Factors Influencing the Concentration of Negative Air Ions in Urban Forests of the Zhuyu Bay Scenic Area in Yangzhou, China

Abstract

Negative air ions (NAIs) are an important indicator of air cleanliness in an area, and play a vital role in promoting the psychological and physiological functions of the human body. However, there are few studies regarding the relationship between NAI concentration and various environmental factors in urban forests. Therefore, we established an observation point in the Zhuyu Bay Scenic Area in Yangzhou City and continuously measured concentration changes in NAIs for three years. At the same time, we also monitored 14 meteorological factors. A random forest model was used to determine the important environmental factors that affected changes in negative air ion concentrations from a non-comprehensive perspective, determine the prediction accuracy of the model, and screen out environmental factors that have a significant impact on negative air ions. The results showed that the environmental factor that NAIs were the most sensitive to in the Zhuyu Bay urban forest was humidity, followed by PM2.5, then wind direction, methane gas, and finally, temperature. Humidity was the most critical factor primarily because it directly affects the formation of NAIs in the environment and vegetation. We used big data to analyze the relationship between NAIs and environmental factors in forest parks. The results help deepen our understanding of NAIs characteristics and their application in urban forests.

1. Introduction

Negative air ions (NAIs) refer to air ions that are negatively charged after oxygen molecules in the air preferentially acquire excess electrons [1]. NAIs are widely distributed in the natural environment, especially in forests and wetlands [2]. NAIs are currently used as an important indicator to evaluate the ecological services of a given area, and are called the “air vitamin” [3]. The production and influencing factors of NAIs are complex and changeable; NAIs are primarily derived from the plant effect, the Lenard effect, radioactive substance induction, etc. [4,5]. Negative oxygen ions have good biological activity and are called “air vitamins and growth factors.” Studies have found that approximately 20% of negative air ions enter the human body through respiration, and approximately 80% of negative air ions are absorbed through human skin, and then produce a series of biological effects, playing a health care role. Negative air ions also have functions in sterilizing, reducing dust, and cleaning the atmosphere. Therefore, the concentration level of negative air ions is one of the important indicators for evaluating air quality [6,7,8,9,10]. Some scholars have defined the concentrations of negative air ions used for health care and therapeutic effects as 1000/cm³ and 8000/cm³ respectively [11,12].

In the late 1880s, German scientists discovered and confirmed the existence of negative air ions [13,14,15]; from the 1920s to the 1930s, German scientists discovered the impact of negative air ions on the human body and affirmed the biology of negative air ions [15,16]. V. P. Tikhonov’s 2004 study linked negative air ions to plant activity, affirmed the relationship between negative air ions and plants, explained that the primary generation site of negative air ions is at the leaf tip, and proposed that the production of negative air ions depends on weather conditions [17]. Researchers have also studied carbon balance as a whole, studying negative air ions and the carbon cycle in the forest [18,19,20]. Researchers have also conducted research regarding negative air ions and wind speed and direction, and concluded that negative air ions are negatively correlated with dust particles such as PM2.5, and that the concentration of negative air ions increases with an increase in light intensity and decreases with a decrease in light intensity [20,21,22,23,24,25,26].

Studies have shown that NAIs in forest environments can inhibit bacteria, remove dust, improve cell activity, and purify ambient air [4,27,28,29]. NAIs in higher concentrations can regulate the body’s homeostasis by regulating the serotonin content level in the body. By stimulating hormone secretion, NAIs improve the body’s metabolic cycle and delay cell aging [4]. Vegetation in forest environments ionizes the air to produce negative oxygen ions through leaf tip discharge, photosynthesis, and the release of volatile aromatic hydrocarbons. Different community structures, tree species, and plant species have significantly different effects on the concentration of negative oxygen ions in the environment [30,31]. Forest stand canopy density can reflect a forest’s environment and structure, and is a decisive factor that affects the light conditions of a forest. It has a significant impact on the diversity of plant species and the stability of community structures in forest environments. Therefore, the study of NAI concentrations in environments with different forest canopy densities can better reflect NAI concentration levels and changing patterns in forest environments [27,32].

Negative air ions are an important indicator of air cleanliness in an area, and play an important role in regulating the psychological and physiological functions of the human body. With the rise of forest eco-tourism, the generation process and impact mechanism of negative air ions have become a research hotspot. This study was based on meteorological data of forest vegetation and negative air ion concentration observation data from the Zhuyu Bay Scenic Area in Yangzhou City. It used a random forest model in machine learning to comprehensively analyze and determine the important environmental factors that affect changes in negative air ion concentrations from a nonlinear perspective. This study’s random forest model was constructed through independent sample pairs, and was simulated and tested to determine its prediction accuracy, and at the same time, environmental factors that had the greatest impact on negative air ions were screened out.

2. Materials and Methods

2.1. Study Site and Plant Conditions

The study site was located in the Zhuyu Bay Scenic Area, Guangling District, Yangzhou, in Jiangsu province. It is the main site of the National Positioning Observation and Research Station of Urban Ecosystem in Yangzhou City, Jiangsu Province. This area is located in the transition zone from a subtropical humid monsoon climate to a temperate monsoon climate, with an average altitude of 2 m, an average annual temperature of 15.8 °C, and an average annual precipitation of 864 mm. The vegetation in this area is zonal secondary forest vegetation. The forest stands grow well and are typical and representative of the plain water network area in the lower reaches of the Yangtze River. The area is rich in forest resources, and the primary types of vegetation are Cedrus deodara, Metasequoia glyptostroboides, and Phyllostachys pubescens.

2.2. Measurement Methodology and Data Collection

The sensors used in the experiment are shown in

Table 1. Types of environmental factor variables observed in this experiment.

Factors	Abbreviation	Measurement Range	Model	Brand	Origin
Particulate matter 2.5	PM2.5	0~999 μg/m3	PH-PM-999	Puhou	Nanjing, China
Negative air (oxygen) ion	NAI	0~500,000 ion/cm3	AN-200	Anion	Chongqing, China
Air pressure	AP	300~1100 hpa	PH-APRE-101	Puhou	Nanjing, China
Wind direction	WD	0~360	PH-WDIR-360	Puhou	Nanjing, China
Wind speed	WS	0~30 m/s	PH-WSPD-30	Puhou	Nanjing, China
Air temperature	T	−20~80 °C	PH-ATERH-165	Puhou	Nanjing, China
Air moisture	AM	0~99.9%	PH-ATERH-165	Puhou	Nanjing, China
Nitric oxide	NO	0~2000 ppm	42i	Thermo Scientific	Waltham, MA, USA
Nitrogen dioxide	NO2	0~2000 ppm	42i	Thermo Scientific	Waltham, MA, USA
Oxynitride	NOX	0~2000 ppm	42i	Thermo Scientific	Waltham, MA, USA
Sulfur dioxide	SO2	0~2000 ppm	43i	Thermo Scientific	Waltham, MA, USA
Ozone	O3	0~2000 ppm	49i	Thermo Scientific	Waltham, MA, USA
Carbonic oxide	CO	0~2000 ppm	48i	Thermo Scientific	Waltham, MA, USA
Carbon dioxide	CO2	0~2000 ppm	410i	Thermo Scientific	Waltham, MA, USA
Methane	CH4	0~2000 ppm	5900-A	Thermo Scientific	Waltham, MA, USA

. Automatic observation equipment was installed in the Zhuyu Bay Scenic Area in the center of the park. The observation equipment had an air inlet 3 m above the ground. The instrument was maintained and calibrated twice each month. Maintenance included replacing the filter and desiccant, and adding water to the tank. The calibration process included preparing standard gases, setting instrument parameters, running blank measurements, running standard gas measurements, calculating standard curves, calibrating instruments, verifying calibration results, and periodic calibration. From January 2020 to December 2022, continuous monitoring and data collection were conducted. In addition to NAIs, nine types of atmospheric pollutants (PM2.5, CO, CO₂, SO₂, NO, NO₂, NO_x, O₃, and CH₄) and five meteorological factors (air temperature, air moisture, air pressure, wind direction, and wind speed) were monitored. All data were measured every hour, and automatically stored on the server.

2.3. Data Filtering

We used R language to conduct preliminary screening of negative air ions data. The screening process was as follows. (1) Screen the time series to exclude time series discontinuities and abnormal data caused by equipment storage interruptions and faults. (2) Compare each value with its preceding and following values. Compare the values. If the value is less than 3 times or 1/3 of the previous value, discard it and record it as NA. (3) Determine 6 or more consecutive identical data values as abnormal values and record them as NA. (4) Calculate the difference for values less than 10, take the average of the two data before and after, round them, and record the interpolated value at that moment. (5) Filter again after assigning values, remove values that are still less than 10, record them as NA, and output all valid data; while filtering factors, only the first two parts are filtered, and outliers are eliminated. After eliminating outliers, approximately 282,390 sets of valid data were finally selected for analysis.

2.4. Data Processing

Through Pearson analysis, the atmospheric pollution and meteorological factors automatically collected by the Zhuyu Bay long-term observation station throughout the year were monitored. We monitored 13 environmental factors in four seasons (spring, summer, autumn, and winter), and generally discussed the correlation between each influencing factor and NAIs. As NAIs are affected by a variety of environmental factors, combined with the results of Pearson correlation analysis, multiple linear regression was performed on the relevant factors in each season to analyze the contribution of different influencing factors to NAIs. A matrix analysis was conducted on the correlation coefficients of 13 environmental factors to explore correlations between these factors. In order to eliminate the similarities between these factors, we used a random forest algorithm to analyze the importance of influencing factors in different seasons.

2.5. Establishment and Analysis of the Random Forest Model

A random forest is an ensemble learning algorithm that performs classification and regression tasks by combining multiple decision trees. It is a forest composed of multiple decision trees. Each decision tree independently randomly selects a portion of samples from the training set for training, and each decision tree randomly selects a portion of features for evaluation when dividing nodes. The final classification or regression result is obtained by voting on the results of all decision trees. Random forest models can be used to handle non-linear relationships, classification, regression, higher-order correlations, as well as to assess the importance of variables, interpolate missing data, and more [33]. Existing studies have shown that random forests have accurate predictive capabilities when screening interference from complex and changeable factors [34,35]. The model fitting effect uses the coefficient of determination (R²) to test the accuracy of the simulation results, and uses the importance score to rank the selected variables. The calculation formula is as follows:

$${\mathrm{R}}^2=1-{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{p}}_{\mathrm{i}}-{\mathrm{o}}_{\mathrm{i}}\right)}^2∕{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{o}}_{\mathrm{i}}-\overline{{\mathrm{o}}_{\mathrm{i}}}\right)}^2$$

$$\mathrm{V}{\mathrm{I}}_{\mathrm{n}}\left({\mathrm{x}}_{\mathrm{j}}\right)=\left\{{\sum }_{\mathrm{i}=1}^{{\mathrm{N}}_{00\mathrm{B}}}\mathrm{I}\left[\mathrm{f}\left({\mathrm{X}}_{\mathrm{i}}\right)={\mathrm{f}}_{\mathrm{n}}\left({\mathrm{X}}_{\mathrm{i}}\right)\right]-{\sum }_{\mathrm{i}=1}^{{\mathrm{N}}_{00\mathrm{B}}}\mathrm{I}\left[\mathrm{f}\left({\mathrm{X}}_{\mathrm{i}}\right)={\mathrm{f}}_{\mathrm{n}}\left({\mathrm{X}}_{\mathrm{i}}^{\prime }\right)\right]\right\}/{\mathrm{N}}_{00\mathrm{B}}$$

In the formula, o_i and o_i are the observed values and model fitting values of negative air ions, respectively; $\overline{{\mathrm{o}}_{\mathrm{i}}}$ is the mean of the observed values. N_OOB is the number of out-of-bag samples; f(X_i) is the i-th observation value in out-of-bag data; f_n(X_i) is the result of the i-th observation value of out-of-bag data on the n-th tree before the observation value of the random replacement variable X_i. The corresponding predicted value f_n(${\mathrm{X}}_{\mathrm{i}}^{\prime }$) is the predicted value corresponding to the i-th observation value of out-of-bag data on the n-th tree after randomly replacing the observed value of variable X_i; I[f(X_i) = f_n(X_i)] and I[f(X_i) = f_n(${\mathrm{X}}_{\mathrm{i}}^{\prime }$)] are discriminant functions. When f(X_i) = f_n(X_i) or f(X_i) = f_n (${\mathrm{X}}_{\mathrm{i}}^{\prime }$), the value is 1, otherwise it is 0.

3. Results

3.1. Seasonal Variations

We imported three-year monitoring data from 2020 to 2022 into EXCEL, removed some abnormal data, and sorted all measured data from Zhuyu Bay Park into the four quarters of three years. We analyzed negative air ions and PM2.5 concentrations in different seasons. We divided the seasons according to the meteorological division method, classifying March to May as spring, June to August as summer, September to November as autumn, and December to February of the following year as winter. Negative air ion concentrations were arranged from high to low by season as autumn > winter > summer > spring, but the difference between spring and summer concentrations was not obvious (Figure 1a). Changes in PM2.5 concentrations in Zhuyu Bay were also recorded during the four seasons. PM2.5 concentration reached its maximum value in spring, but the difference in data between spring and autumn was not obvious (Figure 1b).

Temperature and humidity results show that overall change trends were in an inverted “V” shape, with an upward trend from spring to summer and a downward trend from summer to winter. Both reached maximum values in summer, but the difference was that the temperature change trend was more obvious than that of humidity (Figure 2a,b). From the perspective of air pressure, there was little change in spring and autumn, with a downward trend from spring to summer and an upward trend from summer to winter (Figure 2c). The wind direction results show that change of wind direction reached their maximum value in autumn and their minimum value in summer (Figure 2d). From the perspective of wind speed, there was a downward trend from spring to autumn and an upward trend from autumn to winter (Figure 2e).

By analyzing gas pollutants data, we found that there were significant differences in nitrogen-related gases between summer and winter (Figure 3a–c), and there were also significant differences between nitric oxide and nitrogen oxides between autumn and winter. Nitric oxide and nitrogen oxides showed a downward trend from spring to summer and an upward trend from summer to winter, and concentration changes were not obvious in summer and autumn. The sulfur dioxide concentration fluctuated little among the four seasons, with no significant difference noted (Figure 3d). Seasonal ozone concentrations, from high to low, were as follows: spring > summer > autumn > winter. There was an overall downward trend, and a significant difference between spring and winter concentrations (Figure 3e). Nitrogen dioxide, carbon monoxide, and methane concentrations all showed a downward trend from spring to autumn, and showed an upward trend from autumn to winter (Figure 3b,f,h). Seasonal carbon dioxide concentrations, from high to low, were as follows: summer > autumn > spring > winter. The overall trend was an inverted “V” shape, with an upward trend from spring to summer and a downward trend from summer to winter (Figure 3g).

3.2. Correlation Analysis

In order to further study the relationship between environmental factors and NAIs, we conducted a correlation analysis, as shown in Figure 4. The results show that not only were there significant correlations between environmental factors and NAIs, but there were also significant correlations between environmental factors. For example, while NAIs had a significant positive correlation with NO_X, NO₂ and AP had an even more significant positive correlation; CO had a significant negative correlation, and CH₄, CO₂, and WS had an even more significant negative correlation; AP had a very significant positive correlation with NO_X and NO₂, and CO₂ showed a very significant negative correlation.

3.3. Random Forest Model

Based on a factor analysis, we used NAIs as the dependent variable and these 14 environmental factors as the independent variables to conduct RF regression. We selected the parameters ntree = 200 (number of trees in the RF model), mtry = 2 (number of variables tried at each split in the RF model) according to the guidance of the “randomForest” package in R. The importance ranking scores used are shown in

Table 2. Random forest importance scores of factors.

Factors	% IncMSE	% IncMSE p-Value
AM	67.24008	0.00990099
PM2.5	56.18474	0.00990099
WD	52.99107	0.00990099
CH4	42.50171	0.00990099
AT	37.14405	0.03960396
AP	31.10199	0.27722772
SO2	30.62325	0.34653465
WS	28.60907	0.71287129
CO2	28.36984	0.58415842
CO	27.04638	0.76237624
O3	24.49242	0.93069307
NO2	18.82889	0.64356436
NO	14.8271	0.99009901
NOX	11.76385	0.93069307

and Figure 5.

The random forest model was based on the 14 monitored variables. We established the IncMSE model to study these variables, and comprehensively evaluated the prediction results using an average-based method. The final random forest model, with 14 variables, is shown in Figure 5. Among them, total explanatory variables equaled 81.91%, spring explanatory variables equaled 54.22%, summer explanatory variables equaled 89.46%, and winter explanatory variables equaled 82.71%. It could be seen in the IncMSE model that the importance score of the impact of environmental factors on NAIs changed with the seasons throughout the year. In spring, PM2.5, AP, WD, AT, and AM had high importance scores, and in summer PM2.5, WD, and SO₂ had the highest importance scores. In autumn, environmental factors WD, AT, AM, NO₂, NOX, O₃, CO₂, and CH₄ had the highest importance scores. In winter, PM2.5, AP, WD, and AM had the highest importance scores. Environmental factors CO₂ and CH₄ had high importance scores.

4. Discussion

According to the correlation analysis results, there were significant seasonal differences in the impact of environmental factors on NAIs. And we do not yet understand the interactions between environmental factors. In addition, some uncontrollable factors were unavoidable during the measurement process, such as extreme weather, equipment damage, forest signal transmission, etc., which may have led to the loss or deletion of measurement values. Therefore, this study used the IncMSE model in the random forest model to study the results of the model to more comprehensively and accurately determine the important environmental factors that affected NAIs changes. From this, we concluded that PM2.5, WD, AT, AM, and CH₄ were typical influencing factors.

AM had a certain importance in spring, autumn, and winter. Only in summer was its importance score 0, which should be related to the factors of hot weather and dry air during summer, because water molecules in the air are an important component of NAIs. AM affects the size of OH– in the air. The hot climate during summer causes the air humidity to be lower in summer than in other seasons, which affects the presence of water molecules in the air, thereby affecting the concentration of NAIs. Bowers found that near a waterfall NAI concentrations were affected by surrounding water vapor [2]. Reiter also showed that an increase in air humidity can change the agglomeration effect of particles, reforming smaller particles into larger particles, thereby reducing NAIs and maintaining NAIs at a higher concentration [7].

Although our results show that PM2.5 was not highly correlated with NAIs, we also found that PM2.5 had an important impact on negative air ions concentration, which was consistent with studies such as those of Shi et al. [36] and Li et al. [37]. The results were consistent, which may be related to the fact that particle matter can change its physical properties by binding to NAIs, thereby generating macromolecular precipitates, which can limit the existence and retention of NAIs, thereby reducing the concentration of NAIs [1]. Scholars studying the urban heat island effect have found that the concentration of NAIs decreases as the degree of air pollution increases. At the same time, atmospheric pollutants, such as PM2.5, determine differences in NAIs on the urban–rural gradient [38,39].

WD had importance scores in four seasons, and changes in wind direction affected the generation of NAIs. When the wind does not blow toward Zhuyu Bay, the urban forest park may be less affected by the wind. From a microscopic perspective, the formation of NAIs requires a shift from the molecular state to the ionic state, and this process can be accelerated under strong wind conditions [40]. The conditions of strong wind are based on the wind direction, which also determines to a certain extent the source of negative ions in the air of Zhuyu Bay. For example, wind blowing from the ocean to the land will bring negative ions of the ocean, and wind blowing from the mountainous area to the plains brings negative ions from mountainous areas.

AT is also an important factor affecting the concentration of NAIs in urban forest parks. General studies show that NAI concentrations are negatively correlated with air temperature, which is similar to our research results [41,42,43,44,45]. This may be due to the impact of different monitoring environments. This study introduced the random forest IncMSE model and revealed the primary factors affecting NAI concentrations through continuous on-site monitoring and correlation analysis of measurement data. The monitoring of NAI concentrations was affected by environmental factors, the research area, and other factors. The results of this study are based only on monitoring data from Zhuyu Bay Scenic Forest Park in Yangzhou, Jiangsu, China, and the generalizability of these conclusions requires further confirmation.

5. Conclusions

This study found that environmental factors have important effects on changes in negative air ions concentrations, and that the environmental factor negative air ions concentrations in Zhuyu Bay urban forest were most sensitive to was humidity, followed by PM2.5. Humidity was the most critical factor, because it directly affects the formation of negative air ions in the environment and vegetation. An increase in air humidity enhances the possibility of intermolecular collision between water molecules and promotes the process of molecular transformation into ionic states. These results contribute to deepening our understanding of the characteristics of negative air ions and their applications in urban forests.