Shimen National Forest Park (23°36′50″N–23°39′20″N, 113°46′16″E–113°49′17″E) is located in the northeast of Conghua District, 95 km away from Guangzhou central city (Wushan Street), which is selected as the study area of exurban forest. It is close to Guangdong Longmen Nankunshan Provincial Nature Reserve and Guangzhou Baishuizai Scenic Spots, and the adjacent forest area is concentrated and contiguous, as shown in Figure 1. Shimen National Forest Park has a total area of 2,636 hm² with 98.91% forest coverage, and a forest area of 2,607 hm² with an altitude ranging from 270 m to 1,210 m. It pertains to the south subtropical monsoon climate, with average temperature between 19.5°C and 21.4°C, and an annual average rainfall of about 1,800 mm (Zhu et al., 2021a,b).

Guangdong Longyandong Forest Farm (23°16′41″N–23°18′31″N, 113°25′44″E–113°28′30″E), situated in the northeast suburb of Baiyun District, Guangzhou City, is 34 km away from Guangzhou central city, as the study area of suburban forest. It is adjacent to Maofengshan Forest Park and Causeway Bay Forest Scenic Area, surrounded by forest environments, as shown in Figure 1. Guangdong Longyandong Forest Farm covers an area of 524 hm², with a forest coverage of 94.56%, a forest area of 495 hm², and an altitude of 101–536 m. It has a subtropical maritime monsoon climate with an average rainfall of 1,694 mm and an average temperature of 21°C (Hong et al., 2020).

Changgangshan Nature Reserve (23°09′29″N–23°09′44″N, 113°21′53″E–113°22′12″E), as the study area of downtown forest, is seated in the campus of South China Agricultural University in Tianhe District, Guangzhou City, Guangdong Province, seeing Figure 1. The total area is about 15 hm²,with 89.47% forest coverage, and the forest area is about 13 hm², with an altitude of 40–50 m. The shape of Changgangshan Nature Reserve is quadrangular, with one side adjacent to the South China Expressway, and the other three sides surrounded by the campus buildings of South China Agricultural University, forming a small-scale isolated urban forest. It belongs to the subtropical monsoon maritime climate, with an average annual temperature of about 22.6°C and the annual precipitation of about 2,050 mm (Tong and Xiao, 2014).

Wushan Street (23°09′22″N–23°09′26″N, 113°21′29″E–113°21′53″E), as the study area of urban control, is located near Wushan Station of Metro Line 3 in Tianhe District, Guangzhou City, Guangdong Province. The street area is long and narrow, with a study area of about 5–6 hm² and an altitude of about 40 m, as shown in Figure 1. It is surrounded by a commercial area with dense buildings, heavy pedestrian and vehicular traffic, and no vegetation other than trees along the roadside.

From September 2019 to January 2020 and from May to August 2020, the field data was measured from the 20th to 30th of each month, where the downtown forest and urban control were experimented commonly on the 20th, 21st, and 22nd; suburban forest was tested primarily on the 23rd, 24th, and 25th; and exurban forest was investigated ordinarily on the 27th, 28th, and 29th. The weather conditions during the sampling period were similar, either sunny or cloudy. Test times were appropriately adjusted for unexpected weather conditions. Three indicators of negative air ion concentration, air oxygen content, and forest microclimates were measured in three urban forest environments and an urban control. The negative air ion concentration was measured at a sampling height of 1.5 m above the ground using a COM-3200PRO II negative air ions monitor produced in Japan. This was repeated 3 times for a total of 12 data. The air oxygen content was measured by a TD6000-SH-O₂ oxygen content tester manufactured in China, which was repeated 3 times. Kestrel 5500 hand-held meteorological instrument fabricated in the United States was used to measure forest microclimates, including temperature (T), relative humidity (RH), and wind speed (V), which were each repeated 3 times at a sampling height of 2 m (Zhu et al., 2021a,b).

The CRITIC method proposed by Diakoulaki et al. (1995) and it is primarily adapted to ascertain the objective weights of criteria, based on the contrast intensity and conflicting relationships of the decision criteria. The CRITIC method uses standard deviation and Pearson correlation to measure the contrast intensity and conflicting relationships of each criterion, respectively (Krishnan et al., 2021). The main steps of this trick can be described as demonstrated:

1. Normalization of each indicator

The above three indicators are classified as positive or negative, depending on the impact of each indicator on environmental quality. Positive indicators signify that the greater the value, the better the healthcare benefits, such as NAIC. In keeping with medical measurements, when the AOC is 20.9%, it is normal; when it is more than 23.5%, it appertains to an oxygen-enriched environment, which is conducive to relieve brain fatigue and ameliorate body function (Hong, 2015). In this study, the AOC in different geographical spaces is below 23.5%, so the AOC can be considered as a positive indicator. The negative indicators reveal that the greater the value, the worse the healthcare benefits. The negative indicator in this study is HCI. In order to bring it into a unified evaluation system for comparison, the range normalization method is adopted for standardization, which is calculated as follows:

where xij¯ is the normalized score of alternative i with respect to indicator j, xij is the actual score of alternative i with respect to indicator j, xjmax is the maximum score of indicator j, and xjmin is the minimum score of indicator j.

2. Calculate the standard deviation of each indicator

where xj¯is the mean score of indicator j and n is the total number of alternatives, Sj is the standard deviation of indicator j.

3. Calculate the correlation coefficient of every pair of indicators

where Rj is the correlation coefficient between indicator j and other indicator, rij is the correlation coefficient between indicator i and j.

4. Compute the information content of each indicator

where Cj denotes the information content of indicator j.

5. Determine the objective weight of each indicator

where Wj reveals the objective weight of indicator j.

The objective weight of each healthcare index was determined by CRITIC method mentioned above (Table 4), thus, the formula of urban forest comprehensive healthcare index (UFCHI for short) in different geographical spaces is shown:

In which, NAICi is the normalized value of i-th negative air ion concentration, AOCi is the normalized value of i-th air oxygen content. HCIi is the normalized value of i-th human comfort index.

In SPSS22.0, the UFCHI values were systematically clustered using Ward’s method and the squared Euclidean distance, and the UFCHI was divided into four classes based on the results of the cluster analysis, see Figure 2. The integrated healthcare benefits expressed in terms of different numerical intervals were explained in Table 5.

SPSS22.0 was implemented for variance analysis, multiple comparison, CRITIC calculation, and cluster analysis; Origin 2022 was executed to draw the diagram.

The NAIC of three urban forests were distinctly higher than urban control, which was in line with the research conclusions of most scholars: NAIC in forest areas were immensely greater than that in forestless areas (Feng et al., 2015). A possible explanation for this might be that in the forest environment, radioactive elements escaping from rocks and soil, photoelectric effects on plant branches and leaves, and volatile aromatic substances released by trees can ionize the air in the forest (Zeng et al., 2006). And the dust removal effect of forests can boost NAIC, along with extending their lifetimes. Nevertheless, the urban environment had less vegetation, dense population and heavy traffic. Previous studies have unfolded that human activity had a prominent impact on NAIC, and NAIC was negatively correlated with the flow of population and vehicles (Yuan et al., 2014; Feng et al., 2015). The dense population would consume a great deal of O₂ and generate a large amount of CO₂, and NAIC was passively linked with oxygen consumption (Pu et al., 1993). Simultaneously, CO₂ absorbed negative ions to form heavy ion deposition, leading to a further decline in NAIC. More importantly, automobile exhaust would build up the concentration of suspended particulate matter and CO₂ in near-surface air, which enormously diminished the NAIC and shortened their existence.

The average annual of NAIC in exurban forest was about 1.2 times and 2.4 times than that in suburban forest and downtown forest, respectively. The forest area and forest coverage of exurban forest and suburban forest were large and contiguous, barely influenced by the outside, and had a stable ecosystem. It was detected that NAIC was better at high forest coverage than at low forest coverage (Cui et al., 2020). In addition, exurban forest was rich in plant species, complex in community structure, large in leaf area index (LAI) of community, and relatively high in canopy density. Earlier research suggested that the higher the complexity of community structure, the greater the NAIC (Gao and Zhang, 2018), and NAIC had a positive correlation with LAI of community (Liu et al., 2011). Additionally, there was a positive correlation between NAIC and canopy density, and NAIC tended to heighten with the escalation of vegetation canopy density (Du et al., 2018; Zheng et al., 2021).

Seasonal variations of NAIC in exurban forest and suburban forest were summer > spring > autumn > winter, while downtown forest and urban control were autumn > winter > spring > summer. The reason for the difference may be closely related to the temperature. The downtown forest was small-scale and surrounded by dense buildings, developing an isolated urban forest. In hot, muggy summer, the downtown forest had no noticeable cooling and humidifying effect, and the high temperature inside the forest lessened the stomatal conductance and photosynthetic rate of plants. Former studies have discovered that NAIC was positively correlated with stomatal conductance and photosynthetic rate (Zhao et al., 2018). Stomata was the channels for the release of negative air ions produced by plants, and the larger the stomata, the higher the NAIC (Zhang et al., 2016); the higher the photosynthetic rate, the greater the ability of plants to release oxygen, and the higher the NAIC. At the same time, the high-temperature urban environment given rise to the dispersion of pollutants (Zhang et al., 2022), and pollutants adsorbed loads of negative air ions in the process of diffusion, resulting in a lower rate of negative air ions production than consumption. Therefore, the NAIC in downtown forest and urban control were broadly lower in summer than in other seasons, which was also coherent with the seasonal variation of NAIC in Hangzhou (Qi et al., 2011) and Shenzhen (Wang et al., 2020).

The AOC in different geographical spaces was conspicuously different, exurban forest > suburban forest > downtown forest > urban control, and the variation of AOC in exurban forest and suburban forest were larger. It may be due to the fact that exurban forest and suburban forest were far away from the city center, having lower concentrations of air pollutants. And their forest areas were larger and concentrated, with a larger unit biomass and LAI, so the ability to release oxygen through photosynthesis was notably better than those of downtown forest and urban control. In addition, the variation of AOC was closely related to the growth and development cycle of forest communities (Wen, 2020), varying with the change of plants into the peak growing season or dormant state.

It has been established that AOC was most affected by temperature (Gu, 2013; Wang et al., 2014; Wen, 2020). The seasonal variations of AOC in different geographical spaces were summer > spring > autumn > winter, which was the same as the temperature. This illustrated that, under the impact of temperature, there was an outstanding rise in oxygen release in spring and summer, due to intense physiological activity during the peak period of plant growth. On the contrary, after entering autumn and winter, the physiological activity of plants gradually decayed and slowly entered a dormant period, which considerably reduced the oxygen release. Hence, the seasonal variations of AOC in different geographical spaces were higher in summer and spring than in autumn and winter, which was identified with the findings of Wen (2020).

The HCI of exurban forest and suburban forest was at Level I very comfortable degree, which were one level higher than that of downtown forest and urban control. This reflected that the HCI in forest environments was superior to that in urban environments, which matched the conclusion of most scholars (Gu et al., 2019; Rahman et al., 2020; Teshnehdel et al., 2020): the function of regulating microclimate played by forest environments was strikingly better than that of non-forest area. This may have a lot to do with their altitude, mountain topography, forest area, forest coverage, surrounding environment, and their location away from urban centers. As can be seen from the study area introduction and Figure 1, the forest areas of exurban forest and suburban forest were concentrated and contiguous on a large scale with high forest coverage. Moreover, they had higher elevations, more diverse and complex terrain, and were adjacent to other forest parks, nature reserves or scenic spots. In contrast, the downtown forest had lower altitudes, less pronounced terrain variations, and smaller forest areas. Meanwhile, it was surrounded by dense buildings or the South China Expressway, and more susceptible to the effects of urban microclimates. Numerous studies have manifested that HCI was affected by altitude, vegetation cover, and terrain shading effects; HCI was positively related to altitude (Wen, 2020), and the mountainous terrain shading effect and higher vegetation cover were conducive to the HCI (Yao et al., 2019). Thus, the HCI in exurban forest and suburban forest was better than that in downtown forest and urban control. Furthermore, some scholars have claimed that HCI was positively correlated with NAIC and canopy density (Du et al., 2018). Exurban forest and suburban forest had high NAIC and canopy density, which have achieved remarkable cooling and damping effect and improved HCI. However, the lack of vegetation in urban environment resulted in large variations in temperature and humidity, and its HCI was poor.

The HCI of forest environments in different geographical spaces were all the best in autumn. In contrast, HCI was worst in winter in exurban forest and suburban forest, and in summer in downtown forest and urban control. Precedent studies have shown that all microclimate factors had a certain impact on HCI, with air temperature having the greatest impact, followed by relative humidity and wind speed (Liu C. et al., 2017; Huang et al., 2020). Forest environments can tremendously drop air temperature and rise relative humidity (Yan et al., 2012), and its ability of regulating local microclimate was relatively strong. Thus, exurban forest and suburban forest had the effect of reducing heat and improving comfort during the hot summer months, while in winter, forest environments were colder than urban environments and their HCI was poor. This was matched with the findings of Zhang (2014) that urban forest in regulating human comfort was the worst in winter. Nevertheless, temperature, relative humidity, and wind speed were all at very comfortable level in fall, so the HCI in urban forest environment in autumn was the best. It was perceived that urbanization contributed to the increase of maximum temperature in summer and minimum temperature in winter in urban areas with 57.8% and 46.1%, respectively. The HCI in urban areas moved in the direction of hot and uncomfortable in summer and warmer and more comfortable in winter (Huang et al., 2020). At the same time, downtown forest was more susceptible to urban microclimate and their ability to regulate microclimates was limited. Therefore, the downtown forest and urban control had better HCI in winter than in summer.