China has abundant forests and grasslands. According to the seventh national forest resource inventory, the forests and grasslands cover about 20.36% and 40.00% of the total land area (Ji et al., 2011; Wang et al., 2018). According to the climatic conditions and the province boundary, the study area was divided into seven regions: Northeast, North, Central, East, South, Southwest, and Northwest China (Figure 1). The largest region was Northwest China, contains the Ningxia Hui Autonomous Region, the Xinjiang Uygur Autonomous Region, Qinhai, Shaanxi and Gansu province, the total area is about 3.08 × 10⁶ km². The smallest region was South China, contains the Hong Kong Special Administrative Region, the Macao Special Administrative Region, Guangdong Guangxi and Hainan province. The division of the seven regions was commonly used in China. The inset is the border of the South China sea islands.

For different regions, South and Northeast China have higher forest coverage rates of 44.06% and 41.26%, respectively. The forest cover in Northwest China was the lowest (6.33%). The forest cover rate in Southwest, Central, East, and North China was 28.71%, 39.90%, 36.02%, and 17.51%, respectively. For grassland, Southwest and Northeast were also higher than other regions; the cover rate reached 44.03% and 39.29%. The lowest region was Central China, where the grass cover rate was a mere 5.52%.

A bottom-up method (Seiler and Crutzen, 1980) was employed to estimate the emissions caused by vegetation fire. The model is as follows: E = B A × F L × C E × E F (1) where BA is the burned area (km²), FL is the fuel loading factor (represented by biomass density, kt/km²), CE is combustion efficiency (represented by the proportion of burned biomass), and EF is the emission factor (represented by the mass of emissions per mass of dry biomass burned, g/kg).

As shown in Figure 2, the vegetation system in China was severely disturbed by fire. In total, the burned area reached 1.54 × 10⁶ hm² in the 5 years studied, with an annual average value of 3.08 × 10⁵ hm². The burned areas vary widely from year to year. The most and least severe years were 2005 and 2015, when they reached 5.04 × 10⁵ and 9.42 × 10⁴ hm², respectively. The burned area in other years was 4.68 × 10⁵ (2000), 2.99 × 10⁵ (2010), and 1.74 × 10⁵ hm² (2018).

Also, the burned area differs significantly among the regions (Figure 4 and Table 3). Among the seven regions, South China, Southwest China, and Northeast China were substantially impacted by fire. High vegetation coverage, climatic conditions, and frequent human activities were the main reasons for the high fire occurrence in these regions. South China had the largest burned area (mean = 9.11 × 10⁴ hm²), accounting for 29.59% of the total burned area. Southwest China was the second most affected area by fire. The mean burned area reached 2.15 × 10⁴ hm², accounting for 26.08% of the total burned area. Next in magnitude were the burned areas in Northeast China (mean = 4.79 × 10⁴ hm²) and East China (mean = 5.07 × 10⁴ hm²), accounting for 15.54% and 16.46%, respectively. The impact of fire on North and Central China was relatively small, resulting in 5.15% and 6.97% proportions, respectively. Overall, Northwest China had the smallest burned area of 637 hm² during the 5 years, contributing 0.21% of the total burned area.

Statistics have shown that the fire data and burned area from 2000 to 2005 were much higher than in other years, similar to our findings. The factors affecting fire occurrence are complex. Relevant studies showed that climate variables explained 37.1–43.5% of the fire occurrence variability, while human activities described 27.0–36.5% of variability (Wu et al., 2019).

As shown in Figure 5, forest fire is the primary fire type; the proportion was >60% during the study period. The largest and the least proportion of forest fire was in 2000 and 2010, accounted for 69.77% and 60.43% respectively. The average proportion of forest fire was 63.97%. The second is shrubland fire, the average proportion was 19.50% in the study period. The largest year was 2005, accounted for 22.26% of the total burned area. The least year was 2018, the proportion was 16.23%. Grassland fire had the least proportion with an average proportion of 16.54%. The largest and the least year of grassland fire was 2018 and 2000, accounted for 22.43% and 12.04% respectively.

Using the method in Section 2.2, eleven pollutant emissions were estimated at a spatial resolution of 30 m × 30 m. Figure 6 depicts the distribution of CO₂ in 2018. The spatial distribution and patterns of the emissions were similar to the burned area. However, the emission amount of each grid was different. Take CO₂ as an example, the value range in 2018 was 0.98–20.92 kt/km².

The total vegetation burning emissions of CO, CH₄, NO_x, non-methane VOC (NMVOC), SO₂, NH₃, PM_2.5, PM₁₀, OC, BC, and CO₂ are listed in Table 4. The 5-year cumulative emission reached 1.21 × 10⁵ Gg, averaging 2.42 × 10⁴ Gg. The highest (4.21 × 10⁴ Gg) and lowest (7.84 × 10³ Gg) emissions were recorded in 2000 and 2015, respectively, indicating a significant difference in the emissions after a decade. Whereas the emission amounts in 2000, 2010, and 2018 were 4.22 × 10⁴, 2.23 × 10⁴, and 1.08 × 10⁴ Gg, respectively. CO₂ was the most abundant pollutant from the fires, totaling 1.12 × 10⁵ Gg, which accounted for 92.46% of the total emissions. It was followed by CO (5.63 × 10³ Gg), accounting for 4.64%. The proportions of other pollutants were below 1% (Table 4).

The quantity of pollutant emissions in the various regions changed significantly (Table 5). The highest and lowest regions were South China (5.69 × 10⁴ Gg) and Northwest China (122.20 Gg), accounting for 46.85% and 0.10% of the total emissions, respectively. Southwest China is the second largest region with a total emission of 2.87 × 10⁴ Gg, accounting for 23.67% of the total emissions. The emissions of other regions were between 3.98 × 10³ and 1.57 × 10⁴ Gg, and the proportions were between 3.28% and 12.94%.

Of the various vegetation types, shrubland contributed the highest emissions (Figure 7). The average amount of emissions from shrubland was 1.29 × 10⁴ Gg in the studied 5 years, with an average proportion of 53.07%. Still on shrubland, South China contributed the highest emissions in the study period, accounting for 60.70%, 60.27%, 62.76%, 59.55%, and 43.30%, respectively. The second-largest contributor from shrubland fires was Southwest China, contributing 17.26%, 21.60%, 12.13%, 17.09%, and 28.39% for the aforementioned years.

Forest fire was the second-largest emission source throughout the study period in the study location (Figure 7). The average emission amount from forest fires was 1.04 × 10⁴ Gg and the average proportion was 42.17%. For the forest area, Southwest China was the largest emission source in 2005, 2010, 2015, and 2018, contributing 40.54%, 36.02%, 44.99%, and 45.43%, respectively. Northeast China was the second most significant source, and it was the largest emission source in 2000, accounting for 31.68% of the total emission. In 2005, and 2018, Northeast China was the second-largest source, and the proportions were 22.65% and 29.58, respectively. South China was also important in forest fire emissions. In 2010 and 2015, South China was the second-largest source, and the proportions were 19.53% and 28.88%, respectively.

Overall, grasslands contributed the least emissions (Figure 7). The average amount of grassland fire emissions was 0.98 × 10³ Gg in the study period, resulting in an average contribution of 4.76%. Northeast China was the largest emission source of grassland fires for the study period. They accounted for 57.10%, 29.28%, 53.97%, 26.06%, and 42.11% of the total emissions, respectively. Second on order was North China, responsible for 31.06%, 32.84%, 21.49%, and 12.46% in 2000, 2005, 2010, and 2018, respectively. Southwest China was also crucial to grassland fire emissions, accounting for 15.30%, 21.17%, 33.00%, and 29.16% in 2005, 2010, 2015, and 2018, respectively. For all the vegetation types, Northwest China was the lowest contributor during the study period.

Most studies have estimated fire emissions in China at a regional scale. Yang et al. calculated emissions of six pollutants (CO, CO₂, NO_X, CH₄, NMVOC, and PM_2.5) in the southern provinces of China from 2000 to 2016 (Yang et al., 2018). The result showed that South China’s total average annual pollutant emission was 1.61 × 10⁴ Gg, higher than observed in the current study (1.12 × 10⁴ Gg) (Figure 8). For each pollutant, our results were also lower than those of Yang et al.

Wei et al. estimated the average annual emissions of CO₂, CO, and CH₄ in the Heilongjiang province of China from 1953 to 2012 (Wei et al., 2014) as 3.15 × 10³, 177, and 10.5 Gg, respectively. Their results are similar to those of the current study, i.e., 2.72 × 10³, 171, and 8.1 Gg, respectively. Both compared studies were based on statistical calculations, while our result was based on satellite-derived data. There are some differences in the burned area, which could account for the discrepancy.

Elsewhere, Wang et al. estimated forest fire emissions in Southwest China from 2013 to 2017 based on MODIS burned area product (MCD64A1) (Wang et al., 2020). The result showed that the average emissions of CO₂, CO, and CH₄ were 1.42 × 10³, 91.66, and 4.52 Gg, respectively. Their result is much lower than our result; i.e., 5.26 × 10³, 297.25, and 12.94 Gg, respectively. The low spatial resolution of burned area data was majorly responsible for the difference. It caused the small fires (less than 500 m × 500 m) not considered in their study.

Fire is an important disturbance of vegetation in China. Three different fires, including forest fires, shrubland fires, and grassland fires were analyzed in the study. The result showed that forest fires had the largest burned area, exceeding 60% of the total burned area in each studied year. For different regions, South China was mostly influenced by fire, as identified by a previous study (Wu et al., 2019). South China is an essential agroforestry region in China where the forest borders the farmland. Agricultural fire has become an important cause of fire such as field fires for land reclamation. Furthermore, the main causes of fires were related to human activities in China (Zhong et al., 2003), the population density in South China is large which increasing the risk of human-caused fires.

Although forest fire-burned area accounted for a much larger proportion, shurbland fire was the main contributor to air pollution. The reason is that shrubland was much higher than forest for other factors in calculating fire emissions, including the following: 1) Except for Northeast and Northwest China, in other regions, the FL of shrubland is larger than forest. Especially in South China (with the largest burned area), the FL of shrubland is 17.93 kt/km², 2.33 times that of forest (7.71 kt/km²) (Qiu et al., 2016); 2) The CE of shrubland is 2.43 times that of forest, set with 0.68 and 0.28, respectively, according to the existed studies; 3) For most pollutant, the EF of shrubland is less than that of forest, such as CO, CH₄, NMVOC, etc. But for CO₂, which accounted for 92.43% of the total emissions, the EF of shrubland is larger than that of forest.

In this study, GABAM was employed to improve the accuracy of burned area estimation. However, the other three factors, set by conventional methods, need to be improved.

Fuels in natural fires are often composed of different vegetation and non-plant matter. Vegetation includes forests, shrubland, grassland, dry branches, and fallen leaves, while non-plant matter predominantly includes humus and peat (Bennett et al., 2017). Most studies on evaluating FL have typically set a biome-averaged value for each land cover type at a large scale (Wooster et al., 2005; Wiedinmyer et al., 2006; Padilla et al., 2015). But this method has just considered forests, shrubland, and grassland; other fuel types are not considered, which could underestimate the fire emissions.

Remote sensing is an effective tool for mapping FL. For a high-spatial resolution remote sensing, more influencing factors (such as meteorological and environmental factors) should be employed in the FL estimation model. Furthermore, the fuel model in China is inadequate; it is necessary to construct a reliable and practical fuel model based on the domestic fuel distribution in China, toward meeting the needs of remote sensing applications (Wu et al., 2016).

CE is affected by fire intensity, fuel type, fuel load, and meteorological factors, such as wind speed, relative humidity (De Santis et al., 2010). CE is crucial to estimate emissions from fires accurately. At present, in the research of fire emission estimation, CE is often set as an empirical value (Wooster et al., 2005; Guo et al., 2020; Shi et al., 2021), which could cause a potential error in fire emission estimation (Sá et al., 2005). Another method to obtain CE has two stages viz. the fire intensity should be extracted first, and then, the fixed CE value is adjusted according to the relationship between fire intensity and CE (Veraverbeke and Hook, 2013). Fire intensity can be measured by spectral indexes calculated via remote sensing, such as the difference in normalized burn ratio and normalized burn ratio. However, this method is mainly used on a local scale and is difficult to apply to a large scale (Wu et al., 2016). Hence, new methods to calculate CE on large scale should be developed to reduce the uncertainty in fire emission estimation.