Research data

Most of the analyses were based on officially published government data and public datasets. As shown in Table 1, the scope of forest resources was determined using the Forest Function Classification Map and species-specific forest type maps downloaded from official portals. The data were processed and extracted using QGIS (version 3.40.7).

Table 1.Details of the geographic information system (GIS) data used in this study.

View Fullscreen

To estimate the projected volume of larch log use, we used data from the National Survey on Timber Utilization (Korea Forest Service 2011– 2023) and conducted statistical analyses using IBM SPSS Statistics (version 29.0.1.0, build 171).

The 7th National Forest Inventory (Korea Forest Service 2024b) was used to estimate the growing stock volume of larch forests in the base year of 2020. The periodic growth increment by age class was estimated based on the 2023 National Stand Volume, Biomass, and Stand Harvest Tables (Korea Forest Service and National Institute of Forest Science 2023).

Estimation of resource area

In recent forest resource management and monitoring applications, geographic information systems (GIS) have emerged as core analytical tools enabling spatial data to be used to quantify forest area, location, and growing stock. In particular, GIS is widely applied for the comprehensive spatial analysis of stand structure, zoning function, and age class information. For instance, GIS has been effectively used to delineate forestland areas and evaluate resource volumes based on forest type or management objectives (Kim et al. 2000, Park and Lee 2018). Chong (2007) used GIS to classify six types of forests, including natural and artificial stands such as pine, oak, and larch, and calculated the area of each forest type.

Following a similar approach, this study utilized the Forest Function Classification Map and applied QGIS filters to differentiate between primary and secondary forest functions. This enabled the extraction of forested areas designated as timber-production forests. In the species-based forest type map, the areas of larch-dominated forests were extracted and classified according to age.

Within the primary functional forests, the spatial extent and age distribution of larch-dominated timber-production forests were validated through a visual overlay procedure and finalized through spatial correction. In this way, the final area of larch forests within the timber-production zones was established. Age-class layers were revised, and larch forests with a harvesting age of 30 years or older were defined as mature stands (Korea Forest Service 2024b). Age classes were categorized into 5-year intervals based on stand age as follows: Class I (≤10 years), Class II (11 to 20 years), Class III (21 to 30 years), Class IV (31 to 40 years), Class V (41 to 50 years), and Class VI or higher (≥51 years).

Accordingly, as illustrated in Figure 1, the spatial extent of mature larch forests within the timber-production zones was calculated.

View Figure

• Download as Powerpoint
• Download Figure

Average growing stock volume and growth estimation

To accurately analyze the standing timber resources of domestic larch forests in South Korea, we estimated the total and per-hectare growing stock volume by age class using age-class forest area data and stock volume information extracted via QGIS.

The total stock volume (T_total) was computed using the 7th National Forest Inventory, which provides the growing stock volume and forest area of larch forests by age class. However, the per-hectare growing stock volume (m³/ha) is not directly published, and thus it had to be calculated separately.

The average growing stock volume per hectare for each age class c was calculated according to Equation 1: (1) $$T_{\mathit{total}}^{(c)}=\frac{S_{\mathit{total}}^{(c)}}{A_{\mathit{total}}^{(c)}}$$

The initial growing stock volume for each age class, S^(c)(0), was then given by Equation 2: (2) $$S^{\left(c\right)}\left(0\right)=A^{\left(c\right)}·T_{\mathit{total}}^{\left(c\right)}$$

The periodic annual growth volume, g^(c)(t), based on the legal stand growth table for Site Index 16, was estimated as: (3) $$g^{(c)}(t)=\{\begin{array}{ll}\hfill A^{(c)}·PG{I}_t·5,& if5\mathit{divides}t\mathit{and}t>t_0^{(c)}\\ \hfill 0,& \mathit{otherwise}\end{array}$$where $t_0^{(c)}$ = 10(c − 1), and PGI_t is the 5-year periodic growth increment provided in Table 2.

Table 2.Periodic growth increment by 5-year age intervals for larch forests.

View Fullscreen

Accordingly, the cumulative growing stock volume at time t is expressed: (4) $$S^{(c)}(t)=S^{\left(c\right)}(0)+{\sum }_{k\le t{g}^{(c)}}(k)$$where the summation is performed over all valid intervals k up to time t. This framework enables a concise and systematic quantification of dynamic stock changes by age class, considering only the class area, per-hectare stock, and age-dependent periodic increments. However, it should be noted that this approach relies on the assumption of a constant forest area per age class over time, which may limit the applicability of the results under scenarios of land-use change or natural disturbance such as fire or pests.

The following notation refers specifically to domestic larch forests in South Korea:

• c: age class at the initial point in time (e.g., Class I to VI or higher),

• t: time in years, measured in 5-year increments (age),

• k: cumulative number of 5-year intervals (steps),

• total: entirety of the forest area,

• $S_{\mathit{total}}^{(c)}$: total growing stock volume (m³) of age class c,

• $A_{\mathit{total}}^{(c)}$: total forest area (ha) of age class c,

• $T_{\mathit{total}}^{(c)}$: average growing stock volume per hectare (m³/ha) of age class c,

• $A^{(c)}$: area (ha) of timber-producing forests in age class c,

• $S^{\left(c\right)}(0)$: growing stock volume (m³) of timber-production forests in age class c at the initial time t₀,

• $S^{\left(c\right)}(t)$: growing stock volume (m³) of age class c at time t,

• $g^{\left(c\right)}(t)$: 5-year periodic growth increment (m³) in age class c at time t, and

• $PG{I}_t$: average annual growth increment (m³/ha/yr) between time t − 5 and t.

To capture the dynamics of stock volume change in stands over 30 years of age, the increase in growing stock was divided into two components. The “New entry volume (≥30)” represents stands that have just matured into the ≥30 age class, while the “Growth from existing ≥30” reflects the net natural increment within already mature stands. This distinction enabled estimation of volume changes and assessment of each component’s relative contribution, providing a basis for both short-term projections and long-term evaluation of forest resource availability.

Forecasting domestic larch log demand

To predict long-term changes in domestic demand for larch logs, this study employed an ARIMA time-series forecasting model. ARIMA is a widely used statistical approach that identifies patterns in historical time-series data and forecasts future values based on these trends. It incorporates three core components: autoregression (AR), differencing (D), and moving average (MA), making it a representative tool for demand forecasting based on past behavior (Box et al. 2015, Jing 2022, Gopu et al. 2023).

Previous studies have successfully applied ARIMA models to forest resource statistics, including forecasts of timber production, export, and consumption volumes, based on historical time-series data. This includes analyses of export trends in wood product categories such as veneers, fiberboard, and wood-based panels (Upadhyay 2013, Liu et al. 2022, Varshini et al. 2024).

In line with these prior studies, we conducted time-series forecasting of domestic larch log demand based on past utilization data. In South Korea’s forest management context, the supply volume of domestic timber, collected primarily through logging permits and other official authorizations, is presented as species-specific log supply data (Korea Forest Service 2023a). To reflect practical resource utilization rather than licensed harvest supply data, this study used the actual quantity of larch log purchases by product type in the timber-production industry as a proxy indicator (Korea Forest Service 2024a).

ARIMA (p, d, q) models were applied to the larch log utilization data from 2010 to 2022 shown in Table 3. To verify model suitability, time-series stationarity was assessed using the augmented Dickey-Fuller (ADF) test, and autocorrelation structures were further examined through the autocorrelation function (ACF) and partial autocorrelation function (PACF) analyses. Multiple evaluation criteria were used for model selection, including Bayesian information criterion (BIC), root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE), and the coefficient of determination (R²) (Seo and Chung 2024). Also, to examine whether the assumption of linearity was violated, regression diagnostic tests were conducted.

Table 3.Annual larch log usage by wood product category in Korea.

View Fullscreen

The dataset used in this study is annual in frequency and relatively short in length, with trends characterized by gradual long-term changes rather than abrupt nonlinear fluctuations. Under these conditions, linear time-series models were judged to be more stable and interpretable. In contrast, nonlinear or machine learning approaches often require longer time series or higher-frequency data to achieve reliable performance, conditions that were not met in this study. Considering these factors, the ARIMA model was deemed more appropriate.

Based on this analysis, ARIMA (5,1,1) was selected as the optimal specification, and it was applied to forecast the annual domestic demand for larch logs out to 2060. Nevertheless, long-term projections extending to 2060 inevitably involve uncertainty, and ARIMA models may accumulate error and bias over extended horizons. To address this, the study did not rely on a single trajectory; instead, forecast uncertainty was quantified by constructing 95 percent confidence intervals, with the upper confidence limit used as an indicator of the potential upper bound of demand. This approach was designed to ensure that sustainability assessments could be interpreted within a more realistic range.

Scenario analysis

We evaluated the long-term supply sustainability of larch resources aged 30 years or older in timber-production forests. The evaluation was based on a time-series forecast of the growing stock and larch log demand projections obtained from ARIMA modeling.

At time step $t_0+5n$, the remaining available resources $R(t_0+5n)$ were calculated by adding the growth volume $G(t_0+5n$) to the existing stock $R(t_0+5[n-1])$ and then multiplying the total by the bucking rate (b = 0.80) to determine the practical availability. From this practical availability, the scenario-based projected demand $U(t_0+5n)$ was subtracted to obtain the remaining net resources. If the result was less than or equal to zero, the state of resource depletion was concluded.

This threshold represented the condition for resource depletion and served as a decision criterion for judging sustainability in the subsequent interpretation of the results.

The equation used for this scenario evaluation is expressed as follows: (5) $$R(t_0+5n)=\mathit{max}\left\{0,(R(t_0+5(n-1))+G(t_0+5n))·b-U(t_0+5n)\right\}$$

The scenario-based growth volume, denoted by $G(t_0+5n)$ and defined in Equation 6, was calculated by combining the new entry volume, E, with the contribution from the remaining resources in the previous 5-year step $R(t_0+5\left(n-1\right))$, scaled by the growth contribution ratio $p(t_0+5n)$. (6) $$G(t_0+5n)=(E(t_0+5n)+R(t_0+5(n-1))·p(t_0+5n)$$

The projected log demand $U(t_0+5n)$ was computed as the accumulated volume over the preceding 5-year period using ARIMA model forecasts. To reflect the uncertainty in the demand estimation, three scenarios were defined: a base scenario using the mean forecast $U^{\mathit{base}}(t_0+5n)$, a maximum scenario using the upper confidence limit ( $U^{\mathit{UCL}}(t_0+5n)$), and a minimum scenario using the lower confidence limit $U^{\mathit{LCL}}(t_0+5n)$. The accumulated demand was calculated using Equation 7. (7) $$U(t_0+5n)={\sum }_{i=0}^{4}u(t_0+5n-i)$$where

• $t_0$: the initial reference year of the analysis (set to 2020 in this study),

• $R(t_0+5n)$: remaining standing volume (m³) of larch resources aged ≥30 years at time ( $t_0+5n$),

• $G(t_0+5n)$: scenario-based growth volume (m),

• $E(t_0+5n)$: new entry volume (m³) at time ( $t_0+5n$),

• $p(t_0+5n)$: contribution ratio (%) of growth from existing resource stock,

• b: bucking rate (set to 0.80 in this study),

• i: an index representing the position of each year within the 5-year cumulative demand window, and

• $U(t_0+5n)$: accumulated 5-year demand forecast from the ARIMA model, calculated as the sum of annual demands $u(t_0+5n-i)$ for i = 0, 1, 2, 3, 4. Three demand scenarios were constructed: $U^{\mathit{base}}(t_0+5n),U^{\mathit{UCL}}(t_0+5n),\mathrm{and}{U}^{\mathit{LCL}}(t_0+5n)$.

This scenario-based approach allowed for a quantitative assessment of the ways in which varying levels of projected log utilization affect the risk of resource depletion, providing a foundation for evaluating the sustainability of forest resource management under both optimistic and conservative assumptions (for a schematic overview, see Fig. 2).

View Figure

• Download as Powerpoint
• Download Figure

Estimation of resource extent

Figure 3 illustrates the spatial distribution of timber-production and larch forests across the Republic of Korea, with a specific focus on stands aged 30 years or older (i.e., age Class IV and above). This figure shows the area potentially available for utilization and helps to clarify the scope of spatially distributed larch resources.

View Figure

• Download as Powerpoint
• Download Figure

The spatial distribution of timber-production forests totaled approximately 2,144,580 ha (Fig. 3a). This forms a foundational reference for setting boundaries for future resource assessments and enables sustainable forest resource management in South Korea (Kwon et al. 2024). Specifically, larch forests covered approximately 258,400 ha (Fig. 3b). The breakdown by age class was as follows: Class I, 4,395 ha; Class II, 2,395 ha; Class III, 10,185 ha; Class IV, 37,627 ha; Class V, 9,434 ha; and Class VI or higher, 445 ha. Compared to the estimated 440,000 ha of larch forest area in 2005, which was based on designated economic forest units across national and private lands, this represents a substantial decrease (Chong 2007).

Figure 3c shows the area of larch forests within timber-production forests, totaling 64,482 ha. The age class distribution was as follows: Class I, 4,395 ha; Class II, 2,395 ha; Class III, 10,185 ha; Class IV, 37,627 ha; Class V, 9,434 ha; and Class VI or higher, 445 ha. Stands aged 30 years or older (i.e., age Class IV and above) were used as the threshold for defining mature larch forests, totaling 47,507 ha, representing approximately 73.7 percent of the total larch forest area and 18.4 percent of the larch forests within timber-production forests.

This classified age-based area was established as the reference dataset for the growing stock and growth estimation of larch forests used in this study’s resource analysis and scenario modeling.

Results of growing stock and growth projection

Table 4 presents the forest area and growing stock volume of the larch forests by age class, along with the average growing stock per hectare [ $T_{\mathit{total}}^{\left(c\right)}$]. The results show a clear upward trend in growing stock per hectare with increasing age class: Class II, 43 m³/ha; Class III, 163 m³/ha; Class IV, 218 m³/ha; Class V, 272 m³/ha; and Class VI or higher, 341 m³/ha. This trend reflects the cumulative nature of forest growth, indicating that the higher age classes produce higher productivity per unit area (Korea Forest Service 2025a, 2025b).

Table 4.Larch forest area by age class.

View Fullscreen

Table 5 presents the forest area and growing stock volume by age class at the initial time point and projects the accumulation of growing stock over 40 years at 5-year intervals. The initial growing stock volume [ $S^{\left(c\right)}(0)$] in timber-production forests was estimated at 12,683,619 m³. Age Class IV accounted for the highest area, at 8,202,686 m³ (64.7%), followed by Class V (2,566,048 m³), Class III (1,660,155 m³), Class VI or higher (151,745 m³), and Class II (102,985 m³). This distribution reflects the effects of intensive afforestation during the 1970s and the 1980s, which resulted in a structural concentration of growing stock in age Classes IV and V, suggesting that timber-production forests in Korea are highly concentrated in harvestable age groups (Ahn et al. 2019).

Table 5.Projected cumulative changes in stand volume (m³) by age class over 40 years in timber-production forests^a.

View Fullscreen

The total growing stock volume in timber-production forests is projected to reach approximately 21,255,274 m³ by year 40, reflecting a 67.7 percent increase. Within this, the volume held by stands older than 30 years is projected to increase significantly from 10,920,479 m³ to 21,255,274 m³, accounting for 94.8 percent of the total increase. This indicates that stands exceeding 30 years of age will dominate more than half of the projected growing stock, which reflects the emergence of a mature forest structure. Notably, Class IV, which initially accounts for approximately 60.6 percent of the total volume, will age into a 70-year-old cohort within 40 years, potentially exacerbating the structural imbalance in forest age distribution (Kwon et al. 2024).

Table 6 presents the total stock volume increase, new entry volume, and forest growth volume for stands aged over 30 years, modeled over successive time steps. The overall increase in growing stock peaked in year t₀ + 10, reaching 3,044,714 m², after which a gradual decrease was observed.

Table 6.Decomposition of stock volume (m³) increase (≥30 years) into new entry and growth from existing larch forests over time.

View Fullscreen

New entries occurred at only three time points: 2,178,572 m² at t₀ + 10, 369,070 m² at t₀ + 20, and 677,115 m² at t₀ + 30. These correspond to the model’s minimum age requirement for new cohort entry (≥ 30 years), which was satisfied at that time. In contrast, for t₀ + 15, t₀ + 25, and t₀ + 35, the age criterion was not met, and thus, no new entries occurred. By t₀ + 40, zero new entries were projected due to the model’s exclusion of reforestation beyond that point.

In contrast, growing stock derived from existing stands was observed across all time steps but showed a continuous decline. At t₀ + 5, forest growth contributed 965,400 m² (8.84% of the total increase), which subsequently decreased as follows: 866,142 m² (7.29%) at t₀ + 10, 1,000,262 m² (6.70%) at t₀ + 15, 911,137 m² (5.72%) at t₀ + 20, 887,255 m² (5.16%) at t₀ + 25, 821,642 m² (4.54%) at t₀ + 30, 857,637 m² (4.38%) at t₀ + 35, and 800,563 m² (3.91%) at t₀ + 40.

These findings indicate that as the simulation progressed over time, the relative contribution of existing forests to the total stock increase diminished. This is particularly evident at later time points, when the lack of new entries and the aging of existing cohorts begin to constrain productivity. Consequently, the overall rate of forest stock accumulation declined, revealing a clear stagnation trend. These outcomes underscore the necessity for systematic harvesting and regeneration efforts for overaged stands to mitigate structural imbalances and ensure a sustainable forest resource cycle (Ryu et al. 2016).

Based on this analysis of growing stock dynamics, the next section examines projected timber demand in detail, in order to assess the long-term sustainability of resources across the time series.

Timber demand forecasting

Stationarity and residual diagnostics were conducted to evaluate the appropriateness of the ARIMA model. The initial ADF test results indicated the presence of unit roots at the 5 percent significance level (p = 0.354), and autocorrelation was observed in both ACF and PACF plots, necessitating first-order differencing. Following first-order differencing, the ADF test (p = 0.004) and both ACF and PACF plots showed no evidence of autocorrelation (Fig. 4), thereby establishing a stationary time series.

View Figure

• Download as Powerpoint
• Download Figure

In terms of predictive performance, while information criteria remained relatively stable across varying autoregressive orders, the optimal performance was achieved at AR order 5, where MAPE (7.64%) and MAE (32,059.28) reached their lowest values, and R² (0.525) attained its highest level. Conversely, increasing the moving average order resulted in a sharp increase in RMSE, introducing unnecessary noise without improving model fit. Consequently, the moving average order was set to 1, corresponding to an RMSE of 59,256.28. As illustrated in Figure 5, both ACF and PACF plots demonstrated the absence of autocorrelation. Furthermore, regression diagnostic tests confirmed that the null hypothesis could not be rejected, indicating no violation of linearity assumptions (p = 0.172), and the annual frequency data showed no evidence of seasonality. Collectively, these findings justify the adoption of the ARIMA(5,1,1) model, which optimally balances predictive accuracy with model parsimony.

View Figure

• Download as Powerpoint
• Download Figure

These findings align with those of previous studies predicting export volume and timber yield using ARIMA-based methods, where similar metrics (R², BIC, and MAPE) were used to validate the model performance (Upadhyay 2013, Guo 2024). Accordingly, this study applied the model to project larch timber consumption from 2023 to 2060.

Figure 6 illustrates the time-series projection of the domestic larch timber demand using the ARIMA model. The forecast shows a gradual upward trend. Demand is projected to reach approximately 560,000 m³ by 2030 and exceed 800,000 m³ by 2050. This indicates a long-term increase in demand, driven by expanding industrial use and improved resource utilization.

View Figure

• Download as Powerpoint
• Download Figure

The forecast graph also includes 95 percent confidence intervals represented by the upper confidence limit (UCL) and lower confidence limit (LCL) bounds. The future demand is expected to fall within these bounds at the 95 percent confidence level. For example, the 2030 projection of 560,000 m³ includes a UCL of approximately 720,000 m³ and LCL of 400,000 m³. These intervals reflect forecast uncertainty and can serve as critical references for policy development and long-term resource planning (Yoo et al. 2013, Fattah et al. 2018, Jeong and Kim 2023).

Based on the forecasted values and 95 percent confidence intervals, this study defined three demand scenarios for larch timber. The baseline scenario was established using the predicted value, whereas the high-demand scenario adopted the UCL, and the low-demand scenario used the LCL. These scenarios were designed to support strategic resource-use planning by accounting for variability and uncertainty in future timber demand, thereby providing a framework to evaluate sustainability under different demand conditions.

Scenario analysis results

Figure 7 presents scenario-based projections of future timber demand, assuming the base year (t₀) as 2020, using the ARIMA-based forecast results (base, UCL, and LCL). This study assessed the sustainability of larch resources over 30 years of age in timber-production forests under three different demand scenarios. For each 5-year interval, the available resource volume was calculated by adding growth increments to the standing volume and subtracting the projected cumulative utilization. If the resulting value dropped below zero, the time point was considered to indicate resource depletion.

View Figure

• Download as Powerpoint
• Download Figure

1. Under the base scenario, which applied the central forecast values from the ARIMA model, the analysis suggests that peak resource availability will occur by approximately 2030. While growth and utilization remain balanced in the early period, cumulative utilization eventually surpasses growth, leading to resource exhaustion. In particular, larch stands reforested after 2020 will not reach the minimum rotation age until 2050, resulting in an anticipated supply gap of about 5 years. To bridge this gap, the immediate deployment of engineered wood products using domestic substitute species and wood-based materials is required. Strategies include accelerating the development of hybrid structural products incorporating species such as Korean red pine (Pinus densiflora) and Hinoki cypress (Chamaecyparis obtusa), as well as structural plywood and particleboard, while strategically reducing low-value uses such as fuelwood and pulp (Fujimoto et al. 2021, Lee et al. 2023, Yang et al. 2023).

2. Under the high-demand scenario (UCL), which reflects the upper limit of the 95 percent confidence interval for the utilization forecast, resource depletion is estimated to occur in 2040. During the first 10 years, the growth volume either surpasses or is comparable to the utilization volume, maintaining a relatively stable resource stock. However, from 2035 onward, the utilization volume exceeds the growth volume, resulting in a sharp decline in available resources and complete depletion by 2040. This suggests that under high-demand conditions, the current larch timber resource recycling system cannot maintain a sustainable supply. Given the 30-year rotation requirement, short-term reforestation cannot offset demand, resulting in a severe supply gap of about 10 years. Accordingly, urgent measures are needed, including diversification of timber imports, expedited approval of foreign species with equivalent structural properties, establishment of a strategic timber stockpiling system, and regulatory reforms to expand the application of substitute species in construction (Cheng et al. 2015).

A recent study by Subramanian et al. (2025) found that wood demand from Swedish forests increased across all policy scenarios, with the Bioeconomy scenario particularly demonstrating that total demand surpassed the sustainable supply potential by 2070, leading to the unsustainable practice of utilizing sawn timber and pulpwood for energy purposes.

3. Under the minimum scenario (LCL), which assumes the lowest timber demand level within the 95 percent confidence interval, resource depletion is projected to be extended to 2055. During the early and middle periods, the 5-year average growth volume either exceeds or remains at levels similar to the utilization rate, resulting in a relatively stable stock volume throughout the timeline. In particular, until 2035, growth contributions remain effective, maintaining stock volumes above a certain level, with no abrupt depletion. Moreover, larch stands reforested after 2020 will reach the minimum rotation age by 2050, thereby preventing supply shortages. However, this continuity depends on sustained reforestation aligned with projected demand. To ensure long-term sustainability, adaptive management strategies are essential, including the establishment of annual reforestation quotas based on long-term projections, adjustment of planting schedules in response to demand fluctuations, and the integration of multifunctional forest management with carbon credit schemes and ecosystem service objectives (Carle and Holmgren 2008).