Enhanced stream greenhouse gas emissions at night and during flood events

Headwater streams play a large role in aquatic greenhouse gas emissions. Carbon dioxide (CO2) and dissolved oxygen in streams often undergo changes through diel cycles. However, methane (CH4) and nitrous oxide (N2O) have unknown diel dynamics. Here, we reveal consistent patterns in CO2, CH4, and N2O over diel cycles and during flood events using high‐frequency continuous observations in a subtropical headwater stream. Diel cycles were most pronounced during baseflow. Increased nighttime discharge due to higher groundwater inputs enhanced gas transfer velocities and concentrations. Overall nocturnal emissions were 31%, 68%, and 32% greater than daytime for CO2, CH4, and N2O, respectively. Floods dampened diel signals. If both flood events and diel patterns are neglected, estimates of greenhouse gas emissions from headwaters may be greatly underestimated. Overall, CH4 and N2O emissions from headwater streams may be underestimated by ~ 20–40% due to a lack of observations during nighttime, floods, and in warmer climates.

Headwater streams are biogeochemical hotspots and play a major role in global riverine greenhouse gas budgets (Li et al. 2021;Marzadri et al. 2021).Estimates based on traditional discrete samples suggest that headwater streams (Strahler stream orders 1-3) contribute about 75% of global riverine CO 2 , CH 4 , and N 2 O CO 2 -equivalent (CO 2 -eq) emissions (Li et al. 2021).Because they are often difficult to access and their abundance increases with decreasing stream order, the global significance of headwater streams has been difficult to quantify (Noto et al. 2022).Accurate estimates of stream fluxes are needed to reduce uncertainty in regional and global extrapolations.Large uncertainties in global summaries and models (e.g., Rocher-Ros et al. 2023;Stanley et al. 2023) in streams remain due to scarcity of measurements, geographical bias against warm climates, and a lack of high temporal resolution observations covering extreme hydrological events and diel cycles.
Stream waters reflect hydrological inputs with different biogeochemical and spatiotemporal dynamics.Stream morphology, precipitation event intensity, and surrounding ecosystems all affect greenhouse gas dynamics in headwaters (Rocher-Ros et al. 2023).Anthropogenic nitrogen and carbon inputs often enhance instream greenhouse gases (Andrews et al. 2021;Ho et al. 2022).Storms mobilize and transfer significant loads of C and N from catchment soils to headwater streams (White et al. 2021a) and increase downstream aquatic N 2 O emissions (Woodrow et al. 2022).With climate change enhancing intense rainfall events in the subtropics (Clarke et al. 2022) and increasing agricultural and urban development, stream greenhouse gas emissions will remain difficult to predict (Yao et al. 2019;Battin et al. 2023;Rocher-Ros et al. 2023).
Stream greenhouse gas observations often rely on discrete sampling during daylight hours; however, nighttime gas dynamics can be fundamentally different than during the day.Metabolic processes such as gross primary production and ecosystem respiration are driven by day-night cycles and have a strong direct control on dissolved oxygen (DO) and CO 2 dynamics (Odum 1956;G omez-Gener et al. 2021).Evapotranspiration may also be lower at night, raising the water table and enhancing groundwater discharge (Hill 2019).Groundwater inputs are important greenhouse gas, carbon, and nutrient sources in headwater streams (Hotchkiss et al. 2015;Lupon et al. 2019).Thus, greater groundwater connectivity can enhance gas transfer to the atmosphere.Consistently high nighttime emissions have been shown to be important for global estimates of river CO 2 emissions (G omez-Gener et al. 2021), but little is known on diel patterns and magnitude of nighttime N 2 O and CH 4 emissions.
Here, we hypothesize that diel cycles of oxygen, groundwater discharge, and terrestrial connectivity will enhance nocturnal emissions of CH 4 and N 2 O.To overcome sampling bias against storm events and nighttime, we performed high temporal resolution observations of dissolved stream CO 2 , CH 4 , and N 2 O across diel cycles and during major rainfall events in a subtropical headwater stream.We build on the literature by (1) assessing drivers of greenhouse gases over diel and hydrological cycles, (2) focusing on less studied subtropical systems with high nitrate loads that may modify greenhouse gas production, and (3) contrasting the potential climate implications of stream CO 2 , CH 4 , and N 2 O emissions.

Methods
High-resolution temporal measurements (1 h) were performed to cover variability both on diel time scales and under contrasting hydrological conditions between 06 February and 08 April 2019 (62 d) in the subtropical catchment of Double Crossing Creek, Australia (Supporting Information Fig. S1).The average annual rainfall is 1685 mm, with $ 40% falling between February and April (Australian Government Bureau of Meteorology 2022) (Supporting Information Fig. S2).The catchment is comprised of intensive arable horticulture (60%) with large nutrient amendments.Stream flows in this subtropical region are dominated by episodic rain events (White et al. 2021b) rather than seasonal cycles as in most temperate regions.Our 62 d of observations capture regional hydrological extremes, and the multiple diel cycles provide replication to day-night comparisons.
DO, pH, temperature, salinity, electrical conductivity (EC), depth, current velocity, the natural groundwater tracer radon ( 222 Rn), and the greenhouse gases N 2 O, CH 4 , and CO 2 were measured with a wide array of sensors and data stored using automated dataloggers, along with discrete measurements of dissolved organic carbon (DOC), nitrate (NO 3 À ), and ammonium (NH 4 À ) reported in a companion manuscript (White et al. 2021a)  The k 600 values were then normalized to gas-specific k values and water temperature using equations for the kinematic viscosity of water (Siedler and Peters 1986)  diffusion of the gas in water (Jähne et al. 1987).Air-water greenhouse gas vertical fluxes (f a/w ) were estimated from: where k is the gas transfer velocity coefficient (m d À1 ), α is the solubility coefficient for the specific gas (Weiss and Price 1980), G w is the concentration of the gas in a water sample, and G atm is the equilibrium gas concentration for that temperature, conductivity, and pressure.CO 2 -eq emissions were computed over 20-and 100-yr sustained global warming potentials to determine the relative importance of each gas and compare it to IPCC estimates, following Neubauer and Megonigal (2019).

Results
Our observations spanned a dry summer-to-autumn transition with only 239 mm of rainfall, less than half the longterm average for this time of year in the region.The last 10 d of observations accumulated $ 50% of the total rainfall.Two significant rain events (60 mm on 30 March and 40 mm on 02 April) increased soil moisture from $ 30% to $ 60% and stream discharge (Fig. 1).The median stream discharge was 0.01 m 3 s À1 (0.05 quantile = 0.002 m 3 s À1 , 0.95 quantile = 0.05 m 3 s À1 ) with peaks in flow following rain events > 30 mm.Water temperature ranged from 19.2 C to 24.2 C (21.9 AE 0.9 C; mean AE standard deviation).DO was consistently undersaturated at 70.0% AE 5.9%.The pH (7.0 AE 0.08) and specific conductivity (487 AE 11.5 μS cm À1 ) values were within the range often found for these regional streams. 222Rn ranged from 199 to 587 dpm L À1 (392 AE 54 dpm L À1 ).DOC, NO 3 À , and The stream was always a greenhouse gas source to the atmosphere with CO 2 , CH 4 , and N 2 O saturations ranging between 614% and 1290% (984% AE 145%), 169% and 1176% (413% AE 147%), and 361% and 729% (523% AE 60%), respectively (Fig. 1).Using transfer velocities from Raymond et al.  2019) were, on average, 30% higher than when using Raymond et al. (2012) (Table 1).All transfer velocities and emissions values reported herein are derived from Raymond et al. (2012).Comparisons from Li et al. (2021) are included in tables and Supporting Information Material.
Observations were grouped into three hydrological periods (Fig. 1; Table 1).Dry conditions represented the first 24 d of sampling (February), accumulating 33 mm of rainfall and a mean runoff of 0.02 AE 0.02 mm d À1 .Relatively wet conditions occurred in the following 28 d in March, with 104 mm rainfall and multiple events < 25 mm (mean runoff 0.1 AE 0.08 mm d À1 ).Flood represented the last 10 d of sampling with 102 mm rainfall and rain events ≥ 40 mm (mean runoff 0.3 AE 0.2 mm d À1 ).During the flood, the stream's peak discharge increased by one order of magnitude, and depth increased from 4 to 31 cm, overtopping the bank by 1.4 m.Mean CO 2 and N 2 O saturations reduced during flood while mean CH 4 saturation increased (Fig. 1) and transfer velocities increased by $ 160% (Table 1).Runoff from these rainfall events described and captured herein Fig. 3. Scatterplots of drivers of greenhouse gas production over differing hydrological conditions in the day and the night in Double Crossing Creek, NSW, Australia.Circles represent samples collected at night between 00:00 h and 05:00 h, while triangles represent samples collected in the day from 12:00 h to 17:00 h.Orange, green, and blue shading indicate periods of dry, wet, and flood, respectively.also increased catchment nutrient inputs into the stream (White et al. 2021a).Failing to capture the flood would underestimate overall emissions by 6%, 28%, and 10% for CO 2 , CH 4 , and N 2 O.
Stream CO 2 , CH 4 , and N 2 O saturations and emissions varied over diel cycles, with a clear pattern of greater values during the night (Figs. 2, 3).Diel cycles were most pronounced during dry conditions, and dry period parameter changes are reported herein to highlight extremes (see Supporting Information Fig. S1; Supporting Information Table S1 for additional day-night comparisons).Lower temperature and DO occurred at night.Mean stream discharge increased from 0.003 AE 0.001 to 0.0045 AE 0.001 m 3 s À1 during the night, reflecting increases in stream current velocity and depths.Radon (a natural groundwater tracer) also experienced diel cycles with nocturnal values $ 10% greater than daytime values, implying enhanced nighttime groundwater inputs.Nocturnal mean CO 2 , CH 4 , and N 2 O emissions were 49%, 109%, and 50% higher than daytime emissions.Daytime runoff during the flood period reduced overall diel differences, with mean nocturnal emissions encompassing all hydrological periods 31%, 68%, and 32% greater than daytime for CO 2 , CH 4 , and N 2 O, respectively.Neglecting flood events and nocturnal emissions underestimates overall emissions by 19%, 41%, and 21% for CO 2 , CH 4 , and N 2 O, respectively.

Drivers of enhanced nocturnal greenhouse gas emissions
We show enhanced greenhouse gas nocturnal concentrations and emissions of under varying hydrological conditions in a headwater stream.G omez-Gener et al. ( 2021) reported a $ 30% increase in nighttime CO 2 from a global compilation of highfrequency observations.Here, we highlight potential drivers of CH 4 and N 2 O diel variation in a subtropical stream across contrasting hydrological conditions.Instream production, anthropogenic or natural inputs, groundwater inputs, discharge, stream depth, slope, and temperature can all influence greenhouse gas emissions (Raymond et al. 2012).Higher greenhouse gas concentrations and emissions are often found in lower order compared to higher Strahler order streams (Li et al. 2021), reflecting higher connectivity to terrestrial ecosystems.
The higher nocturnal greenhouse gas emissions in this stream observed during the dry and transition to relatively wet periods were explained by greater gas transfer velocities (14-32%) and greenhouse gas saturations (3-26%) at night (Supporting Information Table S1).The elevation in nocturnal gas transfer velocities reflected changes in depth (6% and 21% higher at night) and current velocity (13% and 24% higher at night) (Supporting Information Fig. S3).The diel variation was most pronounced during the dry period in February when the 10% increase in 222 Rn indicates greater nighttime groundwater inputs.The diel variation was intensified by the higher 222 Rn concentrations during baseflow (Fig. 3), indicating a higher ratio of groundwater input relative to overall flows (Kaule and Gilfedder 2021) that enhances CO 2 , CH 4 , and N 2 O saturations (Fig. 3).The groundwater contribution became less evident in terms of diel cycles of greenhouse gases and radon as flows through surficial soil horizons made a greater contribution in the flood period.
Enhanced nocturnal CO 2 production is typically explained by the cessation of photosynthesis at night (Attermeyer et al. 2021;G omez-Gener et al. 2021).In this stream, the short residence time and large groundwater inputs may minimize instream respiration as the key driver (Hotchkiss et al. 2015), even if the high C and N inputs (White et al. 2021a) can fuel metabolic processes.Rather, the sustained stream greenhouse gas supersaturation was likely due to the gas export from soil respiration through shallow groundwater pathways (Lupon et al. 2019).Groundwater inputs are also higher at night, as revealed by consistently higher 222 Rn concentrations and stream discharge.Taken together, higher nighttime groundwater contributions, higher water flows, and higher gas transfer velocities create a multiplicative effect, enhancing emissions to the atmosphere at night.
Oxygen availability is a major driver of instream greenhouse gas production (Wu et al. 2018).Lotic N 2 O concentrations are expected to be higher at night when DO is generally lower (Rosamond et al. 2012).In agriculturally impacted streams, denitrification and related N 2 O production are stimulated under low DO and high NO 3 À conditions (Rosamond et al. 2012;Liu et al. 2017).Concentrations of DOC and NO 3 À were overall high and increased with discharge due to high anthropogenic loading from surrounding horticultural activities (Wadnerkar et al. 2021;White et al. 2021a).Both DOC and NO 3 À exports may contribute to increased emissions, reflecting a correlation between both the export of DOC and NO 3 À and the export of CO 2 and N 2 O from soil respiration, which accumulates in the biochemically reactive surface area of pore waters and is exported when discharge increases (Marzadri et al. 2017).Here, DO saturations were 10% lower during the night during dry and negatively correlated with N 2 O saturations (Fig. 3; Supporting Information Figs.S3, S4).Our observations of increased nocturnal N 2 O coupled with low DO and high NO 3 À are consistent with observations of diel cycling in agricultural catchments in China (Wu et al. 2018), the United States (Laursen and Seitzinger 2004), Canada (Rosamond et al. 2011), and in larger rivers (Huang et al. 2013).In this stream, N 2 O emissions increased with discharge and directly after the second major rainfall event, consistent with the highest NO 3 À and reducing N 2 O concentrations as expected for regional agricultural streams (Andrews et al. 2021).
In contrast to N 2 O, CH 4 emissions found here were at the lower end of global ranges (Rosentreter et al. 2021;Stanley et al. 2023).As a comparison, the highest CH 4 emissions in our study (0.7 mmol m À2 d À1 ) were one order of magnitude lower than temperate agricultural headwaters (Schade et al. 2016) and negligible in comparison to temperate peatland headwaters (Taillardat et al. 2022) ).The higher concentration of N 2 O, coupled with lower CH 4 observed over the dry period (Fig. 3), also suggests that denitrifiers outcompete methanogens.Other potential loss pathways may be some instream oxidation of CH 4 , with estimates suggesting that half of the dissolved CH 4 in supersaturated lowland headwater streams in the United States is oxidized to CO 2 before evasion (Robison et al. 2022).
Groundwater is an important driver of river CH 4 emissions (Lupon et al. 2019;Rocher-Ros et al. 2023).Overall saturations of CO 2 and N 2 O were consistent with 222 Rn and diluted as flow increased.The inverse relationship with discharge suggests that lateral groundwater was the dominant source of CO 2 and N 2 O.However, CH 4 , while saturated in the stream, had no relationship with 222 Rn and an overall positive correlation with discharge (Fig. 3; Supporting Information Fig. S4).CH 4 in regional streams increases with catchment forest cover during intense runoff periods (Andrews et al. 2021).Hence, we suspect CH 4 sources from upstream forests under high runoff following rainfall, widening the catchment area where methanogenesis may occur.

Implications
Anthropogenic emissions of CO 2 , CH 4 , and N 2 O contribute to 76%, 16%, and 6% of global warming potentials, respectively (IPCC 2022).Translating stream emissions to warming potential, we found that CO 2 , CH 4 , and N 2 O contributed 92.7%, 0.2%, and 7.1% of total aquatic CO 2 -eq emissions (Fig. 4).Overall, average stream N 2 O and CO 2 emissions were $ 22% and $ 19% higher than their anthropogenic global mean contributions, and CH 4 was negligible.Our findings on low CH 4 emissions contribute to the emerging literature on stream CH 4 dynamics (Stanley et al. 2016;Li et al. 2021;Rosentreter et al. 2021).Our CO 2 -eq 20-and 100-yr potential emissions show important variations in N 2 O between dry and flood periods.
Default emissions factors (EFs) have been established by the IPCC for waterways that contribute to indirect N 2 O emissions via leaching and runoff (EF 5r ).The ratio of dissolved N 2 O concentration to dissolved inorganic nitrogen concentration (EF) was used to calculate an in situ EF (Hu et al. 2016).We employ the modified default EF 5r emission factor (0.0075 AE 0.025-0.0005kg N 2 O kg À1 ) to compare our observations to the IPCC indirect emissions model (Hergoualc'h et al. 2019).We found an in situ EF of 0.0003 over the observation period, well below the IPCC default EF 5r emission factor.Our estimate capturing day-night cycles and hydrological extremes is at the lower end of a compilation of 52 studies in modified waterways, revealing EFs ranging from 0.00005 to 0.1133 (Webb et al. 2021) and consistent with summer values found in agricultural headwaters in the United Kingdom (Hama-Aziz et al. 2017).Agriculturally impacted streams in Canada had EFs of 0.0044 at night and 0.0034 in the day (Baulch et al. 2012), consistent with our observations.Despite our inclusion of higher nighttime and flood-time measurements, the emissions observed in this study are still

Conclusions
Our analysis has implications for global estimates of greenhouse gas emissions from headwater streams.Global estimates of riverine CO 2 , CH 4 , and N 2 O emissions exhibit high spatial and temporal variability and are largely based on discrete sampling during the day (Marzadri et al. 2021;Liu et al. 2022;Rocher-Ros et al. 2023), potentially missing higher emissions that occur during the night and floods.In our study, not only were CO 2 emissions 31% higher at night in line with global averages (G omez-Gener et al. 2021), but CH 4 (68%) and N 2 O (32%) emissions were also higher at night.Even though our results are from a single stream, the consistent patterns and the drivers suggest that those mechanisms may be prevalent across other headwater streams around the globe.Therefore, global estimates of riverine CH 4 and N 2 O emissions are likely underestimated for overlooking enhanced emissions during the night and during floods.

Fig. 1 .
Fig. 1.Time series of hydrological and chemical parameters over 62 d of continuous observations in Double Crossing Creek capturing contrasting hydrological conditions and day-night cycles.Orange, green, and blue shading indicate periods of dry, wet, and flood, respectively.

Fig. 2 .
Fig. 2. Day/night comparisons and differences between CO 2 , CH 4 , and N 2 O in Double Crossing Creek, Australia.The first column (plots a, d, and g) is high-resolution 24-h cycles highlighting the within-day variability, using the current velocity of discharge to show that floods can disturb that diel signal.The red lines are averages.CV represents the coefficient of variation of stream discharge (SD/mean) of each day to highlight days when the diel pattern may be altered due to flood events.The second column (plots b, e, and h) is an aggregation of Column 1 data, with the average night and day values for each complete 24 h cycle.Most of the concentrations/fluxes are higher at night.The solid black line is a 1 : 1 ratio.The third column (plots c, f, and i) is an extra step of aggregation, with density plots showing the magnitude of difference from day-night, and across hydrological regimes.The vertical, colored lines show the mean values.Orange, green, and blue shading indicate periods of dry, wet, and flood, respectively.

Fig. 4 .
Fig. 4. CO 2 -equivalent (CO 2 -eq) emissions on a 20-yr SGWP timescale over the 62-day time series observation period in Double Crossing Creek, NSW, Australia.White, light gray, and dark gray shading indicate periods of dry, wet, and flood, respectively.Percentage contributions to total emissions over 20-and 100-yr timescales for CO 2 , CH 4 , and N 2 O (Neubauer and Megonigal 2019).

Table 1 .
and the Mean values with standard deviation and value range for water parameters, transfer velocities, and vertical fluxes over dry (n = 24 d, accumulated rainfall 33 mm), wet (n = 28 d, accumulated rainfall 104 mm), and flood (n = 10 d, accumulated rainfall 102 mm) conditions in Double Crossing Creek, Australia.
. These low CH 4 emissions are expected with oxygen and NO 3 À acting as preferential terminal electron acceptors, inhibiting methanogenesis.Low CH 4 emissions have been observed in agriculturally impacted headwater streams with high DOC and high NO 3 À (Schade et al. 2016) due to