This study was conducted on peatlands within the Sriwijaya Botanical Gardens located in the Musi-Belida River Peat Hydrological Unit (KHG), Ogan Ilir District, South Sumatra Province (Figure 1). The geographically coordinates are 104⁰31′23.26″–104°33′9.16″E and 3°8′58.46″–3°9′48.64″S. The Sriwijaya Botanical Gardens constitute a Special Purpose Forest Area (KHDTK) established under the Decree of the Minister of Forestry No. 485/Menhut-II/2012. This area is designated for conservation, research, development, environmental education, ecotourism, and environmental services (Defriyanti et al., 2018; Herawati and Maryani, 2018). The botanical gardens cover approximately 100 ha of peatland. The topography is predominantly flat with a slope of less than 1% and an elevation ranging from 17 to 23 m above sea level (Herawati and Maryani, 2018).

The Sriwijaya Botanical Gardens were established in peatlands formerly designated as production forests. The initiative began in 2010, with planting preparations commencing in 2014. Prior to development, the area functioned as a four-year-old oil palm plantation and had experienced considerable drainage (Herawati and Maryani, 2018). Currently, the botanical gardens are revegetated with various tree species, including Fagraea fragrans (tembesu), Shorea belangeran (belangeran), Alstonia pneumatophora (pulai), and Dyera lowii (jelutung), all of which exhibit different stand ages. The age differences were primarily attributable to recurrent fires in the area, particularly in 2014 and 2019, which led to replantation efforts in affected plots. The site is characterized by seasonal flooding during the rainy season, with groundwater levels ascending to 40 cm, and highly susceptible to fire during the dry season (Defriyanti et al., 2018).

The most effective revegetation species that thrived and established forest stands following the fires in 2014 and 2019 at the Sriwijaya Botanical Garden were Tembesu and Belangeran; in contrast, areas where revegetation efforts failed tended to develop into secondary forests naturally dominated by Gelam. In further developments, the vegetation in the Sriwijaya Botanical Gardens developed into Belangeran stands planted in 2015 and 2020, Tembesu stands planted in 2015, and secondary forest of Gelam. Each stand group exhibited varying densities and growth in diameter, height, and volume (Table 1). The most substantial tree dimensions, including diameter at breast height, height, and volume per hectare, were observed in Tembesu 2015 stands, whereas the least were recorded in Belangeran 2020 stands. Soil characteristics, including soil moisture, bulk density, organic material, and C-organic content, varied among the stands (Table 2).

The Sriwijaya Botanical Gardens were established on degraded peatlands with peat depths ranging from 207–454 cm. Upon the commencement of construction, the area experienced extensive drainage, followed by the cultivation of palm oil. The canal network was structured by constructing a conservation reservoir centrally and installing water gates at the southern edge of the area (Figure 2). Water gates in peatland conservation areas regulate water levels and maintain the ecological balance of peat ecosystems. The water gates are kept closed under normal conditions and opened only during flooding events, however, they are not in optimal condition. Consequently, the gates are unable to effectively retain the water level, particularly during prolonged dry season.

Emission measurements were conducted on four types of peatland cover within the Sriwijaya Botanical Gardens: Shorea belangeran (belangeran) stands planted in 2015 after the fire in 2014 and planted in 2020 after the fire in 2019; Fagraea fragrans (tembesu) stands planted in 2015 after the fire in 2014; and secondary forest dominated by Melaleuca leucadendra (gelam), which also emerged after the fire in 2014. Measurements were performed using an infrared gas analyzer (IRGA; LI-830, LI-COR, NE, USA). The measurement chamber was constructed using a 25.4 cm diameter polyvinyl chloride (PVC) pipe truncated to a length of 25 cm. Each chamber was embedded in the peat soil to a depth of 5–10 cm, with 15–20 cm remaining above the ground surface for gas sampling and emission assessments.

Emission measurements were conducted for each peatland cover type. Measurements were conducted in triplicate for each cover type. Emission measurements were conducted for each peatland cover type using three spatial replicates (n = 3). Each replicate consisted of a permanently installed chamber collar representing an independent sampling plot within each cover type. Chamber collars were installed at fixed locations and remained in place throughout the monitoring period. CO₂ fluxes measurements were repeatedly conducted at the same fixed collars on a weekly basis, such that temporal variation was captured through repeated measurements at each replicate plot. Spatial replication was defined by the three independent chamber collars per cover type, whereas temporal replication was represented by repeated weekly measurements at each collar.

The initial measurement points for each peatland cover type were located 20 m from the nearest drainage canal, with successive measurement points spaced 25 m apart (Figure 2). Measurements within each plot were performed in a clockwise direction, alternating the starting point to minimize temporal bias from 07.00 to 12.00. Measurements were performed weekly over a span of 22 weeks from September 18, 2024, to February 12, 2025. Annual CO₂ fluxes were estimated by extrapolating weekly measurements under the assumption that the measured fluxes are representative of prevailing non inundated conditions during the observation period. Weekly mean fluxes were integrated over time to derive annual estimates. Periods when collars were inundated were excluded from the extrapolation, and the resulting annual values therefore represent estimates of soil–atmosphere CO₂ fluxes under non inundated conditions. A key limitation of this approach is that the measurement period may not fully represent the full range of seasonal conditions occurring over an entire year and thus introduces uncertainty in the extrapolated annual estimates.

CO₂ emissions were calculated using an equation based on the ideal gas law (Husnain et al., 2014; Marwanto and Agus, 2014) as shown in Equation 1:

Where ʄ_C is the CO₂ efflux (μmol m⁻² s⁻¹), which was convert to tons ha⁻¹ year⁻¹ by using the molar mass of CO₂ (44.01 g mol⁻¹, area conversion (m² to ha), time conversion (s to year), and unit conversion factors (g to ton); P is the atmospheric pressure (Pa); h is the chamber height (cm); R is the gas constant (8,314 J mol⁻¹ °K⁻¹); T (°K), and dC/dt is the change in CO₂ concentration.

Simultaneously with the emission measurements, groundwater level depth was assessed using a piezometer, whereas air, soil, and in-chamber temperatures were recorded using a solid stem mercury water thermometer, Indonesia. The piezometer was constructed using a 7.62 cm diameter PVC pipe with a total length of 2.75 m. It was installed vertically in the peat soil, with approximately 50 cm protruding above the ground surface. The submerged portion of the pipe was perforated with holes spaced 10 cm apart on all sides to facilitate groundwater ingress and precisely reflect the water-table level.

Data on CO₂ emissions, along with associated environmental variables, groundwater depth, soil temperature, and air temperature, were analyzed and tabulated according to peatland cover type. All statistical analysis was conducted in RStudio version 2025.05.1 + 513 Posit Software. Variations in CO₂ emissions and environmental parameters across various peatland cover types were analized using linear-mixed effects models-based Anova (LMMs) with peatland cover type treated as a fixed effect and collar identity nested within stand included as a random effect to account for repeated measurements. This model structure appropriately reflects the nesting of observations within collars and avoids pseudo-replication arising from temporal autocorrelation.

Significance difference of variables between peatland cover were tested by Least Significance Difference at p < 0.05. To examine the relationships between CO₂ emissions and environmental variables, linear mixed-effects regression models were applied, allowing both fixed effects of environmental predictors and random effects associated with repeated observations to be accounted for.

The average annual CO₂ emissions in the Sriwijaya Botanical Gardens reached 31.9 ± 19.9 t CO₂ ha⁻¹ year⁻¹, representing soil-atmosphere CO₂ fluxes. This value is obtained from the average emissions of four peatland cover types found in the area during the six-month periods of 18 September, 2024, to 12 February, 2025, ranging from 25.2 ± 20.4 t CO₂ ha⁻¹ year⁻¹ in the Belangeran stand 2015 to 36.6 ± 16.3 t CO₂ ha⁻¹ year⁻¹ in the Belangeran stand 2020. No significant differences were observed in the magnitude of emissions among peatland cover types, and mean values are presented with its standard deviation (Figure 3, left).

The temporal dynamics of CO₂ emissions across the peatland cover types in the Sriwijaya Botanical Gardens exhibited high initial fluctuations, followed by a consistent decrease from the onset to the end of the weekly measurement period (Figure 3, right). Emission measurements commenced during the dry season in September 2024 and continued throughout the rainy season in February 2025.

The relationships between CO₂ emissions and environmental factors, including groundwater level, soil, and air temperature, were analyzed using linear-mixed effects regression models with collar identity treated as a random effect. Reported effects therefore reflect differences among cover types while accounting for repeated measurements at the same collars over time. Separate mixed-effects regression models were fitted to examine the effect of each environmental factor idividually (Figures 7–9), whereas a multiple mixed-effects regression models was performed to evaluate the combined effect of all three environmental factors within a single model (Table 3). In all models, collar identity (nested within stand) was included as a random effect to account for repeated measurements.

The linear mixed-effects model showed that CO₂ emissions from peatland cover types could be moderately to strongly explained by a combination of environmental factors, including groundwater level, soil temperature, and air temperature, with conditional coefficients of determination (R²) ranging from 0.35 to 0.53 (Table 3). Groundwater levels consistently exhibited a strong negative effect on emissions, especially in the Belangeran 2015 and 2020 stands. In contrast, soil temperature had a more significant positive effect on the Belangeran 2020 and secondary forest of Gelam stands, furthermore, air temperature more influential in Belangeran 2015 and 2020 stands.

Linear mixed-effects model analysis showed a negative relationship between groundwater level and CO₂ emissions in all peatland cover types (Figure 7). The conditional coefficient of determination value was relatively low (ranging from 0.28 in the secondary forest of Gelam to 0.52 in the Tembesu 2015 stand), thereby indicating that the groundwater level explained only approximately 28–52% of the variability in CO₂ emissions. However, the relationship remained highly statistically significant (p-value < 0.01). The coefficient of correlation (ranging from 0.53 to 0.72) further suggested that the groundwater level consistently affected CO₂ emissions across peatland cover types.

In contrast to groundwater level, soil temperatures exhibited a significant positive linear relationship with CO₂ emissions across all peatland cover types, with p-value < 0.01 and a coefficient of correlation ranging from 0.52 to 0.69 (Figure 8). The conditional coefficient of determination was relatively low, ranging from 0.27 in the Belangeran 2015 stand to 0.48 in the Tembesu 2015 stand. However, the highly significant difference indicated that soil temperature consistently influenced the magnitude of CO₂ emissions. Similar to soil temperature, air temperature also demonstrated a strong correlation with CO₂ emissions across all peatland cover types, with a p-value < 0.01 and coefficient of correlation ranging from 0.46 to 0.66 (Figure 9). However, the coefficients of determination for all these relationships were relatively low.

The estimated annual CO₂ emissions from the four representative peatland covers in the Sriwijaya Botanical Gardens were 31.9 ± 19.9 t CO₂ ha⁻¹ year⁻¹ (Figure 3). The emissions are relatively low compared with the national average peat emissions of 48.22 t CO₂ ha⁻¹ year⁻¹, as reported by multiple studies on various peatland applications in Sumatera and Kalimantan, encompassing forests, shrublands, burned areas, agriculture, and plantations (Novita et al., 2021). For comparison, CO₂ emissions from a 15-year-old oil palm plantation were 46 ± 30 t CO₂ ha⁻¹ year⁻¹ (Marwanto and Agus, 2014), whereas those from an 8-year-old rubber plantation were 32.93 ± 10.39 t CO₂ ha⁻¹ year⁻¹ (Wakhid et al., 2017). The observed emissions in this study are also lower than the IPPC Tier 1 default emission factors for peatlands plantation, which are 73.3 t CO₂ ha⁻¹ year⁻¹ for short rotation systems and 55 t CO₂ ha⁻¹ year⁻¹ for long rotation systems (IPCC, 2013). Although the emissions in this revegetated peatland under hydrological recovery remain higher than those reported for undrained peatlands in Muara Siran, East Kalimantan, amounting to 11.02 ± 0.49 t CO₂ ha⁻¹ year⁻¹ (Asyhari et al., 2024), it nonetheless indicates that peat management and revegetation efforts in the Botanical Gardens have likely contributed to a reduction in emissions relative to more degraded or intensively drained peatland systems. Given that deep groundwater levels were still observed during the monitoring period, the study area is more appropriately characterized as a recovering or rehabilitated peatland, rather than a fully restored system, and the observed emission levels should be interpreted within this transitional recovery context.

The extrapolated annual CO₂ fluxes should be interpreted as first-order estimates, given that the measurement period may not fully represent the complete range of seasonal hydrological conditions occurring over an entire year. In tropical peatlands, seasonal inundation can alter dominant CO₂ transport pathways, shifting from soil–atmosphere exchange under non inundated conditions to surface-mediated fluxes during inundation. As periods of inundation were excluded from the present analysis, the reported annual estimates primarily reflect non-inundated conditions and may either underestimate or overestimate true annual emissions.

The Sriwijaya Botanical Gardens have been equipped with water management infrastructure, including a central conservation reservoir and a series of water gates along the southern perimeter of the area. The installation of water gates is intended to sustain groundwater at a depth of approximately 40 cm below the surface and preserve the natural ecological balance of the peat ecosystem. While the performance of this system cannot be considered fully optimal, it has demonstrated a measureable capacity to retain the groundwater level. Average groundwater levels in this study area ranged from −41.0 ± 35.3 to −60.0 ± 24.4 cm indicating a moderate groundwater level compared with more degraded peatland. These levels are higher than those reported by (Wakhid et al., 2017), who recorded an average of −69 cm with a range from −3 to −171 cm, suggesting that water management interventions in the Sriwijaya Botanical Garden contributed to reducing extreme groundwater table drawdown. The primary objectives of water management systems in peatland areas are to maintain the groundwater table, regulate water flow, prevent peat degradation, support ecosystem rehabilitation, and mitigate fire risk (Wakhid et al., 2017; Dohong et al., 2018).

Based on the differences in peatland cover types, CO₂ emissions in the Sriwijaya Botanical Gardens showed variations in annual average values, however, these differences were not statistically significant (Figure 3). The highest emissions were recorded in the Belangeran 2020 stand, closely followed by the Tembesu 2015 stand, with values of 36.6 and 36.0 t CO₂ ha⁻¹ year⁻¹, respectively. The lowest emissions were recorded in the Belangeran 2015 stand at 25.2 ± 20.4 t CO₂ ha⁻¹ year⁻¹. However, the CO₂ emissions from all peatland cover types in this area were substantially lower than those recorded in degraded peatlands, which amounted to 45.1 t CO₂ ha⁻¹ yr.⁻¹ because of inadequate water management resulting in a reduced groundwater level of −65 ± 17 cm (Deshmukh et al., 2023). This indicates that the water management system at the Sriwijaya Botanical Gardens may contribute to reducing emissions compared with more disturbed, unmanaged peatland conditions.

In contrast to other peatland cover types in the Sriwijaya Botanical Gardens, the elevated emissions in the Belangeran 2020 stand are associated with increased annual averages of soil and air temperatures at 30.2 and 31.1 °C, respectively. This stand was replanted in 2020 following its impact from peat fires in 2019 (Maryani and Novriadhy, 2023), which likely contributed to its elevated values. The occurrence of fires resulted in the loss of peatland canopy cover, rendering the area relatively exposed (Siahaan et al., 2020), whereas the emergence of young plant canopies failed to restore it, resulting in increased solar radiation penetration to the ground surface and elevated soil and air temperatures. Peat combustion during fires eliminates surface organic layers, exposing deeper and more decomposable peat (Astiani et al., 2018; Siahaan et al., 2025).

Meanwhile, the high annual emissions in the Tembesu 2015 stand, despite possessing a relatively dense canopy cover and reduced soil and air temperatures (Figures 5 and 6), may be attributed to a deeper groundwater level of −52 cm (Figure 4). The proximity of the stand to the main drainage channel leads to a more expedited decline in groundwater level. Reducing the groundwater level in peatlands increases CO₂ emissions by exposing additional peat to the atmosphere, thereby intensifying aerobic microbial activity and accelerating the decomposition of organic matter. Under saturated (anaerobic) conditions, decomposition is slower (Pärn et al., 2025; Ritzema et al., 2014) and emissions are lower. However, when the water table drops, oxygen penetrates deeper, thereby stimulating microbial respiration and carbon release (Miettinen et al., 2017; Marwanto et al., 2019).

In addition to being driven by low groundwater levels, the high CO₂ emissions recorded in the Tembesu 2015 stand plausibly be associated with enhanced root-related respiration, as inferred from its larger diameter at breast height and higher stand volume (Table 1). Previous studies have shown that root respiration tends to increase in denser and mature stands (Pacaldo and Aydin, 2023). For example, root based respiration has been reported to contribute approximately 19 and 29% to total soil respiration (Rs) in 6 and 15 year old palm plantation, respectively (Dariah et al., 2014). Furthermore (Batubara et al., 2019) reported that CO₂ emission measured with deep collars (excluding root based respiration) were 29% lower than those measured using shallow collars in a 25 year oil palm plantation. Although no direct partitioning between autotrophic (root derived) and heterotrophic respiration was performed in this study, the structural characteristics of the Tembesu stand, namely its larger tree diameter and higher stand volume, likely supports a more intensive root system that may be contribute to higher soil CO2 fluxes. In peatlands, root respiration, particularly from rapidly developing or densely rooted vegetation is widely recognized as an important component of total soil respiration, alongside microbial peat decomposition (Rh). Consequently, the interplay between reduced groundwater level and potentially intensified root activity in the Tembesu 2015 stand likely accounts for the elevated emission levels.

Compared with the Belangeran 2020 stand, the Belangeran 2015 stand (an older stand) demonstrated the lowest CO₂ emissions, recorded at 25.2 t CO₂ ha⁻¹ year⁻¹ (Figure 3). This reduced emission rate correlates with a shallower groundwater level of −42.7 cm (Figure 4) and slightly cooler soil and air temperatures, cumulatively leading to reduced peat decomposition (Rh). The prevalence of more mature vegetation in the 2015 stand likely enhances microclimatic conditions, such as increased shade and humidity, which aid in maintaining cooler soil conditions and inhibiting microbial activity, thereby limiting CO₂ emission. Conversely, despite having the deepest groundwater level (−60.0 cm), CO₂ emissions in the secondary forest of Gelam remained moderate, surpassing those in the Belangeran 2015 stand but falling short of those in the Belangeran 2020 and Tembesu 2015 stands. This may be attributed to the relatively small stand diameter and reduced stand volume, leading to lower root respiration (Table 1). A further explanation is the relatively low soil temperature observed in this stand, potentially linked to high aboveground litter accumulation, providing insulation and reducing heat penetration into the soil. Additionally, this area did not experience fire in 2019, which aided in the preservation of its forest structure and surface organic layer, thereby further moderating temperature and emissions.

The average groundwater levels in the Sriwijaya Botanical Garden differed significantly between peatland cover types (Figure 4). The averages ranged from −41,0 ± 35.3 cm in the Belangeran 2020 stand to −60 ± 24.4 cm in the secondary forest of Gelam. The difference in the groundwater table depth is associated with the proximity of each plot from the drainage canal; the nearer the plot is to the drainage canal, the deeper its groundwater level tends to be (Marwanto and Agus, 2014). In the context of the Sriwijaya Botanical Gardens, the Tembesu 2015 stand and secondary forest of Gelam were the nearest plots to the main drainage canal (Figure 2), exhibiting groundwater levels of −52.0 ± 35.6 and −60.0 ± 24.4 cm, respectively, both of which were deeper than those of the other two plots.

Despite these spatial differences, groundwater levels across all plots remained relatively stable during periods when rainfall ceased. This temporal stability indicates that short-term fluctuations in groundwater level were not solely controlled by precipitation inputs. Such patterns are consistent with regulated water retention within the study area, potentially associated with the operation of the water gate system that limits water outflow and helps maintain groundwater levels in the peatland.

The average groundwater levels across all peatlands in the Sriwijaya Botanical Garden fell below the critical threshold of −40 cm (Dohong et al., 2017). The deep groundwater level allows for greater oxygen penetration into the peat, thereby enhancing microbial activity, accelerating peat decomposition, and ultimately increasing CO₂ emissions (Albert-Saiz et al., 2025). However, this deep groundwater level did not persist throughout the entire period; as the seasons changed from the dry season at the onset of the measurement period to the rainy season toward the end, the water level rose significantly (Figure 4). At the commencement of the observation (week 1), the groundwater levels in all sites were very deep (approximately −120 to −160 cm), indicating dry conditions. Over the first several weeks (especially between weeks 2 and 5), the groundwater level increased rapidly throughout all peatland cover types. After week 5 or so, the groundwater level fluctuated slightly while maintaining an overall upward trend (less negative) until it reached a relatively shallow depth (approximately −10 to −20 cm) by week 22, indicating that water approached the surface. Notably, the groundwater level approached zero in the Belangeran 2020 and 2015 stands by week 22, indicating conditions conducive to surface saturation and a potential tendency toward inundation.

Variations in groundwater levels resulted in differences in CO₂ emissions across peatland cover types in the Sriwijaya Botanical Gardens; however, the magnitude of this effect varied among them. The linear mixed effects model showed that groundwater level significantly affected emissions in the Tembesu 2015, Belangeran 2020, and Belangeran 2015 stands, whereas its effects were not significant in the Gelam secondary forest stand (Table 3). This model accounted for the combined influence of environmental variables while incorporating random effects to address repeated measurements. Furthermore, a simple factor linear-effect regression models examining the relationship between groundwater level and emissions in each peatland cover type (Figure 7) demonstrated differences in both the conditional coefficient of determination (R²) and slope of the equations, despite the relationship being statistically significant (p-value < 0.01) across all covers. The highest conditional coefficient of determination was observed in the Tembesu 2015 stand, with an R² value of 0.52 (r = 0.72) and a slope of −0.25, represented by the equation y = −0.25 x + 23.2. In contrast, the lowest was observed in the secondary forest of Gelam, with an R² value of 0.28 (r = 0.53) and a slope of −0.20, represented by the equation y = −0.20 x + 17.68.

The equation delineating the relationship between groundwater levels and emissions for each peatland cover type can be interpreted in practical terms. For example, the equation in the Belangeran 2020 stand, y = −0.25 x + 23.2, indicates that for every 10 cm decrease in water level, CO₂ emissions increase by approximately 2.5 t CO₂ ha⁻¹ year⁻¹. The regression equation slopes across all peatland cover types range from −0.20 in the secondary forest of Gelam to −0.31 in the Belangeran 2015 stand, indicating that with each 10-cm decrease in groundwater level, CO₂ emissions increase by 2.0 to 3.1 t CO₂ ha⁻¹ year⁻¹. These values are smaller than those reported in two previous studies, which identified increases in CO₂ emissions of 6.5 and 7.2 t CO₂ ha⁻¹ year⁻¹ for every 10-cm decrease in the groundwater level (Wakhid et al., 2017; Deshmukh et al., 2023; Prananto et al., 2020). Within the groundwater level range observed in this study, the relationship between CO₂ fluxes and water table depth appears largely linear, consistent with previous findings. However, the scatter pattern suggests that over a broader range of groundwater level this relationship could become non-linear, aligning with bell-shaped responses documented in peatland studies by (Norberg et al., 2018; Mäkiranta et al., 2009).

A clear pattern suggests that deeper groundwater levels are typically correlated with higher CO₂ emissions resulting from increased aerobic peat decomposition (Rh). Therefore, water table management is important for the conservation and restoration of degraded peatlands (Irfan et al., 2025). Upgrading canal blocking systems and intensifying water gate regulations are urgently needed to elevate groundwater levels, mitigate peat degradation, and thereby decrease CO₂ emissions. Hooijer et al. (2024) reported that the construction of 257 peat dams throughout a degraded peatland area of 4,800 ha successfully elevated the groundwater level from −60 cm to −30 cm and reduced subsidence rates by 50%.

Soil temperature is an environmental factor influencing CO₂ emissions (Marwanto et al., 2019); however, its effects are not always significant (Pacaldo and Aydin, 2023). Recent research has shown a significant variation in soil temperature across the various peatland cover types in the Sriwijaya Botanical Gardens, with measurements ranging from 26.7 ± 1.6 °C in the Tembesu 2015 stand to 30.2 ± 3.0 °C in the Belangeran 2020 stand (Figure 5). The magnitude of the soil temperature in a peatland is influenced by the extent of sunlight penetration to the ground surface, increased sunlight penetration results in elevated soil temperatures. The elevated temperature in the Belangeran 2020 stand may be attributed to several factors, including fire occurrence in 2019 that eradicated the undergrowth and lower canopy cover due to the young age of the stand. These conditions facilitate increased sunlight penetration into the peat soil. In contrast, the reduced soil temperature in the Tembesu 2015 stand may be attributed to its dense canopy cover, which restricts sunlight penetration to the peat surface.

Soil temperatures across all peatland cover types in the Sriwijaya Botanical Gardens exhibited significant relationships with CO₂ emissions, with p-values < 0.01 (Figure 8). However, the conditional coefficients of determinant (R²) were relatively low, ranging from 0.27 (r = 0.52) in the Belangeran 2015 stand to 0.48 (r = 0.69) in the Tembesu 2015 stand. The strongest effect of soil temperature on CO₂ emissions was noted in the Tembesu 2015 stand, represented by the equation y = 5.64 x – 114.58 (R² = 0.48; r = 0.69; p-value < 0.01). This equation indicates that for each 1 °C increase in soil temperature, CO₂ emissions rise by approximately 5.64 t ha⁻¹ year⁻¹; conversely, a 1 °C decrease in soil temperature results in a corresponding decline in emissions by the same amount. As plant growth and canopy development progress, sunlight penetration to the soil surface diminishes, resulting in decreased soil temperatures and reduced CO₂ emissions (De Frenne et al., 2021). While these linear relationships provide a useful approximation within the observed soil temperature range, the relatively high scatter and low R² values suggest that the temperature CO₂ emission relationship may not be strictly linear across broader thermal gradients. In peatland ecosystems, temperature effects on respiration often exhibit non-linear or threshold responses due to interactions with soil moisture, oxygen availability, and substrate limitation.

Weekly fluctuations in soil temperatures across four peatland cover types indicated that the Belangeran 2020 stand consistently exhibited the highest temperatures, with peaks exceeding 30 °C, especially between weeks 6–12. This thermal profile was correlated with its elevated CO₂ emissions, especially during the early measurement period. In contrast, the Belangeran 2015 stand maintained cooler and more stable soil temperatures, predominantly within the 26–30 °C range, and exhibited the lowest overall emission rates. The mature vegetation in this stand likely enhances shading, buffers temperature extremes, and inhibits decomposition. In contrast, the Tembesu 2015 stand showed relatively low soil temperatures, similar to those in the Belangeran 2015 stand; however, its emissions remained high and variable. This suggests that other factors, such as root respiration and low groundwater table depth, play major roles in CO₂ emissions from this stand. The secondary forest of Gelam exhibited intermediate soil temperatures and displayed moderate emission levels.

Similar to soil temperatures, the variation in air temperature between peatland cover types in the Sriwijaya Botanical Gardens correlated with the amount of sunlight penetrating the canopy cover and soil surface. Therefore, the fluctuation in air temperatures between the peatland cover types and their dynamics during the measurement period mirrored the pattern observed for soil temperatures. The highest air temperature recorded in the Belangeran 2020 stand may be attributed to substantial sunlight penetration within the stand, resulting from a fire in 2019 that rendered this stand younger than its counterparts. In contrast, the lowest air temperatures recorded in the Tembesu 2015 and Belangeran 2015 stands may be attributed to their denser canopy, which reduces solar radiation and wind speed, thereby impacting the air temperature. Adaptive species, such as Belangeran, Tembesu, and other native species, have been reported to be associated with more favorable microclimate conditions in degraded peatlands, including lower temperatures and higher humidity following their establishment (Dharmawan et al., 2024; Jaya et al., 2024).

The relationship between air temperature and CO₂ emissions across all peatland cover types in the Sriwijaya Botanical Garden was statistically significant (p-value < 0.01); however, it exhibited a low magnitude of effect, with conditional coefficients of determination (R²) ranging from 0.21 to 0.43 and coefficients of correlation varying between 0.46 and 0.66 (Figure 9). Linear mixed-effects models indicates that air temperature significantly affects CO₂ emissions in the Belangeran 2015 and 2020 stands, while its effect is not significant in the other stands. This suggests that, similar to soil temperature, the influence of air temperature on peatland emissions is context-dependent and may be influenced by other interacting environmental factors. In some cases, the air temperature exhibits a strong positive correlation with CO₂ emissions, whereas in other cases, the relationship appears weak or statistically insignificant (Pacaldo and Aydin, 2023; Yahya et al., 2019).

The low R² values and inconsistent significance across sites suggest that the air temperature and CO₂ emissions relationship may not be strictly linear and could involve non-linear or threshold-type responses, particularly when evaluated over broader temperature ranges or under varying hydrological conditions. The observed inconsistency may be attributed to the complex interplay between temperature and various environmental factors, such as microbial activity, peat and water table depth, oxygen availability, and substrate quality. While these findings provide valuable insight, the limited of sample size restricts the extent to which them can be generalized.

This study provides empirical evidence that the hydrological condition is the dominant control of CO₂ emissions from peat soil, while vegetation type primarily modifies emission magnitude through its effects on microclimate and root respiration. These findings have direct relevance to Indonesian’s FOLU Net Sink 2030 framework, which prioritizes peatland rewetting and fire prevention as key mitigation strategies for achieving net carbon sequestration in the land use sector. The consistently strong negative relationship between groundwater level and CO₂ emissions across peatland cover types indicates that maintaining a high level of groundwater table is more critical for emissions reduction than vegetation composition alone. This supports national mitigation strategies under FOLU Net Sink 2030 that emphasize canal blocking, water gate management, and hydrological rehabilitation as the primary interventions for reducing peat carbon losses.

Although vegetation types influence soil and air temperature and contributed to variation in CO₂ emissions, particularly through canopy openness and root respiration, these effects were secondary to hydrological control. This finding has important implications for national greenhouse gas (GHG) inventory development. Current emission factor approaches often stratify peat emissions by land cover or vegetation classes, however, the results of this study suggest that a common emission factor for rewetted or hydrologically managed peatlands may be appropriate, with vegetation type incorporated as secondary or modifying factor rather than a primary determinant. Such an approach would better reflect the underlying biogeochemical drivers of emissions and reduce uncertainty in national reporting, particularly for peatlands under rehabilitation or conservation management.

It is also important to clarify the terminology used to describe peatland condition in this study. While the Sriwijaya Botanical Gardens have undergone extensive revegetation and partial hydrological intervention, groundwater levels during the early observation period (−120 to −160 cm) indicate that the site does not meet widely accepted definitions of fully restored peatlands which typically require sustained water table depths near the surface (approximately −40 cm) to suppress aerobic decomposition and fire risk. Therefore, the site is more accurately described as rehabilitated or revegetated peatland under ongoing hydrological recovery, rather than fully restored peatland. This distinction is critical for both scientific accuracy and policy relevance, as emission factors and mitigation outcomes differ substantially between restored, rehabilitated, and degraded peatlands.

The limited spatial replication in this study (n = 3 per peatland cover type) constrains statistical power and may not fully capture the fine-scale heterogeneity typical of peatland ecosystems. Microsite variability, including hummock–hollow microtopography, vegetation clumping, root density, and peat physical structure, can strongly influence CO₂ fluxes; therefore, the reported mean values may be sensitive to the specific placement of collars and plots. In addition, revegetated peatlands undergoing recovery are often spatially patchy in both hydrological conditions and vegetation structure, reflecting differences in water-table dynamics, species composition, and management history. As a result, the current sampling design may not represent the full range of environmental conditions present across the peatland, and the findings should be interpreted as representative of local conditions within the monitored plots rather than of the entire landscape.

Beyond spatial considerations, the weekly temporal resolution of flux measurements may not capture short-lived but potentially important emission events, such as rapid post-inundation degassing, storm-driven pulses, transient temperature extremes, or management-related disturbances. The omission of such episodic events could influence annual emission estimates if their contributions are not adequately represented by the sampling frequency. Future studies could reduce this uncertainty by increasing measurement frequency or by pairing flux observations with continuous groundwater-level monitoring.

The extrapolation of weekly measurements to annual CO₂ fluxes further requires several assumptions, including temporal stationarity within sampling intervals, representativeness of sampling days, and interpolation across hydrological transitions. These assumptions introduce additional uncertainty into annual estimates. Incorporating uncertainty propagation approaches, such as confidence intervals around annual totals, would strengthen future assessments and improve comparability across studies.

Finally, mechanistic interpretation of the observed CO₂ flux variability is constrained by the absence of partitioning measurements. Without separating autotrophic and heterotrophic respiration, the study cannot definitively attribute observed flux patterns to root activity, microbial decomposition, temperature effects, or water-table dynamics. Accordingly, mechanistic explanations presented here should be regarded as hypotheses supported by existing literature rather than confirmed causal relationships. Moreover, because peatlands differ widely in restoration stage, vegetation composition, drainage intensity, climate, and management regimes, the findings should not be generalized beyond the study site. Instead, they provide site-specific evidence and pathway-aware flux estimates within the context of a revegetated peatland undergoing hydrological recovery.