD products based on Sentinel data, and these are mostly limited to the USA6-11. 

By combining VOD data with plant functional traits and climate variables, researchers can identify critical tipping points — thresholds where vegetation becomes so dry that the likelihood of ignition spikes. This approach would transform our understanding of wildfire from reactive to proactive and will predict they’re likely to start — tailored to the ecology of the landscape itself.

Existing wildfire warning systems such as Digital Earth Australia Hotspots and ongoing project FireSat focus on real-time detection and monitoring of active fires using infrared radiation. Digital Earth Australia hotspots provides updates every 10 minutes to detect fire hotspots12. Whereas, FireSat is a satellite equipped with Artificial Intelligence (AI) under development by Google Research and Earth Fire Alliance can detect fire with a 5m resolution at the early stage of fire spread, with updates every 20 minutes13. In contrast, the proposed framework (Fig. 1) of early warning system focuses on predicting wildfires before they occur by identifying VOD ‘tipping points’ that increase fire risk. By innovatively applying microwave remote sensing, this proposed system can quantify PFT specific VOD over time to monitor susceptibility of vegetation to fire, offering a proactive rather than reactive detection. Unlike existing systems that rely primarily on thermal cues to identify fires, this proposed approach targets the underlying ecological factors driving fire risk. 

Fig. 1. Conceptual diagram showing how microwave backscatter can be modelled with climate and PFTs to develop a wildfire early warning system based on VOD, a proxy of vegetation water content (VWC). Here, PFTs in this context include examples such as plant successional status: PS (pioneer species such as grass), ESS (early secondary species), LS (late secondary species), and CS (climax species). WC denotes water content. After extracting pixel specific VOD/fire occurrence data tipping point presence will be tested using a standard approach.  Theoretical application of five diagnostic plot14,15 need to use to identify thresholds distinguishing between fire occurrence responses with VOD changes. Peach shading represent disturbance phase. The presence of tipping point in the ecosystem state indicators (in this case it is fire occurrence) over time could suggest a threshold response class (A). However, to confirm a true threshold, this must be driven by a non-abrupt or linear increase in VOD (B). In such cases, the response curve (C) would display an abrupt non-linear relationship at the tipping point. This assumes that other influencing factors (i.e., climate, elevation & PFT) remain constant over the same time period across landscape. Verification of a threshold can also be identified through increased variance in ecosystem state indicators (i.e., fire occurrence) during or just before the threshold event (D) and a bimodal frequency distribution of ecosystem state indicators across the study landscape (E) [Source: Re-created and modified from Marshall et al. 2021].

Why this matters globally

As wildfires increasingly threaten biodiversity, health, infrastructure, and livelihoods, nations urgently need early warning systems that reflect ecological complexity. From the boreal forests of Canada to the savannas of Africa and the Amazon’s tropical canopies, the need for functionally-informed fire prediction is both local and global.

Integrating plant traits with high-resolution remote sensing can help:
•    Improve the accuracy of fire risk assessments
•    Improve fire behaviour models
•    Improve fuel load estimation
•    Guide targeted land management and mitigation
•    Inform resource allocation for firefighting
•    Save billions in economic, environmental, and social costs

A Call to Action

We’re calling on governments, researchers, and environmental agencies worldwide to invest in next-generation fire forecasting systems that incorporate plant functional trait-based approach for early detection of fire. The technology exists. The science is ready. What we need now is the global will to act — before the next fire starts.

Shawkat Sohel, Research Fellow at the FRI

References

1. Mitchell, R.M., Martin, A.R. Fire, flammability and functional traits at the forefront of global change ecology. Function. Ecol. 37, 2767–2769 (2023).
2. Sohel, M.S.I., Herbohn, J.L., Zhao, Y., McDonnell, J.J. Sap flux and stable isotopes of water show contrasting tree water uptake strategies in two co-occurring tropical rainforest tree species. Ecohydro. 16, e2589 (2023).
3. Yebra et al. A fuel moisture content and flammability monitoring methodology for continental Australia based on optical remote sensing. Rem. Sen. Environ. 212, 260-272 (2018)
4. Rao et al. Plant-water sensitivity regulates wildfire vulnerability. Nature Ecol. Evol. 6, 332–339 (2022)
5. Rao et al. Satellite-based vegetation optical depth as an indicator of drought-driven tree mortality. Rem. Sen. Environ. 227, 125–36 (2019)
6. Konings, A.G., Rao K, Steele-Dunne SC. Macro to micro: microwave remote sensing of plant water content for physiology and ecology. New Phytol. 223, 1166–1172 (2019); 
7. Tian et al. Coupling of ecosystem-scale plant water storage and leaf phenology observed by satellite. Nat. Ecol. Evol. 2, 1428–1435 (2018).
8. Bernardino et al. Estimating vegetation water content from Sentinel-1 C-band SAR data over savanna and grassland ecosystems. Environ. Res Let 19, 034019 (2024). 
9. Monteiro et al. Remote sensing of vegetation and soil moisture content in Atlantic humid mountains with Sentinel-1 and 2 satellite sensor data. Ecol. Indic. 163, 112123 (2024). 
10. Holtzman et al. L-band vegetation optical depth as an indicator of plant water potential in a temperate deciduous forest stand. Biogeosci 18, 739-753 (2021)
11. Zhong et al. Quantitative assessment of various proxies for downscaling coarse-resolution VOD products over the contiguous United States. Inter. J App. Earth Obs. Geoinform. 130,103910 (2024)
12. Geoscience Australia. (n.d.). Digital Earth Australia Hotspots
13. Google. (n.d.). Google AI wildfire detection
14. Marshall et al. Conceptualising the Global Forest Response to Liana Proliferation. Front. For. Glob. Change 3, Article 35 (2020)
15. Bestelmeyer et al. Analysis of abrupt transitions in ecological systems. Ecosphere 2, 1–26. (2011)