High spatiotemporal variability of methane concentrations challenges estimates of emissions across vegetated coastal ecosystems

Abstract Coastal methane (CH4) emissions dominate the global ocean CH4 budget and can offset the “blue carbon” storage capacity of vegetated coastal ecosystems. However, current estimates lack systematic, high‐resolution, and long‐term data from these intrinsically heterogeneous environments, making coastal budgets sensitive to statistical assumptions and uncertainties. Using continuous CH4 concentrations, δ13C‐CH4 values, and CH4 sea–air fluxes across four seasons in three globally pervasive coastal habitats, we show that the CH4 distribution is spatially patchy over meter‐scales and highly variable in time. Areas with mixed vegetation, macroalgae, and their surrounding sediments exhibited a spatiotemporal variability of surface water CH4 concentrations ranging two orders of magnitude (i.e., 6–460 nM CH4) with habitat‐specific seasonal and diurnal patterns. We observed (1) δ13C‐CH4 signatures that revealed habitat‐specific CH4 production and consumption pathways, (2) daily peak concentration events that could change >100% within hours across all habitats, and (3) a high thermal sensitivity of the CH4 distribution signified by apparent activation energies of ~1 eV that drove seasonal changes. Bootstrapping simulations show that scaling the CH4 distribution from few samples involves large errors, and that ~50 concentration samples per day are needed to resolve the scale and drivers of the natural variability and improve the certainty of flux calculations by up to 70%. Finally, we identify northern temperate coastal habitats with mixed vegetation and macroalgae as understudied but seasonally relevant atmospheric CH4 sources (i.e., releasing ≥ 100 μmol CH4 m−2 day−1 in summer). Due to the large spatial and temporal heterogeneity of coastal environments, high‐resolution measurements will improve the reliability of CH4 estimates and confine the habitat‐specific contribution to regional and global CH4 budgets.


| INTRODUC TI ON
Methane (CH 4 ) is the second most important greenhouse gas (GHG) driving global climate change (Shindell et al., 2009). Past research has shown that coastal marine environments dominate the global ocean CH 4 budget and contribute 5-28 Tg CH 4 yr −1 to total global CH 4 emissions (Rosentreter et al., 2021b;Weber et al., 2019). However, a scarcity of systematic, high-resolution, and long-term measurements has hampered our ability to confine CH 4 emissions from a wide range of heterogeneous and dynamic coastal environments impeding efforts to evaluate the potential of coastal ecosystems to mitigate climate change by storing carbon (Rosentreter et al., 2021a).
Particularly in coastal sediments, CH 4 can be produced in large amounts due to the organic carbon surplus of primary production from submerged (e.g., seagrass and macroalgae) and partially emerged (e.g., mangroves and salt marshes) vegetation (Duarte et al., 2005;Ortega et al., 2019;Snelgrove et al., 2018) and the accumulation of allochthonous particulate organic matter (Barnes & Goldberg, 1976;Reeburgh, 1983;Wallenius et al., 2021). In such environments, CH 4 emissions can offset or even negate the value of coastal ecosystems as "blue carbon" storage reservoirs by counteracting carbon fixation and burial (Rosentreter et al., 2018). However, stretching over a global coastline of ~1,600,000 km, these environments are intrinsically heterogeneous with a mosaic of habitats with varying substrate composition (Holland & Elmore, 2008;Koch, 2001), associated species communities (Dias et al., 2018;Stein et al., 2014), and ecosystem processes across space and time (Cardinale et al., 2006;Hewitt et al., 2008). Thus, the inherent properties that make coastal environments so diverse and heterogenous also complicate our ability to narrow down carbon dynamics in these areas sufficiently (Rosentreter et al., 2021a).
In this context, global estimates of coastal CH 4 emissions presently do not sufficiently reflect the heterogeneous and dynamic nature of coastal environments. In fact, the three classical blue carbon ecosystems, seagrass meadows, salt marshes, and mangrove forests (Mcleod et al., 2011) have been the focal point for global coastal CH 4 assessments due to their large carbon sequestration potential (Mcleod et al., 2011). It is only recently that tidal flats, coastal aquaculture, and inner estuaries have been added to the global CH 4 budget (Rosentreter et al., 2021b), but measurements from other coastal areas are still pooled without further habitat discrimination (Weber et al., 2019). For example, highly productive but less conspicuous coastal ecosystems with mixed-macrophytes, algal dominance, or bare sediments are common but not explicitly included in these estimates. Yet, given their high carbon turnover rates (Attard et al., 2019a(Attard et al., , 2019b, these habitats may contribute significantly to the total coastal CH 4 emissions (Lundevall-Zara et al., 2021). Moreover, the majority (85%) of studies quantifying CH 4 emissions from vegetated coastal areas have been performed south of 45 degrees North (70% when excluding mangroves that only occur around the tropics) (Al-Haj & Fulweiler, 2020). Northern temperate and high-latitude coastal systems are highly productive and also experience climate change at an accelerated rate compared to low-and mid-latitude areas (Screen & Simmonds, 2010;Serreze et al., 2009), increasing the demand for studies assessing temperature-sensitive CH 4 dynamics from these regions (Yvon-Durocher et al., 2014). There is also a major knowledge gap in our understanding of the variability of CH 4 in surface waters over short spatial scales reflecting the ecosystem mosaic typical for the coastal environment (Sheaves, 2009), as has been shown relevant for seafloor gross primary production and community respiration in shallow areas (Rodil et al., 2021).
Emissions of CH 4 are, furthermore, particularly variable in time, and narrowing the uncertainty in the global coastal CH 4 budget requires a methodology capable of quantifying natural variations arising from biotic and abiotic drivers across multiple timescales (Rosentreter et al., 2021a). For example, 74 of 98 studies (75.5%) used to calculate the global median CH 4 flux from vegetated coastal ecosystems in Al-Haj and Fulweiler (2020) employed flux chamber measurements or discrete sampling. Chamber measurements produce time-averaged flux estimates (often for a period between 24 and 48 h). In contrast, discrete samples have no time-weighted average, but due to logistical reasons, are usually taken at frequencies of no more than one to five samples per day and location (Banerjee et al., 2018;Dutta et al., 2015;Nirmal Rajkumar et al., 2008). These studies resulted in significant advances in our understanding of CH 4 emission from coastal systems. Yet, the strong influence of physical forcing (e.g., wind, waves, currents, tides) on the main CH 4 emission pathways (diffusion and ebullition) over short timescales (minutes to hours) can lead to a high CH 4 concentration and flux variability within one diel cycle, as has been shown in lake environments (Sieczko et al., 2020) and tidal influenced estuarine systems (Rosentreter et al., 2018). In the past decade, methods have been developed to improve the spatial and temporal resolutions of CH 4 concentration and flux measurements in aquatic systems.
For example, using real-time in situ measurements based on a gas equilibrator coupled to cavity ring-down spectroscopy (CRDS), Call et al. (2015) and (2019) showed variability across days to weeks and Rosentreter et al. (2018) documented seasonal CH 4 variability in mangrove creeks. These high-resolution efforts have facilitated an improved understanding of different pathways, sources, and sinks in mangrove forests, yet the amplitude and underlying mechanisms of this variability in other coastal marine ecosystems are largely unknown. In addition, seasonal sampling becomes especially important for annual estimates from underrepresented northern temperate and high-latitude regions, but time-series measurements are often discontinued in winter due to harsh weather conditions. Although high-resolution measurements are critical for reliably capturing the magnitude of the coastal CH 4 variability, these sampling campaigns can be time-consuming and expensive. As such, it is desirable to determine the sampling effort required to obtain a high-accuracy, representative mean dissolved CH 4 concentration for various coastal environments.
We explored the spatial and temporal variability of CH 4 across various heterogeneous coastal environments by systematically measuring CH 4 concentrations in three widely distributed yet understudied northern temperate coastal habitats (Figure 1a). The CH 4 distribution in shallow (<4 m water depth) mixed-vegetated, algae-dominated, and adjacent bare sediment habitats was assessed during five sampling campaigns spanning an entire year (Figure 1b), including an ice-covered period in late winter/early spring. We performed in situ real-time monitoring of CH 4 concentrations using CRDS to account for the temporal variability by diel cycles and peak events (Call et al., 2015;Maher et al., 2013;Rosentreter et al., 2018).
This state-of-the-art technique also permits high temporal resolution measurements of stable carbon isotope ratios of CH 4 (δ 13 C-CH 4 ) that help elucidate the controls and formation and removal pathways of the coastal carbon cycle . All measurements were complemented with benthic vegetation and physicochemical data to (a) provide spatially and temporally resolved CH 4 distribution and emission data from major northern temperate nearshore benthic environments; (b) identify potential biotic and abiotic drivers in shaping the temporal variability of CH 4 ; and (c) test whether current methods are sufficient in resolving the CH 4 distribution within highly heterogeneous and dynamic coastal settings both spatially and temporally.

| Study area
This study compares three distinct nearshore shallow (<4 m water depth) coastal habitats located on the island of Askö in the Baltic Sea (58°49'15.4"N 17°38'08.8"E). The habitats are representative for globally pervasive coastal ecosystems and were identified according to their dominant type of substrate and vegetation: (1) Mixed-vegetated communities of vascular plants and algae on sediments (hereafter "mixed-vegetated" habitat); (2) mixed turf-and macroalgae on rocks with pockets of sediments (hereafter "algaedominated" habitat), and (3) surrounding soft sediments without major macrovegetation cover (hereafter "bare sediments"). Each habitat was assessed visually, and the percent cover of the underlying substrate and macrovegetation was recorded within a 5-m radius. Taxa that could not be identified underwater were sampled and confirmed in the laboratory. Benthic surveys were repeated in April and September 2020. Overall, the mixed-vegetated habitat was characterized by coarse sediments with 60-90% total vegetation cover. The vegetation was dominated in equal parts by vascular plants (e.g., Phragmites australis, Stuckenia pectinata, and Ruppia spiralis) and benthic algae (e.g., Chara aspera and heterogenous assemblages of filamentous algae). The "algae-dominated" habitat was situated on rocks and boulders with pockets of permeable sediments with 80-95% total vegetation cover comprised of the macroalgae Fucus vesiculosus, and Ulva spp., the encrusting Hildenbrandia rubra, and various filamentous algae. No vascular plants were identified in this habitat. The surrounding bare sediment habitat with fine soft sediments had 7-10% total vegetation, of which were mainly dislodged F. vesiculosus and filamentous algae. The study was conducted at the SW facing side of the island, which is dominated by rocky cliffs and shallow embayments and is relatively open to the Baltic Sea. The habitats were fully submerged at all times due to the absence of tides in this region of the Baltic Sea (Medvedev et al., 2016). The average of measured salinities (i.e., per sampling period and habitat) in the studied area ranged from 6.2 to 7.0 over the course of the year, and, thus, reflected brackish water conditions typical for the central Baltic Sea. F I G U R E 1 Study location and habitat types (a), surface water temperature at the study location (b) and histograms of the density distributions of surface water methane (CH 4 ) concentrations (c) and stable carbon isotopes of CH 4 (d) across habitats and different sampling months. The five sampling campaigns are depicted as grey bars in (b); the light blue bar indicates the period of coastal sea ice cover. Temperature-coded points are individual measurements at 15 min intervals, and the black line denotes the daily running mean temperature. CH 4 concentrations in (c) >300 nM, which represent <1% of the data, were omitted for graphical representation but can be found in Table 1. The asterisk denotes under-ice sampling in March While the Baltic Sea receives freshwater inflows from land and has limited saltwater inflows from the Danish straits, locally at the study site on the island in the outer Stockholm archipelago, there were no major freshwater inputs from rivers or streams, which is reflected by relatively constant salinity throughout the measurement period.
For the measurements, we used an adapted version of the Water Equilibration Gas Analyzer System (WEGAS) (details in Humborg et al., 2019) coupled to a CRDS. The system consists of four major components: (i) a submersible seawater intake pump at around 0.3 m water depth mounted to a movable raft that can be deployed noninvasively over the various habitats; (ii) a water handling system comprised of a showerhead equilibrator (1 L headspace volume) and a thermosalinograph (Seabird TSG 45) fed via a hose by the seawater intake pump; (iii) a gas handling system with circulation pumps for the showerhead and ambient air; and (iv) the CRDS gas analyzer for CH 4 and CO 2 concentration and related C-isotope measurements (model G2201-i, Picarro Inc.). The use of a large seawater intake pump results in the combined measurement of CH 4 from ebullition (bubbles) and the dissolved form in the water. The individual contribution of the two forms can, however, not be resolved using the current system. For CH 4 and CO 2 analyses, gas in the showerhead of the equilibrator was measured for 35 min, followed by gas measurements of ambient air for 10 min (i.e., one complete cycle was 45 min).
These measurement cycles (i.e., 35 min, water; and 10 min, air measurements) ran continuously during the five measurement periods mentioned above. The raft with the water intake pump was moved between the defined habitats every 24 h from the shore with ropes.
Measurements in March were distinct from the other sampling periods due to the ice cover that had been present for 4-6 weeks prior to the time of sampling. Here, holes were drilled into the ice and the pump lowered to measure "under-ice" concentrations. We validated the CRDS analyzer's performance according to the manufacturer's instructions with "ALPHAGAZ TM Stable Isotope Ratio Gases" for Picarro instruments. Specifically, before each deployment period, we injected three standards with varying CO 2 and CH 4 bulk concentrations, and varying δ 13 C-CO 2 and δ 13 C-CH 4 isotope values (i.e., low = 1.00 ppm CH 4 , −24.20‰ δ 13 C-CH 4 , 250.00 ppm CO 2 , −5.00‰ δ 13 C-CO 2 ; natural = 1.77 ppm CH 4 , −48.30‰ δ 13 C-CH 4 , 399.00 ppm CO 2 , −8.50‰ δ 13 C-CO 2 ; and high = 10.00 ppm CH 4 , −68.60‰ δ 13 C-CH 4 , 1000.00 ppm CO 2 , −20.10‰ δ 13 C-CO 2 ). Measurements with each standard ran for 10 min, and three-point calibration lines were constructed whose regression coefficients were used to scale the unknown sample data if needed.
Concentration and isotope measurement at 1 Hz frequency were averaged and logged every 10 s. The recorded data were filtered by removing data from the transition period between ambient air and water measurements due to the response time of CRDS to sharp changes in concentrations of CH 4 and CO 2 . Data were also removed during improper functioning (e.g., low water flow). For this study, we used 210,059 (averaged from 2,100,590 measurements at 1 Hz) data points each for CH 4 , CO 2 , δ 13 C-CH 4 , and δ 13 C-CO 2 for statistical purposes. CH 4 concentrations in water (in ppm obtained by the CRDS) were converted to molar concentrations (i.e., CH 4 in nM) and CO 2 was converted to pressure units (i.e., pCO 2 in μatm) . Alongside CRDS measurements, several other environmental and meteorological variables were recorded. Surface water temperature, pH, and dissolved oxygen concentrations at the point of water intake were logged every 15 min using a multiparameter sonde (model EXO2, YSI) that was calibrated prior to each deployment. Water temperature and salinity were also recorded with every CRDS measurement with a thermosalinograph (Seabird TSG 45) that was positioned before the showerhead equilibrator. Wind data observations (wind speed and direction) and air temperature were obtained from a Metek uSonic-3 heated 3D sonic anemometer, and a Vaisala HMP155 shielded temperature probe mounted on a 1.5-m high meteorological mast. The mast was located at the waterline in a coastal bay, approximately 400 m to the northwest of the sampled habitats. Mean winds were adjusted to a 10-m reference height assuming a logarithmic profile with neutral stability (Haugen, 1973): where U is the measured wind speed at height z u , u* is the measured friction velocity by the 3D sonic anemometer, and κ is the von Karman constant (0.4).

| Exploration of the CH 4 distribution variability
We used a generalized linear model (GLM) to examine differences across habitats within each month. Due to positive-skewed data and overdispersion, a quasi-Poisson model was constructed using the glm() function in r (R Core Team, 2021) with "Month" (i.e., March, May, July, August, December) and "Habitat" (i.e., Mixedvegetated, Algae-dominated, and Bare sediments) as factors. We used the R package "emmeans" (Lenth et al., 2019) for pairwise post hoc multiple comparisons with Bonferroni-adjusted p-values.
Results and model details are presented in Table S1. The relationships among CRDS and environmental data were initially assessed using principal component analysis (PCA) using the R packages "Factominer" (Husson et al., 2016) and "Factoextra" (Kassambara & Mundt, 2017). PCA is a multivariate technique used to emphasize variation and to visualize patterns in a dataset, particularly when there are many variables. Upon the visual inspection of the PCA, we calculated Spearman coefficients for correlations between U10 = U + u * × log 10 z u surface water CH 4 concentration and potential environmental drivers (i.e., water temperature, salinity, dissolved oxygen and CO 2 concentrations, and pH).
The thermal sensitivity of the CH 4 distribution was further ex- We applied the Rayleigh model to estimate the fraction of CH 4 that was oxidized in surface water in each habitat and sampling month, as: where δ 13 C CH4 (CRDS) is the isotopic composition of surface water CH 4 measured with the CRDS system, δ 13 C CH4 (S) is the isotopic value of the CH 4 source in sediments, −67‰ that has been measured in local sediments, ε is the isotope fractionation factor for CH 4 oxidation of −20‰ (Bastviken et al., 2002), and ƒ represents the fraction of remaining CH 4 in surface water, that is, 1−f is the fraction of oxidized CH 4 . The Rayleigh model assumes a closed system when CH 4 oxidation occurs, which means CH 4 oxidation is the only sink of CH 4 in water column and is faster than the refreshment of CH 4 supplied to the surface water. This is an oversimplification given the high variability of coastal systems. The true fraction of CH 4 oxidized in surface water could, thus, be underestimated due to the contribution of 13 C-depleted CH 4 source mixing with surface water CH 4 with higher δ 13 C CH4 values in a partially open system.

| Sampling effort evaluation of dissolved CH 4 concentrations
We used a bootstrapping exercise to determine the minimum number of concentration measurements per day required to obtain a high-accuracy, representative daily mean dissolved CH 4 concentration across the various coastal habitats and sampling months. The assumption of these simulations is that our high-resolution sampling effort (i.e., one sample per second) can sufficiently capture the temporal variations of surface water CH 4 concentrations for each habitat. We randomly sampled the population of measured CH 4 concentrations assuming a variable sample size, from 1 to 600 samples a day (with sample replacement). This sampling was repeated 200 times for each sample size, and for each simulation, we calculated the resulting mean CH 4 concentration.  The saturation of CH 4 is relative to the dissolved equilibrium with ambient air.
Abbreviations: CV, coefficient of variation; IQR, interquartile range; N, number of individual observations (10 s average of 1 Hz measurements); SD, standard deviation. The asterisk denotes under-ice sampling in March.

| Sea-air flux computation
The sea-air flux (F) of CH 4 was calculated as: where k (m s −1 ) is the gas transfer velocity, K 0 (mol m −3 atm −1 ) is the aqueous-phase solubility of CH 4 , and pCH4 sea and pCH4 air are the measured partial pressures (atm) of CH 4 in the near-surface water and in the air, respectively. We compared our site-specific atmospheric CH 4 concentration measurements to data of the closest ICOS atmospheric monitoring station (i.e., Utö-Baltic Sea station; sta- where β is the dimensionless (mL of gas dissolved per mL of H 2 O) Βunsen solubility coefficient, A1, A2, A3, and B1, B2, B3 are constants from Table 1 in Wiesenburg and Guinasso (1979), T is the measured water temperature (K), and S the measured salinity (‰). Assuming CH 4 behaves as an ideal gas, K 0 is related to β by K 0 = β (R × T STD ) −1 , where R (m 3 atm K −1 mol −1 ) is the ideal gas constant and T STD (K) is the standard temperature in Kelvin.
The gas transfer velocity (k) used is that determined by Wanninkhof (2014) as: where U is the wind speed (m s −1 ) at 10 m height and Sc balticsea is the Schmidt number at the measurement site, which is dependent on temperature, salinity, and gas molecule. Sc was corrected for the corresponding temperature that was measured simultaneously with partial pressures of CH 4 (pCH 4 ) according to coefficients taken from Table 1 in Wanninkhof (2014). Furthermore, the Schmidt number for Baltic Sea brackish water (i.e., Sc balticsea ) with measured salinity (S balticsea ) was calculated by interpolation of Sc for fresh water (salinity 0‰) and seawater (salinity 35‰) following (Gülzow et al. 2013) and (Jähne et al. 1987): All fluxes are expressed in μmol CH 4 m −2 day −1 . Other variables (e.g., currents, waves, water depth) can also be used to predict k in   (Table S2).
Sea-air fluxes of CH 4 were then calculated based on equations provided above, assuming an average salinity of 6.6 (i.e., the average of measured salinities ranging from 6.2 to 7.0 over the course of the year). Wind speed data from the study location matching the CH 4 concentrations (measured and calculated) was available for 21,445 out of 34,932 data points (61%). The remaining wind speed data were estimated using a Monte-Carlo simulation on the distribution (mean ± SD, 2.25 ± 2.01 m/s) of available wind speed data from that year.

| CH 4 concentrations and δ 13 C-CH 4 values across coastal habitats
We report a high spatial and temporal variability of surface water CH 4 concentrations across the mixed-vegetated, algae-dominated, and bare sediment habitats that span two orders of magnitude, ranging from 6 to 460 nM ( Figure 1c; Table 1). During all sampling periods, the highest concentrations were always observed in the mixed-vegetated habitat, followed by algae-dominated, and surrounding bare sediment habitats (Table 1). A generalized linear model (GLM) with pairwise post hoc multiple comparisons confirmed that CH 4 concentrations differed significantly across habitats within each sampling month (Table S1), with an exception of the algae-dominated and bare sediment habitats in May. In addition, differences between the algae-dominated and bare sediment habitats were minor (expressed by odds ratios close to 1 as effect size statistics) in May, July, and August (Table S1). There were strong seasonal variations of CH 4 concentrations with similar patterns across all habitat types. In general, the highest CH 4 concentrations were observed in August, followed by July, March, May, and December ( Figure 1c; Table 1). The δ 13 C-CH 4 values of surface water varied by >7‰ over the sampling months in all habitat types.
Across all habitats, CH 4 was most enriched in 13 C in December (average of −55‰) and became most depleted in March, approaching −63‰ (Figure 1d). Differences in δ 13 C-CH 4 values across habitats in the same month occurred only in some cases and were smaller than the annual temporal variation (Table S3).  (Table 1).
An exception to the seemingly random CH 4 variability within one diel cycle was the mixed-vegetated habitat in August, when CH 4 consistently peaked during midday (mean ± SD, 333 ± 93 nM at 13:00 h local time), and was lowest at night (141 ± 24 nM at 02:00 h local time; Figure 2a).

| Correlation of surface water CH 4 with environmental variables
Principal component analysis (PCA) revealed distinct separation of the CH 4 and environmental data across months and to a lesser extent across habitats (Figure 3a). The first two principal components The δ 13 C-CH 4 signatures provided an additional dimension to reveal the main processes controlling CH 4 variability given the isotope fractionation associated with CH 4 production and consumption (i.e., oxidation) (Barker & Fritz, 1981). δ 13 C-CH 4 values as a function of CH 4 concentrations reflected temporal variations across seasons (Figure 4). In all habitats, the lowest CH 4 concentrations with the highest δ 13 C-CH 4 values were observed in December, while the highest CH 4 concentrations and the lowest δ 13 C-CH 4 values were found in August and March. The Rayleigh model, assuming that the supply of CH 4 is much slower than oxidation in water column, was applied to estimate the fraction of CH 4 that was oxidized in surface water, suggesting 20% of CH 4 loss through oxidation in August and March, and up to 50% in December in all habitats (Table S3).

F I G U R E 4
Stable carbon isotopes of methane (δ 13 C-CH 4 ) as a function of the log CH 4 concentrations of surface water in three shallow coastal habitats (a-c) of the Baltic Sea. The data is represented as a nonparametric bivariate surface to describe the density of all data pairs (n = 210,059 in total). The contour lines are quantile contours in 20% intervals. The asterisk denotes under-ice sampling in March

| DISCUSS ION
Our high-resolution measurements revealed differences in the distribution of surface water CH 4 concentrations across neighboring coastal habitats over short spatial (within meters) scales and exceptionally high temporal variability that could only be detected with continuous measurement techniques during several seasons.
Acknowledging this high spatiotemporal variability is critical to confine CH 4 emissions from coastal environments and the variability associated with their habitat heterogeneity.

| Temperature sensitivity of coastal CH 4 distribution
Median CH 4 concentrations measured across the here-studied habitats were 4-13 times higher than those observed in deeper waters of the open Baltic Sea (Schmale et al., 2010;Wilson et al., 2018), up to three times higher than previously published data for coastal Baltic areas Ma et al., 2020), and substantially higher than globally compiled nearshore CH 4 concentrations (Weber et al., 2019) (Table S5). The magnitude highlights that vegetated coastal ecosystem are characterized by excessive organic matter loads from primary production, trapping and accumulation of allochthonous organic matter, and sedimentary conditions that can favor CH 4 production (Dale et al., 2019;Wallenius et al., 2021).
However, we also report an exceptionally high spatiotemporal variability of surface water CH 4 concentrations.
A first major source of this variability was attributed to seasonal differences in CH 4 concentrations. The significant correlation between CH 4 concentrations and temperature over the sampling months suggests that temperature mainly regulates seasonal variations. Like most other forms of metabolism, methanogenesis is temperature-dependent, and the calculated apparent activation energies (EA = ~1 eV, across all habitats) were in line with previous global estimates of ecosystem-scale CH 4 fluxes with an EA of 0.96 eV (Yvon-Durocher et al., 2014). Thus, the higher CH 4 concentrations in late summer are likely due to increased production under warmer water temperatures. Historical data from the nearby oceanographic observation station "2507 Landsort Norra" between 2010 and 2020 confirmed that the annual surface water temperature curve from our study area is representative of previous years (Sveriges meteorologiska och hydrologiska institut, 2022). We infer that the observed temperature sensitivity is primarily driven by natural temperature variations rather than a warming effect. Both aerobic CH 4 oxidation together with anaerobic CH 4 oxidation in sediments may also increase in summer due to temperature controlling their rates (Treude et al., 2005;Zehnder & Brock, 1980) and the increased supply of CH 4 supply by molecular diffusion. However, in summer, the overall production of sedimentary CH 4 likely outweighed the relative contribution CH 4 oxidation pathways. In support of this, parallel mea-

| Ice-cover effects on CH 4 dynamics
An exception to the overall seasonal trend was observed in March (i.e., late winter/early spring). Measurements during this month were marked by ice cover that, to this point, had been present for 4-6 weeks. Analogous to many northern lakes (Denfeld et al., 2018), we observed an accumulation of CH 4 under the ice, with mean concentrations six times higher than in December (last month without ice cover). More negative δ 13 C-CH 4 values in March (−62 to −64‰) suggest CH 4 supply with overall low oxidation. This observation corroborates studies showing suppressed methanotrophic activity at very cold temperatures (e.g., Phelps et al., 1998). Calculations of the Rayleigh model confirmed that <20% of the surface water CH 4 was oxidized during this period. However, the CH 4 depleted in 13 C could also be a result of varying fractionation during methanogenesis at lower temperatures or mixed CH 4 formation pathways. The CH 4 accumulation under ice will likely result in enhanced outgassing events following ice break (Ducharme-Riel et al., 2015;Karlsson et al., 2013). Whereas under-ice CH 4 accumulation is a well-studied feature of northern lakes, these dynamics have not been described for northern temperate coastal regions with regular sea ice every year.
Our data suggest the necessity to include the ice-covered period and CH 4 outgassing during ice breakup in future coastal CH 4 sampling strategies and the annual CH 4 budget of northern temperate and high-latitude regions (Omstedt et al., 2004).

| Physical forcing may drive short-term CH 4 variability
A second major source of variability in the CH 4 concentrations was short-term variations that occurred within hours (Figure 2d-f).
Most of this variability was independent of the time of the day and without an apparent and reoccurring diel pattern. However, fluctuations of the CH 4 concentrations were so strong that the minimum and maximum values within one habitat and sampling campaign (time window max. 12 days) could differ by up to one order of magnitude (Table 1). The dispersion of the CH 4 probability distribution around the mean concentration was on average 30% during the ice-free months and, thus, much higher than the reported global open ocean CH 4 variability with CVs ranging between 2% and 11% (Wilson et al., 2018). While we could not find any direct correlation to the available environmental data, one possible explanation for the high variability could be the physical influence of the open coastal setting through wind and/or wave action. A wave-induced pumping effect on the pore water pressure can transport solutes from deeper to surface layers (Precht & Huettel, 2004;Yang et al., 2019); Thus, varying CH 4 release rates from permeable coastal sediments in very shallow waters may cause variable near-surface CH 4 concentrations, as has been shown relevant even for lake systems (Hofmann et al., 2010). In support of this, the CVs of the CH 4 distribution were much lower across all habitats in March (mean CV = 10%), when, due to ice cover, the influence of waves and winds on the water column and sediments was likely minor and no CH 4 escaped to the atmosphere.

| Reoccurring diel CH 4 patterns in summer
A reoccurring diel pattern in CH 4 concentration changes was only observed in the mixed-vegetated habitat in August, with the highest concentrations consistently toward midday and lowest at night. This marked diel variation may be attributed to plant-mediated transport of CH 4 by convective throughflow from rooted submerged plants, which were only present in the mixedvegetated habitat. The convective transport through pressure gradients can account for up to 60% of the total CH 4 transport from sediments during daylight hours and high photosynthetic activity (Kim et al., 1998;van den Berg et al., 2020). In the early stages of plant growth, molecular diffusion through dead/live plants into the standing water column can be the primary transport mechanism (Kim et al., 2001). Most plants at the mixed-vegetated site were fully submerged; thus, a sediment-plant-water flux is likely.
However, Phragmites stems (comprising ~10% of the total vegetation in the mixed-vegetated site) possibly facilitated a sedimentplant-air flux of CH 4 (van den Berg et al., 2020), which will have remained undetected with our approach. Abiotic CH 4 photoproduction from organic matter degradation may also play a role in shaping site-specific CH 4 dynamics in oxygenated surface waters (Li et al., 2020;Zhang & Xie, 2015). However, given that reoccur-

| Spatial distribution of CH 4 reflects coastal ecosystem mosaic
Shallow coastal habitats are heterogeneous, and the variation in spatial structure and temporal change of benthic communities defines the expression of ecosystem functions in form and magnitude (Snelgrove et al., 2014). Reflecting the coastal ecosystem mosaic (Sheaves, 2009), some of the measurements across the studied neighboring habitats were not further than 30-50 m apart.
Yet, despite their proximity, we observed significant differences in the distribution of CH 4 in the water column and the magnitude of the resulting sea-air fluxes. Surface water CH 4 concentrations are likely related to variable CH 4 production and oxidation rates, as indicated by varying δ 13 C-CH 4 values across sites during some months ( Figure S3). These differences may be ascribed to different quantities and qualities of organic matter deposited within local sediments and differences of sediment properties (e.g., porosity) (reviewed in Rosentreter et al., 2021a). The presence of rooted vegetation may also play a role in the small-scale variability, as roots provide substrate via root litter and exudates and transport oxygen into the sediments. In addition, while the employed system measures CH 4 in the dissolved form and from ebullition (bubbles), the individual contribution of the two phases cannot be resolved but may contribute to differences between the habitat types. It becomes apparent that more research is required to determine the spatial scale of this variability and to understand better the controls on substrate availability for methanogenesis. In particular, links between biodiversity metrics (i.e., abundance and biomass) of primary and secondary producers and CH 4 production and consumption pathways need to be better constrained as has been shown relevant for seafloor metabolism (i.e., gross primary production and community respiration) in shallow waters (Rodil et al., 2021). Likewise, integrating knowledge on the structure of sediment microbial communities associated with the different habitats is imperative to improve the prediction of CH 4 production and oxidation pathways from different coastal habitats (Wallenius et al., 2021).

| High sampling intensity is required to capture coastal CH 4 variability
Particularly the high temporal variability on timescales from hours to days complicates our ability to generalize the distribution of  (Figure 5a).
A similar pattern was apparent in all other habitats and sampling periods (Table S6). Consequently, the data collection and sampling strategy are detrimental to accurately capturing the temporal variability and assure justified mean CH 4 concentrations that are the basis for flux computations. Thus, near-continuous measurements using CRDS (Hartmann et al., 2018;Humborg et al., 2019;Maher et al., 2013)  Bootstrapping results of all other months in Table S5 area in waters of <5 m depth is almost 30.000 km 2 (HELCOM, 2013; Jakobsson et al., 2019), and, thus, equals 22% of the global areal extent of mangroves (Bunting et al., 2018)

| Uncertainties in coastal CH 4 distribution and future research directions
Variations of surface water CH 4 concentrations and resulting sea-air fluxes reflecting the heterogeneous nature of coastal environments currently complicate generalizing regional patterns and upscaling attempts globally. Given the CH 4 distribution patterns identified in this study, we encourage several aspects to be considered to refine large-scale coastal CH 4 emission budgets.
First, studies currently used for global coastal CH 4 budgets have a site-selective bias due to their particular relevance in providing a service (e.g., they are interesting from a blue carbon perspective) and for other practical reasons like the accessibility of the study area. Here, we provided evidence that northern temperate coastal habitats, which are presently understudied for their contribution to CH 4 fluxes (e.g., algal communities on rocky shores), can be seasonally relevant sources of atmospheric CH 4 . Similar measurements should be extended to additional coastal environments and geolocations to confirm the global relevance of their CH 4 emissions. The spatial heterogeneity of coastal habitats provides an opportunity for measurements along environmental gradients, with great potential to increase inference across scales (Snelgrove et al., 2014).
Second, new technical approaches have to be embraced to better understand the high temporal variability of the CH 4 distribution and the underlying processes in coastal environments. The use of continuous rather than time-averaged measurements helps to account for short-term temporal variations by diel cycles or peak events (Call et al., 2015;Maher et al., 2013;Rosentreter et al., 2018), and reduces uncertainties when establishing diel budgets. The highresolution measurements across multiple seasons and the identification of dependencies on environmental variables have also bearings for predicting future CH 4 emissions under various changing environmental conditions.
Third, net annual CH 4 fluxes are influenced by temporal variations throughout the year. Thus, to increase confidence when compiling data for global coastal CH 4 budgets, better seasonal coverage of coastal CH 4 needs to be combined with the recognition that reported mean values (both CH 4 concentrations and emissions) might be biased toward sampling in a particular period only. As the seasonal behavior of CH 4 is highly site-specific, the variations need to be considered for each habitat type and geolocation.
Lastly, measurements of CH 4 emission from northern temperate and high-latitude coastal habitats should be acknowledged in future emission budgets. Climate change occurs particularly fast in northern hemisphere mid-latitude (Cohen et al., 2014) and high-latitude (Screen & Simmonds, 2010;Serreze et al., 2009)

| CON CLUS ION
We conducted seasonal sampling campaigns of dissolved CH 4 concentrations and δ 13 C-CH 4 values using a fast-response automated gas equilibrator and CRDS system across three globally pervasive vegetated and nonvegetated coastal habitats. As the first study to compare high-resolution measurements across neighboring habitats, we highlight unprecedented spatiotemporal variability of the CH 4 distribution driven by habitat-specific CH 4 production and consumption pathways, seasonal temperature dependencies, and shortterm fluctuations. A bootstrapping analysis on the continuous data revealed that scaling the CH 4 distribution from few samples involves large errors, and at least ~50 samples per day are needed to achieve accurate emission estimates. Failing to include such high-resolution measurements in future global CH 4 assessments may result in a continued systematic bias of regional and global estimates due to the lack of measurements representative for the coastal ecosystem mosaic-a highly heterogenous environment in space and time.
Ultimately, a better understanding of the habitat-specific contribution to the global CH 4 emission budget would improve efforts to address climate change, such as by revealing the net potential of coastal blue carbon habitats to sequester carbon.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.