A Case Study on the Effects of Weather Conditions on Forest Fire Propagation Parameters in the Malekroud Forest in Guilan, Iran

Abstract

The present study investigates the effect of climatic parameters, such as air relative humidity and wind speed, on fire spread propagation indexes in the Malekroud Forest, Iran using the FARSITE simulator based on Rothermel’s original fire spread equation. Standard fuel models are used to calibrate the vegetation cover. Sorensen (SC) and kappa ($\kappa$) coefficients, as well as the Overestimation Index (OI), are used to estimate the simulation’s accuracy. The results confirm that using both ambient condition data and appropriate fuel models is crucial to reaching reasonable results in fire propagation simulations. The values of the Rate of Fire Spread (ROS), Flame Length (FML), and Fire Line Intensity (FLI) are reported for each particular scenario. The simulation results show that the Sorensen and Kappa coefficient for situations most similar to the real fire reached 0.82 and 0.80, respectively. The investigated fire’s severity is categorized as low-condition fire behavior. The simulation shows that fire propagation falls harshly in the case of air relative humidity by more than 72%, and we will not witness natural fire propagation on a large scale.

1. Introduction

Forest fires are a widespread, frequent, and critical disturbance in worldwide terrestrial ecosystems [1,2,3]. As an increasingly common natural disaster, wildfire hazards have affected all the world’s countries [4,5,6,7]. In addition, the potential of climate change and global warming, an ongoing challenge in today’s world, can increase the risk of forest fires due to a combination of rising forest temperatures and the drying of forest fuels [8,9,10,11]. A similar increase in the global trend of forest fires is happening in Iran. Several studies have been conducted to detect high-risk fire areas, especially in the northern forests of Iran [7,12,13]. Therefore, validating simulation systems before an actual fire can effectively boost the performance of firefighting operations and strategic planning [14]. Additionally, investigating the effect of weather conditions and vegetation and the nature of fire on the regime of fire spread is of great importance [15,16].

Wildfire propagation studies have been one of the most critical research topics in recent years [17,18]. Researchers in this field study the region’s fuel type [19,20], vegetation [21,22], and wind speed [23,24]. The Fire Area Simulator, FARSITE, developed in 1998, is one of the leading fire simulation systems used to describe the spread and behavior of wildland fires [25]. FARSITE was developed based on Rothermel’s semi-empirical fire prediction model [26]. The FARSITE fire growth model projects the growth of prescribed natural fires for wilderness areas and requires the support of a geographic information system (GIS) to provide landscape data [25,27]. Unlike stochastic forest fire growth models [28,29], FARSITE generally stimulates the fire’s spatial growth based on Huygens’ elliptical wave propagation principle [30,31]. Although FARSITE is a Global Simulator and does not need to change the governing equations or solution method for different geographical conditions, it requires the local adjustment of climatic parameters to produce reliable results [32,33,34,35]. In several different studies, the capabilities of the FARSITE simulator have been evaluated to investigate the rate of fire spread in topological and climatic conditions [36,37]. In the studies conducted, it is recommended to analyze the sensitivity of the used model. Naturally, this method increases the accuracy and validity of the study results [38,39].

Approximately 130 fires occur in the nature of Iran annually, leading to the burning of more than 5400 hectares of pastures and forests [40]. Fires burn more than 300 ha annually in the 1.2 million ha of temperate forest in the north of Iran [41]. The human factor plays a major role in these northern Iran fires, which is the study site [42,43]. Besides climatic parameters, topography and anthropogenic factors affect wildfire occurrence [44,45]. Forest fires primarily occur due to low vegetation humidity and high wind speeds, which are frequent from August to December, when the forests also have a high vegetation cover. The Malekroud forest around the city of Siahkal in the north of Iran is one of the areas affected by fires at different times [42,46]. This area has been selected as the study area.

To the best of the authors’ knowledge, no studies have been conducted on the simultaneous effect of the climatic parameters of wind speed and ambient humidity. Therefore, this study simultaneously investigates these parameters in a real fire that occurred in the northern forests of the Alborz mountains. The FARSITE model was used to analyze the environmental parameters, such as air humidity and wind speed, in the Malekroud forest, a remnant of the Caspian Hyrcanian mixed forests. The main aims of this study were to investigate the effects of environmental parameters on a fire spread simulation in the SIAHKAL forest using the FARSITE simulator. In addition, the ability of the FARSITE simulator to evaluate the impact of climatic parameters, such as humidity and wind speed, was also investigated. Moreover, wildfire propagation variables, including the rate of fire spread (ROS), fire line intensity (FLI), and flame length (FML), were analyzed based on standard fuel models in the study domain.

2. Materials and Methods

A fire that occurred in the Malekroud forest near the city of Siahkal on 17 December 2010 was selected as the case study. General information about this fire is given in

Table 1. Information required to simulate the Malekroud forest fire [42].

Item	Simulation Information	Input Parameter
1	Latitude of fire starting point	37.10°
2	Longitude of fire starting point	49.85°
3	Height of fire starting point (m)	120
4	Maximum air temperature (°C)	23
5	Minimum air temperature (°C)	5
6	Precipitation (mm)	0
7	Maximum Wind Speed (km/h)	33.79
8	Average Wind Speed (km/h)	19.42
9	Average wind direction	S
10	Average air relative humidity (%)	39
11	The moment the fire starts	17 December 2010—(17:00)
12	The moment of the end of the fire	18 December 2010—(08:00)

. As explained in the previous section, Sorensen and Kappa coefficients were used to evaluate the results of the simulators. To simulate the desired fire in the FARSITE simulator, all the landscape maps (elevation, slope, ground aspect, vegetation fuel, and canopy cover) of the actual state of the study area must first be prepared and processed in ArcGIS software [33,47,48,49]. Then, using the software, different climate parameter scenarios are considered to investigate potential fire propagation in the Malekroud forest. By comparing the Sorensen and Kappa coefficients in selected scenarios, the scenario with the highest degree of agreement in terms of wind speed and air relative humidity is determined [50,51]. The scenarios with the most significant impact on the spread of fire were chosen as an example of the studied scenarios [50]. Simulated fire behavior maps, including fire spread rate, fire line intensity, and flame length, can be fully operationalized by local fire management officials without further processing [52,53].

The location of the Malekroud forest and a map of the fire barriers in the area are shown in Figure 1. Figure 2 shows the vegetation classification of the Malekroud forest code.

2.1. Evaluation Indexes

The Sorensen and Kappa coefficients, numbers between zero and one, were considered as criteria for simulation accuracy. As the values of these two coefficients approach one, the simulation results will be closer to the real state [56,57]. The Sorensen coefficient (SC) is an asymmetric index indicator of the exclusive association between the burned areas (observed and simulated). The Sorensen relation is calculated as follows:

$$\mathrm{S}\mathrm{C}=\frac{2\mathrm{a}}{2\mathrm{a}+\mathrm{b}+\mathrm{c}}$$

(1)

where a is the common area burned in the real and simulated fire, b is the area burned in the simulated fire and the unburned area in the real fire, and c is the area burned in the real fire that is not burned in the simulated fire, all in hectares.

Cohen’s Kappa coefficient ($\kappa$) is a standard nonparametric measure of classification accuracy, which allows for evaluating the overall agreement between simulated and actual areas after removing the random agreements by chance [32]. The Kappa statistical coefficient shows the similarity level of observed spots and simulated fire, calculated in Equation (2) [57]. This coefficient is a standard non-dimensional parameter for classifying the accuracy of simulations and evaluating the overall agreement between simulated and actual areas [32].

$$\mathrm{\kappa }=\frac{\mathrm{N}\sum _{\mathrm{i}=1}^{\mathrm{r}}{\mathrm{x}}_{\mathrm{i}\mathrm{i}}-\sum _{\mathrm{i}=1}^{\mathrm{r}}({\mathrm{x}}_{\mathrm{i}+}\ast {\mathrm{x}}_{+\mathrm{i}})}{{\mathrm{N}}^2-\sum _{\mathrm{i}=1}^{\mathrm{r}}({\mathrm{x}}_{\mathrm{i}+}\ast {\mathrm{x}}_{+\mathrm{i}})}$$

(2)

where r is the number of rows in the matrix, x_ii is the number of observations in row i and column i, x_i+ is the marginal totals of row i, x_+i is the marginal totals of column i, and N is the total number of observations.

Table 2. Interpretation of matching indexes of simulated fire and the real fire stain for Kappa ($\kappa$) and Sorensen (SC) coefficient values [33,51].

Level	$\mathit{\kappa }$ and SC Value Range	Interpretation
1	<0	No agreement
2	[0.0–0.2]	Slight agreement
3	[0.2–0.4]	Fair agreement
4	[0.4–0.6]	Moderate agreement
5	[0.6–0.8]	Substantial agreement
6	[0.8–1.0]	Almost perfect agreement

is presented to help understand the accuracy of the simulated fire results with the real fire spot fire results and the interpretation of the Sorensen and Kappa coefficients [33,51].

Salis et al. derived an overestimation index (OI) to analyze the primary source of overestimation or underestimation errors in the FARSITE simulator. The OI Index is calculated by Equation (3) and varies between 1 and −1. The highest values show significant overestimation errors, and the lowest value shows underestimation. OI values close to 0 indicate limited differences among overestimated and underestimated cells [33].

$$\mathrm{O}\mathrm{I}=\frac{\mathrm{b}-\mathrm{c}}{\mathrm{b}+\mathrm{c}}$$

(3)

These introduced indexes will be very effective in management decision making to support current fires or the analysis of past incidents [58,59,60,61]. When researchers doubt the best simulation condition with similar SC and κ coefficients, the Overestimation Index can help them choose the best scenarios.

2.2. Metrological Data

To effectively manage the fire and implement preventive measures to reduce the damage caused by the fire, it is necessary to know the type and density of vegetation in the area. Vegetation is a combustible material that affects the ease of fire spread. Vegetation is the most studied of all fire spread agents, and determining all its features is complicated due to its heterogeneity and discontinuity. Among environmental parameters, vegetation is the only factor directly regulated or controlled by management operations. Since it is difficult to describe all vegetation’s physical characteristics, such as load (weight per unit area), size (particle diameter), and mass (weight per unit volume) for all fuels in each area, different classification systems that group different types of vegetation based on fire behavior are used to prepare fuel maps. Since accurate information about local forest fuels is not always available, fire management specialists must rely on standard fuel models. However, it is possible to modify these models according to local conditions based on the experience of forest managers or necessary measurements to obtain more logical results. For example, using the Northern Forest Fire Laboratory system (NFFL), Anderson presented 13 combustible models of standard fire behavior in four flammable fuel groups (Grass, Shrub, Timber, and Slash) [62,63]. Scott and Burgan also presented 40 new types of fuel models in six categories, including Non-burnable (NB), Grass (GR), Grass–Shrub (GS), Timber–Understory (TU), Timber Litter (TL), and Slash–Blowdown (SB) [64].

Fuel models for the study site were chosen by noting the similarities between the existing vegetation characteristics and the estimated components of the available fuels and comparing them with the description of the standard fuel models. Figure 2 shows the general vegetation classification in the study area (Malekroud forest surrounding Siahkal city). As shown in Figure 2, the predominant vegetation in the fire zone is broad-leaved medium-density afforestation. To code the vegetation map into FARSITE, each fuel in the Malekroud Forest was assigned a unique code according to the standard Anderson, Scott, and Burgan fuel classification (

Table 3. Malekroud Forest Fuel Code [64].

Item	Fuel	Code
1	Residential Area	NB1
2	Agricultural Land	NB3
3	Uncovered Land	NB9
4	Natural Forest (Timber–Grass–Shrub)	TU3
5	Low-Density Broadleaf Forestry	TL2
6	Medium-Density Broadleaf Forestry	TL6
7	Coniferous Forests	TL8
8	High-Density Broadleaf Forestry	TL9
9	Hardwood litter	FM9
10	Timber (litter and understory)	FM10

).

One of the factors preventing the spread of fire in the region is the existence of roads, lakes, railways, etc.; therefore, it is necessary to identify these fire barriers in the simulation. In some cases, it is possible to estimate the effect of barriers on the rate of fire spread and the fire propagation pattern by considering fire breaks in the FARSITE Simulations. To complete the fire barrier map, the point, line, and polygon maps were separately determined by locating the portion of the Malekroud forest near Siahkal city in Global Mapper software. Then, the files were prepared by Utilizing ArcGIS software shapefiles. Figure 1 shows the area’s road map and other fire barrier factors. The proximity of the combustion starting point to the road and the location of the Malekroud forest between the roads were important refactors for preventing and controlling the spread of the fire in real time. Since the moisture content of live and dead fuels is not available before a fire, the moisture content of dead fuels in 1 h, 10 h, and 100 h classes based on fire environment parameters (climatic factors, topography, and vegetation) was determined using the Rothermel method, according to

Table 4. Parameters of the method used to determine the initial moisture content of dead fuel based on the Rothermel method [55,65,66].

Item	Simulation Information	Value
1	Average air temperature (°C)	20
2	Average air relative humidity (%)	39
3	Fuel Moisture Correction	7
4	Fire month	December
5	Shading status of vegetation in the area	Shady
6	Height of the fire point relative to the meteorological station	Above the meteorological station
7	Slope orientation	Southern
8	Ground slope	0–30%
9	Dead Fuel Moisture Content Correction (%)	6
10	The initial moisture of the dead fuel	13

.

The reference number for the moisture content of the fuel was selected based on the climatic parameters, temperature, and air relative humidity of the study area at the time of the fire [66]. Additionally, the percentage of the correction of fuel moisture was determined based on other parameters, such as the month and time of the fire, the condition of fuel cover and shading, and the topographic factors elevation, aspect, and slope of the ground. Defined classes for predicting fire behavior severity are presented in

Table 5. Defined classes for predicted fire behavior [64].

ROS (m/min)	Fire Line Intensity (kW/m)	Flame Length (m)	Fire Behavior Severity
0–0.6	0–10	0–0.3	Very low
0.6–1.6	10–100	0.3–1.2	Low
1.6–6.6	100–1000	1.2–2.4	Medium
6.6–16.6	1000–10,000	2.4–3.6	High
16.6–50	10,000–100,000	3.6–7.5	Very High
>50	>100,000	>7.5	Intense

.

In addition to the vegetation type and the moisture content of combustible material, which are among the most effective factors in determining the spread of fire in nature and are mostly fixed for a forest area based on the weather conditions, wind speed and air relative humidity are two very influential factors in the forest fire spread. In this case study, these two factors have been examined in detail regarding the Malkroud forest.

3. Results

3.1. Validation Data

The validation section investigates the effects of changing different fuels on fire behavior and the rate of fire spread in the Malekroud forest surrounding Siahkal city. In this section, five simulation scenarios are defined. Scenarios were created by changing the fuel type in the medium- and high-density broadleaf forest. The fuel codes replaced in the real fire scenario were FM9 and FM10. These two types of standard fuels were developed by the Anderson codes and are used for forest areas with high flammability.

The statistical results of the simulation of these scenarios are shown in

Table 6. Statistical results of the Malekroud forest plant type validation. (a) Common area burned in the real and simulated fire (hectares), (b) Area burned in the simulated fire and unburned in the real fire (hectares), (c) Area burned in the real fire and not burned in the simulated fire (hectares); SC: Sorensen coefficient, $\kappa$: Kappa statistical coefficient, and Overestimation Index.

Scenario	Flammable Fuel Groups	a (ha)	b (ha)	c (ha)	SC	$\mathit{\kappa }$	OI
V1	(TL6, TL9)	16.21	5.29	7.84	0.71	0.68	−0.19
V2	(FM9, TL9)	20.14	6.06	3.91	0.80	0.78	0.22
V3	(TL6, FM9)	15.77	4.63	8.28	0.71	0.68	−0.28
V4	(TL6, FM10)	16.74	4.26	7.31	0.74	0.72	−0.26
V5	(FM9)	18.94	5.16	5.11	0.79	0.76	0.01

. After comparing the Sorensen and Kappa coefficients of the different scenarios, we concluded that the simulation with the highest percentage of agreement with the real fire spot is scenario-V2. Table 5 shows that scenario-V2 produced Sorensen and Kappa coefficients equal to 0.8 and 0.78, respectively, according to the simulation output results. A map of the spread of fire in the study area between the time intervals specified in the FARSITE Simulator shows that scenario-V2 has the highest percentage of agreement with the actual fire spot (Figure 3). It should be noted that in the validation procedure, the hourly wind speed and direction recorded in the meteorological center of Siahkal city have been used for simulation, which is the closest meteorological data regarding the wind speed and direction to the wind spatial data during the fire moment in Malekroud forest, located in Siahkal city. The wind speed and direction are considered fixed through the fire domain for each hour.

In Table 6, the intensity of the fuel types of the study area is compared with the fire behavior classes predicted by Scott and Burgan’s standard fuel model. The model was developed and used to appraise the intensity of fire spread in the fire rating system. The mathematical model applies to a wide range of fuels and is used for predicting the rate of fire spread and intensity.

The values of fire spread rate, fire line intensity, and flame length for different fuel scenarios are provided in

Table 7. Output values of Rate of Fire Spread (ROS), Fire Line Intensity (FLI), and Flame Length (FML) for validation scenarios.

Fuel Type 1	ROS (m/min)	Fire Line Intensity (kW/m)	Flame Length (m)
(TL6, TL9)	0.41 ± 0.16	42.51 ± 28.43	0.41 ± 0.11
(FM9, TL9)	0.47 ± 0.18	40.87 ± 28.43	0.41 ± 0.11
(TL6, FM9)	0.37 ± 0.14	26.88 ± 10.44	0.34 ± 0.06
(TL6, FM10)	0.46 ± 0.16	47.24 ± 37.78	0.42 ± 0.14
(FM9)	0.43 ± 0.17	27.33 ± 11.31	0.34 ± 0.06

¹ Flammable Fuel Groups. (±) shows the range between the minimum and maximum value over the simulated area around the average value.

. The maximum value of these parameters was found in scenario V2 with fuel types FM9 and TL9. The intensity of fire behavior, according to Table 5, will be in the low range. The values of the rate of fire spread, flame length, and fire line intensity in the Malekroud forest for the simulation condition of scenario-V1 conditions are presented in Figure 4.

Figure 4 shows that the typical values of the rate of fire spread, flame length, and fire line intensity in the simulated fire area are in the range of 0.13 < ROS < 0.50, 0.14 < FML < 0.36, and 3.35 < FLI < 63.46, respectively.

Figure 4. Results for real simulated fuel (TL6, TL9). (a) ROS: Rate of Fire Spread (m/min), (b) FML: Flame Length (m), and (c) FLI: Fire Line Intensity (kW/m).

Figure 4. Results for real simulated fuel (TL6, TL9). (a) ROS: Rate of Fire Spread (m/min), (b) FML: Flame Length (m), and (c) FLI: Fire Line Intensity (kW/m).

All selected scenarios are discussed in Appendix B. In Appendix B, the wind-rose for the Malekroud forest (from 2003 to 2019 from October 26 to January 26) is drawn. Additionally, the annual maximum, average, and minimum temperature values and the yearly maximum, average, and minimum air relative humidity values in the Malekroud forest from 2003 to 2019 are illustrated.

Table A2. Temperature and air relative humidity scenarios for simulation (minimum and maximum air temperature are, respectively, 5 °C and 17 °C).

Item	Scenario	Objective Variable	Average Wind Speed (km/h)	Average Air Relative Humidity (%)
1	Ref.	-	19.42	39
2	Sc.01	Air Relative Humidity	19.42	2
3	Sc.02	Air Relative Humidity	19.42	22
4	Sc.03	Air Relative Humidity	19.42	42
5	Sc.04	Air Relative Humidity	19.42	62
6	Sc.05	Air Relative Humidity	19.42	72
7	Sc.06	Air Relative Humidity	19.42	82
8	Sc.07	Air Relative Humidity	19.42	92
9	Sc.08	Air Relative Humidity	19.42	99
10	Sc.09	Wind Speed	4.82	39
11	Sc.10	Wind Speed	16.09	39
12	Sc.11	Wind Speed	38.62	39
13	Sc.12	Wind Speed	120.70	39
14	Sc.13	Wind Speed	4.82	22
15	Sc.14	Wind Speed	16.09	22
16	Sc.15	Wind Speed	19.31	22
17	Sc.16	Wind Speed	28.97	22
18	Sc.17	Wind Speed	38.62	22

is the reference scenario table throughout this study.

3.2. Simulation Result

The FARSITE simulations are reported in a range of eight ambient humidity ratio scenarios and nine wind speed scenarios, as defined in

Table 8. Sorensen Number (SC), Kappa Number ($\kappa$), Fire Spread Rate (ROS), Fire Line Intensity (FLI), and Flame Length (FML) for different scenarios.

Scenario	a (ha)	b (ha)	c (ha)	SC	$\mathit{\kappa }$	OI	ROS (m/min)	Fire Line Intensity (kW/m)	Flame Length (m)
Sc.01	22.68	10.22	1.37	0.80	0.77	0.76	0.58 ± 0.24	74.54 ± 56.61	0.54 ± 0.16
Sc.02	17.32	5.68	6.73	0.74	0.71	−0.08	0.44 ± 0.19	50 ± 37.04	0.44 ± 0.13
Sc.03	14.35	4.75	9.70	0.67	0.63	−0.34	0.39 ± 0.18	41.71 ± 30.26	0.4 ± 0.12
Sc.04	12.04	3.66	12.01	0.61	0.57	−0.53	0.32 ± 0.14	33.86 ± 24.37	0.36 ± 0.11
Sc.05	10.11	3.19	13.04	0.55	0.52	−0.61	0.27 ± 0.13	29.66 ± 23.77	0.34 ± 0.12
Sc.06	6.45	2.35	17.60	0.39	0.36	−0.76	0.23 ± 0.1	25.28 ± 17.62	0.32 ± 0.11
Sc.07	4.45	1.85	19.60	0.29	0.26	−0.83	0.16 ± 0.06	15.07 ± 8.13	0.25 ± 0.07
Sc.08	5.03	2.07	19.02	0.32	0.29	−0.80	0.19 ± 0.08	21.11 ± 12.86	0.29 ± 0.09
Sc.09	10.08	3.32	13.97	0.54	0.50	−0.62	0.25 ± 0.14	31.51 ± 24.4	0.35 ± 0.11
Sc.10	10.73	3.37	13.32	0.56	0.53	−0.60	0.26 ± 0.13	31.37 ± 22.27	0.35 ± 0.1
Sc.11	14.20	4.00	9.85	0.67	0.64	−0.42	0.33 ± 0.11	35.82 ± 21.04	0.38 ± 0.09
Sc.12	22.19	7.61	1.86	0.82	0.80	0.61	0.55 ± 0.25	53.39 ± 28.08	0.46 ± 0.1
Sc.13	13.66	4.74	10.39	0.64	0.61	−0.37	0.31 ± 0.13	14.95 ± 7.99	0.37 ± 0.12
Sc.14	17.34	5.76	6.71	0.74	0.71	−0.08	0.39 ± 0.13	14.80 ± 8.02	0.40 ± 0.12
Sc.15	19.39	6.61	4.66	0.77	0.75	0.17	0.44 ± 0.14	15.06 ± 8.01	0.42 ± 0.12
Sc.16	23.75	14.75	0.3	0.75	0.72	0.96	0.64 ± 0.17	15.37 ± 8.12	0.49 ± 0.13
Sc.17	23.75	37.95	0.3	0.55	0.52	0.98	0.88 ± 0.28	15.76 ± 7.71	0.60 ± 0.16

Symbol (±) means the parameter varies in a range plus/minus the statistical average value.

. The Sorensen and Kappa coefficients, rate of fire spread (ROS), fire line intensity (FLI), and flame length (FML) for all 17 considered scenarios are reported.

The simulation results show that as the air relative humidity increases, the values of the rate of fire spread, fire line intensity, and flame length and the percentage of its compatibility with the real fire stain decrease. The Sorensen and Kappa coefficients for the first scenario are 0.8 and 0.77, respectively. This scenario shows the highest Sorensen and Kappa coefficients among the eight air relative humidity scenarios.

As explained, the simulations are developed based on the scenarios defined in Table A2. The graphs and contours in Figure 5, Figure 6, Figure 7, Figure 8, Figure 9, Figure 10, Figure 11, Figure 12 and Figure 13 are plotted based on the data in Table 8 and corresponding simulations.

Generally, it should be noted that scenarios are divided into two main categories, fixed air relative humidity and fixed wind speed situations, which are widely demonstrated in Appendix B. Scenarios 1 to 8 belong to simulations with the wind speed fixed at 19.42 km/h, while the relative humidity varies from 2% to 99%. Scenarios 9 to 17 belong to wind speed change simulations for two different air relative humidities equal to 39% (Sc.9 to Sc.12) and 22% (Sc.13 to Sc.17).

The diagrams of the Sorensen coefficient, Kappa coefficient, Overestimation Index, Flame Length, Rate of Spread, and Fire Line Intensity, as well as the fire spread contours for the three main scenario categories, are illustrated in Figure 5, Figure 6, Figure 7, Figure 8, Figure 9, Figure 10, Figure 11, Figure 12 and Figure 13.

Figure 5 and Figure 6 belong to air relative humidity change scenarios with wind speeds equal to 19.42 km/h. Figure 12 and Figure 13 belong to air relative humidity change scenarios with wind speeds equal to 38.62 km/h. In the other category, Figure 7 and Figure 8 belong to wind speed change scenarios with an air relative humidity equal to 39%.

However, the two Sorensen and Kappa coefficients alone cannot be a measure for determining the degree of conformity of the simulated fire with real conditions. Therefore, in addition to these two parameters, the Overestimation Index (OI) should also be taken into account. As seen in Table 8 and Figure 5, Scenario 1 cannot be the optimal scenario due to the high value of the OI index. Although Scenario 2 has slightly lower Kappa and Sorensen coefficients (about 7% less), it is considered a more suitable scenario due to the small value of OI (−0.08). As a result, Scenarios 13 to 17 are chosen with the fixed value of air relative humidity at 22%, while the wind speed changes from 4.82 km/h to 38.62 km/h.

Figure 5. Result of air relative humidity change scenarios in Malekroud forest; (a) Sorensen coefficient (SC), Kappa coefficient ($\kappa$), and Overestimation Index (OI); (b) Flame Length, (c) Rate of Spread, and (d) Fire Line Intensity. (Error bars mean the parameter varies in a range plus/minus the statistical average value).

Figure 5. Result of air relative humidity change scenarios in Malekroud forest; (a) Sorensen coefficient (SC), Kappa coefficient ($\kappa$), and Overestimation Index (OI); (b) Flame Length, (c) Rate of Spread, and (d) Fire Line Intensity. (Error bars mean the parameter varies in a range plus/minus the statistical average value).

Figure 6. Fire spread lines between specified time intervals in air relative humidity change scenarios; wind speed is fixed at 19.42 km/h (based on Table A2).

Figure 6. Fire spread lines between specified time intervals in air relative humidity change scenarios; wind speed is fixed at 19.42 km/h (based on Table A2).

By juxtaposing the shape of the fire spread lines at regular intervals in the scenario Sc.01 simulation of moisture changes, we see that the fire spread lines go beyond the actual fire spot, and the number of fire spread lines decreases as the humidity increases (Figure 6). On the other hand, as the wind speed increases, the simulation of the simulated fire increases compared to the real fire stain. The Sorensen and Kappa coefficients for scenario Sc.12, with a wind speed of 120.70 km/h, are 0.82 and 0.80, respectively. According to Figure 8, the fire spread lines between the specified time intervals for scenario Sc.12 show the highest percentage of agreement with the actual fire stain.

Figure 5 shows the results for air relative humidity change scenarios in Malekroud forest. The Sorensen coefficient, Kappa coefficient, and Overestimation Index have been plotted versus air relative humidity in Figure 7a. Flame Length, Rate of Spread, and Fire Line Intensity are respectively plotted in Figure 7b–d.

Based on the results, according to Figure 5, the rate of spread, flame length, and fireline intensity decrease with the increase in air humidity. The slope of this reduction is sharper at a low humidity, especially for an air relative humidity of less than 22%. For an air relative humidity of more than 72%, the defined class for predicting fire behavior falls into “very low” fire behavior severity based on Table 5 [64].

Figure 6, Figure 8 and Figure 12 show the fire spread lines between specific time intervals for different scenarios. These figures especially demonstrate the fire propagation pattern during different climatical conditions qualitatively.

Figure 6 shows the fire spread lines between specified time intervals in the air relative humidity change scenarios for a wind speed of 19.42 km/h (based on Table 8). As shown in Figure 6, and also according to Figure 5a, the Overestimation Index (OI) for the air humidity equal to 22% reaches the nearest value to zero (−0.08). At the same time, Sorensen (SC) and Kappa ($\kappa$) coefficients have relatively high values (respectively, 0.74 and 0.70).

Figure 7. Result of wind speed change scenarios in Malekroud forest; (a) Sorensen coefficient (SC), Kappa coefficient ($\kappa$), and Overestimation Index (OI); (b) Flame Length, (c) Rate of Spread, and (d) Fire Line Intensity. (Air relative humidity is fixed at 39%.) (Error bars mean the parameter varies in a range plus/minus the statistical average value.)

Figure 7. Result of wind speed change scenarios in Malekroud forest; (a) Sorensen coefficient (SC), Kappa coefficient ($\kappa$), and Overestimation Index (OI); (b) Flame Length, (c) Rate of Spread, and (d) Fire Line Intensity. (Air relative humidity is fixed at 39%.) (Error bars mean the parameter varies in a range plus/minus the statistical average value.)

Figure 8. Fire spread lines between specified time intervals in wind speed change scenarios (based on Table A2). Air Relative Humidity is fixed at 39%, and wind speed values are 4.89, 16.09, 38.63, and 120.70 km/h.

Figure 8. Fire spread lines between specified time intervals in wind speed change scenarios (based on Table A2). Air Relative Humidity is fixed at 39%, and wind speed values are 4.89, 16.09, 38.63, and 120.70 km/h.

Although in the simulation of scenario number 12, a speed of 120.7 km/h is considered, according to Table 1, the maximum speed observed during the fire in Malekroud forest is much lower than this number and is equal to 33.79 km/h. Therefore, in addition to the very high Overestimation Index (0.61), this scenario cannot be considered a suitable analysis of the simulation results, and researchers should analyze the consequences of Figure 9 and Figure 10 without considering this scenario. Additionally, we can conclude that the rate of spread, flame length, and fire line intensity increases with the air humidity. Among Scenarios 9 to 12, Scenario 11 has the highest consistency with the real condition.

Figure 9. The rate of fire spread (m/min) for the scenarios with the highest percentage of compliance; (a) changes in wind speed, and (b) changes in air relative humidity.

Figure 9. The rate of fire spread (m/min) for the scenarios with the highest percentage of compliance; (a) changes in wind speed, and (b) changes in air relative humidity.

Similar to Figure 9, Figure 10 shows the result of wind speed change scenarios in Malekroud forest, with an air relative humidity of 39%. Figure 11 shows the fire spread lines between specified time intervals in the air relative humidity change scenarios based on the simulation results.

Figure 9, Figure 10, and Figure 11, respectively, show the rate of fire spread, the flame length, and the fire line intensity values for the scenarios with the highest percentage of consistency with the real fire.

Figure 10. The Flame Length (m) for the scenarios with the highest percentage of compliance; (a) changes in wind speed and (b) changes in air relative humidity.

Figure 10. The Flame Length (m) for the scenarios with the highest percentage of compliance; (a) changes in wind speed and (b) changes in air relative humidity.

Figure 11. The Fire Line Intensity (kW/m) for the scenarios with the highest percentage of compliance; (a) changes in wind speed and (b) changes in air relative humidity.

Figure 11. The Fire Line Intensity (kW/m) for the scenarios with the highest percentage of compliance; (a) changes in wind speed and (b) changes in air relative humidity.

Scenarios 13 to 17 are defined for wind speed change at an air relative humidity of 22%, and Figure 12 and Figure 13 belong to these scenarios. Among Scenarios 13 to 17, Scenarios 14 and 15 have the highest consistency with the real condition, with high Kappa ($\kappa$) and Sorensen (SC) coefficient values and a low Overestimation Index (OI). Despite the high-value Kappa coefficient for scenario 15 (0.72), it is ignored because of the high Overestimation Index (0.96). According to Figure 13, although the rate of spread and flame length increase by wind speed, the fire line intensity does not experience a noticeable boost.

Figure 12. Fire spread lines between specified time intervals in wind speed change scenarios (based on Table A2). Air Relative humidity is fixed at 22%, and wind speed values are 4.82, 16.09, 19.31, 28.97, and 38.62 km/h.

Figure 12. Fire spread lines between specified time intervals in wind speed change scenarios (based on Table A2). Air Relative humidity is fixed at 22%, and wind speed values are 4.82, 16.09, 19.31, 28.97, and 38.62 km/h.

Figure 13. Result of wind speed change scenarios in the Malekroud forest; (a) Sorensen coefficient (SC), Kappa coefficient ($\kappa$), and Overestimation Index (OI); (b) Flame Length, (c) Rate of Spread, and (d) Fire Line Intensity. (Air relative humidity is fixed at 22%.) (Error bars mean the parameter varies in a range plus/minus the statistical average value.)

Figure 13. Result of wind speed change scenarios in the Malekroud forest; (a) Sorensen coefficient (SC), Kappa coefficient ($\kappa$), and Overestimation Index (OI); (b) Flame Length, (c) Rate of Spread, and (d) Fire Line Intensity. (Air relative humidity is fixed at 22%.) (Error bars mean the parameter varies in a range plus/minus the statistical average value.)

The contours of the essential parameters in the simulation are plotted independently. Figure 12 shows the rate of fire spread for the scenarios, with the highest percentage of compliance in scenarios with changes in wind speed and humidity. Similarly, Figure 13 show the flame length and fire line intensity, respectively. The fire spread lines between specified time intervals in wind speed change scenarios with an Air Relative Humidity fixed at 22% are demonstrated in Figure 12 (Scenarios 14 to 18). Additionally, the results for the same scenarios for the Sorensen and Kappa coefficients, flame length, spread rate, and fire line intensity are indicated in Figure 13. Comparing the rate of fire spread in scenarios with changes in humidity and wind speed under the same simulation conditions shows that changes in ambient moisture have the most significant effect on the rate of surface fire spread compared to other parameters.

4. Discussion

According to the simulation results of the air relative humidity scenarios, the high air relative humidity in Malekroud forest at the time of a fire can be one factor preventing the rapid fire spread. In the case of an air relative humidity of more than 42%, the probability of a fire spread rate decreases, and for a relative humidity of air of more than 72%, the fire propagation falls harshly, and we have not witnessed natural fire propagation on a large scale. Considering the high humidity level in the natural broadleaf forests of northern Iran, the possibility of a fire occurring in it due to a single environmental factor alone is minimal. It requires a set of climatic conditions to cause a fire. Among other factors affecting the fire spread in the region, we can mention wind speed and direction parameters. Due to the fire breaks, such as roads and urban areas around the studied area, the effect of these parameters on fire spread is not wholly noticeable. If there were no fire breaks, the effect of the prevailing wind direction in the area on the fire spread could be observed in more detail. The authors’ simulations show that despite the rolling effect of fire breaks on fire propagation patterns, the simulation results are thoroughly conditioned by climatic parameters, especially wind speed and wind direction.

Overall, scenarios are divided into two main categories: wind speed change and air relative humidity change scenarios. The results of air relative change scenarios in Malekroud forest, demonstrated in Figure 5 and Figure 6, show that the rate of spread, flame length, and fire line intensity decrease with the increase in air humidity. The reduction acceleration is much higher at less than 22% air relative humidity. For an air relative humidity of more than 72%, the defined class for predicting fire behavior falls into “very low” fire behavior severity based on Table 5.

The results of wind speed scenarios are demonstrated in Figure 7 and Figure 8. As expected, the rate of spread, flame length, and fire line intensity increase with wind speed. On the other hand, the Sorensen (SC) and Kappa (κ) coefficients, as well as the Overestimation Index (OI), also increase with high acceleration. Increasing the overestimation index to the highest level in all scenarios is not desirable. Three scenarios have the highest consistency with the real condition, with high Kappa (κ) and Sorensen (SC) coefficient values and a low Overestimation Index (OI) (14, 15, and 16). Despite the high-value Kappa coefficient for scenario 16 (more than 0.72), it is ignored because of the high Overestimation Index (0.96).

In the studies conducted mainly in the field of forest fires in the north of Iran, especially regarding Hyrcanian forests, the effect of wind speed and direction parameters, as well as relative air humidity, on the parameters of fire spread has not been investigated so far. Of course, these studies have been conducted in a decentralized manner in other parts of the Middle East, but what distinguishes this study from other studies is the examination of all the essential parameters for describing the behavior of forest fires using simulation evaluation indicators. The combined use of the mentioned indicators and parameters makes it easier for researchers to predict and analyze future fire behavior in Hyrcanian forests.

5. Conclusions

Considering the gap in studies on forest fires in northern Iran, this study simultaneously investigates the simultaneous effect of the climatic parameters of wind speed and ambient humidity in a real fire in the Alborz Hyrcanian forests in northern Iran. This study investigated the effects of climate scenarios, including changes in air relative humidity and wind speed, on the fire behavior and patterns in the Malekroud forest surrounding Siahkal city. The simulation results, including the fire spread rate, flame length, and fire line intensity for each scenario, were calculated and separately compared in the tables embedded in the results section.

Comparing the rate of fire spread in the scenarios of changes in air relative humidity and wind speed under the same simulation conditions shows that changes in air relative humidity have the most significant effect on the rate of fire spread compared to other parameters, including flame length and fire line intensity. The simulation results show that the maximum Sorensen and Kappa coefficients for the scenarios most similar to the real situation were 0.82 and 0.80. The actual fire in the Malekroud forest was classified as low fire behavior severity. Although less commonly used in studies, the authors recommend using the Overestimation Index (OI) to analyze the primary source of errors in the simulations, especially in cases where researchers are uncertain as to which scenarios with relatively similar Kappa and Sorensen numbers they should select.

Recorded atmospheric data such as air temperature, air humidity, and wind speed and direction are recorded hourly (and not in shorter periods) for the Siahkal Meteorological Center (and not for different parts of the region), and this issue leads to uncertainty in the simulation. On the other hand, natural fire breaks limit the possibility of accurately evaluating the simulation results. Simulations show that among a set of conditions that affect fire propagation, considering the high humidity level in the Hyrcanian forests of northern Iran and limitations cussed by natural fire breaks, fire propagation patterns are thoroughly conditioned by climatic parameters.