Biomass Burning in Northeast China over Two Decades: Temporal Trends and Geographic Patterns

Abstract

Despite the significant impacts of biomass burning (BB) on global climate change and regional air pollution, there is a relative lack of research on the temporal trends and geographic patterns of BB in Northeast China (NEC). This study investigates the spatial–temporal distribution of BB and its impact on the atmospheric environment in the NEC region during 2004 to 2023 based on remote sensing satellite data and reanalyzed data, using the Siegel’s Repeated Median Estimator and Mann–Kendall test for trend analysis, HDBSCAN to identify significant BB change regions, and Moran’s Index to examine the spatial autocorrelation of BB. The obtained results indicate a fluctuating yet overall increasing BB trend, characterized by annual increases of 759 for fire point counts (FPC) and 12,000 MW for fire radiated power (FRP). BB predominantly occurs in the Songnen Plain (SNP), Sanjiang Plain (SJP), Liaohe Plain (LHP), and the transitional area between SNP and the adjacent Greater Khingan Mountains (GKM) and Lesser Khingan Mountains (LKM). Cropland and urban areas exhibit the highest growth in BB trends, each surpassing 60% (p < 0.05), with the most significant growth cluster spanning 68,634.9 km². Seasonal analysis shows that BB peaks in spring and autumn, with spring experiencing the highest severity. The most critical periods for BB are March–April and October–November, during which FPC and FRP contribute to over 80% of the annual total. This trend correlates with spring planting and autumn harvesting, where cropland FPC constitutes 71% of all land-cover types involved in BB. Comparative analysis of the aerosol extinction coefficient (AEC) between areas with increasing and decreasing BB indicates higher AEC in BB increasing regions, especially in spring, with the vertical transport of BB reaching up to 1.5 km. County-level spatial autocorrelation analysis indicates high–high clustering in the SNP and SJP, with a notable resurgence of autocorrelation in the SNP, suggesting the need for coordinated provincial prevention and control efforts. Finally, our analysis of the impact of BB on atmospheric pollutants shows that there is a correlation between FRP and pollutants, with correlations for PM_2.5, PM₁₀, and CO of 0.4, 0.4, and 0.5, respectively. In addition, the impacts of BB vary by region and season, with the most significant impacts occurring in the spring, especially in the SNP, which requires more attention. In summary, considering the escalating BB trend in NEC and its significant effect on air quality, this study highlights the urgent necessity for improved monitoring and strategic interventions.

1. Introduction

Biomass burning (BB) is a widely used process that is typically initiated by humans for activities such as forest clearing, rotational agriculture, and vegetation clearing after crop harvesting [1,2]. However, it releases substantial particulates and trace gases, impacting weather, climate change, public health, and air quality [3,4,5]. These particulates act as cloud condensation nuclei and alter the reflectivity and longevity of clouds, thereby potentially suppressing rainfall, exacerbating water scarcity, and adversely affecting agricultural productivity, which could further amplify regional and even global climate variability [6,7,8]. BB affects not only the local area but also neighboring areas in many cases, and even long-distance transmissions have been reported [9,10,11].

With global warming, the spatial and temporal distribution of BB and its impact on air quality have increasingly attracted the attention of scholars [12,13,14,15,16]. Over the past 20 years, global fire points have shown significant regional and seasonal variations [17]. In particular, in regions with high BB emissions, through analyzing the spatiotemporal distribution of BB in South Asia and Southeast Asia from 2001 to 2018, it was found that crop residue burning (CRB) in South Asia increased by 844 fire spots per year [18]. In China, the most severe fire points are in Northeast China (NEC), comprising the provinces of Liaoning (LN), Jilin (JL), and Heilongjiang (HLJ), which is the country’s largest grain-producing region, contributing about 25% of the country’s active fires and 40% of its CRB [19,20]. To reduce uncertainty in the multi-year spatial and temporal variability of BB in NEC and its impact on the atmospheric environment, several scholars have utilized satellite data to investigate the characteristics of BB across different vegetation types in the region. Wang et al. [21] discovered that from 2003 to 2017, the fire points in NEC exhibited a fluctuating yet generally rising trend, peaking during 2013–2017. BB was primarily concentrated in March–April (37%) and October–November (46%). Moreover, agricultural fires constituted most of all fires (90.8%), followed by forests (5.3%) and grasslands (3.1%). Between 2002 and 2016, emissions from forests, organic, and biogenic carbon declined in NEC, whereas emissions from agricultural fields increased, signaling a shift from natural forest fires to human-induced agricultural fires as the primary BB source [22]. Zhao et al. [23] observed that greenhouse gases and particulate matter from BB in NEC were predominantly concentrated in the major grain-producing areas of Songnen Plain (SNP) and Sanjiang Plain (SJP) from 2012 to 2019. Throughout this period, total fire emissions first rose, then fell in a fluctuating manner, peaking in 2015 when additional emissions regulations were enacted. Additionally, the main burning season transitioned from autumn to spring due to stricter burning and crop residue usage controls. The contribution of spring burning grew from 20.5% in 2013 to 77.1% in 2019, an annual increase of 20%. However, with global climate change, accelerated urbanization, and policy changes in recent years, BB patterns in different vegetation types in NEC may be changing; therefore, longer time-series are needed to explore the associated trends and changes, including annual, seasonal, and long-term changes. This would help us understand whether there are gradual increasing or decreasing trends in fire activity and what may be driving these trends. Moreover, Zeng et al. [24], through an analysis of the satellite-observed aerosol vertical structure, discovered a significant increase in upper-level aerosol optical depth (AOD) over the Greater Bay Area during severe BB days in SEA. Therefore, the vertical distribution of BB in the Northeast is of greater concern. In particular, there have been few studies examining whether there are patterns of clustering or dispersion in the distribution of BB. Investigating the clustering of BB characteristics could assist government authorities and policymakers in formulating more forward-thinking policies and regulations to effectively address the challenges posed by BB. Additionally, previous studies have mainly focused on vegetation BB, whereas fires in urban areas pose greater harm [25]. NEC is the largest old industrial base of China, with the urban coal heating season lasting for 4–6 months [26,27]. Therefore, it is necessary to analyze urban fire along with vegetation BB in the area.

Our initial analysis of the fire point distribution across different regions in China (Figure 1a) revealed that the NEC region indeed experiences the highest concentration of BB events, prompting a focused study on this area. Despite governmental measures since 2008 to reduce straw incineration and encourage commercial energy, NEC has experienced severe atmospheric pollution from BB in recent years [4]. Therefore, it is essential to investigate the temporal and spatial distribution patterns of BB in NEC and its impact on the atmospheric environment. In this paper, we analyze the spatial and temporal distribution characteristics of BB in NEC during the 20-year period from 2004 to 2023 based on remote-sensing observational data and reanalyzed data—mainly including fire point counts (FPC), fire radiative power (FRP), and aerosol extinction coefficient (AEC)—and further explore the spatial clustering of BB and its correlation with atmospheric pollutants. Section 2 describes the data set used and the research methodology. Section 3 describes the spatial and temporal distribution characteristics of BB, as well as spatial aggregation and correlation with atmospheric pollutants in the NEC region. Section 4 and Section 5 contain the discussion and conclusions, respectively.

2. Materials and Methods

2.1. Study Area

Figure 1a depicts the distribution of mean fire points in China from 2004 to 2022, with the most severe BB occurring in the study area—Northeast China (NEC)—indicated by the red border. Figure 1b depicts NEC, spanning from approximately 118.85° to 135.10°E longitude and 38.64° to 53.56°N latitude, with elevations ranging from 0 to 2691 m. It mainly consists of three major plains and three major mountain ranges. The plains include the Songnen Plain (SNP), Sanjiang Plain (SJP), and Liaohe Plain (LHP), while the mountain ranges are the Greater Khingan Mountains (GKM), Lesser Khingan Mountains (LKM), and Changbai Mountains (CBM). Figure 1c illustrates the four-category classification based on the International Geosphere-Biosphere Programme (IGBP) vegetation types, including forest, cropland, urban, and other. Detailed subcategories for each category can be found in Table S1.

2.2. Materials

NASA’s Moderate Resolution Imaging Spectroradiometer (MODIS) fire product, known for its long-time span and which is freely available, has been widely utilized in the study of BB [18,25,28,29]. The MOD14A1 [30] data set was utilized in this study to acquire fire-related data, including FPC and FRP. FPC was derived from the FireMask variable and reflects the confidence level of fire detection, while FRP was derived from the MaxFRP variable and indicates the intensity of the fire [31]. A $0.1°\times 0.1°$ grid was created to aggregate and present the BB spatial distributions. MCD12Q1 [32] provides global maps of land-cover types, and was used in this study to classify BB. MOD14A1 and MCD12Q1 data sets are available for access from the Earthdata website (https://www.earthdata.nasa.gov/, accessed on 20 May 2024). Considering that BB aerosols constitute a significant portion of primary combustion aerosol emissions [33], in order to investigate the vertical distribution characteristics of aerosols during significant changes in BB areas, Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) Level 2 satellite products (CAL_LID_L2_05kmAPro [34]), including AEC, were obtained from https://www-calipso.larc.nasa.gov/, accessed on 20 May 2024. Additionally, to further explore the impact of BB emissions on the atmospheric environment, aerosol optical depth (AOD) and some trace gas pollutants (ozone (O₃), carbon monoxide (CO), and sulfur dioxide (SO₂)) were collected from the Earthdata website. Carbon dioxide (CO₂) and methane (CH₄) data were collected from the Atmosphere Data Store (available at https://ads.atmosphere.copernicus.eu/, accessed on 20 May 2024). Rainfall, snowfall, and atmospheric boundary layer height data were obtained from the Climate Data Store (available at https://cds.climate.copernicus.eu/, accessed on 20 May 2024).

Table 1. Remote sensing satellite data and reanalysis data.

Products	Names	Temporal	Spatial Resolution
Satellite data	MOD14A1	2004–2023	1 km
	MCD12Q1	2004–2022	500 m
	CAL_LID_L2_05kmAPro	2013–2023	30 m
	OMAEROe	2013–2023	0.25°
	OMDOAO3e	2013–2023	0.25°
	MOP03JM	2013–2023	1°
	OMSO2e	2013–2023	0.25°
Reanalysis data	CAMS global greenhouse gas reanalysis (EGG4) monthly averaged fields	2013–2020	0.75°
	ERA5-Land monthly averaged data from 1950 to present	2004–2023	0.1°
	ERA5 monthly averaged data on single levels from 1940 to present	2004–2023	0.25°

summarizes all the data sets utilized in this study.

To validate the impact of BB on the atmospheric environment as derived from satellite data, air quality data from ground monitoring stations across NEC were incorporated, which span the period from 2015 to 2023. These include measurements of PM_2.5, PM₁₀, and CO, and were sourced from the China National Environmental Monitoring Centre (https://air.cnemc.cn, accessed on 20 May 2024).

2.3. Methods

2.3.1. Trend Analysis

Trend analysis has been widely used for investigating BB trends [17,19,35]. Fan et al. [17] and Yin [18] have used the Siegel’s repeated median estimator and Mann–Kendall trend test to study the spatiotemporal variation characteristics of global fires and the spatiotemporal distribution of BB in South Asia and Southeast Asia, respectively.

Theil [36] first proposed using the median of all paired slopes as a regression estimator, which is robust to outliers. Sen [37] extended this method to handle ties. The Theil–Sen estimator has become a popular non-parametric statistical technique and has been widely applied in various fields [38,39,40]. Siegel [41] used nested medians to replace the single mean, with a breakdown point of 50% compared to 29.3% with the Theil–Sen estimator. Siegel’s Repeated Median Estimator is particularly suitable for long-term time-series analysis of BB, due to its robustness to outliers and ability to accurately capture trends in highly variable data sets. In the binary linear regression model $Y=A+BX$, the slope of Siegel’s Repeated Median Estimator is

$$\widehat{B}=\underset{i}{\mathit{median}}\underset{j\ne i}{\mathit{median}}\left(\frac{y_i-y_j}{x_i-x_j}\right)$$

(1)

where $\left(x_i,y_i\right)$ represents the point pair of years and FPC (or FRP). If $\widehat{B}>0$, it indicates that the FPC (or FRP) has an increasing trend over time; if $\widehat{B}<0$, it indicates a decreasing trend; and if $\widehat{B}=0$, there is no change.

The Mann–Kendall (MK) trend test [42,43] is a non-parametric method based on the rank of data rather than the data, which has been widely used to detect trends in time-series [44,45,46]. MK analysis is suitable for studying BB because it effectively identifies trends without requiring the data to follow a specific distribution, making it robust against the irregularities and outliers common in BB data. For a time-series $\left(x_1,x_2,\dots ,x_n\right)$, the statistics for the MK trend test are as follows:

$$S=\sum _{i=1}^{n-1}\sum _{j=i+1}^{n}sgn\left(x_j-x_i\right)$$

(2)

where $n$ is the number of data points, and the $sgn$ function is defined as follows:

$$sgn\left(\theta \right)=\left\{\begin{array}{c}\begin{array}{cc}1& \theta >0\end{array}\\ \begin{array}{cc}0& \theta =0\end{array}\\ \begin{array}{cc}-1& \theta <0\end{array}\end{array}\right.$$

(3)

The variance of $S$ is

$$var=\frac{1}{18}\left[n\left(n-1\right)\left(2n+5\right)-{\sum }_{p=1}^q{f}_p\left(f_p-1\right)\left(2f_p+5\right)\right]$$

(4)

where $q$ is the number of tied groups and $f_p$ is the amount of data in in each group. MK standardized statistics are calculated according to the following formula:

$$Z_{\mathrm{M}\mathrm{K}}=\left\{\begin{array}{cc}\frac{S-1}{\sqrt{var}}& S>0\\ 0& S=0\\ \frac{S+1}{\sqrt{var}}& S<0\end{array}\right.$$

(5)

With a given significance level α = 0.05, when $p<0.05$, it is considered to pass the significance test.

2.3.2. Correlation Analysis and Cluster Analysis

Cluster Analysis

Hierarchical Density-Based Spatial Clustering of Applications with Noise [47,48] (HDBSCAN), a state-of-the-art clustering algorithm [49], is used for clustering analysis in this study. This algorithm is an extension of the well-known DBSCAN algorithm [50]. DBSCAN has been widely used to identify dense fire clusters [51,52,53,54,55]. Unlike DBSCAN, which relies on a single global density parameter ε (epsilon), HDBSCAN operates by performing clustering over varying ε values and integrating the results to identify the clustering that demonstrates the best stability across these values [56]. This feature enables HDBSCAN to detect clusters of varying densities, which is particularly advantageous for handling the non-uniform densities often present in satellite-derived BB data.

The algorithm begins by calculating the core distance for each data point, which is defined as follows:

$$c_k\left(x\right)=d\left(x,N_k\left(x\right)\right)$$

(6)

where $N_k\left(x\right)$ refers to the $k^{th}$ nearest neighbor of the data point x. Furthermore, the mutual reachability distance is defined as

$$d_{mreach,k}\left(a,b\right)=max\left\{c_k\left(a\right),c_k\left(b\right),d\left(a,b\right)\right\}$$

(7)

where $d\left(a,b\right)$ represents the Euclidean distance between points $a$ and $b$. A minimum spanning tree is then constructed using these mutual reachability distances as weights. The tree is traversed and its edges are re-sorted, forming a hierarchical structure that is subsequently compressed and segmented. During this process, the minimum cluster size parameter determines the smallest size that a cluster must be to be considered valid. The extraction of clusters and evaluation of their stability are quantified using the stability measure λ, which is determined according to the inverse of the distance. The stability of these clusters, $s_{cluster}$, is evaluated using the following formula:

$$s_{cluster}={\sum }_{p\in C_{cluster}}\left({\lambda }_p-{\lambda }_{death}\right)$$

(8)

where $C_{cluster}$ is the set of points in a cluster, and ${\lambda }_p$ and ${\lambda }_{death}$ are the values representing when a point separates from the parent cluster and when the cluster splits into two sub-clusters, respectively.

In our application, HDBSCAN was applied to remotely sensed data of BB, specifically to analyze the increases and decreases in fire points and their radiative power. The minimum cluster size parameter was set at 15, chosen based on a careful understanding of the data characteristics and extensive testing. This setting optimally balances the level of detail in clustering against statistical significance, effectively avoiding excessive clustering while capturing significant clusters. Furthermore, a sensitivity analysis of the minimum cluster size parameter was conducted. The results are presented in the Supplementary Materials (Figure S1), demonstrating how different settings affect the clustering outcomes and underscoring the critical nature of parameter selection in spatial clustering.

Spatial Autocorrelation

Spatial autocorrelation refers to the degree to which a set of spatial data points is correlated with itself in space. Moran’s Index is a measure of spatial autocorrelation and has been used to identify pollution distribution patterns [57,58,59,60,61]. In this study, Moran’s Index is used to analyze the spatial autocorrelation of BB. The Moran’s index includes the global Moran’s index [62] and local Moran’s index [63]. The global Moran’s index can indicate whether clustering or outliers have occurred in the overall space, while the local Moran’s index can indicate where clustering or outliers have occurred.

Global Moran’s index is defined as

$$I=\frac{n}{S_0}\frac{\sum _{i=1}^n\sum _{j=1}^n{w}_{ij}{z}_i{z}_j}{\sum _{i=1}^n{z}_i^2}$$

(9)

where $n$ is the total number of spatial units, $z_i=x_i-\overline{x}$, $z_j=x_j-\overline{x}$, $i$ and $j$ are indices of spatial units, $x_i$ and $x_j$ are the attribute values of the corresponding spatial units, $\overline{x}$ is the mean value of all spatial unit attribute values, $w_{ij}$ is the spatial weight value, and $S_0$ is the sum of all the weights. With row-standardized weights, $S_0$ is equal to $n$. As a result, the global Moran’s index can be simplified to

$$I=\frac{\sum _{i=1}^n\sum _{j=1}^n{w}_{ij}{z}_i{z}_j}{\sum _{i=1}^n{z}_i^2}$$

(10)

The global Moran’s index ranges between −1.0 and +1.0. When the value is greater than 0, it indicates that the data show a spatial positive correlation; When the value is less than 0, it indicates that the data exhibit a spatial negative correlation; When the value is 0, it indicates that the data exhibit randomness.

The Local Moran’s index is defined as

$$I_i=c·z_i\sum _{j=1}^n{w}_{ij}{z}_j$$

(11)

where $c=1/\sum _{i=1}^n{z}_i^2$, and $\sum _{j=1}^n{w}_{ij}{z}_j$ is the spatial lag of $z_i$.

The Moran scatter plot provides a visual representation of spatial autocorrelation through plotting the spatially lagged variable against the original variable. The slope of the resulting linear fit indicates the global Moran’s index. The Moran scatter plot divides spatial autocorrelation into four categories based on four quadrants: high–high, low–low, high–low, and low–high. The local indicators of spatial association (LISA) cluster map combines significance information obtained from the Moran scatter plot with the spatial location of each observation, categorizing spatial autocorrelation into five groups: high–high, low–low, high–low, low–high, and not significant.

3. Results

3.1. Spatiotemporal Distribution Characteristics and Cluster Analysis

Figure 1a shows the distribution of fire points across various regions in China, with NEC exhibiting the highest concentration of BB activities.

Figure 2 describes the 20-year temporal trends for four types, annual average spatial distribution, and MK results of BB in NEC. The first column represents FPC, and the second column shows FRP. The BB exhibits a fluctuating but generally rising trend, with annual growth rates of 759 spots for FPC and 12,000 MW for FRP (Figure 2a,b). Wang et al. [21] also indicated that FPC in NEC from 2003 to 2017 demonstrated a fluctuating but generally increasing trend. The 20-year period can be divided into four periods: the first period (2004–2012) is characterized by a gradual increase; the second period (2013–2017) is marked by a rapid increase; the third period (2018–2020) is characterized by a sharp decline; and the fourth period (2021–2023) indicates a period of recovery. The general upward trend in BB is predominantly attributed to cropland, possibly due to increased grain production resulting in more straw. This general upward trend may also be linked to socio-economic and environmental factors. Fang et al. [64] discovered a positive correlation between CRB and the level of economic development. As economic conditions improve, farmers are less concerned with the relatively low marginal benefits of soil enhancement from straw recycling or its value addition, while the short-term economic benefits of burning straw for rapid field clearance become more appealing. Zhang et al. [65] have studied the changing environment of SNP cropland, revealing that increased temperature and precipitation have positively influenced crop cultivation. Chen et al. [66] found that the warming climate has led to increased yields of rice, corn, and soybeans in NEC. Correspondingly, the increase in crop cultivation leads to an increase in straw. In the first period, the gradual rise in BB could be linked to the yearly increase in the proportion of grain crops planted [67]. In the second period, it might be associated with factors such as the improvement of farmers’ living standards and the aging of the agricultural workforce [64,68]. This could have potentially led to a decreased need, ability, and willingness to retain straw for purposes such as fuel and feed, resulting in a significant amount of straw being burned directly in the fields. As for the third period, there was a sharp decline, possibly due to intensified enforcement of straw burning bans [69]. The fourth period may be related to the relaxation of policies in some regions. Overall, the trends reflected by FPC and FRP are generally consistent, except for 2006, where FPC exhibited a distinct peak (possibly due to the occurrence of three forest fires in HLJ), resulting in FPC presenting a trough while FRP peaked [70,71]. The spatial distributions of the 20-year FPC and FRP averages are depicted in Figure 2c,d, respectively. The areas of high BB values are concentrated in the SNP, SJP, and LHP, as well as in the transition zone between the northern part of the SNP and the GKM and LKM. These regions are the main grain-producing areas in China [72,73], with significant associated crop residue burning (CRB) [74]. Conversely, the low-value areas of BB are concentrated in the high-altitude areas of the GKM, LKM, and CBM. To further explore the spatial changes of BB in the NEC region over the 20-year period, MK trend tests were conducted on the FPC and FRP at each grid point, as depicted in Figure 2e,f. The trend results are classified into five categories: significant increase, non-significant increase, unchanged, significant decrease, and non-significant decrease (as detailed in Table S2). Regions exhibiting a significant upward trend in BB are primarily concentrated in most areas of the SNP, the LHP, and the southwestern part of the SJP. In contrast, the significant decrease areas are mainly concentrated in the SJP and the transition zone between the northern part of the SNP and the GKM and LKM.

To further explore the land-cover types leading to BB changes in NEC, the MK trend results were statistically analyzed. The proportions of various trends in FPC and FRP, respectively, are shown across different land-cover types in Figure 3a,b, respectively. The results showed that the change trends were mainly concentrated in cropland and urban areas, and were dominated by increases. The proportion of increasing trends in FPC reached 61.7% for cropland and 63.3% for urban areas, while for FRP, these proportions were 62.8% for cropland and 67.2% for urban areas. In particular, the trend of increasing BB was more significant in cropland and urban areas (p < 0.05), as shown in Figure S2. Furthermore, cluster analysis was conducted on the significant increase and decrease points of BB in cropland and urban areas. Figure 3c presents the results of the BB clustering analysis, revealing three significantly increasing BB clusters (BBI: ①, ②, and ③) and two significantly decreasing BB clusters (BBD: ④ and ⑤). The largest significantly increasing clustering area, ①, encompassed most of the SNP with an area of approximately ${\mathrm{68,634.9}\mathrm{km}}^2$, followed by areas of $\mathrm{18,246.2}{\mathrm{km}}^2$ (②, located in the hinterland of the LHP) and $\mathrm{16,150.9}{\mathrm{km}}^2$ (③, located in the southwestern part of the SJP). The two significantly decreasing clustering areas were situated in the hinterland of the SJP (④) and the transition zone between the northern part of the SNP and the GKM and LKM (⑤), with areas of ${\mathrm{48,099.1}\mathrm{km}}^2$ and ${\mathrm{51,764.7}\mathrm{km}}^2$, respectively. Region ① is a main agricultural hub in NEC, including two of the three provincial capitals: Harbin and Changchun [65]. It has a high population density, possibly leading to frequent BB activities [75]. Region ② is located in the hinterland of the LHP, characterized by a high population density and relatively scarce per capita cropland [76]. As a result, a portion of the crop residues is used domestically, which may lead to a lower density of fire points. Regions ③ and ④ are both part of the SJP, but Region ④ has a less habitable environment and is further from the provincial capital cities. This may have led to more severe population loss, reducing the effective cultivated area [77]. Region ⑤ is a high-risk area for forest fires. In NEC, fire prevention policies have become stricter [78].

3.2. Seasonal Pattern

BB in NEC is primarily concentrated during the spring and autumn, specifically before spring planting and after autumn harvesting, with spring being the most intense (Figure 4), as the region tends to do a lot of CRB in these seasons. In spring and autumn, high BB concentrations are mainly concentrated in four regions: the SNP, SJP, LHP, and the transition zone between the northern part of the SNP and the GKM and LKM. Meanwhile, in summer and winter, the spatial clustering of BB is relatively dispersed. Sun et al. [79] also revealed that the high FPC was mainly concentrated in the SNP in the spring. In autumn, there is a noticeable absence of BB aggregation in the southern part of the SJP, which is present during spring. The clustering of BB in the transition zone between the northern part of the SNP and the GKM and LKM exhibits a more pronounced clustering in autumn. Figure 5 illustrates the monthly and seasonal statistical results of the FPC and FRP, respectively. In particular, March–April and October–November are two critical periods contributing the most to BB, accounting for 80.0% and 82.6% of the annual proportion of FPC and FRP, respectively. During these four months, FPC consistently exceeds 2000, while FRP remains above 50,000 MW. The highest BB occurs in April, with FPC and FRP reaching 7753 and 20,000 MW, respectively, accounting for 31.5% and 32.2%. Previous studies have also found that BB in the NEC region is mainly concentrated in March–April and October–November [21]. In addition, BB in the NEC region is primarily driven by croplands across the four seasons. Cropland has the highest proportion (FPC total = 71%) of BB in

Table 2. Seasonal statistics of annual average fire pixel counts (FPC) and fire radiative power (FRP) in northeast China (NEC). The red font indicates the proportion of the dominant land-cover type in the total BB.

Category	Season	Region	Forest		Cropland		Urban		Others
			Value	%	Value	%	Value	%	Value	%
FPC	Spring	LN	182	15.3	772	64.7	212	17.8	26	2.2
		JL	539	18.2	2361	79.6	57	1.9	10	0.3
		HLJ	1950	22.4	6511	74.7	162	1.9	99	1.1
		Total	2671	20.7	9644	74.9	432	3.4	135	1.0
	Summer	LN	56	11.5	189	38.8	226	46.5	16	3.2
		JL	68	42.4	79	49.5	13	8.1	0	0.0
		HLJ	360	36.1	548	55.0	78	7.8	10	1.0
		Total	483	29.4	816	49.7	317	19.3	26	1.6
	Autumn	LN	64	6.3	700	68.4	248	24.3	11	1.0
		JL	203	14.3	1186	83.5	25	1.8	5	0.4
		HLJ	1938	32.7	3867	65.2	83	1.4	38	0.6
		Total	2206	26.4	5753	68.7	357	4.3	54	0.6
	Winter	LN	48	10.4	342	74.1	51	11.0	21	4.5
		JL	28	18.1	126	80.7	1	0.7	1	0.4
		HLJ	74	22.0	256	76.4	3	0.7	3	0.9
		Total	150	15.7	724	76.0	55	5.7	24	2.6
	Total	LN	351	11.1	2004	63.3	738	23.3	73	2.3
		JL	838	17.8	3752	79.8	97	2.1	16	0.3
		HLJ	4321	27.0	11,182	70.0	326	2.0	149	0.9
		Total	5511	23.1	16,938	71.0	1160	4.9	238	1.0
FRP ($\times {10}^3\mathrm{MW}$)	Spring	LN	5.3	22.3	14.7	62.3	3.0	12.7	0.6	2.7
		JL	12.2	18.1	54.0	80.0	1.2	1.7	0.1	0.2
		HLJ	78.0	31.6	162.4	65.7	3.3	1.3	3.3	1.3
		Total	95.5	28.2	231.1	68.4	7.5	2.2	4.0	1.2
	Summer	LN	0.7	12.3	2.1	35.2	2.9	48.2	0.3	4.4
		JL	0.8	47.6	0.8	45.2	0.1	7.2	0.0	0.0
		HLJ	7.7	44.9	8.3	48.6	0.9	5.4	0.2	1.0
		Total	9.2	37.2	11.2	45.1	3.9	15.9	0.4	1.8
	Autumn	LN	1.0	6.5	10.9	70.6	3.4	21.7	0.2	1.1
		JL	3.0	13.7	18.5	84.6	0.3	1.4	0.1	0.3
		HLJ	90.1	49.9	87.8	48.7	1.3	0.7	1.2	0.7
		Total	94.1	43.2	117.2	53.8	4.9	2.3	1.5	0.7
	Winter	LN	1.9	13.7	9.8	71.1	1.0	7.2	1.1	8.0
		JL	0.6	19.1	2.6	80.0	0.0	0.5	0.0	0.5
		HLJ	2.1	21.9	7.2	76.7	0.0	0.4	0.1	1.0
		Total	4.6	17.3	19.7	74.2	1.0	3.9	1.2	4.6
	Total	LN	8.9	15.1	37.5	63.8	10.2	17.4	2.2	3.7
		JL	16.6	17.6	75.9	80.5	1.6	1.7	0.2	0.2
		HLJ	177.9	39.2	265.7	58.5	5.5	1.2	4.8	1.1
		Total	203.4	33.5	379.1	62.4	17.4	2.9	7.2	1.2

. Ke et al. [80] also indicated that cropland dominates BB in NEC, surpassing 70% from 2013 to 2017.

To help government departments understand the situation of BB in the NEC region and further develop a reasonable BB management program, the BB statistics in the NEC region according to the three provinces are detailed in Table 2 and Figure S3. It is worth noting that compared to the other two provinces, LN has a relatively high winter BB (Table 2), especially in urban areas, where the FPC and FRP proportions both exceed 90%. This may be attributed to emissions from factories. Particularly, a different situation was observed in LN Province, where the peak of BB occurs earlier, being concentrated in spring from February to April, rather than from March to April (see Figure S3). This difference could be attributed to LN’s location in the southern part of NEC, leading to earlier spring plowing. Additionally, forest BB in summer and autumn in HLJ and JL provinces is also noteworthy. Additionally, forest BB during autumn in HLJ province and summer in JL province warrants attention, as the forest FRP exceeds that of cropland.

3.3. Vertical Spatial Distribution Characteristics

BB is a large source of atmospheric particulate matter, contributing to a majority of primary organic aerosols [81,82]. Therefore, analysis of the aerosol extinction coefficient (AEC) contributes to understanding the vertical distribution characteristics of BB. Considering Tobler’s first law of geography [83], we compared two typical areas that are relatively close geographically and both have the cropland type, but which present very different BB trends (i.e., ① with significant BBI and ④ with significant BBD in Figure 3c), aiming to minimize the influence of factors other than BB.

The seasonal average AEC profiles of the above two regions (BBI and BBD) from 2013 to 2023 are shown in Figure 6. Between 150 m and the planetary boundary layer height (PBLH), the AEC in the BBI is larger than that in the BBD region for all four seasons, and is the most significant in the spring. In spring, the maximum inflection point of the AEC profile is mainly concentrated below the PBLH about 800 m, such that the BB has a relatively large impact on the near-surface air quality. However, the BB is also transported vertically with the atmosphere up to 1500 m, which suggests that the BBI area in spring may also affect the air quality in the surrounding areas. During the summer season, despite weakened BB activities, the AEC of BBI still exceeds that of BBD (150 m—PBLH range). In autumn and winter, coal-fired heating has a significant effect on AEC in NEC [84,85,86]. The AEC was more significant in the vertical 0–150 m range in autumn in the BBD region compared to the BBI region, which may be due to earlier heating in the BBD region. During the peak heating period in winter, the higher population density in the BBI area inevitably increases the amount of coal burned, resulting in a significantly larger near-surface AEC than in the BBD area.

Considering that AEC is highly influenced by summer meteorological conditions and is affected by autumn and winter heating, the spring season—when BB is most severe—was chosen to study interannual variability. We chose the period starting from 2013, when BB experienced rapid growth (as mentioned in Section 3.1), spanning from the second to the fourth periods (2013–2023). Figure 7 illustrates the interannual variability of AEC in areas with significant BBI and BBD. Below the PBLH (<800 m), the AEC of the BBI area ranges from 0.18 to 0.37, while that of the BBD area ranges from 0.05 to 0.36. Furthermore, compared to the BBD region, the AEC in the BBI region is larger in most years, especially in the past two years (2022–2023). Despite vertical transport of BB smoke at high altitude, the main impacts were concentrated below 1 km of the PBLH in both regions, and so, the near-surface air quality in the NEC region is likely to be more severely affected by BB. In particular, the PBLHs are higher in all years in the BBI regions than in the BBD regions, which may be attributed to the fact that the heat generated by large amounts of BB leads to a localized increase in temperature, which increases the instability of the atmosphere and, consequently, the height of the PBLH [87].

3.4. The Spatial Autocorrelation of BB within the County

The spatial distribution of BB in the NEC region was identified in Section 3.1, but whether there is spatial autocorrelation of BB in neighboring areas (possibly due to policies, environment, human habits, and so on) remains to be further studied. In China, county-level environmental departments are responsible for on-site environmental law enforcement, and the Chinese government initiated vertical environmental management reform in 2016 to enhance the independence of local enforcement [88]. Therefore, spatial autocorrelation analysis was conducted separately for FPC and FRP at the county level in NEC.

The global Moran’s index results indicated that BB in NEC exhibits a positive spatial autocorrelation. Figure S4 displays the results of the global Moran’s index for FPC and FRP in NEC from 2004 to 2023. The global Moran’s index of FPC ranges from 0.31 to 0.55, with an average of 0.45, and all p-values are less than 0.01. Similarly, the global Moran’s index of FRP ranges from 0.23 to 0.56, with an average value of 0.45, and all p-values are less than 0.01 as well. This consistency in global Moran’s index of FPC and FRP indicates the significant clustering characteristics of the BB spatial distribution. Wang et al. [74] also found that the spatial patterns of CRB in NEC present a significant spatial autocorrelation. To investigate the spatial autocorrelation clustering of BB, local Moran’s index analysis was conducted on the annual average FPC and FRP, with the results shown in Figure 8. Both the local indicators of spatial association (LISA) clustering maps of FPC and FRP exhibit three types of clustering: high–high, low–high, and low–low, with similar distribution areas. The difference lies in the larger high–high clustering range in NEC indicated by FPC, attributed to FPC reflecting the confidence level of fire detection, which differs from the representation of fire intensity by FRP. High–high clusters are observed in three regions: the SNP, SJP, and the transition zone between the northern part of the SNP and the GKM and LKM. Low–low clusters mainly occur in the eastern and western parts of LN Province, as well as in the city of Yichun in HLJ Province. The low–high clusters are less frequent, with FPC indicating their presence at the northern edge of the SNP and FRP, and indicating them in the GKM. The inconsistency can be attributed to the autocorrelation analysis being conducted based on the 20-year average BB. From the county-level high–high clustering, it is evident that the SNP and SJP require attention from the government; especially the SNP, which necessitates joint prevention and control efforts between JL and HLJ provinces.

To further explore interannual variations, annual LISA cluster maps were generated for FPC and FRP from 2004 to 2023, as illustrated in Figure 9. The spatial autocorrelation in the SNP exhibits dramatic changes. From 2004 to 2017, the high–high cluster areas in the transition zone between the northern part of the SNP and the GKM and LKM gradually moved southward, reaching the center of SNP in 2017, at which time BB reached its peak. After 2018, high–high clustering gradually decreased from north to south and, by 2021, high–high clustering was only concentrated in the SNP hinterland. This southward shift may be linked to the migration of rural labor to urban areas, as rural labor in this region has shown a trend of moving to rapidly industrializing regions in pursuit of higher income and better employment opportunities [89,90]. However, it is noteworthy that, after 2021, there is a trend of clustering expanding northward again. The high–high clustering center has gradually emerged along the border of the HLJ and JL provinces, encompassing clustering areas that span both provinces. In contrast to the recent resurgence observed in the SNP, the spatial clustering of BB in the SJP has largely dissipated over the past four years, whereas prior to 2019, it consistently exhibited a pattern of high–high clustering. The disappearance of high–high clusters in the SJP may be related to the rapid population decline in the sparsely populated SJP in recent years. Table S3 presents the population data of the SNP and SJP from 2019 to 2022, while Figure S5 illustrates the population changes. SJP experienced a 14.58% decrease in population from 2019 to 2022, with an average annual decline rate of 5.12%. In contrast, SNP saw population decreases of 5.67% and 1.93%, respectively, over the same period. Moreover, the population of the SJP is only about 20% of that in the SNP, yet its area is nearly half that of SNP, indicating that the SJP population is more dispersed. Tobler’s First Law of Geography states that “everything is related to everything else, but near things are more related than distant things” [83]. Unlike the SJP, the recovery of BB autocorrelation in the SNP appears to be associated with the gradual lifting of COVID-19 restrictions. During the easing of these restrictions, the enforcement of straw burning bans also saw a gradual resurgence. However, the temptation to quickly clear land through straw burning and the deeply ingrained belief that ash benefits the soil led farmers to observe each other’s actions, making the densely populated SNP more likely to engage in straw burning simultaneously. This highlights the need for greater attention in the SNP, where JL and HLJ provinces need to collaborate in developing joint prevention and control measures for managing BB.

3.5. The Impact of BB on Atmospheric Environment

In addition to the spatial autocorrelation of BB, its influence on the atmospheric environment is also crucial. BB is one of the largest sources of many aerosols and trace gases in the Earth’s atmosphere [82,91,92], while the produced trace gases and aerosols show large variations in time and space [93]. Based on Andreae’s research on the emission of trace gases and aerosols from BB [91,94], we selected several significant and environmentally impactful ones as representatives, including AOD, O₃, CO, SO₂, CH₄, and CO₂, in order to investigate the impact of BB on the atmospheric environment in the NEC region. Therefore, a correlation analysis was conducted on the impact of FRP on trace gases and aerosols in the NEC region, with data selected from 2013 (when BB was rapidly increasing). The results are depicted in Figure 10.

The regions showing a positive correlation are primarily concentrated in the SNP, with BB impacting the six atmospheric constituents differently. The influence of BB on AOD is distributed across the SNP and SJP. Yu et al. [95] found that CRB highly contributed to a large-scale aerosol increase in NEC. The distribution of positive correlation between BB and CO is consistent with the significant increase region of BB (evident in three regions: the SNP, LHP, and the transition zone between the northern part of the SNP and the GKM and LKM). The correlation distribution between BB and CO₂ and CH₄ is quite similar, being primarily concentrated in the SNP. BB is also a significant source of greenhouse gases such as CH₄ and CO₂ [91,94]. Zhao et al. [23] has revealed that BB contributed to the greenhouse gases emissions of NEC in large quantities. The influence of BB on O₃ is widespread, scattered throughout the entire SNP. BB emissions contain O₃ precursors, and research has shown that BB has a significant impact on O₃ [96,97]. The relationship between BB and SO₂ appears to lack significant correlation. Ren et al. [98] have discovered that the contribution of SO₂ from BB exhibits major spatial disparities, with varying proportions in different regions. Seasonally, the most significant impact of BB on the six atmospheric constituents is observed in spring (as illustrated in Figure S6, which displays the distribution of correlation coefficients between BB and the six major pollutants for each season).

To substantiate the conclusion that BB significantly impacts the atmospheric environment in the NEC region, particularly influencing AOD and CO, this study integrated air quality data from ground monitoring stations in NEC, covering a period from 2015 to 2023. These data were utilized to evaluate the correlations between FRP and pollutants including PM_2.5, PM₁₀, and CO. The correlation coefficients are presented in

Table 3. Correlation between fire radiative power (FRP) and major pollutants at various ground monitoring stations.

Stations	Longitude (°E)	Latitude (°N)	PM2.5	PM10	CO
1760A	123.71	41.84	0.46	0.34	0.32
2213A	123.17	41.25	0.79	0.85	0.92
2219A	123.72	42.22	0.16	0.23	0.30
2235A	122.79	45.61	0.76	0.68	0.29
1130A	126.56	45.82	0.47	0.38	0.42
1780A	123.95	47.34	0.34	0.48	0.27
1781A	123.95	47.32	0.18	0.05	0.48
1788A	129.64	44.55	0.35	0.24	0.67
2258A	130.36	46.80	0.36	0.07	0.47
2263A	127.48	50.25	0.46	0.56	0.76

, while Figure S7 illustrates the time-series comparison between air quality parameters and FRP. The results show that the average correlation coefficients of satellite-inverted FRP with PM_2.5, PM₁₀, and CO provided by ground-based observation sites for 2015–2023 are 0.4, 0.4, and 0.5, respectively, which further reveals that BB has had an impact on the air quality in the NEC region over the years.

4. Discussion

Through an analysis of 20 years of FPC and FRP data in the NEC region, it was observed that BB exhibits a fluctuating but generally rising trend, which is consistent with previous research findings [21,25]. Due to the extended study period, we observed that after 2017, BB showed a declining trend, followed by a recovery. It is noteworthy that the recovery trend since 2021 is evident not only in the time-series of FPC and FRP, but also in the vertical distribution and regional autocorrelation of BB. The recovery is particularly pronounced in the hinterland of the SNP, with a tendency to expand northward again. In this region, cropland contributes the most to BB, with the SNP being the largest grain-producing area among the three major plains in NEC [99]. Seasonally, BB is most severe in spring and autumn, specifically in March–April and October–November. This has also been revealed in previous studies [74,80]. Before 2016, the most severe seasons for BB alternated between spring and autumn. Since 2016, however, the most severe seasons for BB have all occurred in spring (as detailed in Figure S8). Wang et al. [21] revealed that the amount of open BB in March–April was higher than that in October–November during 2016–2017. Analysis of the vertical distribution of BB also indicated that its impact is most severe in spring. Comparative analysis revealed that, in areas where BB exhibits a significant increase compared to those with a significant decrease, the height of AEC reaches up to about 1.5 km. Moreover, analysis of the impact of BB on the atmospheric environment also indicated that the correlation is highest in spring. This could result in pollutants emitted from BB in the SNP being more prone to high-altitude diffusion during spring, potentially affecting other areas. The long-range transport of BB emissions is not uncommon [10,100]. Yang et al. [11] have found that BB in South Asia plays a crucial role in causing heavy springtime air pollution over the Tibetan Plateau. Therefore, the future CRB in the SNP in spring is more worthy of attention. The high-altitude diffusion and potential long-range transport of BB from the SNP could affect the health of not only local people, but also those living in neighboring areas. BB generates black carbon and brown carbon, which can penetrate deep into the respiratory system, causing a range of health issues [1]. Karanasiou et al. [82] have highlighted that PM_2.5 and PM₁₀ originating from BB were associated with all-cause and cardiovascular mortality. Additionally, this study focused solely on the impact of BB on the atmospheric environment. However, BB also has broader and longer-term effects on the global climate through influencing atmospheric components and the carbon cycle [101]. In the future, more attention should be given to the broader and longer-term impacts of BB on the global climate through its influence on atmospheric components and the carbon cycle.

The spatial autocorrelation of BB in the NEC region is evident, primarily concentrated in croplands, which may be closely related to local government policies banning straw burning. The beginning of the intensive implementation of the straw burning ban occurred in 2013 [69,102]. After 2013, most regions in China witnessed a decline in CRB, while the NEC region presented a sharp upward trend [35]. Consequently, local governments enacted a series of policies to implement straw burning prohibition and comprehensive utilization, as detailed in Table S4. Combined with Figure S9, it is evident that years with low values of FRP and FPC correspond to the implementation of straw burning prohibition and comprehensive utilization policies. This indicates that local policies have played a significant role in reducing BB. Particularly in 2018, following the issuance of the State Council’s Three-Year Action Plan for Winning the Blue Sky Defense Battle, corresponding policies were introduced by the local government, leading to a 65.1% decrease in FPC and a 73.5% decrease in FRP compared to 2017. Yang et al. [103] have estimated that the burning ban enforced in 2018 have caused the PM_2.5 concentrations to decrease from the 2015 level by 67.10%, 53.23%, and 10.06% in the HLJ, JL, and LN provinces, respectively. Furthermore, the transition of the most severe BB seasons in the NEC region from alternating between spring and autumn to solely spring after 2016 may be related to government efforts to control haze pollution. Winter haze is most severe in NEC [27], where coal combustion, vehicle emissions, BB, and secondary inorganic aerosols have been identified as major contributors to PM_2.5 [104]. At the end of 2015, the Ministry of Environmental Protection initiated inspections for heavy pollution weather in the NEC region, with the prohibition of CRB being a significant component [105]. Therefore, the effectiveness of the local government’s BB control measures is quite evident. However, it is noteworthy that in the past three years, there has been a recovery trend—especially considering the expanding spatial autocorrelation of BB in the SNP. This not only highlights the importance of paying attention to the SNP, but also provides a technical basis for future policy adjustments. First, the geographic location of SNP necessitates joint prevention and control by the provinces of JL and HJL for effective policy implementation. Second, agricultural areas exhibiting a low–low clustering trend of BB spatial autocorrelation can draw on specific BB management experiences. Third, considering that the rapid decline of BB in the third phase was due to years of stringent policies, the impact of sustainable policy should be considered. Expanding comprehensive utilization of straw and subsidizing corresponding agricultural machinery (e.g., straw crushers) could be better approaches, as stricter policies might hinder economic development [64]. Compared to other regions in China, farmers in NEC require the lowest incentives (287 RMB per hectare) to shift from burning to retaining crop residues [106].

Although several fire emissions databases have been developed, including GFAS [107], GFED [108], and IS4FIRES [109], uncertainty in BB emissions persists due to factors such as location, time, and atmospheric conditions [110,111]. Based on satellite and reanalyzed data, we conducted an analysis of the impact of BB on the atmospheric environment in NEC. We discovered that BB has varying degrees of impact on AOD, CO, CO₂, CH₄, and O₃, with the most prominent correlation observed with CO. Zhang et al. [112] have shown that the interannual variation in atmospheric CO in Asia is sensitive to BB. This could be related to the fact that the lifespan of CO in the air is relatively long [113]. Due to its poor mixing in the troposphere, CO is regarded as a significant indicator of BB activities [114]. In this study, due to the utilization of data from multiple sources including satellite and reanalyzed data, there exists a potential for misinterpretation in comparative analysis due to varying data quality. We also attempted to utilize the commonly available GOSAT data to acquire CO₂ and CH₄. However, due to lower spatial resolution and data gaps within the study area, the results were unsatisfactory.

We acknowledge the inherent limitations of remote sensing data, including potential inaccuracies in fire detection. Giglio et al. [28] have identified targeted improvements for the MODIS C6 active fire detection algorithm, noting a reduction in omission errors by approximately 1.2% for large fires. This enhancement led to relatively effective MODIS fire detections in Central Asia (including NEC) out of 14 global regions, with a commission error rate of about 4%. However, MODIS detections have higher omission rates and lower fire radiative power in Northeastern Asia, compared to VIIRS [115]. Unfortunately, as VIIRS data have only been available since 2012, MODIS fire products have been extensively used in China for long-term, high-quality estimates due to their long temporal coverage [20,35,116,117,118]. Therefore, the MODIS fire product for the years 2004–2023 were selected for the study of BB in this paper. In the future, we will fuse multi-source remote sensing data, such as Landsat or Sentinel series, to help us perform indirect validation of MODIS fire products. In addition, we should aim to improve data accuracy through incorporating ground-based observations and high-resolution satellite data to enhance the accuracy and resolution of BB monitoring. Additionally, applying the methods and findings of this study to other similar regions will help to expand the geographic scope and allow for comparisons of BB patterns and their impacts at different scales. Moreover, more attention should be given to the broader and longer-term impacts of BB on the global climate through its influence on atmospheric components and the carbon cycle.

5. Conclusions

This study used satellite and reanalysis data to investigate the spatiotemporal distribution of BB in the NEC region over a 20-year period (2004–2023), exploring its spatial autocorrelation and impact on the atmospheric environment. Notably, BB presented a fluctuating but overall increasing trend, which was especially pronounced in key agricultural areas such as the SNP, SJP, and LHP. Seasonal analysis highlighted that BB peaks during the spring and autumn in a manner closely linked with the region’s agricultural practices, particularly CRB.

In terms of vertical spatial distribution, the impact of BB on atmospheric aerosols was more pronounced in areas where BB is increasing, especially in the spring. The analysis of AEC revealed significantly higher concentrations of atmospheric particulates below the PBLH in these regions, indicating a substantial impact on near-surface air quality. Additionally, the vertical transport of smoke and pollutants suggests that springtime BB not only affects local air quality, but may potentially impact surrounding areas as well.

County-level spatial autocorrelation analysis further revealed significant clustering in the distribution of BB. This spatial autocorrelation suggests that not only do natural factors (e.g., wind direction and topography) play crucial roles in shaping the distribution of BB, but that local policies, environmental conditions, human activities, and inter-regional interactions do too. Notably, in recent years, there has been a resurgence trend in the high–high clustering of BB in the SNP, highlighting the necessity for joint prevention and control efforts between JL and HLJ provinces.

Moreover, the study on the impact of BB on the atmospheric environment revealed a complex relationship between BB activities and air quality. BB is a major source of various aerosols and trace gases in the atmosphere, with its impact varying across different regions and seasons. In particular, BB was found to significantly influence atmospheric components in spring, especially AOD and CO. The emissions of these gases and aerosols from BB activities not only affect air quality locally, but can also have broader regional implications.

Overall, the intricate nature of BB in NEC and its significant impact on air quality underscore the importance of implementing effective environmental management strategies and policies to mitigate the adverse effects of BB. Joint prevention and control efforts in various regions should garner government attention. These measures are vital for improving air quality and protecting public health, not only in NEC but also in other regions experiencing similar BB activities.