Characterizing Forest Fuel Properties and Potential Wildfire Dynamics in Xiuwu, Henan, China

Abstract

As global climate change and human activities increasingly influence our world, forest fires have become more frequent, inflicting significant damage to ecosystems. This study conducted measurements of combustible materials (moisture content ratio, ignition point, and calorific value) across 14 representative sites. We employed Pearson correlation analysis to ascertain the significant differences in combustible properties and utilized entropy methods to evaluate the fire resistance of materials at these sites. Cluster analysis led to the development of four combustible models. Using BehavePlus software, we simulated their fire behaviors and investigated the effects of wind speed and slope on these behaviors through sensitivity analysis. The results revealed notable differences in the moisture content ratios among different types of combustibles, especially in sites 2, 3, 8, 9, and 13, indicating higher fire risks. It was also found that while humus has a higher ignition point and lower calorific value, making it less prone to ignite, the resultant fires could be highly damaging. The Pearson analysis underscored significant variations in the moisture content ratios among different combustibles, while the differences in ignition points and calorific values were not significant. Sites 5 and 6 demonstrated stronger fire resistance. The simulations indicated that fire-spread speed, fireline intensity, and flame length correlate with, and increase with, wind speed and slope. Sensitivity analysis confirmed the significant influence of these two environmental factors on fire behavior. This study provides critical insights into forest fire behavior, enhancing the capability to predict and manage forest fires. Our findings offer theoretical support for forest fire prediction and a scientific basis for fire management decision-making.

1. Introduction

Forests are an important part of the global ecosystem, not only as a habitat for wildlife but also in regulating the climate, purifying the air, nourishing water, maintaining biodiversity, and maintaining ecological security However, globally, forest fires occur frequently, causing significant ecological damage and economic losses. These fires not only destroy valuable forest resources and threaten wildlife habitats but also pose serious challenges to the security and development of human societies. According to statistics, millions of forest fires occur globally every year, burning tens of millions of hectares of forests and causing significant ecological and economic losses [1]. For example, from 1950 to 2018, more than 800,000 fires occurred in China, affecting a forest area of more than 3815 × 10⁴ hm, with more than 30,000 casualties caused by the fires [2]. Data from the China Bureau of Statistics show that from 2004 to 2021, there were more than 100,000 fires in China, and the affected forest area reached 95.8 × 10⁴ hm² (e.g., Figure 1a). The total fire area reached 224 × 10⁴ hm², and the discounted loss due to forest fires reached about 23.64 × 10⁴ Yuan (e.g., Figure 1b). These data not only reflect the severity of forest fires but also highlight the urgent need for effective prediction and control of forest wildfires.

In order to more effectively predict and manage forest fires, an in-depth understanding of the properties of forest fuels is critical. Combustibles are the material basis and important agents of fire occurrence [3]. The properties of combustibles, such as their dry-to-fresh ratio, ignition point, calorific value, combustible load, surface area-to-volume ratio, fuel bed depth, etc., are critical to understanding combustibles, predicting and managing forest fires [4].

Scholars at home and abroad have conducted extensive research on the characteristics and fire behavior of combustibles and have developed a variety of forest fire behavior models [5,6,7,8,9,10]. In the 1930s, U.S. forest fire professionals initiated pioneering investigations into combustible typologies, stratifying them based on fire propagation velocities and suppression complexities. However, it was not until the 1990s that American wildfire researchers unveiled the FARSITE model [11], a progressive forest fire behavior paradigm grounded in the foundational Rothermel model. This novel approach delved deeper into the stratified anatomy of forest combustibles and postulated innovative taxonomies for combustible classification, thereby enhancing the pre-existing corpus [12]. Numerous scholars have since architectured wildfire spread models rooted in the Rothermel framework. Notably, Andrews pioneered the BehavePlus system, an offshoot of the Rothermel model, furnishing realistic simulations of wildfire dynamics, which subsequently became instrumental for U.S. forest fire stewardship [13]. Meanwhile, Hahn et al. harnessed the capabilities of the BehavePlus5.0 software, fine-tuning six determinants of fire dynamics, including fuel load and fuel bed depth. Their empirical findings accentuated the prolonged ramifications of varied logging intensities on fire behaviors [14].

China’s inquiry into forest combustible materials commenced in the 1980s. Though this initiation was somewhat belated compared to global counterparts, the ensuing advancements have been commendable. Zheng et al. pioneered a taxonomy of forest combustibility in Northeast China, focusing on influential parameters such as the fuel load [15]. Subsequently, Yuan et al. presented a comprehensive review encompassing the classification of forest combustibles, modeling approaches, and the prevailing trajectory of predictive studies. Their work offered a quantitative exploration of the combustion properties of forest materials while also highlighting pertinent challenges and evolving research trends in the Chinese context [16]. Tian et al. harnessed urban land use maps in conjunction with Landsat TM imagery, employing a supervised classification technique to categorize combustibles. Their detailed analysis elucidated the distinct attributes of various combustible types. Their findings suggested a six-fold classification for combustibles in terms of forest fire risk prediction. However, in the realm of coniferous forests, a nuanced classification was underscored, emphasizing the combustibility, vertical stratification, and spatial distribution of materials [17]. In a distinct study, Wang et al. scrutinized fire dynamics across six pivotal forest typologies within Kunming’s Xishan National Forest Park, utilizing the BehavePlus modeling framework. Their empirical evidence underscored that under inclinations of 25° and 35°, the Pinus yunnanensis forest manifested the apex metrics for heat per unit area, fireline intensity, and flame magnitude. In contrast, the most subdued indices were recorded in the Schima wallichii (Dry winter melon) forest [18]. Moreover, Wu et al. integrated cluster analysis to discern combustible categories in southern Jiangxi and employed BehavePlus for fire dynamics forecasting. Their research culminated in the creation of four canonical combustible archetypes, with the ensuing projections bolstering forest fire management strategies [19].

While researchers have made certain achievements in classifying and modeling forest combustibles, most of these studies and models are confined to specific geographical areas. Their primary focus on the combustible types of particular regions limits the scope and accuracy of their application. Furthermore, research on forest fire issues at the county level, both domestically and internationally, is relatively scarce. Xiuwu County, with its rich natural resources and diverse vegetation types, possesses 32,332 hectares of forestry land, a forest coverage rate of 26.35%, and a tree coverage rate of 50.23%. Therefore, studying the characteristics and fire behaviors of combustibles in Xiuwu County is essential.

This paper selects representative vegetation sites in Xiuwu County and employs typical sampling methods for field investigations. In establishing the BehavePlus model, we consider not only the basic factors such as the load, moisture content, ignition point, and calorific value of combustibles but also integrate the impact of the slope and wind speed on combustible fire behavior. This comprehensive approach offers a broader perspective and more precise recommendations for fire risk management and improvement, holding significant practical importance for understanding and preventing forest fires.

2. Models and Methods

2.1. Regional Overview

2.1.1. Overview of the Study Area

Xiuwu County is located in the northwest part of Henan Province, with a total area of 611 km². The geographical coordinates of Xiuwu County range from approximately 35°07′39″ to 35°28′32″ north latitude and 113°08′17″ to 113°32′03″ east longitude. The annual average temperature is 14.5 °C, and the average annual rainfall is 560.4 mm. The region experiences a warm, temperate continental monsoon climate with four distinct seasons. Summers are hot and rainy, while winters are cold and dry.

Xiuwu County has a total area of 632.9 hm² and is primarily divided into Chengguan Town, Qixian Town, Xunfeng Town, Yuntai Mountain Town, Zhouzhuang Town, Xicun Township, Wangtun Township, Wuliyuan Township, and the Xiaoying industrial and commercial area. The total forest area in Xiuwu County is 3 × 10⁴ hm², as shown in Figure 2a,b. The forest coverage rate is 26.35%, with a total of 16,000 hectares of tree woodland, 38 hectares of sparse woodland, and 14,500 hectares of shrubland. The remaining 1734 hectares consist of non-afforested land, nursery land, land without standing trees, land suitable for afforestation, and land for forestry auxiliary production. The proportions of forested areas in Chengguan Town, Qixian Town, Xunfeng Town, Yuntai Mountain Town, Zhouzhuang Town, Wangtun Township, Wuli Source Township, Xicun Township, the Xiaoying industrial and commercial zone, and other regions are as follows: 0.36%, 10.3%, 3.57%, 23.69%, 0.45%, 0.82%, 2.65%, 50.45%, 0.11%, and 7.6%, respectively. Among these regions, Xicun Township, Yuntai Mountain Town, and Qixian Town have the largest proportion of forested areas.

2.1.2. Overview of Sampling Sites

The 14 sample plots surveyed are 6 arbor forest plots and 8 shrub forest plots (as shown in

Table 1. Survey methodology for arbor and shrub forest plots.

Plot Number	Dominant Tree Species	Age Group	Canopy Closure	Average Tree Height (m)	Average DBH (cm) (P1–P6)/Average Ground Diameter (cm) (P7–P14)	Soil Type	Altitude (m)	Slope-Aspect-Slope
P1	Quercus	Young forest	0.9	10.75	20.5	brown soil	468	Fast-East-Downhill
P2	Quercus	Young forest	0.58	3.86	8	yellow brown soil	1124	Slow-Southeast-Uphill
P3	Paulownia	Young forest	0.5	12.3	24.7	yellow soil	0	Flat-no aspect-flat
P4	Paulownia	Young forest	0.55	8	12.5	yellow soil	563	Urgent-Southeast-Downhill
P5	Poplar	Young forest	0.55	17.68	17.5	yellow cinnamon	88	Flat-no aspect-flat
P6	Poplar	Young forest	0.55	9.4	16.3	yellow soil	81	Flat-no aspect-flat
P7	Vitex	Middle-aged forest	0.59	2.1	1.7	yellow soil	452	Slow-Northeast-Downhill
P8	Vitex	Young forest	0.75	2.2	1.13	yellow cinnamon	468	Steep-Southwest Downhill
P9	Vitex	Young forest	0.55	0.8	0.4	yellow brown soil	280	Oblique-Southeast-Uphill
P10	Sumac	Young forest	0.58	1.6	1.43	Sandy soil	505	Flat-northwest-uphill
P11	Sumac	Young forest	0.58	2	1.73	Sandy soil	426	Urgent-Northwest-Uphill
P12	Sumac	Middle-aged forest	0.85	1.7	1.23	yellow cinnamon	806	Slow-Southeast-Uphill
P13	Vitex	Young forest	0.5	2.35	1.8	yellow cinnamon	403	Slow-Northeast-Uphill
P14	Sumac	Young forest	0.5	0.7	0.67	Sandy soil	375	Slow-Northeast-Uphill

), The dominant tree species, age group, canopy density, stand density, average tree height, average diameter at breast height, average ground diameter, soil type, altitude, slope, aspect, and slope position of the plot were investigated.

2.2. Research Framework

As shown in Figure 3, Beginning with the characteristics of forest combustibles and their fire behaviors, this study has clearly defined its objectives and analytical questions. By assessing the properties of combustibles and their fire behaviors, we aimed to develop a model capable of predicting the occurrence of fires and proposed effective risk management strategies. The section on combustible characteristics focuses on analyzing the fundamental physical properties of combustibles, including the dry-to-fresh ratio, ignition point, and calorific value. This part employed Pearson analysis to identify the correlations among these physical properties and utilized the concept of entropy from physics to measure the fire resistance of the samples and their significant differences. In the aspect of combustible fire behavior, our focus was on simulating actual fire behaviors, considering factors such as combustible material load, fuel bed thickness, wind speed, and slope. Cluster analysis techniques were used to identify the key factors influencing fire behavior, and BehavePlus 5.0 software was employed for the model simulations to accurately predict combustible fire behaviors under various conditions. The entire research framework was dedicated to uncovering the dynamic relationship between combustible characteristics and fire behavior, providing a scientific basis to assist Xiuwu County in formulating more effective fire prevention measures. The anticipated results of the study promise to enhance the understanding of fire behavior and offer empirical support for decision-makers in the fields of fire risk management and forest conservation.

2.3. Research Methods

2.3.1. Sample Plot Setup and Sampling Method

In this study, based on the National Forest and Grassland Fire Census Implementation Plan, 14 representative forest sample plots in Xiwu County, Henan Province, were selected for sampling in December 2020, including six arboreal forests and eight shrub forests. As shown in Figure 4a, using GPS with the carrier phase difference technique (RTK) technology, we accurately located the coordinates of the southwest corner of the arboreal forest sample plots and, using this as a reference, we planned a square arboreal forest standard plot with a side length of 25.82 m and an area of 0.067 hm² clockwise. Ensuring that the sample plot perimeter closure difference met the 1/200 accuracy standard, we set up four 2 m² shrub layer sample squares along the boundary at 3 m from either side of the apex of the arborvitae standard plot for investigating the combustible load of the shrub layer. In addition, within each shrub layer sample plot, four additional 1 m² deadfall (herbaceous, humus) sample plots were divided separately to assess the herbaceous, deadfall, and humus loadings. For the shrubland sample plot, a similar method was used to locate the southwest corner coordinates, as in Figure 4b, and a shrubland sample plot with a side length of 5 m and an area of 25 m² was established. The center point was located at 50 m intervals along the northeast and southwest axes, and a shrubland sample plot was set up to investigate combustible loads over a wider area. (If the intervals resulted in samples extending beyond the shrubland boundaries, they were adjusted appropriately). Samples were numbered 1 to 3 in a southwest-to-northeast direction). In the southwest corner of each shrubland sample, a 1 m² deadfall (herbaceous, humus) sample was set up for further combustible loading surveys. This methodology precisely defined the sample plot boundaries and planned in detail the sub-sample plots for combustible load measurements, providing a clear and replicable sampling framework suitable for conducting forest fire risk assessment studies of international caliber.

• Collection of Type I (shrub) samples: For each of the four shrub plots, three standardized shrubs were selected. From each shrub, a 10–20% mixture of stems, branches, and leaves of equal mass proportions was obtained. Within a standard plot, samples from the four plots were mixed separately according to the shrub species. The mixed sample should not be less than 500 g.
• Collection of Type Ⅱ (herbaceous) samples: all parts of herbaceous plants of all samples in the standard site were fully mixed, and about 300 g of fresh herbaceous plants were taken after mixing.
• Sample sampling of Ⅲ (humus): the harvesting method [20] was used to harvest the humus within each sample plot, and non-humus such as gravel, dirt, and obvious roots in the humus were removed, after which the samples were mixed to take about 300 g.
• Sample sampling for IV, V, VI (litter): We investigated the thickness of the litter, collected all litter in the quadrat, and divided it into 1 h (twigs with d < 0.64 cm), 10 h (0.64 cm ≤ d < 2.54 cm), 100 h (twigs with 2.54 cm ≤ d < 7.62 cm), and another three types. The whole harvest mixed sampling was carried out, and each type took about 300 g. Taking the sample back to the laboratory, we determined its dry-to-fresh ratio, ignition point, and calorific value.

2.3.2. Determination and Calculation Methods for Samples

• Determination of the Dry-to-Fresh Ratio

The fresh weight of the samples was measured, and then the samples were placed in an 80 °C drying oven until their mass no longer changed. The dry mass of the samples was recorded. The formula for calculating the dry-to-fresh ratio is as follows (Equation (1)):

$$\mathrm{D}\mathrm{Q}\mathrm{R}=\raisebox{1ex}{$m_d$}\!\left/ \!\raisebox{-1ex}{$m_f$}\right.\times 100\mathrm{\%}$$

(1)

where DQR (dry-to-fresh quality ratio) is the dry-to-fresh ratio (%), $m_d$ is the final dry weight of the sample, and $m_f$ is the fresh weight of the sample.

2.

Ignition Point and Calorific Value

The ignition point and calorific value were categorized according to the dominant tree species (group) in the tree standard site and the dominant tree species (group) in the shrub standard site, and the sample plots were taken in compartments proportional to the area. In each sampled plot, a mixed sample of litter 1 and litter 2 was obtained by combining all litter samples. Similarly, a sample of humus was extracted using the same method. Each sample weighed around 200 g. The combustible material to be tested was taken, removing surface dust, and placed in a cool and ventilated area of the laboratory or dried at 60–80 °C using air- or oven-drying methods. The combustible material sample was pulverized using a plant grinder and passed through a 40-mesh sieve. The ignition point of the combustible material was determined using an ignition point temperature tester (ZXDW-02), with 3–5 measurements conducted for each sample. The calorific value of the combustible material sample was measured using a calorimeter (KY-LR-9000B type), with 3–5 measurements conducted for each sample.

3.

The combustible material load refers to the absolute dry mass of all combustible materials per unit area. The formula for calculating the combustible material load is as follows (Equation (2)):

$$\mathrm{C}\mathrm{o}\mathrm{m}\mathrm{b}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{b}\mathrm{l}\mathrm{e} \mathrm{M}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{a}\mathrm{l} \mathrm{L}\mathrm{o}\mathrm{a}\mathrm{d} (\mathrm{t}/{\mathrm{h}\mathrm{m}}^2)=\left(\frac{\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{d}\mathrm{r}\mathrm{y} \mathrm{w}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t} \mathrm{o}\mathrm{f} \mathrm{t}\mathrm{h}\mathrm{e} \mathrm{s}\mathrm{u}\mathrm{b}\mathrm{p}\mathrm{l}\mathrm{o}\mathrm{t}}{(\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r} \mathrm{o}\mathrm{f} \mathrm{s}\mathrm{u}\mathrm{b}\mathrm{p}\mathrm{l}\mathrm{o}\mathrm{t}\mathrm{s}\times \mathrm{s}\mathrm{u}\mathrm{b}\mathrm{p}\mathrm{l}\mathrm{o}\mathrm{t} \mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a})}\right)\times \frac{\mathrm{10,000}}{1000}$$

(2)

In the formula, the total dry weight of the subplot is measured in kilograms (Kg).

4.

The fuel bed thickness

The combustible bed thickness is the average depth of the combustible material at the surface and has an important influence on fire behavior. At the site of the sample plot survey, the maximum thickness of leaves or stems of herbaceous plants from different sites was measured using a tape measure, and the maximum thickness of herbaceous leaves or stems in the four combustible models was obtained to be 0.657 m, 0.386 m, 0.43 m, and 0.086 m, respectively. It is generally accepted that the thickness of combustible beds of herbaceous plants is about 70 percent of the thickness of the thickest part of their leaves or stems [21]. Therefore, the thickness of the four combustible beds in this study was set at 0.46 m, 0.27 m, 0.3 m, and 0.06 m.

5.

Forest fire resistance:

We used the entropy weight method to rank the fire resistance of the forest stands based on the objective information reflected by the physicochemical indicators of understory fuels [22]. Among the three physicochemical indicators (dry-to-fresh quality ratio, ignition point, and heat value), the greater the information-carrying capacity of a particular indicator on fire resistance, the smaller its entropy value, indicating a greater influence of that indicator on the fire resistance ranking of the forest stands [23]. The calculation process for the entropy weight method is as follows: First, the raw data for the three indicators were standardized. The ignition point has a positive relationship with fire resistance, while the dry-to-fresh ratio and calorific value have a negative relationship. The extreme value method was used for data standardization. The standardization formulas are as follows (Equations (3) and (4)):

$${\mathrm{y}}_{\mathrm{i}\mathrm{j}}=\frac{{\mathrm{x}}_{\mathrm{i}\mathrm{j}}-{\mathrm{x}}_{\mathrm{j}\mathrm{m}\mathrm{i}\mathrm{n}}}{{\mathrm{x}}_{\mathrm{j}\mathrm{m}\mathrm{a}\mathrm{x}}-{\mathrm{x}}_{\mathrm{j}\mathrm{m}\mathrm{i}\mathrm{n}}}$$

(3)

(positive relationship between fire resistance and physical and chemical indicators).

$${\mathrm{y}}_{\mathrm{i}\mathrm{j}}=\frac{{\mathrm{x}}_{\mathrm{j}\mathrm{m}\mathrm{a}\mathrm{x}}-{\mathrm{x}}_{\mathrm{x}\mathrm{j}}}{{\mathrm{x}}_{\mathrm{j}\mathrm{m}\mathrm{a}\mathrm{x}}-{\mathrm{x}}_{\mathrm{j}\mathrm{m}\mathrm{i}\mathrm{n}}}$$

(4)

(inverse relationship between fire resistance and physical and chemical indicators).

In the formulas, ${\mathrm{y}}_{\mathrm{i}\mathrm{j}}$ is the standardized data, ${\mathrm{x}}_{\mathrm{i}\mathrm{j}}$ is the indicator value of combustible materials in the jth category of the ith stand, ${\mathrm{x}}_{\mathrm{j}}$ max is the maximum value of the jth indicator, and ${\mathrm{x}}_{\mathrm{j}}$ min is the minimum value of the jth indicator.

Second, the information entropy of the indicators is calculated (Equations (5)–(7)):

$$E_j=-k{\sum }_{i=1}^n{P}_{ij}{lnp}_{ij} (i=1,2,···,n;j=1,2,···,m)$$

(5)

$$\mathrm{k}=1/\ln \mathrm{n}$$

(6)

$$P_{ij}=Y_{ij}/{\sum }_{i=1}^n{Y}_{ij}$$

(7)

In the formulas, $Y_{ij}$ is the standardized data, $E_j$ is the entropy value of the jth indicator, $P_{ij}$ is the proportion of the jth indicator value of the ith stand, k is the information entropy coefficient, n is the number of stand types, and m is the number of indicators.

Third, the weights of the indicators are calculated (Equation (8)):

$${\mathrm{W}}_{\mathrm{j}}=\frac{1-E_j}{\sum E_j}$$

(8)

In the formula, ${\mathrm{W}}_{\mathrm{j}}$ is the weight of the jth indicator, and m is the number of indicators.

Finally, based on the weights of the indicators, the comprehensive scores of the stand-fire resistance are calculated using a comprehensive evaluation method (Equation (9)):

$$Z_i={\sum }_{j=1}^n{\mathrm{W}}_{\mathrm{j}}{\mathrm{Y}}_{\mathrm{i}\mathrm{j}}$$

(9)

In the formula, $Z_i$ is the final score of the ith forest type, ${\mathrm{W}}_{\mathrm{j}}$ is the weight of each indicator, and ${\mathrm{Y}}_{\mathrm{i}\mathrm{j}}$ is the normalized value of the indicator.

2.3.3. Classification Method for Combustible Material Types

Combustible types were categorized by hierarchical cluster analysis. The hierarchical cluster analysis used Euclidean distance as a similarity measure and clustering based on Ward’s method [24,25] to standardize the variables of stand density, mean diameter at breast height, mean tree height, tree load, and litter load, and the mean tree height, tree load, and litter load were standardized using the standardized Kruskal–Wallis, and then tested for differences between the variables using the Kruskal–Wallis method [26]. After classifying several combustible types through systematic cluster analysis, the combustible parameters of all the samples within the same type were averaged and assigned to the corresponding combustible type, resulting in a number of combustible types and their attributes, which are represented by a clustered spectrogram (commonly known as a dendrogram).

2.3.4. Simulation of Potential Fire Behavior

BehavePlus is an advanced fire behavior simulation software based on a personal computer program system. It integrates models describing the environment and utilizes Rothermel’s fire spread model as its core. The software combines multiple models and algorithms, allowing users to input various parameters related to fuel, weather, and terrain. These parameters include but are not limited to fuel moisture, wind speed, slope, and fuel type. BehavePlus is capable of predicting a range of fire behavior indicators, such as the rate of fire spread, fireline intensity, flame length, and heat release [27,28,29,30].

This article selects the BehavePlus software for the model establishment primarily due to its flexibility and the abundance of fuel models it offers. The software can utilize varying terrain and slope data and allows for the input of specific weather conditions, such as temperature, humidity, and wind speed. Through simulating different scenarios, it can aid local managers in assessing wildfire risks. A significant volume of international research [31,32,33,34,35,36] has employed BehavePlus, demonstrating its efficacy as a powerful and applicable tool. It is particularly suitable for studying the characteristics of forest fuels and wildfire dynamics in Xiuwu County.

The input parameters for the fire behavior model, including slope and meteorological conditions, were based on historical data from Xiuwu County. Considering the similarity of herbaceous plant attributes, default values were used. The default value for the moisture of herbaceous and shrub fuels in BehavePlus is 30%. The surface area-to-volume ratio for grass and shrub fuels follows the default values of the standard model, with a ratio of 6562 m²/m³ for 1 h lag fuels, 5906 m²/m³ for live herbaceous fuels, 5906 m²/m³ for live woody fuels in arbor forest models, and 5249 m²/m³ for live woody fuels in shrubland models. Fuel moisture refers to a set of fuel moisture conditions specific to each fuel type [37]. The fuel moisture series includes 1 h lag fuel moisture, 10 h lag fuel moisture, 100 h lag fuel moisture, live herbaceous fuel moisture, and live shrub fuel moisture (

Table 2. Scenario values of low, medium, and high combustible humidity (%).

Combustibles	Combustible Humidity Scenario
	Low	Middle	High
1 h time lag combustibles	3	6	9
10 h time lag combustibles	4	7	10
100 h time lag combustibles	5	8	11
Live herbal combustibles	30	60	90
Living wood combustibles	60	90	120

). Forest fires in Xiuwu County mostly occurred under low-humidity conditions. Considering the historical humidity conditions during fire events, this study simulated the models of sumac and paulownia under low-humidity scenarios. For the poplar-oak and vitex-sumac models, low-humidity scenarios were selected for 1 h lag, 10 h lag, and 100 h lag fuels, while live herbaceous and live woody fuels were simulated under moderate humidity scenarios.

Currently, the proposed Rothermel model is widely applied and was established by Rothermel in 1972 [38] based on the research in Fentesson (Equation (10))

$$R=I_R\xi (1+{\varphi }_w+{\varphi }_s)/\left(P_b{\epsilon }_0{Q}_{ig}\right),$$

(10)

$R$ represents the rate of fire spread, $I_R$ is the fire reaction intensity; $\xi$ is a coefficient related to heat transfer; ${\varphi }_w$ and ${\varphi }_s$ are coefficients representing the influences of wind and the terrain slope, respectively; the gas density after the drying of the combustible composite is denoted as $P_b$; ${\epsilon }_0$ is the heating coefficient related to bulk density, and the heat required for ignition per unit mass of combustible material is represented as $Q_{ig}$.

Using the formula proposed by the American physicist Byram in 1959 to calculate the fireline intensity [39] (Equation (11)):

$$I=1.667HWR,$$

(11)

I represents the fireline intensity, measured in kJ/m/s; $H$ denotes the calorific value of the combustible material, expressed in kJ/g; $W$ represents the effective combustible load, measured in t/hm⁻²; and $R$ represents the rate of wildfire spread.

Byram proposed a formula for calculating the flame length based on the fireline intensity, converted to international units as follows [40] (Equation (12)):

$$L_f=0.237I^{0.46},$$

(12)

$I$ represents the fireline intensity (kJ/m/s), and $L_f$ represents the flame length (m).

2.3.5. Data Processing and Analysis

Data were organized and analyzed using excel2021 and spss27.0, ANOVA was performed using Duncan′s method for multiple comparisons [41], and the interclass mean chain-linking method was used for the systematic clustering of combustible dry-to-fresh ratios, ignition points, and calorific values. Three physicochemical indexes in different stand types were used to preliminarily evaluate the fire resistance of the stands, and the Pearson correlation coefficient method was used for correlation analyses to simulate fire behaviors of different stand types. The Pearson correlation coefficient method was used to correlate the dry freshness ratio, ignition point, and calorific value of different forest stand types, and by using BehavePlus software, we established a model to simulate the fire behavior of different forest stand types and fire behavior grading. We used Matlab to conduct sensitivity analysis on the fire behavior model. Origin2021 was employed for plotting.

3. Results and Analysis

3.1. Characteristics of Combustibles

3.1.1. Dry-to-Fresh Ratio Characteristics of Combustible Materials in Plots

In the forest research plots, deciduous woodland plots are designated as P1 to P6, while shrubland plots range from P7 to P14. As depicted in Figure 5a,b, the dry-to-fresh weight ratios of combustibles are as follows:

• For combustible Type I (shrubs), the ratio ranges from 53% to 92%.
• For Type II (herbaceous), it varies between 26% and 80%.
• For Type III (humus), the ratio lies between 52% and 94%.
• Type IV (litter with diameter < 0.6 cm) exhibits a ratio ranging from 50% to 94%.

Combustible I (shrubs) had a maximum dry-to-fresh ratio of 92 percent for sample site 9 and a minimum of 53 percent for sample site 6. Combustible II (herbaceous) had a maximum dry-to-fresh ratio of 80% for sample site 13 and a minimum of 26% for sample site 6, respectively. Combustible III (humus) has a maximum dry-to-fresh ratio of 94% for sample site 8 and a minimum of 52% for sample site 2, respectively, and combustible IV (litter 1) had a maximum dry-to-fresh ratio of 94% for sample site 2 and a minimum of 50% for sample site 11. Combustible V (litter 2) had a maximum dry-to-fresh ratio of 90% and 97% for sample site 3. The only combustible material V (Deadwood 2) was seen in sample sites 3 and 9, with dry-to-fresh ratios of 90% and 97%, respectively.

The higher the dry–fresh ratio, the lower the moisture content of the combustible material, and, in this case, combustible material Ⅰ in sample plot 9, combustible material Ⅱ in sample plot 13, combustible material Ⅲ in sample plot 8, combustible material Ⅳ in sample plot 2, and combustible material Ⅴ in sample plots 3 and 9 are more likely to be ignited, prone to fires, and, the fires are also more likely to spread after fire occurs, becoming more violent and endangering forest resources.

Combustibles Ⅰ and Ⅱ in sample plot No. 6, combustible Ⅲ in sample plot No. 2, and combustible Ⅳ in sample plot No. 11 have the smallest dry-to-fresh ratios and the highest moisture contents; they are more difficult to ignite, and the fire after starting a fire is smaller, where the fire spreads slower, and effective prevention and control can be carried out in a short period of time. Therefore, attention should be paid to the daily cleaning of combustibles in the areas of sample plots 2, 3, 8, 9, and 13 to prevent the accumulation of combustibles from starting fires.

3.1.2. Ignition Point and Calorific Value Characteristics of Combustibles

The ignition points and calorific values of humus (III) and litter IV–V (including a mixture of IV samples with diameters less than 0.6 cm and V samples with diameters between 0.6 cm and 2.5 cm) are shown in Figure 5c,d. The range of ignition points is between 265 °C and 288.5 °C and 245.5 °C and 277.5 °C, respectively. In combustible material III the highest ignition point of combustible material III was found in sample plot 8, and the lowest ignition point was found in sample plot 10. In the deadfall (sample IV–V mixture) layer, sample site 12 had the highest ignition point, while sample sites 9 and 3 had the lowest ignition points. Combustible III in sample site 8 and the combustible IV–V mixture in sample site 12 require higher temperatures for ignition, which can delay the onset of a fire, and the ambient temperature alone is not usually sufficient to start a fire. Be aware that combustible materials in sample sites 8 and 12 were ignited, which may produce higher flame temperatures; thus, the fire could spread faster and start a large fire.

Combustible III in sample site 10, and the combustible IV–V mixture in sample site 3 and sample site 9 have a low ignition point, which may lead to a fire at very low temperatures or in the presence of a small ignition source and the fire will spread in a short period of time.

The calorific values in the 14 sample plots ranged from 2.28 kJ/g to 14.33 kJ/g and 13.23 kJ/g to 18.98 kJ/g, respectively. In the combustible material (III) layer, the highest calorific value was found in sample site 10, while the lowest was found in sample site 8. In the mixed layer of deadfall (mixed samples IV–V), broadleaf forest sample site No. 4 had the highest calorific value, while shrubland sample site No. 14 had the lowest calorific value. The calorific value of combustible material III in sample plot 10 and combustible material IV–V in sample plot 4 have the highest calorific value, which indicates that the combustible materials in these sample plots will release more heat during combustion and also produce more carbon dioxide and other pollutants, which may develop into a big fire in a short time during the process of starting a fire. The calorific value of combustible material III in sample plot 8 and the combustible material IV–V mixture in sample plot 14 have smaller calorific values, which indicates that the heat released during the combustion process is small, not easy to produce a large fire, and easier to prevent and extinguish. Although combustible III and combustible IV–V with higher ignition points are not easily ignited, once ignited, they will produce more powerful fires and therefore need to be handled with care regarding fire prevention and fire management, and effective fire prevention and control strategies need to be developed.

As discerned from Figure 5f, there is a negative correlation between the ignition temperature and calorific value, suggesting that as the ignition point increases, the calorific value decreases. An analysis based on calorific values and ignition temperatures suggests that the ignition temperatures of forest humus and litter are influenced not only by primary tree species characteristics, such as their wood density, fiber, and lipid contents, but also by the forest structure, species diversity, topographical factors, and fuel load [42].

Figure 5. (a) Dry-to-fresh ratio of combustibles; (b) characteristics of combustible dry-to-fresh ratio; (c) ignition point of combustible; (d) calorific value of combustible; and (e) characteristics of ignition point and calorific value. Note: Ⅰ represents the shrub layer, Ⅱ represents the herbaceous layer, Ⅲ represents the organic matter layer, Ⅳ represents litter (d < 0.6 cm), Ⅴ represents litter (0.6 cm ≦ d < 2.5 cm), and Ⅳ–Ⅴ represents a mixture of litter-type 1 (d < 0.6 cm) and litter-type 2 (0.6 cm ≤ d < 2.5 cm). The letters in (b,e) represent the results of the analysis of variance. Capital letters indicate significant differences at the 0.01 level, and lowercase letters indicate significant differences at the 0.05 level.

Figure 5. (a) Dry-to-fresh ratio of combustibles; (b) characteristics of combustible dry-to-fresh ratio; (c) ignition point of combustible; (d) calorific value of combustible; and (e) characteristics of ignition point and calorific value. Note: Ⅰ represents the shrub layer, Ⅱ represents the herbaceous layer, Ⅲ represents the organic matter layer, Ⅳ represents litter (d < 0.6 cm), Ⅴ represents litter (0.6 cm ≦ d < 2.5 cm), and Ⅳ–Ⅴ represents a mixture of litter-type 1 (d < 0.6 cm) and litter-type 2 (0.6 cm ≤ d < 2.5 cm). The letters in (b,e) represent the results of the analysis of variance. Capital letters indicate significant differences at the 0.01 level, and lowercase letters indicate significant differences at the 0.05 level.

3.1.3. Analysis of Correlation in Combustible Material Characteristics

According to the results of the Pearson correlation analysis, there is a highly significant positive correlation (p < 0.01) between the dry-to-fresh ratio of combustible materials Ⅰ and Ⅲ, as well as between Ⅲ and Ⅳ. Additionally, the dry-to-fresh ratio of combustible materials Ⅰ, Ⅲ, and Ⅳ shows a highly significant correlation (p < 0.01) with the ignition point of combustible material Ⅲ. Furthermore, there is a significant correlation (p < 0.05) between the dry-to-fresh ratio of combustible material Ⅱ and the calorific value of combustible material Ⅲ. In terms of negative correlations, the dry-to-fresh ratio of combustible material Ⅱ is negatively correlated (p < 0.05) with the dry-to-fresh ratio of combustible material Ⅴ, and the ignition point of the combustible material Ⅳ–Ⅴ mixture is negatively correlated (p < 0.05) with the dry-to-fresh ratio of combustible material Ⅴ. In the forest stands of these 14 plots, there is a positive commonality between the dry-to-fresh ratios of combustible materials Ⅰ and Ⅲ, as well as between the dry-to-fresh ratios of combustible materials Ⅲ and Ⅳ. This indicates a positive correlation between the dry-to-fresh ratios of different combustible material types. From Figure 6, it can be observed that due to the absence of combustible material Ⅲ in some plots, there is a positive correlation between the ignition point and calorific value of combustible material Ⅲ across different plots. On the other hand, the ignition point of the combustible material Ⅳ–Ⅴ mixture shows a negative correlation with its calorific value.

3.1.4. Fire Resistance Evaluation of Forest Stands in Plots

A forest stand refers to a relatively homogeneous community of trees that are somewhat similar in species, age, size, and density. Forest stands are usually the basic unit of forest management and forest ecology studies [43]. In this paper, the fire resistance of forest stands in different sites was obtained by studying the dry-to-fresh ratio, ignition point, and calorific value of forest stands in different sites. According to

Table 3. Comprehensive fire resistance scores of different stand types.

Plot Number	Fresh-to-Dry Ratio/%					Ignite/℃		Calorific Value/(kJ/g)		Final Score
	Ⅰ	Ⅱ	Ⅲ	Ⅳ	Ⅴ	Ⅲ	Ⅳ	Ⅲ	Ⅵ
P1	0.024	0.123	0.049	0.068	0.047	0.051	0.020	0.060	0.026	0.468
P2	0.030	0.017	0.086	0.000	0.047	0.052	0.021	0.017	0.011	0.281
P3	0.013	0.123	0.029	0.032	0.003	0.052	0.032	0.000	0.044	0.328
P4	0.016	0.023	0.058	0.021	0.047	0.051	0.028	0.024	0.000	0.268
P5	0.136	0.123	0.193	0.068	0.047	0.000	0.049	0.085	0.049	0.750
P6	0.058	0.083	0.193	0.178	0.047	0.000	0.049	0.034	0.050	0.692
P7	0.025	0.044	0.072	0.063	0.047	0.050	0.013	0.050	0.014	0.378
P8	0.021	0.123	0.000	0.004	0.047	0.055	0.041	0.022	0.026	0.338
P9	0.000	0.123	0.002	0.032	0.000	0.051	0.034	0.000	0.007	0.249
P10	0.021	0.032	0.062	0.074	0.047	0.050	0.000	0.060	0.018	0.364
P11	0.021	0.018	0.019	0.083	0.047	0.051	0.022	0.056	0.024	0.342
P12	0.025	0.014	0.006	0.006	0.047	0.051	0.021	0.109	0.051	0.330
P13	0.019	0.000	0.014	0.006	0.047	0.051	0.027	0.015	0.070	0.250
P14	0.015	0.020	0.008	0.004	0.047	0.051	0.024	0.051	0.110	0.331

Note: “I” represents the shrub layer, “II” represents the herb layer, “III” represents the organic layer, “IV” represents litter 1 (d < 0.6 cm), and “V” represents litter 2 (0.6 cm ≤ d < 2.5 cm).

, the final fire resistance scores of the sample plots are as follows: Plot 5 has the highest score (0.750), then Plot 6 (0.692), and so on in descending order: Plot 1 (0.468) > Plot 7 (0.378) > Plot 10 (0.364) > Plot 11 (0.342) > Plot 8 (0.338) > Plot 14 (0.331) > Plot 12 (0.330) > Plot 3 (0.328) > Plot 2 (0.281) > Plot 4 (0.268) > Plot 13 (0.250) > Plot 9 (0.249).

Applying the intergroup average chain-linking method for systematic clustering of stand-fire resistance indicators in the 14 sample plots, the fire resistance classification of forest combustibles in Xiuwu County was derived, as shown in Figure 7a. The forest plots were classified into three categories based on the fire resistance indicators (dry-to-fresh ratio, ignition point, and calorific value), with the first category being plots 5 and 6, the second being plot 1, and the third being the rest of the plots. Out of the 14 plots, plots 5 and 6 were more resistant to fire, plot 1 was in the middle of the list, and the rest of the plots were less resistant to fire. The principal component analysis of the three physicochemical indicators of flammability in the 14 sample plots showed that the principal component variance of axis 1 was 42.4% and that of axis 2 was 24.1%, as shown in Figure 7b. The closer the samples are to each other in the graph, the more similar the samples are to each other, and the farther away the samples are, the less similar they are. From axis 1, there is a positive correlation between sample plots 2, 3, 4, 7, 8, 9, 10, 11, 12, 13, and 14, and a negative correlation between sample plots 5 and 6. From axis 2, there is a positive correlation between sample plots 2, 4, 5, 7, 10, 11, 12, 13, and 14, and a negative correlation between sample plots 3, 6, 8, and 9.

3.2. Potential Fire Behavior of Combustible Materials

3.2.1. Classification of Combustible Material Types

The stand factors, such as the mean tree height, mean diameter at breast height, closure, tree layer load, shrub layer load, 1 h time lag load, herb load, and humus load, were used as clustering variables in different stands within the same plot, and the variability of the clustering variables was examined by using a non-parametric test of variance using the Kruskal–Wallis method, as shown below in

Table 4. Kruskal–Wallis test of fuel parameters in different stands.

	Canopy Closure	Stand Density	Average Chest Diameter	Herbal Load	Average Tree Height	Arbor Load	Shrub Load	1 h Time Lag Load	Humus Load
H (K)	3.4929	10.4810	11.9838	2.8502	9.7190	12.2466	2.2667	10.3905	2.7787
asymptotically significant	0.4790	0.0331 *	0.0175 *	0.5832	0.0454 *	0.0156 *	0.6868	0.0343 *	0.5955

Note: * Values denoting asymptotically significant < 0.05, significant difference.

. From the table, it can be seen that there are significant differences in the stand density, average diameter at breast height, average tree height, arbor layer loading, and 1 h time lag loading. Taking the variables with significant differences as the clustering factors, the 14 sample stands in Xiuwu County were systematically clustered, and the clustering results are shown in Figure 7c,d. The woodland types of Xiuwu County’s arborvitae forests can be subdivided into those with poplar-oak forests as the dominant species as one category (Q-1), those with paulownia as the dominant species as one category (Q-2), those with vitex-sumac as the dominant species as one category (G-1), and those with sumac as the dominant species as one category (G-2).

3.2.2. Simulation Analysis of Surface Fire Behavior for Different Types of Combustible Materials

In this study, we employed the BehavePlus software to simulate the key characteristics of forest fires, including the rate of spread, fireline intensity, and flame length, which are critical indicators of wildfire development and propagation. Fuel models corresponding to four types of combustible materials were selected, into which a series of pivotal parameters were inputted. These parameters included the one-hour fuel load, live herbaceous and woody fuel loads, along with their specific surface area/volume, fuel bed depth, and the moisture content of both dry and live fuels for the extinguishing points and calorific values. While the impacts of wind speed and slope on fire behavior have been extensively studied, our research focused on analyzing the fire behavior variations within the ranges of 0–6 m/s wind speeds and a 0–40° slope, utilizing meteorological data from 2011–2020 and field slope data from Xiwu County to yield results with practical applicability. We discovered that increases in wind speed and slope significantly intensified the fire behavior indicators. This finding aligns with the research outcomes of Eftekharian E [44] and Edalati-nejad A [45], corroborating our study and providing invaluable insights into forest fire management and prevention strategies under the specific environmental conditions of Xiwu County.

• Rate of Spread

A forest fire’s spread speed is one of the key parameters in predicting forest fire behavior and the spreading process. The maximum rate of spread of a forest fire is the highest possible rate of spread of a forest fire under certain conditions [46]. By calculating the maximum rate of spread of forest fires, we can provide firefighting commanders with timely and effective information on the dynamics of the fire field and the pattern of change, improving the efficiency of firefighting and ensuring the safety of firefighters. Additionally, the study of the rate of spread provides a reliable reference for planned fires. In areas far from rescue supplies and suppression personnel or in complex terrain, the spread rate of forest fires directly affects whether a fire develops into an uncontrolled fire. With the BehavePlus5.0 fire behavior prediction model, it is possible to derive fire spread rates for different scenarios. From Figure 8, it can be observed that the rate of fire spread increases with the slope and wind speed. Under different slope conditions, when the wind speed increases from 0 m/s to 2 m/s, the rate of fire spread shows a gradual increase. However, when the wind speed increases from 2 m/s to 6 m/s, the rate of fire spread exhibits a significant rise. When the wind speed is constant, the influence of the slope on the rate of fire spread is relatively minor. Under the same slope and wind speed conditions, the rate of fire spread for surface fires in the tree layer is higher for Q-1 (poplar-oak) than for Q-2 (paulownia), while in the shrub layer, the rate of fire spread for G-2 (sumac) is higher than for G-1 (vitex-sumac). This was due to the fact that the 1 h time-lagged fuel load of the poplar-oak forest was greater than that of the paulownia, resulting in a significantly higher rate of surface fire spread than that of the paulownia forest. The 1 h time-lagged fuel load of sumac was also much larger than the 1 h time-lagged fuel load of wattle sumac in the shrub forest sample, resulting in a higher rate of spread [47,48]. The study also found that the sampling sites where Q-1 and G-2 combustibles were located had greater slopes than those where Q-2 and G-1 were located, which is another reason for the higher rate of fire spread between them.

2.

Fireline Intensity

The fireline intensity refers to the amount of heat released by the fire front per unit time in a 1 m wide combustible bed in the direction of its spread. The magnitude of fireline intensity is closely related to the extent of its impact on the forest ecosystem, as well as the firefighting methods and the corresponding firefighting resources. The assessment of fireline intensity allows for an evaluation of its effects on surface materials, soil, microorganisms, and wildlife, among other components of the ecosystem. Based on the destructive intensity, fireline intensity can be classified as low intensity (below 750 kW/m), moderate intensity (750 to 3500 kW/m), and high intensity (3500 to 4000 kW/m). When the fireline intensity exceeds 4000 kW/m, almost all organisms within the forest will be killed by the fire. Only when the fireline intensity is below 4000 kW/m does it hold ecological significance [49].

Figure 9 displays the surface fireline-intensity variation curves for different forest types under varying wind speeds and slopes. From the graph, it can be observed that the surface fireline intensity increases with increasing wind speed and slope. The order of surface fireline intensity from high to low for different forest types, including tree forests and shrub layers, is as follows: poplar-oak > paulownia, sumac > vitex-sumac. In areas with significant terrain variations and high wind speeds, the fireline intensity variation in the Q-1 (poplar-oak) is the most pronounced. The highest fireline intensity is recorded in the Q-1 (poplar-oak), reaching 8247 kW/m, which falls into the high-intensity category. The lowest fireline intensity is observed in the G-1 (vitex-sumac), measuring 5.6 kW/m, which belongs to the low-intensity category. The fireline intensities of other combustible materials fall within the moderate intensity range. The difference between the fireline intensities of the G-1 (vitex-sumac) and the Q-2 (paulownia) is relatively small.

3.

Flame Length

Flame length is the distance from the midpoint of the burning area to the average tip of the flame in the head and is one of the important indicators of fire behavior, which can be converted into fire intensity and is important for the protection of the personal safety of firefighting personnel [50]. Based on the findings in Figure 10, under different slope and wind speed conditions, the flame length of surface fires ranged from 0.17 m to 4.9 m. When the wind speed remained constant, the change in flame length was not significant as the slope increased from 0° to 10°. However, a substantial increase in flame length was observed within the range of 10° to 40° slopes. When the slope remained constant, the flame length gradually increased with the rise in wind speed from 0 m/s to 6 m/s. Among different forest types, the mixed forest of Q-1 (poplar-oak forest) exhibits relatively longer flame lengths and the most significant variation. The flame lengths of Q-2 (paulownia) and G-1 (bramble-sumac) increase similarly, but their increments are relatively smaller.

3.2.3. Sensitivity Analysis

Utilizing Matlab R2020b software, we conducted sensitivity analyses and validated the performance of four combustible fire behavior models, as shown in Figure 11. This sensitivity analysis delved into the effects of wind speed and slope on fire behavior, especially on key indicators like the rate of fire spread, fireline intensity, and flame length. From Figure 11a, it is evident that under varying wind speeds, the Q-1 and G-1 models exhibit exceptionally high sensitivity, particularly regarding fireline intensity, with values significantly higher than those for the flame length and spread rate. This indicates that under changing wind conditions, fireline intensity is the most impactful parameter on the model’s performance.

In contrast, the G-2 model shows higher sensitivity in flame length, while the Q-2 model demonstrates lower sensitivity across all three indicators. Regarding slope, as Figure 11b illustrates, the Q-1 model is highly sensitive to slope changes, especially in fireline intensity, suggesting that fire behavior prediction becomes more complex in areas with steep slopes. Although the G-1 and G-2 models have lower sensitivity in fireline intensity, their significant responsiveness indicates good adaptability to slope variations.

As shown in

Table 5. Effectiveness of combustible model fit.

Model	Spread Rate			Fireline Intensity			Flame Length
	MSE	RMSE	R2	MSE	RMSE	R2	MSE	RMSE	R2
Q-1	0.1111	0.3333	0.9995	2872.3203	53.594	0.9995	0.0052	0.0719	0.9968
Q-2	0.0007	0.0256	0.9995	7.4924	2.7372	0.9995	0.0002	0.015	0.9973
G-1	0.1308	0.3616	0.9835	302.9413	17.4052	0.9835	0.0006	0.024	0.9946
G-2	0.0558	0.2362	0.9996	700.0769	26.459	0.9992	700.0769	26.459	0.9992

, by analyzing the statistical data of MSE, RMSE, and R², it is evident that all models demonstrate high accuracy in predicting the rate of fire spread, with the R² values generally exceeding 0.99. Despite this, notable prediction errors in fireline intensity were observed, particularly for the G-1 model, which exhibited substantially higher MSE and RMSE values compared to other models. This underscores the extreme sensitivity of fireline intensity to factors such as wind speed and slope. The varying sensitivity of these four combustible models to wind speed and slope highlights the performance differences under diverse conditions. Notably, fireline intensity, being highly responsive to changes in wind speed and slope, plays a crucial role in fire management and prediction. Therefore, special attention to the impact of wind speed and slope is necessary in the application of wildfire behavior prediction models to enhance their accuracy and reliability.

3.2.4. Classification of Fire Behavior

Fireline intensity is a function of the rate of spread and heat release per unit area and is directly related to the flame length. Therefore, flame length and fireline intensity are used as fire behavior indicators for classification purposes [51].

Based on the findings in

Table 6. Classification of fire behavior based on flame length and fireline intensity.

Grade	Flame Length	Firebrand Intensity	Explanation of Fire Behavior
1	<0.8	<170	The fire is weak and can be controlled using hand tools
2	0.8–1.1	170~350
3	1.1–2.5	350~1730	Stronger fires require specialized mechanical tools for fighting
4	2.5–3.5	1730~3460	Strong, hard-to-fight fires, forming crown fires and flying fires
5	>3.5	>3460	The fire was very intense, with crown fires, flying fires and large fires

, the fire behavior of the four fuel models was classified and summarized in

Table 7. Fire behavior classification of combustible material models.

Wind Speed	Slope	Type of Combustible
		Q-1	Q-2	G-1	G-2
0	0	1	1	1	1
	10	1	1	1	1
	20	3	1	1	1
	30	3	1	1	2
	40	4	1	1	3
2	0	3	1	1	3
	10	3	1	1	3
	20	3	1	1	3
	30	4	1	1	3
	40	4	2	2	3
4	0	4	2	2	3
	10	5	2	2	3
	20	5	2	2	3
	30	5	2	2	3
	40	5	2	3	4
6	0	5	2	3	4
	10	5	2	3	4
	20	5	2	3	4
	30	5	3	3	4
	40	5	3	3	4

. Q-1 (poplar-oak) exhibited consistently higher fire behavior than the other models under various slope and wind speed conditions. Once a fire occurs, the fire can rapidly escalate, leading to crown fires, spotting, and large-scale flames, especially under conditions of wind speeds at 4 m per second and slopes at 10 degrees. Since the Q-1 (poplar-oak forest) is located near villages and towns, it poses a significant threat to the lives and properties of nearby residents. On the other hand, Q-2 (paulownia) and G-1 (vitex-sumac) demonstrated lower fire behavior under different conditions, allowing for effective suppression using mechanical tools. By managing the understory fuels, the likelihood of fire can be significantly reduced. However, the fire behavior of G-2 (sumac) is relatively intense, especially when the wind speed reaches 6 m per second, leading to escalated fire conditions and highly challenging firefighting operations. Therefore, it is crucial to strengthen the clearing and regulation of this fuel type. Additionally, the fire behavior of this forest type shows an increasing trend with increasing slope, which may be attributed to minimal human interference in steep terrain, resulting in the accumulation of combustible materials and the release of a substantial amount of heat during the burning process, thus triggering severe wildfires [52].

4. Recommendations and Conclusions

4.1. Conclusions

In our detailed study of combustible properties across 14 sites in Xiuwu County, we found that sites 2, 3, 8, 9, and 13 exhibit high dry-to-fresh ratios, indicating lower moisture levels and a heightened propensity for fire outbreaks. Fine litter in site 2, coarse litter in site 3, humus in site 8, shrubs in site 9, and herbs in site 13 all demonstrated increased flammability. Conversely, humus in site 2, shrubs and herbs in site 6, and fine litter in site 11 showed lower moisture content ratios, suggesting lower fire risks.

Further, the Pearson analysis highlighted significant variations in the moisture content ratios among different forest types, with the limited impact of ignition points and calorific values on predicting fire behavior. This insight is crucial for understanding how forest types affect fire spread. The simulation results indicate that the wind speed and slope are key factors that affect the rate of fire spread, fireline intensity, and flame length, providing a basis for risk management.

4.2. Recommendations

For sites with high moisture content ratios, such as sites 2, 3, 8, 9, and 13, priority should be given to preventive measures aimed at reducing the accumulation of combustibles. Maintaining or increasing humus content is a key strategy to lower the likelihood of fires. We recommend the intensified management of flammable stands, such as Q-1 (poplar-oak forests) and G-2 (vitex forests), to prevent potential major fires. Additionally, regular monitoring of fuel moisture and clearing activities are essential fire prevention methods for all forest types.

Ultimately, our study underscores the importance of scientific predictions of fire behavior and effective management strategies for forest conservation amidst global forest fire challenges. Although this research did not fully explore the complexities of the forest understory, our findings provide valuable insights for fire management decision-making and offer new directions for future research.