Soil greenhouse gas fluxes from tropical coastal wetlands and alternative agricultural land uses

. Coastal wetlands are essential for regulating the global carbon budget through soil carbon sequestration and greenhouse gas fluxes (GHG: CO 2 , CH 4 and N 2 O). The conversion of coastal wetlands to agricultural land alters the magnitude and direction (uptake/release) of these fluxes. However, the extent and drivers of change of GHG fluxes is still unknown for many tropical regions. We measured soil GHG fluxes from three natural coastal wetlands: mangroves, saltmarsh, and 15 freshwater tidal forests, and two alternative agricultural land use, sugarcane farming and pastures for cattle grazing (ponded and dry conditions). We assessed variations throughout different climatic conditions (dry-cool, dry-hot and wet-hot) within two years of measurements (2018-2020) in tropical Australia. The wet pasture had by far the highest CH 4 emissions with 1,231 ± 386 mg m -2 d -1 , which were 200-fold higher than any other site. Dry pastures and sugarcane were the highest emitters of N 2 O with 55 ± 9 mg m -2 d -1 (wet-hot period) and 11 ± 3 mg m -2 d -1 (hot-dry period, coinciding with fertilisation), respectively. 20 Dry pastures were also the highest emitters of CO 2 with 20 ± 1 g m -2 d -1 (wet-hot period). Comparatively, the three coastal wetlands measured had lower emission, with saltmarsh up taking -0.55 ± 0.23 of N 2 O and -1.19 ± 0.08 g m -2 d -1 of CO 2 during the dry-hot period. During the sampled period, sugarcane and pastures had higher total cumulative soil GHG emissions (CH 4 + N 2 O) of 7,142 and 56,124 CO 2eq kg ha -1 y -1 compared to coastal wetlands with 144 to 884 CO 2-eq kg ha -1 y -1 . Converting unproductive sugarcane land or pastures (especially ponded ones) to coastal wetlands could provide significant GHG 25 mitigation.

will inform emission factors for converting wetlands to agricultural land uses and vice versa, filling in a knowledge gap identified in Australia (Baldock et al., 2012) and tropical regions worldwide (IPCC, 2013). 60

Study sites
The study area is located within the Herbert River catchment in Queensland, northeast Australia (Fig 1a). The region has a tropical climate with a mean monthly minimum temperature from 14 to 23˚C and mean monthly maximum temperature from 25 to 33˚C (Australian Bureau of Meteorology, ABM, 2020;1968 Table S3). The average rainfall is 2,158 mm y -1, with 65 the highest values of 476 mm during February (ABM 2020;1968 Table S3). The Herbert basin covers 9,842 km 2 , from which 56% is grazing, 31% is conserved wetlands and forestry, 8% is sugarcane, and 4% is other land uses (Department of Science and Environment, QLD, DES, WetlandInfo, 2020). Wetlands in this region were heavily deforested in the past century  due to rapid agricultural development, primarily for sugarcane farming (Griggs. 2018). Before clearing, the land was mostly covered by rainforest and coastal wetlands, mainly Melaleuca forest, grass and sedge swamps (Johnson, Ebert, & 70 Murray, 1999).
We selected five sites, including three natural coastal wetlands (Fig. 1): a mangrove forest (18º 53' 42ʺ S, 146º 15' 51ʺ E), a saltmarsh (18º 53' 43ʺ S, 146º 15' 52ʺ E) and a freshwater tidal forest (18º 53' 45ʺ S, 146º 15' 52ʺ E), and two common agricultural land use types of the region, a sugarcane farm (18º 53' 44.6ʺ S, 146º 15' 53.2ʺ E) and a pasture for fodder 75 grazing. The pasture had different levels of inundation; some areas were covered with shallow ponds (50-100 cm depth), wet grassy areas (hereafter "wet pasture"; '18º 43' 8ʺ S, 146º 15' 50ʺ E) and drier areas (hereafter, "dry pasture"; 18º 43' 7ʺ S, 146º 15' 50ʺ E). The natural coastal wetlands and the sugarcane site were located within the same property at Insulator Creek, while the ponded pasture was 20 km north at Mungalla Station. The mangroves were dominated by Avicennia marina with few plants of Rhizophora stylosa, and the saltmarsh was dominated by Sueda salsa and Sporobolus spp. Landwards, the 80 freshwater tidal freshwater forest, a wetland commonly known as "tea tree swamp", was dominated by Melaleuca quinquenervia trees. While the mangroves and saltmarsh are directly submerged by tides (5-30 cm), the tidal freshwater forest is indirectly affected by tidal fluctuations, such as during large spring tides, when tidal water can push groundwater above the forest. The coastal wetlands were adjacent to a sugarcane farm with an area of ~110 ha (Fig. 1b). The sugarcane is fertilised once a year with urea at a rate of 135 kg N ha -1 and harvested during May-June, while the foliage is left on the soil surface 85 (trash blanket) after harvest. The ponded pastures in Mungalla Station extend over 2,500 ha and support ~900 cattle throughout the year by providing fodder to cattle during dry periods. The selected ponded pastures were covered by Eichhornia crassipes (water hyacinth) and Hymenachnae amplexicaulis (Fig. 1g-h). Each of the five sites was sampled during three periods drycool (May-September), dry-hot (October-December) and wet-hot (January-April; Table 1). During each time, soil physicochemical properties and GHG fluxes were measured as detailed below. 90  Soil texture analysis (% sand, silt, clay) was done with a simplified method for particle size determination (Kettler et al, 2001). Soil electrical conductivity (EC) and pH were measured using a conductivity meter (WP-84 TPS, Australia) in soil/water slurry at 1:5. Soil subsamples were air-dried, sieved (2mm), ground (Retch™ mill) and analysed for %N and %C using an elemental analyser connected to a gas-isotope ratio mass spectrometer (EA-Delta V Advantage IRMS, Griffith 115 University). Additionally, soil samples from the top 10 cm were collected during each sampling to measure gravimetric soil moisture content and bulk density.

Greenhouse gas fluxes
We measured GHG fluxes (CO2, CH4 and N2O) at each site for three consecutive days during each sampling period except for 120 the dry-cool period of 2018, when mangroves, saltmarsh and sugarcane were surveyed for one day. The sampling was done between 9:00 to 11:00 am, representing the mean daily temperatures, thus, minimising variability of cumulative seasonal fluxes based on intermittent manual flux measurements . Additionally, we assessed the variability of our measurements with tidal inundation in mangroves and saltmarsh, which were regularly inundated (~10-30 cm). For this, we measured GHG emissions during a low (0.7m) and a high tide (2.8m; Lucinda, 18° 31' S; 146° 23 'E) in the dry-cool period 125 of 2019. We found that CH4 fluxes did not significantly vary between the low and high tide within all coastal wetlands.
We used static, manual gas chambers made of high-density, round polyvinyl chloride pipe, which consisted of two units: a base (r =12 cm, h =18 cm) and a detachable collar (h =12 cm;Hutchinson and Mosier, 1981;Kavehei et al, 202). The chambers had lateral holes that could be left covered with rubber bungs at low water levels and left open at high water levels 135 to allow for water movement between sampling events. When the wetlands were inundated for the experiments, we used PVC extensions (h = 18 cm). Five chambers were set ~ 5cm deep in the soil at random locations one day before sampling to minimise the disturbance of installation during the experiment (Rashti et al, 2015). The chambers were selectively located on soil with minimal vegetation, roots, and crab burrows. We were careful not to tramp around the chambers during installation and sampling. The fact that emissions were not significantly different among days (p >0.05) provided us with confidence that 140 disturbance due to installation was not problematic.
At the start of the experiment, gas chambers were closed. A sample was taken at time zero and then after one hour with a 20 ml syringe and transferred to a 12 mL-vacuumed exetainer (Exetainer, Labco Ltd., High Wycombe, UK). During the dry-hot season, linearity tests of GHG fluxes with time were conducted by sampling at 0, 20, 40 and 60 min (Rashti et al, 145 2016). For the rest of the experiments, linearity tests were performed in one of the five chambers at each site; R 2 values were consistently above 0.70. During each experiment, soil temperature was measured next to each chamber. At the end of the experiment, the depth of the base was recorded from five points within each chamber to calculate the headspace volume. The obtained volumetric unit concentrations were converted to mass-based units using the Ideal Gas Law (Hutchinson and Mosier, 1981). 150 The GHG concentrations of all samples were analysed within two weeks of sampling with a gas chromatograph (Shimadzu GC-2010 Plus). For N2O analysis, an electron capture detector was used with helium as the carrier gas, while CH4 was analysed on a flame ionisation detector with nitrogen as the carrier gas. For CO2 determination, the gas chromatograph was equipped with a thermal conductivity detector. Peak areas of the samples were compared against standard curves to 8 determine concentrations (Chen et al, 2012). Seasonal cumulative GHG fluxes were calculated by modifying the equation described by Shaaban et al. (2015;Eq. 2):

Equation 2
Where; 160 Ri = Gas emission rate (mg m -2 hr -1 for CO2 and μg m -2 hr -1 for CH4 and N2O), Di = number of the sampling days in a season, 17.38 = number of weeks in each period, assuming these conditions were representative of the annual cycle (see Table 1).
Annual cumulative soil GHG fluxes (CH4 + N2O) were calculated by integrating cumulative seasonal fluxes. These estimations did not account for soil CO2 values as our methodology with dark chambers only accounted for emissions from 165 respiration and excluded uptake from primary productivity. The CO2-equivalent (CO2-eq) values were estimated by multiplying CH4 and N2O emissions by 25 and 298, respectively (Solomon, 2007), which represent the radiative balance of these gases (Neubauer, 2021).

Statistical analyses 170
GHG flux data were tested for normality through Kolmogorov-Smirnov and Shapiro-Wilk tests. The data was then analysed for spatial and temporal differences with one-way Analyses of Variance (ANOVA), where site and season were the predictive factors and the replicate (chamber) was the random factor of the model. When data were not normal, they were transformed (log10 or 1/x) to comply with the assumptions of normality and homogeneity of variances. Some variables were not normally distributed despite transformations and were analysed with the non-parametric Kruskal-Wallis test and Mann-Whitney U Test. 175 A Pearson correlation test was run to evaluate the correlation of GHG with measured environmental factors. Analyses were done with SPSS (v25, IBM, New York, USA), and values are presented as mean ± standard error (SE).

Soil physicochemical properties
Soil physical and chemical parameters (mean values 0-30 cm) varied among sites (Table 2, see full results of statistical analyses 180 in Supplementary Material). As expected, gravimetric moisture content was highest in the coastal wetlands and wet pasture (> 26%) and lowest in the sugarcane and the dry pasture (< 14%). All soils were acidic, especially the freshwater tidal forest and the wet pastures with values < 5 throughout the sediment column; mangroves had the highest pH with 6.0 ± 0.1. The lowest EC was recorded in the pastures (247 ± 38 and 190 ± 39 µS cm -1 for the dry and wet pasture, respectively), and highest in the three natural coastal wetlands with 1,418 ± 104, 8,049 ± 276 and 8,930 ± 790 µS cm -1 for tidal freshwater wetland, saltmarsh 185 and mangroves, respectively.
Soil bulk density was highest in sugarcane (1.5 ± 0.1 g cm -3 ) and lowest in the freshwater tidal wetland (0.6 ± 0.1 g cm -3 ). For all sites, %C was highest in the top 10 cm of the soil and decreased with depth, with highest values in the freshwater tidal wetland (5.1 ± 0.6%) and lowest in the saltmarsh (1.2 ± 0.1%). Soil %N ranged from 0.1 ± 0.0 to 0.4 ± 0.1% at all sites, except in the freshwater tidal wetland, where it reached values of 0.6 ± 0.0% in the top 10 cm (Table 2). 190

Greenhouse gas fluxes
Soil emissions for CO2 were significantly different among sites and times of the year (t =155.09, n =237, p < 0.001; Fig. 2a).
The highest CO2 emissions were measured during the wet-hot period in the dry pasture, where values reached 20.31 ± 1.95 g m -2 d -1 while the lowest values were measured in the saltmarsh, the only site that acted as a sink of CO2 with an uptake rate of -0.59 ± 0.15 g m -2 d -1 . In the pastures, CO2 emissions were twice as high when dry with cumulative annual emissions of 5,748 195 ± 303 g m -2 y -1 compared to when wet, with 2,163 ± 465 g m -2 y -1 . For the coastal wetlands, cumulative annual CO2 emissions were highest in freshwater tidal forests with 2,213 ± 284 g m -2 y -1 , followed by mangroves with 1,493 ± 111 g m -2 y -1 and lowest at the saltmarsh with uptake rates of -264 ± 29 g m -2 y -1 .
The wet pasture had the highest total cumulative soil GHG emissions (CH4 + N2O) with 56,124 CO2eq kg ha -1 y -1 followed by dry pasture 23,890 CO2eq kg ha -1 y -1 and sugarcane 7,142 CO2eq kg ha -1 y -1 . While coastal wetlands had mangroves and freshwater tidal forests, respectively. Overall, the three coastal wetlands measured in this study had lower total cumulative GHG emissions at 1,263 CO2-eq kg ha -1 yr -1 compared to the alternate agricultural land uses, which emitted 87,156 CO2-eq kg ha -1 yr -1 .

Greenhouse gas emissions and environmental factors 220
Overall, we found that not one single parameter measured in this could explain GHG fluxes for all sites except land-use. The   In this study, we found that the three coastal tropical wetlands measured in this study (mangroves, saltmarshes and freshwater tidal forests) had significantly lower GHG emissions compared to two alternative land uses common in tropical 240 Australia (sugar cane and grazing pastures). Notably, we found that coastal wetlands had 200 times lower CH4 emissions and seven times lower N2O compared to wet pastures and sugarcane soils, respectively. While future studies should measure GHG from other wetlands, land uses, and within other tropical regions, these results support the idea that the management or conversion of unused agricultural land could be converted to coastal wetlands could result in significant GHG mitigation.

245
The variability of GHG fluxes was best explained by land use and wetland type; however, some trends with seasons were evident. For instance, CO2 and N2O emissions were lowest during the dry-cool periods. Reduced emissions at low temperatures are expected as the temperature is a main driver of any metabolic process, including respiration and nitrificationdenitrification. Mangroves tend to have higher CO2 emissions as temperature increases (Liu and Lai 2019), and terrestrial forests have significantly higher N2O emissions during warm seasons (Schindlbacher et al, 2004). Emissions of CH4 also tend 250 to increase with temperature as the activity of soil methane-producing microbes (Ding et al, 2004) and the availability of carbon is higher in warmer conditions (Yvon-Durocher et al, 2011). However, as most of the studies on GHG fluxes, were conducted in temperate and subtropical locations where differences in temperature throughout the year are much larger than those in tropical regions. For tropical regions, increased GHG emissions are likely to be strongly affected by the "Birch effect", which refers to short-term but a substantial increase of respiration from soils under the effect of precipitation during the early 255 wet season (Fernandez-Bou et al, 2020).
The main factor associated with GHG fluxes was land use and type of wetland. Notably, coastal wetlands, even the freshwater tidal forests, had much lower emissions compared to the wet pastures. This large difference could be attributed to the presence of terminal electron acceptors in the soils (e.g. iron, sulphate, manganese) of the coastal wetlands, which could 260 inhibit methanogenesis (Kögel-Knabner et al, 2010;Sahrawat, 2004). Sulphate reducing bacteria are also likely to outcompete methane-producing bacteria (methanogens) in the presence of high sulphate concentrations in tidal wetlands, resulting in low CH4 production. Competition between methanogens and methanotrophs may result in a net balance of low CH4 production despite freshwater conditions (Maietta et al. 2020). Additionally, microorganism living within the bark of Melaleuca trees can consume CH4 (Jeffrey et al, 2021), so it is possible that similar bacteria within the soil could reduce CH4 emissions. 265 Interestingly, variability within CH4 fluxes among sites was very high, despite them being very close to each other (Fig. 1b).
These differences highlight the importance of land use in GHG fluxes, which are likely to significantly alter the microbial community composition and abundance, which can change rapidly over small spatial scales (Martiny et al, 2006;Drenovskyet al, 2009).
Our results are consistent with other studies, which have shown the importance of land use in GHG emissions. For instance, in a Mediterranean climate, drained agricultural land use types, pasture and corn, were larger CO2 emitters compared to restored wetlands (Knox et al. 2015). Clearing of wetlands for agricultural development, such as the drainage of peatlands, results in very high CO2 emissions (Nieveen et al, 2005;Veenendaal et al, 2007;Hirano et al, 2012), and restoration of these wetlands could decrease these emissions (Cameron et al, 2020). Additionally, some of the wetland types, such as marshes, 275 were occasional sinks of CO2 and CH4, consistent with previous studies where intertidal wetlands sink of GHG at least under some conditions or during some times of the year (Knox et al, 2015;Maher et al, 2016).
The fluxes measured in the coastal wetlands of this study (-1,191 to 10,970 mg m -2 d -1 for CO2, -0.2 to 3.9 mg m -2 d -1 for CH4, and -0.2 to 2.8 mg m -2 d -1 for N2O) are within the range of those measured in other wetlands, worldwide. For CO2, 280 fluxes can range between -139 and 22,000 mg m -2 d -1 (Stadmark and Leonardson 2005;Morse et al. 2012), for CH4, from -1 to 418 mg m -2 d -1 (Allen et al. 2007;Mitsch et al 2013;Cabezas et al. 2018), and for N2O, from -0.3 to 3.9 mg m -2 d -1 (Hernandez and Mitsch 2006;Morse et al. 2012). Despite being in tropical regions, the fluxes from this study were not notably higher compared to wetlands in other climates. The general lower nitrogen pollution in Australia's soils and waterways compared to other countries may partially explain the lower emissions. However, the GHG flux measurements from this study did not 285 account for the effects of vegetation, which can alter fluxes. For instance, some plant species of rice paddies (Timilsina et al., 2020) and Miscanthus sinensis (Lenhart et al., 2019) can increase N2O emissions, and some tree species can facilitate CH4 efflux from the soil (Pangala et al. 2013). Finally, changes in emissions between low and high tides were detected for CO2 and N2O. Thus, future studies that include vegetation and changes within tidal cycles will improve GHG flux estimates for coastal wetlands. 290

Management implications
Under the Paris Agreement, Australia has committed to reducing GHG emissions 26 -28% below its 2005 levels by 2030.
With annual emissions of 153 million tonnes of carbon dioxide equivalent (Mt CO2-eq y -1 ), Queensland is a major GHG emitter in Australia (~ 28.7% of the total in 2016; DES, 2016). Of these emissions, about 18.3 Mt CO2-eq y -1 (14%) are attributed to 295 agriculture, while land-use change and forestry emit 12.1 Mt CO2-eq y -1 (DES, 2016). Production of CH4 from ruminant animals, primarily cattle, contribute 82% of agriculture-related emissions (DES, 2016). Therefore, any GHG mitigation strategy from land-use change could be important for Australia to achieve its national goals.
This study supports the application of three management actions that could reduce GHG emissions. First, the 300 conversion of ponded pastures to coastal wetlands is likely to reduce soil GHG emissions. Our results showed that wet pastures emit 56 ton CO2-eq ha -1 y -1 of total GHG (CH4 + N2O) compared with 0.2 ton CO2-eq ha -1 y -1 , 0.1 ton CO2-eq ha -1 y -1 and 0.9 ton CO2-eq ha -1 y -1 from mangroves, freshwater tidal forest, and saltmarshes, respectively. This implies that about 55 ton CO2-eq ha -1 y -1 emissions from the soils could be potentially avoided by converting wet pastures to coastal wetlands. The carbon mitigation for GHG emissions from soil solely could provide ~ AUD 860 ha -1 yr -1, assuming a carbon value of AUD 15.37 per 305 ton of CO2-eq (Australian Government Clean Energy Regulator, 2018). This mitigation could be added up to the carbon sequestration through sediment accumulation and tree growth that results from wetland restoration. Legal enablers in Queensland are in place to manage unproductive agricultural land this way (Bell-James and Lovelock 2019), and could provide an alternative income source for farmers.

310
A second management option would be to reduce the time pastures are kept under water. Dry pastures produced significantly less CH4 with ~0.005 kg ha -1 d -1 than wet pastures with 6 kg ha -1 d -1 . For comparison, an average cow produces 141 g CH4 d -1 (McGinn et al, 2004), and our study area supported around 900 cattle over 2,500 ha throughout the year, equivalent to 19 kg ha -1 y -1 compared to 2 kg ha -1 y -1 and 2090 kg ha -1 y -1 CH4 from dry and wet pasture respectively. This 315 implies that nearly 99% of the CH4 emissions came from wet pastures, while dry pasture and grazing cattle had a low share in total CH4 emissions. Therefore, land use management of wet pastures which are used to feed grazing cattle in Queensland may be a significant opportunity to reduce agriculture-related CH4 emissions. Future studies should increase the number of sites of ponded pastures to account for variability in hydrology, fertilisation, and cattle use. However, the very high difference (2-3 orders of magnitude) between dry and ponded pastures provides confidence that pasture management could provide significant 320 GHG mitigation throughout the year.
Finally, fertiliser management in sugarcane could reduce N2O emissions. Higher N2O emissions of 17.63 mg m -2 d -1 were measured in sugarcane following fertilisation during the dry-hot season. Comparatively, natural wetlands had low N2O emissions (0.16 to 2.79 mg m -2 d -1 ); even the saltmarsh was an occasional sink. Thus, improved management of fertiliser 325 applications could result in GHG emission mitigation. Some activities include split application of nitrogen fertiliser in combination with low irrigation, reduction in fertiliser application rates, the substitution of nitrate-based fertiliser for urea (Rashti et al, 2015), removing mulch layer before fertiliser application (Pinheiro et al, 2019;Xu et al, 2019 Zaehle andDalmonech, 2011) or conversion of unproductive sugarcane to coastal wetlands.

Conclusion 330
The GHG emissions from three coastal wetlands in tropical Australia (mangroves, saltmarsh and freshwater tidal forests) were consistently lower than those from two common agricultural land use of the region (sugarcane and pastures) throughout three climatic conditions (dry-cool, dry-hot and wet-hot). Ponded pastures, which emitted 200 times more CH4, and sugarcane emitted seven times more than any natural coastal wetland. If these high emissions are persistent in other locations and within other tropical regions, conversion of pastures and sugarcane to similar coastal wetlands could provide significant GHG 335 mitigation. As nations try to reach their emission reduction targets, projects aimed at converting or restoring coastal wetland can financially benefit farmers and provide additional co-benefits derived from coastal wetland restoration.

Competing interests
The authors declare that they have no conflict of interest.