The FWI is the world’s most widely adopted benchmark for assessing atmospheric conditions that promote wildfire ignition, spread, and intensity. Developed by the Canadian Forest Service, it integrates four meteorological variables: air temperature, relative humidity, precipitation, and wind speed. Through a chain of intermediate components, including the Fine Fuel Moisture Code, the Duff Moisture Code, the Drought Code, and the Initial Spread Index, the index captures how readily surface fuels dry out and how fast flames could travel once ignited. Because these components interact nonlinearly, small concurrent shifts in temperature or moisture can produce disproportionate jumps in fire danger during prolonged dry spells.

A central challenge for regional fire-risk studies is that global climate models are too coarse and too biased to represent local fire weather reliably. To overcome this, the researchers turned to the NASA Earth Exchange Global Daily Downscaled Projections dataset, known as NEX-GDDP-CMIP6, which applies statistical bias correction and spatial disaggregation to CMIP6 simulations, producing daily data on a 0.25-degree grid. The team evaluated 19 models from this ensemble against ERA5-Land reanalysis data for the historical period 1980 to 2014, using mean bias, root mean square error, and Pearson’s correlation to judge how faithfully each model reproduced the four variables feeding the FWI.

The validation results were revealing. Temperature emerged as the best-simulated variable, with correlations between roughly 0.92 and 0.94 across individual models, reflecting the fact that surface temperature is governed mainly by the surface energy balance and large-scale circulation, processes that general circulation models handle well. Wind speed and relative humidity performed moderately, while precipitation proved the weakest link, with correlations of only about 0.52 to 0.58. This is unsurprising: tropical precipitation depends on deep convection, moisture transport, cloud microphysics, and interactions with the complex Andean topography, all of which must be approximated through parameterizations that vary from model to model.

Crucially, no single model outperformed the others across all variables and metrics. Instead, the multi-model ensemble mean delivered the lowest errors and the highest correlations for every variable, reaching 0.95 for temperature, 0.79 for wind speed, 0.73 for relative humidity, and 0.63 for precipitation. By averaging out the idiosyncratic errors of individual simulations, the ensemble provided the most robust foundation for the fire danger projections, though the authors caution that uncertainty in precipitation and moisture-related variables still propagates through the nonlinear structure of the FWI and should temper interpretation of the results.

The historical baseline confirmed a pronounced seasonal rhythm in fire weather. During December through February, the dry season, Moderate and High hazard categories covered 48 percent and 31 percent of the study area respectively, with the highest index values concentrated over La Guajira, the Colombian Caribbean coast, and northern Venezuela. By June through August, the Low category dominated 96 percent of the domain as the wet season peaked, and September through November remained similarly benign. This cycle is orchestrated by the seasonal migration of the Intertropical Convergence Zone and the northeasterly trade winds, which together control when fuels dry and when moisture returns.

The projections for 2025 to 2100 show that climate change amplifies this seasonal cycle rather than erasing it. Under the low-emission SSP1-2.6 scenario, the Low category still covers 87 percent of the region on average, but under the high-emission SSP5-8.5 pathway it shrinks to 64 percent, while the Moderate category expands to 35 percent and High hazard appears for the first time in the mean conditions. The contrast is even starker for extreme events: the Very High category dominates 66 percent of the study area during extreme weather under SSP1-2.6, rising to 74 percent under SSP2-4.5 and 83 percent under SSP5-8.5, a clear signal that radiative forcing expands the territory exposed to the most dangerous fire weather.

Seasonally, the changes concentrate almost entirely in the December-February dry season, with smaller shifts in March through May. Under SSP5-8.5, High-risk conditions during the dry season reach 49 percent of the region and Very High risk emerges at 17 percent, while the wet-season months remain overwhelmingly Low-risk across all scenarios. The authors attribute this pattern to a reorganization of hydroclimatic processes over the tropical Atlantic: anthropogenic warming enhances tropospheric stability and suppresses deep convection, while an intensifying Caribbean Low-Level Jet and an expanding North Atlantic Subtropical High promote the advection of dry, subsiding air. The result is a growing propensity for flash droughts, in which abrupt increases in vapor pressure deficit strip moisture from fine fuels to critical flammability thresholds even without extreme rainfall deficits.

The study’s implications reach well beyond climate modeling. Because the FWI does not account for vegetation continuity, topography, or ignition sources, which in the Neotropics are predominantly human-caused, the projections should be read as an intensification of fire-favorable meteorology rather than forecasts of individual fires. Even so, the persistent identification of La Guajira, the lower Colombian Caribbean basin, and the northern arc of Venezuela as danger hotspots offers a quantitative basis for shifting from reactive firefighting toward anticipatory risk management, including early-warning infrastructure, preventive prescribed burns, and pre-season mobilization of resources. The urgency is underscored by recent history: carbon emissions from fires in Venezuela during the 2024-2025 seasons were roughly 50 percent above the historical average, and about 15 percent of the region’s páramo habitat burned between 1985 and 2022.

This is the first comprehensive FWI assessment for northern South America built on NEX-GDDP-CMIP6 projections, and it delivers a quantitative baseline of atmospheric fire hazard independent of the stochastic variability of human ignitions. The message for the region is sobering but actionable: the climate signal for wildfire danger emerges early, meaning even moderate warming can expand fire-favorable conditions before any dramatic shift in annual rainfall. As the authors note, integrating these meteorological projections with dynamic vegetation models, fuel loads, land-cover change, and socioeconomic ignition patterns remains the next frontier, but the framework now exists to put climate-informed fire danger at the center of land-use planning and cross-border adaptation strategy in one of South America’s most ecologically diverse corners.

Subject of Research: Projection of forest fire hazard in northern South America using the Fire Weather Index driven by NEX-GDDP-CMIP6 climate models

Article Title: Projection of Forest Fire Hazard in Northern South America Using the Fire Weather Index Driven by NEX-GDDP-CMIP6 Climate Projections

Article References: Guerra, Y., Arregocés, H. A., & Rojano, R. (2026). Projection of Forest Fire Hazard in Northern South America Using the Fire Weather Index Driven by NEX-GDDP-CMIP6 Climate Projections. Environmental Challenges, 25, Article 101659. https://doi.org/10.1016/j.envc.2026.101659

Image Credits: AI Generated

DOI: 10.1016/j.envc.2026.101659

Keywords: wildfire, Fire Weather Index, CMIP6, NEX-GDDP, climate change, Colombia, Venezuela, La Guajira, SSP scenarios, precipitation, drought, fire danger

The research, led by Michele Salis and Liliana Del Giudice of Italy’s National Research Council Institute of BioEconomy, together with colleagues from Spain, Iran and other institutions, focused on two time windows: a historical baseline from 1985 to 2014 and a near-future period from 2035 to 2064. The team drew on three state-of-the-art climate models from the Coupled Model Intercomparison Project Phase 6, known as CMIP6, namely ACCESS-ESM1-5, EC-EARTH3 and MPI-ESM1-2-HR, each run under three Shared Socioeconomic Pathways spanning low to very high greenhouse gas emissions. What sets this study apart is the resolution: the raw climate model output, which arrives at a coarse scale of tens of kilometers, was statistically downscaled to a grid of just one square kilometer using a two-step technique combining parametric quantile mapping with geostatistical regression, before feeding into fire danger calculations and fire spread modeling.

To quantify fire danger, the researchers applied the Canadian Forest Fire Weather Index System, the same harmonized method used across Europe by the European Forest Fire Information System since 2007. The FWI integrates daily maximum temperature, minimum relative humidity, wind speed and cumulative precipitation into a single rating of potential fire intensity, and it depends on three fuel moisture codes that track the dryness of fine fuels, duff and deep organic layers. The team also calculated the Daily Severity Rating, a metric that amplifies the influence of extreme FWI values and estimates how difficult a fire burning that day would be to suppress. Both indices were computed for every pixel of the island on every day of the two study periods, for each climate model and each emissions scenario.

The signal that emerged was consistent across all three models, though its magnitude varied. Under the historical baseline, days with very high FWI values were rare, concentrated almost entirely in the May-to-October fire season, when high and very high danger conditions already accounted for roughly 43 to 46 percent of days. By the 2035-2064 period, the picture shifts dramatically. Under the highest emissions scenario, SSP5-8.5, the EC-EARTH3 model projected that very high FWI days would more than double during the fire season, climbing from about 4.6 percent to 10.8 percent of days, while the ACCESS model showed a rise from 4.0 to 10.1 percent. Even the more conservative MPI model projected a near doubling, from 4.5 to 8.2 percent. The Daily Severity Rating told a parallel story, with very high severity days under EC-EARTH3 rising from 8.4 percent to 17.1 percent of fire-season days under the worst-case pathway.

Underlying these danger trends is a straightforward climatic mechanism. All models project warmer and drier conditions for Sardinia, with maximum temperature increases reaching up to about 3.3 degrees Celsius annually and 3.9 degrees during the fire season under high emissions. Precipitation declines across the board, particularly in western and central sectors of the island, and the number of rainy days drops markedly, especially during the hottest months, pointing to longer and more frequent dry spells that desiccate fuels. Intriguingly, the analysis of wind conditions on extreme danger days showed that the projected increase in very high fire weather comes mostly from more days with low and moderate winds rather than stronger winds, suggesting that rising temperatures and reduced rainfall, not windier conditions, will be the dominant drivers of intensifying fire risk.

But the study went a crucial step beyond danger indices. Weather alone does not make a wildfire; fires need ignitions, flammable fuels and terrain that allows spread. To capture those interactions, the team built a probabilistic fire simulation framework using the Minimum Travel Time algorithm, which computes the fastest possible paths of fire growth across a landscape and calculates spread rates with the Rothermel surface fire model, while crown fire initiation follows the Van Wagner approach. The researchers ran tens of thousands of simulated fire footprints, 50,000 under baseline conditions alone, at a fine 150-meter resolution, randomly drawing weather streams, wind directions and fire durations from frequency distributions derived for three weather zones covering northern, central and southern Sardinia. Ignition points were placed according to probability grids built from two decades of historical fire records, and fuel moisture inputs were anchored in field measurements collected on the island, including live fuel moisture values between 65 and 100 percent depending on vegetation type.

The hazard outputs are striking. Under historical conditions, the models simulated an average annual burned area of roughly 11,700 to 15,400 hectares, consistent with the observed mean of about 14,700 hectares per year, or 0.65 percent of the island’s surface. In the future simulations, that figure nearly doubles or more. Under EC-EARTH3 and SSP5-8.5, the average annual area burned jumps to approximately 33,590 hectares, a 250 percent increase over the baseline, while ACCESS projects more than 30,000 hectares per year under the two higher-emission scenarios. Crown fire area, a measure of the most intense and dangerous fire behavior in forests, rises from between roughly 730 and 1,060 hectares per year historically to as much as about 2,020 hectares per year under the worst-case combination of model and scenario.

Perhaps most important is where these changes unfold. Historically, burn probability hotspots have been concentrated in southern Sardinia, with annual burn probabilities exceeding 4 hectares per square kilometer in localized zones. The future projections show those hotspots expanding in both magnitude and extent, pushing into western and interior regions of the island. Under SSP5-8.5, much of the island shifts toward higher burn probability classes, with localized increases above 10 hectares per square kilometer in the main southern hotspot zones. Crown fire probability rises in parallel, with the most pronounced gains in southern and western Sardinia, where some areas reach values exceeding 8.5 hectares per square kilometer. This westward and inland expansion matters because those regions, including areas hit by major fires such as the Montiferru blaze of 2021, are already among the most fire-affected parts of the island, and inland zones lack the moderating influence of the sea that cushions coastal areas.

One of the study’s most thought-provoking findings is a caveat about what fire danger indices can and cannot tell us. Despite the systematic rise in FWI values, the simulated average flame length and average fire size did not increase consistently from the historical to the future period, partly because the extra extreme-danger days occurred disproportionately under low and moderate winds. The authors stress that extreme fire weather is necessary but not sufficient for large wildfires: a blaze must also start, escape initial suppression and find continuous fuels to spread across. Simply equating projected FWI maps with future fire risk, as some earlier studies have done, therefore overstates the certainty of the picture. Probabilistic spread modeling, which integrates weather, ignitions, fuels and landscape connectivity, provides a far more realistic estimate of where fire exposure will actually grow.

The researchers acknowledge limitations, notably that fuel layers and ignition patterns were held constant to isolate the climate signal, an assumption that could either understate or overstate future hazard depending on how land abandonment, fuel accumulation and fire activity themselves evolve. Still, the message for the Mediterranean is unambiguous. The authors argue that suppression-oriented strategies alone cannot cope with the projected intensification, and they call for proactive, integrated risk management: landscape-scale fuel treatments, grazing, prescribed burning, fire-resilient land-use mosaics and spatial planning that accounts for expanding fire-prone zones. As record seasons sweep southern Europe, with more than one million hectares burned across the EU in 2025 alone, Sardinia’s high-resolution glimpse of the future offers both a warning and a practical roadmap for landscapes on the front line of a warming world.

Subject of Research: Projected effects of climate change on wildfire danger and hazard in Sardinia, Italy

Article Title: Climate change will reshape future wildfire danger and hazard in Sardinia, Italy

Article References: Climate change will reshape future wildfire danger and hazard in Sardinia, Italy. (n.d.). https://doi.org/10.1007/s11069-026-08320-5

Image Credits: AI Generated

DOI: 10.1007/s11069-026-08320-5

Keywords: wildfire, climate change, Sardinia, Mediterranean Basin, Fire Weather Index, burn probability, fire spread modeling, CMIP6, SSP scenarios, crown fire, Daily Severity Rating, fire management