Ä¢¹½´«Ã½

In March 2025, southeastern Korea experienced its largest and most destructive wildfires on record. More than a dozen fires broke out on March 22nd and 23rd and spread rapidly over the following days. More than 48,000 hectares burned - over 20,000 hectares more than the second most devastating wildfires, in April 2022, and more than ten times the annual average burnt area.

With 32 casualties, the fires are also South Korea's deadliest wildfires on record, additionally injuring 45 people, and displacing about 37,000 residents (. Most victims were in their 60s and 70s. Uiseong was hardest hit, with 26 deaths, while four occurred in Sancheong. Around 5,000 buildings, including homes, factories, and farms, were destroyed ().

Rural and suburban areas, especially those where homes border forests (WUI zones), were hit hardest. Uiseong also lost the historic Gounsa Temple, built in 618 AD, and 30 other cultural heritage sites, including Joseon-era relics ().

While March to April is the season commonly experiencing wildfires, this year was marked by exceptionally low rainfall in the months preceding the fires and very high temperatures, more than 10C above average, on the days of the outbreak. To analyse whether and to what extent human-induced climate change exacerbated these conditions and what role vulnerability and exposure played in the historic event, researchers from South Korea, the Netherlands, Sweden, Denmark, Mexico, the United States and the United Kingdom undertook an attribution study on the fire weather conditions as well as the preceding dryness.

To capture the event, the team looked at the 5-day maximum Hot-Dry-Windy Index (HDWI) in March, over all of South Korea (see figure 1). The HDWI multiplies the maximum wind speed with the maximum vapour pressure deficit - which combines temperature and humidity to define the drying power of the atmosphere - which reflects high fire risk. In addition rainfall between February and March over the same region was also analysed.

Main findings

• The fires particularly impacted areas where forests border residential, industrial, and heritage sites, which are crucial for risk reduction and land-use planning. Many of the affected people were older adults or individuals with limited mobility, especially in rural and peri-urban areas, making evacuation difficult.
• Even in today's climate, that has warmed by 1.3°C due primarily to the burning of fossil fuels, the combination of high temperatures, low humidity and high wind speeds (HDWI) observed over the 5 days following March 22nd, when the fires broke out, was very unusual. In the current climate they are expected on average about once every 300 years.
• However, independent of how the HDWI is calculated, based on weather observations, the event would have been extremely rare if the climate hadn't warmed and the intensity of the peak 2025 March HDWI is about 25% more intense in today's climate compared to the cooler pre-industrial climate.
• To determine the role of climate change in this observed trend we combine the observation-based estimate with climate models. Most climate models also show an increase in the recent past, but weaker than the observed trend. Combining both, we find an increase of about 15% in the intensity of the HDWI and a doubling in likelihood. An increase of similar magnitude is observed in data from 9 weather stations in the region most affected by the fires.
• We further estimate how the peak March HDWI would change in a 1.3°C warmer climate from today, that is 2.6°C above pre-industrial and estimated by around 2100 under current policies. We find that all climate models project a further increase in the peak March HDWI with continued warming of about 5% in intensity and a further doubling of the likelihood. This is stronger than the change simulated by the models alone up until today, suggesting that the observed trend is indeed attributable to human-induced climate change but has not emerged in all models yet.
• We also analysed potential changes in the 5-day maximum temperatures in March, finding that while rare in today's climate, the return time of 75 years is not as exceptional as for the combined index. The changes in both observation-based products and models are stronger than in the HDWI, even though the climate models again show a much lower trend than observations. This suggests that the trend in the HDWI is primarily driven by the strong increase in temperature, but that the extremeness of this year's event, with a return time of more than 300 years in today's climate, is not just due to the high temperatures, which only have a return time of 75 years.
• We further analyse the rainfall in the months preceding the outbreak of the fire. Here we find a slight drying trend in the observation-based products as well as most weather stations in the country. However, this drying is not represented in climate models, which may be due to Korea's complex geography as a relatively small landmass surrounded by ocean, as well as rainfall which experiences long-term decadal changes. Thus, in contrast to the warming and changes in fire weather, we cannot attribute it to human-induced climate change or natural variability.
• South Korea has planted billions of trees since the 1970s to reverse historic deforestation, improve biodiversity, reduce soil erosion, absorb carbon emissions and provide numerous social and cultural benefits. This year's fires have renewed concerns about the effect of tree-planting on wildfire risk, with continuous forest cover resulting in high fuel loads and increased fire risk. While afforestation provides many benefits, the fuel loads near human settlements need to continue to be carefully managed to mitigate dangerous fires, particularly as climate change increases the likelihood of intense fire-prone weather conditions.
• South Korea has made significant progress in wildfire detection, suppression, and early warning. As fire seasons become longer and more intense, there is a growing opportunity to build on these strengths by further aligning wildfire preparedness with infrastructure, land-use, and emergency planning frameworks.