Influence of Climatic Factors on Lightning Fires in the Primeval Forest Region of the Northern Daxing’an Mountains, China

Forest fires lead to permafrost degradation and localized drought, and regional droughts increase the probability of forest fires, leading to a positive feedback loop between climate change and fires. However, the relationship between fire occurrence and climatic factors change is unclear for boreal forests, which represent the largest land-based biome and stock of carbon. Here, we analyzed the relationship between lightning fire occurrence and meteorological and topographic factors based on the fire frequency, burned area, and meteorological data from the primeval forest region of the northern Daxing’an Mountains in China. We found that lightning fires occurred most frequently at an altitude of 600 to 700 m. From 1999 to 2019, the frequency of lightning fires showed an overall upward trend, whereas the affected area had no obvious change. It can be attributed to fire suppression efforts and greatly increased investment in fire prevention in China. Snow cover had a strong regulatory effect on the start and end dates of lightning fires for seasonal cycle. The frequency of lightning fires was positively correlated with the average temperature, maximum temperature, and surface evaporation and negatively correlated with precipitation and surface soil moisture (0–10 cm). The result will be useful in the spatially assessment of fire risk, the planning and coordination of regional efforts to identify areas at greatest risk, and in designing long-term lightning fires management strategies.


Introduction
Climate warming is faster at higher latitudes and altitudes. Boreal forests, which are the largest biome and largest living stock of carbon on land, show climate warming faster than the global mean level and have experienced unprecedented interference from large fires during the past 10,000 years [1,2]. The increase in fire disturbances is mainly due to the increased lighting ignition associated with warming [3][4][5][6], extended fire seasons [7,8], and higher fuel aridity with better fuel composition [9,10]. Forest fires are important for controlling the structure, succession, and functions of boreal forests [11,12]. At the same time, forest fires may cause boreal forests to change to a net carbon source from a net carbon sink by emiting large amounts of organic carbon to the atmosphere [13,14]. Forest fires emit carbon dioxide, black carbon, and aerosols, which aggravate global warming [15] and thus increase fire risk [16,17]. Currently, despite extensive anthropogenic suppression efforts, the observed forest fire intensity is still increasing significantly [18][19][20][21].
Lightning fires are characterized by randomness, concurrency, and strong concealment. Lightning fires occur all over the world. For example, in Ontario, Canada, lightning fires accounted for 43% of forest fires in the past 26 years but caused 81% of the burned area [22]. In the past 30 years, 52% of all forest fires in Alaska in the United States were caused by lightning. In Spain, lightning fires only account for 3.9% of all fires but 10.7% of the total burned area [23]. Lightning fires are common in China's forested areas, and fires caused by lightning strikes in China account for approximately 1% to 2% of the country's forest fires, which are mainly distributed in Heilongjiang Daxing'an Mountains, Inner Mongolia Daxing'an Mountains, and Xinjiang Altai Mountains [24]. However, until 2020, in the Daxing'an Mountains forested area of Inner Mongolia, lightning fires accounted for 21% of all forest fires in the region, with a maximum of 64%. Since 2010, the Daxing'an Mountains have been affected by El Niño and La Niña phenomena, resulting in abnormal weather events such as high temperatures, strong winds, and uneven precipitation (rainy period lags, obvious rainfall periods, and drought periods) [25]. Frequent dry thunderstorms have also occurred, increasing the likelihood of forest fires [26].
Meteorological factors such as rainfall, wind speed, temperature, and humidity are critical factors for studying the formation of lightning fires [27,28]. Nash and Johnson found that under low-rainfall conditions, higher air pressure is associated with greater ignition probability of cloud-to-ground flashes [29]. Since precipitation data are easy to obtain from conventional weather stations, researchers often use daily or annual precipitation as the reference threshold for lightning fires. It has been found that in Arizona and New Mexico, 2 mm is the threshold of daily precipitation for ignition, while due to the influence of terrain, rainfall, and other factors in Spain, the fluctuation range of the daily precipitation for ignition is 9~22.5 mm [30]. In addition, Shu et al. found that the number of lightning fires in a region is directly related to the amount of precipitation in China. The longer the drought days, the more lightning fires. When the annual precipitation is more than 600 mm, the occurrence of lightning fire is lower, and when the annual precipitation is 350-580 mm, the occurrence of lightning fire is greater than in previous years [31].
The terrain factors that influence lightning strikes mainly includes slope, slope aspect, and altitude. Shu et al. found that lightning strikes mostly occur in the midslope of mountains at elevations above 800 m above sea level in the Pinus pumila forest and the Pinus sylvestris-Pinus pumila forest [32]. Some studies have suggested that lightning strikes are more likely to occur in areas with low altitude and steep slopes because these areas experience stronger thunderstorms compared to areas with higher altitude and less steep slopes [33][34][35]. It also has been found that in high-altitude areas, low temperatures and heavy rainfall significantly reduce the probability of lightning strikes igniting combustibles [36].
At present, lightning fires are complex and uncontrollable natural phenomena, which cause huge losses to human beings and the ecological environment. We can probably capture an early warning signal of lightning fires based on the geographic information and the high accuracy of meteorological monitoring data of the detection area Lightning fires are the primary origin of forest fires in the Daxing'an Mountains region of China. In the present study, we established a dataset that includes the fire frequency and burned area for boreal forest in the primeval forest region of the northern Daxing'an Mountains of China. By analyzing the occurrence law of existing lightning fires and the relationship with climatic factors, the main reasons for the occurrence of lightning fires are explored, which provides a basis for the reconstruction of the early warning system of lightning fires in the future.

Study Region
The study region was the virgin forest area in the northern part of the Daxing'an Mountains in Inner Mongolia, China. The forest area is 9 × 10 5 km 2 , the forest coverage rate is 95.0%, and the geographic coordinates are 120 • 01 20 -121 • 48 37 E and 52 • 01 42 -53 • 20 00 N. As of 2019, a total of 245 fires were caused by lightning strikes over the preceding 20 years. In this study, the lightning fire was investigated by the local forestry department on the fire scene, and it was confirmed as lightning fire by finding lightning strike traces and lightning strike wood. These lightning fires have occurred throughout the study region ( Figure 1). The forest in the study region mainly includes typical coldtemperate forest species such as Larix olgensis, Pinus sylvestris, Populus davidiana, Betula platyphylla, and their mixed forest. The annual average temperature is −5.5 • C, the extreme minimum temperature is −53 • C, and the extreme maximum temperature is 35.4 • C. The annual average precipitation is 450-550 mm and occurs over approximately 88 days.

Study Region
The study region was the virgin forest area in the northern part of the Daxing'an Mountains in Inner Mongolia, China. The forest area is 9 × 10 5 hm 2 , the forest coverage rate is 95.0%, and the geographic coordinates are 120°01′20″-121°48′37″ E and 52°01′42″-53°20′00″ N. As of 2019, a total of 245 fires were caused by lightning strikes over the preceding 20 years. In this study, the lightning fire was investigated by the local forestry department on the fire scene, and it was confirmed as lightning fire by finding lightning strike traces and lightning strike wood. These lightning fires have occurred throughout the study region ( Figure 1). The forest in the study region mainly includes typical coldtemperate forest species such as Larix olgensis, Pinus sylvestris, Populus davidiana, Betula platyphylla, and their mixed forest. The annual average temperature is −5.5 °C, the extreme minimum temperature is −53 °C, and the extreme maximum temperature is 35.4 °C. The annual average precipitation is 450-550 mm and occurs over approximately 88 days.

Fire and Climate Data
We collected lightning fire data (1999 to 2019) from the forest management and protection bureau of the primeval forest region of the northern Daxing'an Mountains, Inner Mongolia. The collected data included the time, location, and altitude of each lightning fire along with the number of fires per year. We extracted monthly climate data (resolution = 0.5°) including precipitation, temperature, and the self-calibrating Palmer drought se-

Fire and Climate Data
We collected lightning fire data (1999 to 2019) from the forest management and protection bureau of the primeval forest region of the northern Daxing'an Mountains, Inner Mongolia. The collected data included the time, location, and altitude of each lightning fire along with the number of fires per year. We extracted monthly climate data (resolution = 0.5 • ) including precipitation, temperature, and the self-calibrating Palmer drought severity index from the Climate Research Unit TS 4.03 dataset (http://climexp.knmi.nl) (accessed on 1 February 2021). The mean climatic conditions of the study region were estimated by averaging the climate data over the area from 50-56 • N, 118-124 • E. The soil moisture (0~10 cm) data were downloaded from ERA5 (https://cds.climate.copernicus.eu) (accessed on 1 February 2021). The snow cover data were downloaded from Rutgers University Global Snow Lab (https://climate.rutgers.edu/snowcover) (accessed on 1 February 2021). We downloaded monthly relatively humidity data (red triangles in Figure 1) from meteorological stations in the study region, while one-month standardized precipitationevapotranspiration index (SPEI) values were obtained from the Spanish National Research Council (https://www.csic.es/en/csic) (accessed on 1 February 2021).
The years with fire occurrence and SPEI above and below the 95% distribution limits were defined as the years of most frequent fire and drought stress, respectively. The relationship between fire occurrence and drought severity was evaluated by calculating Pearson's correlation coefficients. The level of significance of each correlation was determined using the two-tailed null hypothesis.
Digital elevation map data were used as the base map in ArcGIS 10.8 software to analyze the location of lightning fires along with the area of lightning fires in the study region and comprehensively evaluate the spatial characteristics of lightning fires in the study area.

Spatial Distribution of Lightning Fires
The 245 lightning fires that occurred in the study region between 1999 and 2019 were mainly concentrated from 121 • 00 11 -121 • 10 48 east longitude and 52 • 39 00 -52 • 45 49 north latitude ( Figure 2). The frequency of lightning fires was directly related to altitude. However, there was no lightning fire in the study area in 2009. As shown in Figure 2, all lightning fires occurred in the elevation range of 303-1445 m above sea level and were irregularly distributed. Lightning fires were mainly concentrated in areas with elevations of 600-700 m above sea level; 79 fires or 32% of all fires occurred within this elevation range. A previous study has shown that the active area of lightning strikes is affected by the local sources of humidity and the topography [37]. Du et al. reported that lightning fires occur in the elevation range of 200-1300 m above sea level, with fires being particularly concentrated in areas in the elevation range of 300-800 m, which is consistent with our results [38]. The main reasons for the easy occurrence of lightning fires in the middle-altitude area are the high temperature, the easy evaporation of precipitation, and the low water content of combustibles. In contrast, fewer lightning fires were observed at altitudes above 800 m due to lower temperatures, weaker precipitation evaporation, and higher water content of combustibles. In addition, the spatial discontinuity of combustibles in high-altitude areas has a strong inhibitory effect on the occurrence of lightning fires [36]. All together, there is a certain correlation between the occurrence of lightning fires and the altitude. Lightning fires mainly occur in the middle-altitude areas in the primeval forest region of the northern Daxing'an Mountains, Inner Mongolia.

Frequency of Lightning Fires and the Trend of Burned Area
As shown in Figure 3, the frequency of lightning fires in the study region from 1999

Frequency of Lightning Fires and the Trend of Burned Area
As shown in Figure 3, the frequency of lightning fires in the study region from 1999 to 2019 was different in the early season (May to June), late season (July to September), and the entire fire season (May to September). The lightning fire frequency showed an overall upward trend over time, with increase rates of 1.95, 1.91, and 3.86 times/decade. It has been found that with the global warming in recent years, extreme weather such as high temperatures and drought in the Daxing'anling Mountains have increased significantly, which makes the summer forest lightning fires in this area increase year by year. Simultaneously, the main tree species in this study area are coniferous forests, which have high lipid levels and are flammable. The combustibles of coniferous species and coniferous forests are more susceptible to lightning fire than other types of combustibles [38,39]. Coniferous forests are mostly located at high latitudes in cold temperate zones, where the frequency of cloud-to-ground lightning is higher than in low-latitude regions, which is also an important reason for the increase in the frequency of lightning fires [40]. However, in this study, the frequency of lightning fires did not show a significant increasing trend (p > 0.05). At the same time, there is no obvious change in the area of lightning fires over the past 20 years (p > 0.05). This result may be attributed to fire suppression efforts and greatly increased investment in fire prevention in China. Therefore, strengthening the monitoring and early warning of key periods and key areas of lightning fires can effectively avoid the occurrence of major and extra-large forest fires. Simultaneously, the main tree species in this study area are coniferous forests, which have high lipid levels and are flammable. The combustibles of coniferous species and coniferous forests are more susceptible to lightning fire than other types of combustibles [38,39]. Coniferous forests are mostly located at high latitudes in cold temperate zones, where the frequency of cloud-to-ground lightning is higher than in low-latitude regions, which is also an important reason for the increase in the frequency of lightning fires [40]. However, in this study, the frequency of lightning fires did not show a significant increasing trend (p > 0.05). At the same time, there is no obvious change in the area of lightning fires over the past 20 years (p > 0.05). This result may be attributed to fire suppression efforts and greatly increased investment in fire prevention in China. Therefore, strengthening the monitoring and early warning of key periods and key areas of lightning fires can effectively avoid the occurrence of major and extra-large forest fires.

Effect of Snow Accumulation on Lighting Fires
As shown in Figure 1b, snow cover has a strong effect on the start and end dates of lightning fires. Lightning fires can only occur after the snow cover melts, and the lightning fires end before the snow cover begins to accumulate. When snow and ice cover more than 20% of the woodland surface, the wildfire caused by lightning strikes are almost impossible. Lutz et al. [41] found that the reduction in spring snow cover and the earlier ablation time increased the number of lightning fires by investigating the relationship between spring snow cover and lightning fires, which is consistent with our results. The interactions among wildfire occurrence, climate change, and snow cover also create uncertainty regarding the occurrence of regional forest fires. A previous study has shown that global climate warming leads to vegetation changes characterized by a decline in alpine species, an increase in open woodland taxa, and a substantial reduction in annual snow cover, which contribute to an increase in regional fire activity [42]. Taken together, our results suggested that snow cover monitoring should be strengthened in high-risk lightning fires

Effect of Snow Accumulation on Lighting Fires
As shown in Figure 1b, snow cover has a strong effect on the start and end dates of lightning fires. Lightning fires can only occur after the snow cover melts, and the lightning fires end before the snow cover begins to accumulate. When snow and ice cover more than 20% of the woodland surface, the wildfire caused by lightning strikes are almost impossible. Lutz et al. [41] found that the reduction in spring snow cover and the earlier ablation time increased the number of lightning fires by investigating the relationship between spring snow cover and lightning fires, which is consistent with our results. The interactions among wildfire occurrence, climate change, and snow cover also create uncertainty regarding the occurrence of regional forest fires. A previous study has shown that global climate warming leads to vegetation changes characterized by a decline in alpine species, an increase in open woodland taxa, and a substantial reduction in annual snow cover, which contribute to an increase in regional fire activity [42]. Taken together, our results suggested that snow cover monitoring should be strengthened in high-risk lightning fires areas.
As shown in Figure 4, the Julian dates of summer lightning fires during 1999-2019 were concentrated in 120~280 d. Through the Julian date analysis, the summer lightning fires in this area mainly occurred in late June. With the change in climate, the fire danger period of summer lightning fires was prolonged later. Lightning fires occurred frequently from 152 d to 213 d (around 1 June to 31 July), with a total of 170 fires, accounting for 78% in the past 20 years. However, we found that snow accumulation and snow melt had no significant effect on the lightning-fire period (p > 0.05). At the same time, the high incidence of lightning fires is from June to July each year. By strengthening the aviation patrol during the high-incidence period of lightning fires and carrying out forward garrison, the frequency and the area of lightning fires can be effectively reduced.

Correlation between Lightning Fires and Meteorological Indicators
Weather conditions influence the likelihood of ignition, and the likelihood of lightning fires can be reconstructed based on existing meteorological data [43,44]. In the present study, the frequency of lightning fires in the early season (May to June), late season (July to September), and the entire season (May to September) from 1999 to 2019 was positively correlated with the average temperature, maximum temperature, and surface evaporation ( Figure 5), whereas it was negatively correlated with surface soil moisture (0-10 cm) and precipitation. Among these factors, the correlations with soil moisture and evaporation rate were the most significant, with correlation coefficients of −0.6 and 0.5, respectively. We can infer that the surface evaporation rate increases as the temperature increases. Drought conditions (lack of precipitation) decrease the water contents of surface soil and surface combustibles, thereby increasing the likelihood of lightning fires. It has also been shown that lightning fires in Inner Mongolia were mainly concentrated in spring (March to May) and summer (June to August) [45]. The likelihood of lighting fires and fire severity are probably increased by a long dry season without any change in seasonal thunderstorms [46]. Taken together, the frequency of lightning fires was positively correlated with average temperature, maximum temperature, and surface evaporation and negatively correlated with precipitation and surface soil moisture. We should strengthen the monitoring of meteorological data in this area to provide support for lightning fires early warning.

Correlation between Lightning Fires and Meteorological Indicators
Weather conditions influence the likelihood of ignition, and the likelihood of lightning fires can be reconstructed based on existing meteorological data [43,44]. In the present study, the frequency of lightning fires in the early season (May to June), late season (July to September), and the entire season (May to September) from 1999 to 2019 was positively correlated with the average temperature, maximum temperature, and surface evaporation ( Figure 5), whereas it was negatively correlated with surface soil moisture (0-10 cm) and precipitation. Among these factors, the correlations with soil moisture and evaporation rate were the most significant, with correlation coefficients of −0.6 and 0.5, respectively. We can infer that the surface evaporation rate increases as the temperature increases. Drought conditions (lack of precipitation) decrease the water contents of surface soil and surface combustibles, thereby increasing the likelihood of lightning fires. It has also been shown that lightning fires in Inner Mongolia were mainly concentrated in spring (March to May) and summer (June to August) [45]. The likelihood of lighting fires and fire severity are probably increased by a long dry season without any change in seasonal thunderstorms [46]. Taken together, the frequency of lightning fires was positively correlated with average temperature, maximum temperature, and surface evaporation and negatively correlated with precipitation and surface soil moisture. We should strengthen the monitoring of meteorological data in this area to provide support for lightning fires early warning. (March to May) and summer (June to August) [45]. The likelihood of lighting fires and fire severity are probably increased by a long dry season without any change in seasonal thunderstorms [46]. Taken together, the frequency of lightning fires was positively correlated with average temperature, maximum temperature, and surface evaporation and negatively correlated with precipitation and surface soil moisture. We should strengthen the monitoring of meteorological data in this area to provide support for lightning fires early warning. Figure 5. Correlation coefficients between meteorological factors and lightning fire frequency in May to June, July to September, and May to September. The black and red asterisks represent significant levels of over 95% and 99%, respectively. The seven meteorological elements from left to Figure 5. Correlation coefficients between meteorological factors and lightning fire frequency in May to June, July to September, and May to September. The black and red asterisks represent significant levels of over 95% and 99%, respectively. The seven meteorological elements from left to right are: average temperature, maximum temperature, minimum temperature, precipitation, drought index, soil moisture, and transpiration rate.

Trends in Key Climate Indices over the Past 20 Years
From 1999 to 2019, the overall climate conditions in the study region during the early lightning fire season (May to June), late season (July to September), and the entire season (May to September) became warmer and more humid. Specifically, the average temperature, SPEI index (higher SPEI indicates a wetter climate, whereas lower SPEI indicates a drier climate), and precipitation all increased from 1999 to 2019 ( Figure 6). Fuel aridity is enhanced as drought severity increases, leading to more frequent forest fires and larger burned areas [47]. The warming itself along with the corresponding enhancement in drought stress, increased frequency of extreme hot events, and longer fire season also likely increase the risk for fire [48,49]. Ni et al. found that rainfall and summer temperature in the Greater Khingan Mountains are the main meteorological factors that affect the occurrence of lightning fires. Lightning fires were negatively correlated with rainfall and positively correlated with summer temperature, among which rainfall and lightning fires had the strong correlation, which is consistent with our results [50]. In addition, Fill et al. found that less rainfall over a longer dry season likely increases both the potential for lightning-ignited wildfires and fire severity [46].
Meanwhile, we found that the soil moisture and SPEI index increased most significantly from May to June, whereas the soil moisture decreased significantly from July to September. The reduction in soil surface moisture has a certain promotion effect on the formation and spread of fire after lightning strikes ignited combustibles. If the opposite trend in soil moisture continues, it may lead to lightning fires in the early days in the future. Altogether, we concluded that the changes in soil moisture, SPEI index, and precipitation should be paid more attention during the lightning-fire periods.
Meanwhile, we found that the soil moisture and SPEI index increased most significantly from May to June, whereas the soil moisture decreased significantly from July to September. The reduction in soil surface moisture has a certain promotion effect on the formation and spread of fire after lightning strikes ignited combustibles. If the opposite trend in soil moisture continues, it may lead to lightning fires in the early days in the future. Altogether, we concluded that the changes in soil moisture, SPEI index, and precipitation should be paid more attention during the lightning-fire periods.

Conclusions
We found that lightning fires occurred most frequently at an altitude of 600-700 m in the primeval forest region of the northern Daxing'an Mountains. Snow cover had a strong regulatory effect on the start and end dates of lightning fires, and snow cover monitoring

Conclusions
We found that lightning fires occurred most frequently at an altitude of 600-700 m in the primeval forest region of the northern Daxing'an Mountains. Snow cover had a strong regulatory effect on the start and end dates of lightning fires, and snow cover monitoring should be strengthened in high-risk lightning fires areas. The frequency of lightning fires was positively correlated with average temperature, maximum temperature, and surface evaporation and negatively correlated with surface soil moisture (0-10 cm) and precipitation (Figure 7). Acquiring more accurate advance and real-time meteorological monitor data in the future will be useful in the spatially assessment of fire risk, in identifying areas at greatest risk, and in designing long-term lightning fires management strategies. should be strengthened in high-risk lightning fires areas. The frequency of lightning fires was positively correlated with average temperature, maximum temperature, and surface evaporation and negatively correlated with surface soil moisture (0-10 cm) and precipitation ( Figure 7). Acquiring more accurate advance and real-time meteorological monitor data in the future will be useful in the spatially assessment of fire risk, in identifying areas at greatest risk, and in designing long-term lightning fires management strategies.