Introduction

Oak is remarkably important for forestry in Turkey, which is one of the world’s most important regions for oak in terms of both the area covered and richness of species. After pine forests, oak forests comprise the next largest group in Turkey’s forest areas. Oak forests have a 26% share in the general field of forest assets. A quarter of Turkey’s forest area consists of oak forests (OGM 2015). Except for the steppe regions, oaks have spread to almost every region of Turkey, including maquis and agricultural lands, where these trees form pure high and pure coppice forests and often mix with other tree species (Atay 1987).

Coppice method is an ancient silviculture practice for sustainable forestry. Stump sprouts or root suckers are the main sources of the stand regeneration with rotated clear-cuts (Smith et al. 1997). Coppice management is not suitable for the realization of a forest’s multiple functions, causing the degradation and depletion of forests over time (Makineci et al. 2015). Clear-cuts and rotations under the coppice management of oak forests over a span of decades can negatively affect the forest ecosystem. In Turkey, coppice forests comprise a considerable area of the total forest land, and a large portion of these forests is degraded (Makineci et al. 2011). For this reason, the conversion of coppice forests to high forests was considered an ecological and economic necessity, in addition to being essential for sustainable forestry, and the rehabilitation of damaged oak forests was planned. At the beginning of 2006, the Turkish General Directorate of Forestry decided to abandon coppice management and the conversion of coppice forests to high forests (OGM 2006a, b). The amount of carbon that is sequestered in these forests must be determined so that the contribution from the wide range of oak forests to global climate change can be estimated. For this reason, determining the amount of biomass in these forests is especially important.

In northern Turkish Thrace, where this study was conducted, oak is spread over large areas. Pure oak forests comprise 72% of the Thracian forest area. If mixed stands are included, this rate is even higher. Sessile oak (Quercus petraea (Matt.) Liebl), Hungarian oak (Quercus frainetto Ten), and Turkish oak (Quercus cerris L.) species are widespread in high forests and productive oak coppices (Makineci 2005). Large areas of Thracian forests have been destroyed because of their geographical features, and approximately 90% of former deciduous high forests have taken the form of degraded coppices, while aged trees are rarely found (Makineci et al. 2011). Most of these forests have also been managed by traditional coppice management. The different silvicultural systems in a particular region have special importance in the regulation of biomass production. In this context, recent studies on deciduous forests in Europe have shown increasing interest in forest biomass as a potential renewable energy source (Hochbichler 2008; Hochbichler et al. 2009).

Tree-biomass data can be an important tool to identify carbon stocks that are sequestered by forests (Cienciala et al. 2008) and determine the bioenergy capacities of forests. Individual tree-biomass equations are frequently used for estimating and determining individual tree-biomass values. These equations estimate the biomass components as dry mass based on the diameter at breast height (DBH) and tree height, which are generally easily measurable variables. Other variables such as the crown height, crown width, or tree age can also be used as variables to predict tree biomass (Cienciala et al. 2008).

Biomass equations that are developed from single and mixed species can generate useful estimates for specific site conditions and large-scale regional and global comparisons. Such equations are important for forest inventory data and can provide more accurate estimates of species biomass (Litton and Kauffman 2008; Návar 2009).

Researchers have developed equations for biomass components in different oak species with different modeling methods (Carvalho and Parresol 2003; Cienciala et al. 2008; Návar 2009; Suchomel et al. 2012; Zribi et al. 2016; Vargas-Larreta et al. 2017; Husmann et al. 2018). Very few studies that were concerned with oak biomass have been conducted in Turkey, even though oak species comprise around one-quarter of the country’s forestland (Durkaya 1998; Makineci et al. 2011).

The aims of this study are to identify the different biomass components (foliage, branch, bark, stem, and total aboveground biomass) of the main oak species Q. frainetto (QF), Q. cerris (QC), and Q. petraea (QP) in the oak coppice forests of northern Thrace and develop individual tree-biomass equations.

Materials and methods

Study site

This study was conducted on coppice-originated oak forests with different development stages in seven different regions, including various geological parent materials and meteorological characteristics, at a forest site in northern Turkish Thrace (Fig. 1). The climatic parameters were evaluated from long-term data from the Turkish Meteorological Service. We also used regional meteorological estimations by Irmak et al. (1980) for the same research area (Table 1) (Makineci et al. 2011).

Fig. 1
figure 1

Research area on seven different regions in Turkish Thrace 1: Demirkoy1 (D1), 2: Demirkoy2 (D2), 3: Demirkoy3 (D3), 4: Demirkoy4 (D4), 5: Kırklareli, 6: Vize, and 7: Catalca

Full size image
Table 1 Site characteristics of research area (Makineci et al. 2011, 2015)
Full size table

Biomass sampling

We selected sample plots in each region and on each substrate materials according to the development stages. Six replicated sample plots (20 m × 20 m) in each development stage could be classified according to the Turkish Forest Service’s designations (degraded: canopy cover less than 10%; SDF (small-diameter forest): mean DBH of 0–8 cm; MDF (medium-diameter forest): 8–20 cm; and LDF (large-diameter forest): 20–36 cm).

The selected sample plots, despite being labeled as oak in forest management plans, mainly included three oak species: Q. frainetto (QF), Q. cerris (QC), and Q. petraea (QP). The diameter at breast height and height of all the trees in each sample plot were measured. In each plot, one tree for each species with a mean DBH was sampled. Thus, 124 QF, 116 QC, and 234 QP (totaling 474) sample trees were destructively harvested. The DBH of the sample trees ranged from 0.2 to 38.5 cm, the total height was between 1.4 and 23.6 m, and the ages ranged from 4 to 169 years (Table 2).

Table 2 Descriptive statistics of sample trees in biomass estimation
Full size table

After harvesting, leaves, branches, and stems were separated from the sampled tree and weighed. All the sampled branches were live; no dead branches were observed on the sampled trees. Subsamples of stem disks with at least 5-centimeter thickness were taken from sections that were divided with 2-meter intervals. All subsamples of biomass components (leaves, branches, and stem disks) were promptly weighed in the field. In the laboratory, the leaf, branch, and stem-disk subsamples were constantly dried at 70 °C to measure the moisture content. Bark was removed from the disks, and the biomass value was determined according to the weight ratio of bark to wood. The dry weights of the subsamples were used to calculate the biomass components (Makineci et al. 2011).

Data analysis

Biomass models were developed to estimate the biomass components (foliage, branch, bark, stem, and total aboveground) of individual trees for each specific oak species and without species separation. A power equation was used in the biomass models. Two main equations that are commonly used in biomass studies were fitted: (1) the relationship between the biomass components and DBH alone, and (2) the relationship between the DBH plus the total height (H) of the tree and the biomass components.

$$W_{\text{BC}} = a{\text{DBH}}^{b} + \varepsilon$$
(1)
$$W_{\text{BC}} = a({\text{DBH}}^{2} H)^{b} + \varepsilon$$
(2)

where WBC is the biomass component of the tree (kg), DBH is the diameter at breast height (cm), H is the total height (m), a and b are the regression parameters of the model, and \(\varepsilon\) is the error term. Forty biomass models were developed, with five components (foliage, branch, bark, stem, and aboveground biomass), three oak species and one model without species separation, and two main equations (5 × 4 × 2 = 40).

A nonlinear least-squares method was used to estimate the regression parameters (a and b) without any data transformation or bias correction factor (Smith 1993). All the biomass models were fitted by using the nls function of the R Software Version 3.3.1 (R Core Team 2016).

The adjusted coefficient of determination (Adj.R2), root-mean-square error (RMSE), mean absolute error percentage (MAE (%)), and Akaike’s information criterion (AIC) were used to model the evaluation as statistical criteria.

$$AIC = n + n*\log \left( {2*\pi } \right) + n*\log \left( {\mathop \sum \nolimits (y_{i} - \hat{y}_{i} )^{2} /n} \right)\; + 2*\left( {p + 1} \right)$$
(3)

where \(y_{i}\) is the observed biomass value, \(\hat{y}_{i}\) is the estimated biomass value, n is the number of sample trees, and p is the number of the model parameter.

Results

Table 3 presents the parameter estimates and goodness-of-fit statistics of the models for the individual tree-biomass components on each oak species and a combination of all oak species. All the models were developed at a significance level of p = 0.05. The amount of biomass components in the total aboveground biomass (AGB) according to the development stages and oak species is provided in Figs. 2 and 3. The estimated biomass of the tree components according to the DBH and d2h variables is shown in Figs. 4 and 5.

Table 3 Coefficients of regression equations and goodness-of-fit statistics for biomass components of oak species
Full size table
Fig. 2
figure 2

Rates of biomass components in AGB based on development stages. (SDF (small-diameter forest): mean DBH of 0–8 cm; MDF (medium-diameter forest): 8–20 cm; and LDF (large-diameter forest): 20–36 cm)

Full size image
Fig. 3
figure 3

Rates of biomass components in AGB based on oak species

Full size image
Fig. 4
figure 4

Estimated biomass of tree components based on DBH

Full size image
Fig. 5
figure 5

Estimated biomass of tree components based on d2h

Full size image

Foliage biomass

According to the models based on the diameter (Eq. 1), the variation rates of the foliage biomass were 68% (QP), 78% (QF), 81% (QC), and 71% for all species. The mean absolute error percentages (MAE (%)) ranged from 30 (QF) to 38 (QP). The variation ratios as described by the models that combined the DBH and tree height (Eq. 2) decreased, while the MAE (%) increased (Table 3).

The biomass contribution rate of foliage biomass in the oak species varied from approximately 9% (QP) to 14% (QC). When all the oak types were evaluated together, the foliage biomass ratio was 11%. According to the development stages, the foliage-biomass ratios ranged from approximately 4–21%, decreasing as the diameter increased (Figs. 2 and 3).

Branch biomass

The Adj.R2 value of the diameter-only biomass model for estimating the branch biomass was 0.8 for all oak species and 0.75 (QP), 0.82 (QC), and 0.88 (QF) for individual species. The MAE (%) values were 31% (QF), 32% (QC), 37% (QP), and 34% (general branch-biomass equation). Similar to those of the foliage biomass, the R2 values decreased in the models that combined the DBH and tree height (Eq. 2), while the MAE (%) increased. However, the goodness-of-fit statistics of the models for the QF species were very similar (Table 3).

The total biomass contribution rate of branch biomass was approximately 18% in all oak species and barely changed. According to the development stages, the branch-biomass ratio was the lowest for MDFs (16%) and approximately 20% for SDFs and LDFs (Figs. 2 and 3).

Bark biomass

When estimating the bark biomass from the biomass models according to the diameter only, the QC model had the highest Adj.R2 value (0.94), followed by QF and QP with a value of 0.88 Adj.R2. Meanwhile, the models had MAE (%) values of 18% (QC), 22% (QP), and 23% (QF) (Table 3).

In contrast to the foliage- and branch-biomass models, the Adj.R2 values of the models that combined the DBH and tree height were higher for all oak species than the Adj.R2 values for the models that used only the DBH, while the error percentages were lower. After adding the tree height to the models, the Adj.R2 increased by approximately 4% and the MAE (%) values decreased by approximately 4%. (Table 3).

The total biomass contribution from bark biomass varied between 14 (QF) and 17% (QC) among the oak species. When all the oak types were evaluated together, the bark-biomass ratio was approximately 15%. According to the development stages, the bark ratios showed changes between 12 and 17%. As the diameter increased, the bark-biomass rates decreased, as with the foliage biomass (Figs. 2 and 3).

Stem biomass

To estimate the stem biomass, the biomass models that were generated by the diameter alone had very close Adj.R2 values (QF and QC = 0.92, QP = 0.87). Similar to the coefficients of determination, the MAE (%) values were similar: 23 (QF), 23 (QC), and 24 (QP) (Table 3).

Among the models for all species, the MAE (%) values of the models from the DBH were approximately 23%, while the MAE (%) values of the models that were generated by combining the DBH and tree height were approximately 12%. When the tree height was entered into the model, the MAE (%) values decreased by around half, while the coefficient of determination increased (Table 3).

The total biomass contribution rate from stem biomass in the oak species varied between 51 (QC) and 58% (QP). When all the oak species were evaluated together, the stem-biomass ratio was approximately 56%. According to the development stages, the stem-biomass ratios varied between 42 and 66%, and the total stem-biomass contribution increased as the diameter increased (Figs. 2 and 3).

Total aboveground biomass

Among the models of the biomass components, the aboveground biomass model was the best model to describe the biomass variation. The estimated aboveground biomass, when sorted according to the variation ratios as explained by the diameter-only models, was QP = 90%, QF = 94%, and QC = 96%. The lowest percentage of errors was 17 for QC, followed by 18 for QF and 20 for QP (Table 3).

For all the oak species, the MAE (%) values of the models from only the DBH were approximately 19%, and the MAE (%) values of the models when combining the DBH and tree height were approximately 12%. The MAE (%) values decreased, whereas the coefficient of determination increased when the tree height was entered into the model (Table 3).

Discussion

The forests in this study’s research area have been managed for years by coppice management. Generally, periodic clear-cuts with 20-year rotations in coppice management lead to divergence from high forest-stand structures. Sprouts that are supported by a mature root system exhibit particularly rapid height and diameter development, especially in the initial years. Depending on the increased number of rotations, however, sprout formation and development can be reduced by an aging root system and the changing conditions of the site.

In this study, as in similar studies, the most commonly used individual tree-biomass models were used. The diameter at breast height was used as the most important independent variable to estimate individual tree-biomass components. On the other hand, the inclusion of tree height in addition to the DBH has been shown to increase biomass estimation accuracy (González-García et al. 2013; Pehlivan 2017; Vargas-Larreta et al. 2017). In contrast, however, the addition of tree height reduces the accuracy for some biomass components. In our study, adding the tree height to the model increased the error percentage for the foliage- and branch-biomass components. Similarly, Zhao et al. (2015), Lambert et al. (2005), and Bi et al. (2004) found that including the tree height in models increased the stem biomass-estimation accuracy but did not increase the accuracy of the crown (needles and branches)-biomass components. Thus, as indicated by Ter-Mikaelian and Parker (2000), the influence of the stand structure and site quality on the tree height particularly affects variations in the bark-, stem-, and total aboveground-biomass components. However, researchers noted that different silvicultural treatments that affect the stand structure and tree architecture also affect biomass variability (Vargas-Larreta et al. 2017). Although we obtained better estimates of bark-, stem-, and total aboveground-biomass components by using the combined d2h variable in biomass equations, Carvalho and Parresol (2003) also obtained the best estimates for stem-, crown-, and total tree-biomass components in Quercus pyrenaica (Wild.) by using d2h.

The estimated foliage-biomass values had an error of 34% for all species in the model that depended only on the DBH. Foliage biomass is known to be affected by many factors. In addition, symptoms of oak decline have been determined for both the research area and worldwide. A typical indicator of oak decline is defoliation. Although the samplings from this study were taken from healthy trees, the rates of defoliation, especially in QP and later in QF, have been high in different regions.

Among oak species, Q. cerris is a healthy and resistant species. Therefore, the biomass equation of QC expresses the highest rate (81%) of variation in foliage biomass. Findings on this subject in the same research area were summarized by Kumbasli et al. (2017). One of the most important limitations of our study was a lack of distinction between dead and live branches.

The stem-biomass contribution rates in this study were approximately 20% lower than those from central European oak (Q. robur and Q. petraea)-biomass studies. The contribution rate from bark biomass was approximately 15% in our study, while the bark biomass rate in a central European oak-biomass study was approximately two times higher (Cienciala et al. 2008). Different biomass values and total biomass-contribution rates of biomass components can be expected for different sites, species, and development stages.

The distribution of standardized errors of the models developed for each oak species is more homogenous and randomly distributed than the general model developed for all oak species. Also quite a large part of the standardized error varies from − 2 to + 2 in models developed for each oak species (Figs. 6, 7, 8, 9). Therefore, it may be more appropriate to use models developed for each oak species in biomass estimates.

Fig. 6
figure 6

Residual plots of models developed for biomass components in Q. frainetto

Full size image
Fig. 7
figure 7

Residual plots of models developed for biomass components in Q. cerris

Full size image
Fig. 8
figure 8

Residual plots of models developed for biomass components in Q. petraea

Full size image
Fig. 9
figure 9

Residual plots of models developed for biomass components in all oak species (General)

Full size image

Oak coppice stands have significant increments through a high amount of foliage biomass during younger stages because of developed root systems (Aylak Özdemir 2013), while the ratio of the foliage biomass to the total biomass was significantly reduced as the diameter of the stand increased. This phenomenon can accelerate oak decline during later stages. Indeed, net assimilation can decrease with decreasing foliage biomass, creating nutritional and photosynthesis problems alongside health problems. This effect can be more severe in oak forests that have been subjected to intensive coppice management over many years.

Conclusions

Different coppice-originated oak species have different biomasses depending on their development stages. Allometric functions for estimating biomass according to the species and development stages could provide more reliable results in biomass studies that involve other, primarily carbon and bioenergy calculations. The DBH is an important independent variable in these models for estimating biomass components and can be easily measured and estimated with high accuracy. Indeed, while the accuracy of the foliage- and branch-biomass estimates decreased with the inclusion of the tree height to the model, the accuracy of the bark-, stem-, and total aboveground-biomass estimates increased and the error percentages decreased. Generally, as the diameter increased, the proportion of foliage and bark biomass in the total aboveground biomass decreased and the stem biomass increased. The study area was intensively managed by coppice management, with the advantage of an advanced root system, so high foliage biomass was reached at a young stage.