Introduction

Secondary forests created following intense logging, are an important component of China’s tropical forests, part of the ecotone from tropical Southeast Asia to East Asia with lowland rainforests, seasonal rainforests, and mountain rainforests. In the decades prior to the 2000s, secondary forests or plantations replaced most of primary rainforests. Intense logging gave rise to secondary forests, currently enclosed in nature reserves. Reduced impact logging has limited or minor effects on an ecosystem’s biophysical or biogeochemical processes (Miller et al. 2011). However, it is not the same for severely logged forests, which is what has occurred in tropical China.

Studies on tropical secondary forests have focused on early stages of secondary growth on cleared land (Brown and Lugo 1990). Little information is available for tropical secondary forests resulting from heavy logging. The heavily logged secondary forest has few large trees but still contains several in a medium growth stage with a nearly closed, though not dense, canopy.

Logging of large trees stimulates the growth of understory trees by enhancing available sunlight. To what extent this stimulation occurs is still unknown. Brown and Lugo (1990) suggested that biomass accumulation is rapid in the first 15 years of secondary growth (up to 100 t ha−1). In an Amazon logging experiment, this was in a range from 2.5 to 4.5 t ha−1 a−1 (Figueira et al. 2008). Moreover, small trees predominantly contribute to the increase in biomass. The logging example in the Amazon is moderate, logging 15% of the trees with a DBH above 35 cm. How biomass increment and mortality losses respond to heavy or intense logging, as is the case in tropical China, needs quantification. In this study we report inventory data from 38 permanent sample plots to show how intense logging affects biomass increment and mortality.

Materials and methods

Diaoluoshan National Natural Reserve

This study was carried out in Diaoluoshan National Nature Reserve (DNNR) located in the southeast of Hainan Island (18°40′–49′N, 109°45′–57′E). Climate seasonality is due to a tropical monsoon climate. In general, late May through October is the rainy season and November to early May the dry season. Mean annual temperature is 24.6 °C, with the lowest in January (15.3 °C) and highest in July (28.4 °C) (Hu 1997). Annual rainfall is relatively high up to 2160 mm.

The DNNR was approved as a national nature reserve in January 2008. Its total area is 183.87 km2, with 78.41 km2 as core, 84.56 km2 as buffer, and 20.90 km2 an experimental zone (http://dlsnr.forestry.gov.cn/). The lowest point is approximately 100 m.a.s.l.), and the highes 1499 m.a.s.l. Approximately 2126 species, belonging to 239 families of vascular plants, have been recorded. The design of the DNNR aimed to: (1) conserve tropical rainforests; (2) protect rare or endangered species and their habitats; (3) maintain forest biodiversity; and, (4) conserve landscape and tourism resources.

Before the 1980s, DNNR was a forestry centre famous for timber production and the main vegetation classified as tropical rainforest. Forests below 700 m) flourish with species of the Dipterocarpaceae family, such as Vatica mangachapoi Blanco (Hu 1997). These ecosystems are tropical lowland rainforests; ecosystems on mountaintops are short montane moss forests. Between them lies tropical mountain rainforests. Montane moss forests are difficult to access and have limited available quality timber, and logging is therefore not severe. Lowland rainforests with their valuable timber, have been heavily logged. An old engineer, who lives and works in the DNNR, recalled that trees with a DBH above 20 cm were cut for timber before the establishment of the Nature Reserve (personal communication, Mr Yiwen Liang, a local people). The tropical mountain forests intensely with a few old trees randomly distributed were logged intensely.

The inventory plots

A total of 38 plots (50 × 50 m) were established in 2010 (Fig. 1). A re-census was carried out in plots #1–#15 in 2015. The elevation of these plots ranges from 245 to 1255 m.a.s.l. (Table 1). The forest type was determined as: (1) 17 plots with trees belonging to the Dipterocarp family was tropical lowland forest (TLF); (2) one plot without Dipterocarps but with Dacrydium pierrei Hickel was montane moss forest (MMF); and, (3) 20 plots were tropical mountain forest (TMF). Soil of the TLF and TMF plots is latosol; the MMF plot has mountainous yellow soil. Trees H > 2 cm DBH labeled and identified to species. A total of 26,725 trees were recorded. The DBH was measured with a measuring tape and tree height was also recorded.

Fig. 1
figure 1

Distribution of plots in the Diaoluoshan National Nature Reserve

Full size image
Table 1 General information about the plots
Full size table

Biomass allometric equations

The biomass allometry principle indicates that there is a relationship between biomass and tree size (Niklas 1994; Gower et al. 1999). The most common expression for this relationship is the power function:

$$ y = a(x)^{b} $$
(1)

where y is biomass, x tree size, a and b fitted parameters and may be species specific. It is impractical to collect allometric data for all species, which might be over 200–300 per plot in tropical rain forests. Development of mixed-species allometric equations is necessary for this application. Kira et al. (1967) is the pioneer study to use mixed allometric equations for tropical forests. Some mixed-species equations are also available for our study region. The typical one used is that of Li (1993), which was developed based on samplings of 102 trees and more than 70 species, and diameter sizes from 5 cm to 80 cm, heights from less than 8 to 32 m:

$$ W_{AG} = 0.040213((DBH)^{2} H)^{0.97268} $$
(2)

where WAG is aboveground biomass (kg), DBH diameter at breast height (cm) and H height (m). The parameters are similar to those used by Brown et al. (1989) from worldwide data for moist tropical forests:

$$ W_{AG} = 0.044418((DBH)^{2} H)^{0.9719} $$
(3)

We thus adopted the Li (1993) equation for aboveground biomass estimation.

Belowground allometric equations, including fine roots, however, are rare compared to aboveground ones. We collected four allometric equations for comparison purposes in this study:

  1. (1)

    Belowground biomass allometry derived in the same region (Li 1993):

    $$ W_{\text{BG}} = 0.003612((DBH)^{2} H)^{1.11527} $$
    (4)

    where WBG is belowground biomass (kg).

  2. (2)

    Allometry collected in Xishuangbanan prefecture of China for TMF (Zheng et al. 2006):

    $$ W_{\text{BG}} = 0.003612((DBH)^{2} H)^{1.11527} $$
    (5)
  3. (3)

    Allometry collected in Xishuangbanna of China for a tropical seasonal rainforest (Feng et al. 1998):

    $$ \left\{ \begin{aligned} W_{\text{BG}} = 0.0112((DBH)^{2} H)^{0.9045} , \, DBH \le 20\;{\text{cm}} \hfill \\ W_{\text{BG}} = 0.0069((DBH)^{2} H)^{0.9781} , \, DBH > 20\;{\text{cm}} \hfill \\ \end{aligned} \right. $$
    (6)
  4. (4)

    Allometry collected in Thailand for a tropical seasonal rainforest (Kira et al. 1967):

    $$ W_{\text{BG}} = 0.0264((DBH)^{2} H)^{0.775} $$
    (7)

Biomass increment, recruitment, mortality, and total accumulation

Relative growth rate (RGR) is calculated as:

$$ RGR = \frac{{2(DBH_{{{\text{i}} + 1}} - DBH_{\text{i}} )}}{{DBH_{{{\text{i}} + 1}} + DBH_{\text{i}} }} \times 100 $$
(8)

where DBHi and DBHi+1 are diameter at breast height in the ith and i + 1th censuses.Mortality rate (MR) is calculated as:

$$ MR = \frac{{N_{\text{M}} }}{{N_{\text{T}} }} $$
(9)

where NM is the number of tree mortalities in a specific DBH class and NT is total trees in that class.Biomass increment (ΔB) is calculated as:

$$ \Delta B = W_{\text{i}} - W_{{{\text{i}} + 1}} $$
(10)

where Wi and Wi+1 are total biomass of live trees in ith and i + 1th census We tracked individual trees to calculate ΔB (Clark et al. 2001); thus, only live trees in both the ith and i + 1th censuses were used for calculations. Recruitment (R) and mortality (M) biomass were calculated as biomass of dead trees and recruitment of trees during the census time. Total biomass accumulation (Cb) is calculated as:

$$ C_{\text{B}} = \Delta B + R - M $$
(11)

Results

Stand biomass

The heavily logged forests had occasional large and medium- sized trees. Only 22 trees had a DBH above 70 cm from 13896 individuals for plots #1–#15 (< 0.2%) (Fig. 2). The number of medium size trees (10–35 cm DBH was 132 (< 1%).

Fig. 2
figure 2

Frequency distribution in different diameter classes based on the plots #1–#15

Full size image

Mean biomass for all 38 plots was 218.6 ± 80.8 t ha−1 (Table 1). Mean biomass was higher in TMF (228.6 ± 63.9 t ha−1) than in TLF (212.2 ± 97.1 t ha−1). A t test between TLF and TMF was not significant (p = 0.542). The single MMF plot showed a biomass of 118.2 t ha−1. There was an altitude trend in biomass distribution (Fig. 3). Biomass peaked in middle altitudes, with lower and higher altitudes with low biomass. The ratio of belowground and aboveground biomass is 0.32 on average, and it has a relation to elevation (Fig. 4).

Fig. 3
figure 3

Estimated stand biomass in relation to elevation

Full size image
Fig. 4
figure 4

Ratio of belowground and aboveground biomass in relation to elevation

Full size image

Belowground biomass, estimated with different allometric equations, varied strongly (Fig. 5). The allometry derived in the same region (Li 1993) gave the highest estimated biomass (54.1 t ha−1). Allometric equations from the Xishuangbanna prefecture of China gave medium levels of estimated biomass of 28.5 t ha−1 (Feng et al. 1998), and 27.6 t ha−1 (Zheng et al. 2006)). Biomass estimated with allometry equations calculated for forests further from the site (Kira et al. 1967) was lowest (18.6 t ha−1).

Fig. 5
figure 5

Comparison of belowground biomass estimated with different allometric equations

Full size image

Tree growth and biomass increment

Relative growth rates (RGR) tend to decrease with DBH (Fig. 6). Large trees (> 35 cm) showed low RGR, usually less than 5.0% (Mean RGR was 3.4%), while RGR of smaller trees (< 10 cm) varied strongly with a mean value of 7.3%. The mean RGR of medium-sized trees was 6.1%.

Fig. 6
figure 6

Relative growth rate compared to diameter

Full size image

The distribution of biomass increment (ΔB) under different DBH classes is shown in Fig. 7. Despite accounting for nearly half of the total number of trees, individuals with a DBH < 5 cm contributed slightly to the ΔB of both lowland and mountain forests. The mean contribution to total ΔB was highest in DBH classes 12–18 cm for TLF, whereas the highest contribution for the TMF were 24–35 cm DBHs. The pie chart of the contributions of small ([2,10]), medium-sized ([10,35]) and large trees ([35, ∞]) to total ΔB is shown in Fig. 8. Medium-sized trees contribute most to total ΔB, and this is especially obvious in TMF. Large trees, with only a small portion of the number of trees, contribute comparable to that of small trees.

Fig. 7
figure 7

Biomass increment in different DBH classes

Full size image
Fig. 8
figure 8

Pie chart depicting the contribution of different size classes to stand biomass increment Small: DBH < 10 cm; Medium: 10–35 cm; Large : DBH > 35 cm

Full size image

Mortality rate

Therate of mortality was highly dependent on tree size (Fig. 9). No tree with a DBH > 50 cm died during the investigation period. More than 1100 trees, including five with DBH > 35 cm, died during the study period. The overall mortality rate was 0.015 per year.

Fig. 9
figure 9

Tree morality rate plotted against mean size (Data of horizontal axis based on mean DBH of 10 DBH classes)

Full size image

Total biomass accumulation

There were no trends found in biomass increment, recruitment, mortality losses, or total biomass accumulation due to altitude (Fig. 10). The mean biomass increment was 5.2 ± 0.8 and 5.7 ± 1.5 t ha−1 a−1 for TLF and TMF, respectively. Mean recruitment had little effect on total biomass accumulation with mean values of 0.3 for TLF and for TMF. Mortality loss was similar in both with a value of 1.6 t ha−1 a−1(Fig. 11). Overall, the net biomass accumulation was slightly higher in TMF (4.3 ± 1.5 t ha−1 a−1) than TLF (3.9 ± 1.4 t ha−1 a−1). A t-test of Cb between TLF and TMF was not significant (p = 0.612).

Fig. 10
figure 10

Altitude effects on biomass increment, recruitment, mortality losses, and total biomass accumulation (TLF, open symbols; TMF, closed symbols)

Full size image
Fig. 11
figure 11

Net biomass accumulation rate and its components

Full size image

Discussion and conclusions

Stand vegetation accumulated carbon at moderate rates in the tropical secondary forests under study. If the biomass is converted into carbon with a conversion factor of 0.5, the mean vegetation carbon sink was 2.17 and 1.96 t C ha−1 a−1 for mountain and lowland forests, respectively. This estimation is similar to results from a logged forest in the Philippines (1.4 t C ha−1 a−1; Lasco et al. 2006), and in Borneo (1.2 t C ha−1 a−1; Berry et al. 2010). This is higher than that of primary forests in the same region (0.56–0.62 t C ha−1 a−1; Chen et al.2010), in the Amazon (0.98 t C ha−1 a−1; Baker et al.2004), and in Africa (0.63 t C ha−1 a−1; Lewis et al. (2009), but not as large as secondary forests on cleared land (Brown and Lugo 1990).

Medium-sized trees were major contributors to biomass increment in the forests in this study (Fig. 8). This is different from a logging experiment in the Amazon which suggested small trees play a leading role in biomass increment (Figueira et al. 2008). This could be explained by the history of our Nature Reserve. The last logging operation took place in the DNNR in 2001 (personal communication, Mr Yiwen Liang, a local people). Forests were strictly protected after 2001 which means there was already at least 10 years of secondary growth before we established the plots. Selective logging of large trees could rapidly increase the level of sunlight to understory trees. After more than 10 years of change, most of these small- size trees have grown into medium-sized ones. This was the major reason for the differences between our forests and the Amazon logging experiment.

The degree of logging intensity plays an important role in both biomass amount and its rate of accumulation in secondary forests. Similar dipterocarp forests in Xishuangbanna, Thailand, and Malaysia showed high biomass (Ogawa et al. 1965; Terakunpisut et al. 2007; Tan et al. 2015). In general, the low-elevation site usually has higher temperatures than higher altitude sites. If water is not a limiting factor, warmer sites have higher biomass. In our study, biomass was highest in the middle elevation plot of the mountain forest (Fig. 3). The low elevation lowland forests did not have a high biomass, and possibly experienced more disturbance by logging. The biomass increment across DBH classes was low. The highest DBH class contributing to biomass was smaller in TLF (12–18 cm) than in mountain forest (24–35 cm) (Fig. 7).

Mortality, accounting for approximately 30% of biomass increase, should be in biomass carbon accumulation estimates. The overall mortality was 0.015 trees per year. This is comparable to that of primary forests in Pasoh, Malaysia (King et al. 2006). Mortality is strongly dependent on tree size (Fig. 9). This is consistent with previous inventories (Lieberman et al. 1985) and models (Pinard and Cropper 2000). Most importantly, no tree died with a DBH > 50 cm over the 5-year investigation; however, mortality of seedlings and small trees was relatively high. Logging of large trees may increase the level of understory radiance to seedlings and small trees. This could also result in strong competition and subsequent high mortality.

The ratio of belowground and aboveground biomass is important to estimate total forest biomass. The mean ratio was 0.319, almost twice as high than a similar logged forest in SW China (0.165; Tang et al. 1998) and Malaysia (0.172; Pinard and Putz 1996). This ratio is similar to results from primary tropical forests in Brazil (0.252; Russell 1983; 0.255; Salomão et al. 1996, and Columbia (0.263; Saldarriaga et al. 1988). Methodology is a critical element in estimating belowground biomass, and varied strongly with different allometric equations (Fig. 5). On the one hand, there are less allometric equations available for belowground biomass compared to that for aboveground. On the other, most of the allometric equations are derived for primary forests. The allometric equation based on the same region gave the highest belowground estimate, followed by equations from Xishuangbanna, China, and the lowest with equations from Thailand.

Overall, the logged secondary forest showed some differences in biomass and its rate of accumulation compared to primary forests in the tropics. Biomass recovery is not as rapid as in secondary forests established on cleared land. Given this, mean biomass accumulation rate could be maintained in subsequent years; recovery of 100 t ha−1 needs at least 50 years. In reality, this accumulation rate could hardly be maintained due to disturbance and increasing canopy cover. Therefore, tropical forests are sensitive to logging operations and need careful forest management.