High Emissions of Carbon Dioxide and Methane From the Coastal Baltic Sea at the End of a Summer Heat Wave

The summer heat wave in 2018 led to the highest recorded water temperatures since 1926 – up to 21 C – in bottom coastal waters of the Baltic Sea, with implications for the respiration patterns in these shallow coastal systems. We applied cavity ring-down spectrometer measurements to continuously monitor carbon dioxide (CO2) and methane (CH4) surface-water concentrations, covering the coastal archipelagos of Sweden and Finland and the open and deeper parts of the Northern Baltic Proper. This allowed us to i) follow an upwelling event near the Swedish coast leading to elevated CO2 and moderate CH4 outgassing, and ii) to estimate CH4 sources and fluxes along the coast by investigating water column inventories and air-sea fluxes during a storm and an associated downwelling event. At the end of the heat wave, before the storm event, we found elevated CO2 (1583 µatm) and CH4 (70 nmol/L) concentrations. During the storm, a massive CO2 sea-air flux of up to 274 mmol m-2 d-1 was observed. While water-column CO2 concentrations were depleted during several hours of the storm, CH4 concentrations remained elevated. Overall, we found a positive relationship between CO2 and CH4 wind-driven sea-air fluxes, however, the highest CH4 fluxes were observed at low winds whereas highest CO2 fluxes were during peak winds, suggesting different sources and processes controlling their fluxes besides wind. We applied a box-model approach to estimate the CH4 supply needed to sustain these elevated CH4 concentrations and the results suggest a large source flux of CH4 to the water column of 2.5 mmol m-2 d-1. These results are qualitatively supported by acoustic observations of vigorous and widespread outgassing from the sediments, with flares that could be traced throughout the water column penetrating the pycnocline and reaching the sea surface. The results suggest that the heat wave triggered CO2 and CH4 fluxes in the coastal zones that are comparable with maximum emission rates found in other hot spots, such as boreal and arctic lakes and wetlands. Further, the results suggest that heat waves are as important for CO2 and CH4 sea-air fluxes as the ice break up in spring.


INTRODUCTION
The sheltered coastal waters and estuaries are often oversaturated with CO 2 and CH 4 as a result of high supply of autochthonous and allochthonous riverine organic matter that is partly respired in the sediments and water column (Borges and Abril, 2012). Polar amplification of global warming may lead to more frequent extreme warming events (Mann et al., 2018) in high latitude coastal systems such as the northern Baltic Sea. This, in combination with the observed redistribution of land-derived carbon from land to sea, i.e., increased riverine total organic carbon (TOC) loads Andersson et al., 2015), potentially increases respiration patterns and sea-air fluxes of CO 2 and CH 4 .
The two barriers constraining CO 2 and CH 4 outgassing from marine water bodies are anaerobic and aerobic oxidation of CH 4 in the sediments and water column (Reeburgh, 2007(Reeburgh, , 2013 and limited vertical mixing across density gradients that often lead to an accumulation of CO 2 and CH 4 in deeper parts of the water column (Gentz et al., 2014). The latter barrier is particular important in the strongly stratified Baltic Sea (Jakobs et al., 2014). A portion of the CO 2 can be exported laterally with increasing salinity and pH as dissolved inorganic carbon (DIC) from the inland coastal waters to larger depths, as seen in the shallow East Siberian Sea and the central deep Arctic Ocean .
Ebullition of CH 4 is often discussed as a significant source to sea-air CH 4 fluxes, but the detection and spatiotemporal analysis are poorly constrained (Leifer and Patro, 2002;McGinnis et al., 2006;Rehder et al., 2009;Weber et al., 2014;Lindgren et al., 2016). However, recent studies suggest that the deeper, vertically stratified outer parts of high latitude coastal seas, such as the East Siberian Sea or the Baltic Sea, are moderate sinks for CO 2 (Gustafsson et al., 2014(Gustafsson et al., , 2015Omstedt et al., 2014;Humborg et al., 2017;Schneider et al., 2017) and that net sea-air CH 4 fluxes from the Siberian Shelf have been heavily overestimated (Parmentier et al., 2017). For example, only moderate CH 4 fluxes to the atmosphere have been reported for the open East Siberian Sea (Thornton et al., 2016) and the open Baltic Sea (Jakobs et al., 2014). In contrast, the inner parts of coastal waters and estuaries are potential hotspots of CH 4 and CO 2 outgassing (Bange et al., 1994) due to (i) the high organic matter content in coastal sediments, (ii) short residence time of gaseous compounds in the shallow water column, (iii) sediment gas release that is transported directly, as gas bubbles, to the atmosphere (meaning significantly less CH 4 oxidation within the water column), and (iv) the often high turbidity that potentially lowers primary production and CO 2 uptake.
Even if extreme weather events in coastal waters become more frequent, as predicted for the future for the Baltic Sea (Meier et al., 2012), their sampling remain difficult to schedule due to the randomness of their occurrence. However, modern instrumentation provides new opportunities; the online monitoring coastal observatory MONICOAST, near the Tvärminne Zoological Station (TZS) in southern Finland, has been recently launched to understand and visualize the impacts of long-term climate and environmental change 1 . This highresolution monitoring station is the site of almost 100 years of weekly observations at the same station. Thus, this monitoring station constitutes one of the longest temperature records available for a coastal site in the Baltic Sea. Temperature records at 32 m water depth indicated record high temperatures of >20 • C at the end of the heat wave during the summer of 2018.
Previous studies have shown that the combination of high temperatures and high organic matter contents in sediments and the water column along the Belgian coast in the North Sea led to high concentrations of CO 2 and CH 4 in the water column (Borges et al., 2016). Based on these previous results and the records from TZS, revealing the exceptionally high water temperatures, we conducted our cruise at the end of the extended heat wave in the summer of 2018. The main objective of this study was to quantify emission patterns of CO 2 and CH 4 under these extreme conditions, which can be regarded as a snap-shot of future emission regimes in coastal areas. Further, we wanted to investigate whether a sharp vertical water column stratification as established mainly by a halocline in the Baltic Sea and in the East Siberian Sea but not in other areas as the North Sea act as an efficient barrier for CO 2 and especially CH 4 fluxes also in shallow coastal parts. We followed a west-east transect across the Baltic Sea starting in the Swedish Archipelago and ending in the Finnish Archipelago, crossing the deeper open parts of the Northern Baltic Proper. We investigated CO 2 and CH 4 water column inventories and sea-air exchange patterns in contrasting environments where wind mixing and upwelling/downwelling are physical drivers that may trigger outgassing events of CO 2 and CH 4 . Additionally, we were able follow the break-up of the heat wave and summer stagnation that came after the first autumn storms (in this case, gale force winds) that deepened the surface water layer. By collecting continuous measurements by means of a seawater intake system, an equilibrator set-up and CO 2 and CH 4 analyzers and combing these measurements with acoustic mapping of the water column, we were able to map and follow outgassing patterns in the Finnish coastal waters over several days before, during, and after the storm. This spatially integrated and non-invasive approach allowed us to detect a high flux event across an entire bay related to extreme weather, something that needs to be considered in order to deepen our knowledge of coastal biogeochemistry and carbon processing in a warmer and potentially stormier world.

Study Area and Cruise Track
The Baltic Sea is a semi-enclosed brackish water body consisting of a set of major relatively deep basins, shallower bays, gulfs, and archipelagoes. The most recent digital bathymetric model by the European Marine Observation and Data Network (EMODnet Bathymetry Consortium, 2018) delineates the various sub basins of the Baltic Sea that are used throughout this study and indicate the average depth of the Baltic Sea with 53 m FIGURE 1 | Cruise track. . After the Black Sea, the central Baltic Sea is the second largest anoxic coastal water body in the world. This is due to water residence time of ∼30 years, a strong vertical salinity gradient (halocline) at 60-90 m water depth, and ongoing eutrophication .
The cruise was conducted with Stockholm University's Research Vessel (RV) Electra with the aim of obtaining a comprehensive and representative picture of the distribution of CO 2 and CH 4 concentrations in surface water across the Baltic Sea, covering a broad depth range, at the end of the heat wave during the summer of 2018. We covered archipelago-type coastal seascapes of Sweden and Finland, the Northern Baltic Proper, the Åland Sea, and the Gulf Finland (Figure 1). In fact, these shallower coastal seascapes consisting of hundreds to thousands of semi-enclosed bays and basins make up a huge proportion of the Baltic Sea, i.e., 25% of the Northern Baltic Proper, 70% of the Åland Sea, and 60% of the Gulf of Finland are shallower than 40 m depth . The 40 m boundary used in this study to discriminate shallower from deeper sites is chosen because it separates two concentration regimes in the Baltic Sea and the shallower parts show much more elevated concentration of CO 2 and CH 4 . In general, very little is known about sea-air gas exchange in these shallow areas of the Baltic Sea, and hardly any data are available with only a few exceptions for the Gulf of Finland, i.e., the harbor areas of St. Petersburg and Helsinki . For the northern boreal part of the Baltic Sea consisting of the Bothnian Bay, the Quark, the Bothnian Sea, the Gulf of Finland, and the Northern, Western and Eastern Baltic Proper the total area shallower than 40 m is some 71,000 km 2 . For perspective, impressive data sets on water-air CO 2 and CH 4 exchange are available (Juutinen et al., 2009;Humborg et al., 2010) for Sweden and Finland, where total lake area is about 33,000 km 2 and 30,000 km 2 , respectively (the Swedish lake area was estimated excepting the largest deep lakes Vättern and Vänern where no CH 4 and CO 2 exchange estimate exists).

Statistical Evaluation of Long-Term Temperature Records
The long-term temperature record from the TZS coastal monitoring site (lat: 59.85548N, lon: 23.26033E) covers August 1926 to December 2018. A moving average was adopted to smooth the short-term fluctuations in the temperature data by taking the average value of a 1-year length and shifting forward, i.e., for each average annual value calculation, the first number of the 1-year data series was excluded and the next value was included. Note that the time interval of each temperature record varied slightly, from 10 days in the years before 1997 down to 4-7 days after 1998 and finally daily records in 2018. Thus, the 1-year length used in this study is fixed from a day in a year to the same day in the year after regardless of the time interval between two specific temperature records.

Continuous CO 2 and CH 4 Measurements in Surface Water and Atmosphere
The partial pressures of atmospheric and dissolved CO 2 and CH 4 were measured using the Water Equilibration Gas Analyzer System (WEGAS) coupled to cavity ring-down spectrometer (CRDS). This system consists of three major components: (i) a water handling system comprised of a showerhead equilibrator (1 L headspace volume) fed by the seawater intake, an E&H electrode probe and a thermosalinograph (Seabird TSG 45); (ii) a gas handling system with circulation pumps for the showerhead and ambient air from the bow of the ship; and (iii) CRDS gas analyzers for CO 2 and CH 4 concentrations (model G2131-i, Picarro Inc.); and A detailed description of the WEGAS systems and its performance can be found in the Supplementary Material.
Measurements of surface water CO 2 and CH 4 concentrations were performed using the seawater intake located just below the sea surface. Water was pumped through spray nozzles into the open headspace equilibrator at ∼4.5 L min −1 . By creating a fine spray of droplets, the exchange surface between showerhead and water was maximized and an optimal equilibration is achieved. The ambient pressure was maintained through a ambient air-fed vent flow that was monitored continuously during the measurements. The gas equilibrated in the showerhead was measured using the CRDS analyzer. For the standard CO 2 and CH 4 analysis routine, ambient air is measured for 8 min followed by the gas measurements in the showerhead of the equilibrator for 12 min, i.e., one complete analysis cycle is 20 min. Cycles are performed continuously during the cruise including the station time when RV Electra is not steaming.
Continuous CO 2 and CH 4 measurements in both ambient air and surface water were conducted during September 18-23, 2018; in total, 80,393 data points for both CO 2 and CH 4 in air and water were recorded. The recorded data were filtered by removing data taken during the transition period between ambient air and water measurements (see detailed description in Supplementary Material) and collected during improper functioning of WEGAS (e.g., low water flow, exceptional values due to air contamination by the ship) and thereafter the data were also corrected for the vent flow to the equilibrator that was induced while WEGAS was running. Any remaining data with CO 2 concentrations above 450 ppm were also removed. After filtering, atmospheric CO 2 averaged 401.3 ± 2.08 ppm (2σ). As a result of spatial normalization and excluding data that could have been contaminated by the ship, here we present 15,733 data points for both air and water concentrations.
For reporting the CO 2 content in water, the ppm value obtained by the cavity ring down spectrometer (CRDS) were converted to pressure units (pCO2 in µatm). We assumed ambient atmospheric pressure because water was taken at depth of around half a meter below the surface and the pressure in the showerhead is equal to ambient air as achieved by the vent flow taken at the bow of the ship.
For reporting the CH 4 concentrations in water, the ppm value obtained by CRDS were converted to molar units (i.e., nmol/L) by Henry's law (Equation 1), because the achieved equilibration means that the gas content in the showerhead is equivalent to that in the water phase, i.e., CH 4 measured by CRDS is considered as the real-time concentration of CH 4 in water. This makes Henry's law applicable for dealing with gas/liquid equilibria and the Henry's constant for molar concentration and partial pressure conversion is chosen in this case. As the showerhead in the equilibrator is maintained at ambient pressure and the water temperature is measured by thermosalinograph right before the water reaches the equilibrator, Henry's law constant is corrected for each calculation by the measured temperature at atmospheric pressure.
where C is the concentration of CH 4 (mol/L), p is the partial pressure in unit of atm (1 atm = 10 6 ppm) of CH 4 . K H is the Henry's law constant that is dependent on temperature and is corrected in this study following the van 't Hoff equation (Sander, 2015): Where K * H is the Henry's law constant at the reference temperature, 298 K, T * is the reference temperature, 298 K and T is the measured temperature in water (K). The temperature dependence of the equilibrium constant, − sol H R does not change much with temperature and is tabulated (Sander, 2015). K H also varies with pressure, but in this study, pressure was assumed to be atmospheric pressure because water was taken at depth of around half a meter.
In the Baltic Sea both CO 2 and CH 4 concentrations span an order of magnitude and the highest concentrations reported are >2500 µatm of CO 2 (Fransner et al., 2019) and >1200 nM of CH 4 (Jakobs et al., 2014).

Calculations of Sea-Air Gas Fluxes
The sea-air flux (F) of CO 2 and CH 4 is commonly expressed by the following equation (Liss and Slater, 1974): where k is the gas transfer velocity and C sea and C air are concentrations of gases in the sea and the overlying air, respectively. This is also frequently written as Equation 4 where K 0 is the solubility of CO 2 or CH 4 , p sea and p air are the measured partial pressures of the gases in equilibrium with surface water and in the air, respectively. k is the gas transfer velocity that has been optimized as Equation 5 (Wanninkhof, 2014): where U is wind speed (m/s) and Sc is the Schmidt number, which is highly dependent on temperature and gas. Sc in our calculations was corrected for the corresponding temperature that was measured simultaneously with partial pressures of CH 4 and CO 2 (pCH 4 and pCO 2 ).
Combining Equations 3, 4 and 5, the flux of CO 2 and CH 4 in this study is finally calculated as Equation 6: Box Model Approach to Address Sources and Sinks of CO 2 and CH 4 Using the surface CO 2 and CH 4 concentration in inshore and offshore surface waters and sea-air flux data and the information about the deepening of the upper mixed surface layer after the storm event, we can construct a budget evaluating the sources and sinks of CO 2 and CH 4 (Figure 2). The conservation of mass for the surface layer under the assumptions mentioned is: Where h is the mixed layer depth before and after the storm, respectively, F atmospheric flux and I an internal source term. The source water for the advection term, i.e., CO 2offshore and CH 4offshore are assumed to have properties close to what was observed offshore on September 19 and 20.

Continuous Measurements of Gas Flares and Bubbles
The acoustic water column data presented here were continuously collected with a Simrad EK80 wideband splitbeam scientific echo sounder with a center frequency at 70 kHz that was installed on RV Electra. The vertical range resolution is approximately 2 cm and the ping rate was about 2.5 Hz. The sonar produces a linear frequency modulated acoustic signal which provides improved signal-to-noise ratio and a higher vertical range resolution compared to similar narrow-band systems (Stanton and Chu, 2008). The dataset collected with the EK80 was match filtered with an ideal replica signal using a MATLAB software package version R2017a provided by the system manufacturer, Kongsberg Maritime (Lars Anderson, personal communication). Position and altitude information FIGURE 2 | Scheme over the budget approach evaluating the internal sources and sinks of CO 2 and CH 4 in TZS coastal waters.
Frontiers in Marine Science | www.frontiersin.org were provided to the sonar by a Seapath 330 + motion sensor. Ranges from the transducer were calculated using the cumulative travel times through sound speed profile layers based on the nearest (in time) CTD profile (SEA-Bird SBE 911 plus), following the approach by Stranne et al. (2017).

Temperature Records in the Open Northern Baltic Proper and in the Finnish Coastal Waters Near Tvärminne
Temperatures of up to 23.04 • C (daily mean) were recorded in the upper surface layer in the open Northern Baltic Proper on 6 August at the autonomous measuring stations (Figure 3A).
A similar picture appears when evaluating the coastal monitoring stations near TZS. Since 2017, water temperatures have been continuously recorded using in situ instrumentation, allowing us to follow the effects of the summer 2018 heat wave in even greater detail. The highest bottom-water temperatures (32 m depth) of 20.53 • C were recorded on 30 July at this coastal monitoring station ( Figure 3B) and these are the highest temperatures ever recorded at this depth over the past 93 years. A second warming event in bottom waters above the sediments was recorded in early September 2018 just before our sampling campaign with temperatures reaching nearly 17 • C ( Figure 2B). The record-high bottom temperatures measured in the TZS area in 2018 follow an increasing trend observed over the past century until today. In Figure 3C, the long-term temperature record (moving average) between 1926 and 2018 is presented showing a mean water temperature of 4 • C above the sediments between the 1920s and the 1980s. Thereafter, we see a dramatic increase in bottom-water temperatures starting in the 1990s to about 6 • C in the 2010s.

Spatial CO 2 and CH 4 Patterns in Open vs. Coastal Areas of the Baltic Sea
Surface-water temperatures were between 14 and 18 • C throughout the transect ( Figure 4A) which is somewhat cooler than at the peak of the heat wave during July/August, when temperatures of up to 23 • C were recorded in the surface layer ( Figure 3A). Both pCO 2 and CH 4 concentrations were closer to equilibrium with the atmosphere in the central basin of the Northern Baltic Proper, but they were elevated in coastal areas (Figures 4C,D). However, off the Swedish coast at 70-100 m water depth, we encountered an upwelling event indicated by low temperatures and high salinity at the water surface (Figures 4A,B) and with elevated pCO 2 of up to >800 µatm ( Figure 4C). CH 4 showed only moderately elevated concentrations of 12 nmol/L ( Figure 4D). Toward the Finnish coast, both CO 2 and CH 4 gradually increased to concentrations of ∼700 µatm for CO 2 and over 30 nmol/L for CH 4 . Partial pressure and concentration vs. seafloor depth plots reveal that in coastal areas, with water depths <40 m, both CO 2 and CH 4 (Figures 5A,B) show more elevated pCO 2 and CH 4 concentrations in the upper mixed layer compared to the open Northern Baltic Proper.
At the easternmost end of the transect, we investigated parts of Finnish Archipelago Sea near the TZS in more detail between September 19-23, 2018. pCO 2 and CH 4 concentrations followed a salinity gradient with higher values toward the freshwater end-member (Figure 6) that is formed by the small river Karjaanjoki (runoff 0.59 km 3 y −1 , Räike et al., 2012) that runs into the Pojo Bay and further into the Storfjärden Bay (black dashed box in Figure 8A). The highest pCO 2 (1583 µatm) was measured at the northernmost part of the Storfjärden where salinities were <5 psu. A similar pattern was recorded for CH 4 , which reached its highest concentration of 70 nmol/L a bit south from the position where CO 2 peak concentrations were found. An interesting pattern appeared when plotting pCO 2 and CH 4 concentrations against seafloor depth (Figure 7). pCO 2 in surface water were lowest at the shallowest seafloor depth and continuously increased from 751 µatm at <5 m seafloor depth to 1020 µatm at 20-25 m seafloor depth ( Figure 7A). In contrast, the high CH 4 concentration of 44 nmol/L (median) was recorded at the lowest seafloor depth, while concentrations ranged between 24 and 40 nmol/L at deeper depths between 5 and 25 m ( Figure 7B). Note, that all median pCO 2 values and CH 4 concentrations were calculated before the wind event described in the next section.

Wind-Induced Changes in CO 2 and CH 4 Water Column Inventories and Inferred CO 2 and CH 4 Source and Sink Terms in Coastal Waters
Figures 8B,D show the temporal development of pCO 2 and CH 4 concentrations inshore and offshore the TZS coastal area. Elevated pCO 2 of some 800 µatm were recorded by the WEGAS system on September 19 and 20 within the inshore waters (indicated with the black dashed box in Figure 8A) and the offshore pCO 2 (indicated with the red dashed box in Figure 8A) were also at the same elevated range as the inshore concentrations (median of 740-856 µatm; Figure 8B). During sampling on September 21, the wind gradually increased from 5 m sec −1 to some 10 m sec −1 and pCO 2 in the inshore waters increased to 1,015 µatm, although this increase is not significant when taking the variability of the measurements in the inshore waters that day (between 700 and 1 300 µatm; Figure 8B) into account. Wave height increased simultaneously from 0.6 to 1.6 m. Wind speeds peaked at 20 m sec −1 during September 22 and decreased again down to 3 m sec −1 during the evening hours of September 23 ( Figure 8C). During the same time peak wave height of 1.9 m was recorded at the monitoring buoy near TZS, which decreased to 0.5 m. Thus, during the entire period between September 19-23 the average wind was between 8 and 10 m sec −1 (0.5-0.8 m wave height), disrupted by the peak wind event on September 22 with wave heights of up to 1.9 m. CO 2 sea-air fluxes were greatest during this peak wind event with 274 mmol m −2 d −1 (Figure 8C), whereas at the end of the wind event, CO 2 was undersaturated in the upper surface layer and even negative seaair fluxes were calculated. However, during the entire wind event, the pCO 2 decreased sharply from a 2-3 times oversaturation (median of 1015 µatm) to a slight undersaturation (median of   321 µatm recorded at September 23). Thus, the wind event rather effectively depleted the water column with respect to pCO 2 .
In contrast to the expected depletion of CO 2 from oversaturation caused by the wind event, CH 4 remained oversaturated before, during, and also after the wind event. Elevated CH 4 concentrations of 46 -47 nmol/L (median; Figure 8D) were recorded by the WEGAS system on September 19 and 20 within the inshore waters, whereas for the offshore waters CH 4 concentrations were also supersaturated, but less so, with median concentrations between 13 and 15 nmol/L. During and after the wind event during September 22, CH 4 concentrations decreased in the inshore waters to a median value of 20-24 nmol/L, however, remained oversaturated by 5-8 times. Accordingly, the CH 4 sea-air flux calculations ( Figure 8E) revealed the highest CH 4 average fluxes between 2.8 -3.3 mmol m −2 d −1 before the wind event and lower, but still elevated, fluxes between 1.0 -2.4 mmol m −2 d −1 during and after the wind event.
Although wind is directly related to the gas transfer velocity (k) in Equation 3, wind speed could only partly explain the variation in both CO 2 and CH 4 fluxes. However, the highest CH 4 fluxes were observed at relatively low winds (12 m sec −1 ), whereas the highest CO 2 fluxes were observed at peak wind speeds (Figures 8C,E). Accordingly, the CO 2 and CH 4 fluxes show positive relationship (not shown), however, the highest CH 4 fluxes were recorded at very low CO 2 flux regimes, which indicates that CH 4 flux events are decoupled from CO 2 flux events during low wind regimes (9-12 m sec −1 ).
As indicated in Figures 8C,E, the wind increased significantly after September 21 from less than 10 to nearly 20 m sec −1 . Simultaneously, CTD data collected between September 21 (10:25 UTC) and September 23 (14:00 UTC) show a deepening of the surface mixed layer after the wind event (Figure 9). Note that the temperature (Figure 9B) before and after the wind event (September 22) remained more or less constant at 11 • C and that salinity ( Figure 9C) decreased only slightly from 6.5 to 6.4 psu, whereas the upper mixed layer depth increased from roughly 10-25 m water depth. Thus, the surface mixed layer increased over these 2 days in depth from about 10 m to nearly 25 m and this was mainly caused by a downwelling event triggered by south-southeasterly winds pressing water masses into the Finnish inner coastal waters rather than by vertical mixing. Thereafter, the mixed layer depth remained at the same depth for the remainder of the cruise.
The concentrations, mixed layer depths, and atmospheric fluxes used to estimate the internal sink/source terms are given in Table 1. We assume that the transition occurs during 24 h corresponding to approximately the time of the storm event and the interval between the two CTD profiles mentioned above. The results of the budget calculation indicate a net sink of CO 2 FIGURE 6 | Surface distribution of (A) salinity, (B) pCO 2 (µatm), and (C) CH 4 concentrations (nmol/L) within the TZS coastal waters; the black dots indicate daily cruise tracks within the area. in the water column, however, the sink could be somewhat overestimated because the offshore concentrations decreased due to the massive outgassing. Further, the budget indicates a CH 4 source of about 2.5 mmol m −2 d −1 from below the mixed surface layer.
A possible explanation for the deviating patterns of CO 2 vs. CH 4 fluxes can be found in the geophysical variables investigated and continuously recorded during the cruise. The acoustic midwater mapping revealed vigorous and widespread gas flares from the sediments in the Finnish coastal waters as exemplified in Figure 10 where some 10-15 major flares and numerous trains of individual bubbles are visible along a 3 km transect, indicating intense bubble formation and ebullition at the sea floor. Further, gas bubbles are visible throughout the entire water column up to the water surface, penetrating into the mixed surface layer that was situated at around 10-15 m water depth (based on CTD data, Figure 9). The mixed layer is also discernable in the echogram, although vaguer than observations made in Arctic Ocean coastal waters (Stranne et al., 2017).

DISCUSSION
On a global scale, lakes and rivers are significant sources for CO 2 and CH 4 , and their fluxes are relatively well constrained compared to the evaluation of sinks and sources in the global coastal ocean. CO 2 emissions from lakes and reservoirs are estimated to be 0.32 Pg C yr −1 and from rivers and streams to be 1.8 Pg C yr −1 (Raymond et al., 2013), whereas CH 4 emissions from inland waters are estimated to be 77 Tg C yr −1 (Bastviken et al., 2011). Upscaling exercises using a typology approach including four estuarine types and different types of continental shelves along three climatic zones estimates global CO 2 emissions to be 0.27 Pg C yr −1 , whereas global CH 4 emissions from estuaries to be 5.0 Tg C yr −1 (Laruelle et al., 2010;Borges and Abril, 2012). However, Laruelle et al. (2010) conclude that "the largest uncertainty of scaling approaches remains in the variability of CO 2 data to describe the spatial variability, and to capture relevant scales of variability." Note that the global estimates of coastal gas emissions are biased toward the many investigations focusing on estuaries, lagoons, salt marches, and tidal flats (Borges and Abril, 2012), whereas studies capturing shorelines that are less affected by river inputs are less studied. Further, CH 4 emissions in coastal areas are clearly coupled to temperature (Borges et al., 2016;Sawicka and Brüchert, 2017) and the highest rates of change are to be expected in organicrich shallower parts affected by warming water bodies overlaying marine sediments. In fact, experimental studies show that a temperature increase of only 2 • C can increase anaerobic organic matter degradation by 40% in marine sediments (Roussel et al., 2015). We clearly demonstrate in this study that the temperature increases in Finnish coastal waters correspond to such a 2 • C rise during the last 3 decades. Bottom waters and surface sediments in this sub-arctic setting experience peak temperatures of over 20 • C coinciding with elevated CO 2 and CH 4 concentrations and fluxes that are comparable to the more frequently investigated terrestrial hot spots for CO 2 and CH 4 emissions, such as arctic lakes, wetlands and thawing permafrost areas.
We suggest that the sea-air CH 4 flux is likely dominated by frequent bubbling from sediments, which we observed even during low winds, when CO 2 fluxes were at their minimum, high CH 4 fluxes occurred, knowing that this notion lacks a quantitative approach that has to be developed in the future by new ways of interpreting geophysical data of the water column. However, we have shown that surface CH 4 concentrations were highest at the shallowest seafloor depth (Figure 7) indicating that the proximity to the sediments and gas flares as well as the short residence times that prevent full CH 4 -oxidation in the water column as observed in the deep open ocean (Reeburgh, 2007). In contrast, CO 2 in surface water shows a positive relationship with seafloor depths, indicating that respiration of autochthonous and allochthonous organic carbon in the water column is the predominant source for CO 2 even in shallow coastal systems. This has bearing for the understanding and future model parametrizations especially of CH 4 fluxes in arctic and subarctic shelf systems, i.e., a comprehensive understanding of spatial and seasonal CH 4 accumulation and flare distribution in sediments as stimulated by heat waves is needed to quantify CH 4 fluxes from shallow coastal waters. Acoustic methods directly addressing bubble-mediated methane flux are becoming more available (Weidner et al., 2019) and similar studies using mid-water sonar systems near Spitzbergen have shown that CH 4 partly originating from ebullition and transformed into the dissolved pool during upward transport is efficiently trapped below the halocline at some 200 m water depth for at least part of the year, which prevents outgassing to the atmosphere (Gentz et al., 2014). This is corroborated by our findings in the deeper stratified Northern Baltic Proper where even a strong upwelling event led to strong outgassing of CO 2 (Figure 4C), but only moderate increased CH 4 concentrations were observed ( Figure 4D). Recent findings suggest that methanogenic Archaea in zooplankton guts significantly contributes to the CH 4 supersaturation in the upper surface layer of the Baltic Sea (Schmale et al., 2018), i.e., the role of bubble-mediated CH 4 transport in the central parts of Baltic appear less likely. In  shallow coastal areas the role of such pelagic sources may presumably be less important, because the zooplankton biomass is simply limited by the lower volume of water, but this cannot be ruled out.
We observed a pronounced pycnocline also in the coastal waters around TZS (Figure 9), but in this shallow coastal setting we could observe bubbles rising from the seafloor and throughout the entire water column and penetrating the  pycnocline (Figure 10). Note that the WEGAS system is capturing both diffusive and bubble flux collected by the sea water intake. We do not know whether a CH 4 reservoir has accumulated below the pycnocline in the Finnish coastal waters and thereby contributing to the elevated CH 4 concentration in the water column and significant sea-air fluxes. However, gas bubbles can only survive a certain rise height of some 100 m in the water column before they are depleted in CH 4 (McGinnis et al., 2006). Once CH 4 is dissolved, it is prone to oxidation before reaching the atmosphere and here vertical density gradients such as pycnoclines form a barrier for dissolved CH 4 . This barrier to CH 4 transport from the seafloor to the atmosphere does not work in shallow coastal systems where ebullition from the seafloor can reach the atmosphere directly and our investigations show intensive bubble mediated transport between the sediments and the atmosphere down to 45 m water depth (Figure 10). Future investigations will show whether CH 4 concentrations in the bubbles do decrease with increasing water column travel time, i.e., whether these bubbles still contain significant amounts of CH 4 when reaching the sea-air interface. Further, it appears highly likely that the gas bubbles are of biogenic origin, because we found an inverse pattern, though not statistically significant, between CH 4 concentrations and seafloor depth and/or proximity to the shore (Figures 6, 7) -a coastal effect. Eutrophication, high productivity and organic matter from runoff leads to accumulation of sediments with thick sequences of organic mud nearshore the TZS coastal area (Kauppi et al., 2018), which in turn leads to methane production and seepage. These patterns wouldn't appear when the bubbles were of thermogenic origin, i.e., from petrogenic deposits that may occur in folds that should be more randomly distributed.
A striking result of this study is that CH 4 , which has a lower solubility compared to CO 2 , never equilibrated with the atmosphere, whereas simultaneously the slower reacting gas (CO 2 ) with a higher solubility was depleted within a few hours after the wind event. Our box model approach inferred a source term for CH 4 -that might originate directly from sediment flares, or a diffusive flux from the sediments and a theoretical water column reservoir -that is significant (2.5 mmol m −2 d −1 or 30 mg C m −2 d −1 ). A comparable benthic study from the nearshore shallow coastal areas along the Swedish coast (Sawicka and Brüchert, 2017) reported nearshore sediment-water fluxes between 0.1 and 2.6 mmolm −2 d −1 . Our reported rate is at the high end, indicating that the heat wave played an important role in a massive flux event of CH 4 , although we have to bear in mind that the benthic sediment fluxes and fluxes from a potential reservoir in the water column cannot be disentangled from our CH 4 supply estimates. Although we do not have any quantitative measurements, the observed extensive occurrence of gas flares and bubbles point toward a significant contribution from the sediments. A comparable study addressing sea-air fluxes in the North Sea (Borges et al., 2018) report elevated fluxes between 28 and 126 mmol m −2 per year. Our observed accumulated flux for the 5-day period at the end of the heat wave was 11 mmol m −2 (averagely 2.2 mmol m −2 d −1 ), which is similar to CH 4 emissions observed in thermokarst lakes in subarctic areas (Matveev et al., 2016). This result supplies further evidence that heat waves trigger massive CH 4 fluxes to the atmosphere from nearshore sites.
The sea-air CO 2 exchange of 75-274 mmol m −2 d −1 or 0.9-3.3 g C m −2 d −1 also indicates extreme outgassing during the end of the heat wave and can be compared with recent model studies indicating outgassing of up to 60 g C m −2 yr −1 in the nearshore areas of the northern Baltic Sea (Fransner et al., 2019). It is of course difficult to compare simulated annual rates with our measurements that were taken over a few days but it is, nevertheless, striking that the daily sea-air exchange rates of CO 2 presented here are one order of magnitude higher than the simulated ones. It could be that the effects of heat waves followed by increasing autumn winds for the outgassing of CO 2 haven't been considered, i.e., our 5-day estimated sea-air exchange of 6.5 g C m −2 is already 10% of the highest simulated annual values reproducing the spatial hot spots near the coast. The model has been calibrated/validated with ferry-box observations and clearly indicates the classical CO 2 accumulation of up to 2500 µatm under the ice and values close to equilibrium in ice-out situations. However, this study demonstrates that heat waves produce similar accumulations of CO 2 in the water column and the outgassing events can be as significant as ice-out situations that have been described also for other high latitude coastal systems, such as the East Siberia Sea (Thornton et al., 2016;Humborg et al., 2017).

CONCLUSION
In the past decade, continuously recording CO 2 and CH 4 instruments have become state-of-the-art equipment on large research vessels and on commercial ships which run ferrybox solutions. Thus, we have a fairly good understanding of CO 2 and CH 4 exchange processes in open waters of the Baltic Sea, the open Atlantic, or even in remote areas as the Siberian Sea (Körtzinger et al., 1996;Gülzow et al., 2011;Schneider et al., 2014;Thornton et al., 2016;Humborg et al., 2017). Even sediment-water exchange of CH 4 is investigated by automated lander techniques, and many of these studies focus on deeper sites (Boetius and Wenzhöfer, 2013). For the Baltic Sea and the North Sea, there are only a few datasets on nearshore shallow area emissions available and this situation may be representative for many shallow coastal systems that make up half of the Baltic Sea or a third of the global continental shelfs. In this study we show that these nearshore coastal systems are CO 2 and CH 4 emission hotspots and fluxes are comparable to terrestrial hot spot in Arctic lakes (Huttunen et al., 2006;Repo et al., 2007;Juutinen et al., 2009;Wik et al., 2013) and that shallow coastal systems are highly responsive to ongoing changes in temperature. Coastal seascapes constitute a mosaic of heterogeneous environments and habitats, and sediment-water interactions become more pertinent compared to the well-studied open coastal and ocean waters. For shallow coastal waters, the non-invasive and continuous instruments that record sediment and water column gas inventories and fluxes are still not sufficiently used and must be deployed on small research vessels that can reach the shallowest sites, which often host complex habitats such as algal belts, mussel beds, or seaweeds. The combination of the automated continuous CO 2 -and CH 4 -measurement devices and lander techniques, together with modern geophysics as mid-water sonar, multibeam or sub-bottom profilers, will allow identifying gas accumulations, pock marks, as well as flares and bubbles. As a result, we hypothesize that the sediment accumulation and event driven high gas emissions from nearshore shallow sediments is a major carbon capacitor in coastal areas that is not accounted for by the classical invasive core-incubation methods (Reeburgh, 1968) or by lander techniques (Boetius and Wenzhöfer, 2013) that still govern our view and process understanding of sediment-water CH 4 fluxes, and that these event-driven massive fluxes will increase with rising temperatures.

DATA AVAILABILITY
All data necessary to evaluate and build upon the work in this manuscript can be requested from the corresponding author. Atmospheric and surface water CO 2 and CH 4 data as well as CTD water profile data presented in this manuscript will be archived and freely available from the Bolin Centre Database, http://bolin.su.se/data.

AUTHOR CONTRIBUTIONS
CH wrote the manuscript and performed the measurements. MG developed the WEGAS system. XS, MM, and CM-M performed the carbon dioxide and methane measurements and did the statistical data analyses. CS and MJ did the geophysical measurements. BG did the budget calculations. AS performed the CTD measurements and analyses of wind data. AN and JN responsible for the long-term temperature measurements. FUNDING AN, CH, and BG acknowledge funding from the Academy of Finland (project ID 294853).