Fencing farm dams to exclude livestock halves methane emissions and improves water quality.

Agricultural practices have created tens of millions of small artificial water bodies ("farm dams" or "agricultural ponds") to provide water for domestic livestock worldwide. Among freshwater ecosystems, farm dams have some of the highest greenhouse gas (GHG) emissions per m2 due to fertilizer and manure run-off boosting methane production-an extremely potent GHG. However, management strategies to mitigate the substantial emissions from millions of farm dams remain unexplored. We tested the hypothesis that installing fences to exclude livestock could reduce nutrients, improve water quality, and lower aquatic GHG emissions. We established a large-scale experiment spanning 400 km across south-eastern Australia where we compared unfenced (N = 33) and fenced farm dams (N = 31) within 17 livestock farms. Fenced farm dams recorded 32% less dissolved nitrogen, 39% less dissolved phosphorus, 22% more dissolved oxygen, and produced 56% less diffusive methane emissions than unfenced dams. We found no effect of farm dam management on diffusive carbon dioxide emissions and on the organic carbon in the soil. Dissolved oxygen was the most important variable explaining changes in carbon fluxes across dams, whereby doubling dissolved oxygen from 5 to 10 mg L-1 led to a 74% decrease in methane fluxes, a 124% decrease in carbon dioxide fluxes, and a 96% decrease in CO2 -eq (CH4  + CO2 ) fluxes. Dams with very high dissolved oxygen (>10 mg L-1 ) showed a switch from positive to negative CO2 -eq. (CO2  + CH4 ) fluxes (i.e., negative radiative balance), indicating a positive contribution to reduce atmospheric warming. Our results demonstrate that simple management actions can dramatically improve water quality and decrease methane emissions while contributing to more productive and sustainable farming.

INTRODUCTION

Global methane emissions are rising rapidly, nearly tripling from ca. 700 ppb in pre‐industrial times to 1900 ppb today (Conrad, 2009; Dlugokencky, 2022). The accumulation of artificial water bodies has contributed to the growth in atmospheric methane, with aquatic ecosystems now accounting for half of natural and anthropogenic methane emissions (Rosentreter et al., 2021). With farm dams estimated to cover a surface area >75,000 km2 globally (Downing et al., 2006), these artificial systems are now a key part of aquatic ecosystems globally (Malerba et al., 2021; Swartz & Miller, 2021). Therefore, it is likely that farm dams are an important contributor to global carbon cycles—even though this link is often overlooked in national and global carbon inventories. Indeed, the Intergovernmental Panel on Climate Change (IPCC) recently revised their guidelines to promote the inclusion of agricultural ponds in national GHG inventories and tackle this form of anthropogenic carbon emission (IPCC, 2019).

Farm dams (or agricultural ponds) are small, human‐made freshwater bodies created for the purpose of storing water for livestock or crop irrigation (Malerba et al., 2022). These systems have some of the highest greenhouse gas (GHG) emissions per m2 among freshwater ecosystems (Grinham et al., 2018; Ollivier et al., 2018, 2019) due to their much higher nitrogen and phosphorus concentrations than natural ponds (Westgate et al., 2022), creating the perfect conditions for methanogenesis and GHG emissions (Li et al., 2021; Panneer Selvam et al., 2014; Peacock et al., 2021). Importantly, eutrophication appears to have a disproportionate effect on farm dams. That is, a 25% increase in nitrate concentration was observed to double the CO2‐equivalent carbon flux per m2 of a farm dam (Ollivier et al., 2018). Hence, understanding how to reduce the emissions of millions of farm dams worldwide has the potential to make a substantial difference in mitigating climate change. Yet, there is no evidence of the effects of management practices on reducing these emissions.

Using fences to exclude livestock from farm dams improves water quality by reducing direct depositions of nutrient‐rich manure and urine into the water (Westgate et al., 2022). In addition, fencing a farm dam avoids hooved livestock (ungulates) disturbing soils and promotes higher vegetation cover around the dam, acting as a filter to reduce dissolved nutrients (i.e., “phytoremediation”; reviewed in Pilon‐Smits, 2005). A recent study showed that partially or fully fenced farm dams have higher vegetation cover, higher water quality (i.e., lower nutrients, turbidity, and fecal coliforms), and higher macroinvertebrate richness and abundance than unfenced farm dams (Westgate et al., 2022). Moreover, fencing farm dams is often cost‐effective, with the benefits for livestock health and weight gain from higher water quality often exceeding the costs of this management intervention (Dobes et al., 2021).

In summary: (1) nutrient pollution drives high GHG emissions from farm dams; (2) excluding livestock from accessing farm dams favors vegetation growth and improves water quality; and (3) higher water quality provides benefits to livestock health, biodiversity, and aesthetic value. Based on these premises, installing fences could reduce aquatic GHG emissions from farm dams while improving agricultural productivity and biodiversity. Previous studies have already shown that excluding livestock and reducing grazing intensity can reduce methane emissions and enhance carbon sequestration and storage of freshwater wetlands (Limpert et al., 2021; Oates et al., 2008; Watkins et al., 2017). Yet, the effects of installing fences (or any other management intervention) on farm dam GHG production remain untested. Similarly, there is little evidence of the benefits of farm dam fencing on water quality (Westgate et al., 2022). Hence, we here addressed two key questions:

What are the effects of fencing farm dams on water quality (i.e., total dissolved nitrogen, total dissolved phosphorus, and dissolved oxygen), soil organic carbon, and GHG fluxes of methane and carbon dioxide?

What are the mechanisms linking farm dam management to aquatic GHG fluxes?

To answer these questions, we completed a cross‐sectional field‐based study comparing the effects of fencing farm dams on their water quality and carbon footprint. We surveyed 64 farm dams across 17 farming properties. At each property, we compared farm dams under two management regimes: “unfenced dams” where livestock have free access to the water, and “fenced dams” where water access has been controlled for at least 2 years using either full fencing or partial fencing (with a hardened livestock access point). We predicted that fencing a farm dam would reduce dissolved nutrients, increase dissolved oxygen, and lower GHG emissions. Testing these processes contributes to identifying novel GHG abatement methods to reduce the carbon footprint of farming practices.

METHODS

Study area and experimental design

In April 2021, we sampled farm dams across 400 km of the Australian South West Slopes bioregion in south‐eastern New South Wales. The study region has a warm temperate climate, with hot dry summers and cool humid winters (the largest city of Albury has an annual mean temperature of 22°C and annual rainfall of 691 mm). Most of the area is dedicated to livestock grazing (especially beef cattle and sheep) and dryland cropping (mainly cereals and oilseed). We surveyed 64 farm dams located in pastures on 17 farming properties. Within each property, we established two experimental treatments: “unfenced” farm dams and “fenced” farm dams. For each experimental treatment within a farming property, we measured between 1 and 5 dams (depending on availability) on the same day. Unfenced farm dams (N = 33) received no management intervention to improve their ecological condition. Fenced farm dams (N = 31) were either entirely fenced (with a pump delivering water into drinking troughs) or partly fenced (providing water access through a hardened access point) for at least 2 years prior to sampling. We avoided small (<200 m2) farm dams because they were too ephemeral. We measured farm dam areas by tracing the most recent satellite images on Google Earth Pro (version 7.3.4).

Aquatic greenhouse gas emissions

We measured diffusive emissions of methane and carbon dioxide at each farm dam using the methods described in Ollivier et al. (2018, 2019). Briefly, a white plastic floating chamber (0.021 m3 volume and 0.14 m2 surface area) was sealed and connected to an Ultraportable Greenhouse Gas Analyzer (UGGA, Los Gatos Research, Model 915–0011) through two tubes (influx and outflux) on the chamber roof to create a closed circuit. We sampled methane and carbon dioxide (ppm) at 1‐second intervals for ca. 5 min (ca. 300 data points per sample). We measured each farm dam three times from different locations along the shore ensuring the starting concentration matched atmospheric levels.

Floating chambers can measure constant fluxes (diffusion) and stochastic releases of gas bubbles (ebullition). Here we focused only on diffusive fluxes. To do so, we excluded any trajectories showing sudden increases in gas concentration due to a gas bubble being released inside the floating chamber. We estimated the linear rate of change of diffusive gas flux from the water surface to the atmosphere (; mg m−2 d−1) as: where is the linear rate of change in gas concentrations over time within the chamber (ppm s−1), is the chamber volume (0.021 m3), is the conversion factor from ppm to μg m−3 for methane (655.47), is the conversion factor from minutes to day (86,400), is the conversion factor from μg to mg (1000), is the surface area of the chamber (0.14 m2; Lambert & Fréchette, 2005). We retained all diffusive rates without applying any filtering method (e.g., R2 threshold).

Sediment carbon stocks

At each dam, we collected two cores (45 mm diameter, 50 mm deep, 79.52 cm3 volume) from the edge of the pond within the water (wet sediments). We preserved the cores in a freezer until returning to the laboratory. We dried all cores at 60°C until there was no more weight loss (approx. a week) and measured their dry weights. Finally, we ground the cores and determined the organic carbon content by analyzing 10 mg of each sample using a EuroVector MicroElemental CN Analyser (see Gulliver et al., 2020 for details). We quantified each sample's C:N ratio using Acetanilide as standards (71.09% C, 0.5–1 mg input mass; R2 > 0.98). The carbon density of each core was the product of dry biomass density (g cm−3) and carbon content (i.e., % C/100) in units of tons of carbon per hectare (t C ha−1).

Water quality and nutrient analysis

At each site, we measured dissolved oxygen (mg O2 L−1), conductivity (μS cm−1), and water temperature (°C) using a Hach HQ30D portable Multi Meter. We also filtered 50 ml of water from each farm dam using syringe filters with Filtech 483 Glass fiber filter paper (1.10 μm retention, 25 mm diameter). We froze all filtrated water samples immediately after collection and sent them to ALS Environmental (alsglobal.com, Everton Park QLD 4053 Australia) to analyze total nitrogen following APHA 4500‐Norg / 4500‐NO3 − (method EK062G; mg N L−1) and total phosphorus following APHA 4500‐P (method EK067G; mg P L−1). All analyses followed standard protocols and included quality controls. Finally, we took three pH measurements at each dam using the YSI ProDSS Multiparameter Digital Water Quality Meter (Xylem Analytics, Yellow Springs, OH 45387 USA), taking measurements at 1.5 m from the water's edge and at 20 cm depth. We rinsed the sensors with demineralized water between samples and sites and always calibrated probes before use.

Statistical analyses

First, we used individual linear mixed‐effects models to evaluate whether the management regime (categorical variable, either “fenced” or “unfenced”) affected total dissolved nitrogen (log10 mg N L−1), total dissolved phosphorus (log10 mg P L−1), dissolved oxygen (mg L−1), organic carbon stock (log10 t C ha−1), and rates of methane emissions (log10 mg m−2 d−1 + 2), carbon dioxide emissions (log10 g m−2 d−1 + 1.8), and CO2‐equivalent emissions (carbon dioxide + methane; log10 g m−2 d−1 + 1.8). We added two units to methane emissions and 1.8 units to carbon dioxide and CO2‐eq. emissions to avoid negative values when applying the log10 transformation. We calculated CO2‐equivalent units by combining methane and carbon dioxide fluxes using the 20‐year Sustained‐Flux Global Warming Potential (SGWP) metric from Neubauer and Megonigal (2015), where 1 Kg of CH4 traps as much infrared radiation as 96 Kg of CO2. The SGWP calculates the decay rate assuming a sustained gas flux rate over time, and this approach is more realistic for farm dams than the one‐time pulse assumed in the Global Warming Potential metric. We did not correct the p‐values for multiple statistical testing, yet we ensured that reducing the risk of type I error by adopting more conservative thresholds for statistical significance using the false discovery rate (Benjamini & Hochberg, 1995) did not change any of our conclusions.

Second, we used three linear mixed‐effects models to quantify the statistical association of each environmental variable with fluxes of carbon dioxide, methane, and CO2‐equivalent (carbon dioxide + methane) of a farm dam. In the models, the independent variables were farm dam surface area (log10 m2), dissolved oxygen (log10 mg L−1), pH, conductivity (log10 μS cm−1), water temperature (°C), total dissolved nitrogen (log10 mg N L−1), total dissolved phosphorus (log10 mg P L−1), and organic carbon stock (log10 t C ha−1). The initial fully parameterized model included all main effects and a two‐way interaction term to account for the potential interplay between total nitrogen and total phosphorus. To avoid bias from multicollinearity between main effects, we ensured a cut‐off value of five for the maximum variance inflation factor (VIF) in the model, as recommended by Zuur et al. (2009). As a result, pH and dissolved oxygen could not be included together in the models because they are highly correlated (r = 0.72 and VIF >5). Therefore, we used only dissolved oxygen in the mixed‐effects models as this variable is associated with fluxes of both carbon dioxide and methane (whereas pH is only associated with carbon dioxide). Finally, we quantified the importance of each statistically significant explanatory variable by calculating its contribution to the total model prediction power using a permutation approach (Fisher et al., 2019; Niittynen & Luoto, 2018; Virkkala et al., 2021). This analysis consisted of three steps. First, we extracted the predictions from the best‐fitting model (Predictions ). Second, we created simulated datasets using random permutations of each statistically significant explanatory variable to remove its explanatory power. Third, we re‐fitted the model to each simulated dataset, computed model predictions, and quantified the Pearson correlation coefficient between the predictions of the original model (Predictions ) and the predictions with the explanatory variable being permutated (Predictions ), as:

Values close to −1 or 1 indicate greater importance of the shuffled variable for the model's explanatory power. We repeated this process 100 times for each variable to calculate the average importance and 95% confidence intervals.

We centered and scaled all variables before fitting the linear mixed‐effects models. We also added a random intercept to account for the experimental block design where each of the 17 farming properties contained one or more fenced and unfenced dams. To analyze repeated flux measurements from the same pond, we added a nested random intercept of site within farming property. When standardized residuals showed unequal variances or a systematic trend, we included treatment‐specific variance coefficients (function varIdent) or other variance functions (functions varExp or varPower) in the model. We identified the best‐fitting model using Akaike information criteria corrected for small sample sizes (AICc; Burnham & Anderson, 2004). We used standard diagnostics to ensure normality, homoscedasticity, and the absence of influential points or outliers.

We used the statistical software R version 4.0.3 (R Core Team, 2020) with the packages nlme (Pinheiro et al., 2020) and effects (Fox & Weisberg, 2018, 2019) for the statistical analyses, and dplyr (Wickham et al., 2018), plyr (Wickham, 2011), and ggplot2 (Wickham, 2009) for data manipulation and plotting.

RESULTS

Effects of fencing farm dams on greenhouse gas emissions and organic carbon stocks

On average, methane emissions from fenced farm dams (3.5 mg m−2 d−1) were 56% lower than unfenced farm dams (8.05 mg m−2 d−1; Figure 1a). Conversely, we found no significant difference for carbon dioxide fluxes (p = .2; Figure 1b) or for CO2‐eq fluxes (p = .08; Figure 1c). Finally, there was no effect of fencing on the organic carbon stock in the sediments of the farm dams (p = .42; Figure 1d). See Table 1 for summary statistics and Table S1 for statistical scores.

FIGURE 1

Effects of farm dam fencing on (a) methane fluxes, (b) carbon dioxide fluxes, (c) CO2‐eq (methane + carbon dioxide) fluxes, and (d) organic carbon in the soil. Black point ranges represent the means ±95% confidence intervals from the best‐fitting linear models. Grey points are the raw data. All statistics are calculated on a sample size of 64 farm dams across 17 farming properties. We reported percentage changes only on statistically significant effects (see Table S1 for test statistics).

TABLE 1

Summary of farm dam properties in this study. Water volume was estimated using the model in Figure 1 of Malerba et al. (2021): Water Volume = −3.593 + 1.237 × Water Area. Water depth was estimated using the formula (Water Volume × 1000)/(Water Area × 0.4) (Agriculture Victoria, 2022)

Variable	Unit	Rep	Min	Mean	Median	Max
Area	m2	64	227	1978	886	20,796
Water volume (est.)	ML	64	0.21	3.05	1.13	56
Water depth (est.)	m	64	2.31	3.86	3.19	6.73
Longitude		64	146.79	147.63	147.19	149.45
Latitude		64	−36.10	−35.32	−35.86	−33.51
Total Nitrogen	mg N L−1	64	0.40	2.68	2.10	9.20
Total Phosphorus	mg P L−1	64	0.01	0.12	0.06	0.80
Dissolved Oxygen	mg O2 L−1	64	3.16	6.94	6.52	17.60
Water Temperature	°C	64	12.83	16.20	15.57	22.13
pH		63	6.50	7.79	7.71	9.52
Conductivity	μS cm−1	64	11.89	294.3	227.5	1647
CH4 diffusion	g m−2 d−1	63	0.0001	0.0151	0.0034	0.1639
CO2 diffusion	g m−2 d−1	64	−1.6995	1.4897	0.7887	13.9746
CO2‐eq diffusion	g m−2 d−1	63	−1.2161	2.9364	1.5619	21.3560
Sediment organic C stock	t C ha−	63	0.56	6.26	4.59	28.84

Effects of fencing farm dams on water quality

Fenced farm dams recorded higher water quality than unfenced ones across all parameters measured here. Specifically, water from fenced farm dams had on average 32% less total dissolved nitrogen (from 2.4 to 1.6 mg L−1; Figure 2a), 39% less total dissolved phosphorus (from 0.078 to 0.047 mg L−1; Figure 2b), and 22% more dissolved oxygen than unfenced dams (from 6.32 to 7.74 mg L−1; Figure 2c). We found no difference in the water temperature (Figure 2d) and water pH (data not shown) of fenced and unfenced farm dams (see Table 1 for summary statistics and Table S1 for statistical scores).

FIGURE 2

Effects of farm dam fencing on (a) total nitrogen, (b) total phosphorus, (c) dissolved oxygen, and (d) temperature. All statistics are calculated on a sample size of 64 farm dams across 17 farming properties. We reported percentage changes only on statistically significant effects (see Table S1 for test statistics).

Drivers of greenhouse gas fluxes

Overall, most relationships between greenhouse gas fluxes and environmental variables show a high degree of variability. Yet, the methane flux of a farm dam was statistically associated with dissolved oxygen (Figure 3a), sediment organic carbon stocks (Figure 3b), total dissolved nitrogen (Figure 3c), and total dissolved phosphorus (Figure 3d). In contrast, the carbon dioxide flux of a farm dam only showed a negative association with dissolved oxygen (Figure 3f). The total carbon flux of a farm dam, calculated as CO2‐eq (methane + carbon dioxide) fluxes, showed statistically significant associations with dissolved oxygen (Figure 3k), sediment organic carbon stocks (Figure 3l), and total dissolved nitrogen (Figure 3m). Conversely, farm dam area, conductivity, and a two‐way interaction between dissolved nitrogen and dissolved phosphorus were systematically excluded from the best‐fitting models following Akaike information criteria.

FIGURE 3

Effects of five environmental parameters (columns) on three carbon fluxes (rows) from farm dams. Points in each panel represent the partial residuals of farm dams after controlling for the effects of the other fixed and random variables in the models. Lines indicate statistically significant effects following the best‐fitting mixed effect models (±95% confidence intervals). Labelled in each panel are the p‐values and the importance scores of statistically significant terms (see Table S2 for all test statistics and Table S3 for importance scores).

Dissolved oxygen was the most important variable for explaining all three greenhouse gas fluxes (see Table S3 for importance scores). Specifically, doubling dissolved oxygen from 5 to 10 mg L−1 corresponded to a 74% decrease in methane fluxes (from 6.92 to 1.8 mg CH4 m−2 d−1; Figure 3a), a 124% decrease in carbon dioxide fluxes (from 2.27 to −0.56 g CO2 m−2 d−1; Figure 3f), and a 96% decrease in CO2‐eq fluxes (from 3.77 to 0.13 g CO2‐eq m−2 d−1; Figure 3k). Farm dams with dissolved oxygen levels higher than ca. 10 mg L−1 showed a switch from positive to negative CO2‐eq fluxes (i.e., negative radiative balance; Figure 4).

FIGURE 4

Dashed lines indicate where fluxes equal zero. Also indicated in the figure is the dissolved oxygen concentration (10 mg L−1) associated with a switch from positive to negative CO2‐eq. fluxes. Variables were linearized using log10‐transformations when fitting the model, but they are presented here on an arithmetic scale (see Figure 3k for model fits presented in log scales).

Changes in both dissolved oxygen and carbon dioxide were pH related (Figure 5). Dissolved oxygen was positively correlated with the pH (r = 0.72; Figure 5a) and negatively correlated with the carbon dioxide flux (r = − 0.82; Figure 5b), while carbon dioxide flux was negatively correlated with pH (r = −0.76; Figure 5c). Conversely, we found no significant correlation between pH and methane fluxes (p = .39; data not shown).

FIGURE 5

Statistical associations between (a) the dissolved oxygen and the pH of the water, (b) the dissolved oxygen and the CO2 fluxes, and (c) CO2 fluxes and the pH of the water. Each point represents the average value recorded from a farm dam. Reported in each panel are the correlation coefficient and the p‐value of statistically significant relationships.

DISCUSSION

Farm dams are common in many rural landscapes worldwide and make important contributions to carbon cycles and greenhouse gas (GHG) emissions (Grinham et al., 2018; Ollivier et al., 2018; Peacock et al., 2021). We discovered that simple management practices, such as fencing off livestock from farm dams, increased water quality and dramatically lowered methane emissions. Fenced farm dams were characterized by 32% less dissolved nitrogen, 39% less dissolved phosphorus, 22% more dissolved oxygen, and 56% lower methane emissions than unfenced dams. Dissolved oxygen was the most important variable explaining changes in carbon fluxes across dams, whereby doubling dissolved oxygen from 5 to 10 mg L−1 led to a 74% decrease in methane fluxes, a 124% decrease in carbon dioxide fluxes, and a 96% decrease in CO2‐eq (CH4 + CO2) fluxes. Moreover, farm dams with very high oxygen levels (>10 mg L−1) exhibited a switch from positive to negative CO2‐eq fluxes. Finally, we found a strong negative correlation between the pH of the water and both the dissolved oxygen and fluxes of carbon dioxide.

We found that fencing farm dams, on average, more than halves diffusive methane emissions to 3.56 mg CH4 m−2 d−1 compared to 8.16 mg CH4 m−2 d−1 of unfenced farm dams. Our fieldwork took place in the bioregion of South Western Slopes in south‐eastern Australia, an important agricultural hotspot covering 86,811 km2. This region contains an estimated 172 thousands farm dams with a cumulative surface area of 278 km2 (Malerba et al., 2021), which is equivalent to the surface area of all lakes in the region (277 km2; Crossman & Li, 2015). Assuming our data are representative of average yearly fluxes, we estimated that fencing farm dams in this region would avoid emissions of 468 tonnes CH4 year−1, which corresponds to 44,917 tonnes CO2‐eq year−1 using the 20‐year Sustained‐Flux Global Warming Potential (SGWP) metric. These are only ballpark estimates, and more data are needed to better estimate the opportunity for avoided emissions using farm dam restoration. Considering that most farm dams have broadly similar properties and serve the same purposes (i.e., collect water for agricultural uses), our results and qualitative mechanisms may also apply to other regions of the world—albeit with different magnitudes. Thus, an important next step is to use a cost‐benefit analysis to determine if improving farm dam conditions could be a cost‐effective way to help decarbonize agricultural practices at scale (Figure 6).

FIGURE 6

Effects of fencing farm dams on water quality and methane emissions. Installing fences to exclude livestock from farm dams reduces the direct deposition of nutrient‐rich manure and urine into the water, avoids hooved livestock (ungulates) disturbing soil, and promotes higher vegetation cover around the dam. As a result, fenced farm dams have lower dissolved nutrients, higher dissolved oxygen, and lower methane emissions than unfenced dams. Percentages associated with fencing farm dams represent the relative change compared to unfenced dams (Figures 1 and 2).

The range of diffusive carbon fluxes measured here (1 to 164 CH4 mg m−2 d−1 and ‐1.7 to 14 CO2 g m−2 d−1) is comparable to previously published values for farm dams in Australia, Canada, India, and Sweden (Figure 7; Table S4). Yet, our study (and most others) measured diffusive methane fluxes without accounting for other pathways of methane emissions (e.g., ebullition events; Bastviken et al., 2008; Bastviken et al., 2011). Grinham et al. (2018) quantified both ebullitive and diffusive methane fluxes from Australian irrigation and stock dams and reported higher values than ours (up to 3.6 CH4 g m−2 d−1; Figure 7). It is possible that the benefits of fencing farm dams on carbon emissions are even higher than our estimates after accounting for multiple types of methane fluxes. However, research is needed to establish if fencing farm dams can influence ebullitive methane fluxes.

FIGURE 7

Emissions of (a) CH4 and (b) CO2 from agricultural ponds from the scientific literature divided by land‐use type. Each symbol indicates the arithmetic mean and the range of values. Reported in each panel are the overall average emissions and the average ranges. All studies measured methane diffusion, except for Grinham et al. (2018) who measured both methane diffusion and ebullition.

The two main findings of this study were: (1) that excluding livestock from farm dams improves water quality, and (2) that higher water quality corresponds to lower methane emissions (Figure 6). For the first finding, fenced farm dams recorded 32% less dissolved nitrogen, 39% less phosphorus, and 22% more dissolved oxygen than unfenced farm dams. Westgate et al. (2022) is the only other study on this topic and showed comparable results to ours, with a 45–50% reduction in total nitrogen and phosphorus in fenced farm dams over unfenced farm dams, together with reduced turbidity and lower fecal contamination. The similar results between two field studies from different years (2019 and 2021) and seasons (summer and autumn) suggest that the positive effects of fencing on water quality may be maintained throughout the year.

For the second finding, the higher water quality of fenced farm dams corresponded to 56% lower methane emissions (Figure 6). We found that total dissolved oxygen was a key driver explaining the reduced methane emissions. The strong negative effect of dissolved oxygen is consistent with our understanding of methanogenesis as a microbiological process requiring anaerobic conditions (Segers, 1998). Similarly, the positive effects of total dissolved nitrogen and sediment organic carbon stocks meet the expectation that freshwater environments rich with nutrients and labile organic materials emit more GHG (Beaulieu et al., 2019; Li et al., 2021; Peacock et al., 2021). Instead, a surprising result was the negative effect of total phosphorus on methane fluxes, particularly since phosphorus is thought to promote methane production rates (Peacock et al., 2019; Peacock et al., 2021). Phosphorus concentration only had a weak negative effect on methane fluxes but not on carbon dioxide or CO2‐eq fluxes. As shown by Nijman et al. (2022), one explanation could be that a greater phosphorus availability increases the growth and activity of methane‐oxidizing bacteria, resulting in a reduction of methane emissions through the oxidation of methane to hydrogen and carbon monoxide. Yet more studies are needed to clarify the effects of phosphorus on methanogenesis in farm dams.

We found that farm dams with very high concentrations of dissolved oxygen exhibited negative CO2‐eq GHG fluxes (i.e., negative radiative balance), indicating a positive contribution to reduce atmospheric warming. Most farm dams contribute to climate change by emitting substantial amounts of atmospheric GHG (Holgerson & Raymond, 2016; Ollivier et al., 2018; Peacock et al., 2021). Yet, under certain circumstances, small freshwater systems can remove GHG from the atmosphere and act as a carbon sink (Ollivier et al., 2018; Peacock et al., 2021; Webb, Hayes, et al., 2019; Webb, Leavitt, et al., 2019). While we found negative fluxes in only a minority of cases (13 farm dams out of 64), the effect of oxygen on CO2‐eq fluxes was very predictable: every farm dam recording oxygen levels >10 mg L−1 also showed a carbon drawdown (at up to 1.2 g CO2‐eq m−2 d−1). These negative fluxes are due to aquatic photosynthesis (i.e., net ecosystem production) sequestering carbon dioxide from the atmosphere at higher rates than CO2‐eq methane emissions. This finding further emphasizes the importance of farm dam management, even suggesting that increasing oxygen levels could turn farm dams into carbon sinks. Nonetheless, these results are likely to change during the night phase when plant respiration replaces photosynthesis, highlighting the importance of long‐term studies on carbon dynamics in farm dams.

There is still considerable uncertainty on the net radiative balance of farm dams, as there is little data on the rates of carbon sequestration and storage in dam sediments. Yet, farm dams appear to have the highest burial rates of organic carbon among freshwater systems, ranging from 148 to 17,000 g C m−2 year−1 (Downing et al., 2008; Rogers et al., 2022). Therefore, it is possible that farm dams can sequester more carbon in the sediments than what they emit to the atmosphere. Future studies should investigate if fencing farm dams can increase carbon sequestration together with decreasing methane emissions.

Dissolved oxygen was strongly positively correlated with pH and strongly negatively correlated with carbon dioxide, which is evidence that aquatic primary production is the key process regulating dissolved oxygen in the farm dams of this study. Specifically, photosynthetic activity produces oxygen and consumes carbon dioxide, which results in higher pH from faster dissociation of HCO3 − into CO2 and OH− (Zang et al., 2010). Had there been no correlation between pH and dissolved oxygen (as is often the case with aquaculture systems), other factors unrelated to photosynthesis (e.g., decomposition of organic matter) may have been more likely to drive changes in dissolved oxygen (Zang et al., 2010). Importantly, the pH increase from aquatic photosynthesis is likely to further reduce the carbon emissions of a farm dam by moving the carbonate equilibria toward carbonic acid and away from gaseous CO2. Specifically, as the system becomes more basic, the carbonate system changes from CO2 dominated to CO3 − dominated, with negligible carbon dioxide left at pH >8.5 (Andersen, 2018; Drever, 1997).

CONCLUSIONS

We discovered that fencing to exclude livestock from farm dams improves water quality (i.e., fewer dissolved nutrients and higher dissolved oxygen) and reduces diffusive methane emissions. Our data also revealed a threshold in dissolved oxygen at 10 mg L−1 above which farm dams switch from positive to negative CO2‐eq fluxes, helping mitigate climate change. Considering avoided carbon emissions and additional economic and ecological co‐benefits (i.e., higher biodiversity, increased livestock health, and capital value; Dobes et al., 2021; Hazell et al., 2001; Lewis‐Phillips et al., 2019; Westgate et al., 2022), investing in better farm dam management appears to be a promising strategy for improving farming productivity and environmental sustainability. Nevertheless, carbon cycles in farm dams remain one of the least explored among freshwater systems. Promising avenues for follow‐up studies include environmental work to analyze long‐term cycles for several carbon pathways (e.g., methane ebullition, plant‐mediated methane emissions, rate of carbon sedimentation), economic assessments to determine the best allocation of incentives for sustainable management interventions, and social studies to establish non‐market benefits and farmers' willingness to adopt management interventions. This information will help deliver policy recommendations on the cost‐effectiveness of investing in farm dam management as a novel carbon abatement strategy, as well as for additional co‐benefits.

AUTHOR CONTRIBUTIONS

M.E.M, D.B.L, B.C.S, and P.I.M. designed the research, M.E.M., P.W., and I.N.Y. collected the data, M.E.M. and L.S. analyzed the data, M.E.M. wrote the first draft, and all authors contributed to the final draft.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Tables S1‐S4

Click here for additional data file.