High rates of carbon burial linked to autochthonous production in artificial ponds

Ponds are globally abundant and important to the global carbon cycle. Although ponds have large greenhouse gas emissions, they also sequester carbon in their sediments. Here, we studied organic carbon (OC) burial rates in 22 temperate experimental ponds with negligible watersheds, where carbon sequestration derives solely from autochthonous primary production. The ponds were built identically in 1964 and have since experienced different management strategies, allowing us to test how management actions influence burial rates. On average, the ponds accumulated 67.1 g OC m−2 yr−1 (range 38.3–113.6 g OC m−2 yr−1)—about double the global average burial rate for lakes and similar to global averages for wetlands. Carbon burial rates were higher in ponds with macrophytes, fish, and higher N : P loads. We contend that the global carbon sink for inland waters may be substantially underestimated due to the undervaluation of OC burial in natural and artificial ponds.

that ponds are an important source of greenhouse gases to the atmosphere, ponds also sequester organic carbon (OC) through accumulation and burial of organic material (Downing et al. 2008;Taylor et al. 2019;Goeckner et al. 2022;Jeffries et al. 2022).Quantifying OC burial rates from ponds is therefore critical to understanding the role of these systems in the global C cycle.
Areal OC burial rates may relate inversely to waterbody size, with sequestration greater (Downing et al. 2008;Ferland et al. 2012) and more variable (Kastowski et al. 2011) in smaller waterbodies than large ones.Burial rates from three recent studies on artificial ponds averaged between 113.5 and 215.7 g OC m À2 yr À1 (Taylor et al. 2019;Goeckner et al. 2022;Rogers et al. 2022), which is 3-10 times greater than average burial rates in lakes (22-35.4g OC m À2 yr À1 ; Mendonça et al. 2017;Anderson et al. 2020).Although these recent studies demonstrate the potential for high C burial rates in ponds, they also highlight considerable variability.
Explaining variability in OC burial requires considering OC sourced internally (autochthonous) from primary production as well as OC sourced externally (allochthonous) from the watershed, although untangling the extent to which each mediates OC burial is challenging (Downing et al. 2008).For instance, lentic OC burial rates have been linked to erosion (Rogers et al. 2022), precipitation (Xu et al. 2013), and agricultural intensification (Heathcote and Downing 2012;Anderson et al. 2013Anderson et al. , 2014Anderson et al. , 2020)).These external drivers can increase allochthonous sediment deposition and nutrient loading, fueling autochthonous production and deposition of organic material.Increasing evidence points to the strong role of autochthonous production in mediating OC burial, including in agricultural impoundments (Downing et al. 2008) and at a global scale, where lake OC burial rates are linked with fertilizer use (Anderson et al. 2020).
In this study, we disentangled the effects of OC sourced from autochthonous and allochthonous production by examining OC burial dynamics in 22 experimental ponds with negligible watersheds.The experimental ponds were constructed in an identical manner but received different manipulations over time (e.g., nutrient additions, fish management), providing a unique opportunity to test how management strategies influence autochthonous OC burial rates.Here, we quantify OC burial rates in experimental ponds, evaluate how pond management history influences OC burial, and compare OC burial rates across aquatic waterbodies.

Study site
We sampled 22 experimental ponds at the Cornell Experimental Pond Facility Unit 2 in New York, USA (42.5026N, À76.4369 W).Ponds were constructed in 1964 in the shape of inverted truncated pyramids that are 30 Â 30 m (900 m 2 ) and originally 2.4 m deep, with flat bottoms consisting of a natural hardpan clay base.Over the 57 yr between pond construction and our summer 2021 sampling, the ponds were used discontinuously for various experimental manipulations (Hall et al. 1970;Mulligan et al. 1976;Hambright et al. 2007;Fig. 1).In May 2021 to June 2021, the study ponds had an average pH of 8.7 AE 0.6 (SD), conductivity of 157 AE 55 μS cm À1 and were macrophyte dominated in a clearwater state.

Sediment core collection and characterization
We collected two sediment cores from each pond using a push corer (5.08 cm diameter), which we sectioned into 5 cm increments, and froze until analysis of sediment organic matter content via loss on ignition (LOI; Dean 1974).Briefly, we homogenized thawed sediments and subsampled a known volume, which we weighed before and after drying to a constant mass (100 C), and weighed again after ashing (550 C for 2 h) to calculate dry bulk density (DBD; g dry mass cm À3 ) and LOI (% organic matter).We lost the surface (0-5 cm) sample for one core (pond 242, core 1) in the field; for this sample, we used the average LOI (12.83) and DBD (0.24) from all other surface samples.
To confirm the relationship between LOI and percent OC (%OC), we analyzed 68 (out of 178 total) samples for %OC.Samples were weighed, fumigated with hydrochloric acid for 24 h in silver capsules (Harris et al. 2001), redried, and packed into tin capsules prior to analysis on a NC2500 elemental analyzer (Carlo Erba, Italy) coupled to a Delta V IRMS at the Cornell Isotope Laboratory (COIL).We found a strong positive linear relationship between LOI and %OC (R 2 = 0.873, p < 0.001; Supporting Information Fig. S1), which we used to convert LOI to %OC for all sediment samples: Quantifying sediment accumulation and carbon burial We mapped sediment thickness using a sounding rod from a Jon boat.In each pond, we established three transects with four sampling points each, resulting in 12 approximately evenly spaced sample locations.At each location, we measured water depth with a graduated polyvinyl chloride (PVC) pipe and measured sediment depth using a graduated stainless-steel sounding rod with a pointed end (Downing et al. 2008), which we pushed into pond sediments until we met resistance from the hardpan base.The difference between the surface sediment depth and bottom sediment depth is the sediment thickness.
We digitized sediment thickness measurements and used QGIS (QGIS.org2021) to interpolate sediment thickness (m) as a raster (1 Â 1 m pixels) across each pond.The sum of all cells within the perimeter of each pond provided an estimated total sediment volume (m 3 ).We divided the total sediment volume by pond age (57 yr) to estimate annual sediment deposition rate per pond (m 3 yr À1 ).To calculate linear sediment accumulation rate (w, cm yr À1 ), we divided the annual sediment deposition by pond area.
We examined LOI and DBD across sediment depth, and excluded basal core sections with <2.95% organic matter as those sections likely represented underlying hardpan base (n = 17 samples).For each replicate core, we calculated mean DBD and LOI.We calculated sediment mass accumulation rates (g cm À2 yr À1 ) as w multiplied by DBD.We calculated the annual OC burial rate for each sediment core as follows: We averaged the values for the two cores to estimate OC burial rate for each pond.

Linking past pond management with rates of carbon burial
We took two approaches to link past pond management with OC burial rates.First, we used linear regressions with OC burial as a function of total years where different treatments were applied (e.g., total years of nutrient additions per pond; Fig. 1).We used this approach as the records lack a more detailed history; for example, we do not know the total nutrients added through time or the number of years fish were present in a pond (e.g., if die-offs or colonizations occurred outside of experiments).Plant additions consisted of submerged macrophytes, including Elodea canadensis, Ceratophyllum demersum, Myriophyllum sibiricum, and to a lesser extent Potamogeton spp.The most common fishes added to ponds were fathead minnows (Pimephales promelas) and sunfish (Lepomis gibbosus and Lepomis macrochirus).
Our second approach considered the impact of five specific experiments conducted over multiple years in multiple ponds.The studies included a nutrient and fish manipulation conducted from 1965 to 1967 (Hall et al. 1970), an experiment investigating fertilization effects on phytoplankton and zooplankton conducted from 1968 to 1969 (O'Brien and de Noyelles 1974), an unpublished fertilization experiment conducted from 1985 to 1986, an experiment manipulating nutrient stoichiometry and fish from 1989 to 1991 (Hambright et al. 2007), and trophic cascade manipulations from 1987 to 1989 (Hambright 1994).We compared OC burial rates among the treatment ponds using ANOVA when assumptions of normality and equal variance were met or using Kruskal-Wallis tests when assumptions were violated.As there were no treatments in the unpublished fertilizer experiment, we compared OC burial rates between ponds manipulated in the study and ponds not included in the study.
Lastly, we conducted the same analyses described above to test how pond management practices influenced %OC and sediment mass accumulation rates.These additional analyses provide mechanistic insight on how OC burial rates reflect differences in OC quantity, sediment quantity, or both.
All calculations and figures were made in R version 4.0.2(R Core Team 2020), and the data are publicly available (Holgerson et al. 2023).

Sediment accumulation and carbon burial
Mean sediment thickness was 24 cm (range 16-34 cm) across the 22 ponds (Fig. 3).We estimated that sedimentation rates averaged 0.30 cm yr À1 (range 0.20-0.42cm yr À1 ) per pond.Most sediment accumulation occurred where the littoral shelf met the pond bottom with less accumulation at the pond center (Fig. 3).Across the 22 ponds, OC burial rates averaged 67.1 g OC m À2 yr À1 (range 38.3-113.6 g OC m À2 yr À1 ; Supporting Information Fig. S3).

Effects of past pond management on carbon burial
We examined how the frequency of different management actions (Fig. 1) impacted OC burial rates across the 22 ponds (Supporting Information Table S1).OC burial increased with the frequency of plant additions to the ponds (linear model; F = 7.432, p = 0.013; Fig. 4A) and was inversely related to the number of piscicide applications (i.e., fish removal; linear model; F = 9.892, p = 0.005; Fig. 4B).In contrast, OC burial rates were not impacted by the number of times fish, nutrients, or herbicides (i.e., plant removal) were added to the ponds, nor was there an impact of one-time shade or insect additions (linear models; p > 0.10).Although not statistically significant (linear model; F = 3.636, p = 0.071), OC burial rates were slightly higher in ponds that had 10 cm of sediments added in 1967 (81.4 AE 8.7 SE g OC m À2 yr À1 , n = 6) compared to ponds without sediment additions (61.7 AE 5.4 SE g OC m À2 yr À1 , n = 16).
Of the five specific multiyear experiments we examined, two were linked to OC burial rates (Supporting Information Table S2).Hambright et al. (2007) manipulated nutrient stoichiometry and fish communities in 10 of our study ponds; we found a significant effect of nutrient stoichiometry (high or low N : P) on OC burial (ANOVA, F = 15.071,p = 0.005), and no effect of fish communities.Ponds in the high N : P treatment had higher OC burial rates (104.0AE 6.6 SE g OC m À2 yr À1 ) than ponds in the low N : P treatment (68.8 AE 6.0 g SE OC m À2 yr À1 ; Fig. 4D).We also found a significant effect on OC burial rates from an unpublished fertilization experiment, where the 10 ponds that were heavily fertilized with manure, NH 4 NO 3 , H 3 PO 4 , wheat sorts, and vegetable oil had lower OC burial rates (55.5 AE 4.1 SE g OC m À2 yr À1 ) than ponds not manipulated in the study (76.7 AE 7.3 g OC m À2 yr À1 ; Kruskal-Wallis, χ 2 = 4.452, p = 0.035; Fig. 4C).
The observed changes in OC burial rates may result from pond management practices causing differences in %OC and/or sediment mass accumulation rates (Supporting Information Tables S3-S6).We found that piscicide additions decreased both %OC and mass sediment accumulation, whereas plant additions and N : P loads increased mass accumulation rates but did not influence %OC.The unpublished fertilization study did not significantly impact either %OC or mass accumulation rates, suggesting that the change in OC burial was synergistic.

Discussion
The 22 artificial ponds studied here accumulated substantial amounts of OC (67.1 AE 4.9 SE g OC m À2 yr À1 ) in their sediments since construction, despite having almost no watershed.Autochthonous production in ponds can therefore drive a substantial OC sink, regardless of terrestrial inputs from the watershed.Past management practices can partially explain the large range (38.3-113.6 g OC m À2 yr À1 ) in OC burial across ponds.Specifically, the highest OC burial rates were associated with more plant additions, fewer piscicide additions, and greater N : P loads.Although we cannot tease apart potential interactions, below we discuss mechanisms in which each factor may individually affect OC burial.
First, OC burial rates increased with the number of submerged plant additions to the ponds, driven by increased sediment mass accumulation but not increased %OC.Submerged macrophytes trap sediments, which can explain why sediment accumulation occurred closer to the littoral shelves, rather than focusing to the pond center (Barko and James 1998).An established submerged macrophyte community also promotes a stable clearwater state in shallow waterbodies (Scheffer 2004), which contrasts the phytoplankton-dominated (turbid) state that the ponds were periodically managed for through herbicide additions (Hambright 1994;Hambright et al. 2007).Planting macrophytes may elevate OC burial through greater C fixation or through greater recalcitrance relative to phytoplankton (Enríquez et al. 1993).Only a few studies have examined the effects of primary producer communities on C fixation and OC burial rates, with equivocal results (Brothers et al. 2013;Zimmer et al. 2016;Hilt et al. 2017), indicating that more research is needed to clarify the mechanisms in which primary producer communities influence oxygen, mineralization, and OC burial in aquatic systems.
Second, OC burial rates decreased with the number of piscicide additions to the ponds, driven by both decreased %OC and sediment mass accumulation.Fish presence may alter OC burial by inducing a trophic cascade, where zooplanktivorous fish additions reduce zooplankton populations, and indirectly increase phytoplankton production (Hambright 1994;Hambright et al. 2007), likely enhancing OC loads and burial rates.It is also possible that fish increase nutrient recycling, promoting autochthonous production (Vanni 2002) and subsequent OC burial.It then follows that piscicide additions would reduce OC burial rates.Although piscicides may serve to increase OC burial through the deposition of fish carcasses, we did not see such effect, perhaps because this organic matter was labile and readily decomposed.We also found no reciprocal effect of fish additions, likely because the ponds tended to have fish unless piscicides were recently added.
Two studies that manipulated nutrient loads to the ponds were linked with OC burial.First, Hambright et al. (2007) added nutrients weekly at low (10:1 N : P in year 1, 18 : 1 in year 2) or high (98:1 in year 1, 49:1 in year 2) N : P ratios over 2 yr.OC burial was higher in high N : P ponds compared to low N : P ponds (Fig. 4D), which were both higher than burial rates of ponds not included in the study (53.8 AE 3.64 SE g OC m À2 yr À1 ).Increases in OC burial seem driven by increases in sediment mass accumulation, and not %OC.Although Hambright et al. (2007) did not see an effect of N : P on OC sedimentation during their study, the nutrient additions may have promoted primary production in the ponds for years afterward.For instance, many of the high N : P ponds were not manipulated for 13 yr following the study (Fig. 1), and the added nutrients may have fueled primary production and increased burial rates over time.
The link between N : P stoichiometry and OC burial was further supported by a second fertilization study where 10 experimental ponds received NH 4 NO 3 , H 3 PO 4 , and manure (Fig. 4C).Despite this large nutrient addition, OC burial rates were depressed relative to ponds not included in the study (most of which were the Hambright et al. 2007 ponds;Figs. 1, 4).Nutrient stoichiometry may help to explain the results: without including manure (which typically is low in N : P), the N : P ratio of these additions was $ 1.3.Although the mechanism in which N : P influences OC burial rate is not clear, we suspect it relates to macrophytes.Submerged macrophytes, even those with minor root systems, can uptake sediment P (Barko and Smart 1980), making them more likely to be limited by N than P (Barko et al. 1991).Therefore, high N : P loads may boost macrophyte production and subsequently increase OC burial rates.Nutrient limitation in shallow waterbodies may therefore differ from larger, deeper lakes where P is more likely to be limiting (Schindler 1977) and may be more important for regulating OC sequestration (Anderson et al. 2020).As water column nutrient stoichiometry is not often reported along with sediment OC sequestration, studies that link N : P ratios, primary producer types, and  S1 and ANOVAs in Supporting Information Table S2.Colors represent whether pond N : P ratios were manipulated in (D), with blue for the high N : P treatment, yellow for low N : P, and red indicating ponds not included in that study.
primary producer biomass to OC burial rates is an important research area.
Our experimental ponds stored high rates of OC (67.1 g OC m À2 yr À1 ) in their sediments strictly through autochthonous production, which are double global lake estimates (22-35.4g OC m À2 yr À1 ; Mendonça et al. 2017;Anderson et al. 2020) and similar to global wetland estimates (26-176 g OC m À2 yr À1 ; Villa and Bernal 2018).Artificial ponds with watersheds have even higher OC burial rates (113.5-215.7 g OC m À2 yr À1 ; Taylor et al. 2019;Goeckner et al. 2022;Rogers et al. 2022), emphasizing the importance of small waterbodies in OC burial (Downing et al. 2008).Indeed, artificial ponds sequester OC at rates similar to or higher than any other aquatic ecosystem, including coastal mangroves (Fig. 5).Considering that there may be $ 77,000 km 2 of farm ponds globally (Downing et al. 2006) and assuming each one sequesters 135 g OC m À2 yr À1 (the average of the four pond studies in Fig. 5), we estimate that farm ponds collectively sequester 10.4 Tg OC yr À1 , or about 10% of estimated OC burial in lakes (90-120 Tg OC yr À1 ; Mendonça et al. 2017;Anderson et al. 2020).Although data on OC burial in natural ponds is lacking, if we assume the same rates as artificial ponds (almost certainly an underestimate), and assume natural ponds < 0.01 km 2 comprise $ 500,000 km 2 globally (Downing et al. 2006;Verpoorter et al. 2014;Holgerson and Raymond 2016), our total estimate of OC burial in natural and artificial ponds increases to 77.9 Tg OC, or 65-87% of the current lake estimate.Therefore, the undervaluation of pond OC burial may greatly underestimate lentic OC storage on a global scale.
In summary, artificial ponds store large amounts of OC in their sediments, and management practices influence sequestration rates.As natural and artificial ponds may store 77.9 Tg OC yr À1 in sediments globally, ponds partially offset or could potentially surpass their high CH 4 emissions to act as C sinks across the landscape.

Fig. 1 .
Fig. 1.History of manipulations at the 22 sampled ponds through 2020 based on records from Cornell University Experimental Pond Facility.Carbon burial rates (g OC m À2 yr À1 ) calculated for each study pond are shown on the right-hand side.

Fig. 2 .
Fig. 2. Sediment characteristics across core depth for (A) sediment organic matter content (%) estimated by LOI and (B) DBD averaged across all study ponds.Error bars represent standard error.Pond-specific LOI through depth can be seen in Supporting Information Fig. S2.

Fig. 3 .
Fig. 3. Contour of interpolated sediment thickness across the 22 ponds sampled in this study.Darker blues indicate greater sediment thickness.

Fig. 4 .
Fig. 4. Statistically significant OC burial rates based on prior management actions at the experimental ponds, including (A) number of plant additions, (B) number of piscicide additions, (C) if ponds were included in a 1985-1986 nutrient addition study, and (D) if ponds were included in a 1989-1991 study that manipulated N : P ratios.In (A) and (B), the shaded area around the regression indicates the 95% confidence interval and in (C) and (D), boxes demarcate the 25 th and 75 th percentiles, horizontal lines indicate median values, and whiskers extend to the largest value less than 1.5 times the interquartile range.Results of linear models in Supporting Information TableS1and ANOVAs in Supporting Information TableS2.Colors represent whether pond N : P ratios were manipulated in (D), with blue for the high N : P treatment, yellow for low N : P, and red indicating ponds not included in that study.