Experimental Investigation of Lightning-Induced Ignition and Smoldering–Flaming Transition in Boreal Forest Fuels of the Daxing’anling Region, Northeast China

Abstract

Lightning-ignited wildfires are an increasing hazard in boreal forests, with their frequency amplified by global warming and more frequent thunderstorms. However, the mechanisms governing lightning-induced ignition and the subsequent smoldering–flaming transition remain poorly understood. This study aims to understand the ignition mechanisms of lightning-induced forest fires by combining a physics-based heat-balance model and controlled laboratory simulations. Experiments were conducted using twelve representative surface fuel types collected from six typical forest types in the Daxing’anling region, a lightning fire-prone area in northern China. Three fundamental stages of fire behavior development were systematically investigated, including the lightning-induced ignition, smoldering propagation, and the smoldering-to-flaming transition. Fuel moisture content was varied from 5% to 45%, and wind speed was adjusted between 0 and 5 m/s. The results demonstrated that discharge energy and wind speed significantly increased ignition probability, while fuel moisture content was negatively correlated with smoldering spread rate. Wind speed showed the greatest influence on the smoldering-to-flaming transition. The findings provide new mechanistic insights into the thermal and physical processes driving lightning-induced fires, supporting predictive modeling of ignition thresholds and fire behavior under changing meteorological and fuel conditions.

1. Introduction

Forest fires are an intrinsic component of natural ecosystems, playing vital roles in regulating biodiversity, promoting nutrient cycling, and maintaining ecological balance. However, over recent decades, the frequency, intensity, and spatial distribution of wildfires have undergone profound transformations under the influence of global climate change [1,2]. Among various ignition sources, lightning-induced wildfires (LIWs) account for approximately 15%–20% of the total global burned area, dominating ignition processes in boreal and tropical ecosystems where anthropogenic activity is limited [3,4,5]. In high-latitude boreal and Arctic regions, lightning is an important natural ignition source and has increasingly contributed to wildfire occurrence under recent warming. These fires can cause substantial carbon emissions through the combustion of forest litter, organic soils, peat, and permafrost-affected fuels, with Siberian wildfire emissions reaching or exceeding 150 Tg C yr⁻¹ in several extreme fire years [6,7].

Under extremely dry conditions, “dry thunderstorms”—lightning events with negligible precipitation—have been identified as the primary ignition mechanism for many large-scale wildfires [8]. Moreover, fire weather, characterized by high temperature, low humidity, and strong convective instability, remains one of the most critical drivers of wildfire occurrence and spread [3]. Previous modeling studies suggest that lightning activity is sensitive to climate warming, with global lightning projected to increase by about 11%–12% per degree of warming, while substantially larger increases have been projected for boreal forests and Arctic tundra under future climate scenarios [9,10,11,12]. This increase arises from enhanced convective available potential energy (CAPE) and cloud-top electrification in a warmer atmosphere [13]. As a result, lightning-induced forest fires have become an increasingly important climatic hazard in many regions [14,15,16].

In recent years, the rapid development of lightning detection and location networks has substantially improved the observation of lightning activity and its linkage to wildfire occurrence. In particular, the World Wide Lightning Location Network (WWLLN) has enabled global-scale analyses of lightning stroke density and temporal variability, providing an increasingly important observational basis for studies of lightning-caused fire risk [17,18]. At the same time, regional evaluations have shown that the detection efficiency and location accuracy of lightning networks vary across regions, which should be considered when linking lightning observations to fire ignition records [19].

Beyond lightning climatology itself, several countries have developed operational or semi-operational frameworks to estimate lightning-caused fire potential. For example, a lightning-caused wildfire ignition forecasting model has been developed for operational use in Victoria, Australia, using lightning occurrence, fuel condition, and fire-weather variables to generate spatially explicit ignition forecasts [20]. In the United States, the Wildland Fire Assessment System (WFAS) provides spatial products such as Dry Lightning/Dryness, Dry Lightning/Estimated Rain, and Potential Lightning Ignition based on lightning-location and fire-weather data. More recently, machine learning approaches have been used to model dry-lightning-induced wildfire probability in Tasmania, showing the importance of fuel moisture, vegetation type, topography, and fire weather in determining whether a lightning strike initiates a fire [21].

Under natural conditions, high-energy electrical discharges from cloud-to-ground lightning can instantaneously interact with surface fuels such as dry litter, moss, peat, or duff layers—materials that remain highly flammable even at relatively low temperatures due to their porous structure and low thermal diffusivity [22,23]. The intense but localized energy deposition from the discharge column initiates pyrolysis and smoldering within milliseconds, forming a subsurface ignition kernel that may persist undetected for hours or even days [4,24].

Because of its low thermal signature and the absence of visible flames, smoldering ignition often escapes detection by conventional satellite sensors or aerial surveillance, resulting in a “latent ignition window” during which combustion continues slowly beneath the surface [25,26]. This concealed burning phase can later transition to flaming combustion when environmental conditions, such as wind and oxygen availability, become favorable—thereby amplifying the risk of rapid fire spread and intensification [27,28,29].

Despite the increasing prominence of lightning as a natural ignition source worldwide, most prior studies have primarily focused on spatial distributions, meteorological drivers, or macro-scale statistical modeling [30,31]. At the process scale, a parallel body of research has developed mathematical descriptions of forest fuel drying, pyrolysis, and ignition, including models that resolve coupled heat and mass transfer in combustible materials. Recent reviews have summarized experimental and mathematical approaches to forest fuel drying, pyrolysis, and ignition, emphasizing that ignition is controlled by moisture evaporation, thermal decomposition, heat transfer, and the evolving thermophysical state of the fuel [32]. More detailed numerical studies have further simulated forest fuel pyrolysis and moisture evaporation under high-temperature fire exposure using coupled heat-transfer formulations [33]. Nevertheless, recent reviews of lightning-induced wildfires still point out that the physical mechanism of lightning-strike ignition and its transition from ignition to smoldering or flaming combustion remain insufficiently resolved experimentally and mechanistically [34].

In contrast, the fundamental ignition mechanism—particularly the energy balance between long-continuing-current (LCC) discharges and moisture-dependent heat losses—remains poorly understood [35,36,37]. Although LCC strokes, lasting longer than 100 ms, have often been considered potentially important for lightning ignition because their prolonged energy delivery may favor sustained heating, few studies have systematically examined how variations in fuel type and structure affect the transition from transient heating to self-sustaining combustion [38,39,40]. The processes linking ignition, smoldering spread, and later transition to flaming are especially poorly resolved at the fuel-bed scale [41,42,43].

The Daxing’anling region of northeastern China provides a representative boreal setting in which to investigate these processes. It is the largest boreal forest region in China, and one of the country’s most important areas for lightning-caused fires. The region is characterized by extensive larch-dominated forests, deep organic layers, and porous surface fuels, all of which are relevant to lightning-related ignition and early combustion development under boreal conditions [44,45,46]. Field observations have also shown that many ignition-prone surface fuels in this region are composed of larch litter, moss, and humus layers, whose thermal response depends strongly on moisture content and structural characteristics [23,45]. However, the quantitative relationships among discharge characteristics, fuel moisture, and ignition thresholds remain poorly constrained [4,46].

Against this background, the present study combines analytical heat-balance modeling with controlled laboratory arc-discharge experiments to investigate the sequence of ignition and early combustion development in representative boreal forest fuels from Daxing’anling. The focus is placed on three linked stages, namely ignition, smoldering spread, and transition to flaming combustion, with particular attention to the roles of fuel moisture, ambient temperature, and wind conditions.

Specifically, the objectives of this study are to:

• Develop a dimensionally consistent, moisture-coupled heat-balance model that integrates discharge energy, fuel moisture, and wind-driven convective losses;
• Determine the critical smoldering ignition thresholds of representative Daxing’anling forest fuels under simulated long-continuing-current (LCC) lightning discharges;
• Quantify the effects of moisture content, wind speed, and ambient temperature on smoldering-to-flaming transitions.

Through this combined modeling–experimental approach, the study seeks to elucidate the dominant mechanisms controlling ignition and transition dynamics, providing a physically grounded framework for predicting lightning fire initiation and escalation. This work contributes to a deeper understanding of the thermal response mechanisms underpinning lightning-caused fires and fills a critical knowledge gap in the mechanistic study of lightning-induced ignition behavior.

The remainder of this paper is organized as follows. Section 2 presents the theoretical formulation of the moisture-coupled, dimensionally consistent heat-balance model. Section 3 describes the experimental setup and procedures, including Daxing’anling fuel sampling and the operation of the long-continuing-current (LCC) discharge simulator. Section 4 reports the experimental and modeling results, focusing on ignition thresholds, smoldering propagation, and smoldering-to-flaming transitions. Finally, Section 5 concludes the study by summarizing the integrated experimental and modeling findings, emphasizing the quantitative roles of environmental factors in lightning ignition, and discussing their implications for ignition risk assessment and predictive modeling under variable surface conditions.

2. Materials and Methods

2.1. Study Framework

Figure 1 shows the stage-based framework used in this study to examine ignition and early combustion development in forest surface fuels under controlled laboratory conditions. Three stages are considered: ignition, smoldering spread, and transition to flaming combustion. During the ignition stage, electrical arc heating supplies localized energy to the fuel bed and may initiate combustion under suitable moisture conditions. Once ignition has occurred, smoldering can spread laterally through the surface fuel layer, with its development constrained by fuel structure and moisture status. Under more favorable thermal and ventilation conditions, the smoldering front may further intensify and develop into flaming combustion.

The figure is intended to illustrate the analytical and experimental framework of the present study, rather than the full range of natural lightning-caused ignition pathways in forests. In natural settings, lightning ignitions may involve more complex processes, including current transfer through trees, root-system effects, and the formation of secondary ignition sources. The framework adopted here is therefore a simplified representation of ignition-relevant processes in surface fuels, and it provides the basis for the controlled experiments and heat-balance analysis presented below. Accordingly, it is not intended as a direct physical representation of all cloud-to-ground lightning ignition pathways in forests, but as a simplified basis for examining the response of surface fuels to controlled electrical arc heating.

2.2. Theoretical Analysis on Thermal Balance of the Smoldering Ignition Process

To fundamentally understand the relationship among arc discharge energy (E), fuel moisture content (MC), and ambient wind speed (WS) in the simulated lightning-induced smoldering ignition process, a simplified thermal energy transfer analysis was conducted in this study based on the principle of energy conservation, as illustrated in Figure 2 [22,43]. The present framework is restricted to the surface fuel layer and does not explicitly account for the underlying soil. This is a simplification of the natural system, since soil texture, soil moisture, and mineral composition may affect both fuel condition and the transfer of lightning-related energy near the ground.

The point-source assumption is used to represent the localized energy deposition characteristic of a lightning arc, whose footprint is typically less than a few centimeters. This simplification captures the intense but spatially confined heating within the surface layer, allowing analytical description of heat diffusion and ignition thresholds without resolving the full plasma–fuel interaction. It effectively models the thermal impulse region responsible for initiating smoldering before lateral spread occurs. The point-source approximation is intended to represent the localized energy deposition near the ground-contact region of the discharge, rather than the full spatial extent of the lightning channel. This simplification is most relevant for interpreting near-surface thermal effects at the ignition point, but it does not resolve the full geometry of natural lightning channels and should therefore be regarded as an idealization.

The electric discharge is represented by a localized point heat source corresponding to the arc footprint, and the transient heat transfer within the porous fuel layer is analyzed to determine the minimum ignition energy $E_{\min }$.

This energy needs to not only raise the fuel temperature to its smoldering ignition threshold ($T_S$), but also compensate for the total heat losses occurring over the heating duration t. It includes the energy required to heat the fuel ($E_i$) from ambient temperature T to its smoldering ignition temperature ($T_S$), the energy consumed by moisture evaporation ($E_{MC}$), convective heat loss to the ambient environment ($E_{env}$), and lateral conductive losses to adjacent unignited fuel ($E_{cond}$), which can be expressed by:

$$E\ge E_{\min }=E_i+E_{MC}+E_{env}+E_{cond}-E_{com}$$

(1)

where $E_{com}$ represents the heat released by the combustion reaction during the ignition process, which can be expressed as:

$$E_{com}=m_{com}\cdot H_c\cdot \rho \pi {\gamma }^2\delta$$

(2)

in which $H_c$ denotes the heat of combustion of the fuel (J/kg), and $m_{com}$ represents the combustion rate of the fuel (kg/s). The heat release can be expressed as:

$$m_{com}=\alpha \cdot {\displaystyle \frac{m_{ox}}{\beta }}$$

(3)

where $\alpha$ is the thermal transfer efficiency of the fuel, $\beta$ is the stoichiometric coefficient, and $m_{ox}$ is the oxygen supply rate. The relationship can be expressed as:

$$m_{ox}={\rho }_a\cdot Y_{ox}\cdot v^m$$

(4)

in which ${\rho }_a$ is the air density, $Y_{ox}$ is the mass fraction of oxygen in air, $m$ is an empirical exponent, and $v$ represents the wind speed. Thus, combining Equations (2)–(4), we have:

$$E_{com}=\alpha \cdot {\displaystyle \frac{{\rho }_a\cdot Y_{ox}\cdot v}{\beta }}\cdot H_c\cdot \rho \pi {\gamma }^2\delta$$

(5)

The energy required to heat the fuel from ambient temperature to its smoldering ignition temperature ($E_i$) can be expressed as:

$$E_i=\rho c\pi {\gamma }^2\delta \cdot (T_S-T)$$

(6)

where $\rho$ is the density of the fuel, c is the specific heat capacity, $\gamma$ is the radius of the heated region, and $\delta$ is the thermal penetration depth.

The heat loss due to moisture evaporation in the heated region ($E_{MC}$) can be calculated as:

$$E_{MC}=m_H\cdot L$$

(7)

where $L$ is the latent heat of vaporization of water and $m_H$ is the mass of evaporated water. When the fuel has an oven-dry basis moisture content MC, the heat loss can be expressed as:

$$m_H={\rho }_H\pi {\gamma }^2{\displaystyle \frac{MC}{1+MC}}$$

(8)

in which ${\rho }_H$ is the density of water. Therefore, Equation (7) can be expressed as:

$$E_{MC}={\rho }_H\pi {\gamma }^2{\displaystyle \frac{MC}{1+MC}}\cdot L$$

(9)

The convective heat loss to the environment (E_env) can be estimated as:

$$E_{env}=h\cdot \pi {\gamma }^2\cdot (T_S-T)$$

(10)

where $h$ is the convective heat transfer coefficient (W/m²K), which is related to wind speed (v), and can be expressed as:

$$h=C\cdot v^n$$

(11)

where C is a constant related to fluid properties and surface roughness, and $n$ is an empirical exponent typically ranging from 0.5 to 0.8. Thus, Equation (10) can be rewritten as:

$$E_{env}=C\cdot v^n\cdot \pi {\gamma }^2\cdot (T_S-T)$$

(12)

The conductive heat loss to the surrounding fuel ($E_{cond}$) can be computed as:

$$E_{cond}=q\cdot (2\pi \gamma \cdot \delta +\pi {\gamma }^2)\cdot t$$

(13)

where $q$ is the heat flux, which is determined by the material’s thermal conductivity $k$, the conduction thickness $\Delta L$, and the temperature gradient, as:

$$q=k\cdot {\displaystyle \frac{T_S-T}{\Delta L}}$$

(14)

Thus, Equation (13) can be rewritten as:

$$E_{cond}=k\cdot {\displaystyle \frac{T_S-T}{\Delta L}}\cdot (2\pi \gamma \cdot \delta +\pi {\gamma }^2)\cdot t$$

(15)

Therefore, combining Equations (5), (6), (9), (12) and (15) into Equation (1), the minimum ignition energy required per unit area ($E_{\min }$) can be derived as:

$$\begin{array}{l}{E}_{\min }=W_e\cdot t=\rho c\pi {\gamma }^2\delta \cdot (T_S-T)+{\rho }_H\pi {\gamma }^{2}L\cdot {MC}^{\prime }+C\cdot v^n\cdot \pi {\gamma }^2\cdot (T_S-T)\\ +k\cdot {\displaystyle \frac{T_S-T}{\Delta L}}\cdot (2\pi \gamma \cdot \delta +\pi {\gamma }^2)\cdot t-\alpha \cdot {\displaystyle \frac{{\rho }_a\cdot Y_{ox}\cdot v^m}{\beta }}\cdot H_c\cdot \rho \pi {\gamma }^2\delta \end{array}$$

(16)

in which $W_e$ represents the electrical input power supplied by the discharge and ${MC}^{\prime }={\displaystyle \frac{MC}{1+MC}}$. In the condition when wind speed and moisture content are fixed, Equation (16) can be further transformed into Equation (17), indicating the existence of a minimum ignition duration. Smoldering ignition of the fuel can occur if and only if the arc discharge time exceeds this critical threshold:

$$\begin{array}{l}{t}_{\min }={\displaystyle \frac{v^n\cdot C\cdot \pi {\gamma }^2\cdot (T_S-T)-v^m\cdot \alpha \cdot \frac{{\rho }_a\cdot Y_{ox}}{\beta }\cdot H_c\cdot \rho \pi {\gamma }^2\delta }{W_e-k\cdot \frac{T_S-T}{\Delta L}\cdot (2\pi \gamma \cdot \delta +\pi {\gamma }^2)}}+\\ {\displaystyle \frac{\rho c\pi {\gamma }^2\delta \cdot (T_S-T)}{W_e-k\cdot \frac{T_S-T}{\Delta L}\cdot (2\pi \gamma \cdot \delta +\pi {\gamma }^2)}}+{MC}^{\prime }\cdot {\displaystyle \frac{{\rho }_H\pi {\gamma }^{2}L}{W_e-k\cdot \frac{T_S-T}{\Delta L}\cdot (2\pi \gamma \cdot \delta +\pi {\gamma }^2)}}\end{array}$$

(17)

From Equation (17), it can be seen that when wind speed $v$ is held constant, the minimum ignition time $t_{\min }$ increases with increasing moisture content ${MC}^{\prime }$.

If we define $v^{\prime }=v^n\cdot C\cdot \pi {\gamma }^2\cdot (T_S-T)$, $v^{″}=v^m\cdot \alpha \cdot {\displaystyle \frac{{\rho }_a\cdot Y_{ox}}{\beta }}\cdot H_c\cdot \rho \pi {\gamma }^2\delta$, $v^{‴}=v^{\prime }-v^{″}$, Equation (17) can be reformed as:

$$\begin{array}{l}{t}_{\min }={\displaystyle \frac{\rho c\pi {\gamma }^2\delta \cdot (T_S-T)}{W_e-k\cdot \frac{T_S-T}{\Delta L}\cdot (2\pi \gamma \cdot \delta +\pi {\gamma }^2)}}+{MC}^{\prime }\cdot {\displaystyle \frac{{\rho }_H\pi {\gamma }^{2}L}{W_e-k\cdot \frac{T_S-T}{\Delta L}\cdot (2\pi \gamma \cdot \delta +\pi {\gamma }^2)}}\\ +{\mathrm{v}}^{‴}{\displaystyle \frac{1}{W_e-k\cdot \frac{T_S-T}{\Delta L}\cdot (2\pi \gamma \cdot \delta +\pi {\gamma }^2)}}\end{array}$$

(18)

where $v^{\prime }$ and $v^{″}$ represent the convective heat loss and the heat release from combustion, respectively, both of which are influenced by wind speed.

Equation (18) shows that the increase in wind speed not only enhances the heat release rate of combustion reactions but also increases the rate of heat loss through environmental convection. When wind speed is below a critical threshold, the heat generated by combustion exceeds the convective heat loss, and wind exerts a promoting effect on combustion. Conversely, when wind speed exceeds the critical threshold, convective heat loss surpasses the heat release from combustion, and wind begins to inhibit combustion. Thus, a critical wind speed exists: when wind speed is above the critical value, the minimum ignition time $t_{\min }$ decreases with increasing wind speed; while in the condition that wind speed is below the critical value, the minimum ignition time $t_{\min }$ increases with increasing wind speed. It should be noted, however, that Equation (18) provides a simplified thermal balance framework for interpreting the effect of wind on ignition-related heat transfer, rather than a complete predictive description of the subsequent evolution from ignition to sustained smoldering and flaming transition.

To validate the understanding of the influences of energy input, fuel type, moisture content, and wind speed on the thermal energy transfer, in this study, we performed controlled environment experimental studies. The wind speeds were designed to simulate sub-canopy airflow and fall within the low wind speed range—well below the critical threshold. Therefore, under conditions of constant moisture content ${MC}^{\prime }$, the minimum ignition time $t_{\min }$ decreases with increasing wind speed $v$.

The model adopted here is intentionally simplified. It was developed to describe the first-order balance between heat input and heat loss during ignition under controlled laboratory conditions, rather than to reproduce the full multi-physics complexity of natural lightning-caused ignition. Processes such as transient temperature-field evolution, pyrolysis, electrophysical interactions, and gas-phase chemistry are not explicitly resolved in the present formulation. This simplification should be kept in mind when interpreting the model results. The present framework is useful for examining ignition-related thermal behavior, but it does not represent the full set of coupled processes that may operate during natural lightning-caused fire initiation.

2.3. Study Area and Fuel Sampling

The experimental studies were performed using samples from sampling area in Huzhong Town, Huma County, within the Daxing’anling region of Heilongjiang Province, northeastern China, as shown in Figure 3a,b. This region is representative of a cold-temperate coniferous forest ecosystem, characterized by abundant forest resources and diverse vegetation types. The climate in Daxing’anling is cold and dry, with long, snow-covered winters and short, warm summers. Precipitation is concentrated during the summer months, which are also marked by frequent thunderstorms, conditions that make the region highly susceptible to lightning-induced wildfires [44,47,48].

The experiments described in this study were conducted in an indoor laboratory located in northeastern China. Forest fuel materials used in the experiments were collected from field sites in the Daxing’anling region. As shown in Figure 4, twelve types of forest surface fuels were sampled across six representative forest types and post-fire sites. These sampling sites are representative of lightning-prone coniferous forest ecosystems, ensuring that the collected fuels realistically reflect the physical and chemical characteristics of natural surface layers under frequent lightning strike conditions. For the simulated lightning ignition and smoldering spread experiments, only surface fuels collected from Larix gmelinii (Dahurian larch) forests were used. In contrast, all twelve fuel types were employed in the smoldering-to-flaming transition experiments to assess variability in combustion behavior across different vegetation types.

2.4. Experimental Setup and Design

Figure 5a–d show the experimental setup used to examine ignition and early combustion development in fuel beds with different fuel types, porosities, and moisture contents. The experimental apparatus used for electrical arc-discharge primarily consists of a reactor, an electrical arc-discharge generator, and a video recording system. The reactor is an open-top circular dish with a diameter of 170 mm and a height of 25 mm, constructed from acid-resistant steel as illustrated in Figure 5b. Figure 5b also illustrates the mode of localized electrical arc heating applied to the surface fuel layer in the experiment. This schematic is intended to describe the laboratory heat-input configuration, rather than the full physical pathway of natural cloud-to-ground lightning ignition in forests. The smoldering vertical spread test apparatus consists of a cylindrical open-top reactor and a video recording system as shown in Figure 5c. The reactor measures 52 mm in diameter and 70 mm in height, and features multiple small holes on both the base and sidewalls to facilitate air permeability. The smoldering-to-flaming transition test apparatus consists of a combustion reactor and a video recording system as shown in Figure 5d. The reactor has a square cross-sectional area of 170 × 170 mm² and a height of 50 mm, and is constructed from 10 mm-thick corundum–mullite ceramic.

The arc-induced ignition experiment was designed to investigate the effects of three key control variables—fuel moisture content, wind speed, and arc discharge duration—on ignition behavior. The selection of control variables directly corresponds to the main terms in the thermal balance model and allowing for quantitative validation of the predicted relationships between ignition-relevant electrical input, moisture content, and convective environment. The experimental parameters were set as follows:

• Fuel moisture content was adjusted from 5% (air-dried) to 15%, 25%, 35%, and 45%, with an absolute uncertainty of ±5%;
• Wind speed was varied from 0 m/s (no wind) to 1 m/s, 2 m/s, 3 m/s, 4 m/s, and 5 m/s, with an absolute uncertainty of ±0.5 m/s;
• Arc discharge duration was set at 60 s, 120 s, 180 s, 240 s, and 300 s.

In each trial, 80 g of fuel samples were weighed and placed into the reactor. The long-continuing-current generator produced an arc discharge with an output voltage and current of approximately 1500 V and 0.07 A, respectively, corresponding to an input power of about 100 W. This electrical input was used to impose localized arc heating on the fuel bed. The imposed current duration was not intended to reproduce the full temporal structure of natural lightning exactly, but rather to simulate the sustained thermal input associated with continuing-current components that are considered relevant to ignition processes under laboratory conditions. The entire ignition process was continuously recorded by a high-definition camera.

Wind conditions were simulated using an electric fan placed in front of the experimental platform. The target wind speed was adjusted by varying both the fan power setting and the distance between the fan and the platform. The actual wind speed at the sample location was then measured and verified using an anemometer before each test condition was established. The reported wind-speed values in this study therefore refer to the measured airflow at the level of the fuel bed rather than to nominal fan settings.

After the arc discharge ended, the treated fuel sample was allowed to burn and spread freely, allowing the arc-affected fuel sample to burn and spread freely. Observations were made to determine whether a charred zone developed and whether it exhibited self-sustained smoldering propagation. During the laboratory experiments, ambient conditions were maintained at a temperature of 23 ± 2 °C, relative humidity of 80% ± 5%, and atmospheric pressure of 101.3 kPa (1 atm). The reported relative humidity (80% ± 5%) reflects the ambient laboratory condition during the local rainy summer season. This value was not intended to reproduce the low-humidity atmospheric conditions typically associated with dry thunderstorms; rather, the experimental design focused on the effects of controlled fuel moisture content, wind speed, and discharge input under stable laboratory conditions.

The selected fuel moisture content range (5%–45%) and wind speed range (0–5 m s⁻¹) were chosen to cover a broad but realistic range of fire-season surface fuel and near-surface wind conditions, and to ensure comparability with previous laboratory studies on ignition, smoldering combustion, and fire spread. These two variables have been widely recognized as key controls on fire danger, ignition probability, smoldering limits, and spread behavior in both regional fire-weather analyses and controlled combustion experiments [49,50,51,52,53].

The smoldering vertical spread experiment was designed to investigate the effect of fuel moisture content on vertical smoldering spread. Moisture content varied from 5% (air-dried) to 15%, 25%, 35%, 45%, and 55%, with an absolute uncertainty of ±5%.

The lower layer of the reactor contains a 40 g smoldering carbon puck, approximately 3 cm in height, which serves as a sustained smoldering heat source. The upper layer consists of 50 g of ground surface fuels from Larix gmelinii, approximately 4 cm in height. Heat released from the lower layer initiates smoldering in the upper fuel layer, leading to upward smoldering propagation. The entire combustion and propagation process was continuously recorded using a high-definition camera.

The smoldering-to-flaming transition experiment was designed to investigate the influence of four controlled variables: fuel type, moisture content, wind speed, and ambient temperature. The experimental parameters were defined as follows:

• Twelve forest fuel types collected from six representative forest types in the Daxing’anling region were selected as samples;
• Fuel moisture content was varied from 5% (air-dried) to 15%, 25%, 35%, and 45%, with an absolute uncertainty of ±5%;
• Wind speed was adjusted from calm conditions (0 m/s) to light wind (1 m/s);
• Ambient temperature was increased from 25 °C to 50 °C.

In each trial, the lower layer of the reactor was filled with 90 g of surface fuels from Larix gmelinii that had undergone complete smoldering carbonization, serving as a heat source. The upper layer contained 40 g of forest fuel samples collected from twelve fuel types representing six typical forest types in the Daxing’anling region. The heat released from the smoldering lower layer initiated smoldering in the upper layer, and the subsequent transition to flaming combustion was observed and continuously recorded using a high-definition camera.

2.5. Sample Preparation and Moisture Content Control

Prior to testing, surface fuels from Larix gmelinii used in the lightning ignition and smoldering spread experiments were ground into fine particles using a pulverizer to ensure uniform density, particle size, and organic content. This preparation simulates the deep soil structure of undisturbed forest floors and ensures the repeatability of fire experiments [54,55]. This experimental setup ensured that fuel bed structure and boundary conditions were consistent with the assumptions made in the theoretical model, particularly in terms of uniform porosity and localized energy deposition.

All fuel samples were first oven-dried at 105 °C for at least 24 h to achieve a moisture content below 2% [56,57]. After natural air exposure, the moisture content stabilized around 5%. For fuel samples, oven-dry weight (ODW) moisture content (MC) was calculated as the weight of water (i.e., the difference between the initial and oven-dry fuel) divided by the oven-dry fuel. To obtain samples with specific target moisture levels, dried materials were uniformly mixed with a calculated amount of water, sealed, and allowed to equilibrate for over 24 h. The actual moisture content was verified prior to testing using a moisture meter (model XU-DHS-10A; XIULAB, Shanghai, China). The porosity of the ground Larix surface fuel was determined to be approximately 58.0% ± 2%, as shown in Figure 6a.

For the smoldering-to-flaming transition experiment, twelve typical forest fuel types were tested, including surface fuels from six representative forest types and six in-stand fuel components from Betula–Larix mixed forests. These samples were also oven-dried at 105 °C for at least 24 h to reach a moisture content below 2%, with air-dried samples equilibrating to about 5% moisture content. To achieve other desired moisture levels, the dried samples were mixed with pre-calculated amounts of water, sealed, and homogenized for over 24 h. Moisture content was re-verified using a moisture meter prior to testing. The porosity values of the twelve forest fuel types are summarized in

Table 1. Measured Porosity of Typical Forest Fuel samples.

Forest Type Sample	Porosity (%)	In-Stand Fuel Component Samples	Porosity (%)
Larix gmelinii forest	93.00	Post-fire site	90.00
Rhododendron–Larix forest	86.00	Larix twigs	96.00
Ledum–Larix forest	70.00	Moss	88.25
Pinus pumila–Larix forest	78.50	Surface litter	86.50
Grass–Larix forest	82.50	Coarse woody debris	72.50
Betula–Larix forest	77.50	Humus layer	78.50

* The surface fuel samples from Larix gmelinii forest, Rhododendron–Larix mixed forest, Ledum–Larix mixed forest, Pinus pumila–Larix mixed forest, grass–Larix mixed forest, and Betula–Larix mixed forest were all collected at the same time from their respective forest types. The post-fire site, Larix twigs, moss, surface litter, coarse woody debris, and humus layer were sampled simultaneously as different in-stand fuel components from a single Betula–Larix mixed forest plot.

.

The porosity of the biomass samples was determined using an equal-height water-filling method. Undisturbed soil–litter block samples (9.5 cm × 9.5 cm × H) were placed in a rectangular plastic container (Container A), and the litter layer was gently fixed with a wire mesh to preserve the original packing structure as much as possible. A second plastic container (Container B), identical in size to Container A, was then filled with water to the same height as the sample in Container A. The corresponding water volume, $V_1$, was taken to represent the total sample volume. The water was then gradually added into Container A until the water level reached the same reference height, and the added volume, $V_2$, was taken as the pore volume of the sample. Porosity ϕ was calculated as

$$\varphi ={\displaystyle \frac{V_1}{V_2}}\times 100\%$$

(19)

where $V_1$ is the total sample volume and $V_2$ is the pore volume.

Before each trial, a subsample (1–5 g) was taken from the prepared material, and its moisture content was measured using a moisture analyzer to ensure that the difference between the actual and target moisture contents was within ±5%.

To initiate ignition, small pieces of fruitwood charcoal were cut to appropriate sizes and used as conductive media to channel the arc discharge into the fuel bed. A variable transformer with an output power of 100 W and an output voltage of approximately 1400 V was used to generate the arc. The positive and negative electrodes were applied to the embedded charcoal to sustain a long-duration discharge, during which the formation of a charred zone and its potential for self-sustained propagation were observed.

Smoldering ignition of the fuel was defined as the formation of a visibly blackened char region that propagated outward in a self-sustained manner for more than 15 min [58]. During the tests, the ambient conditions were maintained at a temperature of approximately 25 ± 2 °C, relative humidity of 80 ± 5%, and atmospheric pressure of 101 ± 1 kPa.

2.6. Data Collection, Processing and Analytical Approach

Experimental data from the simulated lightning ignition phase were systematically collected, cleaned and normalized. Key variables were subsequently binarized into Boolean features to be used as effective inputs for a support vector machine (SVM) classification model. In the smoldering propagation phase, the spread data were represented by sequential order to serve as a proxy for vertical propagation speed to evaluate the relationship between fuel moisture content and vertical smoldering propagation rate. During the smoldering-to-flaming transition phase, two primary indicators were recorded, including (1) whether a transition to flaming combustion occurred and (2) the time at which flaming was first observed.

Figure 6 presents representative experimental cases of successful and failed ignition. Each sequence displays images captured at 0, 1, 6, 11, 16, 21, 26, and 31 min after arc discharge ended. In the successful case, a clearly visible charred zone continues to expand spontaneously, accompanied by a steady increase in temperature. In contrast, the failed ignition case shows a charred zone that ceases to propagate, with temperature remaining nearly constant. Figure 6a shows an example of successful simulated lightning-induced ignition under the experimental conditions (MC = 5%, WS = 5 m/s, DD = 3 min), whereas Figure 6b depicts a failed ignition case under conditions of (MC = 35%, WS = 0 m/s, DD = 3 min). Representative examples of successful and failed ignition under controlled laboratory conditions are shown in Videos S1 and S2 in the Supplementary Materials.

When the simulated lightning current device is activated, visible electrical arcs and sparks are observed between the fuel and the electrode probes. As the arc discharge duration increases, noticeable charring develops on both the surface and interior of the fuel material. During this process, electrical energy is transferred to the fuel in the form of heat through the arc, leading to pyrolysis and carbonization of the material. When the amount of energy transferred is sufficient, the sum of electrical input and exothermic pyrolysis exceeds the combined endothermic pyrolysis demand and heat losses, resulting in a self-sustained smoldering combustion state. Conversely, if the energy delivered is insufficient to maintain this energy balance, the smoldering process is not sustained and eventually extinguishes.

The framework was designed to quantitatively validate the theoretical model developed in Section 2. The controlled laboratory setup allowed systematic variation in discharge energy, fuel moisture, and wind speed under repeatable boundary conditions, simulating the lightning-induced ignition environment in natural forest floors. The defined parameters—arc discharge duration, wind speed, and moisture content—were directly linked to the key variables ($E$, $v$, and MC) in the thermal balance equations. The smoldering-to-flaming transition tests further extended the analysis to post-ignition combustion dynamics.

3. Results

3.1. Regional Background of Lightning-Caused Fires in Daxing’anling

The Daxing’anling region of northeastern China is the country’s largest boreal forest and one of its most important regions for lightning-induced wildfires. Long-term fire records from the national forest fire management authorities (1966–2024) reveal a clear upward trend in both the number and proportion of lightning fires (Figure 7). Lightning now accounts for over 90% of all ignitions in several recent years, marking a clear transition from human-driven to lightning-dominated fire regimes. This shift coincides with regional warming trends exceeding 0.36 °C per decade and a documented increase in summer thunderstorm frequency over the past half-century [44,47,48].

These trends highlight the growing importance of lightning as a natural ignition source in Daxing’anling and provide the regional context for the present study. Combined with the region’s deep organic soil horizons and highly porous duff layers, make Daxing’anling an ideal natural laboratory for examining lightning–fuel interactions under realistic boreal conditions [59,60].

3.2. Threshold of Simulated Lightning-Induced Ignition

Figure 8a illustrates the three-dimensional classification of smoldering ignition outcomes using a support sector machine (SVM) model [61]. The SVM approach was adopted because it is suitable for small-sample nonlinear classification and can be used to identify decision boundaries among multiple influencing factors. The analysis was implemented in Python (version 3.9) in the PyCharm environment (version 2024.2.1) using the scikit-learn (sklearn) library. Fuel moisture content (X-axis), wind speed (Y-axis), and arc discharge duration (Z-axis) were used as input variables to construct a linear decision boundary. Red points represent successful smoldering ignitions observed in experiments, while blue points indicate non-ignition cases. The green semi-transparent plane denotes the linear decision boundary fitted by the SVM. The model effectively separates ignition and non-ignition samples in the three-dimensional feature space. Light red and light blue point clouds represent the spatial regions predicted by the model as “ignition” and “non-ignition” zones, respectively. The performance of the model was evaluated by examining the consistency between the fitted classification boundary and the observed sample distribution in the three-dimensional feature space. The hyperplane equation derived from the SVM model is:

$$-5.002\cdot X+0.498\cdot Y+0.008\cdot Z-0.759=0$$

(20)

where X, Y, and Z correspond to fuel moisture content, wind speed, and arc duration, respectively. The normal vector is $w=[-5.002,0.498,0.008]$, and the intercept is $b=-0.759$. It should be noted that the variables used in Equation (20) were not normalized prior to model fitting. Therefore, the relative magnitudes of the coefficients should be interpreted with caution, as they may be influenced by variable scale and units and do not provide a fully scale-independent measure of variable importance. Equation (20) should therefore be regarded as an empirical relationship derived from the present experimental dataset. It does not explicitly parameterize all potentially influential physical factors, such as porosity, discharge intensity, or intrinsic fuel flammability, which may also affect the observed combustion behavior.

In the present formulation, moisture content appeared to have the strongest influence on the classification boundary; however, because the variables were not normalized, this result should be interpreted as indicative rather than as a scale-independent ranking of variable importance. This finding aligns with existing knowledge of smoldering ignition mechanisms, which higher moisture content greatly increases thermal inertia and suppresses the heating of fuels, thereby reducing the likelihood of smoldering ignition. Wind speed has a secondary, slightly positive effect, while arc duration shows the smallest coefficient, suggesting a limited marginal contribution in the decision model.

Visualization of the decision plane and sample distribution further indicates that increasing moisture content shifts the sample distribution toward the non-ignition region. Successful ignitions (red points) are primarily located in regions where the moisture content is below 25%, whereas non-ignitions (blue points) cluster in areas with higher moisture content. Although arc duration has a lower weight in the model, it may still influence classification near the decision boundary, particularly under low-moisture and moderate wind conditions.

Figure 8b presents the distribution patterns of the experimental data in a two-dimensional space. A heatmap was generated using moisture content and wind speed as input variables, displaying the minimum arc duration required for ignition. Color intensity indicates the required discharge time, while hatched areas represent experimental conditions under which ignition was not achieved. The results show that moisture content substantially delays or inhibits smoldering ignition. When moisture exceeds 35%, ignition is rarely achieved even with prolonged arc discharge. In contrast, at low to moderate moisture levels (10%–20%), ignition occurs more readily and requires shorter discharge durations. Wind speed contributes positively to ignition mainly within the mid-moisture range, enhancing oxygen supply, but its effect is limited under high-moisture conditions.

3.3. Effect of Moisture Content on Smoldering Spread

Figure 9a presents a sequence of typical images showing the vertical smoldering propagation process over a 90-min period across fuel samples with varying moisture contents. The results indicate that fuel moisture content significantly affects the vertical smoldering spread rate. Samples with lower moisture content (5%–15%) achieved complete surface carbonization within 60 min, whereas those with higher moisture content (≥35%) exhibited a clear delay in propagation, requiring an additional 30 min to complete surface carbonization. In some cases, smoldering propagation was inhibited, and vertical spread could not be completed within the experimental period, demonstrating the strong suppressive effect of moisture on vertical smoldering behavior.

Figure 9b shows the statistical relationship between fuel moisture content (MC) and the smoldering spread capacity of the samples. Six moisture levels (5%–55%) were tested, with each condition repeated seven times, yielding a total of 42 complete experimental observations. Image recognition and time-series analysis were used to extract the “sequence of spread” for each sample, which served as a proxy for smoldering propagation capacity. Spearman rank correlation analysis was performed to evaluate the relationship between moisture content and spread sequence. The results revealed a strong positive correlation, with a correlation coefficient of $R=0.879$ and a significance level of $p<0.001$. These findings indicate that as moisture content increases, smoldering ignition is increasingly delayed, and the spread rate continues to decline, confirming the dominant inhibitory role of moisture in the smoldering propagation process.

This finding corroborates the ignition classification results derived from the Support Vector Machine (SVM) model shown in Figure 9, indicating that under the experimental conditions of this study, fuel moisture content not only governs the probability of smoldering ignition but also significantly affects the vertical propagation capability. The underlying physical mechanism is attributed to the increase in specific heat capacity and intensified vaporization effects associated with higher moisture content. These factors lead to greater heat losses during the heating phase and reduced thermal diffusivity, which together inhibit the ability of the smoldering front to sustain itself, thereby suppressing stable vertical propagation.

This result is consistent with findings reported in both research in China and international studies. Previous research has identified a critical moisture threshold—typically between 30% and 35%—above which smoldering spread rates decline significantly or fail to occur. In the present study, the highest moisture group (55%) exhibited similar characteristics, with severely limited smoldering spread, further validating trends reported in the literature.

3.4. Analysis of Influencing Factors in the Smoldering-to-Flaming Transition

Figure 10a illustrates the flaming transition behavior of various forest fuels under ambient temperature conditions (25 °C), across a gradient of wind speed and moisture content. The experimental results clearly demonstrate that wind speed plays a dominant role in the flaming transition of forest fuels. Under a wind speed of 1 m/s, almost all forest fuels successfully transitioned to flaming combustion (indicated by circular markers). In contrast, under no-wind conditions (left panel), only a subset of samples from the Larix gmelinii forest and Rhododendron–Larix mixed forest exhibited flaming, particularly at lower moisture content levels. Notably, none of the six forest floor components from the Betula–Larix mixed forest exhibited flaming transitions under no-wind conditions.

Figure 10b further confirms the significance of wind speed while also highlighting the role of ambient temperature in governing flaming transition behavior. A comparative analysis between the left panels of Figure 10a,b reveals that a high-temperature environment (50 °C) substantially enhanced the ignitability of fuels from the Rhododendron–Larix mixed forest. Specifically, samples with moisture contents between 15% and 35%, which failed to transition to flaming under no-wind, ambient-temperature conditions, successfully ignited when subjected to either higher temperature or the presence of wind. This finding underscores the compensatory effect of temperature, whereby elevated thermal conditions can partially offset the limiting influence of stagnant airflow. Furthermore, the results consistently demonstrate that lower moisture content facilitates earlier and more reliable flaming transitions across different fuel types. The relationships shown in Figure 10 should therefore be interpreted as empirical patterns derived from the present experimental dataset under controlled laboratory conditions, rather than as universally predictive formulations. Representative examples of the smoldering-to-flaming transition are provided in Videos S3 and S4 in the Supplementary Materials.

The time to flaming ignition was recorded for all trials to characterize the smoldering-to-flaming transition. A representative example is shown in Figure 11, which compares the time to flaming (STF) among various forest fuel types and forest floor components under two temperature conditions (25 °C and 50 °C). The overall trend demonstrates that fuel type is a critical determinant of flaming efficiency, with significant differences in ignition performance observed between forest stand fuels and component-based fuels.

Notably, all experimental samples from the Larix gmelinii forest were collected under windless (0 m/s) conditions, yet consistently exhibited rapid and stable flaming transitions at both temperature levels, indicating exceptionally high ignitability and ignition consistency. This finding highlights the inherent flaming potential of this forest type even under unfavorable conditions such as low temperature and stagnant air. The observed performance is likely attributable to the loose structure, good ventilation, and abundance of high-energy, low-inertia materials such as needles and fine twigs, which facilitate rapid pyrolysis and sufficient oxygen diffusion. These properties enable the formation of continuous and sustained flame fronts.

In contrast, samples from other forest types and all forest floor components were primarily derived from experiments conducted under a wind speed of 1 m/s. While the presence of wind clearly improved the flaming success rate for these fuels, it also introduced greater variability in the time to flaming. This is reflected in the boxplots as larger interquartile ranges and more frequent outliers. Such variability suggests that, although wind generally plays a facilitative role in flaming, it may also induce localized thermal or oxygen inconsistencies, especially in component fuels with heterogeneous structures, high moisture content, or high thermal inertia (e.g., surface litter, humus layer).

Furthermore, the high-temperature condition (50 °C) generally reduced the minimum ignition times across all fuel types, with several forest types showing markedly faster flaming transitions compared to ambient temperature conditions. For component fuels that typically ignite more slowly, elevated temperature appeared to act as an ignition trigger. However, it is important to note that the combined effects of high temperature and wind speed were not always additive. Some fuels still exhibited delayed or failed flaming transitions due to factors such as dense structure, insufficient porosity, or excessive moisture retention.

Figure 12 presents heatmaps of the time to flaming under varying combinations of moisture content and ambient temperature, with explicit comparison between windy (1 m/s) and windless (0 m/s) conditions. The overall trend clearly indicates that increasing moisture content significantly delays flaming ignition, with the effect being more pronounced in the absence of wind. Under windless conditions in Figure 12a, the time to flaming sharply increases once moisture content exceeds 25%, highlighting the strong inhibitory role of moisture in the ignition process. Notably, at 45% moisture content, a slight decrease in flaming time was observed. However, this does not contradict the general trend. Further analysis shows that only two highly flammable fuel types—Larix gmelinii forest and Rhododendron–Larix mixed forest—successfully ignited under this condition, suggesting a “survivorship bias” of highly combustible samples that masks the overall suppression effect.

In contrast, under windy conditions in Figure 12b, the time to flaming is consistently shorter across all moisture levels. This is especially evident at moderate to high moisture content levels (35%–45%), where enhanced airflow and oxygen availability significantly improve heat-transfer efficiency and promote volatile release, thereby offsetting the dampening effects of moisture. Under the combined influence of 50 °C temperature and 1 m/s wind speed, rapid flaming ignition was achieved even for high-moisture samples. This strongly supports the existence of a “wind-driven ignition compensation mechanism” under humid conditions.

Additionally, elevated temperature also contributes substantially to reducing flaming time, particularly in the 15%–35% moisture range, where high temperature significantly accelerates ignition. However, this temperature effect appears non-linear—its impact diminishes at both very high and very low moisture levels. This suggests that moderately moist fuels benefit most from thermal enhancement, likely due to reduced energy thresholds for ignition.

It is worth noting that despite controlling for environmental variables, flaming time still exhibits variability across multiple data points. This could be attributed to heterogeneity in local moisture content and pore structure within the fuel matrix during the smoldering phase. Wind flow interacting with these microstructural differences may introduce complex dynamics, resulting in localized randomness in ignition behavior.

Figure 13a illustrates the distribution of Time to Flaming (TTF) across different porosity ranges under two wind conditions: 0 m/s (no wind) and 1 m/s (with wind). The overall trend indicates that both porosity and wind speed are critical determinants of flaming efficiency, and that a significant interaction exists between these two variables.

In the high-porosity range (>0.9), fuel samples exhibited rapid and stable flaming transitions under both wind conditions, with TTF values highly concentrated and consistent. This suggests that fuels with loose structures and good ventilation possess intrinsic advantages in heat accumulation and oxygen availability. Even in the absence of external airflow, such fuels are capable of sustaining sufficient convective heat transfer and gaseous diffusion, enabling efficient pyrolysis and ignition—an indication of strong “self-ignition adaptability.”

In contrast, for the low-porosity group (0.7–0.8), all samples failed to transition to flaming under no-wind conditions. This outcome underscores the presence of substantial ignition barriers in compact fuels, where heat buildup and oxygen transport are significantly constrained. However, when wind speed increased to 1 m/s, the same fuels successfully ignited, with TTF values predominantly ranging between 50 and 350 s. This clearly demonstrates the “structural compensation effect” of airflow in overcoming physical limitations to ignition.

For the intermediate porosity range (0.8–0.9), fuels ignited successfully under wind conditions; however, TTF values displayed greater variability, indicating increased dispersion. This variation may be attributed to microstructural heterogeneities in fuel arrangement or spatially non-uniform enhancement of heat and oxygen delivery due to wind, leading to fluctuating ignition kinetics across the sample surface.

Figure 13b provides a clear visualization of the distribution of Time to Flaming (TTF) under different wind speed conditions. When wind speed increases from 0 m/s to 1 m/s, the median TTF decreases significantly—by more than 60%—indicating that wind notably enhances overall ignition efficiency. This accelerating effect is primarily attributed to improved convective heat transfer and increased oxygen availability, which together expedite the pyrolysis of fuel and the release of volatiles, thereby lowering the energy threshold required for flame initiation.

However, under the 1 m/s wind condition, the TTF distribution becomes more dispersed, as evidenced by a wider interquartile range and a greater number of outliers. This suggests that while wind promotes ignition on average, it may also introduce greater stochasticity and uncertainty into flaming behavior. In fuels with complex or heterogeneous structures, airflow can induce uneven heat accumulation and variable oxygen diffusion at the microscale, resulting in a non-uniform disturbance effect that compromises the consistency of ignition transitions.

Figure 13c presents the statistical distribution of Time to Flaming (TTF) under two ambient temperature conditions (25 °C and 50 °C). The overall trend indicates that elevated temperature significantly accelerates the flaming transition process. Under the 50 °C condition, the median TTF is notably lower than at 25 °C, suggesting that higher temperatures facilitate a faster approach to ignition threshold, thereby enhancing the flaming transition rate of fuels.

However, the TTF distribution under high-temperature conditions also exhibits greater dispersion. Both the interquartile range (box height) and the number of outliers increase significantly at 50 °C. This indicates that although temperature serves as an overall accelerating factor for flaming, its actual effect is strongly modulated by fuel microstructural characteristics—such as porosity and oil content—as well as by heterogeneity in moisture distribution. These factors contribute to greater variability and unpredictability in ignition behavior under high-temperature scenarios.

4. Discussion

The results indicate that different factors dominate different stages of lightning-induced fire development. Discharge energy is most directly related to ignition initiation, because it determines whether enough heat is delivered to the fuel bed. After ignition, fuel moisture becomes the main constraint on whether smoldering can be sustained and spread. Wind, by contrast, shows a stronger effect on the transition from smoldering to flaming combustion, likely through its influence on oxygen supply and heat transfer. These findings suggest that lightning-caused fire is not a single-step process, but a sequence of linked stages controlled by different limiting factors.

This interpretation is broadly consistent with previous studies. Feng et al. showed that ignition under impulse-current conditions depends strongly on electrical input and fuel properties, which agrees with our result that higher discharge energy increases ignition probability [62]. Our study extends this understanding by separating ignition, smoldering propagation, and smoldering-to-flaming transition, and by showing that the relative importance of the controlling factors changes across these stages. The role of prolonged discharge characteristics should also be interpreted with caution. While long-continuing-current or continuing-current components have often been considered important because they extend heat delivery, recent work suggests that their contribution is not uniformly dominant. Schultz et al. found that wildfire-associated flashes are diverse and do not support a simple interpretation in which most lightning-caused fires are attributable to LCC strokes alone [39].

A similar pattern is reflected in the results of Pérez-Invernón et al., who reported that strokes with continuing currents may have greater wildfire potential when conditions for initial fire spread are less favorable [40]. Our experiments point in the same direction. Fuel moisture was the primary constraint on ignition success, but arc duration still influenced classification near the ignition threshold, especially under low-moisture and moderate-wind conditions. This suggests that sustained energy input is not always the dominant factor, but its role becomes more apparent when environmental conditions are close to critical ignition thresholds.

The differences observed among biomass samples are likely associated with variations in their flammability-related properties. Characteristics such as particle size, packing arrangement, porosity, moisture-holding capacity, and fuel continuity can influence heat transfer, oxygen access, and the ability of the fuel bed to sustain oxidation after initial heating. Fuel beds with higher porosity and better ventilation are more likely to support smoldering spread and, in some cases, subsequent transition to flaming. By contrast, densely packed materials or fuels that retain moisture more effectively may inhibit these processes by restricting oxygen penetration or increasing the energy required for drying. Ignition success and subsequent combustion development therefore depend not only on discharge characteristics and ambient conditions, but also on the physical receptivity of the biomass itself.

The thresholds reported here are specific to the fuel types and laboratory conditions examined in this study, and should not be transferred directly to other biomass samples or regions. Even so, the results may still be useful for hazard characterization elsewhere, not because the numerical values themselves are universal, but because the underlying analytical framework can be applied more broadly. If local fuel properties—such as porosity, moisture retention, and structural continuity—are considered together with regional lightning and weather conditions, the same framework may help indicate whether another biomass type is more prone to ignition, sustained smoldering, or subsequent transition to flaming.

Several limitations should be acknowledged. The experiments were conducted under controlled laboratory conditions using prepared surface fuel beds, which simplify the heterogeneity of real forest floors. Natural fuel beds vary more in composition, continuity, organic-layer depth, and moisture distribution, and are influenced by short-term weather fluctuations that cannot be reproduced here. The present framework also does not explicitly include the underlying soil and its physical properties, although soil texture, soil moisture, and mineral composition may influence fuel condition, near-surface heat transfer, and the dissipation of lightning-related energy. In addition, the discharge simulator was designed to represent sustained thermal input rather than the full temporal and electrical complexity of natural lightning. The imposed discharge duration should therefore be regarded as an experimental approximation of ignition-relevant energy delivery, not as a direct replication of natural lightning. The empirical formulation in Equation (20) does not explicitly include parameters such as porosity, discharge intensity, or intrinsic fuel flammability, although these factors may also influence the observed response. Moreover, because the variables in Equation (20) were not normalized prior to model fitting, the coefficient magnitudes are not strictly comparable across variables with different units and scales. Accordingly, the results are best interpreted as mechanistic evidence for ignition and early combustion development, rather than as field thresholds directly transferable to all forest conditions. Future research should extend this framework to a wider range of fuel types and field conditions, and should further examine the roles of soil properties, natural weather variability, and more realistic lightning-energy transfer processes. Such efforts would help bridge the gap between controlled laboratory results and operational forest fire danger prediction.

5. Conclusions

This study combined laboratory experiments and a simplified heat-balance framework to investigate the effects of surface environmental factors on lightning-related ignition and early combustion development in forest fuels. The main conclusions are as follows.

(1) Different factors dominated different combustion stages. Discharge energy was most closely related to ignition initiation, fuel moisture controlled smoldering persistence and spread, and wind had a stronger influence on the transition from smoldering to flaming combustion.

(2) Low fuel moisture, higher ambient temperature, and moderate wind conditions favored ignition and increased the likelihood of flaming transition. Among these factors, fuel moisture acted as the main limiting factor, whereas wind and temperature mainly influenced post-ignition combustion development.

(3) The experimental results and heat-balance analysis together clarify how environmental conditions affect the conversion of electrical input into sustained combustion. However, the derived relationships should be regarded as empirical and specific to the tested fuels and laboratory conditions.

(4) Although the reported thresholds are not directly transferable to all forest environments, the analytical framework may still be useful for evaluating ignition hazard in other fuels and regions. Further validation under natural fuel, soil, and weather conditions is needed before direct application to field-scale fire danger assessment or early warning practice.