Seoul, selected as the study area, is the capital of South Korea and the metropolitan city with the population of 10.57 million within the area of 605,960,000 m². Population was rapidly increased from 0.65 million to 10.97 million between 1951 and 1992 (Governmen, 2014). The impervious area rate was 47.28% in 2005 and 47.64% in 2010 and the forest area rate was 25.29% in 2005 and 24.96% in 2010. The urbanized area rate was 60.98% and the green and open area rate was 39.03% in 2010. The total park area was 170,100,000 m² that is 28.09% of Seoul and the park area per person was 16.37 m² in 2014 (Governmen, 2014). An increase in impervious and urbanized areas disrupts green infrastructure networks and causes the urban heat island.

Land cover maps based on satellite imagery and cadastral data were the main data to analyze green infrastructure in Seoul. Land cover maps and cadastral maps provide many land information, but they provide different types of data of land uses, districts, and areas based on their objectives. Land cover maps do not have legal boundaries and are used as references and open data for studies and researches. Cadastral maps are used to tax and manage properties administered by Korean laws (Table 1).

The main elements of green infrastructure are mainly classified into hubs and links, whilst this study classified elements of green infrastructure into three elements that are hubs, potential links within hubs, and links for a detailed analysis of district priorities.

The definitions of these elements within this study are different with classical definitions of elements of green infrastructure. Hubs in a classical definition are large areas of natural vegetation, whilst hubs in this study were extracted after buffer analysis based on forest areas of land cover maps to identify significant green areas. Potential links within a hub in this study indicate areas within a hub except forest areas of land cover maps to identify interaction between green areas within a hub. Links in this study indicate straight lines of the shortest distance between hubs to identify corridors connecting hubs (Table 2).

Land cover maps were used to identify hubs, potential links, and links with GIS techniques. Cadastral maps were used to analyze important levels of hubs, potential links, and links with statistical methods. Identification of hubs was a fundamental element to identify potential links within hubs and links that connect hubs. This study assumed forest areas of land cover maps as significant green areas for hubs, because forest areas have the lowest impedance value for green infrastructure (Figure 1a).

The extracted hubs, potential links, and links were used as districts of elements of green infrastructure and in- corporated with cadastral data to analyze values of green infrastructure. Hubs, potential links, and links have land categories and areas from cadastral data within their boundaries. These parcel data were incorporated on impedance values of land uses, and used to analyze priorities of hubs, potential links, and links with applying calculated impedance values of hubs, potential links, and links (Figure 1b).

Impedance values are weighted based on vegetation coverage and type, the age of the green space, and degree of anthropogenic disturbance (Kong et al., 2010). Impedance values of land uses for green infrastructure from prior two studies were applied to 26 categories of land uses of cadastral data [(Surrey, 2011); (Kong et al., 2010)]. The application of impedance values to cadastral data indicates that forest areas of cadastral data are the most significant green area for green infrastructure because forest areas have a lowest impedance value among 26 categories of land uses, whilst constructed areas, industry areas, gas station, and warehouse have negative effects on green infrastructure with high impedance values (Table 3) (Park et al., 2018).

To analyze green infrastructure values and connectivity values of hubs, potential links, and links, this study created and used followed three equations. The Green Infrastructure Value (GIV) represents the significance of hubs and the condition of green infrastructure of hubs. The higher GIVs indicate more significant green areas for green infrastructure. GIVs are calculated as:

(1)

Where GIV_Hubi is the green infrastructure value of Hubi, TA_Hubi is the total area of parcels of cadastral data within Hubi, A_Hubi is the area of each parcel within Hubi, and I_Hubi is an impedance value of each parcel within Hubi.

Interaction between forest areas indicates the efficiency of potential links within hubs. The Interaction Value (IV) within a hub shows levels to connect forest areas for green infrastructure networks within a hub. The higher IVs indicate more significant potential links within hubs for green infrastructure. IVs are calculated as:

(2)

Where IV_Hubi is the interaction value of potential links within Hubi, TA_Hubi is the total area of parcels of cadastral data within Hubi, A_{PL j} is the area of each parcel that is extracted areas excepting forest areas within Hubi, and I_{PL j} is an impedance value of each parcel.

The Connectivity Value (CV) indicates the efficiency of links between hubs. The higher CVs indicate more significant links to connect hubs for green infrastructure within Seoul. CVs are calculated as:

(3)

Where CVLinkt is the connectivity value of Linkt, TA_Linkt is the total area of parcels of Linkt, A_Linkt is the area of each parcel within Linkt, I_Linkt is an impedance value of each parcel.

The Ministry of Environment provides land cover maps having different resolutions. Within Seoul, the land cover map is separated to 135 tile maps. This study combined the separated 135 land cover maps to a raster land cover map that has 1:5,000 scale, 1 m resolution, and seven categories of land uses. Within the map, Seoul has 660,657,635 m² that is a different area in comparison of the total area of cadastral maps. The map shows that impervious areas have 7,164 parcels and 352,254,286 m² that are 53% of the total area of Seoul. Forest areas have 1,365 land cover parcels and 143,144,713 m².

Cadastral maps provide important information of parcels such as objective IDs, districts, parcel numbers, and land categories. Cadastral data of Seoul indicate 605,711,407 m² of city areas, 25 boroughs and 967,502 cadastral parcels. With analyzing land use categories of cadastral data, constructed areas have 718,070 parcels that are 217,974,644 m² and road areas have 149,051 parcels that are 78,569,109 m². Forest areas have 20,959 parcels that are 140,548,998 m². Vegetation areas have 18,188 parcels that are 11,387,207 m² and agriculture areas have 13,386 parcels that are 12,150,105 m². Cadastral data shows that 49% of Seoul is constructed and road areas that can be impervious areas, whilst 23% of Seoul is forest areas.

Forest areas of land cover maps were buffered with 100 m, 300 m, and 500 m to find an appropriate buffered distance for classification of hubs. With 100 m buffer of forest areas, 172 hubs were extracted with 245,300,404 m². With 300 m buffer of forest areas, 43 hubs were extracted with 352,217,049 m². With 500 m buffer of forest areas, 16 hubs were extracted with 430,667,859 m². This study selected 300 m buffer of forest areas to identify hubs based on Landscape Management Areas of Seoul City. With minus 300 buffer, this study extracted 124 forest areas and selected the top 20 buffered areas in terms of surface areas as the main hubs for green infrastructure in Seoul. Within top 20 hubs, 2,570 potential links were extracted with deleting forest areas of land cover maps within 20 hubs (Figure 2).

To identify links, this study delineated 190 straight lines of the shortest distance between hubs (Figure 3a). In 190 lines, 104 lines that cross hubs (Figure 3b) and 44 lines that cross Han-River were deleted (Figure 3c).

Finally, 42 lines were extracted and buffered with 20 m wide that is a same wide of three green ways of the 2030 Seoul City Park and Green Master. With these processes, this study extracted and used 20 hubs and 42 links as the elements of green infrastructure in Seoul (Figure 4).

The extracted 20 hubs have a total area of 186,156,839 m2 from 103,553 cadastral parcels (Figure 2a). Hub1 has the most parcels (45,652) totalling the largest area of 62,672,277 m2. Hub18 has the least parcels (272) with an area of 1,001,803 m2. Hub20 has the smallest area of 806,311 for 312 parcels. Within 20 hubs, parcels that have forest areas of land categories of cadastral data were comprised of 123,755,018 m2 that is 66% of the total area of hubs. Residential areas have the second largest area of 24,072,382 m2 that is 13% of the total area of hubs. GIVs of hubs indicate conditions of land categories of hubs for green infrastructure. The 20 hubs have an average of an importance value of 1.803. Hub18 that has the third smallest area has the highest GIV of 4.650. Hub8 that has the eighth highest area has the lowest GIV of 0.402. The results show that Hub18, Hub13, Hub3, Hub2, and Hub4 is the top five districts that have higher conditions for green infrastructure than other 15 hubs. Hub8, Hub11, Hub17, Hub16, and Hub19 are the districts that have lower conditions for green infrastructure (Table 4).

Within the 20 hubs, 32% of the total area is extracted as potential links that have 59,729,682 m²from 88,424 cadastral parcels (Figure 2b). Hub5 and Hub6 have the highest rates of potential links of 50% within their areas. Hub1 has the largest area of potential links with 19,419,578 m² that is 31% of the total area of Hub1. Within 2,570 potential links from 20 hubs, residential areas of cadastral data have the largest area of 18,627,780 m² that is 31% of the total area of potential links within hubs. Forest areas have the second largest area of 17,144,331 m² that is 29% of areas of potential links.The IVs of hubs indicate connections between forest areas for green infrastructure within a hub. The average IV of the is 2.554. Although, Hub20 has the smallest area, it has the highest IV of 5.894. Hub3 has the third highest area and the second highest IV of 3.840. Hub11 has the lowest IV of 1.199. Hub20, Hub3, Hub4, Hub18, and Hub8 have higher connections between forest areas for green infrastructure than other hubs.

Hub5, Hub6, Hub13, Hub11, and Hub19 have lower connections for green infrastructure than other hubs (Table 4)

The extracted 42 links have 17,860 cadastral parcels with 4,868,134 m². Although a link between Hub5 and Hub9 has 3 parcels, a link between Hub2 and Hub17 has 1,432 parcels. A link between Hub6 and Hub16 has the largest area of 442,937 m² from 687 parcels. A link between Hub2 and Hub16 has the smallest area of 1,358 m² from 5 parcels. Within 42 links, residential areas of land categories have the largest area of 2,523,686 m² that is 52% of the total area of 42 links. Road areas have the second largest area of 938,667 m² that is 19% of the total area of potential links. The eight links have higher connectivity values than 1 and 34 links have lower connectivity values than 1. A link between Hub2 and Hub3 has the highest connectivity value of 37.614 and a link between Hub2 and Hub16 has the second highest connectivity value of 27.909. A link between Hub2 and Hub9 has the lowest connectivity value of 0.005. The average connectivity value except the 10 highest and 10 lowest values was 0.577 (Table 5).