Since the start of the 20th century, anthropogenic production of N r species has increased by a factor of 10, largely from fossil fuel burning and industrial application of the Haber-Bosch process (Galloway and Cowling, 2002;Galloway et al., 2004). Since the end of the 20th century, anthropogenic fixation of atmospheric N 2 has exceeded that of natural terrestrial N 2 fixation (Galloway and Cowling, 2002). The subsequent accumulation of anthropogenic N r has important ecological implications through eutrophication and enhancing primary productivity in aquatic systems. Additional environmental impacts result from processing of N r in aquatic ecosystems including anoxia, which creates conditions favorable for the microbial transformation of N species such as N 2 O.
N 2 O is a powerful greenhouse gas with an atmospheric lifespan of approximately 121 years and a warming potential about 265 times that of carbon dioxide (CO 2 ) and 10 times that of methane (CH 4 ) (Myhre et al., 2013). In addition to its greenhouse warming potential, N 2 O has been the single most important stratospheric ozone depleting substance emitted in the 21st century (Ravishankara et al., 2009). Since preindustrial times, global N 2 O emissions have increased approximately 20% in parallel with increased N r abundance (Galloway et al., 2008;Butterbach-Bahl et al., 2013), with emissions projected to double by 2050 (Davidson & Kanter, 2014). Nearly 40% of total N 2 O emissions are anthropogenic in nature and originate from a combination of fossil fuel combustion, industrial activities, and agriculture, with the latter estimated to be the largest contributor (Jurado et al., 2018). Although N 2 O is produced by abiotic reactions and microbial metabolisms, production through microbial denitrification and nitrification are the greatest contributors to global N 2 O emissions (Thomson et al., 2012).
Karst aquifers have physicochemical characteristics that are ideal for N r processing, including extensive and rapid surface water-groundwater interactions and highly heterogeneous permeability that leads to spatially variable redox conditions (Ford and Williams, 2007). Karst landscapes cover approximately 20% of Earth's ice-free land surface (Goldscheider et al., 2020), and depending on the extent of N r processing, they could significantly effect global N 2 O cycling. However, few studies to date have investigated N 2 O cycling within carbonate aquifers despite their large spatial distribution and potential for N r processing. This processing would be affected by exchange of surface water and groundwater, which controls availability of electron donors (e.g., organic carbon, reduced Fe, Mn, S; NH 4
) and electron acceptors (e.g., NO 3 -; O 2 ), aquifer redox conditions, and abiotic and/or microbial reactions.
N 2 O is an intermediate of denitrification and is produced when NO 3 is microbially respired and sequentially reduced to N 2 . N 2 O production via denitrification has been reported under anoxic to oxic conditions for karstic groundwater (Albertin et al., 2012;Heffernan et al., 2012;Jahangir et al., 2013;Henson et al., 2019). Nitrification is an obligately aerobic process in which N 2 O is produced during NH 4 + oxidation to NO 2 as a by-product during abiotic decomposition of the intermediate species hydroxylamine (NH 2 OH) (Thomson et al., 2012). Nitrification has been hypothesized to be active in karstic groundwater and contribute to elevated N 2 O concentrations (Ueda et al., 1993;Mühlherr and Hiscock, 1998;Hiscock et al., 2003;Jurado et al., 2018), but direct evidence has been lacking. In this study we addressed three primary questions: 1) could karstic springs be an atmospheric source of N 2 O; 2) do N 2 O concentrations change with space and time across a karst landscape; and 3) how much N r (primarily NO 3 -) is reduced to N 2 O? The study area is the karstic Upper Floridan Aquifer (UFA) in north-central Florida, USA (Fig. 1), where previous studies provide much information on the timing of surface water and groundwater exchange and subsurface residence times of water discharging from multiple springs. Analyses of the chemical compositions of the spring water reveal systematic relationships between the residence times, organic carbon contents, and N 2 O concentrations that point toward potential controls on N 2 O production and consumption.
The study area is within the Suwannee River watershed and is underlain by the carbonate Eocene to Oligocene carbonate strata (Fig. 1A-B). These strata include the Suwannee Limestone (10-30 m thickness), Ocala Limestone (20-80 m thickness), and Avon Park Formation (Sutton et al., 2015). Within the watershed, the UFA is unconfined in the lower reaches and confined in the upper reaches by the Miocene Hawthorn Group. The boundary between the confined and unconfined portion of the UFA, called the Cody Escarpment (Puri and Vernon, 1964), is the erosional edge of the Hawthorn Group and is the site of extensive surface water-groundwater exchange (Fig. 1A). All streams flowing across the escarpment either sink completely into the subsurface or become losing streams.
The UFA is characterized by intergranular primary porosity of about 20% and can be classified as an eogenetic karst aquifer (Vacher and Mylroie, 2002). This primary porosity, with an average matrix permeability of 10 -13 m 2 for the Ocala limestone, can provide storage for recharged surface water during flooding events (Florea and Vacher, 2006). At base flow, the matrix porosity may provide 25-50% of groundwater flow to the numerous springs that discharge from waterfilled caves in the UFA downstream of the Cody Escarpment (Rosenau et al., 1977;Scott et al., 2002;Ritorto et al., 2009;Yang et al., 2019). Regional groundwater flow across the watershed is predominately toward the southwest and includes both slow matrix and fast conduit/ fracture flow. We focus on 13 springs that can be separated into three groups with variable rates of surface water-groundwater interactions, chemical compositions, and groundwater residence times.
One group of springs, the Ichetucknee springs group (Fig. 1C), includes eight named springs and numerous unnamed smaller springs that discharge <1 to 5.6 m 3 /sec to the Ichetucknee River (Scott et al., 2002). This group has the lowest DOC concentrations among the studied spring groups (Table 1). The Ichetucknee springs group can be divided into two sub-groups based on their chemical compositions (Martin & Gordon, 2000) and apparent ages (Martin et al., 2016). Group 1a springs (Head Spring, Blue Hole, Cedar Head, and Coffee Springs) have higher dissolved oxygen (DO) concentrations and more variable temperatures (ΔT 0.3-0.5 • C) than group 1b springs (Mission Springs, Devil's Eye, Mill Pond, and Grassy Hole). The mean apparent ages based on CFC-12 concentrations of group 1a springs (35.08 ± 0.20 years) is younger than group 1b springs (40.47 ± 0.28 years; Table 2). These characteristics suggest that group 1a springs have shorter and shallower flow paths than group 1b springs (Martin & Gordon, 2000).
The second group of springs are referred to as "reversing springs" (Peacock, Madison Blue, Little River, and Gilchrist Blue) because river water with DOC and DO concentrations greater than groundwater values (Table 1) periodically intrudes through the spring vents during high flow conditions (Gulley et al., 2011;Brown et al., 2014Brown et al., , 2019)). These reversals occur once or twice a year and flood water may reside in the aquifer for days to months before draining back to the surface as the floods recede. Of this group, mean apparent age has been measured only for Little River Spring. The average apparent age, which is based on CFC-12 concentrations, is ~21 years (Katz et al., 2001;Heffernan et al., 2012; Table 2). This apparent age is likely much longer than water that discharges immediately following a reversal, when residence times may be days to months (e.g., Gulley et al., 2011;Brown et al., 2014). As DOC and DO are biogeochemically processed in the aquifer, groundwater redox state shifts as reflected by variable Fe and Mn concentrations in water discharging from Madison Blue Spring following reversals (Brown et al., 2019).
The third group of springs are part of a Sink-Rise system that occurs where the Santa Fe River crosses the Cody Escarpment (Fig. 1D). Except during flooding, all river water is captured by a sinkhole called River Sink. The river briefly resurges at several karst windows, most prominently at Sweetwater Lake, before discharging permanently at River Rise Spring approximately 7 km downstream from River Sink (Scott et al., 2002). Water recharged at River Sink requires from 6 to 185 h to flow to River Rise depending on river stage (Fig. S1) and River Rise discharge is commonly greater than discharge into River Sink. The downstream increase in flow reflects water gained from matrix porosity to the conduits when the groundwater hydraulic head exceeds that of the conduits, which should increase average residence times and specific conductivity (SpC) of the discharging water at River Rise (Martin and Dean, 1999;Moore et al., 2009). River Sink discharge can exceed River Rise discharge during high flow events when the conduit hydraulic head is greater than that of the groundwater. During these conditions, water is assumed to be lost from conduits to matrix porosity (Martin and Dean, 1999;Bailly-Comte et al., 2010, 2011).
Water was pumped directly from spring vents using a Geotech peristaltic pump and tubing that extended to the shore. At the Santa Fe Sink-Rise system, water was pumped through an overflow cup containing sondes connected to a YSI ProPlus meter that was calibrated daily. At all other locations, the sondes were deployed directly in the path of discharge above spring vents. Measured parameters were dissolved oxygen concentrations (DO% saturation and mg/L), specific conductivity (µS/cm), and temperature ( • C). Parameters were monitored until values stabilized, typically within a few minutes, after which the physicochemical parameters were recorded, and sampling commenced.
Samples to be measured for total DOC, total dissolved nitrogen (TDN = inorganic-N + organic-N), NO 3 -, and SO 4 2-concentrations were filtered through 0.45-µm trace metal grade canister filters. The DOC and TDN samples were preserved with concentrated HCl and nitrate samples were frozen until analysis. N 2 O samples were collected via the headspace extraction method (Pain et al., 2019e.g., Repo et al., 2007) in a 650 mL vessel with 60 mL of sample water displaced with ultra-high purity grade helium (UHP; 99.999% purity) or N 2 (UHP) and shaken to equilibration for 3 min. The 60 mL of headspace gas was immediately transferred to pre-evacuated 75 mL glass vials and analyzed within 1 week of collection.
The N 2 O concentrations were measured at the University of Florida with an Agilent Gas Chromatograph (7820-A) equipped with a µ-ECD (electron capture detector -63 Ni source, 350 • C, makeup gas 5% CH 4 / Argon mixture) and an Agilent J&W GS-CARBONPLOT column (30 m length, 0.320 mm diameter widebore, 3.00 µm film) regulated at a temperature of 30 • C and UHP N 2 as the carrier gas. Calibration standards were prepared by diluting a 0.9700 ppm N 2 O standard in a He or N 2 matrix to 0%, 25%, 50%, 75%, 100% N 2 O. Dilutions were made fresh before each analysis by injecting gases directly into pre-evacuated 75 mL glass vials. Gas concentrations in the headspace samples were converted to dissolved concentrations according to Weiss et al., 1980 and based on the temperature and salinity of the water. All N 2 O samples were collected in triplicate to assess the relative error of the head-space extraction collection method, which generated a relative standard deviation of <0.2 µg N-N 2 O/L. The saturation of N 2 O was calculated as a percentage relative to atmospheric equilibration with water using the method reported by Cooper et al. (2017)
Other redox sensitive solutes, including ferrous iron (Fe 2+ ) and hydrogen sulfide (HS -), were measured in the field as initial samples were collected from all springs using a field spectrophotometer (Hach DR 890 portable colorimeter). Concentrations were below instrumental detection limits (0.01 mg/L for both Fe 2+ and SH -), suggesting these solutes provide little control on redox conditions of spring waters.
Chemical compositions of the two Ichetucknee spring sub-groups are similar to previous observations (Martin and Gordon, 2000;Martin et al., 2016) trations show no correlation (Fig. 3). Median values for EF NO3-range from 0.0007 to 0.0019 (Eq. ( 1)) and group 1b springs have lower median values (0.0009) than group 1a springs (0.0018). Median values of EF TDN (Eq. ( 2)) are similar to the EF NO3-value, ranging from 0.0009 and 0.0017 for sub-groups 1b and 1a, respectively (Table 1).
The reversing springs group has median SpC values and DO concentrations of 400 µS/cm and 1.83 mg/L, respectively. This spring group has higher NO 3 -concentrations than the other two spring groups, with a range from 1.9 mg N/L at Little River Spring to 4.2 mg N/L at Peacock springs (median = 2.4 mg N/L). DOC and SO 4 2-concentrations range from 0.26 to 4.6 mg C/L (median = 0.49 mg C/L) and from 4.1 to 7.9 mg S/L (median = 7.6 mg S/L), respectively. Although we lack the TDN concentration for the sample with the highest NO 3 -concentration in this group, all other TDN concentrations were higher than the other two groups and range from 1.9 to 3.8 mg N/L (median = 2.7 mg N/L) and are at most ~ 0.4 mg N/L greater than NO 3 -concentrations (Table 1). These springs exhibit N 2 O concentrations that range from 1.55 µg N/L (502% saturation) to 2.90 µg N/L (941% saturation) at Gilchrist Blue and Peacock springs, respectively (Fig. 2). Although N 2 O concentrations were generally lower in samples with higher DO concentrations (Fig. 4), these variables are not significantly correlated. In contrast, NO 3 -, DOC, and SO 4 2-concentrations increase with DO concentrations, although only the NO 3 --N 2 O linear correlation is significant. The average EF values are similar but slightly lower than those of the Ichetucknee springs group. The EF NO3-values range from 0.0007 to 0.0015 (Eq. ( 1)) while the EF TDN values range from 0.0006 to 0.0015 (Eq. ( 2)).
Water compositions at the Sink-Rise system show systematic variations from River Sink to River Rise that depend on discharge. Due to the continuous supply of surface water to the Sink-Rise system, it contains the highest DOC concentrations of all three spring groups. The Sink-Rise system also shows greater variance of all solute concentrations, including N 2 O, compared to the other two spring groups (Fig. 2). SpC values range from 57.2 to 359.7 µS/cm at River Sink, from 70.9 to 476.2 µS/cm at Sweetwater Lake, and 69.3 to 526.0 µS/cm at River Rise (Table 2). DO concentrations range from 2.61 to 9.30 mg/L at River Sink, from 0.96 to 7.83 mg/L at Sweetwater, and 0.65 to 7.16 mg/L at River Rise. Median NO 3 -concentrations increase from River Sink to River Rise, with concentrations of 0.12 mg N/L for River Sink, 0.20 mg N/L for Sweetwater Lake, and 0.23 mg N/L for River Rise. Unlike the other two spring groups, TDN concentrations within the Sink-Rise system reach up to ~ 9 times greater than NO 3 -concentrations, reflecting the presence of organic-N and/or other inorganic N r species, such as NO 2 -and NH 4
. Assuming a molar C:N ratio of 50:1 for terrestrial derived organic-C (Perdue and Koprivnjak, 2007), a median DOC concentration for River Rise of 23.4 mg C/L indicates the presence of ~ 0.6 mg org-N/ L, which indicates most of the excess TDN is organic N rather than NO 2 or NH 4 + , similar to the other spring groups. The highest DOC concentrations of all sample sites occur within the Sink-Rise system due to the continuous injection of surface water at River Sink, with median concentrations of 25.8 mg C/L for River Sink, 25.4 mg C/L for Sweetwater Lake, and 23.4 mg C/L for River Rise. Median SO 4 2-concentrations range from 3.3 mg S/L at River Sink, 6.4 mg S/L at Sweetwater Lake, and 11.5 mg S/L at River Rise. TDN concentrations at River Sink range from 0.2 to 1.3 mg N/L, from 0.4 to 1.1 mg N/L at Sweetwater, and from 0.3 to 1.2 mg N/L at River Rise. Most samples from River Sink, Sweetwater Lake, and River Rise have N 2 O concentrations that are supersaturated relative to the atmosphere, except during elevated discharge, when concentrations are near atmospheric equilibrium. Median N 2 O concentrations increase from 1.14 µg N-N 2 O/L (370% saturation) at River Sink, to 1.36 µg N-N 2 O/L (441% saturation) at Sweetwater, to 2.04 µg N-N 2 O/L (661% saturation) at River Rise Spring. This increase is up to ~ 2 µg N-N 2 O/L during most sampling times. In contrast with trends of increasing N 2 O concentrations along the flow path, four sampling times had N 2 O concentrations that decreased from River Sink to River Rise. These samples also contained the highest N 2 O concentrations measured during this study, ranging from 3.45 to 5.09 µg N/L (1120% to 1738% saturation). Within the Sink-Rise system, significant positive correlations occur between N 2 O and NO 3
-and SO 4 2-concentrations, while significant inverse correlations occur between N 2 O and DO and DOC (Fig. 5). During base flow conditions, when discharge is greater at River Rise than River Sink, EF NO3- values varied more than 10-fold with time, ranging from 0.0032 to 0.0177 with a median value of 0.0086. The EF TDN values are lower and range from 0.0007 to 0.0113 with a median value of 0.0042.
Over the past ~ 50 years, NO 3 -concentrations are estimated to have risen by a factor of 50 in the UFA, from ≤ 0.1 mg N-NO 3 L -1 to values as high as 5 mg L -1 (Katz, 2004;Albertin et al., 2012). Increasing NO 3 concentrations in the UFA have been traced back to multiple anthropogenic sources including fertilizer application (51%), animal waste (27%), septic tank drainage (12%), and natural atmospheric deposition (8%) (Katz et al., 2009). Nitrate was thought to have little or no attenuation in the UFA because of rapid flow combined with aerobic and organic carbon-poor conditions (Katz et al., 2009). This view has changed with observations of excess N 2 concentrations and stable isotope analyses of NO 3 -that indicate denitrification is widespread (Table 2) (Albertin et al., 2012;Heffernan et al., 2012;Henson et al., 2019). Therefore, our observations of N 2 O in springs discharging from the UFA may reflect incomplete denitrification of NO 3 -to N 2 O. Incomplete reduction of NO 3 -and accumulation of N 2 O rather than N 2 has been documented in soils and groundwaters due to prevailing oxic conditions (Osaka et al., 2006;Laini et al., 2011;Jahangir et al., 2013;McAleer et al., 2017) and high NO 3 -concentrations that inhibit N 2 O reduction to N 2 (Blackmer and Bremner, 1978;Weymann et al., 2008). However, the variations in N 2 O concentrations observed in this study indicate that the extent of denitrification may differ across karst landscapes with variable surface water-groundwater interactions, groundwater residence times, and availability of DOC. We evaluate these controls based on the spatial and temporal variations of water chemistry and N 2 O concentrations.
Despite variations in geochemical and hydrologic conditions among the three spring groups, all show a positive correlation between N 2 O and NO 3
-concentrations, implying that elevated NO 3 -concentrations result in greater N 2 O production. The NO 3 --N 2 O relationships differ between locations, however, the reversing springs group had the highest NO 3 concentrations and the Sink-Rise system had the highest N 2 O concentrations. These differences suggest factors other than NO 3 -concentrations contribute to the denitrification rate, and specifically, the reduction of NO 3 -and N 2 O. Within the Ichetucknee springs group, up to 32% of available NO 3 -is denitrified to N 2 as reflected in excess N 2 concentrations measured in spring waters (Heffernan et al., 2012) (Table 2). These concentrations are about 3 orders of magnitude greater than the measured N 2 O concentrations and suggest that within the Ichetucknee springs group, much, but not all NO 3 -is completely reduced to N 2 . Although excess N 2 data are not available for the Sink-Rise system, the negative correlation between N 2 O and DO and DOC concentrations (Fig. 5A,C) suggests denitrification may be the primary N 2 O producing pathway. Identifying timing and locations of denitrification could be complicated by mixing of the surface water and groundwater if the two sources have different N 2 O concentrations. In addition to water recharging at River Sink, groundwater sources to River Rise are water draining from the matrix porosity to conduits at depths ~ 30 m below land surface and a second minor source from about 400 m below the land surface (Moore et al., 2009;Jin et al., 2015). This deep source is likely anoxic with negligible NO 3 -and N 2 O concentrations, and thus unlikely to be a source of N 2 O. We assume controls of N 2 O concentrations are a result of shallow groundwater mixing with river water entering at River Sink coupled with production and consumption caused by varying redox conditions and rates of DOC remineralization within the matrix porosity (Fig. S2).
Distinguishing mixing from biogeochemical activity requires information on the mixing extent, which can be derived from conservative parameters such as SpC. Carbonate mineral dissolution increases groundwater SpC values by up to an order of magnitude more than river water (Gulley et al., 2011) springs group (Fig. 3A) opens the possibility that some of the N 2 O could be produced by denitrifying microbes using inorganic electron donors such as reduced Fe, Mn and/or sulfur species. Nitrate reduction by pyrite has been documented in laboratory-based microbial incubations and flow-through experiments according to: (Torrentó et al., 2010(Torrentó et al., , 2011)). This denitrification pathway is thermodynamically favorable and is hypothesized to be active across a range of geological and hydrological settings (Schwientek et al., 2008;Juncher Jørgensen et al., 2009;Zhang et al., 2009;Hayakawa et al., 2013). Pyrite is common in Suwannee Limestone of the UFA (Tihansky & Knochenmus, 2001), particularly in high-porosity zones (Price & Pichler, 2006) and oxidation of pyrite coupled to NO 3 -reduction is consistent with the inverse correlations between SO 4 2-and N 2 O and NO 3 -concentrations within the Ichetucknee system (Fig. 3B &E). However, dissolution of gypsum in the Avon Park Formation, located a few hundred meters below land surface, may provide additional SO 4 2-to these waters (Miller 1986) without a corresponding reduction of NO 3 -. This source may be relevant to the Sink-Rise system considering its deep groundwater source (Moore et al., 2009;Jin et al., 2015). The SO 4 2-concentrations are greater in Ichetucknee group 1b springs than group 1a springs because of their greater depth of flow (Martin and Gordon, 2000), indicating the possibility of a similar enrichment of SO 4 2-enrichment by gypsum dissolution in the Avon Park Formation. If the inverse correlation between N 2 O and SO 4 2-concentrations reflects gypsum dissolution, then the poor correlation between N 2 O and DOC may reflect DOCremineralization through fermentation or the use of electron acceptors other than NO 3 -during the long subsurface residence times.
The length of time that water and associated reactants involved in nitrogen cycling reactions reside in the subsurface may control N 2 O concentrations in spring waters. These potential drivers can be evaluated based on correlations between reactant and product concentrations. For example, such correlations have been used in a forested headwater catchment in Japan to show that shallow groundwater with high DO concentrations, presumably with short subsurface residence times, had elevated N 2 O and NO 3 -concentrations, whereas deeper groundwater flow paths extending to anoxic portions of the aquifer allowed complete reduction to N 2 (Osaka et al., 2006). A similar relationship was found in a sandstone catchment (McAleer et al., 2017) elevated. For the north-central Florida springs sampled here, the N 2 O concentrations vary inversely across the spectrum of residence times (Fig. 2), with the lowest N 2 O concentrations in the Ichetucknee springs with decades-long apparent ages for the discharging groundwater (Martin et al., 2016) and the highest, and also most variable, concentrations at the Santa Fe Sink-Rise system, where ground water has residence times of hours to days (Martin and Dean, 1999). At Ichetucknee springs, lower N 2 O concentrations in older spring waters of group 1b springs suggest more complete denitrification to N 2 (Table 2) than group 1a springs which have higher N 2 O concentrations and younger apparent ages. The apparent ages of all springs show inverse correlations with DO, NO 3 -, and N 2 O concentrations, which support increased reduction of DO, NO 3 -, and N 2 O with longer residence times (Fig. 6A-C). Although not statistically significant, the apparent ages increase with increasing N 2 concentrations reported in Heffernan et al. (2012) for the Ichetucknee springs. If denitrification is the active N 2 O producing mechanism, then these results suggest that longer subsurface residence times facilitate complete reduction of NO 3 -to N 2 (Fig. 6D).
Although the average age of Little River Spring (reversing springs group) water has been estimated to be ~ 21 years (Katz et al., 2001), at times following reversals, it should have shorter residences times on the order of weeks to months as river water intrudes and then discharges from the aquifer. Because the reversals also deliver elevated DO and DOC to the aquifer, these solutes would be expected to enhance N 2 O production within the aquifer. The only water sampled from Little River Springs for this study was collected immediately after a reversal (Fig. S3) and showed an N 2 O concentration of 2.76 µg N-N 2 O/L. This value is more than double the median value from the Ichetucknee springs group of 1.27 µg N-N 2 O/L, even though apparent ages of both springs are decades long. This greater concentration at Little River Spring suggests N 2 O may have been produced during the reversal that occurred immediately prior to sampling or that cumulative effects of multiple reversals create geochemical conditions that are favorable for N 2 O accumulation. None of the other springs within the reversing springs group was sampled immediately following a reversal, but nonetheless, their N 2 O concentrations are up to 7 times greater than the minimum N 2 O concentration observed within the Ichetucknee springs group. The elevated concentrations in the reversing springs group may thus reflect discharge of N 2 O produced during the weeks to months that reversal water remains in the aquifer.
The short residence times for water flowing from River Sink to River Rise is modulated by gain or loss of water to or from the conduits as recharge into River Sink varies with river discharge. River Rise discharge ranged from 4 to 73 m 3 /s during sampling times and is commonly greater than discharge into River Sink. At residence times of approximately 17 h, conduits switch from gaining to losing water (Fig. 7). Longer residence times (up to ~ 72 h) occur during periods of low flow when River Rise discharge exceeds River Sink discharge, reflecting a gain of groundwater that is enriched in SpC, NO 3 -, N 2 O, and SO 4 2-and low in DO and DOC concentrations. The estimated residence times show significant positive correlations with SpC, N 2 O, NO 3 -, and SO 4 2-concentrations and significant negative correlations with DO and DOC concentrations (Fig. 8). The correlation between SpC and residence time (Fig. 8A) suggests that the primary control of variations in N 2 O concentrations at River Rise stems from exchange of water between conduits and matrix porosity. At intermediate residence times of 22 to 27 h that correspond to periods soon after conduits switch between gaining and losing water (Fig. 7), N 2 O concentrations are anomalously high relative to the linear correlations with residence time at River Rise (Fig. 8D; circled data points). The correspondence of increased N 2 O concentrations at times when DOC-and DO-rich surface waters are delivered to the NO 3 -rich groundwater within the matrix porosity suggests that N 2 O production is enhanced by the exchange of water between the conduits and the surrounding matrix porosity, similar to production of N 2 O following spring reversals. Because of the limited amount of time (hours to days) that the recharged river water resides within the matrix porosity, complete denitrification to N 2 may not occur, as supported by the linear increase in N 2 O concentrations with increasing residence time at River Rise (Fig. 8D).
N 2 O concentrations in discharging spring waters depend on the extent of N r reduction (NO 3 -) and/or oxidation (NH 4 + ) to N 2 O within the aquifer and is commonly evaluated based on emission factor estimates (e.g., IPCC, Hergoualc'h et al., 2019). Emission factors for aquatic ecosystems are divided into three categories depending on the environment and include groundwater, streams and rivers, and estuaries, the most relevant value of which for comparison to the north-central Florida springs is the groundwater emission factor (EF 5g ). Estimates of global groundwater EF values have changed by nearly a factor of ten over the recent decades, reflecting the difficulty associated with estimating this value. The most recently reported mean EF 5g value is 0.0060 (Tian et al., 2019) based on 101 studies of agriculturally impacted groundwater and springs (avg. EF 5g = 0.0079), as well as upstream or surface water drainage (avg. EF 5g = 0.0040). Uncertainties in groundwater emission factors arise from the complex subsurface dynamics in aquifers and heterogeneous production, reduction, and transport mechanisms of N 2 O. These uncertainties contribute to poorly constrained values for the atmospheric evasion of N 2 O from aquatic systems (Yao et al., 2020). Atmospheric evasion from numerous locations around the world are influenced by spring discharge as reflected by decreasing concentrations of N 2 O downstream from springs sources (Ueda et al., 1993;Osaka et al., 2006). Furthermore, N 2 O evasion from springs are 60 to 100 times greater than N 2 O evasion from surrounding soils (Osaka et al., 2006;Hedlund et al., 2011). Evaluation of karst groundwater EF values may thus help constrain the potential for atmospheric evasion of N 2 O from spring waters.
The average groundwater EF NO3-(Eq. ( 1)) values for the sampled north-central Florida springs is 0.0045 mg N-N 2 O/mg N-NO 3 -but they show a large range from lows of 0.0007 in the Ichetucknee springs and reversing springs groups to as high as 0.0177 for the Santa Fe River Rise. nitrogen cycling. However, the highest EF values occur at River Sink with EF NO3-values of 0.0627 and EF TDN (Eq. ( 2)) values of 0.0131. These high EF values for the River Sink water suggest greater reduction and/or oxidation of N r species to N 2 O in surface waters draining from landscapes in the Santa Fe River headwaters. Here, the UFA is semi-confined versus the other spring groups where N 2 O reduction may be more prominent as shown by low EF factors at the reversing springs and Ichetucknee springs groups (Table 1). The semi-confined nature of the region upstream of River Sink suggests additional groundwater-surface water exchange resulting in N 2 O production during certain flow conditions, similar to the production between River Sink and River Rise.
Because most of the UFA springs have EF values much lower than the global average, the Floridan aquifer appears to have a lower potential to contribute N 2 O to surface waters. This low potential for N 2 O contribution may be common to other anthropogenically impacted eogenetic karst aquifers in addition to the UFA due to spring waters that typically have residence times on the order of years to decades (Florea and Vacher, 2006). While the presence of elevated NO 3 -concentrations is inherently required for N 2 O production, these results suggest that the turnover of NO 3 -to N 2 O in anthropogenically impacted eogenetic karst aquifers with high NO 3 -concentrations is ultimately controlled by the amount of DOC and DO delivered to the subsurface, facilitated by surface-groundwater interactions.
All of the UFA springs sampled in this study had N 2 O concentrations that are supersaturated compared to equilibrium with atmospheric concentrations, with saturations reaching up to ~1250%. N 2 O concentrations increase with increasing connectivity between surface water and groundwater that enhances the input of DOC into the subsurface, fueling denitrification. N 2 O concentrations are lower where the discharging water has residence times on the order of decades, as represented by the Ichetucknee springs and reversing springs groups, in comparison to spring water with residence times on the order of hours to days as represented by the Sink-Rise system. The low concentrations in older spring water likely reflect reduction of a larger portion of the N 2 O to N 2 , which is consistent with groundwater EF values for the Ichetucknee and reversing springs groups that are about an order of magnitude lower than the newly refined global average for groundwater of 0.0060 (Tian et al., 2019). In contrast, median groundwater EF values are around 5 times greater where residence times are short as exemplified by the Sink-Rise system. At that location, N 2 O concentrations increase as residence times increase from several hours to several days with decreasing discharge, and is a result of continuous input of DOC and DO-rich surface waters into the system. These results imply that the initial input of redox sensitive solutes and subsequent subsurface processing times affect N 2 O fluxes from karst aquifers, with longer residence times facilitating further reduction of N 2 O to N 2 prior to