Introduction
Tropical forests store over 25% of global terrestrial carbon (Bonan Reference Bonan2008) and account for more than one-third of the world’s net primary production (Beer et al. Reference Beer, Reichstein, Tomelleri, Ciais, Jung, Carvalhais, Rödenbeck, Arain, Baldocchi, Bonan, Bondeau, Cescatti, Lasslop, Lindroth, Lomas, Luyssaert, Margolis, Oleson, Roupsard, Veenendaal, Viovy, Williams, Woodward and Papale2010), thereby playing a crucial role in climate change mitigation. Despite their importance, they have been deforested at an alarming rate of 9.2 million ha year⁻¹ (FAO 2020) primarily due to agriculture and cattle ranching (Keenan et al. Reference Keenan, Reams, Achard, De Freitas, Grainger and Lindquist2015, IPBES Reference Brondizio, Settele, Díaz and Ngo2019). Upon land abandonment, tropical forests can regrow naturally through secondary succession (Poorter et al. Reference Poorter, Amissah, Bongers, Hordijk, Kok, Laurance, Lohbeck, Martínez-Ramos, Matsuo, Meave, Muñoz, Peña-Claros and Van Der Sande2023, Reference Poorter, Van Der Sande, Amissah, Bongers, Hordijk, Kok, Laurance, Martínez-Ramos, Matsuo, Meave, Muñoz, Peña-Claros, Van Breugel, Herault, Jakovac, Lebrija-Trejos, Norden and Lohbeck2024), and these secondary forests currently represent over half of the total tropical forest area (FAO 2015). Secondary tropical forests often exhibit rapid above and belowground carbon accumulation rates through fast rates of structural development, litter production, and decomposition. The main drivers and mechanisms of the aboveground carbon stock have been well studied in secondary tropical forests (Finegan et al. Reference Finegan, Peña-Claros, De Oliveira, Ascarrunz, Bret-Harte, Carreño-Rocabado, Casanoves, Díaz, Velepucha, Fernandez, Licona, Lorenzo, Negret, Vaz and Poorter2015, Poorter et al. Reference Poorter, Bongers, Aide, Zambrano, Balvanera, Becknell, Boukili, Brancalion, Broadbent, Chazdon, Craven, Almeida-Cortez, Cabral, Jong, Denslow, Dent, Dewalt, Hernandez-Stefanoni, Jakovac, Junqueira, Kennard, Letcher, Licona, Lohbeck, Marín-Spiotta, Martínez-Ramos, Massoca, Meave, Mesquita, Mora, Muñoz, Muscarella, Nunes, Ochoa-Gaona, Oliveira, Orihuela-Belmonte, Peña-Claros, Pérez-García, Piotto, Powers, Rodríguez-Velázquez, Romero-Pérez, Ruíz, Saldarriaga, Sanchez-Azofeifa, Schwartz, Steininger, Swenson, Toledo and Uriarte2016, Matsuo et al. Reference Matsuo, Poorter, Van Der Sande, Mohammed Abdul, Koyiba, Opoku, De Wit, Kuzee and Amissah2025). Yet, few studies have focused on the belowground carbon stock (but see Hughes et al. Reference Hughes, Kauffman and Jaramillo1999, de Koning et al. Reference De Koning, Veldkamp and López-Ulloa2003, Powers & Marín-Spiotta Reference Powers and Marín-Spiotta2017), although soil organic carbon (SOC) accounts for approximately 70% of the total carbon stock in early successional forests (Matsuo et al. Reference Matsuo, Poorter, Van Der Sande, Mohammed Abdul, Koyiba, Opoku, De Wit, Kuzee and Amissah2025).
Soil organic carbon stock represents the long-term steady-state balance between aboveground litter input and litter breakdown through decomposition (Blume et al. Reference Blume, Brümmer, Fleige, Horn, Kandeler, Kögel-Knabner, Kretzschmar, Stahr and Wilke2016). In a young secondary forest, SOC stock is a net result of legacies of slash and burning of aboveground vegetation, carbon loss during cultivation, and carbon recovery through carbon input by leaf and root litter and root exudates (Powers & Marín-Spiotta Reference Powers and Marín-Spiotta2017). During the burning phase, burning of aboveground biomass can enrich soil carbon through charcoal deposition, while burning soil organic matter can deplete soil carbon stock (García-Oliva et al. Reference García-Oliva, Sanford and Kelly1999). During the crop production phase, soil tillage and high microbial activity due to high temperature may deplete SOC through microbial respiration and decomposition of organic matter. In contrast, during the fallow phase, increased input of more recalcitrant, lignin-rich litter from regenerating vegetation results in accumulation of SOC (Veldkamp et al. Reference Veldkamp, Schmidt, Powers and Corre2020, Gavito et al. Reference Gavito, Paz, Barragán, Siddique, Arreola-Villa, Pineda-García and Balvanera2021, van der Sande et al. Reference Van Der Sande, Powers, Kuyper, Norden, Salgado-Negret, Almeida, Bongers, Delgado, Dent, Derroire, Santo, Dupuy, Fernandes, Finegan, Gavito, Hernández-Stefanoni, Jakovac, Jones, Veloso, Meave, Mora, Muñoz, Pérez-Cárdenas, Piotto, Álvarez-Dávila, Caceres-Siani, Dalban-Pilon, Dourdain, Du, Villalobos, Nunes, Sanchez-Azofeifa and Poorter2022).
After a major disturbance, soil carbon stock can rapidly recover to its undisturbed state through secondary succession (Poorter et al. Reference Poorter, Craven, Jakovac, Sande, Amissah, Bongers, Chazdon, Farrior, Kambach and Meave2021, van der Sande et al. Reference Van Der Sande, Powers, Kuyper, Norden, Salgado-Negret, Almeida, Bongers, Delgado, Dent, Derroire, Santo, Dupuy, Fernandes, Finegan, Gavito, Hernández-Stefanoni, Jakovac, Jones, Veloso, Meave, Mora, Muñoz, Pérez-Cárdenas, Piotto, Álvarez-Dávila, Caceres-Siani, Dalban-Pilon, Dourdain, Du, Villalobos, Nunes, Sanchez-Azofeifa and Poorter2022), but its recovery rates vary across space and time due to different abiotic and biotic conditions (Martin et al. Reference Martin, Newton and Bullock2013, van der Sande et al. Reference Van Der Sande, Powers, Kuyper, Norden, Salgado-Negret, Almeida, Bongers, Delgado, Dent, Derroire, Santo, Dupuy, Fernandes, Finegan, Gavito, Hernández-Stefanoni, Jakovac, Jones, Veloso, Meave, Mora, Muñoz, Pérez-Cárdenas, Piotto, Álvarez-Dávila, Caceres-Siani, Dalban-Pilon, Dourdain, Du, Villalobos, Nunes, Sanchez-Azofeifa and Poorter2022). The recovery of soil carbon stock is primarily determined by three ecosystem processes: the carbon influx through leaf, branch, and root litter input (litter production rates), carbon transfer rates from litter to soil (litter decomposition rates), and carbon efflux from soil to the atmosphere (soil respiration rates) (Högberg et al. Reference Högberg, Nordgren, Högberg, Ottosson-Löfvenius, Bhupinderpal-, Olsson and Linder2005, van der Sande et al. Reference Van Der Sande, Powers, Kuyper, Norden, Salgado-Negret, Almeida, Bongers, Delgado, Dent, Derroire, Santo, Dupuy, Fernandes, Finegan, Gavito, Hernández-Stefanoni, Jakovac, Jones, Veloso, Meave, Mora, Muñoz, Pérez-Cárdenas, Piotto, Álvarez-Dávila, Caceres-Siani, Dalban-Pilon, Dourdain, Du, Villalobos, Nunes, Sanchez-Azofeifa and Poorter2022). All three processes are, in turn, influenced by a combination of abiotic and biotic factors. For instance, leaf litter production rates are strongly determined by forest structure, as it is associated with higher biomass in foliage and branches, which leads to a regular turnover of these plant organs (Capellesso et al. Reference Capellesso, Scrovonski, Zanin, Hepp, Bayer and Sausen2016, Matsuo et al. Reference Matsuo, Poorter, Van Der Sande, Mohammed Abdul, Koyiba, Opoku, De Wit, Kuzee and Amissah2025). Litter decomposition and soil respiration rates are influenced by environmental conditions that regulate the activity of the soil decomposer community (e.g., earthworms, termites, and fungi). For instance, warmer and wetter climates typically enhance microbial and faunal activity, thereby accelerating both litter breakdown and soil respiration (Hanson et al. Reference Hanson, Edwards, Garten and Andrews2000). In addition, litter decomposition rate is influenced by litter quality. For example, increased litter nutrient concentrations (e.g., nitrogen or phosphorus) can increase the decomposition rate because they increase soil microbial and faunal activity by providing essential nutrients (Berg & Laskowski Reference Berg and Laskowski2005). Yet, we lack a comprehensive study that assesses the roles of abiotic and biotic factors in determining carbon fluxes and SOC stock, especially in the context of young secondary tropical forests.
To this end, we assessed the roles of environmental conditions and forest attributes in determining three ecosystem processes that shape SOC in young secondary tropical forests on abandoned agricultural fields in Ghana. We addressed three research questions and their corresponding hypotheses:
-
1) How does macroclimate (climatic wetness) influence forest attributes (aboveground biomass, microclimate, and litter nutrient concentration)? We hypothesise that wetter forests have a) higher aboveground biomass because of a longer growing season and a more productive environment, b) higher soil temperature because of higher air temperature, c) higher soil moisture because of a higher amount of rainfall and d) higher leaf carbon concentrations but lower leaf nitrogen and phosphorus concentrations because wet forests are dominated by evergreen species that invest more in carbon-rich structural defences to increase leaf longevity (Coley Reference Coley1986, Sterck et al. Reference Sterck, Poorter and Schieving2006).
-
2) How do macroclimate and forest attributes affect ecosystem processes (litter production, litter decomposition, and soil respiration)? We hypothesise that a) litter production is higher in forests with greater aboveground biomass due to increased foliage production, and higher in dry forests for a given aboveground biomass due to faster leaf and branch turnover rates associated with a higher proportion of deciduous species, b) litter decomposition rate increases with i) higher soil temperature and moisture as they enhance microbial activity and ii) litter nutrient concentration (N and P) because of higher leaf decomposability, and c) soil respiration increases with i) microclimate and leaf nutrient concentrations because of higher microbial activity, and ii) with aboveground biomass because of higher (root) autotrophic respiration.
-
3) How do macroclimate and ecosystem processes affect SOC stock? We hypothesise that SOC stock a) increases with higher litter input and decomposition rates as they build up soil carbon stock and b) decreases with soil respiration because it releases carbon from the soil to the atmosphere.
Materials and methods
Study site
The study was carried out in dry and wet Ghanaian tropical forests to study the effect of macroclimate on soil carbon stock (Fig. S1).
Dry semi-deciduous forests. The study sites are located near the town of Abofour, Ashanti region in Ghana (7°11’N, 1°73’ W). The region has a mean annual precipitation of 1290 mm, dry season precipitation of 28 mm/month (Ghana Meteorological Service records, data 1973–2009), maximum monthly temperature of 30.6°C, and minimum monthly temperature of 21.2°C (Amissah et al. Reference Amissah, Mohren, Kyereh, Agyeman and Poorter2018). The soil is sandy loam with patches of clay (Forestry Division 1963) and slightly acidic (pH = 6.2) (Matsuo et al. Reference Matsuo, Poorter, Van Der Sande, Mohammed Abdul, Koyiba, Opoku, De Wit, Kuzee and Amissah2025).
The main crops in this area are maize, tomato, and papaya, along with teak plantations. Land use is relatively intensive compared to the wet forest sites, primarily due to two cultivation cycles of maize per year. Farmers commonly use fire to clear fields of weeds and crop residues, producing ash that serves as a natural fertiliser. In the young secondary dry forests, Broussonetia papyrifera, Chromolaena odorata, and Griffonia simplicifolia were the common species (Matsuo et al. Reference Matsuo, Amissah, Kok and Poorter2023). These forests had a mean maximum diameter at breast height (DBH) of 19.9 cm (± 5.0 SD) and a mean canopy height of 10.4 m (± 2.6 SD) across the 17 plots.
Wet-evergreen forests. The study sites are located near the town of Bonsa, Western region of Ghana (5°10' N, 2°02' W). The region is characterised by a mean annual precipitation of 1808 mm and a dry season precipitation of 82.6 mm/month (Ghana Meteorological Service record, data 1973–2009). The monthly mean maximum temperature is 32.0°C, and the mean monthly minimum temperature is 22.8°C (Ghana Meteorological Service records, data 1973–2011). The soil is sandy loam with patches of clay (Forestry Division 1963). Due to higher weathering and leaching, soils are more acidic (pH = 4.9) and less fertile than in dry semi-deciduous forests (Amissah et al. Reference Amissah, Mohren, Kyereh, Agyeman and Poorter2018, Matsuo et al. Reference Matsuo, Poorter, Van Der Sande, Mohammed Abdul, Koyiba, Opoku, De Wit, Kuzee and Amissah2025).
The main crops are cocoa and rubber tree plantations, as well as cassava fields cultivated with either short-lived (approximately 0.5-year) or long-lived (up to 2-year) varieties that are produced for local consumption or local markets. Cassava is often intercropped with cocoa, yams, and banana trees. Because cassava is typically managed as a perennial crop in these systems, the frequency of burning is low, and the overall land-use intensity prior to abandonment is relatively minimal. In the young secondary wet forests, Harungana madagascariensis, Macaranga barteri, and Rauvolfia vomitoria were the common species (Matsuo et al. Reference Matsuo, Amissah, Kok and Poorter2023). These forests exhibited relatively more developed structure, with a mean maximum DBH of 22 ± 10.8 cm and a mean canopy height of 13.0 ± 3.6 m across the 18 plots.
Sampling framework
In 2021, we established 17 vegetation plots (25 × 25m) on recently abandoned (0-1 year) maize fields in dry forests and 18 vegetation plots on abandoned cassava fields in wet forests (Matsuo et al. Reference Matsuo, Poorter, Van Der Sande, Mohammed Abdul, Koyiba, Opoku, De Wit, Kuzee and Amissah2025). Therefore, when most measurements for the present study were conducted in 2023 and early 2024, the forests were between 2.3 and 3.6 years old since abandonment. The fallow period (year) was determined based on the interviews with the farmers and field observations. In each plot, the DBH (DBH, cm) and height of all woody individuals (> 1cm DBH) were measured annually between 2021 and 2023. Individual tree basal area was then calculated as π*(DBH/2)2. For multi-stemmed individuals, we counted the number of stems and measured the DBH of the largest stem (DBHlarge) and an average-sized stem (DBHaverage). The total number of other stems for each individual (Nstems) was counted, and the basal area was estimated (eq. 1).
$$Individual\,basal\,area = 0.25 \times \pi $$$\times [DBH_{large}^{2} + DBH_{average}^{2} \times (N_{stems}-1)]$
Aboveground biomass. For each individual, aboveground biomass (AGB, kg) was estimated using allometric equations based on DBH and species-specific wood density (WD, g cm-3). For trees and shrubs, we used the allometric equation developed specifically for young secondary forests in Ghana (eq. 2; Matsuo et al. Reference Matsuo, Poorter, Van Der Sande, Mohammed Abdul, Koyiba, Opoku, De Wit, Kuzee and Amissah2025). This equation was derived from the destructive sampling of 335 individuals in forests approximately five years after land abandonment, located in the same dry and wet forest regions as our study sites. Because the model was developed within the same ecological and successional context, it provides high relevance and accuracy for estimating aboveground biomass in our plots. The same equation was applied to both dry and wet forests, as species differences between these forest types were accounted for through WD. When the WD data for some species were not available, available local WD data at the highest taxonomic resolution (at genus or family level) or the average WD of each site were used. Liana diameter was also measured at 1.3 m above the rooting point, but because lianas have different allometric relationships compared to trees and shrubs, their aboveground biomass was estimated using a liana-specific allometric equation developed for secondary tropical forests in Ghana (eq. 3) (Addo-Fordjour & Rahmad Reference Addo-Fordjour and Rahmad2013).
$${AGB\ \left( {tree/shrub} \right) = {\rm{exp }}[ - 1.65 + 2.14\times {\rm{ln}}\ \left( {{\rm{DBH}}} \right) + 0.45\times\left( {{\rm{WD}}} \right)]}$$
$$AGB\ \left( {liana} \right) = - 0.36 + 1.9\times{{ {\rm{DBH}}}}$$
Total aboveground biomass per plot (tonne ha-1) was then calculated by summing the biomass of all stems and multiplying the values by 16, and total aboveground biomass from the 2023 census was used in this study.
Microclimate. A data logger (TMS-4 data logger; TOMST s.r.o., Prague, Czech Republic) was installed at the centre of each plot in 2021 to monitor soil moisture and temperature at 10 cm depth at 15-minute intervals. The recorded data were calibrated using the HOBO Max Soil Moisture and Temperature Data logger (Onset Computer, Bourne, MA) to convert the obtained values to volumetric soil moisture content (cm3 cm-3). Soil moisture and soil temperature data from April to May 2023 were used in the analysis to link microclimate to the decomposition rates.
Litter production. Each plot was subdivided into four equal quadrants (12.5 m × 12.5 m), with a litter trap (50 × 50 cm, placed at 1.0 m height) installed at the centre of each quadrant to quantify litter production. Litter was collected monthly over eight months (January to August 2023), which covers approximately three months of the dry season and five months of the wet season to capture seasonal variability in litterfall. During the peak rainy season (June and July), litter samples were collected every two weeks to minimise the risk of decomposition within the traps. In the remainder of the year, litter collection was done monthly due to a lower risk of decomposition and logistical and financial constraints. Although litter traps were not placed at the very onset of the dry season, they were installed early enough to capture the majority of dry season litterfall. Nonetheless, we acknowledge that the timing of installation may have led to a slight underestimation of total annual litter production, particularly in dry forests dominated by deciduous species. Litter samples were sorted into leaf materials, branches, reproductive parts (flowers, fruits, and seeds), and animal droppings. Afterwards, they were oven-dried at 65°C for 48 hours and weighed. The litter production rate (tonne ha⁻¹ year⁻¹) for each plot was calculated by summing the total dry weight of litter (excluding animal droppings), dividing by the number of collection days to obtain a daily rate, multiplying by 365 to extrapolate to an annual basis, and then multiplying by 10,000 to convert the result to tonne ha⁻¹.
Litter nutrient concentration. Leaf litter samples were pooled per plot, oven-dried at 65°C for 48 hours, and analysed for nitrogen (N), phosphorus (P), and carbon (C) at Wageningen University & Research. Total leaf litter C concentration was measured using a CHN analyser, while total N and P concentrations were determined spectrophotometrically using a segmented flow system (Skalar San++ System) from leaf digests prepared with a mixture of H2SO4–Se and salicylic acid (Novozamsky et al. Reference Novozamsky, Houba, Van Eck and Van Vark1983).
Litter decomposition rate. Litter decomposition rates were estimated using the litter bag method. A plot-specific litter bag was used to capture the variability of leaf nutrient concentrations among and within plots. Each litter bag contained litter collected from its corresponding litter trap. Two grams of the leaf litter that was thoroughly mixed and oven-dried at 65°C for 48 hours, was placed into nylon bags (8 × 14 cm). We oven-dried the litter samples before preparing the litter bags in order to (1) standardise the initial dry weight of litter across all bags, ensuring comparability of decomposition rates, and (2) prevent microbial decomposition during the period between litter collection, bag preparation, and field incubation. However, we acknowledge that the oven-drying process could alter microbial communities and activity associated with the litter, potentially reducing decomposition rates. The bags had a mesh size of 1.03 mm to ensure the involvement of a complete set of decomposing organisms (Pérez-Harguindeguy et al. Reference Pérez-Harguindeguy, Díaz, Garnier, Lavorel, Poorter, Jaureguiberry, Bret-Harte, Cornwell, Craine, Gurvich, Urcelay, Veneklaas, Reich, Poorter, Wright, Ray, Enrico, Pausas, Vos, Buchmann, Funes, Quétier, Hodgson, Thompson, Morgan, Steege, Van Der Heijden, Sack, Blonder, Poschlod, Vaieretti, Conti, Staver, Aquino and Cornelissen2013).
Four litter bags were incubated per plot, yielding a total of 140 bags (35 plots × 4 bags per plot). The litter bags were incubated at a depth of 1-2 cm below ground to reduce moisture heterogeneity among samples (Pérez-Harguindeguy et al. Reference Pérez-Harguindeguy, Díaz, Garnier, Lavorel, Poorter, Jaureguiberry, Bret-Harte, Cornwell, Craine, Gurvich, Urcelay, Veneklaas, Reich, Poorter, Wright, Ray, Enrico, Pausas, Vos, Buchmann, Funes, Quétier, Hodgson, Thompson, Morgan, Steege, Van Der Heijden, Sack, Blonder, Poschlod, Vaieretti, Conti, Staver, Aquino and Cornelissen2013), and all the litter bags were collected after approximately 4 weeks of incubation. Litter bags were placed 50 cm from each litter trap to facilitate the retrieval of the litter bags at harvest. After collecting litter bags from the field, soil and root debris were carefully removed with water and tweezers, and cleaned litter was oven-dried at 65°C for 48 hours and weighed. Then, the decomposition rate was calculated as the weight loss percentage per month (eq. 4).
$$\eqalign{& Weight\, loss\, percentage \cr & = (dry\, weight\, loss\, in\, 30\, days/initial\, dry\, weight) \times 100 \cr} $$
Soil respiration. Plot-level soil respiration was measured using a portable system (LCpro T) connected to a soil chamber (LCPro Soil Hood; ADC BioScientific Ltd., Hoddesdon, UK). A steel collar (external Ø 11.1 cm) was inserted at each sampling point at 1–5 cm below ground to prevent gas diffusion through the soil, while ensuring minimal soil disturbance. The collar was placed into the soil 15–30 minutes before the initial measurement to standardise the impact of chamber installation and facilitate comparison between plots. A pilot study was conducted to determine the minimum waiting time for gas exchange stabilisation before measurement. Spatial heterogeneity was accounted for by measuring soil respiration at three to five locations per plot. Then, the respiration rate of the plot was averaged for the analysis. Soil respiration rates were measured during the daytime from January 2–11, 2024.
Soil organic carbon stock. In 2023, four soil samples per plot were collected using a soil core (5 cm diameter) from each study site to estimate SOC stock. A simple soil ring was chosen as it has a lower sampling impact on the plot (Freschet et al. Reference Freschet, Pagès, Iversen, Comas, Rewald, Roumet, Klimešová, Zadworny, Poorter, Postma, Adams, Bagniewska-Zadworna, Bengough, Blancaflor, Brunner, Cornelissen, Garnier, Gessler, Hobbie, Meier, Mommer, Picon-Cochard, Rose, Ryser, Scherer-Lorenzen, Soudzilovskaia, Stokes, Sun, Valverde-Barrantes, Weemstra, Weigelt, Wurzburger, York, Batterman, Gomes de Moraes, Janeček, Lambers, Salmon, Tharayil and McCormack2021). A total of 140 soil samples were collected at a depth of 0–15 cm (a volume of 295 cm3) because at this depth, soil organic matter is strongly affected by litter production (Feng et al. Reference Feng, Wang, Ma, Fu and Chen2019), and the high presence of rocks beyond 15 cm limited the soil sampling (Matsuo et al. Reference Matsuo, Poorter, Van Der Sande, Mohammed Abdul, Koyiba, Opoku, De Wit, Kuzee and Amissah2025). Soil organic carbon content was then measured from the collected soil samples by the modified dichromate oxidation method of Walkey-Black (Nelson & Sommers Reference Nelson and Sommers1983) at the Soil Research Institute of Ghana (CSIR-SRI) in Kumasi, Ghana. The estimated SOC content is then multiplied by soil bulk density to express it in tonnes ha-1 in the first 15 cm of the soil.
Data analysis
We conducted either a t-test or a Wilcoxon rank-sum test, depending on the distribution and variability of the data, to assess differences in forest attributes, ecosystem processes, and SOC stock between wet and dry forests. Normality was checked using the Shapiro–Wilk test, and homogeneity of variances was evaluated using Levene’s test.
Structural equation modelling (SEM) was employed to investigate the cause-and-effect relationships among climatic wetness, forest attributes, ecosystem processes, and SOC stock to identify the primary drivers of SOC in young secondary forests. We ran a series of 6 different SEMs (with combinations of 2 microclimate x 3 litter nutrient concentration variables). For microclimate, soil moisture and temperature were included, and for litter nutrient concentration, leaf N, P, and C were used. As our best-fitted model, we selected models that were not rejected (P > 0.05, chi-squared test) and that have the highest absolute R2 value for SOC stock. Models were fitted using the statistical package ‘lavaan’ in R (Rosseel Reference Rosseel2012) (R version 4.4.1: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria).
Results
Compared to dry forests, wet forests showed significantly higher aboveground biomass, soil temperature and moisture, leaf litter C concentration, and litter production rates, but lower leaf litter N and P concentrations and litter decomposition rates (Fig. 1, Table S1). There were no significant differences in soil respiration or SOC stocks between the two sites.
Differences in forest attributes [aboveground biomass, soil temperature, soil moisture, and litter nitrogen (N), phosphorus (P), and carbon (C) concentration], ecosystem processes (litter production rate, litter decomposition rate, and soil respiration rate), and soil organic carbon stock between dry (orange, N = 17) and wet (blue, N = 18) secondary tropical forest plots in Ghana. Bars represent mean values ± standard error. P-values indicate the results of statistical comparisons between dry and wet forests based on either a t-test or a Wilcoxon rank-sum test (see Methods for details). Full statistical details are provided in Table S1.
The best-fitting SEM (Fig. 2) explained a moderate to substantial part of the variation in SOC stock (32%), litter production rate (38%), litter decomposition rate (50%), soil respiration rate (18%), aboveground biomass (33%), soil temperature (62%), and litter N concentration (77%). The best-fitting model selected soil temperature, rather than soil moisture, among the microclimatic variables, and leaf litter N concentration, rather than leaf litter P or C concentration, among the leaf litter nutrient variables. According to the best model, wet forests had significantly higher aboveground biomass and soil temperature, but lower litter N concentration. Climatic wetness did not have a direct effect on any of the studied ecosystem processes or on SOC stock. Aboveground biomass positively influenced litter production but negatively affected soil temperature. Litter decomposition rates increased significantly with litter production (quantity), litter N concentration (quality), and soil temperature (microclimate). Soil respiration was not significantly affected by any of the predictors. SOC stock increased significantly only with the litter decomposition rate. Figure 3 presents bivariate scatterplots showing all relationships among variables included in the best-fitting SEM model, and full statistical results are provided in Table S2.
Results of the best-fitting structural equation model (χ2 =16.8, df =9, p =0.052). Standardised path coefficients (β) and their corresponding significance levels (p-values) are shown along the arrows. The explained variance (R²) for each endogenous variable is indicated within the corresponding box. Solid arrows represent statistically significant relationships (p < 0.05), while dashed arrows indicate non-significant relationships.
Bivariate relationships between study variables included in the best structural equation model (SEM). Data are shown for dry (orange, N = 17) and wet (blue, N = 18) secondary tropical forest plots in Ghana. These relationships are illustrated using simple regressions to show pairwise associations and raw data patterns. Note that these are for visualisation purposes only and do not necessarily reflect the results or effect sizes from the SEM analysis.
Discussion
We evaluated the roles of abiotic and biotic drivers in determining the SOC stock in young secondary forests. Soil organic carbon only increased with litter decomposition rate, and was, surprisingly, not affected by litter production rates and soil respiration rates. Climatic wetness had an indirect positive effect on SOC stock by increasing aboveground biomass and soil temperature, and an indirect negative effect through a low litter nitrogen concentration. Below, we discuss the underlying drivers and mechanisms of SOC stock in young secondary tropical forests.
Wetter forests have higher aboveground biomass and soil temperature, but lower litter nutrient concentration
Aboveground biomass increased with climatic wetness (Figs. 1 and 2), likely due to a longer growing season and higher water availability. This is in line with studies from secondary (Poorter et al. Reference Poorter, Bongers, Aide, Zambrano, Balvanera, Becknell, Boukili, Brancalion, Broadbent, Chazdon, Craven, Almeida-Cortez, Cabral, Jong, Denslow, Dent, Dewalt, Hernandez-Stefanoni, Jakovac, Junqueira, Kennard, Letcher, Licona, Lohbeck, Marín-Spiotta, Martínez-Ramos, Massoca, Meave, Mesquita, Mora, Muñoz, Muscarella, Nunes, Ochoa-Gaona, Oliveira, Orihuela-Belmonte, Peña-Claros, Pérez-García, Piotto, Powers, Rodríguez-Velázquez, Romero-Pérez, Ruíz, Saldarriaga, Sanchez-Azofeifa, Schwartz, Steininger, Swenson, Toledo and Uriarte2016) and old-growth (Poorter et al. Reference Poorter, Van Der Sande, Thompson, Arets, Alarcón, Álvarez-Sánchez, Ascarrunz, Balvanera, Barajas-Guzmán, Boit, Bongers, Carvalho, Casanoves, Cornejo-Tenorio, Costa, De Castilho, Duivenvoorden, Dutrieux, Enquist, Fernández-Méndez, Finegan, Gormley, Healey, Hoosbeek, Ibarra-Manríquez, Junqueira, Levis, Licona, Lisboa, Magnusson, Martínez-Ramos, Martínez-Yrizar, Martorano, Maskell, Mazzei, Meave, Mora, Muñoz, Nytch, Pansonato, Parr, Paz, Pérez-García, Rentería, Rodríguez-Velazquez, Rozendaal, Ruschel, Sakschewski, Salgado-Negret, Schietti, Simões, Sinclair, Souza, Souza, Stropp, Ter Steege, Swenson, Thonicke, Toledo, Uriarte, Van Der Hout, Walker, Zamora and Peña-Claros2015) tropical forests across the Neotropics, where water availability is the main driver of biomass accumulation rates and stocks.
Soil temperature decreased with aboveground biomass (Fig. 2), likely because taller and denser canopies with larger crown areas intercept more incoming radiation, thereby reducing air and soil temperature (Lebrija-Trejos et al. Reference Lebrija-Trejos, Pérez-García, Meave, Poorter and Bongers2011, Matsuo et al. Reference Matsuo, Martínez-Ramos, Bongers, Van Der Sande and Poorter2021, Reference Matsuo, Hiura and Onoda2022). Although wet forests have larger aboveground biomass, climatic wetness had an independent positive effect on soil temperature (Fig. 2), indicating that soil temperature is higher in wet forests for a given aboveground biomass, probably because of higher mean annual temperatures in wet forests (Amissah et al. Reference Amissah, Mohren, Kyereh, Agyeman and Poorter2018). Additionally, soil temperatures may be higher in wetter forests because of higher soil thermal conductivity and soil heat capacity that increase with higher soil moisture (Noborio et al. Reference Noborio, Mcinnes and Heilman1996, Abu-Hamdeh & Reeder Reference Abu-Hamdeh and Reeder2000).
Both leaf litter N and P concentrations were higher in dry forests than in wet forests (Fig. 1D, E). This difference is likely because of higher soil fertility in the studied dry forests (Matsuo et al. Reference Matsuo, Van Der Sande, Amissah, Dabo, Abdul and Poorter2024) and a higher proportion of deciduous species (57.8% in dry forests and 8.5% in wet forests based on the community weighted mean weighted by basal area) and nitrogen-fixing species (19.4% in dry forests and 0.43% in wet forests) (T. Matsuo, unpublished). Soil fertility is higher in dry forests because of less rainfall and leaching, which may increase plant nutrient uptake and leaf nutrient concentrations. Deciduous species often have higher leaf nutrient concentrations to maximise photosynthetic capacity for enhanced growth rates during the rainy season, to compensate for the shorter growing season (Pringle et al. Reference Pringle, Adams, Broadbent, Busby, Donatti, Kurten, Renton and Dirzo2011, Vargas G. et al. Reference Vargas G., Brodribb, Dupuy, González-M, Hulshof, Medvigy, Allerton, Pizano, Salgado-Negret, Schwartz, Van Bloem, Waring and Powers2021).
Decomposition rates increased with microclimate and litter quality and quantity
We found that the litter production rate increased with aboveground biomass (Fig. 2), likely because higher biomass is associated with a larger leaf area and, hence, higher annual leaf litter production (Zhang et al. Reference Zhang, Cheng, Dang, Ye, Zhang and Zhang2013, Capellesso et al. Reference Capellesso, Scrovonski, Zanin, Hepp, Bayer and Sausen2016). As expected, litter decomposition rates increased with soil temperature (microclimate), leaf litter nitrogen concentration (litter quality), and litter production rate (litter quantity) (Fig. 2). Microclimatic conditions facilitated litter decomposition by increasing microbial biomass and activities (Kirschbaum Reference Kirschbaum1995). Litter quality enhances decomposition rates by increasing leaf decomposability (Gartner & Cardon Reference Gartner and Cardon2004, Bumb et al. Reference Bumb, Garnier, Coq, Nahmani, Del Rey Granado, Gimenez and Kazakou2018) and by supporting microbial activity by providing essential nutrients to microbes (Berg & Laskowski Reference Berg and Laskowski2005). Litter quantity also positively influenced the decomposition rate, probably because a large carbon input from litter sustains a larger microbial biomass that enhances decomposition.
Soil respiration was expected to increase with aboveground biomass due to increased belowground root respiration and with higher (heterotrophic) microbial respiration facilitated by higher soil temperature, greater litter nutrient concentrations, and faster litter decomposition rates. Surprisingly, none of these variables significantly affected soil respiration rate (Fig. 2), possibly because soil respiration was heterogeneous within sampling plots (with a large 69 % coefficient of variation within plots) and varies over time (Bond-Lamberty et al. Reference Bond-Lamberty, Pennington, Jian, Megonigal, Sengupta and Ward2019). A chronosequence study in Panama showed that soil respiration was two times higher in pastures than in secondary forests but did not change further during forest succession. The higher soil respiration in the pasture is due to higher grass root respiration and grass residual decomposition (Salimon et al. Reference Salimon, Davidson, Victoria and Melo2004). Future studies should use vegetation variables that are more directly related to root respiration, such as root biomass and (specific) root length, and microbial variables such as microbial biomass and activity to identify the potential drivers of soil respiration.
Litter decomposition is the main driver of the soil organic carbon stock
Our best-fitted model (Fig. 2) showed that SOC stock was only positively affected by litter decomposition rate, indicating that litter decomposition is the primary mechanism that builds SOC (Cotrufo et al. Reference Cotrufo, Soong, Horton, Campbell, Haddix, Wall and Parton2015, Berg & McClaugherty Reference Berg and Mcclaugherty2020) in the early successional forests. Climatic wetness did not have any direct effects, although wet forests in our study have higher clay content (Matsuo et al. Reference Matsuo, Poorter, Van Der Sande, Mohammed Abdul, Koyiba, Opoku, De Wit, Kuzee and Amissah2025), which typically contains higher SOC because it binds organic carbon and forms soil aggregates that protect organic carbon from microbial decomposers (Blume et al. Reference Blume, Brümmer, Fleige, Horn, Kandeler, Kögel-Knabner, Kretzschmar, Stahr and Wilke2016). Instead, climatic wetness had indirect effects on SOC stock by shaping forest attributes (Fig. 2). Although litter input is the primary source of SOC, no significant direct effect was observed, perhaps because the rapid decomposition rate in our study system, 38% month⁻¹ in dry forests and 31% month⁻¹ in wet forests, led to less accumulation of organic carbon, or because there is a lag time in the build-up of SOC. Instead, litter production had only an indirect effect on SOC stock by increasing the decomposition rate (Fig. 2). Soil respiration had, surprisingly, no significant effect on soil carbon stock, perhaps because the soil respiration rate was relatively low compared to the size of the SOC stock, which may have made it difficult to capture its effect (Ryan & Law Reference Ryan and Law2005).
Conclusions
We evaluated the roles of environmental conditions and forest attributes in determining three ecosystem processes that affect SOC in the first years of secondary tropical forest succession. We found that SOC stock increased with litter decomposition but was not affected by carbon input (litter production) or carbon loss (soil respiration). Wetter forests had higher soil temperature and litter production, which enhanced decomposition rates, but this was more than offset by lower leaf litter nitrogen concentration, which slows decomposition rates. Our findings, therefore, highlight the direct role of litter decomposition and the indirect role of climate in shaping belowground carbon dynamics in early forest succession.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0266467425100321
Acknowledgements
We thank the owners of the secondary forest sites, the local communities, and all the people who have established and measured the plots. We also thank Arie Bogert, Dieudonne Koyiba, Justice Opoku, Kwame Kumi, Prince Obeng, Oppong Emmanuel, Kwasi Matthew, Alex Baah, Abraham Ayiku, Samuel Ampofo, Christian Owusu, Emmanuel Kugblenu, Frances Opoku, Jonathan Dabo, Justice Kwateng, Justice Mensah, Kofi Kwateng, Seth Naenewotor for their assistance throughout the fieldwork. We would also like to extend our gratitude to Bo Zhou, Hennie Halm, Miho Tsujii, Bas de Wit, and Thomas Meerwijk for leaf and litter nutrient analysis in the lab.
Financial Support
This study was carried out within the framework of the PANTROP project supported by the ERC Advanced Grant (834775) awarded to LP. LK was financially supported by the National Overseas Scholarship (2020-2021) from the Ministry of Tribal Affairs, Government of India and an Internship grant (2024) from Stichting Het Kronendak.
Competing interests
The authors declare no conflict of interest.
Open access
