An assessment of CO₂ and CH₄ emissions in a tropical river: from the Kenyir Reservoir to the estuary

Study site and sampling

Water sampling campaigns were divided into three areas: (i) Kenyir Reservoir (main reservoir only), (ii) Terengganu River (water head and intermediate section, ≈50 km length) and (iii) Terengganu River Estuary (downstream, 15 km length) (Fig. 1). Five sampling campaigns were conducted within two seasons (dry and wet) in seven sampling stations (S1–S7) in Kenyir Reservoir. The reservoir has a maximum depth of 145 m, and the maximum water level drawdown was limited to 10 m; with a minimum dam operating level at 120 m (Rouf et al., 2010). Both dry (March, May, July 2018) and wet (February and November 2018) seasons field sampling campaigns were determined based on local rainfall data obtained from the nearby hydrological station (Sg Gawi station, Kenyir).

Three longitudinal samplings were conducted immediately downstream of the Kenyir Dam under two contrasting river discharge conditions. 13 sampling stations (L1–L13) were chosen depending on land use, logistics and accessibility through jetty and bridges. The sampling campaigns covered ≈50 km in river distance, from the dam outlets to 15 km from the mouth of the Terengganu River. The monthly averaged-discharge rate of the Terengganu River varied from 147 to 224 m³ s⁻¹ (data obtained from the Department of Irrigation and Drainage Malaysia), showing relatively little variation between wet and dry seasons. The river sampling campaigns were designed to capture different discharge conditions rather than focusing on seasonal differences. Specifically, the March and April 2018 sampling campaigns were carried out during the high-discharge season, with river flows exceeding 150 m³ s⁻¹. In contrast, the April 2019 campaign was conducted during the low-discharge season when river flows were below 150 m³ s⁻¹.

The estuarine sampling campaigns were divided into three longitudinal and two time series samplings. The estuarine is a shallow tropical salt-wedge estuary with a depth of 1 to 10 m (average depth: 4.7 m). The longitudinal sampling consists of 15 depth profiling stations (E1–E15) covering a transect distance of 15 km starting from the Terengganu River mouth (Fig. 1). The sampling was conducted during December 2017 (high flow), September (low flow) and December 2018 (high flow). In addition, two 24-hour stationary time series samplings were conducted at different tidal regimes in April 2019 (spring tide) and August 2019 (neap tide). For time series sampling, vertical profile physical water quality measurement was collected every hour, while GHG and nutrients sample collection for surface and bottom depths were collected every two hours.

Data S1 and Table S1 summarise the study sites and sampling activities conducted in the reservoir, river, and estuary during each campaign.

All in situ physical water quality measurements (pH, temperature, conductivity, dissolved oxygen, and hydrostatic pressure) were taken with a calibrated YSI 6-Series Multiparameter Water Quality Sonde (YSI 6600, USA). The vertical water column profile was done by lowering the probe until it reached a safe bottom with the assistance of a portable sonar profiler to avoid underwater equipment entanglement with submerged debris such as branches of dead trees.

Greenhouse gases sampling and analysis

Water samples were taken manually with a 5L Niskin water sampler (General Oceanic, Miami, FL, USA) and carefully transferred via tubing into pre-baked 60 mL borosilicate serum bottles to ensure laminar flow and minimize bubble formation. The samples were immediately preserved with saturated mercury chloride (0.05% vol/vol), sealed with butyl rubber stoppers and crimped metal caps. The samples were transported back to the Universiti Malaysia Terengganu chemistry lab for gas chromatography (GC) analysis, following the protocol described by Lee et al. (2022).The GHG samples were headspaced by replacing one-third of the water with helium gas. The samples were then left to equilibrate at room temperature for 30 min, followed by CO₂ and CH₄ measurements using a gas chromatograph (Agilent 7980 GC system). Five mL of headspace was injected into the GC using a gas-tight syringe. The GC was equipped with 0.9 m by 1.6 cm o.d. column packed with 80/100 mesh HayeSep Q (Agilent J&W). For CH₄ analysis, Helium carrier gas was set at the flow rate of 21 mL min⁻¹. The flame ionization detector operated under the following conditions: N₂ makeup gas two mL min⁻¹, H₂ 48 mL min⁻¹, air 500 mL min⁻¹, and temperature 250 °C. CO₂ concentration was measured by a thermal conductivity detector operated at the column temperature of 120 °C and filament temperature of 200 °C. The GC column oven was operated at an initial temperature of 50 °C for 4 min, then programmed to 90 °C, and held at this temperature for 6 min. The calibration preparation procedure for GC analysis was according to Lee et al. (2022).

The ebullition fluxes experiment was conducted in the shallow part of the Kenyir Reservoir. Three submersible chambers were placed about 0.5 m below the surface. 60 mL serum bottles prefilled with deionized water were used for gas collection. A tube with a 21G needle was connected from the submersible chamber nozzle to the bottom of the serum bottle (inlet), and another 21G needle near the bottleneck served as the outlet. The outlet tube’s end was placed lower than the nozzle’s depth.

The setup required priming to remove air from the chamber and tubing using a syringe attached to a three-way valve. Priming was complete once the setup was filled with water. The experiment ran for 24 h with samples taken in triplicate. The gas volume collected was determined by the water loss in the serum bottle (Gao et al., 2013); CO₂ and CH₄ concentrations were measured using GC as described above. Figure S1 shows the detailed design of the submersible chamber.

Greenhouse gases diffusive flux across the water–air interface

The diffusive flux of CO₂ and CH₄ was obtained by estimating the dissolved GHG concentration of surface water and gas transfer coefficient, k ₆₀₀ (Ferrón et al., 2007; Wanninkhof, 1992). The pCO₂ (or pCH₄) flux is given in Eq. (1). (1)${\displaystyle GasFlux=k_i{K}_i\left(pGa{s}_{2water}-pGa{s}_{2atm}\right)}$ where Gas Flux is the outgassing of CO₂ or CH₄ across air-water interface (mmol m⁻² d⁻¹), k_i (cm h⁻¹) is the gas transfer velocity coefficient, K_i is the solubility constant of CO₂ or CH₄ (mol L⁻¹ atm⁻¹), pGas_water (µatm) is the partial pressure of CO₂ or CH₄ in surface water, and pGas_atm (µatm) is the partial pressure of CO₂ or CH₄ in atmosphere.

Gas transfer velocity, k_i is influenced by water temperature and salinity. Following the method specified by Ferrón et al. (2007), k₆₀₀ (cm h⁻¹) values from each parameterizations was used to calculate k_i for each gas at the recorded temperature and salinity in the field using Eq. (2): (2)${\displaystyle k_i=k_{600}{\left(S{c}_i÷600\right)}^n.}$ For the reservoir, we used n =  − 2/3 for the wind speed <3.7 m/s and −1/2 for higher wind speed (Wanninkhof, 1992). For rivers and estuaries, where the water-air interface is expected to be turbulent rather than smooth, n =  − 1/2 was applied (Guérin et al., 2006; Wanninkhof, 1992). For the Schmidt numbers, Sc_i is the ratio of the kinematic viscosity of water over the diffusivity of the gas (Sc = ν/D). The formulation of Sc_i is derived from Wanninkhof (1992) and expressed in Eq. (3): (3)${\displaystyle Sc=A-Bt+C{t}^2-D{t}^3}$ where t is in degrees Celsius (^∘C), and A, B, C and D are constant for the coefficients listed in Wanninkhof (1992) (refer to Table S2).

Since this study did not measure k₆₀₀, gas fluxes were estimated using parameterizations of k₆₀₀ reported in the literature (Table S3). The final flux determination was obtained by averaging all parameterizations applied for the reservoir and estuary study areas. In contrast, for the lake study area, the Crusius & Wanninkhof (2003) (C&W03) parameterization was exclusively used to estimate fluxes. C&W03 parameterization for the lake was used in the Terengganu River due to the absence of accessible water velocity and depth measurements, which are typically required for accurate k ₆₀₀ estimation in river sections. C&W03, previously applied in a reservoir study, was chosen because it provides the highest transfer velocity (cm h⁻¹) value, offering a conservative upper limit estimate. This approach ensures a comprehensive and cautious estimation of gas fluxes within the constraints of the available data.

In this study, wind data for the u₁₀ (m s⁻¹) in k₆₀₀ was taken from Zippenfenig (2024) global model (Table S4) and applied in the literature k₆₀₀. The local wind speed data was extrapolated to wind speed at 10 m (u₁₀), according to Crusius & Wanninkhof (2003), Eq. (4): (4)${\displaystyle u_{10}=u_z\left[1+\frac{{\left(C_{d10}\right)}^{1/2}}{K}ln\left(\frac{10}{z}\right)\right]}$ where z is the measured wind speed height, C_d10 is the drag coefficient at 10 m in height (0.0013; Stauffer, 1980), and K is the Von Karman constant (0.41).

The ebullition flux was obtained by using the following Eq. (5) from IHA (2010): (5)${\displaystyle \text{Ebullition Flux}\left(\mathrm{mg}{\mathrm{m}}^{-2}{\mathrm{d}}^{-1}\right)=\frac{\text{Gas Con}.\left(\mathrm{mg}{\mathrm{m}}^{-3}\right)\times \text{Gas Vol}.\text{Collected}\left({\mathrm{m}}^3\right)}{\text{Funnel Area}\left({\mathrm{m}}^2\right)\times \text{Sampling Interval}\left(\text{days}\right)}}$ where gas con. is the measured GHG (CO₂ and CH₄), gas vol. collected is the volume of GHG collected in the serum bottle, and funnel area is the surface area of the inverted chamber (refer to Fig. S1).

The GHG emissions were estimated using a surface area-based approach. Surface areas of the Kenyir Reservoir, the Terengganu River, and its estuary were digitized from satellite imagery. The average GHG concentration was determined from field measurements at multiple locations within each waterbody. The total GHG flux was then calculated using the Eq. (6), where the flux density was derived from in situ measurements and expressed in mmol m⁻² d⁻¹. (6)${\displaystyle TotalGHGsEmission\left(mmol{m}^{-2}{d}^{-1}\right)=surfacearea\left(m^2\right)\times \text{mmol}{m}^{-2}{d}^{-1}.}$

Data visualization and statistical analysis

The spatial distribution of GHG and fluxes was mapped using ArcGIS and the Ocean Data Viewer (ODV, AWI) software. A variable marker size approach was employed to display individual data points. Instead of simply showing coloured dots at sample points for visualizing heatmap continuity, the weighted-average gridding spatial interpolation technique available in ODV was used (Schlitzer, 2002). The interpolation X and Y scale-lengths were minimized to prevent overlapping coverage between data points and balance the data structure and smoothness preservation.

The dataset was characterized using descriptive statistics (minimum, maximum, average, median, and standard deviation) performed with SPSS statistical software, and seasonal trends were averaged for comparison. Since the Shapiro–Wilk test (p < 0.05) indicated that the data did not follow a normal distribution, non-parametric methods were used. Specifically, Spearman’s rho was applied to assess correlations, and the Kruskal–Wallis test was employed to compare means across stations, depths, and periods.

Water physical properties

The reservoir was thermally stratified during dry and wet periods (Fig. S2). Surface water temperature ranged from 28.5 °C to 30.5 °C across the study sites. The average temperature difference between the surface and bottom water was significant (p < 0.05, ±5.0 °C). Under stratification conditions, the reservoir maintained a permanent hypoxic hypolimnion. The vertical distribution of oxygen saturation generally became hypoxic after 40 m depth, but several second oxyclines were observed between 20 and 40 m water depth at several shallow sites.

Figure S2 shows that the oxygen saturation levels (DO) differed significantly among the upstream (L1–L4), middle (L5–L8), and downstream (L9–L13) sections of the Terengganu River (Kruskal–Wallis test, p < 0.05). Notably, the DO level immediately downstream of the dam (L1, 1 km) was consistently lower (41 DO%) than other parts of the river (72 DO%). The spatial extent of low DO concentrations in the upstream sections (L1–L4) was directly associated with dam outflow averaged-discharge rates. During the high discharge periods in March 2018 and April 2018, with dam outflow averaged-discharge rates recorded at 163 and 154 m³ s⁻¹, respectively, low DO levels (<50 DO%) extended 11–12 km downstream from the dam (Fig. S2). In contrast, the low flow period in April 2019, with an outflow averaged-discharge rate of 124 m³ s⁻¹, resulted in a considerably reduced spatial extent of low DO, confining the affected area to within one km of the dam outlet.

The longitudinal and vertical profiling of salinity at the estuary indicates that the area is a partially mixed estuary, with tidal amplitudes averaging between 0.8 and 2.9 m. Figure S2 shows that the maximum seawater intrusion occurred approximately eight km from the estuary’s mouth (E8–E15). The shorter saltwater intrusion observed in this study could be attributed to the presence of a breakwater outside the estuary inlet. The Terengganu estuary inlet is protected by a semi-circular coastal breakwater, which has only a small opening to the sea, thereby limiting freshwater outflows.

During September 2018 sampling (low flow), DO levels in the estuary were typically lower near the substrate and higher closer to the water’s surface. In contrast, DO levels were homogeneous throughout the water column during the December 2017 sampling (high flow). The variation in DO concentrations between the two sampling campaigns may be due to seasonal differences in river discharge rates. Although river discharge data were unavailable for the December 2017 sampling campaign, the water level in December 2017 (7.8 m) was higher than in September 2018 (7.3 m). River water level is directly proportional to flow velocity (Mohd Saupi et al., 2018). Consequently, the estuary’s dissolved oxygen levels were often not vertically stratified during high discharge events, as strong river flows tend to homogenize the water column and disrupt any oxyclines. In addition, the DO concentration in the estuary during December 2017 was much lower than during September 2018.

GHG concentrations and fluxes

The CO₂ and CH₄ concentrations in the Kenyir Reservoir were supersaturated, with the CO₂ ranging from 96 to 616 µM, and CH₄ from 0.1 to 127 µM. Seasonally, CO₂ concentrations showed significant differences (Kruskal–Wallis test p < 0.05), while CH₄ concentrations did not show differences within sampling months (Kruskal–Wallis test, p > 0.05). There was a statistically significant difference in CO₂ and CH₄ concentrations between the sampling sites (Kruskal–Wallis test p < 0.05, Fig. S3).

However, CO₂ and CH₄ fluxes in the Kenyir Reservoir did not show significant spatial variations (Kruskal–Wallis test, p > 0.05, Fig. S3). During the dry period, CO₂ fluxes for the Kenyir Reservoir (S1-S7) ranged from 26 to 62 mmol m⁻² d⁻¹, and CH₄ fluxes ranged from 0.02 to 5.0 mmol m⁻² d⁻¹. During the wet period, CO₂ fluxes increased significantly compared to the dry period (Kruskal–Wallis test p < 0.05), ranging from 5.6 to 143 mmol m⁻² d⁻¹, while CH₄ fluxes did not show significant seasonal variation (Kruskal–Wallis test p > 0.05), ranging from 0.1 to 12 mmol m⁻² d⁻¹. Higher CO₂ and CH₄ emissions were detected in the surface water of S3, S4, S6, and S7 (Fig. 2).

Ebullition fluxes during the dry period were significantly higher (Kruskal–Wallis test, p < 0.05), with CH₄ reaching 3.79 × 10⁻² mmol m⁻² d⁻¹ and CO₂: 1.56 × 10⁻² mmol m⁻² d⁻¹, respectively. These rates were at least an order of magnitude higher than those observed during the wet period (CH₄: 3.94 × 10⁻⁸ mmol m⁻² d⁻¹ and CO₂: 1.66 × 10⁻⁴ mmol m⁻² d⁻¹).

Averaged CH₄ concentrations downstream of Kenyir Reservoir (Fig. 3) were significantly higher during the high discharge period (9.1 ± 15 µM, range: 0.2 to 65 µM) than during the low discharge period (1.3 ± 1.2 µM, range: 0.001 to 4.4 µM, Kruskal–Wallis test p < 0.05). High CH₄ and CO₂ concentration anomalies were observed in the upstream (L1–L4) and the middle segment (L7–L10) of the Terengganu River, indicating a potential hotspot for GHG emission.

In the Terengganu River Estuary, CO₂ and CH₄ concentrations across all sites varied between December 2017 (high flow) and September 2018 (low flow, Fig. 4). The CH₄ concentrations ranged from 0.2 to 20 µM during high flow, and between 1.1 to 10 µM during low flow. CO₂ concentrations ranged from 7 to 307 µM during high flow and 27 to 317 µM during low flow.

GHG flux measurements in estuary during high flow showed that CH₄ fluxes averaged 13 ± 5.0 mmol m² d⁻¹ (range: 4.3 to 21 mmol m² d⁻¹), while CO₂ fluxes averaged 20 ± 30 mmol m² d⁻¹, (range: −11 to 112 mmol m² d⁻¹, as shown in Fig. 4). In contrast, CH₄ fluxes remained relatively stable during low flow, averaging 4.1 ± 2.8 mmol m² d⁻¹ (range: 1.6 to 10 mmol m² d⁻¹). However, CO₂ fluxes were significantly higher (Kruskal–Wallis test P < 0.05), with an average of 232 ± 95 mmol m² d⁻¹ (range: 70 to 350 mmol m² d⁻¹).

Time-series measurements during neap and spring tides revealed notable differences in GHG concentrations and fluxes. During neap tide, CH₄ concentrations averaged 1.5 ± 0.6 µM, (range: 0.4 to 2.6 µM), while CO₂ concentrations averaged 56 ± 30 µM (range: from 20 to 109 µM). In contrast, CH₄ concentrations were lower during spring tide, averaging 0.6 ± 0.4 µM (range: below detection limit to 1.4 µM), and CO₂ concentrations were slightly higher, averaging 61 ± 38 µM (range: 24 to 126 µM).

Over a 24-hour sampling period, GHG fluxes during neap tide showed that CH₄ averaged 0.7 ± 0.3 mmol m² d⁻¹ (range: 0.2 to 1.3 mmol m² d⁻¹, Fig. 5), while CO₂ fluxes averaged 23 ± 16 mmol m² d⁻¹ (range: 3.2 to 53 mmol m² d⁻¹) During spring tide, the CH₄ fluxes averaged 0.6 ± 0.4 mmol m² d⁻¹ (range: −0.09 to 1.3 mmol m² d⁻¹), whereas CO₂ fluxes increased, averaging 55 ± 44 mmol m² d⁻¹ (range: 10 to 131 mmol m² d⁻¹).

Greenhouse emissions from Kenyir Reservoir

Temperature and dissolved oxygen profiles indicate that the Kenyir Reservoir is a meromictic system with stable thermal stratification year-round (Fig. S2). This thermal stratification drives oxygen stratification, isolating bottom waters from atmospheric exchange and resulting in anoxic conditions that enhance methane production through methanogenesis (Davidson, Audet & Jeppesen, 2018). Under these low-oxygen conditions, organic matter decomposition also contributes to elevated carbon dioxide levels. As a result, oxygen stratification enhances both CH₄ and CO₂ production in deeper waters (Fig. S3).

Statistically, the average CO₂ emission rate across all sites was significantly higher during the wet season (Kruskal–Wallis test p < 0.05). This increase may be attributed to stronger winds. In this study, monthly average wind speed in the Kenyir Reservoir was 1.3 times higher during the wet season than in the dry season (Table S4). This finding aligns with the findings of Sawakuchi et al. (2017) that windier conditions promote surface water mixing and enhance the diffusion of dissolved GHG into the atmosphere.

Ebullition was observed in the shallow southern region near S7 (Fig. S2), likely due to shallow water. Previous studies (Bastviken et al., 2004; Gerardo-Nieto et al., 2017; West, Creamer & Jones, 2016; DelSontro et al., 2010) have shown that ebullition is more common in littoral zones. In the Kenyir Reservoir, active ebullition occurred near to S7, where water depths were less than 40 m.

The overall water level increase in Kenyir Reservoir during the wet season was associated with a lower ebullition rate, as higher hydrostatic pressure suppresses gas bubble formation, similar to the observation of Ostrovsky et al. (2008) in Lake Kinneret. We assume that GHG production from the bottom sediments remains consistent year-round due to the persistent anoxic conditions at the reservoir bottom (Fig. S4). However, intense wind forcing during the wet season (Table S4) enhances internal lake motion, disrupting the ascent of GHG bubbles. McGinnis et al. (2006) demonstrated that increased horizontal currents break larger bubbles into smaller ones, increasing the total surface area available for bubble redissolution before they reach the surface. Consequently, these processes result in fewer GHG bubbles captured by submersible trapping funnels in S7 during the wet season.

Table 1 presents the CO₂ and CH₄ emissions from 39 freshwater reservoirs and lakes latitudinally from north to south. Kenyir Reservoir ranks 9th and 8th for CH₄ and CO₂ among the global freshwater GHG emissions references in Table 1. The complied data showed that tropical reservoirs release more GHG than boreal, temperate, and subtropical reservoirs. For instance, Tucurui Reservoir in Brazil measured 0–180 mmol CH₄ m⁻² d⁻¹ and 30–3,242 mmol CO₂ m⁻² d⁻¹ in 1993 (Dos Santos et al., 2006). Curuá-Una reservoir measured CH₄ emissions of 0.1–7.0 mmol m⁻² d⁻¹ and CO₂ emissions of 0.6–1,612 mmol m⁻² d⁻¹ in 2017 (Paranaíba et al., 2021), which were nearly ten times higher than Lake Allatoona, USA, in the sub-tropical reservoir that recorded 11.7 mmol CH₄ m⁻² d⁻¹ and 54.8 mmol CO₂ m⁻² d⁻¹ in 2012 (Bevelhimer et al., 2016).

Tropical reservoirs exhibit higher GHG emissions (0–180 mmol CH₄ m⁻² d⁻¹ and −36–3,243 mmol CO₂ m⁻² d⁻¹) than reservoirs in other climates (0–18.3 mmol CH₄ m⁻² d⁻¹ and −5.5–213 mmol CO₂ m⁻² d⁻¹). This difference is largely due to greater organic carbon in tropical reservoirs and higher surface runoff, which continuous supplies organic matter, fueling bacterial metabolism and enhancing GHG production (Tranvik et al., 2009). Table 1 also shows higher GHG emissions are commonly observed in young reservoirs (less than 10 years). For example, in the first year after damming, annual diffusive CO₂ and CH₄ emissions in reservoirs, such as Nam Theun 2, Nam Leuk (Laos), Petit-Saut (French Guiana), and Tucurui (Brazil), were approximately 1 to 10 times greater than in later years. Despite being more than 30 years old, the Kenyir Reservoir continues to exhibit high CO₂ and CH₄ emission than other reservoirs with similar age, such as the Batang Ai Reservoir (Soued & Prairie, 2020) and Reservoir Lokka (Huttunen et al., 2004).

Greenhouse gas emissions downstream of a dammed river

High GHG concentrations in the Terengganu River were observed near the dam discharge outlets (4.4–36.5 µM CH₄ and 30.3–194 µM CO₂), extending up to 12 km downstream (0.7–3.6 µM CH₄ and 80–133 µM CO₂) from L1 to L4 (Fig. 3). In these upstream sections, CH₄ concentrations were negatively correlated with dissolved oxygen (r =  − 0.6, p < 0.05), suggesting that the elevated CH₄ levels originated from the hypoxic bottom waters of the reservoir. Consistently high CH₄ concentrations were recorded near the reservoir’s bottom (mean 621 ± 187 µM, n = 20), where oxygen depletion provides favorable conditions for methanogenesis (Conrad, 2020; Müller et al., 2019). From site L1 to L4, immediately downstream of the reservoir outlets, CH₄ concentration decreased by approximately 80% (Fig. 3). This decline was likely driven by a combination of dilution, atmospheric evasion, and microbial oxidation (DelSontro et al., 2016b; Guérin & Abril, 2007; Kemenes, Forsberg & Melack, 2007). Further downstream at L5 and L6, CH₄ concentrations dropped below five µM.

The substantial increase in CO₂ and CH₄ concentration in the middle segment of the Terengganu River (L7 to L10, Fig. 3) is most likely linked to in-stream sand extraction activities. At least four active in-stream sand extraction operations were observed during the sampling campaign along this river section. These activities involved submersible pumps extracting riverbed materials comprising gravel, sand, silt, and mud, resulting in the mineralization of resuspended particulate organic matter, releasing substantial CO₂ and CH₄ into the overlying water column. In addition, the disturbance of riverbed sediments may have promoted the degassing of CH₄ and CO₂ from porewater (Marcon et al., 2022). Qin et al. (2020) also reported that in-stream sand extraction disrupts the carbon sequestration potential of riparian areas, significantly reducing CO₂-fixing microbial communities and leading to increased GHG emissions. Beyond its biogeochemical impacts, sand extraction disturbs sediment supply and transport equilibrium, triggering morphological alteration that causes irreversible changes in the watershed characteristics (Padmalal & Maya, 2014). The resuspension of sediments increases turbidity in the water column (Ashraf et al., 2011). Over time, reduced light penetration limits photosynthesis and oxygen production, promoting the anaerobic condition in the benthic layer, which enhances microbial decomposition of organic matter, further increasing CO₂ and CH₄ production.

The Terengganu River is a source of CO₂ and CH₄ to the atmosphere with fluxes ranging from 0.7 to 136 mmol CO₂ m⁻² d⁻¹ and 0.06 to 20 mmol CH₄ m⁻² d⁻¹, respectively. In-stream sand extraction is an important contributor to the river’s CH₄,  and CO₂ flux. Overall, sand extraction activity alone has contributed more than 50% of Terengganu River’s total carbon (CO₂+CH₄) emission (Table 2). Downstream discharge water from the dam is the second largest CH₄ source, contributing 30% of the diffusive flux to the atmosphere.

The median concentration of CO₂ (99 µM) and CH₄ (1.6 µM) in the Terengganu River was at least 6 and 8,000 times higher than atmospheric levels (∼17 µM CO₂; ∼0.002 µM CH₄), respectively. Overall, CO₂ and CH₄ levels in the Terengganu River fall within the typical range of global tropical rivers (42–337 µM CO₂; 0.4–350 µM CH₄) but are at least an order of magnitude lower than those in dammed rivers and peat-draining rivers (Table 3). In terms of diffusive flux, Terengganu River CO₂ (0.7 to 136 mmol m⁻² d⁻¹) and CH₄ (0.06 to 20 mmol m⁻² d⁻¹) fluxes are comparable to those reported for Wohlen Reservoir (DelSontro et al., 2016b), River Tay (Harley, 2013), Tianjin River (Hu et al., 2018), and as well as Kariba Dam (Teodoru et al., 2015). However, they are one to two orders of magnitude lower than fluxes reported for tropical reservoirs and rivers in Brazil (Abril et al., 2005; Guérin et al., 2006; Kemenes, Forsberg & Melack, 2007; Kemenes, Forsberg & Melack, 2011), Indonesia (Wit et al., 2015), and Malaysia (Müller et al., 2015).

Greenhouse gas emission dynamics in Terengganu River Estuary

The average CH₄ concentrations in the estuary were significantly higher during high flow (9.2 ± 4.4 µM) than during low flow (3.6 ± 2.1 µM). Elevated CH₄ levels observed during the high flow period in December 2017 were likely driven by substantial rainfall-induced urban runoff. Increased dissolved organic carbon during these events may have contributed to higher CH₄ concentration in the estuary (Fig. 4). Previous studies have shown that river segments draining urban areas are important CH₄ sources (Tang et al., 2021). CO₂ and CH₄ removal were observed from sites E7 to E15 during low flow periods, probably due to mixing with seawater in the lower estuary (Fig. 4).

Seasonally, the largest range of CO₂ fluxes was observed during the dry season (September 2018), ranging from 70 to 350 mmol m⁻² d⁻¹ compared to the wet season (December 2017), which ranged from −11 to 110 mmol m⁻² d⁻¹ (Fig. 4). In contrast, average CH₄ fluxes were higher in December 2017 (13.1 mmol m⁻² d⁻¹) than in September 2018 (4.1 mmol m⁻² d⁻¹) (p < 0.05). Overall, CO₂ and CH₄ efflux in TRE displayed a decreasing trend from the upper to the lower reaches (Fig. 4). The elevated CH₄ emissions in the freshwater portion during December 2017 were likely due to inputs from nearby tributaries, stormwater discharge outlets, and non-point sources. Additionally, breakwater structure and land reclamation near the estuarine inlet may have altered water column mixing, potentially facilitating in situ GHG production (Fig. 4). This observation is consistent with Looman, Maher & Santos (2021), who reported increased localized outgassing due to altered watercourse hydrodynamics. Conversely, the decline in CO₂ and CH₄ fluxes in the lower estuary was likely driven by degassing from tidal oscillation and dilution with seawater.

The time-series CO₂flux in the estuary varied primarily with tidal height rather than diurnal fluctuations (Fig. 5). During the spring tide, the maximum average CO₂ flux reached 40 mmol m⁻² d⁻¹ during the daytime and 74 mmol m⁻² d⁻¹ at night. In contrast, during the neap tide, CO₂ flux ranged from 3 to 53 mmol m⁻² d⁻¹, peaking at 51 mmol m⁻² d⁻¹ at 0400 h. The average CO₂ flux during the daytime was 31 mmol m⁻² d⁻¹, while at night it was 17 mmol m⁻² d⁻¹.

Similarly, CH₄ fluxes were strongly influenced by tidal conditions. During spring tides, the average CH₄ flux reached 0.7 mmol m⁻² d⁻¹ during the ebb tide, significantly higher than the 0.008 mmol m⁻² d⁻¹ recorded during the flood tide. For neap tides, CH₄ emissions showed a comparable amplitude to those during spring tides, ranging from 0.2 to 1.3 mmol m⁻² d⁻¹, with a peak of 1.3 mmol m⁻² d⁻¹at 1400 h. The average daytime emission was 0.9 mmol m⁻² d⁻¹, while nighttime average was 0.6 mmol m⁻² d⁻¹.

Although these differences suggest a possible diel component, tidal height appears to be the dominant factor controlling emissions. Higher tidal amplitudes during spring tides correspond to longer ebb tide duration (Mao et al., 2004), which enhances tidal damping and facilitates more efficient material exchange and seaward transport (Arueira et al., 2022; Geyer, Chant & Houghton, 2008; Siegle et al., 2009). During the ebb tide, as water levels in the estuary drop, the reduced water column allows trapped CO₂ in the organic-rich sediment to be released more effectively.

Similarly, CH₄ emission in estuaries strongly depends on tidal height variation, occurring predominantly at low tide and primarily in shallow water depth systems (Ferrón et al., 2007; Grunwald et al., 2007; Harley, 2013). Harley (2013) showed that a falling tide can partially replace the estuary water column with CH₄-supersaturated freshwater from the upper estuary. Ferrón et al., (2007) and Grunwald et al. (2007) also observed that CH₄ concentration increases as tidal amplitude and salinity decrease, attributing these changes to allochothonous inputs, urban effluent discharge, in situ production, or the drainage of CH₄-rich pore water from tidal flats.

Table 4 compares CH₄ diffusive flux in the studied estuary with other river estuaries worldwide. The CH₄ flux in the estuary is comparable to the Jiulong River Estuary, China (Li et al., 2021), Adyar Estuary, India (Nirmal Rajkumar et al., 2008) and Scheldt Estuary, United Kingdom (Middelburg et al., 1996), spanning sub-tropical, tropical, and temperate regions. However, the CO₂ diffusive flux in Terengganu River estuary exceeds the range reported for tropical river estuaries. It is an order of magnitude higher than the sub-tropical estuaries in China (Yu et al., 2013) and the USA (Jiang, Cai & Wang, 2008). High CH₄-emitting estuaries in Table 4 share two key similarities: elevated suspended particulate matter levels and an external forcing mechanism that enhances CH₄ production. These external factors include tidal induced resuspension (Li et al., 2021), prolonged water residence times due to reduced river discharge (Nirmal Rajkumar et al., 2008), freshwater inflow influced by sewage (Middelburg et al., 1996), and hydrodynamic alterations from breakwater and land reclamation (this study). Additonal contributing factors include pollution, monsoonal influences, and estuarine morphological changes (Borges, Abril & Bouillon, 2018; Gupta et al., 2009).

The total GHG emissions from the Terengganu River catchment are estimated at 572 Gg CO₂-equivalent per year, with 85% CO₂ and 15% CH₄. This estimate applies to a global warming potential of 27 for CH₄, based on a value from the IPCC Sixth Assessment Report (AR6). The catchment spans 386.6 km², with the upper freshwater section (Kenyir Reservoir) covering 369 km² and containing highly supersaturated dissolved GHG that influences the 61.5 km-long Terengganu River and its estuary. The river occupies ∼10 km², while the estuary spans 7.6 km². Average GHG flux from each section was multiplied by their respective surface areas, revealing that 94% of total emissions originated from the reservoir, 5.5% from the estuary, and only 0.5% from the river. Despite its smaller surface area than the river, the estuary contributes a disproportionately high share of emissions due to its higher organic matter content and reduced water turnover, exacerbated by the semi-enclosed breakwater that further elevated CH₄ emissions during the wet season.