Introduction

PM2.5 are tiny suspended particulates with an aerodynamic diameter of less than or equal to 2.5 μm. These generally comprise both poisonous and harmful substances and are difficult to filter. Therefore PM2.5 particles have the ability to stay in the atmosphere for a long period of time and can be transmitted over large distances (Chung et al. 2014; Hossain and Easa 2012). Once inhaled, PM2.5 may cause significant harm to human health, and are one of the major precipitants that both trigger and exacerbate lung and heart diseases in cities across the world (Abbey et al. 1995; Boldo et al. 2006; Che et al. 2015; Fan et al. 2009; Franklin et al. 2007).

In previous studies, factors influencing the concentration of PM2.5 have been generally divided into two categories: natural factors, e.g., weather conditions (Hsu and Cheng 2016; Wang et al. 2013), and human factors. Examples of the latter include the emissions of PM2.5 into the atmosphere through coal-fired power plants and vehicle exhaust pipes. PM2.5 emissions is a particularly serious issue in some densely populated cities since there tend to be both a large number of motor vehicles and significant combustion of coal and other fuels for heating homes in winter (Baker and Scheff 2007; Caggiano et al. 2011; Kulshrestha et al. 2009; Tan et al. 2015; Wang et al. 2012). In response to this situation, the mode of central heating was altered from traditional boiler-type rooms per district to co-generation of heat and power in some cities or industrial parks. Such changes have resulted in simultaneous provision of electricity and heat energy, which not only saves energy but also greatly reduces emissions of PM2.5.

Besides the aforementioned influencing factors, studies have shown that PM2.5 concentrations can be affected by forests through the physical blockage of atmospheric particulates and the absorption of particulate matter through stomata and lenticels of stems and branches. Furthermore, forests can positively affect air temperatures, relative humidity and wind speeds directly or indirectly and other meteorological conditions of the flora microclimate (Becker et al. 2000; Escobedo and Nowak 2009; Escobedo et al. 2008).

Forests play a valuable role in both reducing PM2.5 concentrations in cities and improving urban air quality. Thus, forest cover could be associated with changes in PM2.5 concentrations, and there may be a relationship between PM2.5 concentrations and other factors. Currently, PM2.5 studies mainly focus on the physical and chemical composition of these small particulates (Baker and Scheff 2007; Dan et al. 2004; Khan et al. 2016; Perrone et al. 2012; Sun et al. 2016), or on the impact of basic meteorological factors, other air pollutants, and automobile exhaust emissions on the concentrations of PM2.5. The latter has mainly been analyzed by assessing variation of the PM2.5 concentration in a city or a region where atmospheric pollution is considered to be serious, such as the Jing-Jin-Ji (Beijing, Tianjin and Hebei) region, China, southeast Queensland, Australia, and the northeast and Midwest of the USA as well as San Francisco Bay. A previous study showed that the concentration of PM2.5 in these areas varied by season and geography, and that these regions were characterized by dense populations, high energy consumption, and serious air pollution (Beaver et al. 2010; Ding et al. 2008; Wegesser et al. 2009; Yeganeh et al. 2017). In addition, many studies have demonstrated that the meteorological environment is very complex (Zhang et al. 2015). Meteorological conditions can largely diffuse or accumulate air pollutants (Tai et al. 2010; Yang et al. 2011), which makes the relationship between each meteorological factor and PM2.5 concentration complicated. In fact, Yin et al. (2016) emphasized that the relationship between wind speed and PM concentration was not a simple positive or negative correlation. Indeed, they found in a Beijing study PM2.5 concentrations and daily maximum wind speed exhibited a U-shaped relationship, while concentrations had a nonlinear relationship with PM10 and average air humidity. Analysis of meteorological data for Nagasaki, Japan showed that precipitation was negatively correlated with PM2.5, while temperature was positively correlated (Wang and Ogawa 2015). Studies have also confirmed that PM2.5 concentration and average wind speed have a nonlinear relationship, but there were linear relationships with single related factors such as CO (carbon monoxide) or NO2 (nitrogen dioxide) (Ni et al. 2017). In summary, no studies to date have reported a relationship between forest cover and PM2.5, and the main factors that can affect PM2.5 concentrations. Previous studies have been limited by an analysis of only one city, or have only studied the relationship between one factor and PM 2.5. The objectives of this study were (1) to identify the distribution characteristics of PM2.5 concentrations in Heilongjiang Province; (2) to confirm the relationship between forest cover and PM2.5 concentrations; and, (3) to identify the main factors which affect PM2.5 concentrations in Heilongjiang Province.

Materials and method

Study area

The study was conducted in Heilongjiang Province (E121°11′–135°05′, N43°26′–53°33′), the northernmost province of China and covers 12 cities and the Greater Khingan Mountains area (Fig. 1). It has a forest cover of 46.1%. The main climate characteristics of this Province are dry springs of low temperatures, prolonged cold winters, a short frost-free period. In cold periods the people mainly heat with coal.

Fig. 1
figure 1

Distribution of sampled cities in Heilongjiang Province, China

Full size image

Data

For this study, 8760 measurements of the daily average PM2.5 concentrations were obtained from Heilongjiang Environmental Protection Bureau from January 1, 2015 to December 31, 2016. A further 1065 averaged measurements of other air pollutants (PM10, SO2, CO and NO2 and O3) were also collected for a total of 9825 measurements. Daily average data for average wind speeds, sunshine duration, average atmospheric pressures, mean temperatures, and average relative humidity were collected in 12 cities from October 1, 2015 to April 30, 2016 from http://data.cma.cn and 1065 values were obtained. We also gathered data on populations, land areas, meteorological conditions, heated areas and motor vehicle numbers in each city from the Annals of Statistics of Heilongjiang Province.

Air quality rating standard

According to the State Environmental Protection Standards of the People’s Republic of China: Technical Regulation on Ambient Air Quality Index (HJ633-2012), the relationships between daily average PM2.5 concentration and air quality levels are shown in Table 1.

Table 1 Corresponding air quality levels of daily average PM2.5 concentrations
Full size table

Results and discussion

Analysis of daily average PM2.5 concentrations

PM2.5 concentrations are influenced by many factors and high concentrations mainly develop in winter and spring. According to the climate characteristics of Heilongjiang Province, a cycle is defined by a non-heating period May 1st to September 30th and a period when artificial heating is used October 1st to April 30th. The daily average PM2.5 concentrations during heating and non-heating periods in these 12 cities in 2015 and 2016 as of December 31st, 2016 are shown in Fig. 2.

Fig. 2
figure 2

Daily average concentrations of PM2.5 in the 12 cities, which are Harbin (a), Daqing (b), Suihua (c), Qitaihe (d), Shuangyashan (e), Mudanjiang (f), Hegang (g), Yichun (h), Heihe (i), Jiamusi (j), Qiqihar (k), and Jixi (l)

Full size image

For the two non-heating periods, Fig. 2 shows that the daily average concentrations of PM2.5 in Jiamusi, Mudanjiang, Qitaihe, Suihua, Yichun, Jixi, Qiqihar, and Shuangyashan were less than 75 μg/m3. Figure 2a shows that Harbin had 4 days in 2015 when the daily average PM2.5 concentrations slightly exceeded 75 μg/m3 and PM2.5 concentrations of this city were less than 75 μg/m3 in 2016. Figure 2i shows that there were 2 days in May of both years in Heihe when daily average PM2.5 concentrations exceeded 75 μg/m3. Figure 2b shows that daily average PM2.5 concentrations in Daqing were higher than 75 μg/m3 in both years (2 and 3 days, respectively). Finally, Fig. 2g indicates that the daily average PM2.5 concentrations in Hegang did not affect air quality except 3 days.

Figure 2 also demonstrates that increasing fossil fuel consumption by heating led to increases in pollution sources, especially after November when it became cold. The increasing heating demand and large amounts of coal burning led to a daily average concentration of PM2.5 in 12 cities that sometimes exceeded 75 μg/m3, though specific levels differed by city. In March and April, when the weather turned warmer and heating demand decreased, PM2.5 concentrations concomitantly declined on a gradual basis. Data pertaining to the daily average PM2.5 concentrations in the heating period of 2015 were collected and summarized in Fig. 3. These are in accord with specified standards of the daily average PM2.5 concentration (done by converting the daily average concentration standard of air quality index to PM2.5 daily average concentration standards) in the State Environmental Protection Standards of the People’s Republic of China: Technical Regulation on Ambient Air Quality Index (HJ633-2012). Figure 3 shows that a large proportion of Harbin’s daily average PM2.5 concentrations were at light and medium levels, while heavy pollution also accounted for a large share. Harbin was followed by Mudanjiang, Qiqihar, Qitaihe, Daqing and Suihua with respect to pollution levels.

Fig. 3
figure 3

Proportion of light, medium and heavy pollution days for 12 cities in the heating period of 2015

Full size image

We obtained the following results from the analysis: (1) during the non-heating period, 4 of the 12 cities had daily average PM2.5 concentrations that exceeded 75 μg/m3 but only for a period of 1–4 days. Almost all the other daily average concentrations were considered to be good or excellent. Therefore, we concluded that variables associated with human activities did not increase the daily average PM2.5 concentrations during the non-heating period or affected pollution levels. (2) During the heating period, demand increased significantly which is attributable to human activities and the daily average PM2.5 concentrations often reached a standard of pollution.

Analysis of the relationship between forestry cover and PM2.5 concentrations

Previous studies have confirmed that air quality of many cities and regions (Los Angeles, Beijing, and Santiago Metropolitan Region) can be improved by urban forest cover through eliminating air pollutants and changing microclimates (Escobedo and Nowak 2009; Powe and Willis 2004; Yang et al. 2005). However, little research has focused on the relationship between urban air quality and forests of Heilongjiang province, which has the largest area of forest cover in China at approximately 46.1%. Differences in populations, land area, meteorological conditions, heated areas and motor vehicle levels in different cities cause PM2.5 concentration levels to vary (Hsu and Cheng 2016; Majumdar et al. 2015; Palencia et al. 2016). In the present study, according to forestry cover, relative heating areas, motor vehicle levels (Table 2) and heating method, the 12 cities, can be divided into four groups based on the monthly average PM2.5 concentration data from May 1st 2015 to April 30th 2016 (Table 2; Fig. 4).

Table 2 Forestry coverage rate, relative heating area, and motor vehicle population in 12 cities
Full size table
Fig. 4
figure 4

The monthly average PM2.5 concentration of four groups from May 1st 2015 to April 30th 2016

Full size image

Cities in the first group include Yichun, Heihe, and Hegang and all had a high forest cover. There were relatively small heating areas in Yichun and Heihe, while Hegang did not have large motor vehicle numbers and therefore there were fewer air pollutants due to human factors. Consequently, the average PM2.5 concentration in these cities in one 12-month period was low and the air considered to be of good or excellent quality. Cities in the second group included Qiqihar, Daqing, and Suihua with the lowest forest cover in the Province (Suihua > Daqing > Qiqihar). Daqing and Qiqihar had relatively large heating areas during the heating period while Qiqihar and Suihua had large motor vehicle levels. Consequently, due to more air pollutants caused by human behavior, the monthly average PM2.5 concentrations in November in Daqing and Suihua were within light pollution levels, and Qiqihar was also lightly polluted in November and December (pollution levels were ranked as Qiqihar > Daqing > Suihua). The third group included Harbin, Mudanjiang, and Qitaihe. Although their forest cover was high (Mudanjiang > Qitaihe > Harbin), Harbin and Qitaihe had relatively large heated areas and all three cities had large motor vehicle numbers. Consequently, due to pollutants caused by human activities, the three cities were moderately polluted in November and December (Harbin > Qitaihe > Mudanjiang, which was in contrast to forest cover ranking). However, the PM2.5 concentration data for Harbin was considered to be approaching heavily polluted levels. The fourth group of cities included Jiamusi, Jixi, and Shuangyashan and had low forest cover, at least below the provincial average. However, the heating area in these three cities was relatively small, while thermoelectric networking was used for heating (http://www.xinhuanet.com) in both Jiamusi and Jixi. However, the population of Shuangyashan did not have a large number of motor vehicles and the monthly average PM2.5 concentration was within the scope of being good to excellent in one 12-month cycle.

We can draw the following conclusions from the analysis. A higher forest cover was associated with lower concentrations of PM2.5. However, if society was to discharge pollutants into the atmosphere without any limits, this would result in an increase in PM2.5 concentrations that would reach medium or even heavy pollution levels. In such instances, forest cover would have a limited role in mitigating such high levels of PM2.5. However, our data did suggest that higher forest cover resulted in lower PM2.5 concentrations. In contrast, a low forest cover and significant human activities usually results in high PM2.5 concentrations that reach appreciable pollution levels. When forest cover is low, it may be possible to restrain human discharge of pollutants into the atmosphere which can help control PM2.5 increases.

Determination of the factors affecting PM2.5 concentrations

The formation of PM2.5 is a complex process and previous studies have only focused on PM2.5 concentrations and their association with a single factor, with data lacking on the multi-factorial analysis of PM2.5 concentrations. Furthermore, it is not possible using this type of approach to determine which factor is the most important in influencing PM2.5 concentrations. Traditional classifications and regression statistical models have been used to select the importance of indirect assessment variables through statistical significance and Akaike information criterion (AIC), which resulted in some important variables being incorrectly deleted. As a modern regression method, the Random Forest (RF) methodology has been applied to the complex situation of high-dimensional variables where there are complex relationships. The importance of this method can be extended to each variable where important variables can be identified and be prevented from being deleted (Baccini et al. 2008; Cutler et al. 2007; Prasad et al. 2006; Xie et al. 2009). In this study, based on the theory and merits of a Random Forest model, forest cover, land area, heating area and the number of vehicles were regarded as a constant during the period of 2015. We then determined the important variables affecting the concentration of PM2.5 by data analysis for meteorological factors and air pollutants from the heating periods in 2015. Since both meteorological factors and air pollutants can affect PM2.5 concentrations, and these factors also interact with each other in a complex relationship, it is difficult to employ classical regression to confirm the main factors that are most important in determining PM2.5 levels. This paper adopted the modern regression method of RF and assessed forest cover, land area, heating area and motor vehicle levels which were considered to be constant in one 12-month cycle of 2015. Table 3 shows important variable values from high to low (two decimal fractions were used) in which X1 to X10 represents average wind speed (X1), sunshine duration (X2), average atmospheric pressure (X3), average temperature (X4), average relative humidity (X5), PM10 (X6), SO2 (X7), CO (X8), NO2 (X9) and O3 (X10), respectively. During the cooler months in 2015, cities with high PM2.5 concentrations, such as Harbin, Mudanjiang, Qiqihar, Qitaihe, Daqing, and Suihua, were significantly influenced by several factors that affected PM2.5 concentrations. These air pollutants were PM10, SO2, NO2, and CO and all meteorological factors were ranked after these four pollutants. Although different for each city, the average wind speeds and relative humidity were relatively important. We conclude that the air pollutants PM10, SO2, NO2, and CO were the main influencing factors for changes to PM2.5 concentration levels, and any increase in these pollutants was strongly associated with PM2.5 increases. If forest cover remains unchanged, it is imperative that there is control of population-driven pollution so that there is a concomitant reduction in PM2.5 concentrations in the atmosphere.

Table 3 Importance rank of air pollutants and meteorological factors in 6 cities (10 factors in total)
Full size table

Conclusions

The daily average PM2.5 concentration of 75 μg/m3 was a critical value in Heilongjiang Province. Human activities did not result in concentrations that exceeded 75 μg/m3 during the warm months. However, the daily average PM2.5 concentrations during the cooler months (heating periods) were significantly affected by human activities. Concentrations of PM2.5 were unaffected when cities had high forest cover and PM2.5 levels were always at lower concentrations. Forests play a positive role in decreasing the concentration of PM2.5. Fine particulate matter (PM10), sulfur dioxide, nitrogen dioxide and carbon monoxide were the main factors affecting PM2.5 concentrations, while meteorological factors and O3 levels were secondary. To reduce PM2.5 concentrations, it is necessary to control the discharge of pollutants, especially during the cooler months (heating period).