Introduction

Small lakes and ponds (<0.01 km2) have emerged as important water bodies for the processing and transportation of carbon (C)1,2,3,4. Of an estimated 0.583 Pg C yr−1 emitted from lakes and ponds globally, ~24% is from waterbodies <0.01 km2 despite only accounting for ~15% of total pond and lake surface area2. Similarly, small lentic artificial freshwaters (e.g., ponds, reservoirs, ditches, infiltration basins) are estimated to store large quantities of organic carbon (OC) in sediments globally, more than natural ecosystems and up to 4-times as much as the world’s oceans annually1.

In urban ecosystems, permanently wet stormwater ponds (SWPs) are suspected to play a large role in C processing due to characteristics similar to those of other small lentic systems. For example, both SWPs and natural ponds have individually small sizes (commonly <0.01 km2 2,5), shallow water columns, high sediment and perimeter to water volume ratios, and frequent mixing. These pond features contribute to rapid accumulation of C, high internal productivity during the growing season, frequent heterotrophy, and high C gas evasion, all to a disproportionately higher degree than larger water bodies1,2,3,4. However, SWPs may differ from natural ponds in C cycling properties due to their differences in hydrology (engineered in and outflow structures), elevated allochthonous resource inputs from urban landscapes, and intensive water quality and sediment management practices (i.e., dredging, algaecide application).

SWPs are anthropized ecosystems constructed to provide specific ecosystem services such as flood management and pollutant removal by capturing stormwater runoff from the surrounding watershed via drainage infrastructure (e.g. gutters, curbs, roads, drains). This runoff can carry high loads of biologically significant materials such as dissolved and particulate organic matter with varying lability, C, and essential nutrients (nitrogen (N), phosphorus (P))6,7. Further, SWPs are ubiquitous across urban landscapes, representing the most common stormwater control measure in the U.S8. In Florida, they are estimated to cover an area of 672 km2 5 and are prolific in eastern U.S coastal communities9,10. Despite this widespread presence, SWPs are overlooked and under-researched and little is known about their role in regional C cycling or their net benefit to society (services vs. disservices). Still, the growing body of research on artificial waters and urban ponds has shown that they are capable of emitting substantial quantities of carbon dioxide (CO2) and methane (CH4) greenhouse gases (GHGs) and potentially more so than natural ecosystems, up to 2.5x more compared to natural ponds <10,000 m2 in area11, despite high C burial rates12.

As urbanization and SWPs replace natural uplands and downstream receiving waters (e.g., wetlands, streams, ponds), changing the movement of water, solutes, and energy in the landscape, there is an increasing need to understand the capacity of SWPs to store C in sediments and exchange C gas with the atmosphere. Studies of urban ponds or impoundments around the globe have reported burial rates ranging from 8.7 g C m−2 y−1 in urban SWPs (South Carolina, USA)9 to 2120 g organic C (OC) m−2 y−1 in eutrophic impoundments13. In addition to receiving high terrestrial C loads, SWPs can be highly internally productive, with a significant proportion of the OC pool being autochthonously produced4,6,14. Despite high rates of C accumulation in sediments, small ponds are typically net C sources, driven by CO2 and CH4 emission to the atmosphere. For example, in terms of CO2 equivalents, 96% of artificial ponds and ditches from seven countries were net sources of GHGs to the atmosphere11. Similarly, an investigation of 77 small agricultural impoundments (Victoria, Australia) showed that they accounted for 3.4x higher CO2 equivalent fluxes than temperate reservoirs while having only 0.94 times the comparative area15. Mean CO2 equivalent fluxes have been reported as high as 9.2 g CO2 eq m−2 d−1 in SWPs, with 38% from diffusive CO2 fluxes and 62% diffusive + ebullitive CH4 fluxes (The Netherlands)16. However, C fluxes from small ponds are highly variable17 and urban ponds may be particularly heterogenous due to variation in pond management, urban infrastructure, and watershed activities. While many studies have separately estimated GHG emissions or C burial in artificial ponds, rarely have studies combined both to compare net C budgets (see ref. 16,18), which can allude to their net benefit to society. Understanding the drivers of C source/sink dynamics in SWPs can enhance our ability to manage these ecosystems to reduce the negative impacts stormwater control measures might pose on downstream ecosystems and to regional climate change.

Ponds and lakes can undergo successional development as sediments accumulate over time, which can be associated with successional changes to pond biogeochemistry. For example, sediment organic matter, summer nitrogen limitation, sediment oxygen demand, and net N2-fixation rates have been shown to increase with increasing SWP age19. Similarly, C burial rates were considerably lower in new (3 year old) artificial ponds compared to mature (18–20 year old) artificial ponds12. Ecosystem age has also recently been found to be a strong predictor in empirical models (G-res tool) for the diffusive emission of C gases from reservoirs, especially at higher temperatures20. Observing the temporal C source/sink dynamics of SWPs and patterns over a pond age gradient can provide insight into how these inland urban waters behave over short (annual) and long (decadal) timelines. In this study, we quantify rates of C burial as well as CO2 and CH4 gas fluxes in five SWPs spanning a 20-year age gradient (14, 15, 18, 23, and 34 years old; ponds identified based on pond age hereafter) in southwestern Florida. We measured gas fluxes using a floating chamber method over 10 months (June 2019 – March 2020) and compared fluxes against pond age, measured environmental factors, and C burial rates to identify potential drivers of gas fluxes and estimate the net C flux of urban SWPs. Our results show that subtropical urban SWPs can sequester large quantities of C in their sediments. As ponds matured in age, C burial increased and C emissions (CO2–C + CH4–C) decreased, suggesting that ponds may shift toward a net C sink as they age.

Results and discussion

Carbon burial rates increased with pond age

Areal and annual C burial rates were calculated for four to six sediment cores extracted per pond in June 2019 using total OC (TOC), dry weight, and piston corer dimensions. Mean TOC content in whole cores of ponds ranged from 1.7 to 26.5% (Table 1 and Supplementary Fig. 1). The areal OC stock ranged from 306 ± 8.2 g OC m−2 in Pond 14 (14-year-old pond) to 7380 ± 56.1 g OC m−2 in Pond 34 (34-year-old pond), increasing linearly with increasing pond age (R2 = 0.99, p < 0.001, Fig. 1a and Table 1). Scaling this OC stock by pond age (assuming constant burial rates over time) yields estimated annual burial rates ranging from 21.8 ± 7.5 g OC m−2 y−1 in Pond 14 to 217 ± 50.2 g OC m−2 y−1 in Pond 34, with annual rates increasing logarithmically with pond age (R2 = 0.93, p < 0.01, Fig. 1b). These OC burial rates, especially in the 23- and 34-year-old ponds (141 and 217 g OC m−2 y−1, respectively), are comparable to or exceed previously reported rates from a range of aquatic ecosystems (Supplementary Table 1) including critical blue C sinks such as mangrove forests, salt marshes, and seagrasses21. The OC burial values reported here are also similar to reported rates in other SWPs and constructed systems9,16,22,23, but remain below results for some impoundments around the globe (Supplementary Table 1).

Table 1 Pond morphology, water chemistry, and C burial data (mean ± standard error).
Full size table
Fig. 1: Sediment organic carbon stocks and burial rates in stormwater ponds.
figure 1

Organic carbon stock (g OC m−2, a) and the estimated rate of burial (g OC m−2 y−1, b) increase with pond age. Individual core values (n = 4–6 per pond) are plotted as small gray points and the larger black points represent the pond mean, which was used for regressions. Both linear and logarithmic regressions are statistically significant (p < 0.01).

Full size image

The relationship between C burial and pond age suggests that internal sediment and/or aquatic properties allow ponds to become increasingly stable environments for C deposition and storage. Burial rates were not related to pond morphology (perimeter, area, depth, area:perimeter). SWPs are unique ecosystems that receive a dramatic quantity of sediment from impervious and piped catchments. It is possible that over time and as more particulate material is added to the sediment surface, sediment thickness and anaerobic conditions beneath newly deposited material increase, slowing down the rate of OM decomposition. A higher degree of recalcitrant OM can also be left behind, increasing the pool of stable OM more resistant to degradation. Further, the addition of SWPs to newly developed landscapes is associated with the addition of new landscaping vegetation to the urban watershed, such as trees which can increasingly contribute leaf litter to ponds as they mature over time, suggesting that allochthonous C inputs may increase concurrent with in-pond conditions favorable for C burial. Natural factors such as precipitation, length of the growing season, and littoral vegetation also affect C accumulation in urban SWPs9,22. In addition to allochthonous inputs, SWPs can support highly productive algal communities and phytoplankton turnover represent an internal OC input24. Littoral vegetation may provide a similar autochthonous OC source, but sites in this study contained no littoral vegetation and also had not been dredged (a common management practice). Despite the apparent relationship between age and C burial, our sample size is small, and this relationship may be spurious. A prior study found no relationship between sediment OC and age in an analysis of 45 aquaculture ponds25, but the aquaculture activities may have offset natural successional dynamics. Furthermore, an experimental study of small ponds found that new ponds (3 years old) exhibit significantly less C burial in sediments than mature ponds12. Nonetheless, in the process of preventing urban sediment from moving into downstream waters, constructed SWPs have emerged as significant reservoirs of sediment C, as shown here and in other anthropized ecosystems (reservoirs, impoundments, SWPs) or small ponds3,9,17,22,23.

In Florida, there are at least 76,000 SWPs5. Given the role of SWPs in storing sediment C and their ubiquity on the landscape, these anthropized ecosystems represent a large pool of C in urban aquatic sediments. Using the mean OC burial rates in the youngest and oldest pond combined with the estimated surface area covered by SWPs in Florida (627 km2)5, we estimate that SWPs in Florida can store between 0.2 and 4.6 Tg C annually. The upper end of this estimate would be enough to offset 1.4% of the global CO2 evasion from inland lakes and reservoirs26.

Carbon gas fluxes are controlled by physical and chemical factors within ponds

The mean floating chamber flux of CO2 over the study period (biweekly measurements June 2019 to March 2020) was 5940 ± 857 mg CO2 m−2 d11 with a range of −13600 to 34300 mg CO2 m−2 d −1 (Supplementary Fig. 2). CH4 fluxes were consistently positive but more variable and lower than CO2 fluxes with a mean of 47.8 ± 17.1 mg CH4 m−2 d−1 and range of 0.12–1400 mg CH4 m−2 d−1 (Supplementary Fig. 3). This maximum CH4 flux was accompanied by an unexplained water quality change that coincided with cloudy white water and DO levels depressed to 0–5% throughout the water column. Using median values, annual fluxes equate to 1290 g CO2 m−2 yr−1 and 5.0 g CH4 m−2 yr−1. The median estimate for CH4 is far below the 2019 Intergovernmental Panel on Climate Change estimate for manmade freshwater ponds (18.3 g CH4 m−2 yr−1), although the mean CH4 flux in this study is on par at 17.4 g CH4 m−2 yr−1 27. On a C mass basis, the daily mean flux of C gases combined (CO2-C + CH4-C) equates to 1636 mg C m−2 d−1 (CO2 making up 98%), and an annual flux of 324 g C m−2 yr−1 (using the median). Diffusive CH4 contributed a small percentage of total pond C fluxes in this study. Converting CH4 to CO2 equivalents with a sustained-flux global warming potential of 4528 and summing with CO2 fluxes results in a mean of 8090 mg CO2 eq m−2 d−1, of which CH4 makes up 27%, and annual median of 1660 g CO2 eq m−2 yr−1. This is a smaller fraction of CH4 contribution to overall CO2 equivalent fluxes compared to other urban SWPs such as 94% of 28.2 g CO2 eq m−2 d−1 (excluding ebullitive emission)29 and 62% of 9.2 g CO2 eq m−2 d−1 (including ebullitive emissions)16. Using the median CO2-equivalent fluxes in the oldest and youngest pond combined with the estimated surface area covered by SWPs in Florida (627 km2)5, we estimate that SWPs in Florida can emit between 52 and 204 Tg CO2-equivalents annually (or 12–50 Tg C annually).

Ponds were a source of CO2 over the vast majority of the study period excluding the two oldest ponds, which reported negative flux values during 6 of 17 sampling events each (35%), reaching as low as −507 mg CO2-C m−2 d−1 (Pond 23) and −3730 g CO2-C m−2 d−1 (Pond 34, Supplementary Fig. 2). Ponds 15 and 18 only exhibited negative CO2 gas flux one time each. These results suggest that at some points in the year, older ponds may switch from net heterotrophy to net autotrophy. A gradual decrease in mean CO2 flux with increasing pond age can be seen in Supplementary Fig. 3, but the same pattern is not evident with CH4 fluxes. Similarly, a survey on hydroelectric reservoirs around the globe (n = 85) found that CO2 and CH4 emissions were inversely related to reservoir age30 a pattern that was also found on a temporal study of a single hydroelectric reservoir (CO2 only)18.

The CO2 fluxes from this study were similar to or higher than other aquatic ecosystems, CO2 fluxes were similar or higher, whereas CH4 fluxes were typically lower than other urban studies (Supplementary Table 2). Ebullition is considered an important pathway for methane emission in lentic ecosystems with anoxic sediments31. Because we did not measure ebullition our values for overall CH4 flux are likely underestimated. Small patches of bubbles were occasionally captured underneath our flux chamber, causing the internal concentration of CH4 to spike to ~130,000 ppb. For reference, the global mean atmospheric CH4 concentration for 2019 was 1870 ppb32. There was not a similar spike in CO2 concentrations when bubbles were captured.

Linear mixed-effects models (LMM) were used to estimate the relationship between gas fluxes and environmental variables measured in this study (Supplementary Table 3). For both CO2 and CH4 models, site (five levels) and sampling date (seventeen levels) were set as random effects and pond age was a similar fixed effect. Including the variation from random effects, 68% of the variation in CO2 fluxes was explained by inverse relationships with surface % DO, pH, pond age, and the interaction between %DO and pH. In addition to the negative relationship between CO2 and pH and the observed decrease in mean CO2 flux with age (Supplementary Fig. 3), an associated observation was the slight increase in mean pH with pond age, up to 10.4 in Pond 34 (Fig. 2 and Supplementary Fig. 4). Surface %DO and pH were strongly related to CO2 in the LMM, and when combined can be considered indicators of primary production. Photosynthetic fixation of CO2 from the water column results in an increase in pH and DO. Concurrently, the chemical form of CO2 in the water column (and its concentration) is controlled by pH, speciating into carbonic acid, bicarbonate, and carbonate at higher pH values. According to the kinetics of this reaction, free CO2 in freshwater becomes depleted at pH > 8.3. Of the 14 observations of negative fluxes, all but one occurred above pH 8.3 (Fig. 2). So, while primary production contributes to a decrease in CO2 emissions, the increased pH can exert a compounding effect, rendering high pH ponds and lakes to be weak CO2 sources, similar to observations in in other urban ponds11, saline lakes33, and in agricultural reservoirs34. Other abiotic factors may contribute to an increase in pH. Florida is known to have shallow carbonate bedrock which, coupled with the weathering of urban concrete infrastructure (calcium hydroxide, portlandite, calcium silicate hydrates), can contribute carbonates and calcium salts, increasing alkalinity in aquatic ecosystems35,36. Additionally, in high CaCO3 systems, inputs of CO2 can contribute to increased alkalinity over time via calcium release. Thus, ponds have the potential to become CO2 sinks over time as a result of increasing eutrophication and accumulation of natural- and anthropogenically sourced alkaline ions increasing pH.

Fig. 2: Carbon dioxide fluxes in stormwater ponds as a function of surface water pH.
figure 2

CO2 fluxes decrease with increasing surface water pH. In agreement with CO2-carbonic acid equilibrium, a majority of CO2 fluxes result in negative values at pH > 8.3, the point at which water column CO2 is nearly all converted to carbonate or bicarbonate. The dotted line denotes zero CO2 flux.

Full size image

Compared to CO2, a lower amount of the variation in diffusive CH4 flux was explained by variables in this study. With accounted variance from random effects, 51% of CH4 flux variation was explained by an inverse relationship to benthic %DO and positive relationships to benthic temperature and log-transformed CO2 fluxes (Supplementary Table 3). There was no relationship between CH4 flux and site age. The bottom of most pond water columns remained hypoxic throughout the study period with 65% of the data below 3.0 mg/L DO, creating conditions beneficial for anaerobic methanogenesis, which can produce CH4 alone (by hydrogenotrophic methanogens) or CH4 and CO2 (by acetoclastic and methylotrophic methanogens). The positive relationship between CH4 and CO2 could be due to this simultaneous production, or due to increased production of CH4 using CO2 and H2 as substrates as a result of its increased availability (by hydrogenotrophic methanogens). CO2 may also be attributed to aerobic or anaerobic CH4-oxidation. The negative relationship with water column DO was also indicated during the extreme methane emission during DO depletion (0−5% DO through the water column) in Pond 14. Finally, sedimentary methanogens typically become more active at higher temperatures as a result of increasing use of higher redox potential electron acceptors as well as a higher input of primary production-derived substrates that fuel their metabolic activity37,38.

Younger ponds are net sources of C to the atmosphere

We estimated a C balance for each pond by comparing median organic C burial to median C gas flux (CO2-C + CH4-C flux) (Fig. 3). The combination of C gas flux (C export) and burial (C storage) reveals that three of the five ponds were, on average, net sources of C to the atmosphere. However, the median burial values fall within the 5th−95th quantile range for C efflux in Pond 23 and 34, the net burial ponds (Fig. 3 and Table 1). It should be noted that these estimates compare rate measurements determined at separate time scales (seconds vs multi-decadal). As pond age increased, the difference between C gas efflux and C storage in sediments decreased, implying that ponds may become more C-neutral as they age. The degree of this net benefit in older ponds may change if methane ebullition estimates had been included, as others have found that urban pond ebullition accounted for up to 50% of total CH4 emission16. Previous studies have identified increasing CH4 emission with ecosystem age18 and increasing trophic level39, which can be associated with pond succession, converting ponds into net sources in terms of CO2-equivalents due to increasing CH4 contributions. Regardless of pond age, this study provides evidence that SWPs, which are constructed to provide ecosystem services, may be providing a disservice by acting as net contributors to GHG emissions, despite storing substantial quantities of C in sediments. Other studies of small ponds support this assertion. For example, both Peacock et al. (2019) and Ollivier et al. (2019) observed release of GHGs from urban and small agricultural ponds at rates high enough to call for their inclusion in global carbon budgets.

Fig. 3: Net carbon flux in stormwater ponds based on sediment burial and gaseous emission.
figure 3

A comparison of median C flux (blue circles, CO2–C + CH4–C) as positive values and median OC burial rate (brown circles) as negative values. The net C flux (g C m−2 y1) in stormwater ponds is shown by the red circles and line and represents the difference between gas flux and burial. Bars on points indicate the 5th and 95th quantiles.

Full size image

C burial and C gas fluxes have been compared in other anthropized ecosystems. In a 20-year old urban pond (the Netherlands) the annual flux of CO2-C and CH4-C combined, 391 g C m−2 yr−1, was substantially higher than the pond’s rate of burial at 29 g C m−2 yr−1. The aforementioned pond exhibited a lower burial rate and annual emission than Pond 18 (and younger sites) in our study (burial = 126 g C m−2 yr−1, emission = 429 g C m−2 yr−1). In a 4-year-old (time since flooding) hydroelectric reservoir, daily estimates of flux and burial were 2.3 g C m−2 d−1 emission vs. 0.1 g C m−2 d−1 burial18.

Our results suggest that after urban ponds are constructed, they emit large proportions of C inputs from the landscape and potentially increase storage efficiency over time. We suspect that when C enters urban SWPs, less C is buried in younger ponds compared to older ponds, as seen in burial studies on aging artificial ponds12. Younger ponds may exhibit higher C mineralization rates which simultaneously increases emissions while reducing burial. However, identified trends can be drastically altered depending on how ponds are managed, suggesting management actions can be used to modulate pond C storage and C emission. For example, increasing urban pond depth is associated to reduced CH4 release29, likely due to increased water column contributions to organic matter oxidation prior to sediment settling and CH4 oxidation, and ponds or lakes larger in area exhibit lower C emissions per unit area compared to smaller ponds2. Other suggested properties for anthropized ecosystem C management include water column stratification, water retention time, water source inputs, and landscape position34. It is important to note that drivers of C emission can differ between natural and anthropized ecosystems such that CH4 emissions from reservoirs globally were correlated to productivity and fluxes increased with increasing area40. In contrast, natural lakes were better explained by morphometry and fluxes decreased with increasing area40.

Methods for pond maintenance include sediment dredging and muck removal, algaecide application, aeration systems (fountains, bubblers), fish stocking, and vegetation management41. Although ponds included in this study have not been dredged, dredging could reset the ‘effective pond age’. Dredging may not effectively reset a pond, however, as sediment resuspension that occurs in the process has caused lakes to relapse into a eutrophic state because of the reintroduced availability of sediment nutrients to the water column42. The application of aquatic pesticides (i.e., chemical treatments for algal blooms and invasive macrophytes) in SWPs is likely to impact C cycling such that substantial amounts of labile OC are rapidly contributed to the sediment surface, influencing sediment oxygen demand and the respiration of CO2 and CH4. Fish are added to ponds for recreation as well as mosquito and algal consumption, and can have varying effects on CO2-equivalent fluxes. For example, by grazing on zooplankton that consume both algae and CH4-oxidizing bacteria, fish can elicit a trophic cascade that causes an increase in algal productivity and CH4 flux and reduced CO2 flux39. However, fish can also graze on predators to CH4-oxidizing bacteria, decreasing CH4 flux43. The presence of vegetation may also be beneficial for reducing GHG emission, as seen in previous studies44 and is responsible for increasing C burial rates22. Finally, aeration systems are commonly used to prevent anoxia and could therefore reduce CH4 emissions, as shown in experimental studies on hypolimnetic DO manipulations45.

The number of small constructed ponds was found to increase 18-fold from 1937 to 2005 in the Brandywine watershed of central Pennsylvania and northern Delaware, and land use change associated with urbanization will likely continue to increase the numbers of small constructed ponds regionally and globally46. As ubiquitous anthropized aquatic ecosystems such as SWPs continue to be constructed in urban landscapes and have emerged as significant contributors to GHGs and C storage, more work is required to assess their net C footprint, patterns with ecosystem age over a larger number of sites, and biogeochemical responses to management strategies. An improved knowledge of how management actions interact with natural ecological processes is critical for understanding the role of SWPs and other small, anthropogenic aquatic ecosystems in the global C cycle.

Methods

Study site

The study ponds are in Manatee County, Florida, which is part of the Tampa Bay metropolitan area. It has a subtropical climate, with mean annual temperature of 22.8 °C and mean annual rainfall of 140 cm, 60% of which occurs during a distinct annual wet season from May to September. To assess C processing in pond systems, sites selected were permanently wet ponds and had no littoral or emergent vegetation. Submersed macrophytes were not accounted for in this study. Additionally, each pond was surrounded by homes with similar landscape conditions (i.e., ornamental plants, trees, turf grass) and contained no aeration systems. The only exception is Pond 18, which has drainage connection to a residential golf course in addition to the residential neighborhood. A potentially important difference between each pond is the number of inflow and outflow structures, which may impact a pond’s total incoming water volume from either surface runoff or other ponds that drain into it. All ponds in this study had one outflow structure and at least one inflow structure, but these were not fully accounted for.

The age of ponds was determined using multiple sources such as the Manatee County Property Appraisal and Google Earth. Google Earth was used to observe aerial photos that date back to 1995. For gap years and ponds older than 1995, ages were determined using aerial photos and housing information on the appraisal website, with the assumption that ponds were constructed at the same time as the houses around it.

Water properties

C cycling processes can be affected by various water properties, such as temperature, pH, salinity, conductivity, and dissolved oxygen (DO). Along with each floating chamber measurement, we recorded these water quality parameters from both surface and bottom water column locations (benthic samples) at three locations evenly distributed in open water at least five meters from littoral shelves. Measurements were taken with a YSI ProDSS sonde. Probe calibration was conducted on the day of each sampling event. Surface and bottom measurements were recorded within 30–60 cm of the water’s surface and above the sediment surface, respectively. Samples for dissolved organic carbon and total dissolved nitrogen were filtered with a 0.45 μm filter and analyzed on a Shimadzu TOC-L Total Organic Carbon Analyzer coupled with a TNM-L Total Nitrogen Module.

Sediment OC burial

C accumulation rates were estimated in five ponds over an age gradient to establish C storage within a pond and to identify differences in burial over time. Six sediment cores were collected from each pond. Cores were taken in evenly distributed areas in the center region of the pond away from littoral shelves and banks. Cores were extracted during May and June 2019 using a modified piston corer with a clear sleeve 89.7 cm in height and opening area of 22.88 cm2. The layer of deposited material is de fined as all organic sediment muck found above the distinct pond sand-fill material, which was visually identified in each core through the clear coring sleeves. Due to the difficulty of coring from sand underlain sediments (falling apart before they could be fully collected), some cores were compromised and so replicates range from four to six cores per pond. This muck-sand contact defines ‘time zero’ when the pond was constructed, and material began to deposit. We assume that all muck present has been deposited since the time of construction. Once cores were extracted, all deposited sediment material, excluding sand fill, was extruded and collected. The collected sediment of a core was then homogenized. Sediments were frozen for 24 h and freeze dried to remove moisture. Once dried, each core was weighed for dry weight. Three subsamples of dried sediments were analyzed for total organic carbon (TOC). The three subsamples were averaged to represent the amount of organic C found in each core. TOC was quantified for each core by the UF Wetland Biogeochemistry Lab using a Shimadzu TOC5050A Total Organic Carbon Analyzer. Using a catalytic combustion oxidation method, sediment samples are combusted and analyzed for total carbon (TC) and total inorganic carbon (TIC); TOC is calculated as the difference between the two. Results are reported as g OC per 100 g dry sediment concentrations, or % OC.

Organic carbon stock (g OC m−2) and burial rates (g OC m−2 y−1) were calculated and represented using two equations below. The second considers the cumulative pond age to estimate a burial rate of g OC m−2 yr−1 (Eq. 2). Equation 1 is calculated as the mean C content per core (g C per 100 g dry sediment) times the core dry weight, divided by the core sleeve opening area (m2). Equation 2 is the same as Eq. 1 with the addition of pond age (years) as a multiplier in the denominator.

$${{{{{{\mathrm{Burial}}}}\,{{\mathrm rate}}}}}\left({{{{{{\mathrm{g}}}}}}\,\,{{{{{\mathrm{OC}}}}}}}\,\,{{{{{{\mathrm{m}}}}}}}^{-2}\right)=\frac{{{{{{{\mathrm{TOC}}}}}}}\,\left({{{{{{\mathrm{g}}}}}\,{C}\,{per}}}\,{100}\,{{{{{{\mathrm{g}}}}\,{{\mathrm dry}}\,{{\mathrm sediment}}}}}\right)* {{{{{{\mathrm{Core}}}}\,{{\mathrm dry}}\,{{\mathrm weight}}}}}\,(g)}{{{{{{{\mathrm{Sleeve}}}}\,{{\mathrm opening}}\,{{\mathrm area}}}}}\,\left({{{{{{\mathrm{m}}}}}}}^{2}\right)}$$
(1)
$${{{{{{\mathrm{Burial}}}}\,{{\mathrm rate}}}}}\left({{{{{{\mathrm{g}}}}}}\,\,{{{{{\mathrm{OC}}}}}}}\,\,{{{{{{\mathrm{m}}}}}}}^{-2}{{{{{\mathrm{y}}}}}}{{{{{{\mathrm{r}}}}}}}^{-1}\right)=\frac{{{{{{{\mathrm{TOC}}}}}}}\,\left({{{{{{\mathrm{g}}}}\,{{C}}\,{{per}}}}}\,{100}{{{{{{\mathrm{g}}}}\,{{\mathrm dry}}\,{{\mathrm sediment}}}}}\right)* {{{{{{\mathrm{Core}}}}\,{{\mathrm dry}}\,{{\mathrm we}}}}}{{{{{{\mathrm{ight}}}}}}}\,({{{{{\mathrm{g}}}}}})}{{{{{{{\mathrm{Sleeve}}}}\,{{\mathrm opening}}\,{{\mathrm area}}}}}\,\left({{{{{{\mathrm{m}}}}}}}^{2}\right)* {{{{{{\mathrm{Pond}}}}\,{{\mathrm Age}}}}}\,({{{{{{\mathrm{yrs}}}}}}})}$$
(2)

Floating chamber fluxes of CO2 and CH4

To examine fluxes of CO2 and CH4 from SWPs to the atmosphere we employed a floating chamber method where the vertical rate of flux from the air-water interface was estimated based on the change in gas concentration within the chamber over time and cross-sectional area47. The use of floating chambers is increasingly employed in gas flux studies for its affordability and ease of use, which involves short deployments, and simple operation. Upon deployment, results are only representative of single measurement points but can be used for quantifying spatial variability. CO2 and CH4 gas fluxes were measured biweekly from June 2019 to May 2020 mid-day between 10:00 a.m. and 2:00 p.m. Gas measurements were collected using a LI-COR Smart Chamber as the enclosed floating component attached to a LI-COR LI-7810 CO2/CH4/H2O Trace Gas analyzer that utilizes a non-dispersive infrared method. Detection ranges were 0–100 ppm CH4 (0.25ppb precision) and 0–10,000 ppm CO2 (1.5 ppm precision).

The Smart Chamber (floating chamber) is equipped with an air temperature thermostat (accuracy ± 0.5 °C between 0 and 70 °C) and a barometric pressure sensor (accuracy ± 0.4 kPa between 50 and 110 kPa). It has a volume of 4244.1 cm3 and attaches to a polyethylene collar 11.43 cm in height with an opening area of 317.8 cm2 exposed to the water surface. The collar is fixed onto a buoyant Styrofoam lid with a hole cut in the center, allowing ~14 cm of the collar to extend into the water. The submerged collar walls improve flux accuracy, as the absence of submerged wall extension has reported fluxes 3–5 times greater than when wall extensions are present48.

Ventilation to ambient air is required for the chamber pressure equilibrate with ambient air pressure. The chamber features patented pressure vent technology, characterized as a radially symmetrical vent that allows the chamber to maintain a pressure inside that is equal to that of the ambient air pressure outside under low and high wind conditions regardless of wind direction. This improves the ability of measurements to represent fluxes under natural conditions. Additionally, to account for decreased gas diffusion gradients upon chamber enclosement, an exponential function is used to determine initial slope conditions as soon as the chamber is closed to represent fluxes under ambient conditions. When ebullitive bubbling of sediment gases were caught under the chamber during a measurement, the chamber was reset and measurements re-done to maintain consistency in chamber exponential or linear slopes used to calculate fluxes. Supplementary Fig. 5 shows R2 values for CO2 and CH4 chamber flux measurements. For CH4 measurements, 82% of individual measurements reported R2 of 0.95 and above, and only two observations were below 0.50. CO2 slope R2 were more variable, where only 66% of values were above 0.80. We kept all measurements with significant exponential or linear regressions (p < 0.05), removing thirteen non-significant CO2 measurements and ten for CH4.

On each sampling date gas fluxes were measured at three locations within the pond. The mean of these values was used to represent whole pond fluxes. At each point, the chamber was allowed to rest on the water’s surface for a few minutes before taking measurements. Measurements in the closed chamber were taken for 200–300 s. As gas accumulates in the chamber the sample air is mixed by both the hemispherical shape of the chamber and the position of the inlet and outlet tubing. Sample air is pumped into the trace gas analyzer by a 1.2-m-long tube and detected for CO2, CH4, and H2O by optical feedback – cavity-enhanced absorption spectroscopy. Gas accumulation within a chamber often exhibits exponential increases, therefore an exponential fit of the concentration measured over time is used to determine slope. We compared the R2 of exponential fits to linear fits in case there were occasions where a linear fit was more appropriate. For CO2, there were thirteen out of 236 observations where a linear fit was used and ten out of 239 for CH4. The equation used here to estimate flux rates is a slight modification of that by Duc et al. (2013)49(Eq. 3):

$$F=\frac{{{{{{{\mathrm{d}}}}}}X}}{{{{{{{\mathrm{d}}}}}}t}}\frac{{p}_{a}{Vb}}{{RTA}}$$
(3)

Where F represents the rate of gas flux at the air-water interface (μmol [C gas] m−2 s−1), is the slope of the exponential fit to the change in concentration within the chamber (dC) over time (dt) from the moment of closure (μmol mol−1 s−1). pa is ambient pressure (atm), V is the volume of the Smart Chamber (mL), A is the area of chamber coverage (m2), R the universal gas constant (82.056 mL atm mol−1K−1), and T is ambient temperature (K). b is the molecular weight of either CO2 or CH4 (g μmol−1) and converts μmol m−2 s−1 to g C m2 s−1.

Statistics

All statistical analyses were conducted in the statistical software ‘RStudio’ (version 1.2.5) and ‘R’ (version 3.6.4). Significance in all analyses were results that reported p-values below 0.05. Because of the large sample size in data observations, we relied on histograms and QQ plots to determine univariate normality. In the case of a variable containing non-normally distributed data, log transformation was applied. In the case of CH4, normality was achieved after log-transformation. The same was not true for CO2 which was left untransformed for further statistical analyses. Multicollinearity was assessed by computing pairwise Pearson correlation coefficients between specific outcome variables using the cor_test() or cor_mat() function and homogeneity of variances was assessed using Levene’s test (levene_test() function), all from the R package rstatix50.

To assess the existence of trends in C burial with ponds of differing ages, a linear and logarithmic regression was conducted on aerial rates (g C m−2) and rates using pond age (g C m−2 yr−1), respectively. In both cases, regressions were calculated as burial rate as a function of pond age (yrs). A linear mixed-effects model (LMM) was conducted for each gas to summarize the full effect of correlated variables using the lmer() function of the R package lme451. In both the CO2 and CH4 LMM, pond site and sampling date were set as random effect. Correlated variables for each respective gas were initially determined using Pearson correlation coefficients and then included as fixed effects. Slightly modified versions of each CO2 and CH4 LMM were compared using the Second-order Akaike information criterion (AICc). Surface and benthic DO were correlated to CH4 fluxes in the LMM and were determined to be acting independently on fluxes by reviewing their plots against one another and comparing linear models with CH4 flux that did and did not include an interaction between surface and benthic % DO.