| 30 Sep 2026
Quantifying the heterogeneous impacts of drought across the distribution of wildfire scale
Abstract. Wildfire research has traditionally viewed drought as a meteorological factor, but multi-hazard frameworks highlight it as a critical driver of fire characteristics. Despite this conceptual shift, existing studies have largely focused on mere association between drought and wildfire, leaving a critical gap in quantifying drought’s impact on wildfire scale. Addressing this limitation, we analyzed observational records of nearly 12,000 wildfires in South Korea from 2000 to 2024 using the Quantile Treatment Effect (QTE) approach to quantify drought impacts on burned area and wildfire duration across the distribution of wildfire scales (0.05 to 0.95 quantile range). Our analysis showed that the influence of drought increased markedly toward the upper quantiles of the distribution of wildfire scale, leading to a maximum increase in burned area of 510 ha and wildfire duration of 1.2 days. At the upper tail, the drought-induced relative increase in burned area was up to 13.7 times higher and strongly exceeded the increase in wildfire duration. To our knowledge, this is the first study to quantify the quantile-dependent impacts of drought on wildfire characteristics using observed wildfire scale data. These findings advance our understanding of drought-wildfire interactions and provide an empirical foundation for multi-hazard wildfire risk assessments.
This study investigates the effects of drought on wildfire burned area and duration. The authors need to address the following points to improve the clarity and reproducibility of the methods and results.
(Introduction)
An important aspect of this study is its examination of the relationship between drought and wildfire from a multi-hazard perspective. The authors need to provide a more detailed review of the main analytical methods used in previous drought–wildfire and related multi-hazard studies, including their inherent limitations. Specifically, the authors must explicitly articulate the added value and unique benefits of applying the QTE approach over the previous methodologies introduced in the Introduction.
The authors also need to introduce any previous applications of Quantile Treatment Effect (QTE), the central analytical method employed in this study, to drought or other natural hazards. Although the purpose of using QTE is understandable, the rationale for selecting this method should be strengthened by linking relevant applications to the statistical characteristics of the wildfire data.
(2 Study area and data)
Since most readers are unfamiliar with local geographical names in Korea, it would be more appropriate to relocate the study area map—currently presented in Figure S1—into the main text or integrate it into Figure 1.
(3.1 Classification of wildfire events: multi-hazard vs single-hazard)
SPI is a key indicator used to determine drought exposure and classify wildfire events into multi-hazard and single-hazard groups. The authors need to provide the basic SPI calculation procedure using equations and specify the precipitation data period, standardization reference period, probability distribution, parameter estimation method, and missing-data treatment.
In line 96, the authors stated that “Following McKee et al. (1993), we defined a drought event as a period of at least three consecutive months with SPI-6 ≤ -1.0.” However, the authors should verify whether McKee et al. (1993) actually defined drought events based on SPI-6. If this is not the case, an explicit rationale for adopting SPI-6 must be provided. Additionally, the authors need to explain how minor drought events were handled or filtered in their analysis.
(3.2 Quantile Treatment Effect (QTE) approach)
Although the manuscript defines potential outcomes, propensity scores, and QTE, the authors need to clarify how these concepts are implemented in the analysis of the observational data in this study. The authors also need to explain the rationale for selecting the confounding variables, their measurement timing, the methods used to assess comparability between the two groups, and the potential influence of unobserved confounders on the results.
To facilitate reproducibility, the authors need to provide information illustrating the analytical data structure or specific examples showing how the location and timing of each wildfire event, drought indicator values, group classification, covariates, and outcome variables are linked and used to estimate QTE.
In Eq. (2), is already defined as the multi-dimensional vector of pre-treatment confounders. The authors need to verify whether X in Eq. (3) is the same variable. If not, a different notation should be used to avoid confusion.
(4.1 Empirical distributions of total burned area and wildfire duration: multi-hazard vs single-hazard)
In line 207, the authors stated that “We compared the Empirical Cumulative Distribution Functions (ECDFs) of single- and multi-hazard events to capture their distributional differences.” The authors must provide statistical significance test results to substantiate these observed differences.
(4.2 Role of drought across the distribution of wildfire scale using QTE approach)
The maximum QTE increased substantially after events with burned areas below 1 ha were excluded. However, the same quantile corresponds to different wildfire magnitudes in the two datasets. The authors need to report the observed burned area and duration corresponding to key quantiles in each dataset and discuss how the change in the analyzed sample affects the comparison and interpretation of the two sets of results.
The authors also need to discuss the fluctuations in the QTE estimates for wildfire duration around the 0.8 quantile in Fig. 4(d).
(4.3 Comparative analysis of drought impacts on total burned area and wildfire duration)
In line 277, the authors cite Fig. S6; however, Fig. S6 is not in the Supplement.