Oxygenation of a karst subterranean estuary during a tropical cyclone: mechanisms and implications for the carbon cycle

Seasonal precipitation affects carbon turnover and methane accumulation in karst subterranean estuaries, the region of coastal carbonate aquifers where hydrologic and biogeochemical processes regulate material exchange between the land and ocean. However, the impact that tropical cyclones exert on subsurface carbon cycling within karst landscapes is poorly understood. Here, we present 5‐month‐long hydrologic and chemical records from 1 and 2 km inland from the coastline within the Ox Bel Ha Cave System in the northeastern Yucatan Peninsula. The record encompasses wet and dry seasons and includes the impact of rainfall during the development of Tropical Storm Hanna in October 2014. Methane accumulated in highest concentrations at the inland site, especially during the wet season preceding the storm. Intense rainfall led to episodic increases in water level and salinity shifts at both sites, indicating a spatially widespread hydrologic response. The most profound storm effect was a ~ 0.8 mg L−1 pulse of dissolved oxygen that declined to zero within 2 weeks and corresponded with a reduction of methane. A positive shift in methane's stable carbon isotope content from −62.6‰ ± 0.6‰ before the storm to −44.0‰ ± 2.4‰ after the storm indicates microbial methane oxidation was a mechanism for the loss of groundwater methane. Post‐storm methane concentrations did not recover to pre‐storm levels during the observation period, suggesting tropical cyclones have long‐lasting (months) effects on the carbon cycle. Compared to seasonal effects, mixing and oxygen inputs during storm‐induced hydrologic forcing have an outsized biogeochemical influence within stratified coastal aquifers.

The region of a coastal aquifer where seawater and terrestrial-borne meteoric waters intersect, the subterranean estuary, supports biogeochemical processes that affect coastal elemental cycling (Moore 1999;Moore and Joye 2021;Santos et al. 2021).Karst subterranean estuaries are found worldwide in karstic (eroded carbonate) platforms, accounting for at least 15.7% of all coastlines (Goldscheider et al. 2020), but their influence on groundwater-derived nutrient and trace element concentrations in seawater is likely to be far greater (Gonneea et al. 2014;Rodellas et al. 2015;Mayfield et al. 2021).
Dissolution and diagenesis of carbonate bedrock during low sea-level stands formed abundant caves and sinkholes that have become subsequently inundated through sea-level rise, creating a globally dispersed environment at the land-sea aquatic continuum for characteristic groundwater-dependent ecosystems (a.k.a., anchialine ecosystems) inhabited by invertebrates with subterranean adaptations (Bishop et al. 2015;van Hengstum et al. 2019).These permeable conduit networks enhance the land-to-sea transport of freshwater and the intrusion of seawater that impacts groundwater resources (Perry et al. 2002;Ferguson and Gleeson 2012).Groundwater movements are orchestrated by intertwined meteorological, hydrologic, and oceanographic factors that are sensitive to anthropogenic alterations (e.g., groundwater extraction), semidiurnal forces (e.g., tidal), seasonal changes (e.g., precipitation), and extreme weather events (e.g., hurricanes and tropical storms).
Subterranean estuaries discharge substantial dissolved methane into shallow coastal waters (Cable et al. 1996;Burnett et al. 2006;O'Reilly et al. 2015), perhaps surpassing rivers (e.g., de Angelis and Scranton 1993; Middelburg et al. 2002).Shallow coastal areas are the primary source of natural methane emissions from the oceans to the atmosphere (Weber et al. 2019).However, regional and global methane flux estimates from subterranean estuaries to coastal surface waters remain uncertain, mainly because of the spatial and temporal variability of methane production and consumption resulting from climatic, hydrogeological, and biogeochemical factors associated with coastlines (Pain et al. 2019).For example, anaerobic oxidation of methane removed $ 1466 mg CH 4 m À2 yr À1 from the groundwater before its discharge to the sea from a siliciclastic subterranean estuary (Schutte et al. 2016).By contrast, aerobic methane oxidation in a partially oxic karst subterranean estuary setting consumed only $ 25 mg CH 4 m À2 yr À1 for the catchment region (Brankovits et al. 2018), yet this methane contributes substantial carbon and energy to the groundwater ecosystem (Brankovits et al. 2017).
In surface estuaries, floodwaters accelerate the turnover and land-to-sea transport of organic carbon stored in coastal watersheds and, thus, have a major influence on coastal carbon cycling (Bianchi 2011;Bauer et al. 2013;Ward et al. 2017).Terrestrial organic matter inputs affect the composition and bioavailability of dissolved organic carbon and dictate microbial community structure within estuaries and adjacent coastal waters (Steichen et al. 2020;Yan et al. 2020), which influences CO 2 emissions and other lingering biogeochemical effects for weeks to months after severe rainfall (Bianchi et al. 2013;Osburn et al. 2019).On karst landscapes, a key process that regulates biodiversity and impacts water quality is increased mixing between fresh and saline waters when rainfall-driven flooding shunts large volumes of water through conduits (Coutino et al. 2017;Kovacs et al. 2017;Calder on-Gutiérrez et al. 2018).It therefore stands to reason that storminduced hydrologic forcing within karst subterranean estuaries will impact groundwater-dependent ecosystems, the internal carbon cycle, and the export of carbon to the coastal ocean.
In this study, we obtained continuous 5-month-long meteorological, hydrological, and chemical records-collected at varying distances inland from the coastline-to identify spatial differences and temporal dynamics in the accumulation and consumption of methane within a karst subterranean estuary in the northeastern Yucatan Peninsula, Mexico.These records document the impact of intense rainfall during the development of Tropical Storm Hanna, a cyclone that tracked from west to east on the southern part of the peninsula in October 2014 (Cangialosi 2014).The concurrent records reveal mechanisms that link storm-induced hydrologic forcing with marked changes in dissolved oxygen concentrations and methane turnover.

Study sites
The Caribbean coast of the Yucatan Peninsula in Mexico (Fig. 1a) has flat topography (5-m mean elevation) and lacks surface flowing streams and rivers.The Caribbean Sea experiences microtides spanning 0.1-0.2m (Kjerfve 1981) that affect groundwater table elevation up to 100 km inland on the peninsula (Coutino et al. 2021).Precipitation evapotranspires or infiltrates the unconfined and density-stratified carbonate aquifers rather than flowing over land to the sea (Perry et al. 2002;Beddows et al. 2007).The freshwater-saltwater transition zone becomes progressively deeper with distance from the coastline (Fig. 1c).Networks of well-developed caves and sinkholes influence hydrologic exchange between the aquifer and the sea depending on seasonal rainfall and storms (Young et al. 2008;Null et al. 2014).The mapped portion of the conduits presently surpasses 1600 km along a 12-km wide section of the coastline between Cancun and Muyil (Kambesis and Coke 2013; QRSS: https://caves.org/project/qrss/qrlong.htm, last accessed on 5 July 2022).
The Yucatan Peninsula's tropical climate is characterized by dry (from December to April/May) and wet (from May to November) seasons (Curtis et al. 1996;Kottek et al. 2006).More than 80% of the mean annual rainfall of $ 1500 mm occurs during the wet season, with maximum rainfall happening during the months of June and September (Curtis et al. 1996).This period overlaps with the Atlantic Hurricane Season (June to November) when the likelihood of cyclones (e.g., tropical depressions, tropical storms, and hurricanes) is greatest.Hurricanes and tropical storms have repeatedly impacted local habitats, coastal groundwater resources, and sediments within flooded cave passages in the Yucatan Peninsula and other parts of the western tropical North Atlantic (Garrido-Pérez et al. 2008;Kovacs et al. 2017;Winkler et al. 2020).
Tropical Storm Hanna was initially a depression that crossed the southern part of the Yucatan Peninsula from west to east on 22-24 October 2014 (Cangialosi 2014).It produced disorganized showers and thunderstorms as it moved eastward.Then rising air developed a strong coupling (deep convection) between the surface and the lower atmosphere, resulting in the formation of a tropical depression (on 22 October) with rainbands that stretched over the peninsula and into the western Caribbean Sea.The cyclone passed $ 220 km south of the study area and achieved tropical storm status on 27 October.Although no storm surge was reported along the Caribbean coast, significant salinity increases within the meteoric lens at multiple locations in the northeastern part of the Yucatan Peninsula indicate regionally widespread mixing as large volumes of water from the intense recharge rushed through the conduits, leading to shear-induced vertical mixing between the deeper saline groundwater and the meteoric waters (Coutino et al. 2017;Kovacs et al. 2017).
We investigated two sites within the Ox Bel Ha Cave System that were accessed through flooded sinkholes (cenotes) by trained scientific cave divers along a transect perpendicular to the coastline (Fig. 1b).Cenote Na'ach Wennen Ha is $ 1 km from the coastline (20.170 , À87.460 ) and Cenote Odyssey is $ 2 km from the coast (20.170 , À87.470 ).To distinguish the study sites, we refer to the one closer to the coastline as coastal site and the other as inland site.The nearby Cenote Bang further inland (6.3 km from the coast) has been a site of previous investigations (Brankovits et al. 2017(Brankovits et al. , 2018)).The caves are flooded and are located beneath dry tropical forest south of the town of Tulum.Because none of these locations are recreational dive sites, access was limited to researchers during the study.The Cenote Na'ach Wennen Ha conduits are somewhat shallower ($ 6 m average and $ 13 m maximum water depth) than the Cenote Odyssey conduits ($ 13 m average and $ 22 m maximum water depth).Biodiversity surveys found cave-adapted invertebrate animals restricted to fresh, brackish, and saline waters at these locations (Alvarez et al. 2015), which are typical for subsurface habitats in the region (Calder on-Gutiérrez et al. 2018;Angyal et al. 2020).

Environmental parameters
Physicochemical water column parameters were measured using a YSI XLM-600 and/or EXO2 multi-parameter datasonde (Xylem, Inc., Yellow Springs, OH) with a measurement frequency of 0.25-1 Hz during the deployment and recovery of the underwater data loggers.This approach enabled measurements of water depth, temperature, salinity, and dissolved oxygen along vertical profiles of the water column.The datasonde was carried by the lead diver slowly descending (2-4 cm s À1 ) and advancing with probes projecting forward to acquire undisturbed profiles.
Between 15 August 2014 and 14 January 2015, precipitation, barometric pressure, and air temperature were recorded at a meteorological station on a secured rooftop (Speleotech facility; 20.207 , À87.468 ) about 4 km from Cenote Odyssey, 15 m above ground level, and well above the forest canopy line in Tulum, Quintana Roo.Precipitation, presented as daily total accumulation (mm d À1 ), was measured with an Onset (Onset Computer Corp., Bourne, MA) RG-3 rain gauge (calibration accuracy of AE 1.0%, up to 2 cm h À1 ; 0.2-mm resolution; time accuracy AE 1 min per month).Barometric pressure (< 0.02-kPa resolution and a AE 0.3% of full scale or 0.62-kPa accuracy as the maximum error) and air temperature (0.1 C resolution, AE 0.44 C accuracy) were recorded at 15-min time intervals with an Onset U20 data logger.
Data loggers measuring water level, groundwater flow, and water temperature were also deployed in the flooded caves at shallow depths (5.0-5.5 m) at each sampling station.Dissolved oxygen was only monitored at Cenote Odyssey.Flow velocity was measured as current speed and direction with a TCM-1 (Lowell Instruments LLC., Falmouth, MA) tilt current meter (TCM; AE 2 cm s À1 + 3% of reading accuracy and 0.1-cm s À1 resolution; bearing AE 5 accuracy for speed > 5 cm s À1 with a 0.1 resolution) at 1-min intervals.Water level measured with an Onset U20 logger (with an error less than AE 1.0 cm and 0.21-cm resolution at 15-min intervals) is reported as the departure from the mean value after compensating for barometric pressure change at the meteorological station.Dissolved oxygen (DO), reported as mg L À1 , was measured with an Onset U26 logger (0.02 mg L À1 resolution and a calibration accuracy of AE 0.2 mg L À1 for the calibrated range 0-20 mg L À1 ).The DO logger was zero calibrated by placing it in sodium sulfite following manufacturer recommendations for low oxygen (< 4 mg L À1 ) environments.The recovered data loggers showed no signs of biofouling, which is consistent with other logger deployments in dark and oligotrophic underwater cave environments (Beddows and Mallon 2018).

Temporal records of water chemistry
Continuous water samples for analysis of methane concentration and chloride concentrations (for salinity) were obtained with OsmoSamplers (Jannasch et al. 2004).Osmo-Samplers continuously pump environmental water into hundreds of meters long copper coils from which chemical timeseries have been obtained at deep ocean hydrocarbon seeps (Lapham et al. 2013), Arctic lakes (McIntosh Marcek et al. 2021), estuaries (Gelesh et al. 2016), and karst subterranean estuaries (Brankovits et al. 2018).For this study, each OsmoSampler (manufactured by TLR, Inc.) had eight semipermeable membranes (2ML1, Durect Co.) that separated a saturated salt solution from deionized water.The eightmembrane setup was chosen to target a pumping rate close to 1 mL d À1 because it allowed for a higher than weekly resolution.The resultant osmotic potential between the chambers drew water across the membranes and generated pumping rates of 0.87 and 1.24 mL d À1 for the Osmo-Samplers in Cenote Na'ach Wennen Ha and Cenote Odyssey, respectively.The osmotic pump was connected to a 300-m coil of 1.59 mm outer diameter gas-impermeable copper tubing that was filled with deionized water prior to deployment.Deionized water drawn from the copper coil was replaced by environmental water through a 0.2-μm hydrophilic filter (Rhizosphere Research Products) at the sample inlet located at the end of the coil.
OsmoSamplers were deployed following established protocols in flooded caves (Brankovits et al. 2018).Briefly, the pump and the sample coil were co-deployed with the data loggers and packed in a plastic laundry basket (42 cm wide, 42 cm long, and 33 cm high) to allow the diver to safely deploy the full set of devices.At both sites, sampling stations were established as close to the cave ceiling as possible in the upper aquifer with low salinity and minimal flow velocity where higher methane concentrations are expected (Brankovits et al. 2017).The sample inlets of the Osmo-Samplers in Cenote Na'ach Wennen Ha and Cenote Odyssey were placed at 5.5 m and 5.0 m water depths, respectively, which also correspond to the low-salinity-halocline between the fresh and brackish waters.The deployment period lasted from 15 August 2014 and 14 January 2015.Upon recovery, the coils were sealed on both ends, stored at $ 7 C in the field and at 4 C thereafter until they were subsampled in the laboratory within 1 month of collection.
Subsampling involved crimping 4.5-m sections of the copper tubing that were later split into a 4-m-long section from which water was extracted for methane concentration analysis, and a 0.5-m-long section for analysis of chloride concentration (Gelesh et al. 2016).Near the end of the 300-m-long coil attached to the OsmoSampler in Cenote Na'ach Wennen Ha, the salinities changed from the environmental water's brackish (2-3 psu) to deionized water values (0 psu), indicating the entire record was sampled (i.e., no overflow of sample from end of tubing).We did not find the salinity interface between the sample water and the deionized water in the OsmoSampler coil recovered from Cenote Odyssey because the copper tubing overfilled.Based on pumping rates, each 4.5-m section represents 2.5 d for the Cenote Na'ach Wennen Ha coil and 2.0 d for the Cenote Odyssey coil.It was therefore possible to ascribe a date/time interval to each copper coil section.All sections were analyzed for chloride, but, on average, only every second section was analyzed for methane.These analyses were carried out at the Chesapeake Biological Laboratory in Solomons, MD.
For methane concentrations, sample water from the 4-m sections ($ 2 mL volume) was expelled into 10-mL syringes using a benchtop tube roller with a gas tight connection between the tubing and syringe.Subsequently, 8 mL of ultrazero air was added to the syringe which was then shaken for 2 min to equilibrate dissolved methane into the inert headspace.This headspace was then injected into a gas chromatograph equipped with a HayeSep D packed column (Alltech Inc., Nicholasville, KY) and a Flame Ionization Detector to measure methane partial pressures.After comparing sample response to certified standards (Airgas, Inc.), the resultant partial pressures were used to calculate dissolved methane concentrations (in nM) in the water using Henry's law according to equations in Magen et al. (2014).Certified standards (Airgas, Inc.) were used for the calibration curve.Analytical precision was 3% and all measurements were above the detection limit of 2 ppmv.
For chloride concentrations, sample water from the 0.5-m sections were diluted (1:135) in Milli-Q water and analyzed on a Dionex ICS 1000 ion chromatograph (IonPac AG22 4 Â 50-mm guard column, IonPac AS22 4 Â 250-mm analytical column, and ASRS 300 4 mm suppressor) with an AS40 Autosampler that has an analytical precision of 2%.Certified IAPSO seawater standard (Ocean Scientific International Ltd.) was used for the calibration curve.Chloride concentrations were used to determine salinity as described by Brankovits et al. (2017).

Water column geochemistry
In addition to temporal sampling, discrete water samples were manually collected in syringes at the times of OsmoSampler deployment and recovery to obtain independent baseline measurements at the beginning and end of the time-series from next to the sample inlet of the Osmo-Samplers.Additional discrete samples were collected from the water column along depth profiles in Cenote Odyssey (from 1.5 to 16 m water depth) and in Cenote Na'ach Wennen Ha (from 3 to 12 m) to cover the full salinity spectrum at each site.
Collection, processing, and analysis of the discrete samples were carried out as described in detail by Brankovits et al. (2017).Briefly, water samples were collected near the deployed OsmoSamplers in plastic 60-mL syringes fitted with three-way stopcocks.The syringes were rinsed with distilled water and dried prior to the dive and flushed with sample water prior to closing the stopcock.Samples were kept cool during transport to the field lab where, within 8 h of collection, they were transferred into appropriate vials (30-mL serum vials for methane analysis and 2-mL screw cap plastic vial for ion concentrations).The serum vials for methane water samples were prepared prior to sample collection by adding 0.2 mL 1 M NaOH as a preservative before sealing the container with 1-cm-thick butyl septa and vacating the serum vial of air with a pump.
The samples were then kept in the fridge 7 C in the field and stored at 4 C until laboratory analyses at U.S. Geological Survey (USGS) Woods Hole Coastal and Marine Science Center and Woods Hole Oceanographic Institution (WHOI) in Woods Hole, MA.Methane concentrations were determined with a Shimadzu 14-A gas chromatograph (GC); stable carbon isotope composition of methane (δ 13 C-CH 4 ) was measured using a Thermo-Finnigan DELTA Plus XL isotope ratio mass spectrometer coupled to an Agilent 6890 GC; and chloride concentrations (used to determine salinity) were measured using a Metrohm 881 Compact Plus ion chromatograph (IC) equipped with a Metrosep A Supp 5-250 anion column.

Water column properties
Physicochemical profiling revealed distinct water masses separated by haloclines that were relatively constant in depth during August 2014 and January 2015 (Fig. 2).Lowest salinity values were consistently measured in the subaerial (open-toair) cenote pool and near the cave ceiling.This shallow water mass was separated by a low-salinity-halocline at $ 5 m (Fig. 2) below the groundwater table from the deeper parts of the meteoric lens.Highest salinities were always recorded in the deepest parts of the caves below a high-salinity-halocline at $ 12 m below the groundwater table at the coastal Cenote Na'ach Wennen Ha and at $ 13 m at the inland Cenote Odyssey (Fig. 2).To differentiate the subterranean water masses, we describe the water mass below the high-salinity-halocline as saline groundwater (> 30 psu); the water mass in the meteoric lens with intermediate salinities above the high-salinityhalocline as brackish groundwater (1-30 psu); and lowest salinities above the low-salinity-halocline as fresh groundwater (< 1 psu), similar to Brankovits et al. (2017Brankovits et al. ( , 2018)).
At Cenote Odyssey ($ 2 km from the coast), salinity increased with depth from as low as 0.8 psu to mesohaline (as high as 7.3 psu) between the two haloclines and reached highest salinity (as high as 35.5 psu) in the saline groundwater.A similar vertical salinity gradient was observed at the At both locations, profiling was carried out in the cenote pool (gray) and within the flooded cave environment (continuous line for salinity and dotted line for DO).At $ 5-m water depth, the low-salinity-halocline separates the fresh groundwater (< 1 psu; FW) from the brackish groundwater (BW), which is separated by the high-salinity-halocline from the saline groundwater (SW).Orange arrow shows the OsmoSampler (OS) deployment depth.
coastal Cenote Na'ach Wennen Ha site ($ 1 km from the coast) with salinities varying from as low as 0.5 psu in shallow depths to polyhaline (24.5 psu) at the high-salinity-halocline at $ 12 m.Here, passage morphology did not permit full access to the saline groundwater.DO concentration always remained below 0.7 mg L À1 in the meteoric lens but reached up to 1.3 mg L À1 in the saline groundwater near the highsalinity-halocline before decreasing again with depth (Fig. 2).

Discrete samples
Discrete samples allowed us to measure the δ 13 C-CH 4 before and after the OsmoSampler deployment at both sampling sites (Fig. 3; Table 1).At the beginning of the August 2014 deployment at the inland Cenote Odyssey, the highest methane concentrations (from 1055 nM to 7894 nM) and lowest δ 13 C values (À62.6‰ to À60.4‰) occurred in the fresh groundwater.By contrast, in January 2015 at the end of the OsmoSampler deployment, methane was 13 C-enriched (À49.0‰ to À38.0‰) and substantially lower in concentration (ranging from 77 nM to 143 nM).In comparison to the inland site for August 2014, methane concentrations at Cenote Na'ach Wennen Ha were considerably lower (ranging from 69 to 177 nM) and 13 C-enriched with δ 13 C values ranging from À57.5‰ to À51.5‰.Based on the one  measurement in January 2015 at the end of the OsmoSampler deployment, the dissolved methane concentration was higher (1057 nM), but the δ 13 C value of À57.0‰ was similar to initially measured values at the coastal Cenote Na'ach Wennen Ha site.

Meteorological and hydrologic records
The rainfall record showed seasonal and sub-seasonal variations consistent with the regional weather patterns (Figs.4a and 5a are identical but shown twice for ease of viewing).From this time-series, August to October is defined as the wet period, and November to January is consistent with dry conditions based on the definition that a typical dry season is characterized by monthly precipitation < 60 mm and < 4% of annual rainfall in successive months (Kottek et al. 2006).The largest measured precipitation event (225 mm on 22 October 2014) throughout the record was associated with the rainbands of the tropical depression during the development of Tropical Storm Hanna, despite the study area not being in the storm track (Cangialosi 2014).The 225-mm rainfall event constitutes $ 30% of the total precipitation measured during the 5-month monitoring period, which was 764 mm (wet season 678 mm and dry season 86 mm), and $ 15% of the 1500 mm mean annual precipitation in the region.
During the development of Tropical Storm Hanna, dramatic changes were registered by data loggers at the inland (Fig. 4a,b) and coastal sites (Fig. 5a,b).Changes in water level occurred on subtidal and tidal to seasonal scales at both sampling locations.At the inland Cenote Odyssey, the water level was on average 0.4 m below the full record's average before the storm, while it was 0.2 m above this average after the storm (Fig. 4b).Despite the rise in water level that inherently increases the hydraulic head, horizontal flow remained below detection limit at Cenote Odyssey throughout the time-series, suggesting the devices were not deployed in the portion of the passage where flow presumably increased.
Anoxia ($ 0 mg L À1 ) prevailed for most of the DO record in the fresh groundwater, but DO pulses occurred during the tropical cyclone and a subsequent precipitation event (Fig. 4b).Dissolved oxygen concentrations reached highest value (0.77 mg L À1 ) on 23 October 2014 following the intense rainfall and then gradually decreased over a 2-week period to undetectable concentrations by 4 November 2014.Another, but significantly smaller, increase in DO (up to 0.17 mg L À1 ) coincided with a 2-day rainfall (35 mm in 48 h) on 19-20 December 2014.Then DO concentrations gradually returned to anoxic within a week.
The water level record from the coastal Cenote Na'ach Wennen Ha site (Fig. 5b) shows similar patterns to the inland site, with 0.4 m below average and 0.2 m above average before and after the intense rainfall, respectively.Here, flow velocity was also below detection limit throughout the record with the exception of a dramatic increase (up to 3.92 cm s À1 ) during the tropical cyclone on 22 October 2014.

Chemical records
The OsmoSampler deployed at the inland Cenote Odyssey location provided a 122-day record of water chemistry (14 September 2014 and 14 January 2015) (Fig. 4c).This record is shorter than the actual deployment period (15 August 2014 and 14 January 2015), because the coil overfilled with cave water thereby expelling the early portion of the record.At the coastal Cenote Na'ach Wennen Ha site, the $ 149-daylong water chemistry record covered the entire deployment period from 18 August 2014 to 14 January 2015 (Fig. 5c).
Methane concentrations and salinities analyzed from discrete samples before and after the OsmoSampler deployment period were consistent with the first and last measurements from the OsmoSampler records at both locations (Figs. 3, 4c, 5c).
There was a striking drop in methane concentration that coincided with a sudden increase in salinity (from 0.67 to 2.96 psu) at the inland site.A co-occurring drop in methane concentration from 2566 nM (before the storm) to 221 nM (after the storm) was also observed at the coastal site, but it is less evident, because methane concentrations were greatly variable here and they started to decrease weeks before the event.
During the drop in methane, the salinity dropped from 0.74 to 0.44 psu, which is the lowest measured salinity value in these records.The sudden changes in salinity provide evidence that the drop in methane coincided with spatially widespread mixing between fresh and saline waters when the hydraulic head increased in October 2014.The records before and after the storm are hereafter referred to as the pre-storm and post-storm period, respectively.
At the inland Cenote Odyssey site, the average salinity was 0.71 AE 0.05 psu (n = 73) and average methane concentration was 4137 AE 888 nM (n = 33).The mean pre-storm methane concentration (10,635 AE 934 nM, n = 11) was significantly higher (p < 0.05, t-test) than the post-storm values (887 AE 291 nM, n = 22) (Table 2).This mean value does not include the single extremely high methane concentration anomaly (44,969 nM) observed in November 2014 during the otherwise low-methane period (Fig. 4c).This outlier is more than three times higher than the highest methane concentration previously measured in flooded caves in the region (Brankovits et al. 2018).We do not have a cogent explanation for a process that would be responsible for this anomaly, but we also do not have any evidence to discount it from the dataset.
Compared to the inland site, the coastal Cenote Na'ach Wennen Ha record had a lower mean methane concentration (1131 AE 238 nM, n = 31), higher mean salinity (1.11 AE 0.07 psu, n = 60), and greater variability (methane ranged from 106 to 5572 nM; salinity ranged from 0.42 to 2.70 psu) throughout the  4).
record (Fig. 5c; Table 2).Mean methane concentration was significantly higher (p < 0.05, t-test) prior to Tropical Storm Hanna at the coastal site (1862 AE 491 nM, n = 13) than after (604 AE 103 nM, n = 18).Salinity-methane plots using the time-series data points demonstrate shifts between the pre-and post-storm mixing regimes at both sites (Fig. 6).Methane concentrations during the storm and the following 2-week period, at the inland Cenote Odyssey site, were between the pre-and post-storm time-series values (Fig. 6a) and discrete sample values (Fig. 3a).Furthermore, we utilized conservative mixing calculations that are typically employed along estuarine (e.g., Chanton and Lewis 1999) or pore water (e.g., Pohlman et al. 2008) salinity gradients to differentiate the effects of physical mixing and removal/addition by biological processes.In this case, we exchanged space for time in our model by assuming the salinity variation that occurred through time in this subterranean estuary represents the same mixing and reaction dynamics that would occur in space in a surface estuary.Using equations described by Brankovits et al. (2017), the endmembers for the model were the highest salinity (salinity = 2.70 psu; CH 4 = 107 nM) and the lowest salinity (salinity = 0.56 psu; CH 4 = 5572 nM) values observed during the pre-storm period at the coastal Cenote Na'ach Wennen Ha site.The resulting mixing model, which shows non-conservative mixing of methane, reveals methane removal as salinity fluctuates at the coastal site in the wet season even before the storm's impact (Fig. 6b).

Methane sources
Methane in the low salinity groundwater at the inland Cenote Odyssey site, located $ 2 km from the coast, had considerably higher concentrations (as high as 7894 nM) and more negative stable carbon isotope values (as low as δ 13 C = À62.6‰)than the methane in Cenote Na'ach Wennen Ha $ 1 km from the coastline (Fig. 3).The 13 Cdepleted stable isotope values are consistent with methane being produced by methanogenic microbes (Whiticar 1999).This observation supports the hypothesis that methane is formed during the microbial decomposition of tropical forest vegetation preceding its accumulation in the cave environment in the region (Brankovits et al. 2017).Diminishing methane concentrations and enrichment of δ 13 C values with increasing salinity (and depth) in 2014 (Fig. 3a,b) are consistent with microbial consumption of methane by methanotrophs in the water column (Brankovits and Pohlman 2020).In the fresh groundwater, the initially high methane concentrations (5400 AE 2181 nM) with depleted δ 13 C values (À 62.6‰ AE 0.6‰) from 2014 were in sharp contrast with the reduced concentrations (109 AE 30 nM) and 13 Cenriched (À 44.0‰ AE 2.4‰) methane from 2015 (Fig. 3; Table 1).The concomitant shift toward lower concentrations and enriched δ 13 C values are most consistent with microbial methane oxidation having consumed (Barker and Fritz 1981;Whiticar 1999) a significant portion of the accumulated methane between August 2014 and January 2015.
Interestingly, methane δ 13 C values in the saline groundwater were always below À50.1‰, even in 2015 when methane in the fresh groundwater was especially 13 C-enriched (ranges from À49.0‰ to À37.1‰).As a result, in 2015, methane became continuously more 13 C-depleted with increasing salinity (and depth) in the water column, a trend opposite to 2014 (Fig. 3b).However, if the methane in the saline groundwater were primarily a residue from the fresh groundwater methanogenic source, methane δ 13 C values in the saline groundwater would be either the same or-in case of oxidation-more 13 C-enriched as those values from the fresh groundwater.Consequently, a top-down methane source from the freshwater may not have been the only source of methane Wennen Ha; NWH).Time-series are marked based on the period relative to the storm: pre-storm (orange circles), storm event and following 2 weeks (yellow triangles), and post-storm (gray squares).A conservative mixing line is shown for the pre-storm period (orange dotted line).
in the saline groundwater in 2015.Rather, a portion of this methane likely originated from a methanogenic source independent of the meteoric waters.It is possible that deep circulation of saline groundwater (e.g., Beddows et al. 2007) introduced microbially produced methane generated within the deeper, inaccessible, and anoxic portions of the aquifer or that in situ aerobic methanogenesis was occurring (e.g., Bogard et al. 2014;Repeta et al. 2016).Although the aerobic methanogenesis is generally associated with elevated primary production in surface waters of lakes and oceans, differentiating these potential sources is a topic that merits further investigation.At the coastal Cenote Na'ach Wennen Ha site, methane was generally lower in concentration and more 13 C-enriched in its isotopic content than at the inland location (Fig. 3; Table 1).Among the discrete samples, even the highest measured concentration of methane (1054 nM) had a more positive δ 13 C value (À57.0‰)than the source methane values further inland (Fig. 3).These concentrations and stable carbon isotope values suggest that methane was a partially oxidized residue from a microbial methane source during these sampling periods; either from the fresh or the saline groundwater, or a mix of both.The time-series, however, reveals elevated methane concentrations (as high as 5570 nM) during October 2014 when salinity dropped below 1 psu (Fig. 5c), demonstrating that environmental conditions can facilitate methane accumulation at the coastal site.Due to its proximity to the ocean, the coastal site is exposed to greater mixing between the meteoric and saline waters, such as the saline groundwater or intruding marine water (Fig. 1c), that likely enhances methane oxidation (and limits methane production) over time.

Periods of methane accumulation
The 5-month-long, 2-to 5-day resolution chemical records are consistent with the vertical geochemical profiles from the water column in that methane concentrations were significantly higher at the inland site (Fig. 4c) than at the coastal site (Fig. 5c).During the $ 2-month-long pre-storm period, the inland site was characterized by low salinity, persistent anoxia, and the highest consistent methane concentrations across all records (Fig. 4; Table 2).The persistently high methane concentration (on average, 10,635 AE 934 nM, n = 11) indicates methane was continuously generated during the prestorm period.Despite occasional rainfall in the wet season, the only measurable flow or water level changes were observed at tidal and subtidal intervals (Fig. 4).
Further inland from the coast (6.3 km) at Cenote Bang and along the same transect in the Ox Bel Ha Cave System (Fig. 1b), methane accumulated (5949 AE 132 nM) in the freshwater at comparable levels to this study's inland site and methane concentrations decreased during heavy rainfall (Brankovits et al. 2018).Collectively, these findings suggest that methane accumulation in the karst subterranean estuary is spatially controlled, with methane accumulating to the greatest extent away from the coastline (> 2 km) in the fresh groundwater during stable hydrologic conditions.
In contrast to the inland region, the average methane concentration was lower (1862 AE 491 nM, n = 13) and substantially more variable through time at the coastal site (Fig. 5).Temporal fluctuation in salinity (between 0.5 and 2.7 psu) implies more active mixing between fresh and saline groundwaters near the coastline, perhaps a result of tidal-pumping and the shallower depth of the high-salinity-halocline.During the pre-storm period, lower salinities were always associated with higher methane concentrations (Fig. 5c).This relationship indicates that fresh groundwater was also the dominant source of dissolved methane at the coastal location.However, non-conservative mixing of methane during the pre-storm period (mixing model on Fig. 6b) indicates continuous methane consumption as salinity fluctuates.This observation is consistent with the discretely collected methane δ 13 C values from the beginning of this period being a partially oxidized residue (Fig. 3b).
An increased input of oxygenated saline water at the coastal site likely shifts redox conditions from net methanogenic to net methanotrophic.In the absence of oxygen, sulfate-rich saline groundwater can inhibit methanogenesis (Jørgensen 1982;Jørgensen and Parkes, 2010) and/or facilitate anaerobic methane oxidation using sulfate as an electron acceptor (Boetius et al. 2000).Recently, sulfatedependent anaerobic methane oxidation has been observed in conditions with low salinity and sulfate concentrations (Mostovaya et al. 2022).However, if the increased salinity is from oxygenated marine-derived water, it will create ideal conditions for methane oxidation by aerobic methanotrophy (Niemann et al. 2006;Steinle et al. 2017).Regardless of whether it was aerobic or anaerobic methanotrophy, the stable isotope evidence from discrete sampling corresponds with the methane time-series in that microbial methane consumption in the subterranean estuary was more effective at the coastal site, likely due to greater marine influence and/or greater level of mixing with saline groundwater, than further inland.For the same reason, more methane built up at the inland regions during hydrologically stable periods.

Effects of a tropical cyclone on carbon cycling
The storm-induced hydrologic mixing was followed by a decline in methane from 15,651 to 685 nM in less than 10 d at the inland site (Fig. 4) and a drop from 2566 to 221 nM in less than 5 d at the coastal site (Fig. 5).In the Ox Bel Ha Cave System, temporary reduction in methane is known to be driven by a combination of physical processes (i.e., dilution and hydrologic transport) and biogeochemical processes (i.e., microbial methane oxidation) in response to seasonal rainfall-induced mixing of different groundwater regimes (Brankovits et al. 2018).The current chemical records, however, reveal a dramatic shift in methane concentration that did not recover to pre-storm levels during our observation period (Figs. 4,5;Table 2), indicating that the large-scale rainfall associated with the development of Tropical Storm Hanna, which tracked more than 220 km south of the study area (Cangialosi 2014), had a lasting impact on methane accumulation in the subterranean estuary.
At the inland site, the storm-induced increase in water level and dissolved oxygen coincided with a salinity pulse and decreasing methane concentrations (Fig. 4).The salinity increase (from consistent $ 0.67 psu to as high as 2.96 psu) implies shear-induced vertical mixing of saline groundwater into the meteoric waters (Coutino et al. 2017).This process suggests methane in the brackish water was at least partially diluted from mixing with the saline groundwater.The continuous decline in salinity to near pre-storm levels ($ 0.70 psu) in the approximately 2 weeks following the event is consistent with the reestablishment of stratified water masses.Although methane increased slightly during the same period, its concentration remained well below pre-storm levels in the entire post-storm period (Figs.4c, 6a), suggesting methane removal by oxidation, in addition to dilution.
The spike in DO from non-detectable values (i.e., $ 0 mg L À1 ) to as high as 0.77 mg L À1 (Fig. 3) provides clear evidence that oxygenated waters were forced into the anoxic methane-charged fresh groundwater during the tropical storm.Because such hypoxic DO levels are sufficient for aerobic methanotrophs to be active (Steinle et al. 2017), it likely facilitated a hot moment of methane oxidation using oxygen as the electron acceptor.Following the event, DO concentrations vanish to baseline anoxic levels in about 2 weeks.The continuous decline in DO most likely results from the microbial oxidation of methane and other organic carbon forms.
The episode of enhanced methane oxidation, driven by the temporal availability of oxygen, provides a plausible biogeochemical explanation for the shift in geochemical parameters within the water column from August 2014 to January 2015 (i.e., reduction in methane concentration coupled with 13 Cenrichment; Fig. 3).The magnitude and spatial extent of the oxidation event in combination with the prolonged dry conditions after the storm could have restructured microbial communities, like in surface estuaries (Steichen et al. 2020;Yan et al. 2020), and, thus, limited the replenishment of methane in the months following Tropical Storm Hanna.
Considering the co-occurring spike in DO and salinity during the event (Fig. 4), it is reasonable to assume the saline groundwater contributes oxygen to the otherwise anoxic and fresh portion of the upper aquifer.Indeed, highest DO concentration in the cave environment's water column was documented in the saline groundwater at the high-salinityhalocline at both sampling sites (Fig. 2).Similar DO profiles were recorded in consecutive seasons further inland in the Ox Bel Ha Cave System (Fig. 1b) (Brankovits et al. 2017) and in karst subterranean estuaries in other geographical regions (Sket 1996).Dissolved oxygen transferred by tidal-pumping into cave conduits from the marine environment is quickly depleted to lower concentrations due to respiration within the karst subterranean estuary (Young et al. 2018).Nevertheless, the continuous influx of marine water just below the highsalinity-halocline may reach far inland on the Yucatan Peninsula (Beddows et al. 2007), which could serve as a hydrologic mechanism that maintains elevated DO concentrations compared to overlying and underlying water layers in this oligotrophic setting.
A scenario of conservative mixing with respect to oxygen between the anoxic freshwater (DO = 0 mg L À1 ; salinity = 0.7 psu) and the hypoxic saline groundwater (DO = 1.3 mg L À1 ; salinity = 36 psu) (Fig. 2) would have increased DO by 0.1 mg L À1 when salinity reached a maximum of 3 psu during the mixing event.The hypothetical 0.1 mg L À1 DO value corresponds to only about 13% of the total 0.8 mg L À1 DO measured during mixing and serves as a maximum estimate to what the saline groundwater could have contributed to the DO spike following the tropical storm.The remaining $ 87% (0.7 mg L À1 ) of the DO must have come from an external source.The magnitude of the recharge likely reached a threshold that allowed atmospheric oxygen to be injected into the coastal aquifer.
At the coastal site, the episodic drop in methane concentration coincided with a decline in salinity during the storm (Fig. 5).Salinity went from oligohaline (0.74 psu) to fresh (0.44 psu), which is the lowest measured salinity among all records in the study.Simultaneously, water level rose, and flow velocity increased (Fig. 5b).These observations are consistent with recharge of air-equilibrated rainwater into the aquifer during the storm.
An increased lateral flow likely serves as a mechanism for solutes, including methane, to be emitted to the coastal ocean by discharging groundwater (Taniguchi et al. 2002;Burnett et al. 2006).Methane in the adjacent sea, near the coastline, is elevated in concentration (121 AE 28 nM, n = 6) and 13 Cdepleted (À 59.0‰ AE 2.1 ‰, n = 5) relative to atmospheric equilibrium (Brankovits et al. 2017).These values are comparable to methane at the nearby Cenote Na'ach Wennen Ha site (Table 1) and shallow coastal waters in the Gulf of Mexico where groundwater discharges methane (Bugna et al. 1996;Cable et al. 1996).Therefore, the northeastern Yucatan Peninsula subterranean estuary is a likely source of methane to the coastal ocean.Nevertheless, evidence provided here demonstrates that increased episodic mixing accelerates microbial methane oxidation that will in fact limit methane fluxes with groundwater discharge.
Considering that methane-charged fresh groundwater is also a source of terrestrially derived dissolved organic carbon (DOC = as high as 834 μM) and these constituents are key carbon and energy sources for the subterranean ecosystem in the region (Brankovits et al. 2017), enhanced oxidation events due to tropical cyclones may have a substantial role in accelerating the transfer of carbon into the coastal aquifer food web.Analog processes occur in surface estuaries, where organic matter mobilized during flooding events expedites the seaward transport of carbon while it enhances microbial respiration and, thus, accelerates the transformation of terrestrial dissolved organic matter (Bianchi et al. 2013;Osburn et al. 2019;Yan et al. 2020).However, unlike in surface estuaries, the months-long drop in methane concentration observed here suggests that karst subterranean estuary habitats may experience extended periods of food and energy shortage.Whereas these findings are supported by the presence of cavedwelling shrimp and other specialized invertebrate fauna that are adapted to sustained periods of starvation, they contradict the long-held idea of steady environmental conditions related to coastal caves (Sket 1996).Taken together, our findings demonstrate that storm-induced biogeochemical changes impact both the carbon cycle and ecosystem functioning in karst subterranean estuaries.
Given that karst platforms dominate the Caribbean regions coastal and island landscapes, where tropical cyclones are increasingly prevalent (Lehmann and Coumou 2015;Emanuel 2017), our work highlights the need to better understand critical thresholds required to mobilize organic carbon and, at the same time, drive oxygen into the coastal aquifer from the atmosphere or the marine environment.How the karst subterranean estuary ecosystem, including its carbon cycle, would react to storm surge (e.g., Giambastiani et al. 2017) or a direct hit from a tropical cyclone remains unknown.Further research efforts could focus on assessing the balance between enhanced oxidation of methane, and other forms of dissolved organic carbon, and the increased export of these reduced carbon compounds from the karst subterranean estuary in response to varying magnitude of meteorological events.

Conclusions
Concurrent 5-month hydrologic and chemical records from 1 and 2 km inland from the coastline in the northeastern Yucatan Peninsula's coastal aquifer reveal higher methane accumulation further inland during hydrologically stable periods.Lesser accumulation near the coastline is most likely due to methane consumption supported by tidally enhanced mixing of oxygenated saline waters with methane-enriched anoxic fresh groundwater.Both records captured the wet and dry season and documented the effects from a tropical cyclone.The intense rainfall resulted in a spatially widespread hydrologic event that forced and mixed oxygenated waters into the methane-charged portion of the aquifer, stimulating microbial methane oxidation.Cave conduits are corridors for the distribution and transport of dissolved oxygen and organic carbon in coastal aquifers on karst landscapes.Oxygen added into the anoxic low-salinity upper aquifer during the tropical cyclone originated from both saline groundwater by upward vertical mixing and rainwater by recharge.In the 2.5 months following the storm, methane concentrations did not recover to pre-storm levels, suggesting that large-scale rainfall events have lasting effects on the karst subterranean estuary carbon cycle.Considering that methane and other forms of reduced carbon are food and energy sources for coastal aquifer food webs, tropical cyclones may have a critical role in accelerating carbon transfer into the food web and, thus, dictating pronounced fluctuations in nutrient availability within karst subterranean estuaries.

Fig. 1 .
Fig. 1.(a) The Holbox Fracture Zone and Xel Ha Zone (gray area with dotted line) of the northeastern coast of the Yucatan Peninsula in Mexico contains complex networks of shoreline-perpendicular cave passages perforated by surficial sinkholes (cenotes).(b) The Ox Bel Ha Cave System has more than 319 km of mapped conduits and 150 associated cenotes.(c) These solutional karst features influence the subsurface hydrology where the meteoric lens (low salinity groundwater) overrides the marine-derived saline groundwater in the unconfined aquifer.Cenote Na'ach Wennen Ha (NWH; 1 km inland) and Cenote Odyssey (ODY; 2 km inland) are the sites for this investigation.

Fig. 2 .
Fig.2.Salinity and dissolved oxygen (DO) water column profiles from (a) Cenote Odyssey (ODY) and (b) Cenote Na'ach Wennen Ha (NWH) in August 2014 (purple) and January 2015 (green).At both locations, profiling was carried out in the cenote pool (gray) and within the flooded cave environment (continuous line for salinity and dotted line for DO).At $ 5-m water depth, the low-salinity-halocline separates the fresh groundwater (< 1 psu; FW) from the brackish groundwater (BW), which is separated by the high-salinity-halocline from the saline groundwater (SW).Orange arrow shows the OsmoSampler (OS) deployment depth.

Fig. 3 .
Fig. 3. Concentrations and δ 13 C values of methane (CH 4 ) from discrete samples obtained from Cenote Odyssey (ODY) and at Cenote Na'ach Wennen Ha (NWH) in the wet season in August 2014 (light orange) and the dry season in January 2015 (dark blue).The collection times overlap with the beginning and end of the chemical records obtained with the OsmoSamplers.The samples were collected from the stratified water column along depth profiles that cover the full salinity gradient.Salinity is used as a proxy for depth to show the relative position of the fresh (FW), brackish (BW), and saline (SW) groundwater masses.Only one sample was collected from the coastal site in 2015.

Fig. 4 .Fig. 5 .
Fig. 4. (a) Precipitation record from the Tulum meteorological station during OsmoSampler deployment at the inland Cenote Odyssey site.(b) Water level, dissolved oxygen (DO), and flow velocity.(c) Methane concentration (orange circle) and salinity (blue-filled circle).The gray open circles are discrete samples collected near the OsmoSampler before and after the deployment.

Fig. 6 .
Fig. 6.Salinity vs. methane plots using time-series data points from (a) the inland site (Cenote Odyssey; ODY) and (b) the coastal site (Cenote Na'ach

Table 2 .
Brankovits et al. (2022)records.Records were obtained from the inland site (Cenote Odyssey) and the coastal site (Cenote Na'ach Wennen Ha).Pre-storm period represents the chemical record prior to Tropical Storm Hanna while post-storm represents the period after.Data fromBrankovits et al. (2022).
*Average concentration excludes the 45 μM methane anomaly post-storm in Cenote Odyssey (see Fig.