Investigating the Optimal Location of Potential Forest Industry Clusters to Enhance Domestic Timber Utilization in South Korea

South Korea has abundant forest resources capable of supplying the domestic wood demand. Despite the extensive forest resources, there is continued uncertainty about the nature, quantity, and quality of the timber contained in any particular forested area. Additionally, some technical, logistic, and economic challenges act as barriers to the expansion of domestic timber utilization. To overcome these limitations and to enhance the domestic timber utilization in South Korea, this study investigated the optimal location of potential forest industry clusters. The potential forest availability was estimated based on localized allometric equations. The integration of the analytical hierarchy process and GIS modeling, including a supply chain that minimizes transportation costs, allowed the identification of optimal forest industry clusters locations that balanced the economic, environmental, and social dimensions within the forest industry supply chain. The study reveals that the estimated potential forest resources availability presented approximately 1 billion m3, including sawlog (474 million m3) and pulpwood grade (541 million m3). Additionally, 45 percent of the sawlogs and 48 percent of the pup grade wood were produced from the Gangwon and Gyeongsangbuk-do regions. Furthermore, the logistic analysis indicates that ten potential forest industry clusters are best aligned with the optimal socio-economic impacts with minimized timber transportation costs. To identify the optimal size and number of potential forest industry clusters, further studies that consider fixed and variable costs for maintaining the forest industry clusters are required.


Introduction
The forest industry has faced various challenging issues in the overall forest supply chains [1]. There are issues with the integrated networks composed of raw material suppliers, manufacturing facilities, and transportation providers that work across organizational boundaries to deliver forest products to consumers [2]. Recently, the most developed forest countries in Europe, such as Sweden, Finland, and the Baltic countries, have improved forest supply chains by implementing cost-effective logistics on given facility location [3][4][5]. The forest industry clusters (FIC) concept has been introduced and applied to examine an efficient forest supply chain that minimizes transportation costs in each Despite the abundant forest resources, South Korea is heavily dependent on imported wood to supply its domestic wood demand [14]. According to the statistical book of forestry in 2019, 85% of the total wood consumption in South Korea was imported [15]. One of the factors that make this imbalance between timber supply and demand in the domestic market, is that South Korea has lower cost competitiveness in the wood industry compared to other countries, such as Austria and Japan, due to its low density of forest roads and low proportion of planted forests for timber production [16]. Although several previous studies have found evidence for potential domestic timber utilization, some technical, logistic, and economic challenges act as barriers to the domestic utilization expansion [6,[17][18][19][20].
These challenges point out that: (i) forest sites are highly variable, of unknown timber quality, and widely distributed all over Korea in such a way that it imposes harvesting, processing, and transportation problems that have implications on the economic viability of domestic timber utilization from any specific location; (ii) despite the large volume of forest resources, continued uncertainty exists about the timber nature, quantity, and quality in any particular forest location; (iii) the lack of robust planning prompt concerns about the decreasing profitability of domestic forest entities, highly affected by low productivity and high production cost due to their small business scales [21].
Since transportation is the most critical factor impacting the overall cost of forest supply chains, site selection for FIC in proximity to unutilized forest resources currently available will offer an opportunity to minimize the timber supply cost by providing an optimal location of potential forest industrial areas [22]. Selecting optimal locations of FIC needs to consider environmental and socio-  [9]. The "Non-stocked" indicates the areas that include bamboo forests, eroded lands, roads, etc.
Despite the abundant forest resources, South Korea is heavily dependent on imported wood to supply its domestic wood demand [14]. According to the statistical book of forestry in 2019, 85% of the total wood consumption in South Korea was imported [15]. One of the factors that make this imbalance between timber supply and demand in the domestic market, is that South Korea has lower cost competitiveness in the wood industry compared to other countries, such as Austria and Japan, due to its low density of forest roads and low proportion of planted forests for timber production [16]. Although several previous studies have found evidence for potential domestic timber utilization, some technical, logistic, and economic challenges act as barriers to the domestic utilization expansion [6,[17][18][19][20].
These challenges point out that: (i) forest sites are highly variable, of unknown timber quality, and widely distributed all over Korea in such a way that it imposes harvesting, processing, and transportation problems that have implications on the economic viability of domestic timber utilization from any specific location; (ii) despite the large volume of forest resources, continued uncertainty exists about the timber nature, quantity, and quality in any particular forest location; (iii) the lack of robust planning prompt concerns about the decreasing profitability of domestic forest entities, highly affected by low productivity and high production cost due to their small business scales [21].
Since transportation is the most critical factor impacting the overall cost of forest supply chains, site selection for FIC in proximity to unutilized forest resources currently available will offer an opportunity to minimize the timber supply cost by providing an optimal location of potential forest industrial areas [22]. Selecting optimal locations of FIC needs to consider environmental and socioeconomic constraints that involve determining the cluster number and scale within the geographical network of connected forest resources, processing facilities, and end-markets [23]. GIS modeling is a tool broadly used to investigate forest volume estimation and inventory assessment [24][25][26][27], as well as logistic analysis to identify minimal transportation costs based on the least cost distance matrix [28][29][30]. Additionally, GIS integration with the analytical hierarchy process (AHP) is a preferred approach for investigating the potential location of forest-related facilities and for assessing the relative importance of economic, environmental, and social criteria affecting site selection [30][31][32][33][34][35][36]. The AHP analysis is a specific technique that is applied to estimate the factor weighting [37]. The influencing effect of FIC varies from region to region, and to reflect the regional interests, it is necessary to assign them a different relative weight [38].
This study aims to investigate the optimal locations of prospective FIC in South Korea. Integrated GIS and AHP models with logistics simulations were used to identify the best FIC candidate locations with optimal logistic supply chain cost. The results could contribute to the future Korean FIC planning and investment decisions with high efficiency of domestic timber utilization.

Methodology
The overall research process is presented in Figure 2. Forest resources were estimated by the potential grade of product types, such as sawlog and pulp, based on wood grading standards in Korea [39]. The restriction and suitability models were developed to investigate the potential FIC available land areas. In this phase, integrated GIS and AHP were applied to assign a weight to each main and sub-criterion [40]. All GIS and spatial data were collected or provided by the government and industrial partners, including the Korea Forest Service (KFS), National Institute of Forest Science (NIFOS), Ministry of the Interior and Safety, Ministry of Environment, National Geographic Information Institute, and National Transport Information Center. FIC optimal locations were investigated through three analytical phases, including land availability, land suitability, and location-allocation analysis. In the first phase, the restriction area was removed applying developed restriction model. Then, the potential FIC locations were identified using the suitability model according to a set of specific main and sub-criteria. In the last phase, identified potential FIC locations were simulated using network analysis to investigate the minimization of total transportation cost between the FIC candidate locations and forest resources. The following section describes, in detail, the phase of each analysis.  Table 1 shows the species composition of South Korea's forests. The potential forest availability was estimated using the forest growth and yield model developed by the NIFOS [41]. The details and key assumptions were described in [41][42][43]. However, estimating the forest resources at the local to regional scales has been hampered by the uncertainty of environmental conditions of the regional forests [44]. To overcome these limitations, the models were developed on local-based forest   Table 1 shows the species composition of South Korea's forests. The potential forest availability was estimated using the forest growth and yield model developed by the NIFOS [41]. The details and key assumptions were described in [41][42][43]. However, estimating the forest resources at the local to regional scales has been hampered by the uncertainty of environmental conditions of the regional forests [44]. To overcome these limitations, the models were developed on local-based forest inventory data, including four major conifer species in the country: Pinus densiflora Siebold & Zucc., Pinus koraiensis Siebold & Zucc., Larix kaempferi (Lamb.) Carr., and Pinus rigida Mill. The developed models are presented in Table 2. Equation (1) presents the model estimation of forest resource availability, where Y is the estimated total volume of forest resources available in the country, X ij is forest stand age for species i in province j, and α ij , β ij are parameter values to estimate forest volume for species i in province j ( Table 2). The required data for forest resource availability estimation were sourced from the Korean Forest Geospatial Information Service's database [45].  Additionally, to account for forest regimes and harvesting efficiency, the estimation of forest resources availability was limited in the diameter at breast height (DBH) and area by the Korean Forest Harvesting Regulation (KFR) [42]. In this research, the DBH over 18 cm was only considered as actual usable forest resource as sawlogs in natural forest and plantation forests. Furthermore, forest areas that are less than 5 ha were not included in the estimation due to the inefficiency of forest harvesting and forest ownership. Most of the small scale of forest lands (less than 5 ha) are owned by private sectors, including individual, company, and corporation. In this context, there are some limitations to manage and access the forest land owned by private owners. For these reasons, this research did not include small scale (less than 5 ha) forest lands. Lastly, estimated forest availability was separated into two types of forest products-sawlogs and pulp. The definitions of those two forest products are presented in Table 3. Table 3. Definition of forest product types and associated wood grades [42].

Restriction Model to Identify FIC Available Area
The potential location of FIC needs to consider several criteria and constraints, such as geological and environmental factors, which are imposed by national and/or local government regulations. Primarily, land-use conversion and development of industrial areas are strictly regulated within national land-use conversion development regimes [46]. Thus, we considered the geological and environmental constraints to select the potential locations of FIC. These constraints could identify protected and reserved areas where industrial area construction is not allowed (e.g., Baekdudaegan mountains reserve, national park, forest reserved area, wildlife protection area, and urban natural park areas). Restricted buffer zones were then created through proximity analysis. The potential FIC locations were designed only outside the buffer zone.
To develop a restriction model, a list of GIS datasets is described in Table 4. Each GIS restriction layers combined into one single layer as a restriction map, including specific information of constraint values. The vector layer was converted and reclassified into a binary raster data format (0 or 1 value). Reclassified binary raster data consisted of binary values, thus areas consisted of "zeros" were classified as restricted area, and areas consisted of "ones" were classified as FIC available areas. The final restriction map was generated by multiplying all reclassified binary raster layers using a raster calculation tool in ArcGIS 10.5 [30,47]. The output of the reclassified map was called "restriction map".

Developing a Suitability Model Using AHP
The potential locations for FIC candidates were investigated using integrated AHP and GIS techniques to assign the factors a different relative weight [38,40]. To evaluate the relative weighted values, the first step was to construct an AHP hierarchical model consisting of two phases, including the main criteria (economic, environmental, and social), each having three sub-criteria. Nine sub-criteria were considered in the suitability analysis. The selected sub-criteria were categorized according to three different characters of main criteria called "economic", "Environmental", and "Social".
For the main criteria's economic factor, timber availability, local timber demand, and transport logistics were analyzed. They were regarded as high correlated sub-criteria with economic influence on the FIC construction. It was assumed that the abundant forest resources within an area would have a higher weighted value compared to an area less abundant in forest resources. The areas with a high local demand for forest resources were also given a high weighted value because areas with high demand for forest resources are more likely to attract FICs. The close distance to roads and FICs also led to high weighted value due to the minimum distance to roads, which can reduce logging transportation costs substantially.
To avoid land-use change issues in the estimation of relative weights of sub-criteria included in the environmental factor, it was preferred to locate FICs in available areas that are legally authorized under Korea's land-use law and regulations to construct FIC. Furthermore, FIC candidate locations closer to industrial areas were preferred because they would have a higher weighted value compared to distant potential FIC candidates. The topographical preference for FIC location was determined to be in a flat area (0 to 3 degree) to avoid landslides or erosion that could occur on steep slopes and sites with easily eroded soils. The local employment rate and population density in the local government area were considered in the weighting assignment by the social criteria. A low employment rate was preferred in the construction of FICs that contribute to the local government job creation. Low and middle population density regions were prioritized for improving local economic development. Overall weights were obtained from 15 forestry experts, consisted of the academic, forest industry, forest research institution, and government officer in KFS.
Once the weights were assigned to GIS layers, they were reclassified into raster format (100 × 100 m cell size), and the corresponding final suitability map was generated by a weighted linear combination of the overall GIS layers [30]. The applied equation was as follow (Equation (2)): where S i is suitability value in the ith cell at the final grid, w n is the weight allocated to the nth criterion from the AHP analysis, C in is the value of the ith cell in the grid of the nth layer, and p is the total criteria number in the suitability analysis [48]. The overall criteria values were standardized before using them in the equation. The estimated values Equation (2) in the final output were reclassified into five-level classes where higher values indicated a more suitable location for the potential FIC location. Computed reclassified raster value cells in levels 5 and 4 were identified as the most suitable FIC candidates for potential FIC construction.

Identifying the Optimal Location of FIC Using Network Analysis
The optimal location of FIC was identified using the Network Analysis tool in ArcGIS 10.5. The road network, location of forest resources, and location of potential FIC candidates were needed to simulate location-allocation analysis. The location-allocation analysis calculated transportation distances using an actual road network between FIC candidates and forest resources rather than a radius or straight-line distances. The aim of the analysis (p-median problem solver) was to locate n candidate FICs among m potential selected suitable FIC candidates (m > n) while satisfying several constraints; thus, the distances between each forest resource and FIC location were minimized [49].
It automatically identifies the shortest travel distance so that the overall distance between the FIC candidate points s and the set of forest resource points f is minimized (Equation (3)) [48]. where w f = weight associated with each forest resource location f, d sf = distance between forest resource point f and the potential FIC location s. x sf = one if forest resource point f is allocated to FIC location s, otherwise equals zero [48]. The location-allocation tool creates the least cost matrix with the shortest travel distances between FIC candidates and the available forest resource points. This tool was developed with multiple techniques to achieve a near-optimal solution, including vertex substitution heuristic and refining metaheuristic [50]. Logistic costs were calculated based on optimal route travel distance derived from the location-allocation analysis (Equation (4)).
where C ij is the total transportation cost (United States Dollar-USD per green ton) for the optimal route between the FIC candidate j and forest resource point i; FC is the fixed cost related to harvesting cost (USD per dry ton) at forest resource point i; VC is the variable cost associated with traveling distance (USD per ton-km); d k is the distance of the traveled road segment (km), and M is the total number of segments along the optimal pathway between the FIC j and the forest resource point i [30,48]. In this study, the transportation cost from the logging site to the FIC was only considered to investigate the optimal FIC location and numbers. Other cost components, including fixed and variable costs, were kept at the same rate in the whole country due to the uncertainty of the future harvesting operation and cost. Finally, to investigate the most efficient number of potential FIC construction, six scenarios with a different number of optimal FIC locations and transportation limits were simulated using the location-allocation analysis (Table 5).

Forest Resources Availability in South Korea
The estimated forest availability indicated that a large amount of forest resources are available in the form of sawlogs and pulp/energy wood to supply FIC in South Korea. The distribution of potential sawlogs and pulp/energy wood availability map is shown in Figure 3. The total estimated sawlog volume is 474,949,000 m 3 . Pulp/energy wood equals 541,876,000 m 3 . The largest number of potential forest resources was located in the Gangwon-do region. Gangwon and Gyeongsangbuk-do regions were able to produce about 45% of the sawlogs and 48% of the pulp/energy wood (Figure 4). Forest resources availability analysis shows that most forest resources could be produced in Northeast Korea, including Gangwon-do, Chungcheongbuk-do, and Gyeongsangbuk-do regions (Figure 4). Compared to previous research [41], the estimated total volume of forest availability was not significantly different. However, this research adapted Korean forest harvesting regulation and examined available forest resources as sawlog and pulp grade wood, respectively. In the near future, there is required to investigate forest resource availability considering socio-economic and environmental constraints, a barrier to the harvesting operation.

Forest Resources Availability in South Korea
The estimated forest availability indicated that a large amount of forest resources are available in the form of sawlogs and pulp/energy wood to supply FIC in South Korea. The distribution of potential sawlogs and pulp/energy wood availability map is shown in Figure 3. The total estimated sawlog volume is 474,949,000 m 3 . Pulp/energy wood equals 541,876,000 m 3 . The largest number of potential forest resources was located in the Gangwon-do region. Gangwon and Gyeongsangbuk-do regions were able to produce about 45% of the sawlogs and 48% of the pulp/energy wood (Figure 4). Forest resources availability analysis shows that most forest resources could be produced in Northeast Korea, including Gangwon-do, Chungcheongbuk-do, and Gyeongsangbuk-do regions (Figure 4). Compared to previous research [41], the estimated total volume of forest availability was not significantly different. However, this research adapted Korean forest harvesting regulation and examined available forest resources as sawlog and pulp grade wood, respectively. In the near future, there is required to investigate forest resource availability considering socio-economic and environmental constraints, a barrier to the harvesting operation.

Identified Locations of Forest Industry Clusters Candidates in South Korea
As a result of the AHP analysis and pairwise comparison, different weights were given to each main and sub-criteria [30,51]. The estimated weights and the corresponding consistency ratios are presented in Table 6.

Identified Locations of Forest Industry Clusters Candidates in South Korea
As a result of the AHP analysis and pairwise comparison, different weights were given to each main and sub-criteria [30,51]. The estimated weights and the corresponding consistency ratios are presented in Table 6. The economic factor had the highest weight among the three main criteria. This was most prominent at the second level (sub-criteria), where "Timber availability" was ranked the highest weight (0.289) among all sub-criteria. In the environmental criterion (main criteria level), the "Distance from the industry" was given the highest weight (0.118). Social aspects had the lowest weighted main criterion, and the "Local employment rate" had the highest weight (0.095) compared to "Population density" (0.031) and "Employment rate by age class" (0.035). The consistency ratio showed a high degree of consistency confirming the adequacy of the estimated weight values in Table 6. The AHP results highlighted that the cost-effective timber supply chain is highly required to develop the FIC. Additionally, the role of FIC was expected to create local jobs in the rural area and to revitalize the local economy.
The corresponding weights were adopted in the suitability model development, where the reclassified weights were applied to investigate the final suitability map for the FIC candidates' potential location. Integrated restriction and suitability maps were then computed from raster calculations, which allowed the investigation of final FIC candidate locations. All identified potential locations of FIC candidates, which had the highest suitability value (≥5), were selected. Figure 5 presents the identified FIC candidate locations. Five hundred and sixty-five points were identified as suitable locations to construct FIC. The excluded areas were mainly protected, reserved, and restricted areas where construction of FIC is not allowed, such as Baedudaegan mountain reserves and a national park. White cells in Figure 5  protected, reserved, and restricted areas where construction of FIC is not allowed, such as Baedudaegan mountain reserves and a national park. White cells in Error! Reference source not found. represent those restricted areas where FICs cannot be established. The number of final FIC candidates by regions are as follows: Gangwon-do regions (n = 393), Jeollabuk-do (n = 52), Gyeongsangnam-do (n = 47), Chungdheongbuk-do (n = 32), Gyeongsangbuk-do (n = 26), Gyeonggido (n = 8), Jeollanam-do (n = 5), and Chungcheongnam-do (n = 2).

Identified Optimal Locations of Potential FIC in South Korea
Error! Reference source not found. presents the 12 identified FIC locations that were computed from the candidate locations (565 candidates) by solving the p-median problem that had 164,850 forest resource points. The location-allocation analysis was simulated based on six scenarios, and the detailed information of the scenarios was described in Table 5.  Figure 6 presents the 12 identified FIC locations that were computed from the candidate locations (565 candidates) by solving the p-median problem that had 164,850 forest resource points. The locationallocation analysis was simulated based on six scenarios, and the detailed information of the scenarios was described in Table 5. Table 7 shows the simulated transportation costs under different scenarios. Scenario 1 indicated Goesan-gun as the optimal location of FIC. The average transportation distance of this scenario (141 km) was the longest compared to the other scenarios. Furthermore, scenario 1 was able to reach up to 99% of forest resource locations available throughout the country. The transportation cost from forest resource locations to the selected FIC, however, was highest among other scenarios due to long transportation distance. The increase of FIC constructions led to a dramatic decrease in the average transportation distance, thus the associated cost decreased. Scenario 2, in which three FICs with a 200 km transportation radius limit was assumed to be constructed, had a coverage rate reaching approximately 98% of the total forest available resources. The average distance was about 64% (91 km) of that computed in Scenario 1 (141 km). The total transportation cost significantly decreased by adding two more FICs in the timber supply chain of the country ( Table 7). The most cost-effective number of FIC in the country is 10. If more than 10 FICs were to be constructed, the transportation costs would increase. This is due to the overall transportation distance connecting forest resources and FIC locations, which increases over 10 FICs, thus decreasing the transportation cost efficiency. As expected, the transportation cost was highly influenced by the distance between the FICs and the location of forest resources.    There have been several previous research studies investigating the optimal location of forestrelated industries including biomass facility [22,30,51] and forest industry clusters [6,53]. Especially in South Korea, Koo et al. (2016) investigated the suitable location of FIC using GIS and economic assessment. However, they did not consider logistic cost from potential forest resources to prospect FIC and constraints of FIC constructions. To overcome these limitations, this research provides the optimal location of prospective FIC in South Korea using potential forest resource availability. Integrated GIS and AHP, along with a timber supply chain cost analysis, were developed to identify the optimal location of FIC.

Identified Optimal Locations of Potential FIC in South Korea
Internationally, Woo et al. (2018) investigated the optimal location of biomass energy facilities in the Tasmania region by integrating MCA and GIS. They investigated the impact of moisture content on transportation costs using local truck costs per tonne and kilometer. Furthermore, Delivand et al. (2015) examined optimal locations of bioenergy facilities with minimized transportation costs and greenhouse gas emissions in South Italy. They simulated greenhouse gas emission along the timber supply chain using location and allocation analysis results. Compared to both cases, a major weakness of this research is that there were no pre-existing data along the timber supply chain, including greenhouse gas emission, moisture contents, and cost of truck transportation in South Korea.
In future studies, considerable efforts in greenhouse gas emission along the timber supply chain and economic assessment of FIC are required to widely investigate the efficiency and impact of FIC for the South Korea conditions. Furthermore, the cost analysis model has a limit because it does not consider the facility investment cost (i.e., construction cost) and FIC capacity that may affect the optimal number and FIC locations. Thus, the construction cost and FIC size need to be taken into account for more practical design of FIC locations in the future.

Conclusions
This research investigated the optimal location of FICs in South Korea. It aimed to improve the national forest supply chain and increase the country's domestic timber value. The potential locations of prospective FICs were investigated using restriction and suitability models. To determine the optimal locations of FICs, an integrated AHP and GIS model were developed. The results show that South Korea has enough forest resources to operate a number of FICs. The scenario-based analysis highlighted that 10 FICs with a 100 km radius was the best option to minimize overall transportation costs from the forest resource to cluster location. From the supply chain perspective, we found that the distance between forest resources and potential FIC locations was identified as the most impacting factor on transportation cost. However, a practical management planning approach considering FIC operation and investment costs, as well as harvesting cost, is further required in future research to adequately quantify the impact of FIC location and number on overall supply chain costs.