Carbon stock in Japanese forests has been greatly underestimated

National forest Inventory based on direct field measurements: m-NFI

In one campaign of the current m-NFI survey, direct field measurements were conducted for 5 years at more than 10,000 plots throughout Japan²⁰. These plots formed a 4-km grid interval across Japan. Measurements were conducted at given plots including forests. The third m-NFI survey was conducted in 2009–2013 (Fig. 1); the data can be downloaded from the Japan Forestry Agency website (http://www.rinya.maff.go.jp/j/keikaku/tayouseichousa/). The number of planned plots was initially 14,838; actual forest plots were then selected, which resulted in 14,719 plots (14,522, if eliminating bamboo plots). Practically, however, the number of the target plots was 13,357 because of their accessibility.

Each plot comprises triple concentric circles with areas of 0.01, 0.04 and 0.1 ha. In the largest (0.1 ha), medium (0.04 ha) and smallest (0.01 ha) circular areas, the DBHs of trees ≥18, 5 and 1 cm DBH, respectively, are measured. Tree height is measured for each of the selected 20 trees in the plot and tree heights of the remaining trees are estimated using the DBH–tree height relationships derived from the 20 tree measurements. Stem volume of a single tree is calculated from the empirical relationship between DBH and tree height for each tree species and its growing region^54,55. Note that the publicly available data do not include individual tree scale information as mentioned above, but rather the plot scale totalized information.

This study used only the published data, that is, stem volume per hectare (Vs; m³ ha⁻¹), name of dominant tree species and forest age for a given plot. Here, the dominant tree species denotes a tree species with the largest total cross-section area at breast height in the plot. The total number of plots used for the survey was 12,457 because we excluded 703 plots in which not all of the three types of the data were identified and 197 bamboo plots for which the C stocks were difficult to estimate. Measurement validation commenced in 2010. The annual mean differences in Vs between the practical measurements and the validation surveys were 4%–8% in 2010–2013²⁹.

National forest inventory based yield table predictions: p-NFI

The yield tables used in p-NFI show empirical relationships among stand-scale stem volume, age class (ranked from 1 at 5-year intervals) and land-productivity index for each major tree species. Although the land-productivity index is critical for estimating the stem volume in a given stand, no clear standard existed for many target forest stands on how the land-productivity index was determined. As far as we investigated, no one in national and local forestry agencies knows how to revise the land-productivity index, which can be affected by various environmental changes, which are a large source of error for the stem volume estimation. Note that the currently-used yield tables were constructed using the actual measurements from around 1970.

Yield tables for Japan differ depending on the type of forest-ownership. For national forests (ca. 30% of the total forested area), yield tables created by the national government have been used. However, for local-government-owned and private forests (ca. 70%), each of 47 prefectural governments to which the given forests belong created its own yield tables. Surprisingly, most yield tables have not been published and even the national forestry agency, that is, the Japan Forestry Agency, cannot know all the information on their yield tables. The Japan Forestry Agency has aggregated the data submitted by administrators for national and prefectural forests and published estimations of the stem volume with some relevant information (http://www.rinya.maff.go.jp/j/keikaku/genkyou/index1.html).

The forests were categorized into two types in p-NFI: plantations and natural forests. Data for plantations comprises stem volume and forest area, which are further categorized into 10 forest types, 47 prefectures and 19 age classes. Data for natural forests include only stem volume categorized into 11 forest types and 47 prefectures. Notably, no information on forest area for each forest type is included for natural forests.

Carbon density estimation using biomass expansion factor (BEF)

BEF is the coefficient derived from allometry measurements, which enables us to convert Vs to the stand scale C stock. BEF has several definitions, and one type of BEF calculates total biomass (sum of stems, branches, leaves and roots) and another type calculates aboveground biomass (sum of stems, branches and leaves). Previous studies in Japan have used both types of BEF; here, we refer to these as BEF₁ and BEF₂. BEF₁ was suggested in a study examining long-term changes in Japanese forest C stocks²⁵ We used the past forest C stock values that were calculated with BEF₁ in a previous study²⁵; therefore, BEF₁ is indispensable for the present study. BEF₂ was used in the official reports from the Japanese government⁵⁶ and is probably a more commonly used BEF in Japan than BEF₁. Therefore, we estimated all forest C stocks with BEF₁ to clarify the temporal changes in forest C stock and used both BEFs for m-NFI in 2009–2013 to quantify the uncertainty derived from the difference in BEF. The conversion methods using these BEFs are shown below.

BEF₁ (Mg m⁻³) is for the conversion from Vs to total biomass per hectare and was presented for 10 plantation-forest types in p-NFI. According to the evidence that BEF₁ tends to increase with decreasing Vs, the following equation is reached:

where a (Mg m⁻³) and b (Mg ha⁻¹) are empirical parameters obtained for a given forest type. In the case that Vs is extremely small in Eq. 1, BEF₁ becomes unrealistically large. Therefore, we set an upper limit of 1.5 for BEF₁ according to data obtained in a previous study²⁵. In addition, in previous studies using BEF₁, C content was defined as 0.5²⁵. As a result, forest stand scale C density (Cd; MgC ha⁻¹) is given as follows:

BEF₂ (unitless) is the conversion factor to expand stem biomass to the aboveground biomass (that is, stems, branches and leaves) and is presented for two age classes (below and above 20 years old) of 32 tree species, some alien species, other conifers and other broadleaf trees (see Table S1). Cd using BEF₂ can be written as follows:

where WD is woody density (Mg m⁻³) and R (unitless) is the ratio of root biomass to aboveground biomass for a given tree species. Stem biomass ((V{rm{s}}times {WD})) is converted into aboveground biomass by BEF₂, then multiplied by ((1+R)) to calculate total biomass. α is C content: 0.51 and 0.48 for conifers and broadleaf trees, respectively⁵⁶.

Carbon stock estimation using m-NFI

The m-NFI-derived C stocks in 1961 and 1966 have been already reported in a previous study²⁵; therefore, we estimated C stock for m-NFI in 2009–2013. Again, the numbers of actual forest and bamboo plots were 14,522 and 197, respectively. We calculated Cd at 12,457 of the 14,522 forest plots using both BEF₁ and BEF₂, which were determined for the dominant tree species at the plot. We then obtained plot-averaged Cd for each of the 11 forest types categorized in p-NFI. The total forest C stock was estimated as a total of each forest type C stock, which was calculated by multiplying the forest type averaged Cd and forest type area for the 14,522 forest plots.

Surveyors could not reach or tree species could not be identified in 2,065 of the 14,522 forest plots. We assumed the 2,065 plots to be natural forests but could not apply a forest type (using the p-NFI categories). This means that while we could estimate forest type areas for 12,457 observation plots, those for 2,065 plots are unknown, which resulted an incorrect total forest type area in the 14,522 total forest plots (see Table S2).

Therefore, it is necessary to estimate the forest type areas for these 2,065 plots and the total for the 14,522 plots (Table S2). Assuming that the areas of each forest type in the p-NFI natural forest category are proportional to each forest type’s estimated stem volume in the natural forest category in the 2012 p-NFI, we assigned 11forest types to the 2,065 plots. The sum of the assigned plots for each forest type in the 2,065 non-observation (natural forest) plots and the 12,457 observation plots can be assumed to represent the area ratio of each forest type in the 14,522 total forest plots (Table S2). The forest area of each forest type can be calculated by multiplying Japan’s total forest area and this area ratio. Using aerial photographs, satellite images and field surveys, Japan’s total forest area was estimated to be 25.86 million ha (http://www.rinya.maff.go.jp/j/keikaku/tayouseichousa/tikuseki.html). The ratio of the bamboo forest plots to the actual forest plots (~1%) was subtracted from the total forest area, leaving 25.6 million ha as the total forest area.

Carbon stock estimation using p-NFI

The p-NFI-estimated C stocks until 1995 have been already reported in previous studies^25,28; therefore, we estimated C stock for the subsequent three p-NFIs in 2002, 2007 and 2012 using the same method as the previous studies. We calculated Vs at plantations by dividing the p-NFI-estimated stem volume by the forest area for each factor, that is, prefecture, forest type and age class. Then, the Cd was obtained by using the BEF₁ calculation with the Vs (see Eqs. 1 and 2). The C stocks at plantations were estimated by multiplying the Cd and the forest area for each factor and summing the products.

No information is given in the p-NFI natural forest category about forest area; therefore, we estimated the C stock in natural forests without using the Vs. We first obtained plantations’ BEF₁ values for each prefecture and forest type and used these in place of those for natural forests. Then, the C stock in natural forests was estimated by multiplying the BEF₁ and the total stem volume for each prefecture and forest type in natural forests and summing the products. Japan’s total forest C stock is determined by the sum of the C stocks in plantations and natural forests.

Calculation of 95% confidence interval (CI)

We calculated the 95% CI for m-NFI-derived C stocks. No information on the C stock uncertainty was available in the previous study, which used the 1961 and 1966 m-NFIs to calculate the forest C stocks²⁵. Thus, we assumed that the widths of the 95% CIs of the total forest C stocks calculated from the m-NFI of 1961 and 1966 were equal to that of total stem volume in the 1961 m-NFI, i.e., 5.4% of the mean value⁵⁷. For m-NFI in 2009-2013, the 95% CI of the area-based C stock was calculated for each of the 11 forest types (Table S2) as follows:

where ({varepsilon }_{i}), ({sigma }_{i}), and ({n}_{i}) represent the half width of 95% CI (MgC ha⁻¹), standard deviation (MgC ha⁻¹), and sample size of area-based C stock of forest type i, respectively. Then the half width of 95% CI of total forest C stock (({varepsilon }_{{rm{T}}}); MgC) can be calculated with the forest area of forest type i (({A}_{i}); ha) as follows:

Reproducing past carbon stocks of Cr. japonica and Ch. obtusa

Assuming that for a given forest type i (such as Cr. japonica) the net forest C uptake flux (F_i; MgC ha⁻¹ year⁻¹) is proportional to the area-based C stock (V_i/A_i, in which V_i and A_i are C stock (MgC) and forest area (ha), respectively), the net C uptake of the forest type i (F_i × A_i; MgC year⁻¹) may be approximated as follows:

where F_T and V_T are net total forest C uptake (MgC year⁻¹) and total forest C stock (MgC), respectively. Thus, the past C stock of i at the year Y₁ (Cst_{Y1, i}; MgC) can be calculated as:

where Y₀ is the reference year and Cst_{Y0, i} is the C stock (MgC) at Y₀.

Source: Ecology - nature.com