Satellite-based Assessment of Ecosystem Services Considering Social Demand for Reduction of Fine Particulate Matter in Seoul

Abstract

Fine particulate matter (PM2.5) has been the biggest environmental problem in Korea since the 2010s. The present study considers the value of urban forests and green infrastructure as an ecosystem service (ES) concept for PM2.5 reduction based on satellite and spatial data, with a focus on Seoul, Korea A method for the spatial ES assessment that considers social demand variables such as population and land price is suggested. First, an ES assessment based on natural environment information confirms that, while the vitality of vegetation is relatively low, the ES is high in the city center and residential areas, where the concentration of PM2.5 is high. Then, the ES assessment considering social demand (i.e., the ESS) confirms the existence of higher PM2.5 values in residential areas with high population density, and in main downtown areas. This is because the ESS of urban green infrastructure is high in areas with high land prices, high population density, and above-average PM2.5 concentrations. Further, when a future green infrastructure improvement scenario that considers the urban forest management plan is applied, the area of very high ESS is increased by 74% when the vegetation greenness of the green infrastructure in the residential area is increased by only 20%. This result suggests that green infrastructure and urban forests in the residential area should be continuously expanded and managed in order to maximize the PM2.5 reduction ES.

1. Introduction

In metropolitan areas, the health and psychological impact of environmental pollution on citizens is very significant (von Schneidemesser et al., 2019; Lu, 2020). In particular, the sensitivity to air pollution concentrations in cities has increased significantly since the World Health Organization (WHO) designated fine particulate matter (PM2.5) as a class 1 carcinogen in 2013 (WHO, 2017). Although pollutants have not increased significantly, except in certain developing countries such as India and China, social concern and citizens’ needs levels for clean environment have increased (Apte, 2018; Lim et al., 2020). According to future climate change scenarios, it is predicted that air pollution will still be a problem in the future if greenhouse gases are not actively reduced (Park et al., 2020a; Silva et al., 2017).

Since the Korean particulate matter forecast began, many citizens have become concerned and have discussed particulate matter (Ryu and Min, 2020). Accordingly, even before the COVID-19 pandemic, it was not difficult to find citizens wearing masks during the season of high PM2.5 concentrations. The actual PM2.5 concentrations in Korea are higher than the global average, and are frequently very bad during the winter-spring period, when the northwest wind is a factor, due to external contributions(Park et al., 2020b; Lim and Park, 2022). In addition, Korea’s direct emissions of air pollutants are higher than the global average due to heavy chemical industry and dense cities, although domestic emissions have significantly improved due to pollutant regulations and technological improvements (Kim et al., 2018; Moon et al., 2018). Overall, the level of civil society has changed more than the improvement in the PM2.5 concentration, and Korea’s representative air pollution indicator is presently PM2.5.

In the midst of various environmental problems such as PM2.5, the concept of nature-based solutions or ecosystem services (ES) has attracted recent attention for calculating the direct effects or benefits of functional ecosystem use by humans (Daily, 2008; Gopalakrishnan et al., 2019). In particular, forest ecosystems are in the spotlight for various services such as carbon storage, air purification, and water supply (Lee et al., 2018; Lim and Choi, 2021; Lee and Lim, 2022). However, the physical mechanism of the reduction effect of forest ecosystems on PM2.5 has not yet been clearly elucidated, and previous studies on the ES concept are insufficient (Gaglio et al., 2022), with particulate matter and ES not being directly addressed due to several constraints. For instance, while satellite information provides high spatial resolution data regarding vegetation which is an indicator of ES, the spatial resolution of data on the atmospheric environment is relatively low (Park et al., 2020a; Lim and Park, 2022). In addition, the physical mechanism cannot be evaluated quantitatively yet; only a qualitative conceptual evaluation is possible. Moreover, social problems such as particulate matter can be directly dealt with only when indicators relating to urban social demand are considered. Hence, given the high concentration of PM2.5 in cities, and high demand for improvement from citizens, it is necessary to deal with this issue from the ES perspective.

Therefore, the present study approaches the ecosystem services of urban forests and green infrastructure in Seoul as a representative Korean city that is greatly affected by particulate matter. In particular, key indicators are constructed using satellite-based data, and the ES of urban green space is evaluated with consideration of the civic social demand. In addition, in consideration of the urban forest management plan, a simple scenario is set up to apply the expansion of the urban forest in the residential area, and the effect of the expansion of green infrastructure in improving the ES is confirmed. Finally, the value of urban forests and green infrastructure in responding to social problems such as particulate matter in a space with high social demand is discussed.

2. Data and Methods

1) Study Area

The target area of the present study is the entire area of Seoul, the capital of Korea, which is the economic, social, and cultural center. Seoul is located at latitude 37.25° N to 37.43° N and longitude 126.45° E to 127.18° E, with a total area of 605.21 km² (Fig. 1). For many decades, the population of Seoul was over 10 million, but it has decreased since 2015 as the population flowed into a new city in Gyeonggi-do. Currently, approximately 9.5 million people live in Seoul. However, many residents of Gyeonggi-do either commute to Seoul or spend a lot of time there, so their value as a floating population must not be ignored.

Fig. 1. Study area: (a) land use of Seoul, (b) administration boundary of Seoul.

2) Methods

Herein, a methodology is devised to assess the reduction of PM2.5 due to vegetation such as green infrastructure and urban forest areas from the ES perspective. Here, the concept of ES assessment is focused not only on the PM2.5 reduction function of vegetation, but also on social demand and the role of vegetation (Lee et al., 2017).In other words, the present study evaluates the ecosystem services considering societal demand (ESS) of urban green areas.

Although previous studies have been conducted on the PM2.5 reduction function of forests, quantified values for individual processes such as absorption, sedimentation, and blocking were not identified. Hence, the present study approaches the ES of urban green areas as a spatial evaluation concept by measuring the vegetation greenness in urban areas with high resolution and substituting this value into the PM2.5 concentration (Song et al., 2016). Although it is reasonable to conduct an ES assessment that considers urban forest information such as species, age, and grade, there is only limited information on green infrastructure in residential areas. Therefore, a quantitative evaluation is performed via the available high-resolution vegetation greenness index.

The ES assessment method presented herein is defined by Eq. (1) and (2):

\(E S={ }_n V G \times{ }_n P M 2.5\)     (1)

\(E S S={ }_n V G \times{ }_n P M 2.5 \times\left(\left({ }_n \text { Pop }+{ }_n \text { Cost }\right) \times 0.5\right)\)      (2)

where the ES considers the vegetation by using only natural environment information, while the ESS considers social demand; _nVG is the normalized vegetation greenness(VG) calculated from a 10-m grid of the constructed monthly Sentinel-2 satellite images, _nPM2.5 is the normalized average annual PM2.5 concentration calculated from a high-resolution grid, _nPop is the normalized population density from the statistics for each district of the target area at the level of the minimum-unit local government, and _nCost is the normalized information on the individual official land prices of the target area, which refers to the economic feasibility of each plot of land.

The two indicators of social demand are _nPop and _nCost. The former is the population living in the area, which can receive services due to urban green space, and is thus an indicator of the numerical impact, while the latter includes the floating population, which is high in commercial areas, and is an indicator of the social needsfor health and the environment due to the cost of living. Thus, _nPop and _nCost are applied together to consider the resident population, the floating population, and the social demand equally. However, this assessment method was designed in terms of relative assessment rather than numerical evaluation. The calculated spatial indicator was divided into 5 levels from very high to very low (i.e., a quintile classification) via Natural Break Classification. For statistical comparison, very low was digitized as 1 and very high as 5, etc., and the regional average was calculated.

In addition, two policy scenarios for strengthening the PM2.5 reduction ES of the urban green infrastructure were set up, and the resulting changes analyzed. While these policy scenarios conceptually utilized the spatial realization of major related items such as urban forest creation in the living area and recovery of urban forest health in the 2nd Seoul Urban Forest Management Plan, the creation of new green space in Seoul has significant spatial limitations. Therefore, the scenarios were run in the spatially-limited green infrastructure area in Seoul as a method of improving and reassessing the relatively low-quality green infrastructure. For this purpose, in a grid with a VG of 0.5 or less, Scenario 1 was set to increase the VG by 10%, and Scenario 2 was set to increase the VG by 20%. The effect of green infrastructure improvement was analyzed by evaluating the ESS to which the strengthened green infrastructure was applied. This spatial unit scenario was constructed with reference to Park et al. (2020b).

3) Data and data processing

The global high-resolution annual average PM2.5 concentration data developed by van Donkalaar et al. (2016) was downscaled to the city scale for the present study. This included the global atmospheric observation network data obtained via the aerosol robotic network , and the aerosol optical thickness data from the orbital multi-angle imaging spectroradiometer, the moderate resolution imaging spectroradiometer, and SeaWiFS instruments using the Dark Target, Deep Blue, and multiangle implementation of atmospheric correction algoritms. These were used together with the predicted values of the GEOS-Chem model to construct representative PM2.5 data spread over a wide area. These datasets provide the additional advantage of having considered the regional environmental characteristics via the geographic weighted regression model (Hammer et al., 2020). The data set is provided by the research team’s web page of Washington University (https://sites.wustl.edu/acag/datasets/surfacepm2-5/).

The co-kriging spatial statistical technique was applied to downscale the high-resolution data obtained from the atmospheric environment observation network of Seoul, including explicit spatial information and background PM2.5 data. The target spatial resolution of the downscale is 30 m² . Co-kriging is a proposed method that compensates for the weaknesses of univariate kriging, which cannot take into account other factors affecting the main variables. If additional data are available, the predictive ability at unsampled points can be improved via multivariate kriging (Jo et al., 2018). Thus, while univariate kriging performs interpolation by using autocorrelation of the same variable according to the location, with only one variable value, co-kriging performs interpolation by linearly combining two or more variables. In co-kriging, the variable to be predicted is called the main variable, and one or more other variables are each referred to as secondary variables. The main variable (Z) is then given by Eq. (3):

\(\boldsymbol{Z}=\sum_{i=1}^{N(x)} \alpha_i z_i+\sum_{i=1}^{N_s} \sum_{i=1}^{M_j} \alpha_{j k} \boldsymbol{u}_j\left(\boldsymbol{x}_{j k}\right)\)      (3)

where N is the total number of data points in the main variable, N_s is the total number of secondary variables used, u_j is the j-th secondary variable, M_j is the total number of data in the j-th secondary variable, α is the weight, and x is the position of each data point. A total of Ns + 1 variables and N + (N_s × M_j ) data points are used to predict the value of the main variable (Jang et al., 2015).

The urban atmosphere and roadside atmosphere data from the observation network of the atmospheric environment in Seoul was synthesized and used as the main variable, and the data produced by van Donkelaar (2016) was used as the main secondary variable (background data) for downscaling. A total of 25 urban atmospheric observation stations and 15 roadside atmospheric observation networks were used in this work (Fig. 2(a)). The period was 2018, and the target spatial resolution was set to 30 m2. To downscale the data to the target spatial resolution, high-resolution spatial information that affects the generation, movement, and spread of PM2.5 in an area was included. In accordance with previous studies, the distance from the source of PM2.5, and the altitude ratio of each point to the average altitude, were utilized herein. The distance from the PM2.5 emission source (Fig. 2(b)) was calculated as the distance from roads, industrial, commercial, and residential areas in the mid-class land cover map produced in 2013. The elevation ratio of each point to the average elevation was calculated as the ratio of the elevation value of each grid to the average elevation of Seoul (Fig. 2(c)). Here, a value of less than 1 is lower than the average elevation of Seoul (64 m), making it an advantageous area for PM2.5 movement. This concept was also used n the study by van Donkalaar (2016).

Fig. 2. The spatial distributions of the input variables for PM2.5 that were used for downscaling in the study area: (a) the PM2.5 background data (color gradation) and observation stations (yellow points), (b) the distance from the PM source, and (c) the elevation ratio.

To evaluate the ES for reducing PM2.5 in green infrastructure and urban forests, it is necessary to utilize high-resolution VG information at the regional level. Therefore, a 10-m scale time-series vegetation index was constructed using data from the European Space Agency’s Sentinel-2 satellite, as described by Kim et al. (2021). Basic data acquisition and processing were performed using the Google Earth Engine platform, with monthly images being obtained from January to December 2018, and atmospheric correction was performed via machine learning of the values for the surrounding area during cloudy periods. Band 4 (red, around 0.66 μm) and Band 8 (NIR, around 0.86 μm) were used to calculate the normalized difference vegetation index (NDVI), which was used as the VG herein, in accordance with Lim and Yeo (2022). The seasonal characteristics were precisely derived from the monthly VG, and this was calculated as the maximum value for each grid unit to derive the 2018 VG. Those green infrastructure areas that could provide ES were separately extracted from this data, where areas having a VG of 0.3 or higher were selected as the extraction criterion. Below that criterion, the vegetation was considered to be artificial covering, rather than a terrestrial ecosystem.

The official land price information, which describes the floating population and local values, and the population density, which describes the actual consumers, were constructed in spatial units in order to evaluate the ESS. For the population data, the demographic data for each counting district provided by the Statistical Geographic Information Service of Statistics Korea was used. However, although the aggregated district information has the advantage of being detailed, the deviation of the values is large when there are urban parks, rivers, and mountains. Meanwhile, the individual official land prices for Seoul were obtained from the national geospatial information portal, and were used for the analysis by averaging with the boundaries of the least basic administration level, such as the population density.

3. Results and Discussion

1) Spatial characteristics of social and environmental indicators

In Fig. 3(a), the high-resolution spatial distribution (30 m² ) of PM2.5 concentration based on the statistical technique described in Section 2.3 is similar to that of the existing 1 km² background data (Fig. 2(a)), and is well explained in terms of the detailed topography and regional characteristics. In particular, the eastern part of Seoul, Yeongdeungpo-gu, and Gangseo-gu, exhibit the same high concentrations as were observed in the existing background data (Fig, 2(a)), and differences according to altitude and cover are also clearly revealed in the outlying mountainous areas. Hence, while the actual PM2.5 concentration at any one time might not reflect such geographical characteristics, it is judged that the long-term average distribution is generally similar to the actual concentration.

Fig. 3. The spatial distributions of the main indicators in Seoul: (a) PM2.5, (b) VG, (c) land cost, (d) population density

When the 30-m spatial resolution of the Landsat satellite data used in previous studies is increased to the 10-m resolution used herein, it is confirmed that the characteristics of the mountainous areas in Seoul are prominently displayed in the combined VG information (Fig. 3(b)) Moreover, even the landscaping trees in the residential area are closely revealed, thereby indicating that the monitoring of small-scale green infrastructure is now feasible. In particular, VG values comparable to those of urban parks or mountainous areas are observed in the large-scale apartment complexes that have been progressively built over a long period in Gangnam-gu and other areas. In addition, the detailed vegetation differences are confirmed, for example, in areas where roads have been created in mountainous areas. This suggests the data analysis should be performed in a spatial unit with as high a resolution as possible in order to reveal the urban-scale ecosystem functions and services.

The spatial distributions of the social variables, i.e., land cost and population density, are revealed in Fig. 3(c) and (d), respectively. The land cost values are significantly high in Gangnam-gu, Seocho-gu, Yeouido, and Jongno, and there is a clear distinction between residential and mountainous areas. Meanwhile, the population density is higher in areas where the residential complex is concentrated than in areas where land costs are high. In other words, in Seoul, the population density is high in the medium land-cost areas. If the floating population is reflected by the high land cost in the commercial district, this is explained by the high population density in the residential district. It is thought that spatialization of the two data sets(i.e., land cost and population density) and application of ES evaluation can reveal citizens’ demand for ES.

2) Spatial assessment of ecosystem services considering social demand

In this section, the ES is first evaluated based on the natural environment data (i.e., PM2.5, VG) that were constructed in the previous section. Here, a higher PM2.5 concentration and a higher VG is regarded as a more significant ES. Accordingly, the urban parks and small mountain areas within the city show higher ES values than do large mountain areas such as Bukhansan, Dobongsan, and Gwanaksan, outside of Seoul (Fig. 4(a)). In particular, the green areas in the northeast of Seoul, Yeongdeungpo, and Gangseo have high PM2.5 concentrations and, hence, significant ES values. As do the urban forests in Seocho and Gangnam. In the sub-regional ES statistics (Table 1), Seongdong-gu, Dongdaemun-gu, Yangcheon-gu, Gangseo-gu, Yeongdeungpo-gu, and Gangnam-gu are seen to have high ES values, corresponding to a high PM2.5 concentration, although the VG is relatively low.

Fig. 4. The spatial distribution of (a) ES and (b) ESS.

Next, an ESS is assessed by including the social data (i.e., population density and land cost). The results in Fig. 4(b) indicate that large mountain areas far from downtown have relatively low ESS values, whereas higher ESS values are observed in major downtown areas and in residential areas with high population densities. Of particular note are the high ESS values of urban forests and green infrastructure in the residential areas; especially the landscaping trees in a large-scale apartment complex that has been constructed a long time ago. In Table 1, the sub-regional ESS statistics are comparable to those of the ES, although regional differences are clearly seen. In addition, areas with high land prices and high population, such as Songpa-gu, have high ESS values. Areas with high average ESS values generally have high populations, high land prices, and above-average PM2.5 concentrations. These include Dongdaemun-gu, Yangcheon-gu, Songpa-gu, Seongdong-gu, and Dongjak-gu (Table 1). Moreover, the differences between regions and sections are more pronounced in the ESS assessment than in the ES assessment, even when their PM2.5 concentrations and VG levels are similar. These showed also differences according to social demand even with similar PM2.5 and VG levels.

Table 1. The sub-regional statistics of the various socio-environmental indicators, the evaluated ES, and ESS

3) Spatial assessment of ES by applying a green infrastructure improvement scenario

Various qualitative changes in the green infrastructure close to the residential areas due to each of the two scenarios described in Section 2.2 are revealed in Fig. 5. In particular, significant improvements are observed in the high-accessibility green infrastructure of most residential areas and roadside vegetation, including the large apartment complexes in Nowon-gu. As shown in Table 2, this is accompanied by remarkable changes to the evaluated ES, both spatially and numerically. Furthermore, those residential areas that already had very high ESS values in the pre-scenario (present-day) assessment are further improved after the green infrastructure scenario, such that the ESS value of green infrastructure is very high in most residential areas. The results in Table 2 indicate that the percentage of areas with very high ESS values increases from 7.91% for the present day to 9.80% after Scenario 1, and to 13.82% after Scenario 2. This corresponds to numerical area increases from 26.89 km² to 33.3 and 46.97 km², respectively. Hence although 20% increase in the vitality of green infrastructure, a 74% increase was indicated in the areas with very high ESS. Meanwhile, the percentage of areas with High ESS values increased from 20.23% present day to 23.77% after Scenario 1, and 24.94 after scenario 2.

Fig. 5. The spatial distributions of ESS values after green infrastructure Scenario 1 (a) and Scenario 2 (b).

Table 2. The ESS statistics before (present-day) and after green infrastructure Scenarios 1 and 2

In Fig. 6, the major regions of Dongdaemun-gu and Seongdong-gu, where the PM2.5 concentrations and population densities are very high, have been magnified in order to reveal the difference in the quality of green infrastructure. Here, it can be seen that the present-day ESS values of low-rise residential areas and the campus (University of Seoul, Hanyang University) are relatively low (Fig. 5(a)), but are significantly improved after green infrastructure Scenario 2 (Fig. 5(b)). Overall, this result suggests that green infrastructure and urban forests should be continuously expanded and managed in order to maximize the PM2.5 reduction ES.

Fig. 6. Comparison of local ESS values (a) before (present-day) and (b) after green infrastructure Scenario 2.

4) Implication and limitations

Until now, ecosystem services have been assessed using either spatial data-based tools such as InVEST and ARIES or surveyed data (Lee et al., 2017; Lee et al., 2018; Lim et al., 2019). The present study proposed a methodology for ES evaluation via a simplified equation using satellite imagery as the main data source.In particular, terrestrial vegetation information, which is the basis of ES, was utilized herein as a satellite-based high-resolution vegetation index that can be readily applied to a wide area, including the third world. The PM2.5 was suggested herein as a type that can be handled via satellite-based ES assessment. Nevertheless, other air pollutants and indicators such as the land surface temperature will be available in the future.

In the present study, PM2.5 data were downscaled as high-resolution information that is suitable for ES evaluation at the regional level. However, there was a limitation in the direct analysis of this high spatial resolution data because there is generally a large spatial resolution difference between atmospheric environment informationsuchasPM2.5andthevegetationinformation in general (Lim et al., 2020; Son and Kim, 2021). To address this issue, statistical techniques were used to downscale the variables that affect the generation and movement of PM2.5 to a level suitable for vegetation or urban ecosystems. Although the downscaling process always has inherent uncertainty, the spatial distribution of the target indicator can be reasonably accurately simulated if the value of the data accumulated over a certain period (e.g., one year) reflects the actual spatial characteristics of the region. This technique can be applied in various forms of topics dealing with terrestrial ecosystems and atmospheric environments.

The presentESSassessmentsuggested the importance of green space in areas where population and social infrastructure are concentrated. It confirmed the importance of the vitality of vegetation in places with high PM2.5 concentration, and suggested that green infrastructure is more valuable in areas with large floating and resident populations. In particular, in the green infrastructure scenario considering Seoul’s urban forest management plan, if the vitality of the green space in the residential area is strengthened, the ES is expected to increase significantly. This suggests that, in the long term, many roles are needed in the expansion and creation of green infrastructure in the residential area in the urban environment plan.

The present study has several limitations. First, the ES and ESS assessment techniques presented herein are simplified equations for relative evaluation within a region. In other words, it is not a numerically accurate evaluation; therefore, it is necessary to pay attention to the interpretation of the results and their future use. In the long term, it is hoped that a simple method for evaluating the physical ecosystem functions will be presented. In addition, the vegetation index was used as an indicator of ecosystem functions, and the land cost and population density were used as indicators of social demand. While the use of these alternative indicators is sound, it can also be a limitation. Moreover, there are many different views, such as political demand or economic perspectives, that refer to the demand for ecosystem services. This study also has limitations considering only residential and commercial areas for the ecosystem service demand. In addition, while the present study ended with the spatial assessment, the ES value needs to be converted into economic value. It is expected that spatial assessment and economic evaluation will be performed together in the future.

4. Conclusions

Herein, a thorough satellite-based spatial assessment of the ecosystem services (ES) aimed at responding to PM2.5 concentrations improved our understanding of the role of green infrastructure in urban regions. First, an ES evaluation based on natural environment information confirmed that although the VG is relatively low, the ES is high in the city center and residential areas where the concentration of PM2.5 is high. In the ES evaluation reflecting the social demand of the city (i.e., the ESS), higher values were confirmed in residential areas with high population density, and in main downtown areas. In particular, the high ESS values of urban forests and green infrastructure in the residential areas stood out. After all, the ESS of urban green infrastructure is high in areas with high land prices and population density, where the PM2.5 concentrations are above average. By applying a future green infrastructure improvement scenario based on the urban forest management plan, assuming a 20% increase in green infrastructure, a 74% increase was indicated in the areas with very high ESS. This result suggests that green infrastructure and urban forests in the residential area should be continuously expanded and managed in order to maximize the ES for PM2.5 reduction. This present study is significant in that it utilized satellite data to perform an assessment of the ESS in response to PM2.5. However, there is a limitation in that a simplified equation was used to calculate the ES values, and the results should be interpreted only as a relative evaluation. Further, it is not possible to calculate the ES as an economic value. It is hoped that this will be supplemented through future research.