Estuaries play an important role connecting the carbon cycle across the land-to-ocean aquatic continuum, processing large amounts of allochthonous and autochthonous carbon1. This is despite estuaries constituting only a small fraction of the world’s surface (0.2%)2 compared to continental shelf seas (5%)3 and the open ocean (64%)4. Carbon from upstream rivers and associated coastal wetlands entering estuaries is either buried (and potentially stored long-term), emitted to the atmosphere in the form of greenhouse gases, or exported to the ocean5,6. Estuarine CO2 emissions are estimated to equate to the size of the CO2 sink in shelf seas (0.268 ± 0.225 Pg C yr−1), or 19% of the CO2 sequestration in the open ocean7, but early estimates were mostly based on studies in the northern hemisphere (e.g. refs. 5,6,8,9). CO2 emissions from estuaries can differ between estuary geomorphic types10,11, but the mechanisms by which geomorphology affects estuarine CO2 emissions in Australia have not been determined. There is also limited knowledge of how disturbance impacts CO2 emissions and how different geomorphic estuary types modify any disturbance effect.

Low to high disturbance and land-use changes in the upper catchment have the potential to alter the quantity and quality of carbon delivered to estuaries5,12, and hence the associated estuary CO2 emissions12,13. Dissolved inorganic carbon (DIC)14, dissolved organic carbon (DOC)15, and particulate organic carbon (POC)16 inputs typically increase in impacted estuaries and tend to be associated with increased CO2 emissions7,17,18. However, the effect of land-use on CO2 emissions from estuaries can vary12,19. For instance, moderately and highly disturbed estuaries in Australia have been reported to emit more CO2 per unit area (37 ± 10 mmol CO2-C m−2 d−1) than less disturbed estuaries (6.3 ± 4 mmol CO2-C m−2 d−1)12, whereas very small coastal estuaries with high land-use changes (>90% of catchment modified) had lower CO2 emissions than estuaries with low land-use changes (~21% of catchment modified)15. Moreover, changes in land-use along an estuarine gradient can influence nutrient cycling (e.g., decomposition) and change the quantity and quality (labile or refractory) of organic matter inputs20,21,22, resulting in increases or decreases in CO2 emissions between riverine-upstream, mid-estuary, and near-marine regions23,24.

Estuaries of different geomorphology are the result of varying influences of river discharge, tidal amplitude, and wave energy, which determine estuarine hydrological characteristics such as water depth, current velocity, and water residence times2,25,26. In turn, water depth and current velocity influence the gas transfer velocity (k) which controls the rate of CO2 emission from the water into the atmosphere27,28,29. Water residence times control CO2 emissions in estuaries by determining the direction and intensity of estuarine water-air CO2 fluxes2,7, because long water residence times allow for more carbon decomposition, resulting in increased DIC that can be emitted as CO27,28,30. Shorter water residence times accelerate DOC export to the ocean, resulting in lower CO2 emissions17,18. Stratification of the water column can also influence CO2 emissions31,32. Photosynthetic CO2 uptake occurs in the surface layer, whereas CO2 respired in the bottom waters is isolated from atmospheric exchange at the surface32,33,34, resulting in an overall increase in CO2 partial pressure (pCO2) but unaffected water-air CO2 flux rates32,34. Stratification occurs in estuaries with weak tidal forcing, which leads to the separation of water layers of different densities (i.e. salinities)31,35 or temperatures (thermohaline stratification)32,33,34. In estuaries with stronger tidal influence, tidal pumping can also increase CO2 emissions through the lateral import of DIC and DOC from coastal wetlands to the estuary36,37. Tidal pumping can be a significant driver of CO2 emissions, with groundwater-derived export accounting for 93% to 99% of DIC and 89% to 92% of DOC exported from mangroves into tidal creeks36. In very large river systems2 such as the Amazon River38 in Brazil and the Yangtze39 River in China, riverine transport from the estuary to the ocean can result in extensive estuarine plumes that can act as either a source or a sink of CO2. However, such systems do not exist in Australia. A recent global analysis showed that fjords predominantly act as CO2 sinks, while tidal systems and deltas emit more CO2 than lagoons11.

Australia makes a significant contribution to the total number and surface area of estuaries globally. Australia’s coastline of 36,700 km includes 971 estuaries40, accounting for 1.82% of the total number of estuaries globally41 and 2.35% of global estuarine surface area2. More importantly, the majority of Australia’s estuaries (70.6%) are classified as low or moderately disturbed40, contrasting with predominantly disturbed estuaries in Europe and the United States where the majority of estuarine CO2 emissions have been measured9,10. This reflects Australia’s population density of only 3.3 persons km−242, the 3rd lowest in the world. Despite Australia’s contribution to global estuary number and surface area, CO2 emissions have been measured in only a few Australian estuaries (e.g. refs. 12,23,43,), and there are no estimates of total CO2 emissions from Australian estuaries. In this study, we (1) calculated CO2 emissions from pCO2 measurements from 36 estuaries in different climate zones in Australia and combined these CO2 emission estimates with previously published CO2 emissions from 11 other Australian estuaries12,30 (total of 47 estuaries), and (2) assessed the interaction effects of anthropogenic disturbance and geomorphology on CO2 emissions for these 47 estuaries. Based on disturbance and geomorphology classifications of Australian estuaries, we then scaled up CO2 emissions from the 47 estuaries to all 971 assessed estuaries to better constrain Australia’s contribution to global estuarine CO2 emissions. We hypothesised that estuarine geomorphic type and disturbance level (including land-use change) would significantly impact estuarine water column pCO2 and CO2 emissions, and that there would be an interaction between geomorphic type, disturbance level, and CO2 emissions. We further hypothesised that relative CO2 emissions from Australian estuaries would be lower than global estuary CO2 emissions because of generally lower disturbance found in estuaries in Australia.


Physical differences between estuary types

Mean (min-max) tidal range was highest in tidal systems (n = 14) (3.7 m (1.1–6.0 m)), moderate in small deltas (n = 12) (1.5 m (1.1–2.1 m)), and generally lower in lagoons (n = 21) (0.6 m (0–1.4 m)) (Fig. 1E). Water depth (n = 92, 75, and 121, respectively) (Fig. 1A) and current velocity (n = 101, 67, and 112, respectively) (Fig. 1B) significantly increased (p = 0.001) from lagoons to small deltas to tidal systems. Wind speed was significantly higher in lagoons (n = 88) than in small deltas (n = 79) (p = 0.004) and tidal systems (n = 126) (p = 0.001), but was similar between small deltas and tidal systems (p = 0.137) (Fig. 1C). The mean gas transfer velocity normalised to the Schmidt number of 600 (k600), was highest in tidal systems and significantly lower in small deltas (p = 0.001). Although lagoons had the lowest k600 (n = 751), it was not significantly different from small deltas (n = 667) and tidal systems (n = 1036) (negative t-values) (Fig. 1D and Supplementary Fig. 1). Tidal range significantly increased from the lowest in lagoons (n = 21) to the highest in tidal systems (n = 14, small deltas: n = 12) (p = 0.001) (Fig. 1E). Temperature differed significantly between estuary types (lagoons: n = 751) (p = 0.001) with the lowest mean temperature in small deltas (n = 719) and the highest mean temperature in tidal systems (n = 1138) (Fig. 1F and Supplementary Table 3). Across all estuary types, except tidal systems, and within estuary types, temperature did not significantly correlate with pCO2 and water-air CO2 flux. In tidal systems, there was a significant increase in water-air CO2 flux with temperature (r = 0.254, p = 0.01).

Fig. 1: Physical parameters in the three estuary types.
figure 1

Median (red line), mean (red asterisk), 1st and 3rd interquartile ranges (box caps), minimum, and maximum values (whiskers) of (A) water depth, (B) current velocity, (C) wind speed, (D) mean gas transfer velocity normalised to Schmidt no. 600 (k600) calculated from the five parameterisations (Table 2), (E) tidal range, and (F) temperature in the lagoons (blue, n: A = 92, B = 101, C = 88, D and F = 3789, and E = 21), small deltas (green, n: A = 75, B = 67, C = 79, D = 3362, E = 12, and F = 3622), and tidal systems (yellow, n: A = 121, B = 112, C = 126, D = 5207, E = 14, and F = 5720). Outliers were omitted from the graphs. Letters above figures denote statistical differences among estuary types, with letters that are the same indicating no significant difference (PERMANOVA, two-tailed, and at 95% confidence interval). Source data are provided as a Source Data file.

Estuary pCO2 and water-air CO2 fluxes

The majority of the estuaries studied were a source of CO2 to the atmosphere (Fig. 2A). Mean (±SE) pCO2 and water-air CO2 fluxes were 799 ± 13 µatm and 26.4 ± 0.9 mmol CO2-C m−2 d−1 in lagoons (n = 751), 1181 ± 12 µatm and 63.9 ± 1.1 mmol CO2-C m−2 d−1 in small deltas (n = 719), and 1007 ± 6 µatm and 54.8 ± 0.8 mmol CO2-C m−2 d−1 in tidal systems (n = 1138) (Table 1). Six small deltas and three tidal systems had small sections that were weak CO2 sinks (>−3.5 mmol CO2-C m−2 d−1). Four lagoons were overall CO2 sinks, whereas 14 lagoons had sections that were strong CO2 sinks (up to −64.7 mmol CO2-C m−2 d−1). Although pCO2 and water-air CO2 fluxes in the lagoons had the largest range, pCO2 and water-air CO2 fluxes were significantly lower than in small deltas and tidal systems (p = 0.001), with lower means and medians (Fig. 2A). pCO2 (p = 0.693) and water-air CO2 fluxes (p = 0.064) in small deltas were not significantly different to those in tidal systems (Fig. 2A).

Fig. 2: CO2 partial pressure (pCO2) and water-air CO2 flux in estuary types and disturbance groups.
figure 2

Median (red line), mean (red asterisk), 1st and 3rd interquartile ranges (box caps), minimum and maximum values (whiskers) of pCO2 and water-air CO2 flux at per-minute resolution in the (A) lagoons (blue, n = 3789), small deltas (green, n = 3622), and tidal systems (yellow, n = 5720); and (B) low (black, n = 1796), moderate (dark grey, n = 3189), high (grey, n = 3677), and very high (light grey, n = 4469) disturbance groups. Outliers were omitted from the figures. Dotted line along the x-axis represents atmospheric pCO2 and water-air flux CO2 equilibrium. Letters above figures denote statistical differences among estuary types, with letters that are the same indicating no significant difference (PERMANOVA, two-tailed, and at 95% confidence interval). Source data are provided in the Source Data file.

Table 1 Descriptive statistics calculated for pCO2 and water-air CO2 fluxes using data at per-minute resolution of each estuary type, disturbance group, and in the disturbance groups within each estuary type

Disturbance effects on estuary CO2

pCO2 and water-air CO2 fluxes significantly increased with greater disturbance in Australian estuaries (low to very high disturbance, n = 356, 633, 731, and 888, respectively) (Fig. 2B). For example, mean water-air CO2 fluxes across all estuary types increased from 34.1 ± 0.8 mmol CO2-C m−2 d−1 in the low disturbance systems to 63.8 ± 1.2 mmol CO2-C m−2 d−1 in the very high disturbance systems (Table 1). However, the effect of disturbance on pCO2 and water-air CO2 fluxes was estuary type specific (Fig. 3). In the lagoons, pCO2 in the low disturbance systems (n = 41) was below atmospheric equilibrium (Fig. 3A1), with CO2 influx from the atmosphere into the estuarine waters (range: 17 to 436 µatm, −46.2 to 4 mmol CO2-C m−2 d−1; Figure 3A2). pCO2 increased significantly in higher disturbance lagoons (p = 0.001), except between highly and very highly disturbed lagoons (p = 0.934) (moderate to very high disturbance, n = 161, 261, and 288, respectively). Similarly, water-air CO2 flux in lagoons increased significantly with higher disturbance (p = 0.001), but only from low to high disturbance, and was significantly lower in very high disturbance lagoons compared to high disturbance lagoons (p = 0.037).

Fig. 3: CO2 partial pressure (pCO2) and water-air CO2 flux in estuary type disturbance groups.
figure 3

Median (red line), mean (red asterisk), 1st and 3rd interquartile ranges (box caps), minimum and maximum values (whiskers) of (row A) pCO2, and (row B) water-air CO2 flux at per-minute resolution across different disturbance groups within (column 1) lagoons (from low (light blue) to very high disturbance (dark blue), n = 214, 815, 1312, and 1448), (column 2) small deltas (n; high (light green)=1777, and very high (dark green)=1845), and (column 3) tidal systems (from low (light yellow) to very high disturbance (dark brown), n = 1582, 2374, 588, and 1176). Outliers were omitted from the figures. Dotted line along x-axis represents atmospheric pCO2 and water-air CO2 flux equilibrium. Letters above figures denote statistical differences among estuary types, with letters that are the same indicating no significant difference (PERMANOVA, two-tailed, and at 95% confidence interval). Source data are provided in the Source Data file.

In the small deltas, pCO2 was significantly higher in the very high disturbance systems (n = 366) compared to the high disturbance systems (n = 353) (p = 0.001), but water-air CO2 flux was similar between the high and very high disturbance systems (p = 0.101) (Fig. 3B). No measurements were taken in low and moderate disturbance small deltas. In the tidal systems, disturbance effects on pCO2 were insignificant in the low (n = 315) and moderate (n = 472) (p = 0.682), and low and high (n = 117) disturbance systems (p = 0.118) but significantly increased from the moderate to high disturbance systems (p ≥ 0.006) (Fig. 3C). pCO2 in the very high (n = 234) disturbance tidal systems was significantly greater than in low disturbance systems (p = 0.001). Water-air CO2 fluxes significantly increased with higher disturbance (p = 0.001) except between the low and high disturbance systems (p = 0.094). Water-air CO2 fluxes were greatest in very high disturbance systems (Fig. 3C2 and Table 1).

Seasonal CO2 emissions from Australian estuaries

We estimated that Australian estuaries emit a mean (±SE) of 7.62 ± 0.48 Tg CO2-C yr−1 over the summer season, of which tidal systems contributed 93.4%, and lagoons and small deltas contributed 4.4% and 2.2%, respectively. To estimate winter water-air CO2 fluxes, seasonal ratios from published summer and winter water-air CO2 fluxes from 13 estuaries (Supplementary Table 1) were averaged to obtain the seasonal ratio (winter CO2 flux: summer CO2 flux) for each estuary type (Supplementary Table 2) which were then applied to summer water-air CO2 fluxes from the current study. In these 13 estuaries, lagoons in winter had a lower mean CO2 uptake, with seasonal ratios ranging from 0.58 to 0.69 (mean: 0.64; Supplementary Table 2). Small deltas and tidal systems in winter had higher mean water-air CO2 flux rates than in summer, with seasonal ratios ranging from 0.33 to 4.71 in small deltas (mean: 1.49) and 0.43 to 2.17 in tidal systems (mean: 1.3) (Supplementary Table 2). Winter water-air CO2 fluxes in Australian estuaries had a mean (±SE) flux rate of 49.7 ± 7.5 mmol CO2-C m−2 d−1, 25.8% higher than summer flux rates (Table 2).

Table 2 Mean, standard error (SE), and median for summer, winter, and annual water-air CO2 flux rates and annual CO2 emissions from Australian estuaries (Estuary (Est.) types: La: lagoons, SD: small deltas, TS: tidal systems; Disturbance (Dist.) groups: 1: low, 2: moderate, 3: high, 4: very high)

Annual CO2 emissions from Australian estuaries

The mean annual CO2 emission from Australian estuaries was 8.67 ± 0.54 Tg CO2-C yr−1 (Table 2), with tidal systems accounting for 94.4% of annual CO2 emissions, followed by small deltas at 2.5%, and lagoons at 3.1%. These proportions compared with surface areas for tidal systems, lagoons, and small deltas, representing 89.9%, 8.6%, and 1.5% of total Australian estuary surface area, respectively (Table 3). Due to the larger surface area coverage of lagoons with increased disturbance (Table 3), CO2 emissions from lagoons were dominated by the higher disturbance systems. High disturbance lagoons had the greatest CO2 flux rates, but very high disturbance lagoons covered a greater proportion of lagoon surface area (62%) and therefore, as a category, emitted the most CO2. In contrast, lower disturbance small deltas and tidal systems covered the largest proportion of their respective estuary-type surface area and emitted the most CO2 annually (Tables 2 and 3). Low and moderately disturbed tidal systems had the greatest total emissions, driven mainly by the large surface area coverage (33% and 48%, respectively) in remote northern Australia. Very high disturbance small deltas had the highest water-air fluxes and the largest proportion of total small delta surface area (50%) and therefore, emitted the most CO2 annually. Low disturbance lagoons were the only CO2 sinks of all the estuary types and disturbance groups, whereas in tidal systems the very high disturbance systems emitted the least CO2 annually. The low, moderate, and high disturbance small deltas emitted similarly low levels of CO2 annually (Table 2). Moderately disturbed Australian estuaries had the largest CO2 emissions, followed by the high, low, and very high disturbance systems (Table 2).

Table 3 Number of estuaries, estuarine surface area, and percent of total estuary surface area classified by estuary type according to Dürr et al.2 and classified by disturbance in NLWRA40 for sampled estuaries in this study and in all Australia


There was a strong geomorphic effect on measured pCO2 and water-air CO2 fluxes in Australian estuaries (Fig. 2A1 and A2), with the lagoons particularly different from the small deltas and tidal systems. Overall, lagoons had the lowest pCO2 and water-air CO2 fluxes of the three geomorphic types. This was likely driven by higher benthic productivity, which can result in a net autotrophic system with CO2 uptake across the water-air interface43,44 over a diurnal period45,46. Indeed, seagrass meadows cover an average of 18% of lagoon water areas in NSW, compared to only ~6% in small deltas and tidal systems47. Consistent with this, CO2 undersaturation and CO2 uptake have been reported in three Australian lagoons43, as well as non-Australian marine-dominated shallow coastal systems with a large cover of benthic vegetation (e.g. refs. 44,48,49). Freshwater input could also be a driver of pCO2 and water-air CO2 fluxes in estuaries; freshwater is typically supersaturated with CO2 and a source of allochthonous organic matter5,50,51,52 that may subsequently decompose and release CO27,51. However, we found poor relationships between salinity and pCO2 and water-air CO2 fluxes in Australian lagoons (Supplementary Fig. 2), suggesting that freshwater organic matter was not an important source of CO2 in these systems. This may reflect a weak hydrological connection between lagoons and upstream rivers, which would limit input of riverine water. This is consistent with a previous study showing lower CO2 emissions in estuaries with lower riverine input compared to river-dominated estuaries53.

pCO2 and DIC concentration were higher in small deltas and tidal systems compared to lagoons (Figure 2A1, Supplementary Fig. 4B1, Supplementary Results). The inverse relationships between salinity and pCO2 or water-air CO2 fluxes in small deltas and tidal systems indicate that CO2 outgassing from the re-mineralisation of organic matter in upstream waters had a larger contribution in these systems compared to the lagoons31, as seen in other estuaries with higher riverine input53. This is because increasing pCO2 and water-air CO2 fluxes with decreasing salinity indicate that increasing CO2 is linked to freshwater input upstream. The tidal systems and small deltas also have a stronger connection to the river and associated input of CO2 supersaturated water2, which enhances CO2 emissions in the estuary. In our study, all of the small deltas and tidal systems were tropical and sub-tropical (23.5° to 35° latitude) (Supplementary Data 1), where much of the atmospheric carbon uptake and sequestration occurs within their mangrove-lined shorelines54,55. Lateral export from vegetated shorelines into adjacent estuaries can be a significant pathway for the transport of carbon in the form of DOC, POC, DIC, and CO2-rich pore water, or as a result of the degradation of exported organic matter56,57,58. Increased tidal range resulted in an increase in pCO2 and water-air CO2 fluxes (Supplementary Results), suggesting increased pCO2 and water-air CO2 fluxes were due to lateral export in our estuaries56,57,58, although we did not directly measure lateral inputs. Although estuaries in lower latitudes have higher water temperatures that could drive increased water-air CO2 fluxes, water temperature did not correlate with pCO2 and water-air fluxes in our estuaries (Supplementary Results). As such, higher pCO2 and DIC concentrations in small deltas and tidal systems compared to lagoons could likely be attributed to increased lateral inorganic (and organic) carbon export from intertidal coastal wetlands due to stronger lateral exchange by tides compared to lagoons (Fig. 1E).

In contrast to the lagoons, where DIC was positively correlated with pCO2 and water-air CO2 fluxes, in tidal systems and small deltas pCO2 and CO2 fluxes were not strongly associated with DIC concentrations; only CO2 fluxes in tidal systems showed a very weak trend with DIC concentration (Supplementary Results). Removing the effect of salinity in our analysis (as a co-variate), the differences in pCO2 and water-air CO2 flux correlations with DIC between estuary types suggests that other factors specific to small deltas and tidal systems further influence CO2 and DIC dynamics along the estuarine gradient in those systems. For example, shorter water residence times and increased intertidal wetlands are among factors driving processes impacting CO2 and DIC, which may include CO2 emissions of mangrove porewater DIC and mineralisation of exported POC and DOC55,59,60. The positive relationship between pCO2 and DOC (Supplementary Results and Supplementary Table 5) suggests that DOC mineralisation may drive increased pCO2 in small deltas. In tidal systems, stronger tidal influence compared to small deltas could further promote CO2 emissions through the tidal resuspension of sediments, releasing DIC and organic matter for remineralisation61,62,63, with excess DIC and organic matter exported to the coastal ocean58,64.

The sensitivity analysis showed that summer:winter water-air CO2 flux ratios ranged from 0.33 to 4.71 across the 13 estuaries (Supplementary Table 1). The largest range in summer:winter water-air flux CO2 ratios was for small deltas, where ratios were up to 3x larger than in lagoons and tidal systems (Supplementary Table 2). However, when estimating annual emissions, the large range in ratios in small deltas was attenuated by the large surface area of tidal systems compared to the small surface area coverage by small deltas (Table 3). Including mean winter water-air flux rates in annual Australian estuarine CO2 emission calculations only showed 13% greater water-air flux rate and 14% greater annual emissions than from summer measurements alone. Although our study accounts for pCO2 variations in seasonality extremes, it does not account for variations due to diurnal cycles and episodic events such as flooding. However, the diurnal effect on CO2 emissions from estuarine surface waters is likely minimal, with differences between day and night driven more by tidal influence65,66,67. Episodic events are more significant drivers of increased CO2 emissions in estuarine surface waters68, but quantifying the effect of these events on CO2 emissions in Australian estuaries was beyond the scope of this study.

The seasonal variability of water-air CO2 fluxes in Australian estuaries is consistent with other studies globally, showing a range of change between seasons. For example, mean CO2 water-air fluxes were highest in autumn (36.2 mmol C m−2 d−1) (mean water temperature: 11.5 °C) followed by spring (24.1 mmol C m−2 d−1) (7.5 °C) and summer (18.2 mmol C m−2 d−1) (17.5 °C), and lowest in winter (7.9 mmol C m−2 d−1) (2.9 °C) in the temperate Tay estuary (tidal system) in the United Kindom69. In contrast, significantly higher mean water-air CO2 fluxes were found in winter (15.6 ± 5.2 mmol C m−2 d−1) and summer (13.4 ± 22.2 mmol C m−2 d−1, highest discharge rate) than in spring (−13.7 ± 16.4 mmol C m−2 d−1) and autumn (2.7 ± 6.6 mmol C m−2 d−1) in the Delaware Estuary, USA (tidal system) (2015 water temperature range: 0.4 °C to 28.6 °C70, temperature data: In the current study, there was no correlation between water-air CO2 fluxes and temperature across all estuaries and within estuary types (studied estuary temperature range: 16 °C to 34.3 °C), except for a weak, significant correlation in tidal systems. This suggests that water-air CO2 fluxes in our study were likely driven by factors other than temperature, for example, riverine and lateral inputs into the estuaries or residence times.

Importantly, the seasonal variability between summer and winter water-air CO2 flux rates was small compared to the variability within individual estuaries and estuary types. Variability within the estuary types ranged from a maximum per-minute water-air CO2 flux rate 10 times (9 times as an estuary average minimum) larger than the minimum rate in the lagoons, 156 (113) times larger in the small deltas, and 165 (13) times larger in the tidal systems (Table 1). This larger within-estuary type spatial variability in water-air CO2 flux rates was captured in our continuous sampling along each estuary. We also accounted for the likely range of seasonal variability in water-air CO2 flux rates (maximum in summer, minimum in winter) by applying a summer:winter ratio to our summer data. As such, we argue that our annual emissions estimates are fairly robust.

Lagoons had the strongest disturbance signal, with pCO2 and water-air CO2 fluxes increasing with increasing disturbance (Figure 3A1 and B1), mainly driven by changes in the extent of seagrass cover. With increasing disturbance, NSW lagoons had a general decrease in seagrass cover (low = 55%, moderate = 19%, high = 24%, and very high = 3%)47 and mean dissolved oxygen (low = 123%sat, moderate = 97%sat, high = 95%sat, and very high = 108%sat) (Supplementary Fig. 4C2). Despite relying on data that was over 15 years old (mapped in 2007–2009)71,72, percent seagrass cover had a strong, negative association with pCO2 and a weaker, negative association with water-air CO2 fluxes (Supplementary Fig. 5 and Supplementary Results). These relationships suggest that higher pCO2 and water-air CO2 fluxes reflect decreased CO2 uptake by benthic vegetation (i.e. a less autotrophic system), as reported in several seagrass studies (e.g. refs. 43,44,73). High DOC concentrations in the low-disturbance lagoons were also consistent with DOC release from benthic vegetation, as observed in previous studies48,74,75 (Supplementary Fig. 4A2, Supplementary Table 5, and Supplementary Results). High percent O2 saturation in the very high disturbance lagoons (e.g. Curl Curl Lagoon; Supplementary Results, Supplementary Data 2, and Supplementary Table 4) was most likely due to a switch in production from benthic microalgae and macroalgae to phytoplankton76, enabling enhanced pCO2 drawdown and negative water-air CO2 flux rates (Figs. 3A1, B1).

This study estimates that Australia’s estuaries have a mean (±SE) annual area-weighted water-air CO2 emissions of 44.6 ± 6.4 mmol CO2-C m−2 d−1, which is 25% and 110% greater, respectively, than estimates of global means of 35.6 mmol CO2-C m−2 d−1 10 and 21.2 mmol CO2-C m−2 d−19. However, the role of estuary type in CO2 flux rates had a significant impact on our estimates of Australian water-air CO2 flux. Annual mean (±SE) area-weighted water-air CO2 fluxes of the lagoons upscaled to all Australian lagoons (13.4 ± 4.1 mmol CO2-C m−2 d−1) was 68% lower than from global lagoon estimates (41.4 mmol CO2-C m−2 d−1)10. In contrast, annual mean water-air CO2 fluxes (Table 2 and Fig. 4) scaled to all Australian small deltas (80.1 ± 11 mmol CO2-C m−2 d−1) and tidal systems (60.9 ± 11.8 mmol CO2-C m−2 d−1) were 99% and 22% higher, respectively, than from global small deltas (40.3 mmol CO2-C m−2 d−1) and tidal systems (49.9 mmol CO2-C m−2 d−1)10. In addition to the higher water-air CO2 fluxes in Australian small deltas and tidal systems, the lower global mean water-air CO2 flux for all estuaries combined compared to Australia likely also reflects the contribution of fjords and fjards to global estimates9,10. Globally, fjords and fjards have been shown to take up CO2 from the atmosphere (median 66 Tg CO2 yr−1)11, but are absent in Australia.

Fig. 4: Australian and global estuarine CO2 emissions by estuary type.
figure 4

Mean water-air CO2 fluxes in Australian estuaries (n: Lagoons = 21, Small deltas = 12, and Tidal systems = 14) and in global estuaries10 from the three estuary types defined in this study. Error bars represent standard errors. Source data are provided as a Source Data file.

Lower mean CO2 emissions in Australian lagoons compared to lagoons globally are likely due to overall lower disturbance in Australia. In addition, it may also reflect the greater abundance of ICOLLs in Australia (21% of global occurrence41,77,78). Isolation from marine waters, low riverine flow, and long residence times in ICOLLs may enhance autotrophic drawdown of CO2 by abundant seagrasses, resulting in smaller water-air CO2 fluxes (Table 2) than observed in non-Australian lagoons43,79,80. The weak hydrological connection between ICOLLs and rivers would also limit the input of CO2 supersaturated river water.

Higher mean water-air CO2 fluxes in Australian subtropical and tropical small deltas and tidal systems, compared to small deltas and tidal systems globally was an unexpected finding. Subtropical and tropical estuaries have previously been estimated to have lower water-air CO2 fluxes than systems at temperate latitudes9. We were unable to explicitly test for climate as we do not have sufficient estuaries of each geomorphic type and disturbance in each climate zone. However, macrotidal northern Australian small deltas and tidal systems are different from the small deltas and tidal systems in previous studies81 (Fig. 6 in Matthews and Matthews82) as North Australian tidal systems are dominated by extensive mangrove cover and have significantly greater tidal ranges (>4 m). Larger tidal ranges would lead to greater lateral inorganic and organic carbon export from mangroves to the tidal systems56,58,83. Higher mean water-air CO2 fluxes may also reflect the longer residence times resulting from characteristically low Australian freshwater inflows84,85. Long residence times would allow more time for CO2 produced from lateral inputs of DIC, DOC, and POC, and DIC from increased DOC and POC decomposition to be emitted across the water-air interface rather than flushed to the ocean.

Collectively, estuary geomorphic type is more important than disturbance in Australia, resulting in higher mean CO2 emissions from Australian estuaries despite their lower overall disturbance compared to global estuaries. The climate zone also has an important control on estuarine geomorphic type (e.g. tropical and subtropical mangrove-dominated macrotidal estuaries). This study suggests that relative to their surface area, Australian estuaries contribute a disproportionately large amount of CO2 emissions annually to global estuarine emissions. Using surface area estimates for Australian (62,100 km2; calculated from Table 3 in Chen et al.9) and global estuaries (1,012,440 km2) and global estuarine CO2 emission estimates by Laruelle et al.10 and Chen et al.9, Australian estuaries emit a mean (±SE) of 12.1 ± 1.7 Tg CO2-C annually. These emissions account for 12% or 8% of the estimated mean global estuarine CO2 emissions of 0.1 Pg CO2-C yr−19 or 0.15 Pg CO2-C yr−110, despite Australian estuaries accounting for only 6.1% of their calculated global estuarine surface area. This estimate includes the surface area coverage of the estuary types and disturbance groups and is dependent on the accuracy of surface areas for Australian and global estuaries. For instance, total estuarine water area of 39,390 km2 has been reported for Australia (Table 3)40, which is 57% greater than estimated by Dürr et al.2 (25,056 km2) and 37% smaller than estimated by Laruelle et al.10 (62,100 km2). Applying the estuarine surface areas of Australia’s National Land and Water Resources Audit (NLWRA)40 to data collected in the current study, Australian estuaries are estimated to emit (mean ± SE) 8.67 ± 0.54 Tg CO2-C annually (Table 2).

Australian tidal systems contributed the majority of the mean (±SE) annual CO2 emissions (8.18 ± 0.91 Tg CO2-C yr−1, 94.4%), with far smaller contributions from lagoons and small deltas (Table 2). Although lagoons in Australia (8.6% of total area) cover six times the estuarine surface area of small deltas (1.5% of total area;), CO2 emissions from lagoons were disproportionately low (0.27 ± 0.05 Tg CO2-C yr−1, 3.1% of Australian estuarine emissions) compared to small deltas (0.21 ± 0.03 Tg CO2-C yr−1, 2.5%) reflecting smaller water-air fluxes in lagoons (Table 2). The proportions of CO2 emitted by the different geomorphic types of estuaries in Australia were different from the proportions reported globally. For example, lagoons globally account for a larger proportion of CO2 emissions (31%; 0.046 Pg CO2-C yr−1) than small deltas (13%; 0.019 Pg CO2-C yr−1), and tidal systems only contribute 41% of global emissions (0.063 Pg CO2-C yr−1)10. The remaining proportion is made up of limited or non-filtering estuary types such as large rivers, karst-dominated coasts, and arheic coasts2. Furthermore, global CO2 emission estimates incorporate the contribution of fjords and fjards, which have the lowest water-air CO2 flux or show CO2 uptake but account for close to half of the global estuarine surface area10,11. Therefore, differences in Australian and global estuarine CO2 emissions are driven mostly by the geomorphic type (related to the climate zone of the estuary). This highlights the need to include geomorphic types in global CO2 emission assessments.

Geomorphology and disturbance influence water-air CO2 fluxes in Australian estuaries as a result of decreased hydrological connectivity in lagoons, and increased upstream riverine lateral inputs and tidal influence in small deltas and tidal systems. Water-air CO2 flux rates increase with higher disturbance, but geomorphology and disturbance interact, with the strongest disturbance signal in the lagoons, and a weak disturbance signal in the small deltas and tidal systems. Seasonal variations in CO2 emissions were a less important control on water-air CO2 fluxes in Australian estuaries. Previous global estuarine CO2 emission estimates have included geomorphology9,10,11, but not disturbance or the two factors together. CO2 emissions for global lagoons could therefore be over-estimated due to the bias towards more disturbed systems in the northern hemisphere. In contrast, CO2 emissions for global small deltas and tidal systems could be under-estimated due to the bias towards temperate systems in the northern hemisphere. As such, upscaling of global estuarine CO2 emissions should be based on geomorphic estuary-types but also consider land-use disturbance and climate and ideally, their interaction with geomorphic type.


In 36 estuaries around Australia, pCO2, DIC, DOC, physicochemistry, and physical parameters (wind speed, depth, current velocity, and barometric pressure) were measured along the salinity gradient from the marine to freshwater endmember (where possible). Data were combined with published CO2 fluxes and water quality data for 11 other Australian estuaries12,30, giving a total of 47 Australian estuaries. The same survey methods were used in all 47 estuaries. pCO2 and water-air CO2 fluxes were classified according to estuary type (lagoons, small deltas, and tidal systems) and disturbance group (low, moderate, high, and very high) and analysed for significant differences. Finally, the classified water-air CO2 fluxes were upscaled to all of Australia and mean estimates were compared to previous global mean estimates of estuary CO2 emissions. While we provide a full set of statistics in this study, we use the mean (±SE) for global comparison because our high-resolution water-air CO2 fluxes over a range of disturbance classes and geomorphic estuary types was better represented by the means than medians.

Estuary classification schemes

Estuaries were selected to cover a large range of disturbance and geomorphic types according to the classifications of NLWRA40 and Dürr et al.2. NLWRA40 assessed 971 Australian estuaries and described four disturbance classes (low (near-pristine), moderate (relatively unmodified), high (modified), and very high (extensively modified)). These disturbance groups were qualitatively classified based on changes in catchment land-use, estuary use, and ecology (Supplementary Table 6) and provided an assessment that was more relevant than adopting a single set of indicators. This is because the Australian continent covers a large surface area, encompassing over 1000 estuaries and climatic variations, making a single set of disturbance indicators likely misleading86,87. The global estuarine typology of Dürr et al.2 details three geomorphic types found in Australia: (1) lagoons (including Intermittently Closed or Open Lakes and Lagoons (ICOLLs) and estuaries with a central basin morphology), (2) small deltas, and (3) tidal systems (including drowned river valleys and tidal embayments), based around morphological and sedimentation characteristics driven by tidal influence (classification criteria in Supplementary Table 7). However, the existing classification of Australian estuaries2 did not match our observations of satellite imagery, because it was developed with a low spatial resolution (0.5°, or 50 km). Therefore, all 971 Australian estuaries were re-classified into the three estuary types by distinguishing physical characteristics based on the criteria of Dürr et al.2 (Supplementary Table 7) using satellite imagery (Google Earth) (Supplementary Data 3). Water surface area for 108 estuaries with missing surface area measurements were also calculated using satellite imagery (Google Earth). The re-classified estuary database was then combined with the estuarine disturbance database in NLWRA40 (dataset URL:, Supplementary Data 3).

The spatial distribution of estuary types in Australia corresponds to the tidal ranges of their respective coastlines (Fig. 5A). Tidal systems dominate the macro-tidal regions of northern Australia, whereas lagoons are found mostly in the micro-tidal regions of southern Australia. All three estuary types with all four disturbance groups, except for low and moderate disturbance in small deltas, were included in our estuary selection (Table 3). The surface area of estuaries sampled and included in this study represents 12.3% of the total Australian estuarine surface area, consisting of 19.1% of lagoons, 20.1% of small deltas, and 11.6% of tidal systems in Australia (Table 3).

Fig. 5: Distribution of estuary types and study locations in Australia.
figure 5

A Estuaries in Australia40 classified into three estuary types based on conceptual definitions (Supplementary Table 7) by Dürr et al.2 (©Google Earth) and the (B) location of study estuaries in Australia according to estuary type (shapes) and disturbance class (colours) (©OpenStreetMap, H.: Harbour, R.: River, and L.: Lake.

Study sites

Measurements from the 36 estuary surveys and from the 11 published estuary surveys were taken over the austral spring-summer season (Fig. 5B and Supplementary Data 1). The estuaries included in this study were comprised of 21 estuaries in New South Wales (Nov to Dec 2017), one in southeast Queensland (Moreton Bay, Oct 2018), seven along the north Australian coastline (from Karumba, Queensland to Wyndham, Western Australia, Oct to Dec 2018), seven along the south-west coastline of Western Australia (from Albany to Perth, Feb to Mar 2019), three in north-east Queensland30 (late spring Sept to Oct 2014) and eight in southeast Queensland12 (late spring Oct 2016) (Fig. 5B and Supplementary Data 1). Percent seagrass coverage for the New South Wales estuaries was obtained from Roper et al.47. Termeil Lake and Lake Brou were excluded from seagrass coverage analysis because although zero coverage was recorded by Roper et al.47, extensive seagrass cover was observed during our surveys.

Underway data measurements

Using a 6 m research vessel, physicochemical parameters and pCO2 were measured along a transect encompassing the length of each of the 36 estuaries, starting at the river mouth just after high-tide and ending in freshwater (salinity ~2). Although we aimed to finish the surveys at salinity ~2, this was not always possible because of shallow water and/or natural and artificial obstacles. As such, estuary data in this study reflect the spatial variations along the estuarine gradient (marine to upstream-riverine). A cruising ground speed of ~8 km h−1 was maintained where possible to ensure spatial and temporal consistency. The surveys were carried out during daylight hours, typically lasting over the course of a day. Surveys in large estuaries often required 2 to 3 days but never exceeded five days (Supplementary Data 1). pCO2 was recorded at one-minute intervals using an integrated water-gas loop setup (Supplementary Fig. 6).

Water was continuously pumped from beneath the hull (0.5 m to 1 m water depth) at ~1800 l−1 h−1 using a 12 V pump with backflow prevention (800GPH, Rule) to a high-flow filter basket (Ozito) before entering a two-way split. One path led to a flow-through chamber with a multi-parameter sonde (HL4, Hydrolab) measuring salinity (±0.5%), temperature (±0.1 °C), pHNBS (±0.2), and dissolved oxygen (DO%sat; ±2%). The second path entered a loop consisting of a pair of interconnected showerhead exchangers (RAD Aqua, Durridge) equilibrating dissolved gases in the water with the headspace. The dried gas stream was then measured for CO2 concentration, using a LiCOR 840 A CO2 gas analyser (accuracy <1%) and a Picarro G-2508 Cavity Ring-Down Spectrometer (CRDS, ±0.05%)87. In the ICOLL lagoons (indicated in Supplementary Data 1) where a smaller boat was used, CO2 measurements were only taken with the LiCOR 840 A CO2 gas analyser. Measured CO2 was humidity-corrected and in-situ pCO2 was calculated using methods in Pierrot et al.88. The LiCOR 840 A was calibrated using a two-step process with low (250 ppm) and high (8000 ppm) pCO2 gas standards. The CRDS was serviced and calibrated by the manufacturer (Picarro, USA) before each field trip.

Discrete water samples, morphological, and meteorological data

Water samples were collected for DIC and DOC concentrations, along with estuarine (depth and water current velocity) and meteorological measurements at the start and end of survey transects and at salinity intervals of ~5. In cases where salinity did not change much (<5) along the survey, samples were collected every hour instead (i.e. every 8 km of estuary travelled). In the ICOLL lagoons (as indicated in Supplementary Data 1) where no significant salinity gradient was present, discrete water samples were collected from at least 3 points across the estuary. For DOC, 30 ml water samples were filtered through a pre-combusted (500 °C, ~5 h) 0.7 µm GF/F filter (Whatman, Merck) into an acid-washed glass vial containing 100 µl of 85% phosphoric acid (H3PO4). 50 ml water samples for DIC analysis were syringe-filtered (0.45 µm SFCA Minisart, Sartorius) into a crimp-top glass bottle without any headspace and preserved with 30 µl mercuric chloride (HgCl2). DOC concentrations were determined using a total organic carbon analyser (±3%; 1030 W, Aurora)89. DIC concentrations were analysed with a Marianda AIRICA coupled to a CO2/H2O analyser (LI7000, LiCOR), calibrated for accuracy with certified reference material90 at a typical precision of better than 2 µmol kg−1 91. All samples were processed immediately, stored on ice while the survey was underway, and frozen (−20 °C) as soon as possible (typically within five hours of collection) except for DIC, which was stored at room temperature. On the main research vessel and the smaller boat, water current velocity was measured using a differential GPS-assisted Lagrangian method with a neutrally-buoyant drifter (adapted from Wetzel and Likens92). Current velocity measurements likely indicated flow rates of the ebbing tide, as the surveys were carried out after the turn of the high tide. Water depth on the main research boat was measured using a hull-mounted acoustic transducer (Airmar), while water depth was measured using a lead and line on the smaller boat or taken from Roper et al.47. Barometric pressure (±0.5hPa @20 °C), air temperature (±1.1 °C @20 °C), and true wind speed (±5% @10 m s−1) were measured 3 m above the water surface using a vessel-mounted weather station (200WX, Airmar). In the ICOLL lagoons, daily averaged meteorological data were obtained from the closest Bureau of Meteorology (BOM) weather station (Climate Data Online93).

Water-air CO2 flux calculations

Water-air CO2 flux (FCO2; mmol CO2-C m−2 d−1) was calculated at 1-minute intervals using Eq. (1):


where k600 is the gas transfer velocity (m d−1), K0 is the solubility coefficient of CO2 (mol l−1 atm−1), and Cwater and Cair are the partial pressure of CO2 (µatm) in water and air, respectively94. The formula from Weiss95 was used to obtain CO2 solubility coefficients based on salinity and temperature Eq. (2):


where K0 is expressed in moles L−1 atm−1, A1 (-58.0931), A2 (90.5069), A3 (22.2940), B1 (0.027766), B2 (-0.025888), and B3 (0.0050578) are constants, T is absolute temperature, and S‰ is the salinity. CO2 atmospheric concentration was assumed to be 407 µatm96, which was the mean concentration in 2018. Although k600 is a significant variable required for calculating water-air fluxes, measuring k600 in-situ was not feasible due to the large spatial coverage of this study. As such, five empirical k600-models for a range of coastal-marine ecosystems were used from the literature to estimate mean k600 (equations (6) to (10) listed in Table 4), i.e., mangrove-dominated28,97 and tidal27 (using wind speed, water depth, and current velocity), lagoonal53 (using wind speed and water depth), and marine-dominated94 (using wind speed only) coastal ecosystems. Windspeed is corrected for a height of 10 m (U10) by rearranging the formula from Amorocho and DeVries98:


where Uz is the measured windspeed at z height (3 m) in m s−1, C10 is the surface drag coefficient for wind at 10 m (1.3 × 10−3)97, and κ is the Von Karman constant (0.41). In the first four parameterisations (Eqs. (6) to (9) in Table 4), k600 is the gas transfer velocity (cm h−1) normalised to a Schmidt number of 600. The parameterisation in equation 10 (Table 4) by Wanninkhof94 calculated k at the Schmidt number (Sc) of the measured temperature and salinity, converted to k600 using Eq. (4):

Table 4 Gas transfer velocity normalised to Schmidt number of 600 (k600) parameterisations using various methods in published literature

A Schmidt exponent of -0.5 was used to account for higher water turbulences associated with tidal currents in estuaries99. To calculate water-air CO2 fluxes, k was derived from k600, which was calculated using the other four parameterisations (Eqs. (6) to (9) in Table 4) by rearranging Eq. (4). Sc at the measured temperature and salinity was calculated using the formula in Wanninkhof94 (Eq. (5)):


where A, B, C, D, and E are constants for CO2 in fresh water (1923.6, −125.06, 4.3773, −0.085681, and 0.00070284, respectively) and seawater (2116.8, −136.25, 4.7353, −0.092307, 0.0007555, respectively), and t is temperature in °C. A salinity factor was calculated from the difference between freshwater and sea water Sc and applied to calculate Sc at the measured salinity.

To ensure consistency between water-air CO2 fluxes measured for estuaries in this study and those previously reported for eight southeast Queensland estuaries12, water-air CO2 fluxes were recalculated using the five parameterisations. Water-air CO2 fluxes from the previously reported three north Queensland estuaries30 were not recalculated because water depth and current velocity data were unavailable. However, given that these three estuaries were each categorised in a different disturbance group and/or estuary type (moderate and high disturbance tidal system, and a high disturbance small delta, Table 3), this should not introduce any systematic bias.

Data processing and statistical analysis

Per-minute pCO2 and water-air CO2 flux were averaged to 5-minute datapoints to reduce the number of data points whilst maintaining the high resolution and main features of the dataset. Kolmogorov-Smirnov and Levene’s tests for normality and homoscedasticity, respectively, returned significant results (p < 0.05), ruling out parametric methods for statistical analysis. Consequently, significant differences (α = 0.05) between and within estuary types (3 factors) and disturbance groups (4 factors) were tested using the Permutational Multivariate Analysis of Variance (PERMANOVA) procedure with Euclidean distance as the dissimilarly matrix in Primer v7 and PERMANOVA+ add-on (PRIMER-e). The dataset for PERMANOVA analysis was normalised (z-score) but not power-transformed to retain the heterogeneity between the mean and variances, retaining the spatial scale along the estuarine transect. Not power-transforming the data reduces the possibility of an inflated type I error100. Salinity was included as a covariate in the pCO2, water-air CO2 flux, DOC and DIC analyses to remove differences influenced by salinity. The focus of PERMANOVA is on the differences between the data points rather than descriptive statistics (mean, median, etc.). PERMANOVA can therefore, identify significant differences in datasets even where there are similarities in the descriptive statistics. 9999 permutations were performed using residuals under a reduced model using type I sum of squares. Significant results were further analysed with pairwise PERMANOVA. pCO2 and water-air CO2 fluxes were analysed for correlations with salinity using Pearson’s correlation and combined with physicochemical data, DOC, DIC, and percent seagrass cover to analyse for partial correlations while controlling for the effect of salinity (α = 0.05) in SPSS v25 (IBM). Data used for correlation analysis were power-transformed where necessary and normalised (z-score). Partial correlation analysis was chosen over multivariate methods such as Principal Component Analysis (PCA) because targeted testing for correlations between variables and CO2 was more useful than an exploratory approach.

CO2 emission upscaling to the Australian continent

Published summer and winter water-air CO2 flux rates were available for 13 Australian estuaries, including each of the three geomorphic estuary types (Supplementary Table 1). Of these, 10 of the estuaries were included in this study12,30, along with an additional three from a published study43. The summer and winter water-air CO2 fluxes were used to calculate a summer:winter water-air CO2 flux ratio (mean and range) for each of the three estuary types (Supplementary Table 2). Ratios for each estuary type were then applied to the measured summer water-air CO2 fluxes for the 47 estuaries to estimate the mean and range of winter water-air CO2 fluxes (Table 2). The summer and winter mean water-air CO2 flux rates from each estuary were averaged together to derive the annual water-air CO2 flux rates and emissions for the 47 estuaries. To gauge the sensitivity of annual Australian emissions to winter flux rates, the summer mean flux rates were also adjusted up and down by the minimum and maximum of winter:summer ratios (Supplementary Table 2). Upscaled fluxes determined based on these minimum and maximum ratios allow an upper and lower limit to be placed on these estimates.

Annual CO2 emissions from the 47 estuaries were upscaled to all Australian estuaries (n = 971) by multiplying the estuary type-specific and disturbance-specific water-air CO2 fluxes (mmol CO2-C m−2 d−1) by the total estuarine surface area of the relevant systems40 (Table 3). Small deltas with low to moderate disturbance were not available for this study. However, even though measured low and moderately disturbed small delta water-air CO2 fluxes were likely different from the mean high and very high disturbed small delta water-air CO2 flux, their impact on total Australian estuary emissions is low. This is because small deltas only make up 1.5% of Australia’s estuarine surface area (Table 3).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.