Nitrate Availability Modulates the Temperature Sensitivity of N2O and N2 Production From Denitrification

TL;DR

This study shows that nitrate levels strongly influence how warming affects the production of nitrous oxide and dinitrogen during denitrification in sediments.

Contribution

The study demonstrates that nitrate availability can override temperature effects on N2O and N2 production ratios in denitrification.

Findings

Under nitrate-replete conditions, warming increases N2 production while decreasing net N2O production.

When nitrate is limited, temperature changes do not significantly affect N2O or N2 production.

Nitrate availability can override temperature controls on the N2O:N2 production ratio.

Abstract

Nitrous oxide (N2O) can be both produced and reduced to dinitrogen (N2) during microbial denitrification, with the balance between these steps controlling the net flux of this potent climate gas. Here, we first used a meta‐analysis of published studies to predict how warming may regulate N2O and N2 production in soils and sediments. However, as most of these former studies used nitrate at far higher than ambient concentrations, the applicability of these predictions to ambient conditions may be limited. In addition, few studies separated denitrification from other microbial pathways contributing to N2O and N2 production. To address these limitations, we used 15N‐isotope labelling experiments in freshwater sediments to test how temperature sensitivity varies with limited (10 μM) and replete (100 μM) nitrate. Temperature affected N2O and N2 production only when nitrate was replete, where N2 production increased but net N2O production declined with warming, leading to a lower N2O:N2 production ratio at higher temperatures. These results show that substrate availability can outweigh temperature in controlling the balance between N2O and N2 production, indicating that temperature‐based predictions of N2O emissions alone may overestimate the effects of climate warming.

N2O is a potent greenhouse gas with denitrification in sediments and soils representing a major source, yet its response to warming under varying nitrate conditions remains poorly constrained. Combining meta‐analysis with isotope labelling experiments, we show that under nitrate‐replete conditions, warming increases production of climate‐benign N2 while decreasing net N2O production, lowering the N2O:N2 ratio. In contrast, when nitrate is limited, neither N2O nor N2 changes with warming. These findings show that nitrate availability can override temperature controls on N2O versus N2 partitioning, suggesting that ignoring nitrate availability may misrepresent future N2O emissions under global warming.

Figures

Meta‐analysis of published rates of net N2O and N2 production from denitrification (inferred and distinguished, Table 1) as a function of temperature in aquatic sediments and terrestrial soils. The rate of net N2O (a) and N2 (b) production increased at higher temperatures in both aquatic sediments and terrestrial soils. Data were visualised using the “Visreg” package in R (Breheny and Burchett 2017), plotted as partial residuals (brown and green circles) from the best fitting mixed‐effects models after accounting for the random effects, with the overall temperature sensitivity estimates shown as black lines (Table S3). For net N2O production, n = 100 (aquatic) and 276 (terrestrial) measurements; for N2 productions, n = 330 (aquatic) and 208 (terrestrial) measurements. As normalisation methods differed among studies, all production rates were standardised by centring each study's N2O or N2 production rates on its specific intercept.

Ratio of net N2O to N2 production inferred from denitrification as a function of temperature, based on incubations from aquatic sediments and terrestrial soils. n = 95 and 196 measurements for aquatic and terrestrial datasets, respectively. The black line gives the apparent activation energy derived from the slope of the best fitting linear mixed‐effects model using actual parallel measurements of net N2O and N2 production (Table S3, Model M0 for N2O:N2). The grey line shows the apparent activation energy predicted from the independent net N2O and N2 datasets shown in Figure 1. The two estimates did not differ significantly (χ 2 = 1.23, p = 0.27; linear hypothesis test for fixed effects, ‘car’ package in R).

Rates of 15N2O and 15N2 production and their ratios, from independent sediment incubations amended with varying 15NO3 − concentrations. (a) Peak potential rates of 15N‐N2O production calculated over the first 3 h. (b) Rates of 15N‐N2 production calculated over the full 24 h incubation. Both gases showed increasing production with higher 15NO3 − availability, with rates plateauing at approximately 100 μM, indicating saturation of nitrate‐dependent denitrification rates. (c) Ratios of 15N‐N2O to 15N‐N2 production across 15NO3 − concentrations from all incubation durations. Because of the transient peak N2O production within < 3 h, especially at the two highest NO3 − concentrations, the ratio of N2O to N2 could exceed 1. Dots and vertical lines in each plot represent means and standard errors, respectively (n = 6 ponds per concentration of 15NO3 − at each time point).

Temperature sensitivities of the net production rate (nmol g−1 h−1) of N2O and N2 from denitrification under nitrate‐limited or nitrate‐replete conditions. (a) Net 15N2O production was consistent across temperatures with 10 μM 15NO3 − added. (b) 15N2O production decreased at higher temperatures with 100 μM 15NO3 −. (c) Net 15N2 production was consistent across temperatures with 10 μM 15NO3 −. (d) 15N2 production increased at higher temperatures with 100 μM 15NO3 −. The lines in (b, d) represent the best‐fitting linear mixed‐effect model (Table S4). The full incubation temperature range was 5°C–25°C. n = 37, 39, 38, and 38 incubations, respectively, for panels (a–d), conducted using sediments from eight ponds per 15NO3 − treatment.

Temperature sensitivity of the N2O:N2 production ratio from denitrification under nitrate‐limited or nitrate‐replete conditions. (a) The N2O:N2 production ratio remained consistent across temperatures with 10 μM 15NO3 − added. (b), The ratio decreased at higher temperatures with 100 μM 15NO3 −. The solid line in (b) represents the best‐fitting linear mixed‐effect model (Table S4). The full incubation temperature range was 5°C–25°C. n = 35 and 38 incubations for (a, b), respectively, conducted with sediments from eight ponds per 15NO3 − treatment.

Tables

TABLE 1. Studies included in the meta‐analysis used to characterise the temperature sensitivity of net N2O or N2 production attributed to denitrification in aquatic sediments and terrestrial soils. In most cases, denitrification was the inferred source of N2O or N2, rather than being conclusively distinguished from other potential processes e.g., nitrification (Wrage‐Mönnig et al. 2018; Zhu et al. 2025). The dataset comprised 1250 measurements (j) across 25 studies (i), with 525 measurements in aquatic sediments and 680 in terrestrial soils.

Ecosystem	Measurement	Denitrification	Substrate	Study
Aquatic	N2O	Inferred	NO3 −: 57 μM	Myrstener et al. (2016)
	N2	Inferred	Unspecified DIN	Nowicki (1994) a
	N2	Inferred	Ambient	Seitzinger et al. (1984)
	15N2	Distinguished	15NO3 −: 100 μM	Brin et al. (2014)
	15N2	Distinguished	15NO3 −: 100 μM	Brin et al. (2017)
	15N2	Distinguished	15NO3 −: 50 μM	Rysgaard et al. (2004)
	15N2	Inferred	15NO3 −: 5.9 to 20.2	Veraart et al. (2011) b
	N2O, 15N2	Inferred	NO3 −: 30 μM	Silvennoinen et al. (2008)
	15N2O, 15N2	Distinguished	15NO3 −: 100 μM	Tan et al. (2020)
Terrestrial	N2O	Inferred	15NO3 −: close to ambient	Billings and Tiemann (2014)
	N2O	Inferred	NO3 −: 10204 μM	Dobbie and Smith (2001)
	N2O	Inferred	NO3 −: 8571 μM	Smith et al. (1998)
	N2O	Inferred	NO3 −: 2654 μM	Benoit et al. (2015)
	N2O	Inferred	NO3 −: 10093 μM	del Prado et al. (2006)
	N2O	Inferred	Ambient NO3 −: 286 μM	McKenney et al. (1984)
	N2O	Inferred	NO3 −: > 85 times ambient	Kurganova and de Lopes Gerenyu (2010)
	15N2O	Distinguished	15NO3 −: > 18,000 μM	Duan et al. (2019)
	N2	Distinguished	N2O: 4 μM	Qin et al. (2014) c
	N2	Inferred	NO3 −: 5497 μM	Castaldi (2000)
	N2	Inferred	N2O: 4063 μM	Holtan‐Hartwig et al. (2002)
	N2O, N2	Inferred	NO3 −: 34286 μM	Bailey (1976)
	N2O, N2	Inferred	NO3 −: 8700 μM	Keeney et al. (1979)
	15N2O, 15N2	Inferred	15NO3 −: > 8000 μM	Lai and Denton (2018) b
	15N2O, 15N2	Distinguished	15NO3 −: 7143 μM	Yu et al. (2023)
	N2O/N2	Inferred	Ambient NO3 −: 65–387 μM	Maag and Vinther (1996)

TABLE 2. Trial experiment conducted to determine the optimal substrate concentration and incubation duration for characterising N2O and N2 production from denitrification.

Treatment	15NO3 − (μM)	Time (h)	Targeted product (s)
Control	0	0, 0.5, 3, 6, 12, 24	45N2O, 46N2O, 29N2, 30N2
15NO3 −	10, 20, 50, 100	0, 0.5, 3, 6, 12, 24	45N2O, 46N2O, 29N2, 30N2

TABLE 3. Experimental designs to characterise the temperature sensitivities of N2O and N2 production via denitrification, to test for potential anammox activity, and to quantify potential N2O production from nitrification. All treatments were applied to independent sediment samples collected from eight independent ponds. 15N‐substrate concentrations in the slurries are nominal. Note that in the combined anammox experiment, 14NO3 − was added at a nominal concentration of 100 μM.

Treatment	15NO3 − (μM)	15NH4 + (μM)	Temperature (°C)	Targeted product(s)
Denitrification
Control	0	n.a.	5, 10, 15, 20, 25	45N2O, 46N2O, 29N2, 30N2
15NO3 −	10 or 100	n.a.	5, 10, 15, 20, 25	45N2O, 46N2O, 29N2, 30N2
Anammox
15NH4 +	n.a.	10 or 60	15	29N2
15NH4 ++14NO3 −	n.a.	10 or 60	15	29N2
Nitrification
Control	n.a.	0	15	45N2O, 46N2O
15NH4 +	n.a.	22 or 44	15	45N2O, 46N2O
15NH4 ++ATU	n.a.	44	15	45N2O, 46N2O

Full text

Introduction

1

N_2_O has approximately 273 times the warming potential of carbon dioxide (CO_2_) (Intergovernmental Panel on Climate Change 2021), is the primary driver of stratospheric ozone depletion (Ravishankara et al. 2009), and its atmospheric concentration continues to rise at an accelerating rate (Meinshausen et al. 2011; Tian et al. 2024). Small freshwater bodies are increasingly recognised as disproportionately large sources of greenhouse gases for their size (Holgerson and Raymond 2016; Li et al. 2024; Zhu et al. 2020), and global N_2_O emissions from lakes and reservoirs have increased by 126% since the 1850s (Li et al. 2024).

The microbial denitrification pathway NO_3_ ^−^ → NO_2_ ^−^ → NO → N_2_O → N_2_ (Knowles 1982), is a key process responsible for both the production and reduction of N_2_O in aquatic sediments (Beaulieu et al. 2011; Codispoti et al. 2001; Maavara et al. 2019) and terrestrial soils (Billings and Tiemann 2014; Yu et al. 2023). As denitrification involves both the production and reduction of N_2_O, the balance between these two steps determines net N_2_O emission to the atmosphere. The activity of each step is regulated by environmental factors including nitrate availability, dissolved oxygen concentration, organic carbon supply, and temperature (Codispoti et al. 2001; Ji et al. 2018; Kuypers et al. 2018).

Temperature is a key regulator of total N_2_O and N_2_ production via denitrification, with both generally increasing as temperatures rise (Bailey and Beauchamp 1973; Holtan‐Hartwig et al. 2002; Keeney et al. 1979; Seitzinger et al. 1984; Silvennoinen et al. 2008; Tan et al. 2020). However, reported responses of net N_2_O production to temperature vary among studies. Some have observed a positive relationship, suggesting that N_2_O production exceeds its reduction to N_2_ at higher temperatures (Dobbie and Smith 2001; Duan et al. 2019; McKenney et al. 1984; Myrstener et al. 2016; Smith et al. 1998), whereas others have shown a negative relationship, with N_2_ production increasing more strongly than N_2_O as temperatures rise, resulting in lower net N_2_O production (Silvennoinen et al. 2008). Several studies have also found no significant temperature effect on net N_2_O production (Bailey 1976; del Prado et al. 2006; Lai and Denton 2018). This variability suggests that additional environmental factors may modify the temperature sensitivity of net N_2_O production and the N_2_O:N_2_ balance, introducing uncertainty into predictions of N_2_O‐climate feedback under global warming.

For example, nitrate (NO_3_ ^−^) is a key substrate for denitrification, directly regulating the production of both N_2_O and N_2_ (Baulch et al. 2011; Beaulieu et al. 2011; Palacin‐Lizarbe et al. 2018). As anthropogenic N_2_O emissions largely stem from nitrogen pollution associated with fertilizer use, nitrate availability could mediate how N_2_O emissions respond to the combined pressures of climate warming and nitrogen enrichment (Craswell 2021; Tian et al. 2024; Zhu et al. 2025). However, most existing studies characterising the temperature sensitivity of N_2_O and N_2_ production have applied NO_3_ ^−^ enrichments far above ambient concentrations. In soils, ambient nitrate concentration are typically below 0.4 mM (Dobbie and Smith 2001; Maag and Vinther 1996; McKenney et al. 1984), yet former experimental additions were routinely at millimolar concentrations—approximately 10–100 times higher than ambient (del Prado et al. 2006; Dobbie and Smith 2001; Duan et al. 2019; Keeney et al. 1979). Excess nitrate additions (> 10 times ambient concentrations) have also been used in studies of aquatic sediments, although both ambient concentrations and experimental additions were lower than in soils (Brin et al. 2014, 2017; Myrstener et al. 2016; Rysgaard et al. 2004). Consequently, the effect of nitrate availability on the temperature sensitivity of N_2_O and N_2_ production under closer‐to‐natural conditions is still unclear.

To date, only a limited number of studies have characterised the temperature sensitivity of net N_2_O production in sediments (Lai and Denton 2018; Myrstener et al. 2016; Silvennoinen et al. 2008; Tan et al. 2020), and only one has simultaneously quantified ^15^N‐labelled N_2_O and N_2_ to directly link their production to denitrification (Tan et al. 2020). Because multiple microbial pathways can contribute to N_2_O production and its reduction (Zhu et al. 2025), isolating the denitrification signal is critical. However, how temperature and nitrate availability co‐regulate denitrification‐derived N_2_O and N_2_ production remains elusive.

In this study, we address these knowledge gaps by combining a meta‐analysis of published data with controlled ^15^N‐labelling experiments. We first used existing datasets to predict the apparent temperature sensitivities of N_2_O and N_2_ production, and their ratio, in both soils and aquatic sediments. As most previous studies (16 out of 25; 64%) were conducted under nitrate‐replete conditions, we then experimentally tested whether these temperature sensitivities change with nitrate availability. We quantified ^15^N‐N_2_O and ^15^N‐N_2_ production under varying ^15^NO_3_ ^−^ concentrations to examine how temperature and nitrate availability jointly regulate the relative production of these two denitrification gases in nitrate‐limited freshwater pond sediments. Using ^15^N‐tracers, we confirmed the negligible contribution of anammox and nitrification to N_2_ and N_2_O production, respectively, ensuring that denitrification was the sole source of both gases. We addressed two key questions: (1) How does the net production of N_2_O and N_2_, and their ratio, respond to warming in a nitrate‐limited system? and (2) Does nitrate availability modify the temperature sensitivity of these processes?

Materials and Methods

2

Meta‐Analysis: Temperature Sensitivities of Net N2O and N2

Production From Denitrification

2.1

Here we define “Inferred” as N_2_O or N_2_ attributed to denitrification without confirmatory evidence excluding other sources and “Distinguished” as N_2_O or N_2_ production from denitrification confirmed using isotope‐pairing techniques. We compiled published studies reporting rates of net N_2_O and N_2_ production and their ratios, inferred or distinguished from denitrification across a range of temperatures in both aquatic sediments and terrestrial soils (Table 1), providing the first meta‐analysis of their temperature sensitivities. We summarised the gas analysis methods and nitrate concentrations used in each study and noted whether denitrification was inferred or explicitly distinguished as the sources of N_2_O and N_2_ production. We then evaluated the temperature sensitivities of net N_2_O production, N_2_ production, and their ratio (see Statistical Analysis).

Optimisation of Incubation Conditions for Characterising N2O and N2

Production From Denitrification

2.2

Here we used our well‐established experimental freshwater ponds—that are known to be nitrate‐limited (< 2 μM, Table S1) to experimentally characterise the temperature sensitivity of N_2_O and N_2_ production and how that interacts with NO_3_ ^−^ during denitrification (Si et al. 2023; Yvon‐Durocher et al. 2010). We also know that the pond sediments have well‐developed denitrifying communities (see Data S1 and Table 2), with 43 out of 48 common genera known to harbour denitrifying species present (Zumft 1997). As half of the experimental ponds have been maintained at 4°C above their paired ambient ponds since 2006 (Yvon‐Durocher et al. 2010), we also tested whether this long‐term warming influenced gas production from denitrification.

First, we conducted a trial to determine the optimal nitrate concentration and incubation duration that enabled quantifiable, simultaneous detection of N_2_O and N_2_, ensuring that subsequent temperature sensitivity estimates were not confounded by substrate limitation or product turnover. Plexiglass corers were used to collect sediments from six experimental ponds in December 2020, with three technical‐replicates taken from different locations within each pond (Trimmer and Nicholls 2009). Sediment cores were transported to the laboratory in the dark on ice packs (4 h transport time) and stored overnight at 4°C. Prior to incubation, the cores and pre‐weighed vials (12 mL Exetainer, Labco) were placed inside an anoxic glove box (residual O_2_ < 5 ppm, Belle Technology), which was constantly flushed with oxygen‐free N_2_ (OFN) recycled through oxygen‐scrubbing catalytic cartridges. Anoxic medium was made by flushing N‐free artificial pond water medium (Si et al. 2023) with OFN for 20 min. The top 2 cm of the cores for each pond were homogenized, and 2 mL of sediment slurry and 4 mL of the medium were added to each gas‐tight vial. The slurry was added carefully to prevent sediment from remaining on the vial thread, which could compromise sealing and introduce air contamination. The vials were then sealed and pre‐incubated at 15°C in a temperature‐controlled room overnight to reduce any residual porewater NO_x_ ^−^ and oxygen (Figure S1) (Risgaard‐Petersen et al. 2004; Trimmer et al. 2003).

After pre‐incubation, the vials were amended with 50 μL of ^15^NO_3_ ^−^ stock solutions (98% of ^15^N, Sigma Aldrich) to achieve nominal final concentrations of 10, 20, 50, or 100 μM, with un‐amended vials serving as controls (Table 2, and see Figure S1). Each concentration treatment was incubated for 0.5, 3, 6, 12, and 24 h at 15°C, corresponding to the annual average temperature in the ambient ponds (Si et al. 2023). Independent vials were used for each pond, nitrate concentration, and incubation duration. At each time point, microbial activities were terminated by injecting 100 μL of formaldehyde (37 wt. %) through the septa and the vials then equilibrated at room temperature (22°C) until further analysis.

As formaldehyde interferes with the colorimetric assays for nitrite (NO_2_ ^−^), nitrate (NO_3_ ^−^), and ammonium (NH_4_ ^+^), we prepared parallel samples for separate measurement of gases and dissolved nutrients (Ouyang et al. 2021). At each time point, samples for nutrient analysis were centrifuged immediately, and supernatants were frozen at −20°C until analysis. Nutrient concentrations were measured using an automated wet‐chemistry autoanalyzer (San^++^, SKALAR Analytical B.V.) following standard colorimetric techniques (Kirkwood 1996). Detection limits were 0.05 μM for NO_2_ ^−^, 0.1 μM for NO_x_ ^−^ (NO_2_ ^−^ + NO_3_ ^−^), and 0.2 μM for NH_4_ ^+^.

For ^15^N‐N_2_O concentration measurement, headspace sub‐samples from each vial were transferred into an air‐filled 12 mL gas‐tight vial and analysed on a continuous flow isotope ratio mass spectrometer (CF‐IRMS, Delta V Plus, Thermo Finnigan) coupled to an automated trace gas pre‐concentrator (Precon, Thermo Finnigan). Calibration was performed with air, 0.12 ppm, 1.04 ppm, and serially diluted 96 ppm N_2_O standards (BOC Limited), yielding a linear increase between peak area and N_2_O mole amount (0.08–5.85 nmol per vial). Concentrations of ^15^N_2_ were measured on the CF‐IRMS using 100 μL of sample headspace, bypassing the reduction column to avoid the reduction of N_2_O to N_2_ (Trimmer and Nicholls 2009). Signal drift at mass 30 was corrected by inserting air standards after every 10 samples (Si et al. 2023).

The net production rates of N_2_O and N_2_ in each vial were calculated from the headspace concentrations and gas solubilities at equilibrium temperature, based on (Weiss and Price 1980) for N_2_O and (Weiss 1970) for N_2_ (Si et al. 2023). The production of ^15^N‐N_2_O (^45^N_2_O + 2 × ^46^N_2_O) and ^15^N‐N_2_ (^29^N_2_ + 2 × ^30^N_2_) was calculated as the excess production in the ^15^NO_3_ ^−^ treatments relative to unamended controls. After gas measurements, vials were centrifuged, supernatants removed, and the remaining sediments oven‐dried to constant weight to obtain dry weight‐normalised production rates.

Characterising the Temperature Sensitivities of N2O and N2

Production From Denitrification

2.3

From the trial experiments, we identified the optimal nitrate concentrations and incubation duration for subsequent temperature‐sensitivity measurements: 10 μM representing nitrate‐limited conditions and 100 μM nitrate‐replete conditions (Figure 3 and Figure S2), and a 3 h incubation that enabled robust, simultaneous quantification of N_2_O and N_2_ at both nitrate concentrations (Figure S2). We then collected additional sediment cores from eight experimental ponds in September 2021, to characterise the temperature sensitivities of N_2_O and N_2_ production from denitrification. Following 24 h pre‐incubation, vials were amended with ^15^NO_3_ ^−^ to nominal final concentrations of 0 (control), 10, or 100 μM (Table 3), and each treatment was incubated for 3 h. The samples were incubated across a temperature gradient of 5°C, 10°C, 15°C, 20°C, and 25°C, selected to reflect seasonal variation in daily mean water temperature in the ponds (Si et al. 2023). This temperature range is representative of temperate freshwater systems in the UK and is also consistent with broader patterns across temperate freshwaters globally. For example, an analysis of 635 lakes worldwide showed that lakes located between 30° and 60° latitude have mean annual water temperatures predominantly within this range (Woolway and Merchant 2019). All other procedures for sediment collection, sample preparation, and measurements remained the same as in the trial experiment.

TABLE 3: Experimental designs to characterise the temperature sensitivities of N2O and N2 production via denitrification, to test for potential anammox activity, and to quantify potential N2O production from nitrification. All treatments were applied to independent sediment samples collected from eight independent ponds. 15N‐substrate concentrations in the slurries are nominal. Note that in the combined anammox experiment, 14NO3 − was added at a nominal concentration of 100 μM.

Characterising the Potential of Anammox and Nitrification

2.4

As both denitrification and anaerobic ammonium oxidation (anammox; NH_4_ ^+^ + NO_2_ ^−^ → N_2_) produce N_2_, we distinguished between these two processes by analysing the production of ^29^N_2_ (^14^N^15^N) and ^30^N_2_ (^15^N^15^N) from different combinations of ^15^N‐labeled substrates (Brin et al. 2014; Dalsgaard et al. 2003; Kuypers et al. 2003; Trimmer et al. 2013). We performed a parallel experiment with two additional sets of incubations with the pond sediments: a ^15^NH_4_ ^+^ only treatment, and a ^15^NH_4_ ^+^ plus ^14^NO_3_ ^−^ treatment (Table 3). Each treatment included two concentrations of ^15^NH_4_ ^+^ (nominal final concentration of 10 or 60 μM, prepared from 98% ^15^N‐NH_4_ ^+^, Sigma Aldrich) to test for anammox activity under both low and high substrate availability. In the combined treatment, ^14^NO_3_ ^−^ was added at a final concentration of 100 μM, consistent with the high ^15^NO_3_ ^−^ concentration used in the denitrification experiment. After incubation, ^29^N_2_ concentrations were measured to assess potential anammox activity.

In addition, we tested the potential for N_2_O production via nitrification (NH_4_ ^+^ → NH_2_OH → N_2_O). Sediment cores were collected from eight ponds in February 2022 using the same sampling procedures as in the denitrification experiments but with a headspace of air rather than N_2_. Vials were amended with ^15^NH_4_ ^+^ (final concentration 22 μM or 44 μM) with or without allylthiourea (ATU; final concentration ~80 μM), where ATU inhibited the oxidation of NH_4_ ^+^ (Ginestet et al. 1998). Un‐amended vials served as controls to account for background nitrification (Table 3). Independent vials were used for each pond and treatment and incubated for 0, 3, 8, 18, or 24 h at 15°C. Microbial activities in samples were terminated with formaldehyde at each time point, and N_2_O was measured using the same methods as in the denitrification experiments.

Statistical Analysis

2.5

All statistical analyses and plotting were performed in R version 4.5.0 (R Core Team 2025) using RStudio. For the meta‐analysis, the dataset comprised 1250 temperature–rate observations drawn from 25 independent studies (Table 1), including 525 measurements from aquatic sediments and 680 from terrestrial soils. In the meta‐analysis, we fitted the reported net rates of N_2_O production, N_2_ production, and the N_2_O:N_2_ ratio using linear mixed‐effects models to quantify their temperature sensitivities as apparent activation energies. The apparent activation energies reported here are best interpreted as emergent, sediment‐ or soil‐level properties of denitrification, integrating the combined temperature responses of N_2_O production and reduction within natural systems, rather than as intrinsic kinetic parameters of individual enzymes. Such system‐scale temperature sensitivities are widely used in biogeochemistry and ecosystem ecology to characterise how key processes—such as ecosystem respiration (Yvon‐Durocher et al. 2012), primary production (Yvon‐Durocher et al. 2010), denitrification, anammox (Brin et al. 2014, 2017; Tan et al. 2020), N_2_‐fixation (Welter et al. 2015), and methane production or emissions (Yvon‐Durocher et al. 2014; Zhu et al. 2020)—respond to warming across complex, multi‐pathway systems.

For each observation j (j = 1, …1250) from study i (i = 1, …25), Fij denotes the measured rate or ratio at absolute temperature Tij (K), and the response variable is the natural log‐transformed value, lnFij. Temperature dependence was modelled using a centred Boltzmann–Arrhenius formulation (Yvon‐Durocher et al. 2014; Zhu et al. 2020):

$$[eqn]$$

where k is the Boltzmann constant (8.62 × 10^−5^ eV K^−1^), E¯ is the overall apparent activation energy (eV), ai and bi are study‐specific random slope and intercept terms, lnFTc⃐ is the mean log‐rate (or log‐ratio) at the reference temperature Tc, Ecoij is ecosystem type (terrestrial soil or aquatic sediment) with associated fixed effect βeco, and εij is the residual error term. The centring temperature was defined as the midpoint of the observed temperature range in Kelvin, Tc=Tmax+Tmin/2. Models were fitted using the lme4 package (Bates et al. 2015), with full models including interactions between temperature and ecosystem type and both random slopes and intercepts were ranked by the small sample‐size corrected Akaike Information Criterion (AICc) with the ‘MuMIn’ package. For visualisation only (Figures 1 and 2), rates were standardised by subtracting the study‐specific intercept to remove differences in rate normalisation among studies; this transformation was not applied during model fitting, where between‐study differences were accounted for via the random intercept term.

Meta‐analysis of published rates of net N2O and N2 production from denitrification (inferred and distinguished, Table 1) as a function of temperature in aquatic sediments and terrestrial soils. The rate of net N2O (a) and N2 (b) production increased at higher temperatures in both aquatic sediments and terrestrial soils. Data were visualised using the “Visreg” package in R (Breheny and Burchett 2017), plotted as partial residuals (brown and green circles) from the best fitting mixed‐effects models after accounting for the random effects, with the overall temperature sensitivity estimates shown as black lines (Table S3). For net N2O production, n = 100 (aquatic) and 276 (terrestrial) measurements; for N2 productions, n = 330 (aquatic) and 208 (terrestrial) measurements. As normalisation methods differed among studies, all production rates were standardised by centring each study's N2O or N2 production rates on its specific intercept.

Ratio of net N2O to N2 production inferred from denitrification as a function of temperature, based on incubations from aquatic sediments and terrestrial soils. n = 95 and 196 measurements for aquatic and terrestrial datasets, respectively. The black line gives the apparent activation energy derived from the slope of the best fitting linear mixed‐effects model using actual parallel measurements of net N2O and N2 production (Table S3, Model M0 for N2O:N2). The grey line shows the apparent activation energy predicted from the independent net N2O and N2 datasets shown in Figure 1. The two estimates did not differ significantly (χ 2 = 1.23, p = 0.27; linear hypothesis test for fixed effects, ‘car’ package in R).

Similarly, for the pond sediment experiments, apparent activation energies for net rates of N_2_O production, N_2_ production, and their ratio, were estimated by fitting these natural log‐transformed variables into Equation (1) where i and j are the ponds and their respective observations and TC the average incubation temperature (15°C) in the laboratory. Pond status (warmed vs. ambient) was included as an additional fixed effect in the full model, together with incubation temperature and nitrate concentration, and pond identity was treated with random slope and random intercept terms, as above. We found no significant differences between ambient and warmed ponds in either the slope or intercept of the temperature‐rate relationship at either 10 μM or 100 μM ^15^NO_3_ ^−^ (all p > 0.05, Figure S3). Therefore, the long‐term warming treatment was not retained in the full model and is not discussed further. Model fitting and selection followed the same approach as in the meta‐analysis, with apparent activation energies derived from the slopes of the best‐fitting models.

Results

3

Meta‐Analysis: Temperature Sensitivities of N2O and N2

Production From Denitrification

3.1

From the meta‐data analysis of published studies, net N_2_O production from denitrification increased at higher temperatures, with an apparent activation energy (Ea) of 0.47 eV (95% CI: 0.36–0.58 eV, Figure 1a). The temperature sensitivity of net N_2_O production was consistent across aquatic sediments and terrestrial soils (Table S3, LRT comparing models M0 and M2 for net N_2_O: χ ^2^ = 5.03, d.f. = 2, p = 0.08). Similarly, N_2_ production from denitrification increased at higher temperatures, with an apparent activation energy of 0.62 eV (95% CI: 0.48–0.75 eV, Figure 1b; Table S3, comparing M0 and the null model for N_2_: χ ^2^ = 28.6, d.f. = 1, p < 0.001). The temperature sensitivity of N_2_ production was also indistinguishable between aquatic sediments and terrestrial soils (Table S3, comparing M0 and M2 for N_2_: χ ^2^ = 0.001, d.f. = 2, p = 0.999).

For studies reporting parallel measurements of net N_2_O and N_2_ production, we derived the temperature sensitivity of the N_2_O:N_2_ ratio directly and found it to be negative, with an Ea of −0.22 eV (Figure 2, black line; Table S3, comparing M0 and null model for N_2_O:N_2_, χ^2^ = 11.4, d.f. = 1, p < 0.001). We also used the derived temperature sensitivities for net N_2_O and N_2_ production from the independent (non‐parallel) datasets (Figure 1) to predict the temperature sensitivity of the N_2_O:N_2_ ratio (Figure 2, grey line). The apparent activation energies derived from the two approaches were statistically indistinguishable (p = 0.27; linear hypothesis test for fixed effects using the ‘car’ package). These results indicated that while both net N_2_O and N_2_ production increased at higher temperatures, the lower temperature sensitivity of net N_2_O led to a negative relationship between N_2_O:N_2_ and temperature. Critically, most of these former studies conducted incubations with high NO_3_ ^−^ enrichment, limiting the applicability of these findings to ambient conditions, especially nitrate‐limited systems.

In addition, while these former studies commonly inferred denitrification as the source of N_2_O or N_2_ production, most did not explicitly distinguish it from other microbial pathways involved in N_2_O or N_2_ production (Table 1). As such, the reported temperature sensitivities likely reflect combined effects from multiple processes rather than denitrification alone.

Optimal Incubation Conditions for Characterising Net Production of N2O and N2

From Denitrification

3.2

We could not detect any excess ^29^N_2_ production in either the ^15^NH_4_ ^+^‐only treatments or those amended with ^15^NH_4_ ^+^+^14^NO_3_ ^−^ (Figure S4), confirming the absence of potential anammox activity in our pond sediments. In addition, we observed no significant N_2_O production from nitrification (Figure S5). Together, these results indicate that denitrification was the sole pathway responsible for both N_2_O and N_2_ production in the pond sediments.

In the denitrification trial incubations, ^15^N‐N_2_O production peaked and subsequently declined during all incubations (Figure S2a). At ^15^NO_3_ ^−^ concentrations below 100 μM, ^15^N‐N_2_O reached its maximum within approximately 0.5 h and declined to zero before 24 h in most incubations. In contrast, at 100 μM ^15^NO_3_ ^−^, the ^15^N‐N_2_O peak occurred later (at ~3 h) and reached a higher production than with lower nitrate concentrations. Meanwhile, ^15^N‐N_2_ accumulated continuously over 24 h in treatments with ^15^NO_3_ ^−^ above 10 μM, whereas at 10 μM ^15^NO_3_ ^−^, ^15^N‐N_2_ production plateaued after approximately 12 h (Figure S2b). The rapid rise and subsequent loss of transient ^15^N‐N_2_O, together with the plateau in ^15^N‐N_2_ at 10 μM ^15^NO_3_ ^−^, indicated substrate‐limitation during the incubation. Based on these dynamics, we estimated peak ^15^N‐N_2_O production rates during the first 3 h, corresponding to the latest observed N_2_O peak across all ^15^NO_3_ ^−^ treatments, and ^15^N‐N_2_ production rates over the full 24 h incubation to capture their respective maximum potentials.

Rates of both ^15^N‐N_2_O and ^15^N‐N_2_ production increased at higher ^15^NO_3_ ^−^ concentrations and then plateaued at approximately 100 μM, indicating saturation of nitrate‐dependent denitrification rates (Figure 3a,b). Meanwhile, the ratio of ^15^N‐N_2_O to ^15^N‐N_2_ production increased up to 50 μM ^15^NO_3_ ^−^ and then plateaued (Figure 3c). In addition, 43% of the ^15^NO_3_ ^−^ added was consistently reduced to ^15^N_2_ after 24 h of incubation across all ^15^NO_3_ ^−^ concentrations, on average (Figure S6, p = 0.52, df = 14, Kruskal‐Wallis test). This indicated that the fraction of the ^15^NO_3_ ^−^ reduced to the end‐product N_2_ was not affected by NO_3_ ^−^ availability over the tested range of 10–100 μM.

Rates of 15N2O and 15N2 production and their ratios, from independent sediment incubations amended with varying 15NO3 − concentrations. (a) Peak potential rates of 15N‐N2O production calculated over the first 3 h. (b) Rates of 15N‐N2 production calculated over the full 24 h incubation. Both gases showed increasing production with higher 15NO3 − availability, with rates plateauing at approximately 100 μM, indicating saturation of nitrate‐dependent denitrification rates. (c) Ratios of 15N‐N2O to 15N‐N2 production across 15NO3 − concentrations from all incubation durations. Because of the transient peak N2O production within < 3 h, especially at the two highest NO3 − concentrations, the ratio of N2O to N2 could exceed 1. Dots and vertical lines in each plot represent means and standard errors, respectively (n = 6 ponds per concentration of 15NO3 − at each time point).

Based on these initial trial incubations, to characterise how nitrate availability interacts with the temperature sensitivity of N_2_O and N_2_ production, the sediments were subsequently incubated with either 10 μM or 100 μM ^15^NO_3_ ^−^ for approximately 3 h.

Temperature Sensitivities of N2O and N2

Production From Denitrification

3.3

Net production of both ^15^N_2_O and ^15^N_2_ was detected in 95% of the subsequent 3 h incubations (152 out of 160 incubations) with either 10 μM or 100 μM ^15^NO_3_ ^−^ and the temperature response of N_2_O and N_2_ production differed according to NO_3_ ^−^ availability. At 10 μM ^15^NO_3_ ^−^, net production of both gases showed no significant increase at higher temperatures (Figure 4a,c; Table S4, M10.a compared to M10.b for both gases: p = 0.38 and p = 0.18 for N_2_O and N_2_, respectively). In contrast, at 100 μM ^15^NO_3_ ^−^, both ^15^N_2_O and ^15^N_2_ production were sensitive to temperature (Table S4, M100.a compared to M100.b for both gases: p < 0.05 and p < 0.001 for N_2_O and N_2_, respectively), but with opposite signs. For example, net N_2_O production had a negative temperature sensitivity (−0.25 eV; Figure 4b), while N_2_ production was positive across the 5°C–25°C range (0.43 eV; Figure 4d).

Temperature sensitivities of the net production rate (nmol g−1 h−1) of N2O and N2 from denitrification under nitrate‐limited or nitrate‐replete conditions. (a) Net 15N2O production was consistent across temperatures with 10 μM 15NO3 − added. (b) 15N2O production decreased at higher temperatures with 100 μM 15NO3 −. (c) Net 15N2 production was consistent across temperatures with 10 μM 15NO3 −. (d) 15N2 production increased at higher temperatures with 100 μM 15NO3 −. The lines in (b, d) represent the best‐fitting linear mixed‐effect model (Table S4). The full incubation temperature range was 5°C–25°C. n = 37, 39, 38, and 38 incubations, respectively, for panels (a–d), conducted using sediments from eight ponds per 15NO3 − treatment.

At 10 μM ^15^NO_3_ ^−^, the ^15^N_2_O:^15^N_2_ ratio remained consistent across temperatures (Figure 5a p > 0.05, M10.a compared to M10.b for the ratio, Table S4). In contrast, at 100 μM ^15^NO_3_ ^−^, the ^15^N_2_O:^15^N_2_ ratio decreased exponentially from 5°C to 25°C (Figure 5b; Table S4, M100.a compared to M100.b for the ratio: χ ^2^ = 24.6, d.f. = 1, p < 0.001) i.e., a negative temperature sensitivity (−0.7 eV; −0.93 to −0.47 eV, 95% CI, Figure 5b) which indicated a higher relative accumulation of N_2_O compared to N_2_ at colder temperatures when nitrate was replete.

Temperature sensitivity of the N2O:N2 production ratio from denitrification under nitrate‐limited or nitrate‐replete conditions. (a) The N2O:N2 production ratio remained consistent across temperatures with 10 μM 15NO3 − added. (b), The ratio decreased at higher temperatures with 100 μM 15NO3 −. The solid line in (b) represents the best‐fitting linear mixed‐effect model (Table S4). The full incubation temperature range was 5°C–25°C. n = 35 and 38 incubations for (a, b), respectively, conducted with sediments from eight ponds per 15NO3 − treatment.

Discussion

4

Our meta‐data analysis of published data showed that both net N_2_O and N_2_ production from denitrification increased at higher temperatures, with apparent activation energies of 0.47 eV and 0.62 eV, respectively (Figure 1). These temperature sensitivities were consistent across aquatic sediments and terrestrial soils. Because N_2_ production was more temperature sensitive than net N_2_O production, the N_2_O:N_2_ ratio declined at higher temperatures (Figure 2). However, most of these studies were conducted under high NO_3_ ^−^ enrichment far exceeding ambient concentrations (Table 1)—often by 10–100 times (Brin et al. 2014, 2017; del Prado et al. 2006; Dobbie and Smith 2001; Duan et al. 2019; Keeney et al. 1979; Myrstener et al. 2016; Rysgaard et al. 2004)—meaning that the predicted temperature sensitivities likely apply only when nitrate is replete. Moreover, few of these former studies employed ^15^N‐labelling techniques (Tan et al. 2020), meaning that N_2_O or N_2_ production has often been attributed to denitrification without excluding potential contributions from other microbial pathways.

Here, we addressed these limitations by applying ^15^N‐labelling techniques to directly isolate denitrification and quantify its temperature sensitivity in N‐limited aquatic sediments. While our results showed that both net N_2_O production and its reduction to N_2_ occurred across 5°C–25°C under both 10 and 100 μM ^15^NO_3_ ^−^, the effect of temperature was only evident at the higher nitrate concentration (Figure 4b,d). Under nitrate‐limited conditions (10 μM ^15^NO_3_ ^−^), neither net N_2_O production nor its reduction to N_2_ responded significantly to temperature (Figure 4a,c). Similarly, a previous study reported that both the rate and temperature sensitivity of N_2_ production were substantially higher in inorganic nitrogen‐enriched (NO_3_ ^−^ and/or NH_4_ ^+^, unspecified) estuarine mesocosms (1.1 eV) than in controls (0.4 eV) (Nowicki 1994). Together, these results indicate that nitrate availability modulates the temperature sensitivity of N_2_O and N_2_ production during denitrification, suggesting that nutrient‐polluted systems may be more responsive to climate warming, with shifts in the balance between N_2_O and N_2_ production potentially altering their overall greenhouse gas emissions.

According to Arrhenius kinetics, enzyme‐mediated reaction rates generally increase with temperature. However, when substrate availability is low, microbial activity can be constrained by substrate supply, rather than temperature sensitivity. Beyond the denitrification patterns observed here, such invariant temperature responses under substrate limitation have been reported for other microbial processes, including N_2_O fixation (Si et al. 2023), ammonia oxidation (Horak et al. 2013; Zheng et al. 2020), and methane oxidation (Lofton et al. 2014; Szafranek‐Nakonieczna et al. 2019). For denitrification, incubations with permeable shelf sediments showed that the half saturation constants for nitrate were higher than ambient concentrations at most sites (Evrard et al. 2013). Similarly, cultures of the marine denitrifier Pseudomonas denitrificans exhibited a half saturation constant for nitrate of 55 μM (Kornaros et al. 1996), exceeding nitrate concentrations found in many marine sediments (Brin et al. 2014; Garcia‐Robledo et al. 2016; Rysgaard et al. 2004; Thamdrup and Dalsgaard 2002). Together, these findings indicate that denitrifiers in natural sediments likely operate under chronic nitrate limitation, which likely dampens their temperature sensitivity.

Here, under nitrate‐replete conditions, N_2_ production increased exponentially between 5°C and 25°C (Figure 4d). In contrast, net N_2_O production decreased (Figure 4b), resulting in lower accumulation at higher temperatures. This opposing temperature response of N_2_O and N_2_ production has also been observed in river sediments, where rising temperature enhanced N_2_ production while suppressing N_2_O production (Silvennoinen et al. 2008). Similar patterns have also been found in soils, with N_2_ production increasing and net N_2_O production declining between 10°C and 30°C (Bailey 1976). Moreover, elevated N_2_O emissions have been documented in soils at very low temperatures—for example, at −1°C compared with above 0°C (Wertz et al. 2013), at 0°C compared with 5°C (Holtan‐Hartwig et al. 2002), and at 4°C compared with 20°C (Melin and Nômmik 1983). Together, these findings indicate that under nitrate‐replete conditions, while higher temperatures promote N_2_ production via complete denitrification, they do not necessarily lead to greater net N_2_O production. This suggests an important trade‐off under climate warming: in nitrate‐replete systems, warming may mitigate N_2_O emissions but simultaneously accelerate nitrogen loss, potentially altering nutrient availability.

Under these conditions, because N_2_ production exhibited greater temperature sensitivity than net N_2_O production (Figure 4), the N_2_O:N_2_ production ratio declined at higher temperatures (Figure 5). This pattern aligns with previous findings of lower N_2_O:N_2_ production ratios at higher temperatures in soils (Bailey 1976; Keeney et al. 1979; Lai and Denton 2018; Maag and Vinther 1996; Yu et al. 2023) and river sediments (Silvennoinen et al. 2008), as reflected in our meta‐analysis (Figure 2). Biogeochemical modelling of European forest soils similarly predicted that a 1.8°C increase in temperature would lower the N_2_O:N_2_ product ratio and reduce N_2_O emissions by 6% (Kesik et al. 2006). The differing temperature sensitivities of N_2_O and N_2_ production may emerge because warming can alter the relative abundance and/or the expression of key denitrification genes, particularly norB and nosZ, which encodes enzymes for N_2_O production and its reduction to N_2_, respectively. For example, short‐term (24 h) experimental warming in grassland soils decreased the norB: nosZ relative abundance ratio, and seasonal shifts in norB abundance and norB: nosZ ratios have been linked to longer‐term changes in N_2_O emissions (Billings and Tiemann 2014). Similarly, lower norB: nosZ expression ratios correlate with lower N_2_O emissions in seagrass meadows, supporting a quantifiable link between gene activities and denitrification rates (He et al. 2024). Future work that directly quantifies the temperature‐driven changes in norB and nosZ‐associated expression alongside N_2_O and N_2_ production, and how these responses vary with nitrate availability, may provide stronger mechanistic basis for the differing temperature sensitivities observed here.

Other soil studies have suggested that elevated N_2_O emissions at low temperatures result from the inhibition of N_2_O reductase activity in the cold (e.g., around 0°C) (Holtan‐Hartwig et al. 2002; Öquist et al. 2007). Conversely, while here we worked with anoxic slurries, in natural soils, N_2_O reduction to N_2_ may be further enhanced at higher temperatures by reduced oxygen solubility and/or lower oxygen availability driven by elevated respiration relative to photosynthesis (Smith 1997; Veraart et al. 2011). Importantly, these pronounced temperature effects appear restricted to nitrate‐replete conditions, whereas under nitrate‐limited conditions, neither N_2_O nor N_2_ production, nor their ratio, responded significantly to warming.

Conclusion

5

Here, we combined a meta‐analysis of published data with ^15^N‐tracer experiments to examine how temperature and nitrate availability regulate N_2_O and N_2_ production from denitrification. The meta‐analysis predicted that the differing temperature sensitivities of N_2_O and N_2_ production would lead to a decline in the N_2_O:N_2_ ratio with warming. However, because most previous studies applied nitrate enrichments far above ambient levels, this temperature dependence may not hold under natural conditions. From our pond sediment incubations under both high and low nitrate availability, we show that the decline in the N_2_O:N_2_ ratio with warming occurred only when nitrate was replete, as substrate availability can outweigh temperature in controlling the balance between N_2_O and N_2_ production. These findings underscore the need to consider both nutrient availability and temperature when predicting N_2_O emissions from natural sediments and soils, as neglecting nutrient limitation may overestimate the impact of warming on nitrogen cycling and N_2_O emissions.

Author Contributions

Yueyue Si: conceptualization, methodology, data curation, formal analysis, visualization, writing – original draft, writing – review and editing, investigation, validation. Mark Trimmer: conceptualization, methodology, supervision, funding acquisition, project administration, writing – review and editing, resources.

Funding

This work was supported by Queen Mary University of London.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: gcb70767‐sup‐0001‐Supinfo.docx.