Tropical peatlands provide multiple ecosystem services, including hydrological and climate regulation, biodiversity conservation, and livelihood support from timber and non-timber forest products (Giesen et al., 2018; Kurnianto et al., 2015; Thornton et al., 2018). Tropical peatlands are characterized by persistent water saturation, anaerobic conditions, low pH, and nutrient-poor environments. These conditions limit organic matter decomposition and promote long-term accumulation of carbon-rich soil (Cobb & Harvey, 2019; Page & Baird, 2016). Tropical peatlands can store up to 20 times more carbon than typical tropical forests (Murdiyarso et al., 2019). Given their substantial ecological importance, maintaining the sustainability of these ecosystems is critical for preserving their carbon storage capacities.

However, land-use changes can compromise these functions, resulting in drought and disruption of the carbon storage function (Baird et al., 2017; Evers et al., 2017). Acknowledging the critical role of tropical peatlands in carbon storage, several studies have assessed their ability to store carbon and estimate it to be over 75 Gt C (Cobb et al., 2020; Dargie et al., 2017; Warren et al., 2017). Indonesia possesses one of the largest tropical peatlands globally, estimated at 13.43 million ha (Anda et al., 2021) and holds the majority of carbon storage reserves, with a final estimate of 28.1 Gt C (Warren et al., 2017). These large reserves highlight the strategic significance of Indonesian peatlands for regional and global climate change mitigation.

Building on this context, a portion of Indonesia’s peatlands, approximately 190,000 hectares (Miettinen et al., 2016), are managed by small-scale farmers who use traditional practices. Most of these practices rely heavily on fire as the primary method for land clearing (Medrilzam et al., 2017). Without strict local control, such burning can spread easily, causing extensive forest and peatland fires that exacerbate ecosystem degradation and release significant amounts of carbon into the atmosphere. Over time, this situation has worsened due to the construction of drainage canals, which lower the water table and disrupt the natural hydrological regime of peatlands (Hooijer et al., 2012). Drainage leads to irreversible peat subsidence through a combination of compaction, consolidation, and oxidation and the aerobic decomposition of organic matter that emits CO₂ into the atmosphere (Evans et al., 2019). Although small-scale plots are individually small, their collective area is extensive and significantly contributes to peatland carbon emissions in Indonesia.

Rewetting is a core component of Indonesia’s peatland restoration strategy, aimed at re-establishing saturated conditions to suppress aerobic decomposition and reduce carbon loss (Dohong et al., 2018; Graham et al., 2017). This intervention typically involves the construction of canal blocks and other hydrological infrastructure to maintain GWT in 0.4 m depth (Indonesia’s threshold for maximum GWT in cultivated area) and stabilize degraded peat systems. Restoration efforts gained urgency following the catastrophic 2015 fire event, which had widespread socio-economic and regional impacts. In response, the Indonesian government committed to restoring over two million hectares of degraded peatlands (Wardhana, 2016). To support this, the Indonesia Peatland Restoration Agency (BRG) was established in 2016, promoting the “3R” approach: rewetting, revegetation, and revitalization of local livelihoods (Dohong, 2017).

Although rewetting has been widely reported to reduce CO₂ emissions from managed peatlands and improve the carbon balance of previously drained tropical peat forests (Darusman et al., 2022; Lestari et al., 2022; Novita et al., 2024), site-specific evidence remains necessary to refine the EFs. However, long-term rewetting may also alter peat conditions (Negassa et al., 2019), because deeper GWT expand anoxic zones and reduces oxic peat layers (Girkin et al., 2025), thereby suppressing aerobic decomposition and potentially enhancing methanogenesis (Zhong et al., 2020). Recent evidence indicates that rewetting and GWT management can reduce emissions, although their effects on soil CO₂ flux remain specific and depend on interacting hydrological and environmental conditions (Hooijer et al., 2024; Mahardika et al., 2026; Sinclair et al., 2024).

Expanding measurements across diverse land use and peat conditions will support Indonesia’s transition to higher-tier emission estimates. Without accurate CO₂ flux data, particularly from small-scale systems, national peatland carbon accounting will continue to rely on generalized factors, such as those presented in the IPCC Wetlands Supplement 2013 (Hiraishi et al., 2014). The urgency for obtaining such data lies in the fact that smallholder peatlands constitute a substantial yet underrepresented component of Indonesia’s peatland carbon assessment. This study focuses on soil CO₂ flux in small-scale systems in West Kalimantan, where several parts of their peatlands are extensively managed by local farmers under low-input conditions, providing a realistic context for improving the accuracy of emission estimates in subnational carbon assessments.

Moreover, only a few studies have investigated CO₂ emissions from smallholder peatlands in Indonesia, including those by Agus et al. (2010), Jamaludin et al. (2020), Khasanah and van Noordwijk (2019), and N₂O and CH4 emissions by Jovani‐Sancho et al. (2023), indicating the need for broader and more detailed measurements. Therefore, this study aimed to quantify soil CO₂ fluxes from small-scale agricultural peatlands to complement the existing datasets and improve EF accuracy. The analysis differentiated between peat and root-derived sources to clarify their relative contributions, providing supplementary information for refining national emission estimates and enhancing the understanding of peatland carbon dynamics.

This pattern is consistent with the understanding that peat structural changes result from oxidation, compaction, shrinkage, and long-term subsidence (Hooijer et al., 2012; Sinclair et al., 2020). The soil chemical properties (Table 1) further reflected minimal external inputs and limited soil management across all sites, with consistently low pH based on the Indonesian Soil Research Institute fertility thresholds (Eviati et al., 2023) and higher total nitrogen and C-organic ratios. These results confirmed that soil fertility remained poor, likely due to the absence or limited use of fertilizer and soil amendments.

The PCA results (Fig. 4) indicated that the first two principal components explained 69.2% of the total variance, with PC1 accounting for 42.6% and PC2 accounting for 26.6%. PC1 represents the dominant gradient separating land use according to the degree of disturbance and management intensity, where bulk density shows the strongest positive loading, while CO₂ flux (F_CO₂), soil organic carbon, total nitrogen, and pH exhibited strong negative loadings. This pattern suggests that increased peat compaction and structural degradation in managed land use (HOR, RBB, and OPP) are closely associated with shifts in soil chemical properties and elevated CO₂ fluxes. In contrast, PC2 reflected a gradient of intrinsic peat physical characteristics, primarily driven by porosity and peat depth, both of which showed high positive loadings. This gradient clearly distinguishes DSF from cultivated land use, with DSF characterized by deeper, more porous, and relatively less degraded peat. The proximity of the F_CO₂ vector to pH and organic carbon further indicates that variations in CO₂ flux are mainly governed by changes in organic matter quality and soil chemical conditions induced by land-use conversion rather than by peat depth alone.

These results directly indicate that, even under smallholder management, land-use intensity substantially shapes the magnitude of peat- and root-derived CO₂ emissions (see Table 2). In OPP, the particularly high fluxes are attributed to the dominance of root (autotrophic) respiration and the presence of fibric peat, which together enhance CO₂ release through both biological and physical pathways. The proportion of autotrophic respiration in the OPP was higher than that in Sarawak (34%) (Ishikura et al., 2018) and Riau (47.6% ) (Jamili et al., 2021), further indicating that the combined effects of peat decomposition level, land management practices, and plantation age (Dariah et al., 2014; Matysek et al., 2018) jointly regulate site-specific emission magnitudes. Although the total fluxes in the OPP were lower than those recorded in Jambi (Khasanah & van Noordwijk, 2019), they were still higher than those in similar systems in Kubu Raya (Jamaludin et al., 2020). The higher heterotrophic respiration in the HOR, exceeding the 62% proportion reported for small-scale peatlands in Aceh (Agus et al., 2010), confirms that horticultural peatlands, although limited in size, can contribute disproportionately to the total emissions, particularly under persistent drainage and soil exposure.

Furthermore, the total CO₂ fluxes were higher than those reported for ginger cultivation in Kubu Raya, West Kalimantan (Jamaludin et al., 2020). The RBB site exhibited moderately lower emissions than small-scale rubber agroforestry systems in Jambi (Khasanah & van Noordwijk, 2019) but was closely aligned with those observed in West Kalimantan (Jamaludin et al., 2020).

In contrast, the DSF site showed the lowest soil CO₂ flux among all land uses, comparable to rewetted peat forests in Central Kalimantan (Darusman et al., 2022) but much lower than that of secondary forests in Ketapang, West Kalimantan (Basuki et al., 2021), and remnant forests (Khasanah & van Noordwijk, 2019). These results highlight that even after previous drainage, certain secondary forest sites can retain relatively low emission rates owing to moisture retention, denser vegetation cover, and minimal disturbance history, supporting the effectiveness of rewetting in reducing CO₂ emissions.

Despite the previous drainage, the relatively low fluxes in the DSF suggest that rewetting and vegetation recovery successfully suppressed aerobic decomposition through improved soil moisture retention and canopy cover. The consistently lower emissions in DSF compared with other land uses in this study supports the view that rewetting is an effective mitigation strategy for reducing peat-derived CO₂ while promoting ecological recovery. Similar findings from Mempawah, West Kalimantan (Novita et al., 2024), reported that rewetting reduced heterotrophic and total respiration in palm oil plantations by 34% and 20%, respectively, whereas another study reported reductions of up to 39% in rubber and 36% in oil palm systems (Lestari et al., 2022). However, peat-derived CO₂ emissions remained above 80% post-rewetting, indicating that organic matter mineralization continued even after hydrological restoration. These findings collectively emphasize that rewetting lowers, but does not eliminate, carbon emissions, because its effectiveness depends strongly on local peat properties and moisture dynamics (Grand-Clement et al., 2015; Mahardika, Tarigan, et al., 2024).

The Random Forest regression results demonstrate that hydrological variables dominate the control of soil CO₂ flux across land use types, with GWT consistently identified as the most influential feature at HOR, OPP, and RBB, confirming that moisture-related factors are the primary drivers of the fluxes. An exception was observed at the DSF site, where RHs exerted feature dominance (Fig. 5 & 6). Previous studies have reported a strong relationship between GWT and RHs (Taufik et al., 2019; Widiarso et al., 2020), which emerged as the dominant factor influencing soil CO₂ flux in this study. Partial Dependence Plots revealed clear threshold responses, where the CO₂ flux increased sharply once the GWT declined beyond ~25 cm at the HOR and ~30 cm at the OPP, indicating the enhanced aerobic decomposition of peat organic matter under deeper drainage conditions (Carlson et al., 2015; Husen et al., 2014; Mahardika, Anshari, et al., 2024). At the DSF site, RHs emerged as the dominant control on CO₂ flux, consistent with evidence that peat respiration peaks at intermediate moisture levels and declines under both excessively dry and near-saturated conditions (Comeau et al., 2021). In contrast, RBB exhibited highly variable and less predictable emission patterns, as reflected by low model performance (OOB R^² = 0.216), suggesting greater heterogeneity in management and micro-hydrological conditions. Soil moisture and temperature showed higher relative importance than atmospheric variables (Ta and RHa), which consistently played a minor role, differing from studies that emphasized temperature as a primary control (Marwanto & Agus, 2014; Sjögersten et al., 2018).

When evaluated against reference emission factors, CO₂ fluxes in HOR, RBB, and DSF were lower than or close to both IPCC Tier-1 and Indonesia’s 2nd FREL defaults, whereas OPP exceeded both (Table 4). These comparisons demonstrate that generalized default EFs often fail to represent accurately small-scale systems because they overlook local differences in management intensity, cultivation cycles, and hydrological regulation. In smallholder peatlands, minimal fertilizer use, short cultivation periods, and irregular water control create emission patterns distinct from large-scale, intensively drained plantations that underpin most Tier-1 and Tier-2 estimates. The variability observed across sites reinforces the need for EFs that account for site-specific environmental and hydrological conditions. As shown by Murdiyarso et al. (2024), CO₂ emissions from undrained and rewetted peatlands differ substantially depending on GWT, peat degradation, and site condition—where CH₄ emissions decline rapidly after rewetting, but heterotrophic CO₂ fluxes may take decades to stabilize. Similarly, He and Roulet (2023) refined the IPCC EF of northern peatlands by incorporating seasonal variability and hydrological stratification, reducing uncertainty nearly threefold. These insights align with the findings of this study, which emphasize that the GWT and soil moisture regime act as the dominant environmental control of CO₂ fluxes.

Nevertheless, the observation period may not capture seasonal or interannual variations in CO₂ flux, whereas unmeasured differences in microtopography and site-level management could have introduced spatial variability that was not fully reflected in the regression analysis. Future research should prioritize long-term and continuous measurements across a wider range of smallholder peatland uses to improve emission factor precision and to better evaluate the long-term impacts of rewetting and low-input agricultural practices on peat carbon dynamics.

Among the observed sites, oil palm plantations showed the highest total soil CO₂ fluxes, while secondary forests tended to have the lowest emissions. These results suggest that land use influences the magnitude and sources of soil CO₂ fluxes in small-scale peatland systems, with environmental regulation predominantly controlled by several factors. Furthermore, this study provides additional information that may support future refinement of Indonesia’s emission factors toward higher-tier approaches. Further research covering a wider range of peat conditions, management practices, and seasonal variations is needed to enhance the understanding of CO₂ flux dynamics in small-scale tropical agricultural peatlands.