Introduction

Inland waters are important sites of biogeochemical activity and are often sources of the major greenhouse gases carbon dioxide (CO2) and methane (CH4) to the atmosphere (Cole et al. 2007; Tranvik et al. 2009; DelSontro et al. 2018). While inland waters make up just 4% of the Earth’s non-glaciated land surface area (Verpoorter et al. 2014), they disproportionately contribute to the global carbon cycle where they process, store, transport, and emit large amounts of carbon as CO2 and CH4 (Mendonça et al. 2017; Pilla et al. 2022). Freshwater lakes and ponds are estimated to emit over 0.5 Pg C of CO2 per year (DelSontro et al. 2018) and 0.3 Pg C of CH4 per year (Rosentreter et al. 2021), yet large uncertainties remain in global budgets. Small waterbodies such as ponds remain a major source of this uncertainty, and ponds are increasingly recognized as disproportionately large and variable emitters of CO2 and especially CH4 (Holgerson & Raymond 2016; Rosentreter et al. 2021).

Ponds, defined here as waterbodies < 5 ha in size, < 5 m in depth, < 30% covered in emergent vegetation (Richardson et al. 2022), and including both natural and artificial waterbodies, comprise over 90% of freshwater lakes by number (Downing et al. 2006) and 20% by surface area (Verpoorter et al. 2014). Because ponds often have small volumes, large nutrient loads, and high rates of production and decomposition (Andersen et al. 2017), they tend to have large emissions of CO2 and CH4 (Grinham et al. 2018; Gorsky et al. 2019; Rabaey & Cotner 2022). However, measured emissions among ponds can vary by several orders of magnitude, even amongst ponds within the same region (Peacock et al. 2019; Audet et al. 2020; Rabaey & Cotner 2022). This variation can compound when evaluating pond fluxes at larger scales and leaves whole estimates of freshwater emissions poorly constrained. Many synthesis efforts rely on single measurements of CO2 or CH4 fluxes in ponds, often during the summer (e.g. Holgerson & Raymond 2016) and temporally representative pond greenhouse gas measurements are lacking (Koschorreck et al. 2020).

Seasonality has been shown to be an important factor in determining annual greenhouse gas fluxes from lakes, as summer fluxes of CO2 and CH4 may not be representative of the entire year (Encinas Fernández et al. 2014; Gorsky et al. 2021). In temperate and boreal lakes, mixing periods during the spring and fall can contribute the majority of yearly CO2 and CH4 flux (Riera et al. 1999; Jansen et al. 2019), and these turnover events can contribute significantly to global lake emissions (Johnson et al. 2022). Ponds have often been assumed to be continuously mixed (Scheffer 2004), though recent research has shown this is frequently not the case. Despite small surface areas and shallow depths, ponds can exhibit a variety of mixing regimes from continuously mixing throughout the open-water period to stratifying for weeks and months at a time (Holgerson et al. 2022). Carbon dioxide and methane dynamics vary under mixed and stratified conditions, as stratification allows for the accumulation of gases and is often accompanied by increased anoxia (Søndergaard et al. 2023), both of which can lead to higher diffuse emissions of CH4 (Ray & Holgerson 2023). Stratified conditions in ponds also can lead to reduced CO2 fluxes or CO2 uptake, likely due to separating sediment respiration from the atmosphere and/or increased primary production (Ray & Holgerson 2023). Evaluating the impact of mixing and stratification on pond CO2 and CH4 dynamics over longer time periods is necessary to understand variability in annual greenhouse gas emissions and improve aquatic greenhouse gas budgets.

In this study, we examined annual CO2 and CH4 dynamics in four temperate ponds with varying mixing regimes, including three constructed stormwater ponds and one natural pond. Our objectives were to (a) assess seasonal patterns of both CO2 and CH4 fluxes in ponds, (b) determine how varying mixing regimes influence seasonal patterns, and (c) determine seasonal contributions to annual greenhouse gas fluxes. We predicted stratification would lead to higher emissions overall and that ponds with longer periods of stratification would behave more like lakes, with accumulation of gases during summer and larger emissions during fall mixing. We predicted mixed ponds would have lower emissions during the fall and would have lower total CH4 emissions due to an oxygenated water column.

Methods

Study sites

This study was performed in four ponds located within or near the metropolitan area of St. Paul and Minneapolis, Minnesota, United States, and all fit the definition of ponds (< 5 m maximum depth, < 5 ha surface area) as described in Richardson et al. (2022). The four ponds were chosen based on prior monitoring data to represent varying mixing regimes, and we expected that Alameda would mix rarely during the summer period, Cedar Bog and Materion would mix intermittently, and Cleveland-Roselawn would mix frequently. Three of the ponds (Alameda, Materion, and Cleveland-Roselawn) are functioning urban stormwater ponds all located in residential suburban watersheds. All three ponds represent typical stormwater ponds for the area and were created for water regulation and the reduction of nutrient exports. The other pond, Cedar Bog, is a natural dystrophic pond within an exurban nature preserve (Lindeman 1942). Cedar Bog is directly surrounded by a matrix of forest and swamp and is representative of many natural forested ponds throughout the ecoregion.

All four ponds had similar surface areas and maximum depths (all < 1.5 ha and < 1.8 m), as well as similar annual mean total phosphorus (TP) and total dissolved nitrogen (TDN) concentrations (Table 1). One of the largest differences among ponds was the dominant primary producer type, as Alameda was covered in duckweed (Lemna sp. and Wolffia sp.) for most of the open water period and had little to no macrophyte growth; Cedar Bog had a high density of submerged macrophytes (Ceratophyllum demersum) throughout the year; Materion had high a density of phytoplankton with little to no macrophytes; and Cleveland-Roselawn had high densities of submerged macrophytes (Potamogeton sp.) as well as duckweed and phytoplankton growth.

Table 1 Mean (sd) water chemistry and morphology for each pond
Full size table

To assess mixing regimes, thermistor strings of at least four temperature loggers (HOBO pendant temperature loggers, Onset, Bourne Massachusetts, United States) were placed in each pond in April and May 2021, and measured water temperature every 5 min (30 min in winter). Depths included 0.02, 0.6, 1.2, and 1.75 m for Alameda, 0.1, 0.3, 0.5, 0.7, 1, and 1.5 m for Cedar Bog, 0.02, 0.6, 0.8, and 1.35 m for Cleveland-Roselawn, and 0.02, 0.5, 1, and 1.7 m for Materion. The surface logger was attached underneath a foam float to avoid direct sunlight. Manufacturer reported temperature accuracy is 0.5 °C, and resolution is 0.04 °C. All loggers were intercalibrated in a 20 °C water bath before and after deployment. Both before and after deployment, all loggers read within 0.1 °C and no corrections were made. To assess greenhouse gas dynamics, ponds were sampled bi-weekly from mid-April 2021 through ice-on in late November 2021, as well as twice during the winter ice-cover period and twice after ice-off in April and May 2022. During each sampling event concentrations and fluxes of CO2 and CH4 were measured, as well as profiles of water chemistry variables.

Water chemistry

During each sampling event, profiles were taken at the deepest point of each pond to measure water temperature, dissolved oxygen (DO), pH, chlorophyll a (chl a), and conductivity (Manta probe, Eureka Water Probes, Austin Texas, United States). Measurements were taken every 5 s, and the probe was lowered slowly enough to allow for measurements every 0.05–0.1 m. Dissolved oxygen profiles were used to calculate the anoxic fraction (AF) of the pond, a discrete measurement of the extent of anoxia at the time of sampling. The AF was defined as the fraction of sediment exposed to anoxic conditions (anoxic sediment area/pond surface area; Nürnberg 1995). Dissolved oxygen values under 2 mg L−1 were considered anoxic (Nürnberg 1995). The anoxic sediment area was determined using DO profiles and pond hypsographic curves determined from bathymetry. As an example, in a 1.5 m-deep pond in which the DO profile was less than 2 mg L−1 within 0.5 m of the bottom of the pond, all sediment at a depth of 1 m or greater was considered the anoxic sediment area.

In addition to profiles, surface water samples were taken approximately monthly. One liter surface water samples were collected in the middle of each pond and used to measure total phosphorus (TP), dissolved organic carbon (DOC), and total dissolved nitrogen (TDN). Water samples were immediately placed in a cooler for transport and filtered or frozen the same day. For DOC and TDN analyses, half of the collected water samples were passed through a muffled 0.7 μm pore-size filter and placed in the freezer. The second half of the collected water samples were directly placed in the freezer for later TP analysis. TP was measured using the molybdenum blue reaction with acid persulfate digestion (Murphy & Riley 1962). DOC and TDN were measured using a Shimadzu TOC-L model high temperature carbon-analyzer with a TNM-L module (Shimadzu Corp., Kyoto, Japan).

Water temperature and mixing

To determine mixing and stratification dynamics, water density gradients were calculated from the thermistor measurements and conductivity measurements. Since conductivity was only measured biweekly, we assumed conductivity changed linearly between measurements to estimate conductivity at the same timescale of temperature measurements. Salinity was calculated from conductivity using equations from Fofonoff and Millard Jr (1983), and Hill et al. (1986). Water densities at the top of the water column (5 cm from the surface) and bottom of the water column (at the surface of the sediment) were determined using temperature and salinity following Millero and Poisson (1981) using the rLakeAnalyzer package (Winslow et al. 2016) and then divided by the distance between them to give a density gradient (kg m−3 m−1, Fig. 1). Density gradients are commonly used to assess mixed depth and stratification, however different density gradient thresholds can be used to define whether a water column is stratified vs. mixed (Gray et al. 2020). In our analyses, we considered a density gradient threshold of 0.287 kg m−3 m−1, as this was shown to successfully distinguish mixing regimes in ponds and shallow lakes (Holgerson et al. 2022).

Fig. 1
figure 1

Density gradients (a) and temperature heatmaps (b) of the four study ponds, Alameda (rarely mixed), Cedar Bog (intermittently mixed), Materion (intermittently mixed), and Cleveland-Roselawn (frequently mixed). In (a), gray shading represents times when the pond was stratified (density gradient > 0.287 kg m−3 m−1) while the dashed line represents the density gradient threshold of 0.287 kg m−3 m−1, and blue shading represents the period under ice cover

Full size image

CO2 and CH4 flux

Fluxes of CO2 and CH4 were measured at every open water sampling period using an opaque floating chamber (headspace of 10.02 L) connected directly to a portable greenhouse gas analyzer (DX4040 FTIR Gas Analyzer, Gasmet Technologies Oy, Vantaa, Finland) via inlet and outlet tubing to create a closed loop system (Rabaey & Cotner 2022). To measure CH4 and CO2 fluxes, the float was placed on the water surface and the pressure was allowed to equilibrate through the outlet port, which was then connected to the analyzer (pumping rate of 1.5 L min−1) to begin the incubation. No alterations were made to the water surface before placing the chamber, and the chamber was placed on top of any floating macrophytes (such as duckweed) or algal mats on the surface of the pond as well as open water. Gas measurements were taken every 5 s, and incubations lasted at least 5 min, and up to 10 min if rates were low. Chamber incubations were taken at three locations on each pond, one near the shoreline, one in the center of the pond over the deepest water, and one in between those two points.

Fluxes of CO2 were calculated based on the slope of linear increase or decrease in chamber concentration using a simple linear regression, as the concentrations in the chamber always increased or decreased linearly over time. Linear models were evaluated using R2, and all incubations had R2 values above 0.72. For CH4, ebullition led to a non-linear increase in CH4 concentration when bubbling occurred (example shown in Supplementary Figure S1). With the 5 s measurement resolution, emission from ebullition events could be easily separated from the diffusive flux (Xiao et al. 2014; Rabaey & Cotner 2022). First, the CH4 diffusion rate was calculated by fitting a linear regression to a straight segment of the sample curve (Supplementary Figure S1). Multiplying this diffusion rate by the total sample time gave the CH4 concentration in the chamber due to diffusion. Methane concentration due to ebullition was the surplus CH4 concentration, calculated by subtracting the final diffusion concentration and original background concentration from the total concentration at the sampling endpoint (Supplementary Fig. S1).

Dissolved CO2 and CH4 concentrations

Concentrations of CO2 and CH4 in both the surface water and bottom water were measured using the headspace technique (McAuliffe 1971) at only the center location of each pond, starting in June 2021. Surface samples were collected 5–10 cm below the surface, and bottom water samples were collected approximately 10 cm above the sediment using a Van Dorn water sampler while not producing bubbles or turbulence. One side of the Van Dorn was opened and immediately 125 mL of water was collected in a 140 mL plastic syringe. Any bubbles formed when drawing water were removed by reducing the syringe volume to 105 mL. Atmospheric air (32.5 mL) was introduced into the syringe and then the syringe was vigorously shaken for 2 min. Thereafter, 30 mL of the headspace was transferred into a separate syringe, and immediately injected onsite into the Gasmet portable gas analyzer, equipped with a closed loop injection system (Wilkinson et al. 2019).

To estimate the concentration of CO2 and CH4 in the surface and bottom water, gas concentrations in the sample headspace were first calculated using the following equation (Wilkinson et al. 2019):

$$ X_{{headspace}} = \left( {\frac{{V_{l} - V_{s} }}{{V_{s} }}} \right)\Delta X + X_{0} $$
(1)

where Xheadspace is the concentration of the sample headspace, Vl is the volume of the closed loop system, Vs is the volume of the injected sample, ΔX is the change in gas concertation after sample injection, and X0 is the initial gas concentration. Inputs from the added atmospheric air were subtracted out from the headspace concentration using onsite measurements of atmospheric gas concentrations (usually around 400 ppm CO2 and 2 ppm CH4). The partial pressure of gas in the headspace was used to calculate the moles of dissolved gas in the water (molaq) according to Henry’s law, and this was added to the moles of gas in the headspace (molheadspace) and divided by the original sample volume to find the concentration of gas in the original sample (Cw) (Johnson et al., 1990):

$${mol}_{aq}={P}_{headspace}\times {K}_{H}$$
(2)
$${mol}_{headspace}=\frac{{ P}_{headspace}}{RT}\times {V}_{headspace}$$
(3)
$${C}_{w}=\frac{{mol}_{aq}+{mol}_{headspace}}{{V}_{w}}$$
(4)

where Pheadspace is the partial pressure of gas in the headspace (atm), KH is Henry’s constant (mol L−1 atm−1), R is the universal gas constant (0.082, L·atm mol−1 K−1), T is the temperature (K), Vheadspace is the volume of the headspace (L), and Vw is the volume of the water sample (L). Henry’s constants were calculated for CO2 (Weiss 1974) and CH4 (Wiesenburg & Guinasso 1979) and corrected for water temperature and pressure. Constants were corrected for salinity, as particularly the bottom waters of the urban ponds had high conductivity in the spring due to road salt inputs (up to 6700 µS cm−1, 6.1 psu). Carbon dioxide concentrations were also corrected for carbonate equilibrium following Koschorreck et al (2021).

Seasonal flux calculations and statistics

To assess the effect of different mixing regimes on seasonal emission pathways, fluxes of CO2 and CH4 were estimated for the open water period of spring and summer, the fall turnover period, and release of gases during ice out. Ice-on and ice-off dates were determined by visually checking each pond daily when ice started to form/melt. Both ice-on and ice-off dates were similar for all four ponds. To calculate fluxes from the open water period of spring and summer, we used the measured fluxes from April 2021 through 22 September 2021 (diffusive for CO2, diffusive and ebullitive for CH4). The total flux from this period was calculated using area under the curve functions (using the trapezoidal rule) from the flux R package (Jurasinski et al. 2022). The same procedure was done for the fall turnover period, which we defined as 23 September 2021, through ice-on (25 November).

To estimate the release of stored CO2 and CH4 after ice out, under-ice measurements of CO2 and CH4 were used to estimate the total storage of both gases at ice out (4 April). Surface and bottom concentrations of CO2 and CH4 were similar under the ice, and concentrations at the two depths were averaged for each gas to give a depth-integrated mean. Gases were assumed to accumulate linearly under the ice (e.g. Jansen et al. 2019), and concentrations were extrapolated from the last sampling date (early March) to the ice out date (4 April). While previous studies have estimated that only 25–46% of stored CH4 may get emitted during fall turnover events due to oxidation (Schubert et al. 2012; Encinas Fernández et al. 2014), a higher proportion of gases may escape to the atmosphere during ice out turnover events (64–96%, Jansen et al. 2019). Since the water column under the ice was completely anoxic in each pond, we assumed all the accumulated CO2 and CH4 was released to the atmosphere, and we used the estimated concentration at ice out to calculate a total flux for the ice out period.

To evaluate the total global warming potential of emitted gases, gas fluxes for each period were converted to CO2 equivalents (mg CO2-eq. m−2 d−1) on a mass basis by multiplying the CH4 emissions by 27.2 (100-year warming potential; Calvin et al. 2023). To test the effect of mixing on CO2 and CH4 dynamics, linear mixed effect models were used with “pond” as a random effect using the lme4 R package (Bates et al. 2015). Model significance was evaluated using the lmertest package (p-values; Kuznetsova et al. 2017) and the MuMIn package (R2; Barton 2020). All statistical tests used a significance level of 0.05, and all statistical analyses were performed using R (R 4.4.2, R Core Team 2022).

Results

Water chemistry

The four ponds all had similar annual mean TP (0.12–0.19 mg L−1) and TDN (0.51–0.85 mg L−1) concentrations, while Cedar Bog and Cleveland-Roselawn had higher annual mean DOC concentrations (12–14 mg L−1) compared to Alameda and Materion (8–9.3 mg L−1, Table 1). All ponds had the highest nutrient concentrations (TP, TDN, DOC) in winter under the ice (Table S1).

Differences in primary producers among the ponds also corresponded with differences in water chemistry. Alameda was covered in duckweed for most of the open water period and had little to no macrophyte growth and low Chl-a concentrations. Cedar Bog had a high density of submerged macrophytes and low Chl-a concentrations. Materion had the highest mean Chl-a concentrations (14 µg L−1), and highest surface pH (8.4), while Cleveland-Roselawn had high densities of submerged macrophytes, duckweed and high Chl-a concentrations (Table 1).

Mixing and oxygen regimes

The four ponds exhibited distinct mixing regimes (Fig. 1). Alameda rarely mixed during the summer period and from 1 June through 31 August was mixed only 2.9% of the time. For this same period, Cedar Bog and Materion mixed intermittently, mixing 10% and 13% of the time, respectively and Cleveland-Roselawn mixed the most often at 28%. Starting in September, all four ponds mixed daily, usually at night. Cleveland-Roselawn mixed most nights from July onward (Fig. 1).

Conductivity had the largest impact on the density gradient in Alameda, where bottom water conductivity reached 4000 µS cm−1 in the springtime and slowly decreased throughout the summer (Fig. S2). This elevated conductivity in the springtime kept Alameda from mixing, even though the temperature difference between the top of and bottom of the water column was small. Cedar Bog had slightly elevated conductivity in the bottom waters during the fall and winter, but conductivity never got above 700 µS cm−1 and had little effect on density gradients (Fig. S3). Both Cleveland-Roselawn and Materion only had elevated conductivity under the ice in the winter (Fig. S4, Fig. S5) and the bottom waters of Materion in March had the highest conductivity of over 6000 µS cm−1.

Mixing also impacted the oxygen regimes of the ponds, and across all ponds the anoxic fraction significantly increased with longer stratified periods (p = 0.041, R2 = 0.800; Fig. 2c). Alameda had the lowest mean surface DO and highest mean anoxic fraction throughout the year (Table 1), and only had an anoxic fraction of less than 0.89 in the spring and fall (Table S1). Cedar Bog also had low surface DO and a high mean anoxic fraction, while Materion and Cleveland-Roselawn had high mean surface DO and lower mean anoxic fractions, with Cleveland-Roselawn having a low anoxic fraction throughout the open-water period. All ponds went anoxic under the ice in winter (Table S1).

Fig. 2
figure 2

Relationship CO2 bottom water concentrations (a), CH4 bottom water concentrations (b), and the anoxic fraction (c) with the length of stratification (days since last mixed) during the spring and summer period (15 April–22 September). CO2 concentrations were not significantly related to stratification length (p = 0.159, R2 = 0.62) although there was a positive relationship. Both CH4 concentrations and the anoxic fraction increased significantly with longer stratification length (p < 0.001, R2 = 0.749; p = 0.041, R2 = 0.800). Relationships between variables and length of stratification were assessed using mixed effect models with pond as a random effect and stratification length as a fixed effect

Full size image

Open-water CO2 and CH4 concentrations and fluxes

Bottom water concentrations of CO2 and CH4 had similar seasonal patterns, and CO2 concentrations were generally ~ 10 times greater than CH4 concentrations (Fig. 3). Bottom water concentrations of CO2 and CH4 generally accumulated during stratified periods during the spring and summer (15 April–22 September), though the sampling frequency was not sufficient to capture all patterns during short-term stratified periods. Methane concentrations in the bottom waters increased significantly as stratified periods increased (p < 0.001, R2 = 0.749; Fig. 2b), while CO2 concentrations were not significantly related to stratification length (p = 0.159, R2 = 0.62; Fig. 2a). Across all ponds, CH4 accumulated at a rate of 3.5 µM per day under stratified conditions during the spring and summer.

Fig. 3
figure 3

Open water dynamics of CO2 and CH4 in each of the four study ponds, including fluxes of CO2 and CH4 (a–d), surface concentrations (e–h), and bottom water concentrations (i–l). Surface concentrations were measured 5cm below the surface and bottom water concentrations were taken 10cm above the sediment. Note the different scales for CO2 and CH4. Gray shading represents times when the pond was stratified (density gradient > 0.287 kg m−3 m−1)

Full size image

Alameda had the highest concentrations of CO2 and CH4 in the bottom waters, reaching maximum values of 4.1 mM CO2 and 0.5 mM CH4 in early August. Cedar Bog reached maximum bottom water concentrations of 0.6 mM for CO2 and 0.3 mM for CH4, while Materion reached maximum concentrations of 1.4 mM for CO2 and 0.29 mM for CH4. Cleveland-Roselawn had the lowest build-up of gases, with maximum concentrations of 0.55 and 0.05 mM for CO2 and CH4, respectively.

Fluxes for CO2 and CH4 had diverging seasonal trends (Fig. 3). Across all ponds, CO2 fluxes were lowest during the summer period (mean 3.3, sd 41.9 mmol m−2 d−1) and highest during the fall (mean 60.4, sd 93.5 mmol m−2 d−1). Methane fluxes were highest on average during the summer (mean 19.1, sd 13.1 mmol m−2 d−1). Alameda and Cedar Bog experienced the highest CO2 and CH4 emission peak during the fall while the water column was mixed. Materion and Cleveland-Roselawn experienced no fall peak in fluxes, and CH4 emissions trended towards zero in both ponds during the fall (Fig. 3).

During the spring and summer period (15 April–22 September), CO2 fluxes in each pond were significantly lower under stratified conditions (p = 0.007; Fig. 4a). Methane emissions in each pond were significantly higher under stratified conditions (p = 0.003) and mean CH4 emissions were three times higher during stratification (6.2 vs. 18.3 mmol m−2 d−1; Fig. 4b).

Fig. 4
figure 4

Fluxes of CO2 (a) and diffusive emissions of CH4 (b) under stratified and mixed conditions in the four study ponds during the spring and summer period (15 April–22 September). CO2 fluxes were significantly lower under stratified conditions (p = 0.007), and CH4 emissions were significantly higher under stratified conditions (p = 0.003). Differences between mixed and stratified conditions were assessed using a mixed effect model with pond as a random effect and mixing status as a fixed effect. Box plots depict the minimum, first quartile, median, third quartile, and maximum, overlain with each individual point

Full size image

CH4 ebullition

All four ponds had frequent CH4 ebullition during the open-water period, and 53% of all chamber measurements demonstrated ebullition events. Ebullition increased with increasing water temperature above the sediment and was highest during the summer months (Fig. 5). Ebullition patterns in all ponds had two peaks, a similarly timed peak at the beginning of June, followed by a decrease in ebullition rates and a second peak in late summer, with the timing varying in each pond (Fig. 5b).

Fig. 5
figure 5

Ebullition rates of CH4 in the four study ponds. a Ebullition exponentially increased with water temperature above the sediment. Black line represents exponential equation, R2 represents Efron’s psuedo R2 for non-linear models (formula shown in figure, Tsediment = temperature just above the sediment). b Across the open water period ebullition increased in the warmer months but showed a decrease across each pond in mid-summer. Black line represents a LOESS regression of all points. In both plots grey error bars represent ± SE of the three chamber measurements

Full size image

Under ice CO2 and CH4 concentrations

Ponds had ice cover from 25 November 2021 through 4 April 2022, with several days of partial ice cover around those dates that varied by pond. All ponds were completely anoxic under the ice (Table S1) and had under-ice accumulation of CO2 and CH4 during the winter (Fig. 6). Cedar Bog had the highest accumulation of both CO2 and CH4, followed by Cleveland-Roselawn, Alameda, and Materion. Using a linear model, estimated gas concentrations at the time of ice-out were 1.3 mM CO2 and 0.32 mM CH4 for Cedar Bog, 1.1 mM CO2 and 0.19 mM CH4 for Cleveland-Roselawn, 0.78 mM CO2 and 0.16 mM CH4 for Alameda, and 0.73 mM CO2 and 0.12 mM CH4 for Materion (Fig. 6).

Fig. 6
figure 6

Under ice accumulation of CO2 (triangles) and CH4 (circles) in each pond, measured at three different time points during the winter. Lines represent multiple regression of CO2 (dotted) and CH4 (solid) concentrations for each pond (p < 0.001, R.2adj = 0.98), and red dotted lines represent estimated gas concentrations at the time of ice-out (4 April 2022, depicted by red vertical line)

Full size image

Total CO2 and CH4 fluxes

Estimating total fluxes for each period (ice-out, spring–summer, and fall turnover) showed substantial differences in the contribution of each gas/period to total emissions for each pond (Table 2). Alameda and Cedar Bog had large positive fluxes during fall turnover that contributed significantly to annual totals (33% of annual CO2 and 23% of diffusive CH4 for Alameda, 66% for CO2 and 21% of diffusive CH4 for Cedar Bog), while both Materion and Cleveland-Roselawn had negligible contributions from fall turnover (< 3% of diffusive CH4, Table 2). All ponds had similar positive fluxes during the ice-out period, which ranged from 6.4 to 10% of CO2 fluxes and from 6.4 to 11% of diffusive CH4 fluxes (Materion and Cleveland-Roselawn had negative yearly CO2 fluxes, making percent comparisons impossible for those two ponds). Materion and Cleveland-Roselawn were sinks for CO2 during the spring–summer period, while all ponds were sources of CH4 throughout the year. In terms of warming potential, CH4 fluxes were responsible for the majority of emissions in CO2 equivalents (76–104%, Fig. 7). The spring–summer period contributed most to emissions in CO2 equivalents, with diffusive emissions of CO2 and CH4 contributing the most in Alameda and Cedar Bog, and CH4 ebullition contributing the most in Materion and Cleveland-Roselawn (Fig. 7).

Table 2 Total CO2 and CH4 fluxes for each pond (mol m−2 yr−1)
Full size table
Fig. 7
figure 7

Total emissions in CO2 equivalents from all ponds for each period (Summer = spring–summer period, Fall = fall turnover period, Ice melt = Ice-out fluxes). CH4 fluxes were multiplied by 27.2 on a mass bases to obtain CO2 equivalents (g CO2-eq. m−2)

Full size image

Discussion

Measurements of CO2 and CH4 concentrations and fluxes across four north-temperate, urban-exurban ponds with differing mixing regimes showed that mixing and stratification can influence greenhouse gas emission dynamics. All ponds were large sources of greenhouse gas emissions in CO2 equivalents on an annual basis, with CH4 emissions responsible for the majority of the warming potential (Fig. 7). Periods of stratification led to a higher anoxic fraction and the accumulation of CH4 in the bottom waters (Fig. 2), and stratified periods had higher diffusive CH4 emissions and lower CO2 fluxes (Fig. 4). Diffusive emissions of CO2 and CH4 contributed to the majority of the annual emissions in the two ponds with the lowest mixing frequency (81–91%), while ebullitive CH4 emissions contributed the majority of annual emissions in the ponds with more frequent mixing (76–81%). Prolonged stratification during the summer also led to increased emissions of both CO2 and CH4 during the fall turnover period, whereas little stratification in two of the ponds resulted in low fall releases (Fig. 3, Fig. 7).

Mixing regime influence on CO2 and CH4

Despite the four ponds having similar surface areas and maximum depths (Table 1), they exhibited distinctly different mixing regimes (Fig. 1). The most frequently mixed pond (Cleveland-Roselawn) in this study was mixed 28% of the time from 1 June through 31 August, while the most rarely mixed pond (Alameda) was continuously stratified through most of April-August despite being less than 2 m deep. These results agree with recent studies demonstrating that even small, shallow ponds can be strongly stratified during the summer (Holgerson et al. 2022; Ahmed et al. 2023; Loewen & Jackson 2023). For Alameda specifically, a large conductivity gradient was responsible for stratifying the pond during the spring and early summer (Fig. S2), a phenomenon that can be common in temperate urban ponds that receive road salt inputs (Loewen & Jackson 2023). While the conductivity gradient lessened throughout the summer, the temperature gradient kept the pond stratified until early fall. All three stormwater ponds saw the largest increase in conductivity under the ice in late winter due to road salt runoff (Fig. 1, Fig. S2–S5), though only Alameda had a conductivity gradient persist into the late spring.

Among all ponds, CO2 and CH4 dynamics varied under mixed and stratified conditions during the spring and summer months. Gas concentrations in the bottom waters were higher during stratified periods (Fig. 2), similar to other recent studies in shallow lakes demonstrating the accumulation of gases during stratification (Davidson et al. 2023; Søndergaard et al. 2023). CH4 specifically had significant accumulation during longer periods of stratification (p < 0.001, Fig. 2), and the most stratified pond in this study (Alameda) had the highest CH4 concentrations in the bottom waters (Fig. 3). Periods of stratification also led to anoxic conditions (Fig. 2), which increases CH4 storage by limiting CH4 oxidation (Bastviken et al. 2008; Søndergaard et al. 2023). Due to the shallow depths of these ponds, diffusion across the water column can lead to increased emissions when bottom water concentrations are elevated (Ray & Holgerson 2023), and stratification ultimately led to higher diffusive emissions of CH4 (Fig. 4). Mixing events during the summer likely led to a release of stored CH4 and pulses of higher emissions, however the biweekly sampling frequency in this study was not sufficient to consistently capture emissions during mixing events. Methane emissions were lower during mixed periods in the summer, coinciding with lower surface and bottom CH4 concentrations and higher oxygen levels.

During the spring and summer, CO2 fluxes generally followed an inverse pattern to CH4 (Fig. 3) and CO2 fluxes were lower during stratified periods (Fig. 4a). Stratification isolates the surface waters and can disconnect surface photosynthesis from benthic respiration (Jensen et al. 2022; Ray & Holgerson 2023). Conversely, during mixed periods accumulated gas and CO2 produced from the benthos can vent directly to the atmosphere (Huotari et al. 2009). Stratification during the summer also led to increased CO2 and CH4 emissions during fall turnover, when cooling air temperatures broke down stratification and released stored gases (Huotari et al. 2009; Jensen et al. 2022). The two most stratified ponds, Alameda and Cedar Bog, had substantial fall turnover fluxes of both CO2 and CH4 while Materion and Cleveland-Roselawn did not. These fluxes occurred after the water columns first mixed (Fig. 3), potentially coinciding with cooler temperatures that caused mixing deeper into the sediments.

While mixing regimes impacted intra-pond patterns in CO2 and CH4 dynamics, mixing alone did not control the magnitude of emissions among different ponds. Though the two ponds with the lowest mixing frequency (Alameda, Cedar Bog) had the highest annual CH4 emissions, Cedar Bog had the highest CH4 emissions throughout the year despite mixing more frequently than Alameda. As a natural forested pond, Cedar Bog receives the most terrestrial inputs of the four ponds and had the highest DOC and TP concentrations (Table 1). This, combined with deep, carbon-rich sediments typical of dystrophic lakes and ponds (Tranvik 2009) could have fueled the high rates of CH4 production and emission in Cedar Bog (West et al. 2012; Beaulieu et al. 2019) compared to the three stormwater ponds. Surface CO2 concentrations and fluxes also varied by pond, and CO2 fluxes can be impacted by variability in terrestrial carbon inputs and sediment respiration, as well as carbonate buffering controlling the shift of DIC species (Holgerson 2015; Webb et al. 2019). Though CO2 fluxes were lower during stratified periods, the two ponds with the lowest mixing frequency had the highest CO2 emissions. These two ponds (Alameda and Cedar Bog) had low surface water oxygen concentrations well below saturation even into the fall (Table S1), indicating high aerobic respiration rates and/or methanotrophy that impacted the whole water column. In lakes with high DOC concentrations there can be higher carbon consumption through methanotrophy than aerobic respiration (Reis et al. 2022), and methanotrophy may be an important source of CO2 in these ponds.

CH4 ebullition

In all ponds, CH4 ebullition was a significant contributor to greenhouse gas emissions, accounting for 29–72% of the total CO2 and CH4 emissions in CO2 equivalents (Fig. 7). The pond with the highest mixing frequency (Cleveland-Roselawn) also had the highest ebullition rates despite having the lowest diffusive CH4 flux rates. Previous research has shown that CH4 ebullition is strongly influenced by temperature (DelSontro et al. 2016), and the mixed water column in Cleveland-Roselawn meant that the sediments were warmer than those in other ponds (Fig. 5a, Table S1). Conversely, Alameda had the lowest ebullition rates, stratified in late May, and had the coldest sediment temperatures throughout the summer. Macrophytes may also have impacted ebullition rates by providing plant-derived carbon to the sediment or facilitating bubble formation through the release of photosynthetically produced oxygen (Desrosiers et al. 2022; Hilt et al. 2022). Cedar Bog and Cleveland-Roselawn had high densities of submerged macrophytes and the highest ebullition rates throughout the open-water period, whereas Alameda and Materion were largely devoid of submerged macrophytes, which may have contributed to their lower ebullition rates.

Ebullition is inherently variable and episodic (Baron et al. 2022), yet all four ponds followed a similar trend in ebullition rates throughout the summer (Fig. 5b). As expected, ebullition increased in all ponds in early June correlating with warming temperatures, but then unexpectedly decreased in mid-summer followed by another increase in late summer. While influenced by temperature, ebullition is also positively influenced by nutrient concentrations and productivity (Davidson et al. 2018). The large ebullition peak in June could be a result of both warming temperatures and nutrient availability from winter and spring runoff, followed by a decrease in ebullition once labile organic matter had been consumed. In late summer, internal production leads to an abundance of autochthonous labile organic matter, which has been shown to greatly enhance CH4 production over allochthonous material (West et al. 2012; Emilson et al. 2018). This autochthonous production could be driving the second peak in ebullition, before cooling temperatures cause ebullition fluxes to decrease in the fall.

Seasonal contributions to total CO2 and CH4 fluxes

All ponds were annual sources of greenhouse gases in CO2 equivalents (Fig. 7), though Materion and Cleveland-Roselawn were sinks of CO2. We only measured daytime fluxes of CO2 and CH4, and thus would have underestimated CO2 fluxes if they were lower during the day (when primary production exceeds respiration) compared to night (when respiration continues). However, recent studies have found no significant differences in daytime and nighttime CO2 and CH4 concentrations in small ponds (Bergen et al. 2019).

Only the two most stratified ponds, Alameda and Cedar Bog, had significant CO2 and CH4 contributions during the fall turnover period, contributing 20–21% of yearly fluxes in CO2 equivalents, and even a higher proportion of CO2 fluxes (33 and 66%, respectively). These two ponds behaved similarly to larger lakes, which can have a significant proportion of yearly CO2 and CH4 fluxes occur at fall turnover (Riera et al. 1999; Vachon et al. 2017).

Under the ice, concentrations and accumulation rates of CO2 and CH4 were likely driven by different factors than in the open water period. Macrophyte biomass from summer growth has been shown to fuel under-ice respiration and increase oxygen depletion in shallow lakes (Rabaey et al. 2021), and here the two ponds with dense summer macrophyte growth (Cedar Bog and Cleveland-Roselawn) had the highest accumulation rates of CO2 and CH4 under the ice, despite Cleveland-Roselawn having the lowest CO2 and CH4 concentrations during the open-water period. Alameda and Materion summer growth was dominated by duckweed and algae, respectively, both of which likely degraded quickly before the onset of ice in the fall.

Fluxes of CO2 and CH4 at ice-out contributed 6–11% of total annual CO2 and CH4 fluxes in each pond, which is less than ice-out contributions found in a synthesis of temperate, boreal, and arctic lakes (mean 17% for CO2 and 27% for CH4, Denfeld et al. 2018) and in several ponds and shallow lakes in northern Sweden (15–30% for CO2 and 26–59% for CH4, Jansen et al. 2019). The smaller annual percentage contribution of ice-out flux in the ponds in this study was likely due to the high rates of open-water CH4 flux, which were 1.3–4.1 mol m−2 yr−1 CH4 in these study ponds, compared to 0.13–0.31 mol m−2 yr−1 CH4 in the Swedish ponds. The water and sediment in these temperate ponds reached much higher temperatures compared to the Swedish ponds in the summer, which could explain why ice-out fluxes were similar but summer fluxes were higher in the ponds in this study.

The spring–summer period was responsible for the majority of fluxes in CO2 equivalents in all four ponds, largely due to summer CH4 diffusive and ebullitive fluxes. This seasonal pattern is different than most temperate lakes, which can have the highest CH4 diffusion rates during turnover events, though CH4 ebullition is usually highest in the warmest months across most aquatic systems (Johnson et al. 2022). Summer fluxes of CH4 in these ponds (mean of 27.3 mmol CH4 m−2 day−1 for diffusive and ebullitive fluxes) were extremely high compared to temperate lakes (3.3 mmol CH4 m−2 day−1, DelSontro et al. 2016; 4.6 mmol CH4 m−2 day−1, Johnson et al. 2022), but on a similar scale to measurements of CH4 fluxes in small ponds (20–58 mmol CH4 m−2 day−1, Gorsky et al. 2019; Grinham et al. 2018; Rabaey & Cotner 2022).

Implications

Overall, the shallow nature and warmer temperature of these ponds relative to deeper lakes leads to enhanced greenhouse gas production and release to the atmosphere, particularly for CH4. Rather than CO2 and CH4 accumulating in the hypolimnion, as occurs in deeper systems, these gases are rapidly exchanged with the atmosphere even if gases accumulated during stratification. Furthermore, the ponds that mixed more frequently in this study also released less total CH4 to the atmosphere, likely due to increased DO and more complete CH4 oxidation (Whitman et al. 2006). Although other mechanisms clearly impacted emissions, these results suggest that ponds that stratify extensively can be particularly large sources of CO2 and CH4. For stormwater ponds in particular, increasing urbanization and road salt application is leading to more frequent and prolonged stratification (Loewen & Jackson 2023), and, coupled with warmer temperature under climate change, may increase urban greenhouse gas emissions in the future. Promoting mixing and aeration in stormwater and other shallow ponds with high emissions could be a useful greenhouse gas abatement strategy in urban areas (Hounshell et al. 2021), and more work is needed to see if artificial mixing could reduce emissions from ponds.

Conclusion

Half of global CH4 emissions come from highly variable aquatic sources (Rosentreter et al. 2021), and small ponds have amongst the highest CH4 fluxes and highest variability of all aquatic ecosystems (Holgerson & Raymond 2016). Here, we show that pond mixing regimes play a large role in determining intra-pond variability in the seasonality and type (CO2 vs CH4, diffusion vs. ebullition) of greenhouse gas emissions. Though all four ponds in this study were shallow (< 2 m) and had similar surface areas (0.9–1.5 ha), mixing regimes varied from stratifying for most of the summer (June–August) to mixing almost nightly throughout the open water period. Ponds that stratified behaved more like temperate lakes with significant fluxes of CO2 and CH4 at fall turnover when the water column and sediments mixed, while mixed ponds had almost no fluxes during the fall. Stratification during the open-water period led to the accumulation of gases, more extensive anoxia, and greater diffusive fluxes of CH4. Most of the greenhouse gas fluxes occurred in summer, due to the large contribution from summer CH4 diffusion and ebullition. Overall, considering differences in pond mixing regimes is crucial for upscaling emission rates from ponds, and can help account for the large variation in pond greenhouse gas emissions.