Introduction

Tropical forests have been exposed to drastic land-use changes in the past 50 years (Don et al. 2011; Gibson et al. 2011). The change from traditional shifting cultivation to continuous cropping or oil palm plantations leads to a loss of soil organic matter (SOM) and a rapid decline in plant productivity in some tropical regions (Kimetu et al. 2008). The limited availability of fertilizers and organic resources (e.g., manure) induces land abandonment and further deforestation for small-holder farmers in low-input agriculture settings (Lal 2004, 2006; Smith 2008). Land-use strategies, including fallow systems, need to be optimized to maintain levels of SOM in tropical agroecosystems, considering the availability of resources for farmers.

Carbon (C) cycle in tropical forest soils are characterized by large C fluxes via litterfall and microbial decomposition (Vitousek and Sanford Jr 1986; Fujii et al. 2018). Continuous cropping usually results in a rapid loss of soil organic carbon (SOC) stocks (Fujii et al. 2009a), whereas both gains and losses in SOC stocks under natural grassland and secondary forests or human-induced pastures and plantations have been reported (Veldkamp 1994; Yonekura et al. 2012; Sang et al. 2013). The effects of land-use changes on SOC stocks are inconsistent and variable, depending on the stage of cropping or fallow system and the extent of disturbance or management practice (e.g., fires and tillage) (Don et al. 2011), but both natural and disturbed ecosystems share similarity that SOC dynamics are driven by plant and soil microbial processes (Veldkamp 1994; Hooper and Vitousek 1998). By extracting the dominant plant traits and soil properties regulating turnover and storage of organic matter in these ecosystems, the effects of fallow grassland or forest on SOC stocks can be evaluated.

Vegetation changes could influence SOC stocks directly through litter quantity and quality and indirectly through effects on soil chemical and biological properties (Fujii et al. 2018). Field incubation of standard substrates (litter bag or cellulose test) have shown that litters with high lignin to nitrogen (N) ratios decompose relatively slowly (Berg and MacClaugherty 2003) and that SOM of C4 plant (e.g., Imperata grass) origin can decompose faster than SOM of C3 plant (e.g., trees) origin (Wynn and Bird 2007). Organic matter turnover could also be affected by soil properties that are changeable under different vegetation covers (Yamashita et al. 2008). For example, soil acidification under tropical forests (e.g., Acacia plantation) can reduce microbial activity and retard organic matter decomposition (Hayakawa et al. 2014). The stability of organic matter derived from forest and grassland can be traced using the difference in their litter 13C isotopic signature and the potential importance of grassland in SOM storage has been reported by several studies (Yonekura et al. 2012). Combination of litter bag tests of standard substrates and 13C isotopic signature analysis allow us to extract the effects of plant traits and soil properties on SOC dynamics by tracing the SOC stocks of individual plant origins.

Fires that occurred in 1982–83 and 1997–1998 in East Kalimantan, Indonesia, resulted in land-use changes from primary dipterocarp forest to cropland, Imperata grassland, oil palm or leguminous tree (Acacia mangium) plantation on Imperata grassland, and natural secondary forests regenerated by pioneer species Macaranga spp. (Ohta et al. 2000). We monitored 30-year changes in SOC stocks and analyzed the factors regulating organic matter turnover and storage using litter bag tests and 13C natural abundances. Based on preceding studies (Fujii et al. 2011, Yonekura et al. 2013), we hypothesized that (1) grassland can contribute to initial increase in SOC stocks, but secondary forest or Acacia plantation maximize the long-term gain in SOC stocks due to inputs of lignin-rich litter; and (2) soil acidification in Acacia plantation on Imperata grassland would reduce microbial decomposition of grassland-derived SOC and increase total SOC stocks, compared with continuous Imperata grassland or oil palm plantation. Based on SOC stocks and nutrient availability, we also attempt to propose an optimal fallow system for agroecosystems transformed from dipterocarp forests.

Materials and methods

Site description and sampling design

Experiments were carried out in tropical forests and agroecosystems in Bukit Soeharto (S0°51′, E117°06′; 99 m a. s. l., average inclination 15°), East Kalimantan Province, Indonesia (Fig. 1). The mean annual air temperature was 26.8 °C, and the mean annual precipitation was recorded as 2187 mm yr−1. Soils were derived from sedimentary rocks and classified as Typic Paleudults (Soil Survey Staff 2014). Soil pH is low throughout the profile (3.8–4.3) and clay contents increase with depth (23–31%). The detailed information of soil physicochemical properties was given in Fujii et al. (Fujii et al. 2009a, b). After the fires in 1982–1983, we establish plots of vegetation cover change sequences: continuous primary dipterocarp forest (PF plots), the vegetation cover changed from primary dipterocarp forest to natural secondary forests of the pioneer species Macaranga gigantea regeneration (Macaranga forest; MF plots), and Imperata cylindrical grassland (IG plots). The adjacent Imperata cylindrica grassland of IG plots were further changed to oil palm (Elaeis guineensis L.) plantation (OP plots) during the period 2004–2015 and Acacia mangium plantation (Acacia forest; AF plots) during the period 1992–2015. Each land-use sequence composed of three plots (20 m × 20 m), composite soil samples were collected from three pits at each plot. The distance between each pit was 10 m. Our study is based on two assumptions that (1) initial soil properties of five sites are similar to those obtained from the pristine dipterocarp forest site in 1983 and (2) the soil properties of AF and OP plots before land-use change was same with the those in grassland plots at the time of conversion (IG plots). We monitored changes in SOC stocks from 1986 to 2015 and compared litter decomposition rates using litter bag tests for five sites under vegetation change sequences in 2015.

Fig. 1
figure 1

Changes in soil pH (0–5 cm) under different land-uses. Bars represent standard errors (N = 3)

Full size image

Physicochemical and microbiological properties of soils

All the soil samples were collected, kept in plastic bags at 4 °C prior to analysis, and sieved (<2 mm) to eliminate litter, roots, and stones. For the soil samples collected from the PF, MF, AF, IG, OP sites in 2015, a subsample of field-moist soil was used for measurements of microbiological properties, and another subsample was air dried and used for measurements of physicochemical properties.

The concentrations of total C and N in soils were measured using a CN analyzer (Vario Max CN; Elementar Analysensystem GmbH). Soil pH was measured using a soil to solution (water) ratio of 1:5 (w/v) after shaking for 1 h. The particle size distribution was determined by the pipette method (Gee and Bouder 1986). The concentrations of exchangeable potassium (K+) were measured using batch extraction with ammonium acetate (1 M, pH 7.0) and flame photometry. The concentrations of inorganic N (NH4+ and NO3−) in the field-moist soils were measured by extraction with 2 M KCl for 30 min using a soil to solution (H2O) ratio of 1:5 (w/v) (Mulvaney 1996; Rhine et al. 1998). Available phosphorus (P) concentrations were estimated using the Bray 2 extraction method (Blakemore et al. 1987). The microbial biomasses C (MBC) at the start of litter bag field incubation was determined using the chloroform fumigation-extraction method (Vance et al., 1987) with conversion factor of 0.45 (Wu et al. 1990). The soluble C in the fumigated and non-fumigated soil samples were extracted with 0.5 M K2SO4 (soil to solution ratio of 1:5) and measured using a total organic C analyzer (TOC-V CSH; Shimadzu, Japan).

Soil carbon stocks and calculation of grass and forest origins using 13C natural abundance

The organic horizons were sampled in 30 × 30-cm quadrates in three replicates per plot, oven dried (48 h, 70 °C), and individually weighed. For the mineral soil horizons (<200 g C kg−1), the bulk density (g cm−3) was measured using a 0.1-L core in three replicates per plot. The SOC stocks in each soil horizon were calculated by multiplying the soil C concentrations, the bulk density, and individual depths as follows:

$$ Soil\ C\ stock\ \left(\mathrm{kg}\ C\ {m}^{-2}\right)=\mathrm{Soil}\ C\ \mathrm{concentration}\ \kern0em \ \left(g\ C\ k{g}^{-1}\right)\times \mathrm{Soil}\ \mathrm{horizon}\ \mathrm{depth}\ (m)\times \mathrm{Bulk}\ \mathrm{density}\ \left(\mathrm{Mg}\ {m}^{-3}\right) $$
(1)

The SOC stocks at each site were calculated by summing up to the mineral soil depth of 40 cm. A sample (1 mg litter and 10 mg soil) was weighed into a tin capsule, and δ13C was measured with an on-line C analyzer (NC 2500; Thermo Fishe Scientific) coupled with an isotope ratio mass spectrometer (MAT252; Thermo Electron, Bremen, Germany). All δ13C values were expressed relative to the Vienna Pee Dee Belemnite (VPDB) international standard:

$$ {\updelta}^{13}\mathrm{C}=\frac{\frac{{}^{13}\mathrm{C}}{{}^{12}\mathrm{C}}\left(\mathrm{sample}\right)}{\frac{{}^{13}\mathrm{C}}{{}^{12}\mathrm{C}}\left(\mathrm{VPDB}\right)}-1 $$
(2)

The standard deviation for four replicate combustions of the same standard within a sequence was 0.02%.

Assuming that the soil δ13C profiles in primary dipterocarp forests at the sampling periods are identical to those in 1983 and those unaffected by Imperata grass before and after the fires, grassland-derived and forest-derived C concentrations (mg C kg−1 soil) were estimated using the following equations, respectively (Veldkamp 1994):

$$ {\displaystyle \begin{array}{l}\mathrm{Grassland}\kern0.5em \mathrm{derived}\kern0.5em \mathrm{C}\left(\mathrm{mg}\ {\mathrm{kg}}^{-1}\right)\\ {}\kern3.839998em =\frac{\delta^{13}\mathrm{C}\left(\mathrm{grass}\mathrm{land}\ \mathrm{soil}\kern0.17em \mathrm{sample}\right)-{\delta}^{13}\mathrm{C}\left(\mathrm{forest}\ \mathrm{soil}\right)}{\delta^{13}\mathrm{C}\left(\mathrm{grass}\ \mathrm{litter}\right)-{\delta}^{13}\mathrm{C}\left(\mathrm{forest}\ \mathrm{soil}\right)}\times \mathrm{Soil}\kern0.5em \mathrm{C}\ \left(\mathrm{mg}\ {\mathrm{kg}}^{-1}\right)\end{array}} $$
(3)
$$ \mathrm{Forest}\ \mathrm{derived}\ \mathrm{C}\ \left(\mathrm{mg}\ \mathrm{k}{\mathrm{g}}^{-1}\right)=\mathrm{Soil}\ \mathrm{C}\ \left(\mathrm{mg}\ \mathrm{k}{\mathrm{g}}^{-1}\right)-\mathrm{Grassland}\ \mathrm{derived}\ \mathrm{C}\ \left(\mathrm{mg}\ \mathrm{k}{\mathrm{g}}^{-1}\right) $$
(4)

where Imperata leaf biomass δ13C (−12.3‰) was used for δ13C (grass litter). The calculation was made at each corresponding depth and scaled up to the soil profiles.

Similarly, to examine the effects of Imperata grassland conversion to Acacia or oil palm plantation on SOC storage, Acacia-derived and grassland and primary forest–derived C concentrations (mg C kg−1 soil) were estimated for the Acacia plantation (AF site) and oil palm plantation (OP site) using the following equations, respectively:

$$ {\displaystyle \begin{array}{l}\mathrm{Acacia}\ \mathrm{derived}\ \mathrm{C}\ \left(\mathrm{mg}\ {\mathrm{kg}}^{-1}\right)\\ {}\kern3.839998em =\frac{\delta^{13}\mathrm{C}\left(\mathrm{Acacia}\ \mathrm{soil}\ \mathrm{sample}\right)-{\delta}^{13}\mathrm{C}\left(\mathrm{grassland}\ \mathrm{soil}\right)}{\delta^{13}\mathrm{C}\left(\mathrm{Acacia}\ \mathrm{litter}\right)-{\delta}^{13}\mathrm{C}\left(\mathrm{grassland}\ \mathrm{soil}\right)}\\ {}\kern3.839998em \times \left\{\mathrm{Soil}\ \mathrm{C}-\mathrm{Forest}\ \mathrm{derived}\kern0.37em \mathrm{C}\right\}\kern0.6em \left(\mathrm{mg}\ {\mathrm{kg}}^{-1}\right)\end{array}} $$
(5)
$$ {\displaystyle \begin{array}{l}\mathrm{Grassland}\ \mathrm{derived}\ \mathrm{C}\left(\mathrm{mg}\kern0.24em {\mathrm{kg}}^{-1}\right)\\ {}=\mathrm{Soil}\ \mathrm{C}\ \left(\mathrm{mg}\ {\mathrm{kg}}^{-1}\right)-\mathrm{Forest}\ \mathrm{derived}\ \mathrm{C}\ \left(\mathrm{mg}\ {\mathrm{kg}}^{-1}\right)-\mathrm{Acacia}\ \mathrm{derived}\ \mathrm{C}\ \left(\mathrm{mg}\ {\mathrm{kg}}^{-1}\right)\end{array}} $$
(6)

where forest-derived C concentration in AF site was assumed to be equal to the value obtained in the continuous Imperata grassland (IG site) for each sampling year. Acacia leaf litter δ13C (−32.7%) was used for δ13C (Acacia litter). The same calculation was conducted for the oil palm plantation (OP site) using the oil palm leaf litter δ13C (−30.9%). The changes in forest-derived and grassland-derived SOC stocks were plotted respectively against time after vegetation change (yr). The data were fitted to a single exponential decay function using the least-squares technique in SigmaPlot 11.0 (SYSTAT Software Inc., Point Richmond, CA, USA):

$$ R{}_r={R}_i{e}^{- kt} $$
(7)

where Rr is the remaining proportion of the substrate (%), Ri is the initial proportion of the substrate (i.e., 100%), k is the decomposition rate constant (yr−1), and t is time (yr) since conversion. The mean residence times were estimated from 1/ k, assuming a steady state.

Litterfall carbon input, organic layer carbon stock, and litter decomposition rate factor

Litterfall was collected using circular litter traps (60 cm diameter) between June 2015 and June 2016. The organic layers and fine root (diameter < 2 mm) biomass were collected in the 30 cm × 30 cm quadrats. Fine root biomass in the mineral soil (0–40 cm) was estimated by collecting the roots in 5 cm depth intervals in cores of 0.1 L volume. Roots were rinsed in distilled water to remove soil materials. Five replicates were used for these measurements.

The litterfall, organic layer, and fine root samples were oven dried at 70 °C for 48 h, weighed, and milled. The Klason lignin concentrations in the leaf and root litter samples were determined by digestion with sulfuric acid (Allen et al. 1974). The P and K concentrations in the fresh litter samples were determined using inductively coupled plasma atomic emission spectrometry (ICP-AES; SPS1500; Seiko Instruments Inc.) after nitric-sulfuric acid wet digestion.

Assuming that the organic layer C stock reaches a steady state, the organic C turnover rate (yr−1) in the organic layer was calculated by dividing litterfall C input (Mg C ha−1 yr−1) by the organic layer C stock (Mg C ha−1) (Olson 1963). The lignin (%) to N (%) ratio was calculated and used as an indicator of litter recalcitrance (Aerts 1997).

Measurement of litter and cellulose decomposition rates under field conditions

To examine whether litter decomposition rates depend simply on litter quality, or whether it is also affected by the soil environment, decomposition rates of the standard substrate (cellulose filter paper and leaf and root litters) were also compared. Cellulose is a major constituent of plant materials [10–87% from Berg and McClaugherty 2003] and decomposition rates of cellulose can be a rough indicator of soil microbial activities involved in litter decomposition (Hayakawa et al. 2014). As a standard substrate, the decomposability of Macaranga root litter was compared between PF and MF soils, while the decomposability of Imperata grass root litter was compared between IG and AF soils.

The leaf litter bags were buried at the boundary between the organic horizon and the mineral soil, while root litter bags were buried into surface mineral soil (A horizon, 5-cm depth). Cellulose filter paper (Advantec no. 6, 55-mm diameter) was buried into surface mineral soil (5-cm depth) (Hayakawa et al. 2014). All substrates were packed in the nylon mesh bags (65 × 65 mm, 100-μm mesh pore size) to keep out insects and worms. The fallen leaves collected by litter traps (10 × 10 mm) and fine roots (diameter < 2 mm; length, 10 mm) collected from the surface mineral soil were used for litter bag tests after oven drying at 70 °C for 24 h. In PF plots, leaf litters were composed of Shorea laevis litters (60%) and Dipterocarpus cornutus litters (30%), and others (10%). We used Shorea laevis litters as a representative litters of the PF plots for litter bag test. At each site, five mesh bags of litter or cellulose filter paper were collected at each sampling interval (3 months and 1 month, respectively). The substrate remaining in the mesh bag was dried (70 °C, 24 h) and weighed after soil particles had been carefully removed. The remaining weight of substrates was calculated on an ash-free basis by subtracting the weight of the soil adhering to the substrates, which was estimated by dry combustion (600 °C, 4 h).

To obtain the decomposition rate constant k (yr−1) for litter and cellulose decomposition using Eq. 7, the remaining proportion of substrate (leaf litter, root litter, or cellulose filter paper) (%) relative to the initial weight of the substrates (i.e., 100%) was plotted against time (yr) and fitted to a single exponential decay function (Sparrow et al. 1992).

Measurement of fine root production rates

To estimate root litter input, annual production of fine root (diameter < 2 mm) was measured using the root mesh method (Hirano et al. 2009) at forest sites (PF, MF, AF) in five replicates. A net sheet (width 20 cm × depth 20 cm) with 2-mm openings was inserted vertically into the mineral soil to a depth of 20 cm. After 1 year of incubation, a soil block (width 20 cm × depth 20 cm × thickness 2 cm) containing the net sheet was collected to measure the biomass of fine roots that had grown through the net sheet. Due to the occurrence of fires at the IG site (October 2015), annual fine root production at this site was measured based on the net increase in fine root biomass between October 2015 and October 2016. It should be noted that both methods risk underestimating fine root production due to decomposition of dead fine roots between sampling intervals.

Monitoring of soil temperature and volumetric water content

To determine the effects of soil temperature and moisture on litter and cellulose decomposition rates, we monitored soil hydrothermal conditions at each site. The volumetric water contents of the soils (5-cm depth) were measured in three replicates with amplitude-domain reflectometry probes (Theta probe, ML2x; Delta T Devices). The air and soil temperatures at each depth were measured in three replicates using temperature loggers (Thermochron, SL type). Seasonal fluctuations in soil temperature (5-cm depth) and soil moisture were monitored, with recording at 30-min intervals.

Statistical analyses

All data were expressed as means ± standard errors (SEs), with combined SEs from three to six replicates (Taylor 1997; Zar 1999). The significance of differences in rate constants, k, for litter and cellulose decomposition between sites and substrate types was tested using the F-test and the Tukey method modified for comparison of regression slopes (Zar 1999). The significance of differences in SOC stocks between sites and sampling periods was analyzed using one-way ANOVA and the Tukey method for multiple comparisons. Pearson’s correlation coefficients were calculated to examine relationships between the rate constants and soil properties. All statistics were performed using Sigmaplot 11.0 (SYSTAT Software Inc., CA, USA) and tested at significance level of 0.05, unless otherwise stated.

Results

Chemical and biological properties of surface soils

Soil pH values were significantly lower at the three forest sites (PF, MF, AF) than at the IG and OP sites (Table 1). Thirty-year soil monitoring showed that pH increased after the change from primary dipterocarp forest to Imperata grassland and oil palm plantation due to ash or lime inputs, and decreased by Acacia plantation on Imperata grassland (Fig. 1). The soil C and N concentrations (0–5-cm depth) were significantly lower at the OP site than at the other four sites (Table 1). The microbial biomass C in the OP and IG soils were significantly (P < 0.05) lower than the forest soils (Table 1). Soil nutrient (N, P, K) availability differed among potential fallow vegetation sites (MF, AF, and IG; Table 1). Soil exchangeable K concentrations were significantly lower in MF than in AF and IG. Soil available P concentrations were significantly lower in AF than in MF and IG (Table 1). Soil inorganic N (mainly NH4+) concentrations were significantly lower in IG than in AF and MF (Table 1).

Table 1 Mean (N = 3) chemical and biological properties of soil samples (0–5 cm)
Full size table

Changes in soil organic carbon stocks under land use changes

Thirty-year soil monitoring revealed a net increase in SOC stocks within the initial 10 years following the fire-induced change from primary dipterocarp forest to Imperata grassland (Fig. 2). No increase in SOC stocks was found at the continuous Imperata grassland (IG site) between 1992 and 2015, whereas conversion of Imperata grassland to Acacia plantation (AF site) and Macaranga forest (MF site) resulted in continuous increases in SOC stocks until 2010 or 2015 (Fig. 2). A net decrease in SOC stocks was observed within the initial 10 years after conversion of Imperata grassland to oil palm plantation (2005–2015; Fig. 2).

Fig. 2
figure 2

Changes in soil carbon stocks (0–40 cm) under different land-uses. The C stock in the organic horizon was not counted. Bars represent standard errors (N = 3)

Full size image

The δ13C values of Imperata grass litter differed significantly from those of forest and oil palm litters (Table 2). The change of primary dipterocarp forest to continuous Imperata grassland consistently resulted in the significantly higher soil δ13C values between 0 and 40 cm depth, with higher δ13C values at shallower depths of grassland soil samples (Fig. 3a). In the continuous Imperata grassland (IG site), grassland-derived C rose rapidly to 12–48% between 1983 and 1992, and reached 59% in 2015 (Fig. 4a). Conversion of Imperata grassland to Acacia or oil palm plantation resulted in a decrease in soil δ13C (Fig. 3b) due to the input of litter with higher δ13C values (Table 2). The forest-derived soil C stocks in the continuous Imperata grassland decreased by 9.8 Mg C ha−1 over 32 years (1983–2015; Fig. 4a). The grassland-derived soil C stocks increased by 16.2 Mg C ha−1 over 32 years with SOC accumulation rate of 0.5 Mg C ha−1 yr−1 (Fig. 4a). The conversion of Imperata grassland to Acacia plantation increased SOC stocks by 28.5 Mg C ha−1 over 23 years at the expense of 2.0 Mg C ha−1 loss of grassland-derived SOC (Fig. 4b). The conversion of Imperata grassland to oil palm plantation resulted in a loss of 10.1 Mg C ha−1 grassland-derived SOC over 11 years with SOC loss rate of −0.92 Mg C ha−1 yr−1 (Fig. 4c). Compared to the SOC stocks in primary dipterocarp forest, net increase in SOC stocks in Macaranga regeneration forest after the fires was 23.9 Mg C ha−1 over 30 years, with an average SOC accumulation rate of 0.8 Mg C ha−1 yr−1 (Fig. 5). The accumulation rate of Acacia-derived SOC remains high (1.2 Mg C ha−1 yr−1) for 23 years, while the initial accumulation rate of grassland-derived SOC dropped from 1.6 Mg C ha−1 yr−1 to an average SOC accumulation rate of 0.2 Mg C ha−1 yr−1 for 32 years (Fig. 5). The grassland-derived SOC in Acacia plantation was decomposed at slower rates, compared to grassland-derived SOC in oil palm plantation and forest-derived SOC in Imperata grassland (Table 3).

Table 2 Chemical properties of leaf and root litter samples
Full size table
Fig. 3
figure 3

Changes in soil 13C natural abundance in forest change to Imperata grassland and Macaranga forest (a) and Acacia or oil palm plantation on Imperata grassland (b). Bars represent standard errors (N = 3)

Full size image
Fig. 4
figure 4

Changes in soil carbon stocks (0–40 cm) under different land-uses. Changes from primary dipterocarp forest to the grassland (a), primary dipterocarp forest to grassland, followed by Acacia plantation (b), and primary dipterocarp forest to grassland, and to oil palm plantation (c)

Full size image
Fig. 5
figure 5

Changes in soil carbon stocks derived from primary dipterocarp forest under Imperata grassland, changes in soil carbon stocks derived from grassland under Acacia or oil palm plantation, and soil carbon gain under Imperata grassland, Macaranga, and Acacia

Full size image
Table 3 Decomposition rates of grassland-derived and forest-derived soil organic matter under different land use
Full size table

The SOC stocks in the deeper soil horizons (40 to 100 cm depth) amounted to 16.3 ± 1.1 Mg C ha−1 under primary dipterocarp forests. However, grassland-derived SOC accounted for <20% of SOC stocks (40–100 cm; 18.5 ± 0.5 Mg C ha−1) and continuous grassland resulted in a gain of 3.7 Mg C ha−1 SOC stocks (40–100 cm) for 32 years, which was far smaller than a gain of 16.0 Mg C ha−1 SOC stocks (0–40 cm; Fig. 5). There were no significant changes in SOC stocks (40–100 cm) and their soil δ13C between primary dipterocarp and Macaranga forests (Fig. 3a), while Acacia plantation increased SOC stocks (40–100 cm; 24.3 ± 1.0 Mg C ha−1) and lead to an increase in soil δ13C (Fig. 3b) and a gain of 8.0 Mg C ha−1 SOC stocks (40–100 cm) for 23 years, which was also smaller than a gain of 28.5 Mg C ha−1 SOC stocks (0–40 cm; Fig. 5).

Decomposition rates of leaf and root litters and cellulose paper

Lignin concentrations were significantly higher in forest litters (PF, MF, AF) than in grass and oil palm litters (P < 0.05; Table 2). The Acacia, Macaranga, and Imperata grass leaf litters were rich in N, P, and K, respectively (Table 2). The C/N ratio of Imperata grass leaf litter was highest among the plant samples, but its lignin to N ratio was lower than those in the PF, MF, and AF litters (Table 2). The decomposition rate factors of C in the organic layer in MF and AF were significantly lower than in OP, IG, and PF (Table 4).

Table 4 Litter turnover and decomposition rates of cellulose paper, leaf litter, and root litter
Full size table

The litter bag tests showed that the rates of mass loss differed markedly between substrates (Fig. 6). The coefficient of determination (R2) for fitting to the Eq. 7 ranged from 0.85 to 0.98, and rate constants (k) of litter decomposition were significant (P < 0.05; Table 4). The k values of oil palm litter decomposition were lowest among the five sites, followed by those for Imperata grass and forest litters (Table 4). The results were consistent for leaf and root litters (Table 4), as shown by a positive correlation between the k values of leaf litter and root litter decomposition (Fig. 7). Among forest litters, the k values for Macaranga leaf and root litter decomposition were lower than those for the other two sites (Table 4). The k values for leaf and root litter decomposition were correlated negatively with litter lignin to N ratios (R = −0.76, P < 0.05, N = 10; Fig. 8a), but not with litter C/N ratios (Table 2) nor with microbial biomass C (Table 1). There was no significant difference between the k values of root and leaf litter decomposition for respective plant species, but the slope of linear regression between them significantly higher than the slope of 1:1 line (Fig. 7).

Fig. 6
figure 6

Proportion of litter remaining relative to the initial litter mass in litter bag field incubation. Note that leaf and root litter bags were buried in the organic layer and the surface soil, respectively. Bars represent standard errors (N = 3)

Full size image
Fig. 7
figure 7

Relationship between rate constants of leaf litter decomposition in the organic layers and root litter decomposition in the surface soil (0–5 cm). Bars represent standard errors (N = 3)

Full size image
Fig. 8
figure 8

Relationship between rate constants of leaf and root litter decomposition and litter lignin to nitrogen ratio (a) and relationship between rate constants of leaf and root litter and cellulose paper (b). Bars represent standard errors (N = 3)

Full size image

The k values of cellulose decomposition for the IG and OP sites were significantly higher than those for the forest sites (Table 4). The k values of cellulose decomposition were correlated positively with soil pH (Fig. 8b). Due to the small variation in mean annual soil temperature at 5 cm depth (27 to 28 °C) in five sites, the k values of cellulose decomposition were independent of cumulative soil temperatures. The k value of Imperata grass root litter decomposition was significantly higher in the IG soil than in the AF soil (P < 0.05; Table 5). In contrast, the k values of Macaranga root litter decomposition did not differ significantly between the PF and MF soils (Table 5).

Table 5 Decomposition rates of Macaranga or Imperata root litters in the soils under different vegetation
Full size table

Fine root biomass and annual litter input

Fine root biomass and annual production were significantly higher at the IG site than at the three forest sites (P < 0.05; Table 6). Among the forest sites, there was no significant difference in litterfall C inputs (Table 6). Similarly, there was no significant difference in fine root production between the forest sites (Table 6). Using the changes in SOC stocks under the present fallow or plantation vegetation (Fig. 2), the annual SOC budgets under the present vegetation ranged from −0.6 to 1.2 Mg C ha−1 yr−1 (Table 6). The average proportion of annual SOC gain relative to annual C inputs (litterfall + fine root production) varied from −11.9 to 25.2% (Table 6). Among the four fallow or plantation systems, the proportion of annual SOC gain relative to annual C input was negatively correlated with soil pH (Fig. 9) or the k values of root litter decomposition obtained under the current vegetation, respectively.

Table 6 Aboveground and fine root biomass, annual litter production, and proportion of soil organic carbon gain relative to carbon input
Full size table
Fig. 9
figure 9

Relationship between soil pH and the proportion of annual gain in soil organic carbon stocks relative to annual carbon input under land use change. Data sources are Tables 4 and 6. Bars represent standard errors (N = 3)

Full size image

Discussion

Effects of plant traits on litter decomposability under different land uses

Changes in vegetation types influence SOC stocks through effects on litter input and microbial litter decomposition (Stockmann et al. 2013). Microbial litter decomposition increases with decreasing litter recalcitrance or lignin/N ratios (Taylor et al. 1989). Thus, we hypothesize that the changes from primary dipterocarp forest to lignin-poor grassland or oil palm and N-rich Acacia plantations could accelerate microbial litter decomposition and a loss of SOC stocks. The positive correlation between leaf and root litter decomposition suggests that plant traits strongly influence litter decomposition in both of the organic and mineral soil layers (Fig. 7). The negative correlation between litter decomposability and lignin/N ratios (Fig. 8a) supports the principle that low N availability, as well as lignin abundance, reduces litter decomposition (Taylor et al. 1989; Prieto et al. 2016), and that the lignin/N ratio could be a rough predictor of litter decomposability across major plants in the tropical forests and agroecosystems (Fig. 8a).

The lower decomposability of root litter than of leaf litter (Fig. 7) is consistent with the results of five tree species in temperate forest (Makita and Fujii 2015). The larger contributions of recalcitrant root litters to soil C storage have also been suggested by Uselman et al. (2007), but the decomposability of root litters in our study is lower than leaf litters despite lower recalcitrance or lignin/N ratios of root litters at all sites except for PF (Fig. 8a). Considering the limited physical protection of root litters in the litter bag tests (e.g., soil aggregates), the reduced microbial or enzyme activities due to the increased acidity (Fig. 8b), rather than litter recalcitrance, can contribute to the lower root litter decomposability in the mineral soil (Fig. 7).

In addition to leaf and root decomposability, the differences in pathways of litter supply to the mineral soil can affect SOC storages in forests and grasslands (Qualls 2000). In the forests, majority of aboveground litterfall-C is respired and a proportion of dissolved organic matters [e.g., 30% of litterfall-C from Fujii et al. 2009b] are supplied into the mineral soil. The dissolved organic C fluxes leaching from the organic horizon into the mineral soil [ca. 0.5 Mg C ha−1 yr−1 from Fujii et al. 2009b] are smaller than root litter inputs or fine root production in the Imperata grassland soil (2.6 Mg C ha−1 yr−1; Table 6). The direct and greater inputs of fine root litter in Imperata grassland (Table 6) can contribute to the greater initial SOC gain (Figs. 3 and 4a), compared to three forest sites, where the annual C input is dominated by aboveground litterfall (Table 6).

Effects of soil acidity on litter decomposability under different land uses

The positive correlations between cellulose decomposition rate constants and soil pH (Fig. 8b) support the hypothesis that soil acidity, as well as litter quality (lignin/N ratio; Fig. 8a), affect litter decomposition rates. The limited decomposition of grass litters and preservation of grassland-derived SOC under Acacia plantation (Tables 3, 5 and Fig. 5) is consistent with the limited microbial activities of litter decomposition at low pH (Fig. 8b). This can be explained by (1) the decreased litter-degrading enzyme activities at low pH and (2) the increased Al toxicity to microorganisms (Illmer and Mutschlechner 2004) and deactivation of enzymes by Al (Scheel et al. 2008). Especially, degrading enzyme activities of cellulose, major constituent of plant litters, are sensitive to soil acidification (Fig. 3b) and reduce at pH < 5.5 (optimal pH) (Criquet 2002; Hayakawa et al. 2014). This could retard overall litter decomposition in the PF, MF, and AF forests. Drastic soil acidification due to nitrification of the N fixed by leguminous Acacia trees (Fig. 1) contributes to SOC accumulation by limiting microbial activities of grassland-derived SOC decomposition (Fig. 8b; Table 3). This result contrasts with the lack of difference in Macaranga root litter decomposition rates between primary dipterocarp and Macaranga forest sites (Table 5), where the change in soil pH was smaller (Fig. 1) and the decomposition of lignin-rich substrates were less sensitive to soil acidification, compared to cellulose-rich grass root litters (Table 2). Our previous study also supports that ligninolytic enzyme activities are not reduced under acidic condition (Fujii et al. 2012). The sensitivity of litter decomposition to soil pH change could depend on plant quality (lignin/N) and the magnitude of soil acidification (Fig. 8a,b).

Effects of land-use change on soil organic matter storage

Changes in vegetation cover or land use could cause an increase or a decrease in SOC stocks (Veldkamp 1994; Cerri et al. 2004). The vegetation cover that maximizes soil C stocks within the initial 10 years is Imperata grassland; thereafter, Acacia plantation and Macaranga forest maximize soil C stocks (Fig. 2). Initial increases of grassland-derived SOC [1.3 Mg C ha−1 yr−1 (1983–1992); Fig. 5] are consistent with the reports in tropical grasslands or pastures (Cerri et al. 2004; Yonekura et al. 2013). The relatively high SOC accumulation rates [0.6 Mg C ha−1 yr−1 (1983–1992); Fig. 2], compared to the global average after land use changes [ca. 0.3 Mg C ha−1 yr−1 from Post and Kwon 2000], could be related to the following Imperata grassland characteristics: rapid colonization after frequent fires (Kiyono 2000), high primary productivity (Hartemink 2001), and large C inputs from root litter (Table 6; Astapati and Das 2010). The saturation of grassland-derived SOC stocks (Fig. 5) is explained by the efficient stabilization of initial grass-derived C in aggregates or at sorption sites with limited capacity (Hayakawa et al. 2014), but additional C inputs are not fully protected in aggregates or sorption sites already occupied by forest-derived or grassland-derived SOC (Fig. 5). The variation in mean residence times of grassland-derived SOC and Imperata root litters in Acacia and oil palm plantation (Tables 3 and 5) suggest that decomposition of grassland-derived organic matter can be accelerated by land-use change to oil palm plantation (Table 3 and Fig. 8a). In addition to intrinsic nature of lignin-poor grass litters and high biodegradability of C4 plant (Table 2; Wynn and Bird 2007), higher soil pH compared with forest sites (Fig. 1 and Table 1) are favorable for rapid mineralization of litter and SOM by microbes (Fig. 8b). The SOC accumulation rates in the Imperata grassland–Acacia sequence or Macaranga forest (0.8 to 1.2 Mg C ha−1 yr−1; Fig. 5 and Table 6) are close to the upper limit reported of global dataset (Post and Kwon 2000) and are supported by the low decomposability of forest litters (Fig. 7 and Table 4) and reduced microbial decomposition activities in acidified soils (Table 5 and Fig. 8b). Soil acidification, which is accelerated by nitrification of the fixed N under Acacia plantation (Fig. 1), and input of recalcitrant litter reduce microbial decomposition (Fig. 8a) and contributes to an increase in soil C stocks (Fig. 2 and Table 3). The high rates of SOC accumulation in Macaranga forests (Fig. 5) compared with primary dipterocarp forests (Fig. 2) are caused by inputs of coarse woody debris derived from the fires (Toma et al. 2017) and higher litter recalcitrance (Table 4). Non-myrmecophytic Macaranga gigantea could produce lignin-rich litters compared with the other Macaranga species with ant defense system (Eck et al. 2001; Table 2). In addition to litter recalcitrance (Fig. 8a), the greater SOC gain in the soils with lower pH (Fig. 9) suggests the role of soil acidification in the retardation of microbial decomposition and SOC preservation under fallow and plantation vegetation (Fig. 8b and Table 3).

Fallow systems to maximize soil carbon storage and implications for fertility improvement

The roles of grass fallow and Acacia plantation in restoration of soil organic matter has been recognized (Islam and Weil 2000; Yonekura et al. 2012), while we found that the optimum and feasible land-use strategies could vary depending on the length of fallow period, available resources, and the original acidity of soil. In a short-term (<10 years) fallow system, SOC stocks can be maximized by continuous Imperata grassland, rather than natural secondary Macaranga forest (Fig. 5). The potential importance of grassland or pasture for SOC accumulation has been confirmed in several tropical regions (Cerri et al. 2004; Yonekura et al. 2012; Sugihara et al. 2019), but one of our novel findings is saturation of SOC storage in short-term grassland fallow (Fig. 5). In a relatively long-term (>10 years) fallow system, SOC stocks can be maximized by the Imperata grassland–Acacia sequence or Macaranga forest (Fig. 5). Regarding essential macronutrients (N, P, K) for plants, Imperata grass leaf input leads to high C/N ratios (Table 1). Imperata grass and Macaranga have been regarded as common and notorious weeds or bush by people native to this region, and Acacia trees have invasive, despite commercial merits in plantation, in natural dipterocarp forests in Southeast Asia (Ohta et al. 2000). The high availability of these three litter resources and the difference in nutrient availability among them could allow small-holder farmers to supply NPK in low-input agriculture (Table 2). For example, the amendment of N-rich Acacia leaf litters into N-poor Imperata grassland soil, amendment of P-rich Macaranga leaf litters into P-poor Acacia forest soil, and amendment of K-rich Imperata leaf litters into Macaranga forest could increase soil fertility without high fertilizer costs in tropical low-input agriculture (Table 1). Fallow systems that combine three vegetation covers could be one of land-use strategies that increase soil productivity and potentially reduce deforestation pressure in tropical forests.