Introduction

Nitrogen (N) fertiliser use in agriculture has increased to supplement soil N supply to sustain yields under the pressure of global food demand. However, increasing N surpluses on a farm level have raised concerns about inefficient N fertiliser use and the associated losses of N to the environment. Dissolved N losses cause degradation of sensitive water catchments and coastal lagoons (Fowler et al. 2013). Gaseous N emissions in the form of nitrous oxide (N2O) contribute to global climate change, as N2O is a long-lived atmospheric trace gas with a global warming potential 273 times higher than that of carbon dioxide (CO2) over a 100-year period (IPCC 2021) and the largest remaining threat to the stratospheric ozone layer (Portmann et al. 2012; Ravishankara et al. 2009). Croplands are globally responsible for 66% of total anthropogenic N2O emissions which are predicted to double by 2050 (Davidson and Kanter 2014). Fertiliser N inputs increase the magnitude of N2O emissions and account for 54% of the increase in global N2O emissions in the recent decade (Tian et al. 2019). Sustainable agricultural production therefore requires the adaption of N fertiliser application rates which account for soil N supply and plant N demand, reducing N2O emissions and other N loss pathways while maintaining crop productivity (Liu et al. 2018; Mueller et al. 2014; Thompson et al. 2019).

Sugarcane (Saccharum spp.) often receives substantial N fertiliser inputs to fully leverage its large potential for biomass production (Roy et al. 2006). Fertiliser N inputs in sugarcane systems range from 140 to 220 kg N ha− 1 in Australia (Schroeder et al. 2010), 150 to 400 kg N ha− 1 in India (Robinson et al. 2011) and from 400 to 800 kg N ha− 1 in China (Li and Yang 2015). Surplus N inputs are often applied as economic insurance to protect profits against uncertainty in crop growth (Calcino et al. 2010), likely leading to applied N exceeding crop N demand. Excessive N fertiliser application leads to low N uptake efficiency and large N losses (Thorburn et al. 2017) and substantial N2O emissions from sugarcane systems have been reported (Allen et al. 2010; Wang et al. 2016). The industry-specific emission factor (EF) of fertiliser-induced N2O for sugarcane in Australia assumes 2% of applied N fertiliser lost as N2O (DISER 2020), yet an exponential response of N2O emissions to N rates has been recently reported from an intensive sugarcane system (Takeda et al. 2021a). Losses of N from intensively managed sugarcane systems not only highlight agronomic inefficiencies, but also the large environmental footprint of sugarcane production.

Both the mineralisation of soil organic N and N fertiliser inputs supply N available for plant uptake and loss and thus differentiation of N sources in plant N uptake and N2O emissions is essential to the efficient use of N fertiliser from both agronomic and environmental perspectives. To this end, the difference method has been widely used, calculating the contribution of N fertiliser to plant N uptake or N2O emissions by its difference from an unfertilised control (Antille and Moody 2021; Scheer et al. 2012). This method gives apparent plant fertiliser N uptake efficiency and N2O EFs when dividing fertiliser-derived N uptake and N2O respectively by the N rate. This method assumes that the contribution of soil N in fertilised treatments is equal to plant N uptake or N2O emissions in the non-fertilised control. However, N fertiliser application can affect soil N mineralisation, termed the priming effect (Kuzyakov et al. 2000), and thus the contribution of soil N, leading to a bias regarding N source partitioning by the difference method. The use of 15N labelled fertiliser can trace the fate of fertiliser N in crop and soil and provides information regarding the amount of fertiliser N lost to the environment as the balance between recovered and applied 15N fertiliser (Dourado-Neto et al. 2010). Fertiliser 15N uptake in sugarcane commonly accounts for only 20–40% of the N applied (Boschiero et al. 2018; Chapman et al. 1994; da Silva et al. 2020; Franco et al. 2011; Prasertsak et al. 2002; Takeda et al. 2021b; Vallis et al. 1996), which is lower than values typically reported from other field crops such as maize (65%), wheat (57%), and rice (46%) (Ladha et al. 2005). Subtracting the recovered fertiliser N from the total N in the respective pool enables evaluation of soil N contribution to this N pool. Soil-derived N usually dominates plant N uptake despite the high N inputs in intensive sugarcane systems (Franco et al. 2011; Joris et al. 2020; Oliveira et al. 2017; Prasertsak et al. 2002), and fertiliser N application has recently been shown to increase soil-derived N uptake in an intensive sugarcane system (Takeda et al. 2021b), implying that N turnover is dominated by soil-derived N which can contribute to N2O emissions.

However, studies partitioning fertiliser and soil-derived N2O emissions under field conditions are limited to grassland (Buckthought et al. 2015; Friedl et al. 2017; Lampe et al. 2006; Phillips et al. 2013) and grain systems (Liao et al. 2021; Linzmeier et al. 2001; Xu et al. 2019, 2021). Incubation and greenhouse studies showed increases in soil-derived N2O production in response to N fertiliser addition (Daly and Hernandez-Ramirez 2020; Roman-Perez and Hernandez-Ramirez 2021; Schleusner et al. 2018; Thilakarathna and Hernandez-Ramirez 2021), highlighting the importance of interactions between the native soil N pool and N fertiliser inputs to N losses. However, these interactions are affected by the presence and type of plants, and studies from different cropping systems are therefore unlikely to be representative of intensive sugarcane systems, where ratooning, a large root system, N fertiliser banding at the beginning of the season and large amounts of irrigation and rainfall create unique conditions for N turnover. Differentiating the contribution of soil and fertiliser N to both N uptake and N2O production in response to N rates is essential to understand the significance of soil N to N turnover with increasing N fertiliser rates towards efficient N management in sugarcane systems.

This study investigated the contribution of fertiliser and soil N and their interaction to plant N uptake and N2O emissions in response to different N rates in two intensive tropical sugarcane systems. Combining high temporal resolution monitoring of N2O emissions with in-situ 15N2O measurements and recovery of fertiliser 15N quantified the contribution of fertiliser and soil N to both plant N uptake and N2O emissions over the growing season under field conditions. The study was undertaken on two different sugarcane farms, allowing the demonstration of the combined effects of fertiliser application methods (two-side banding vs. stool splitting), irrigation management (furrow irrigation vs. overhead sprinkler) and trash management (burnt vs. green cane trash blanketing; GCTB) on sources of N for plant N uptake and N2O emissions. We hypothesised priming of soil organic matter mineralisation by N fertiliser addition, and that mineralised soil N, not fertiliser N dominates plant N uptake and N2O emissions.

Materials and methods

In the current study, we combined additional in situ 15N2O measurements together with high temporal resolution measurements of N2O presented in Takeda et al. (2021a) and the recovery of 15N-labelled fertiliser presented in Takeda et al. (2021b) conducted on a commercial sugarcane farm in Burdekin, QLD (19° 37’ 4’’ S, 147° 20’ 4’’ E) from October 2018 to August 2019 to attribute N2O losses to fertiliser and soil (non-fertiliser) N sources. We conducted a complementary field experiment on a commercial sugarcane farm in Mackay, QLD (21° 14’ 4’’ S, 149° 04’ 6’’ E) from October 2019 to August 2020 and present the results together.

Study site

The climate is tropical in both Burdekin and Mackay and the average annual rainfall is 945 mm (Bureau of Meteorology, Site 033002, Ayr) and 1598 mm (Bureau of Meteorology, Site 033119, Mackay), respectively. The annual average daily minimum and maximum temperatures are 18.0 and 29.2 °C in Burdekin and 19.1 and 26.5 °C in Mackay. The soil is classified as Brown Dermosol and Brown Kandosol in the Australian Soil Classification (Isbell 2016), or Luvisol and Fluvisol in the World Reference Base (WRB) Classification (IUSS Working Group 2014) at the Burdekin and Mackay sites, respectively. Sugarcane varieties Q240 and Q208 were planted in 2015 and 2016 and the crop was the 3rd ratoon during the experiment at the Burdekin and Mackay sites, respectively. Irrigation was applied by furrow irrigation at the Burdekin site and overhead sprinkler at the Mackay site. Sugarcane is burnt before harvest to remove the leaves at the Burdekin site, leaving little trash (crop residues) on the ground. ‘Green cane trash blanketing (GCTB)’, a practice where the cane is harvested green and the trash is spread over the ground, is practised at the Mackay site. Site conditions are summarised in Table 1.

Table 1 Climate conditions, soil properties at 0-0.2 m depth and crop management practices at the Burdekin and Mackay sites
Full size table

Experimental design

The experimental design of the Burdekin site was a randomised strip design with four plots across two strips for each N treatment, detailed in Takeda et al. (2021a). The experiment at the Mackay site had a completely randomised block design with three replicates per treatment (around 10 m in length) and three planting rows per plot with a row distance of 1.56 m, accompanied by an unfertilised control (0N) plot with three subplots (20 m of length). Fertiliser N rate treatments included 0 N, 150 kg N ha-1 (150N), 200 kg N ha-1 (200N) and 250 kg N ha-1 (250N), plus 100 kg N ha-1 (100N) at the Mackay site only. The recommended N rate in SIX EASY STEPS best management practice of the Australian sugar industry (Schroeder et al. 2010) was 150N at the Mackay site and 200N at the Burdekin site. Urea was applied by banding the fertiliser 10 cm deep and 30 cm from the bed centre on both sides of the cane row at the Burdekin site and by stool splitting 10 cm deep at the bed centre of the cane row at the Mackay site. For the 15N recovery in the soil and the plant, a 2.0 m section was exempted from the application of N fertiliser in each plot and 15N enriched urea fertiliser (5 atom%) in solution was manually applied at the corresponding rate (i.e. 100N, 150N, 200N and 250N) matching the N fertiliser placement at the respective site.

Measurement of N2O and CO2 emissions using an automated chamber system

Soil-borne N2O and CO2 emissions were measured at a high temporal resolution using an automated chamber system (Grace et al. 2020) from 17 October 2018 to 15 August 2019 at the Burdekin site and from 3 October 2019 to 24 August 2020 at the Mackay site. Details of the automated chamber system are given in Supplementary materials S1.1. Manual gas sampling was conducted for the control plots of the Mackay site by the static closed chamber method (Friedl et al. 2017), detailed in Supplementary materials S1.2. The placement of the chambers accounted for N fertiliser placement and irrigation practice at each site: At the Burdekin site, chambers were installed covering the area from (a) the fertiliser band to the centre of the bed (bed chamber) and (b) the fertiliser band the centre of the furrow (furrow chamber). At the Mackay site, bed chambers (a) were placed at the centre of the bed (i.e., on the fertiliser band) and furrow chamber measurements (b) were substituted with those from the control plots.

15N2O sampling and analysis in the micro plots

Emissions of 15N2O were measured to estimate the contribution of N fertiliser to N2O emissions, as described by Friedl et al. (2017). Micro plots were established alongside the main plots with N fertiliser rates of 150, 200 and 250 kg N ha-1 at the Burdekin site and with 100, 150, 200 and 250 kg N ha-1 at the Mackay site. The micro plots had a completely randomised block design with four replicates. A steel base (0.22 m × 0.22 m at the Burdekin site and 0.2 m × 0.4 m at the Mackay site) was installed in each micro plot and 15N enriched urea fertiliser (70 atom%) was applied inside the base at the corresponding rates. Gas sampling was conducted with static closed chambers at the Burdekin site from November 2018 to February 2019 and with semi-automated chambers at the Mackay site from October 2019 to January 2020, detailed in Supplementary materials S1.3. The gas samples were analysed for the concentration of N2O using a Shimadzu GC-2014 Gas Chromatograph (Shimadzu, Kyoto, Japan) and for 15N2O using an Isotope Ratio Mass Spectrometer (IRMS) (20–22 Sercon Limited, UK). The contribution of N fertiliser to N2O emissions is detailed in "15N calculations" subsection.

Plant and soil sampling and analyses

Plant and soil samples were taken from each of the 2.0 m sections just before the farmer’s harvest (on 27–28 August 2019 at the Burdekin site and 25–26 August 2020 at the Mackay site). The procedure of plant and soil sampling and analyses at the Burdekin site are detailed in Takeda et al. (2021b) and briefly described here. Aboveground sugarcane biomass, trash on the ground, two green leaves at the 3rd node from the section and the adjacent row and remaining stools and major roots of sugarcane were harvested, as described in the Supplementary materials S1.4. Soil samples were taken at three to four points between the bed and furrow centres using a soil corer and a post-hole driver down to 1.0 m. At the Burdekin site, the sampling points were 0, 0.25, 0.50 and 0.75 m away from the bed centre. At the Mackay site, those were 0, 0.12, 0.40, 0.80 m away from the bed centre in 100N and 250N and the three points except for 0.12 m away from the bed centre in 150N and 200N. Each soil core sample of 1.0 m was separated into 0–0.2, 0.2–0.4, 0.4–0.7, 0.7–1.0 m soil depths and subsamples (about 100 g) were taken. The dried plant and soil samples were then finely ground and analysed for N and 15N content via IRMS analysis (20–22 Sercon Limited, UK). The stalk samples were further dried with a vacuum oven at 40 °C for 48 h before fine grinding to avoid aggregation due to sugar.

15N calculations

Fertiliser 15N recovered in the plant, soil and N2O emissions were then calculated by 15N mass balance (Friedl et al. 2017; Rowlings et al. 2016). The percentage of N in the individual plant, soil and N2O samples (‘sinks’) derived from 15N-labelled fertiliser (Ndff plant or soil) was calculated from.

$$Ndff \left(\%\right)=\frac{{\%}^{15}N\; excess \ of \ sink}{{\%}^{15}N\; excess \ of \ fertiliser}\times 100$$
(1)

where the %15N excess used for all sources and sinks was the 15N abundance less an adjustment of %15N measured for the corresponding plant and soil samples in the 0N plots for background enrichment. Calculations for fertiliser 15N recovered in plant and soil are detailed in Takeda et al. (2021b) and given in Supplementary materials S1.5. Overall fertiliser 15N loss was calculated by the difference between the N applied and fertiliser 15N recovered in the soil and plant.

For the gas samples from the 15N micro plots, the measured %15N excess in N2O was corrected to account for the N2O existing in the headspace at 0 min after chamber closure, detailed in Supplementary materials S1.6. After the correction, the proportion of N2O emissions derived from fertiliser (Ndff N2O) at the fertiliser band was calculated by Eq. [1]. Then, Ndff N2O was gap-filled per N rate over the crop growing season on a daily basis at each site. The gap-filled daily Ndff N2O per N rate from the micro plots was then applied to daily N2O emissions per plot measured by the automated chamber system on the main plots to calculate fertiliser-derived N2O emissions as follows:

$${Fertiliser \ derived \ N}_{2}O \ {emissions}_{i, j}= {area \ weighted \ N}_{2}O \ {emissions}_{i, j}\times \frac{{Ndff \ {N}_{2}O}_{i, j}}{100}$$
(2)

where i and j indicate days after fertilisation and chamber position (i.e. bed or furrow). At the Burdekin site, the adjusted Ndff N2O at the fertiliser band in micro plots was used for both bed and furrow chambers because both chambers covered the fertiliser band and N2O emissions did not differ between the positions. At the Mackay site, adjusted Ndff N2O at the furrow was assumed to be zero because the furrow chamber did not cover the fertiliser band. The rationale behind these assumptions is detailed in Supplementary materials S1.7.

Contribution of soil-derived N to plant N uptake and N2O emissions was calculated by the difference between total N and fertiliser-derived N in each N pool per plot. This contribution of soil N includes fertiliser N from the previous seasons, N in the crop residue and other sources such as N deposition or fixation.

Auxiliary measurements

For soil ammonium (NH4+) and nitrate (NO3−) measurements, soil samples (0–20 cm depth) were taken near the fertiliser band in each plot one day after fertilisation, every 3–7 days for the first three months and then monthly thereafter. Soil NH4+ and NO3− were extracted by adding 100 mL of 2 M KCl to 20 g of air-dried soil and shaking the solution for one hour, followed by NH4+ and NO3− content measurements using a Gallery™ Discrete Analyzer (Thermo Fisher Scientific, USA).

Statistical analysis

Statistical analyses and graphical presentations in this study were conducted using R statistical software version 3.5.2 (R Core Team 2018) with a significant level set at P < 0.05.

Daily N2O emissions were calculated by averaging the measured hourly fluxes over a 24-h period from each chamber and multiplying by 24. Missing daily N2O emissions between measurements were imputed by linear interpolation, using a package “imputeTS” (Moritz and Bartz-Beielstein 2017). Area-weighted N2O emissions were calculated by the ratio bed:furrow = 1:1 at the Burdekin site, while bed:furrow = 1:2 at the Mackay site. Cumulative N2O emissions were then calculated by summing the daily average of each chamber over each growing period. Gap-filling of Ndff N2O values between measurements over the crop growing season was conducted by fitting generalised additive mixed models (GAMMs), using a package “mgcv” (Wood 2011) and detailed in Supplementary materials S1.8. Briefly, GAMMs can quantify non-linear relationships without specifying the functional forms (Rosa et al. 2020; Dorich et al. 2020), which were used to analyse Ndff N2O over time in response to days after fertilisation (DAF) and N rates.

Two-way analysis of variance (ANOVA) was performed to assess significant differences in cumulative (fertiliser and soil-derived) N2O emissions, plant N uptake and fertiliser 15N recoveries between the N rate treatments and the sites. Subsequently, the Tukey-Kramer test was performed when the ANOVA yielded P < 0.05, using a package “agricolae” (de Mendiburu and Yaseen 2020). To evaluate the response of fertiliser 15N loss to N rates across sites, (generalised) linear mixed models were fitted with the site as a random factor, using a package “lme4” (Bates et al. 2015).

Results

N2O emissions and contributions from soil and fertiliser N

The response of cumulative N2O emissions to increasing N rates differed between the two sites (Fig. 1), with larger N2O emissions at the Mackay site compared to the Burdekin site overall across all rates (P < 0.001). Cumulative N2O emissions in the control were 0.29 kg N2O-N ha− 1 at the Burdekin site and 1.50 kg N2O-N ha− 1 at the Mackay site. Fertiliser application increased cumulative N2O emissions (P < 0.001), leading to 1.52–3.06 kg N2O-N ha− 1 and 3.72–4.14 kg N2O-N ha− 1 emitted from the fertilised treatments at Burdekin and Mackay sites respectively. Cumulative N2O emissions exponentially increased with N rates at the Burdekin site (P < 0.001) but did not differ between fertilised treatments at the Mackay site (P = 0.984).

Fig. 1
figure 1

Cumulative N2O emissions at the (a) Burdekin and (b) Mackay sites and fertiliser N contribution to N2O (Ndff) at the fertiliser band at the (c) Burdekin and (d) Mackay sites at 0, 100 (only at the Mackay site), 150, 200 and 250 kg N ha− 1. The lines and error bars in (a) and (b) indicate gap-filled mean observed values and standard errors. The points, lines and shaded areas in (c) and (d) indicate the observed values, the fitted values by the generalised additive mixed model (GAMM) and the confidence intervals of the fitted values, respectively. The columns in (c) and (d) show the amounts of rainfall (grey) and irrigation (light blue)

Full size image

Changes in Ndff N2O at the fertiliser band over the measurement period showed the peak of Ndff N2O occurring at around 25 DAF at the Burdekin site and at 10 DAF at the Mackay site (Fig. 1). Highest Ndff N2O was about 70% at the Burdekin site and 90% at the Mackay site. At both sites, spikes of Ndff N2O recurred and diminished over time. The spikes of Ndff N2O occurred even after 90 DAF at the Burdekin site but diminished by 60 DAF at the Mackay site. The difference in Ndff N2O among N rate treatments was smaller than the difference between the sites. Across the two sites, the period with high Ndff N2O > 30% ocurred within two months after fertilisation.

Fertiliser-derived N2O emissions ranged from 0.39 to 0.78 kg N2O-N ha− 1 at the Burdekin site and 0.97 to 1.34 kg N2O-N ha− 1 at the Mackay site (P < 0.001), correlated positively with increasing N rates (P = 0.043) (Fig. 2). Fertiliser N contributed 22.8–25.40% and 23.7–35.7% to cumulative, seasonal N2O emissions at the Burdekin and Mackay sites, respectively. The proportion of the applied N emitted as N2O calculated by the difference from the unfertilised control was termed apparent N2O EF, which is the standard manner to derive N2O EF. Apparent N2O EF ranged from 0.70 to 1.14% and 0.90–2.41% at the Burdekin and Mackay sites, respectively. The 15N recovery in N2O enabled calculation of 15N N2O EFs, which showed 0.19–0.31% and 0.53–1.02% of the N applied emitted as N2O at the Burdekin and Mackay sites, respectively. Apparent EFs exceeded 15N N2O EFs regardless of site and N rate (Fig. 3).

Fig. 2
figure 2

Cumulative N2O emissions over the crop growing season derived from fertiliser (light blue) and soil (brown) at the (a) Burdekin and (b) Mackay sites at 0, 100 (only at the Mackay site), 150, 200 and 250 kg N ha− 1. The error bars indicate standard errors and the percentage in the columns is the proportion of cumulative total N2O emissions

Full size image
Fig. 3
figure 3

Apparent against 15N N2O emission factor at the Burdekin (blue) and Mackay (red) sites across N rates at 0, 100 (only at the Mackay site), 150, 200 and 250 kg N ha-1. The error bars, solid line and dashed line indicate standard error, 1:1 line and regression line, respectively

Full size image

Soil-derived N2O emissions responded positively to N rates (P < 0.001) and were larger at the Mackay site compared to the Burdekin site (P < 0.001). In fertilised treatments, soil-derived N2O emissions ranged from 1.11 to 2.28 kg N2O-N ha− 1 at the Burdekin site and from 2.41 to 2.88 kg N2O-N ha− 1 at the Mackay site (Fig. 2). Emissions of soil-derived N2O continued > 100 DAF while the majority of fertiliser-derived N2O emissions occurred within two months after fertilisation (Fig. S1).

Plant N uptake and fertiliser 15N recoveries

Plant N uptake ranged from 67.2 to 177.2 kg N ha− 1 at the Burdekin site and 91.8 to 140.6 kg N ha− 1 at the Mackay site, positively correlated with N rates (P < 0.001) (Table 2). Fertiliser 15N uptake increased with increasing N rates (P < 0.001), ranging from 38.8 to 65.0 kg N ha− 1 at the Burdekin site and 19.8 to 45.9 kg N ha− 1 at the Mackay site (Table 2). Fertiliser 15N uptake accounted for 24.2–37.7% and 15.6–32.1% of the plant N uptake at the Burdekin and Mackay sites, respectively, which increased with increasing N rates (P < 0.001). Soil-derived N uptake in the fertilised treatments ranged from 110.9 to 121.7 kg N ha− 1 at the Burdekin site and 85.3 to 107.4 kg N ha− 1 at the Mackay site (Table 2). Fertiliser 15N uptake accounted for 25.9–26.3% and 16.2–20.8% of the N applied (15NUpE) at the Burdekin and Mackay sites respectively (Table 2). Apparent NUpE was 46.8–66.9% at the Burdekin site and 17.6–35.4% at the Mackay site (Table 2).

Table 2 Plant N uptake derived from fertiliser and soil, N uptake efficiency (15N and apparent) and fertiliser 15N remaining in the soil at different N rate treatments at the Burdekin and Mackay sites
Full size table

Fertiliser 15N remaining in the soil ranged from 36.2 to 39.4 kg N ha− 1 at the Burdekin site and 30.5 to 61.1 kg N ha− 1 at the Mackay site at 100–250 kg N ha− 1 (Table 2). The majority of fertiliser 15N in the soil was recovered down to 0.3 m soil depth and within 0.2 m from the bed centre at the Mackay site (Fig. S2). There was a significant difference in fertiliser 15N remaining in the soil between 100N and the higher N rates (200N and 250N) at the Mackay site but no difference among N rates ≥ 150 kg N ha− 1 at both sites. The fraction of applied 15N fertiliser remaining in the soil decreased with increasing N rates (P = 0.017), ranging from 14.5 to 26.3% at the Burdekin site and 24.4 to 36.5% at the Mackay site (Table 2).

Fertiliser 15N loss increased exponentially with increasing N rates (P < 0.001), ranging from 71.8 to 148.8 kg N ha− 1 at the Burdekin site and 49.7 to 143.0 kg N ha− 1 at the Mackay site (Fig. 4). Fertiliser 15N loss accounted for 47.8–59.5% and 47.3–57.2% of the N applied at the Burdekin and Mackay sites, respectively, increasing with increasing N rates (P = 0.023).

Fig. 4
figure 4

Fertiliser 15N loss in response to N rates across the Burdekin and Mackay sites. The points and error bars indicate mean values and standard errors. The line and shaded area show the fitted regression line and 95% confidence intervals

Full size image

Auxiliary measurements

Rainfall during the crop growing season totalled 1180 and 1418 mm at the Burdekin and Mackay sites (Table 1). Furrow irrigation was applied at monthly intervals during the growing season and stopped ten weeks before harvest at the Burdekin site with an estimated total amount of 600 mm. Overhead sprinkler irrigation was applied three times in the early growing season at the Mackay site, totalling 180 mm. Rainfall and irrigation water was allocated more in the early growing season at the Burdekin site, totalling 1321 mm over the first four months after fertilisation compared to 808 mm at the Mackay site (Fig. 1).

Soil NH4+ and NO3− contents increased shortly after fertiliser application and peaked at 30–60 DAF at both sites (Fig. S3). Soil NH4+ and NO3− contents at the peak were 10 and 30 mg N kg− 1 at the Burdekin site and > 200 and > 100 mg N kg− 1 at the Mackay site. Mineral N contents then decreased to below 10 mg N kg− 1 by 100 DAF and stayed for the rest of the growing season across sites.

Emissions of CO2 including root respiration were used as a relative measure for heterotrophic soil respiration enabling comparisons between treatments and sites. Cumulative emissions of CO2 over the season were lower at the Burdekin site than at the Mackay site (P < 0.001), ranging from 4.1 to 5.6 Mg C ha− 1 and 10.2 to 11.2 Mg C ha− 1, respectively (Fig. S3). Cumulative CO2 emissions did not differ between N rate treatments.

Cane yield

Cane yield ranged from 59.6 to 147.3 Mg ha− 1 at the Burdekin site and 89.1 to 135.2 Mg ha− 1 at the Mackay site (Fig. S4). The fitted yield response curves to N rates showed that 95% of the maximum yield could have been achieved at 127 and 108 kg N ha− 1 of N rates at the Burdekin and Mackay sites, respectively.

Discussion

This study provides further evidence that N fertiliser losses are high in intensively managed sugarcane systems with 50% of applied N fertiliser permanently lost by the end of the season even at the recommended N rate. Nitrogen fertiliser losses increased exponentially with increasing N fertiliser inputs. These high losses demonstrate that improved N rate recommendations are required, accounting for the different sources of N available for plant uptake and loss and their interaction. Combining high temporal resolution N2O measurements with 15N recoveries in soil, plant and N2O emissions across different N fertiliser rates at two sugarcane sites demonstrated (i) higher cumulative N2O emissions from the site with overhead irrigation under GCTB management and lower soil pH as compared to the site with furrow irrigation, pre-harvest burning and higher soil pH, (ii) that the majority of fertiliser-derived N2O emissions occurred within two months of fertilisation (iii) that 60% of plant N uptake and N2O emissions were derived from mineralised soil N and (iv) that added fertiliser N increased the contribution of soil N to N2O emissions. Our findings show that the magnitude of total N2O emissions across N rates was not driven by fertiliser N availability itself, but by the response of soil-derived N2O emission to N inputs. The observed dominance of soil N for plant N uptake but also for N2O emissions highlights the importance of soil N mineralisation for N turnover in sugarcane systems. Considering the ratio between soil and fertiliser derived N in plant and N2O, the high N fertiliser losses at both sites imply substantial soil-derived N losses from intensively managed sugarcane systems. Therefore, holistic measures to sustain soil fertility including N fertiliser management strategies are important to ensure crop productivity while controlling N losses.

Response of N2O emissions to N fertiliser rates

Cumulative N2O emissions increased exponentially with increasing N rates at the Burdekin site. Emissions of N2O were generally higher at the Mackay site, and fertilisation increased N2O emissions compared to the control, but unlike in the Burdekin, N2O emissions did not increase with N rates. This difference between sites in cumulative N2O emissions can be attributed to differences in N availability, C availability and soil pH as key drivers of N2O production. The difference in the response to increasing N rates however is likely driven by soil water content: Rainfall and irrigation totals were similar across the two sites yet 75% of rainfall and irrigation occurred in the first four months at the Burdekin site compared to 50% in Mackay. Furrow irrigation (Burdekin) causes soil water contents close to saturation (Takeda et al. 2021a), suggesting that this factor was non-limiting for denitrification at the site, in particular during the first four months when the majority of N2O emissions were observed. By contrast, smaller amounts of irrigation supplied via overhead sprinkler (Mackay) and less rainfall likely limited N2O production in Mackay, explaining why increasing N substrate availability for N2O production with increasing N rates was not reflected in cumulative N2O emissions.

Fertiliser N placement, the diffusion of the fertiliser band in response to irrigation and rainfall and resulting soil NO3− availability also differed between sites: Fertiliser N was banded on each side of the cane row in the Burdekin and exposed to higher amounts of rainfall and irrigation at the beginning of the season as compared to Mackay, where a single band of N fertiliser was placed by stool splitting, and subjected to smaller amounts of rainfall and irrigation. These differences suggest less diffusion of the fertiliser N from the band in Mackay, consistent with the higher soil NO3− concentrations reaching > 100 mg N kg− 1 (Fig. S3) and the larger amount of fertiliser 15N stranded (Fig. S2) near the band at the site. Increasing NO3− concentrations not only increase N substrate availability for N2O production but also limit the reduction of N2O to N2 (Friedl et al. 2020; Senbayram et al. 2012, 2019, 2022), explaining the larger N2O emissions at the Mackay site across N rates.

Differences between sites in N2O emissions from the non-fertilised control point towards drivers of N2O emissions other than N substrate availability. Unlike burnt trash management, GCTB provides C through sugarcane residues, increasing N2O emissions (Warner et al. 2019; Weier et al. 1998) by supplying C to heterotrophic N2O producers or stimulating microbial O2 consumption. Larger cumulative CO2 emissions at the Mackay site compared to the Burdekin site (Fig. S3), highlight the importance of trash management for microbial activity and resulting N2O emissions at the site. Aboveground biomass and stalk yield (Fig. S4) did not differ between sites, indicating negligible differences in root respiration, and therefore allowing the use of CO2 emissions as a relative measure of heterotrophic soil respiration driven by C availability.

The soil pH at the Mackay site is slightly acidic (4.1), compared to the neutral pH of 6.9 at the Burdekin site (Table 1). Similar to high NO3− concentrations, low soil pH inhibits the reduction of N2O to N2 (Dannenmann et al. 2008; Russenes et al. 2016; Šimek and Cooper 2002), promoting high N2O emissions even at lower N rates at the Mackay site. Available data do not allow us to disentangle the overlapping effects of soil water, C and N substrate availability and chemical soil properties on N2O emissions. The comparison between sites however demonstrates that the N response of N2O emissions is highly site and soil specific, highlighting the limitation of its estimation by using a single factor or response curve.

Large fertiliser-derived N2O emissions after fertilisation while cumulative N2O are dominated by soil N

The majority of fertiliser-derived N2O emissions occurred within two months after fertilisation with > 30% of N2O derived from applied N fertiliser (Fig. 2 and S1). Fertiliser N contributed to N2O emissions shortly after fertilisation, despite the high concentration of urea in and around the fertiliser band, which can slow hydrolysis and inhibit nitrification (Janke et al. 2019, 2020). At the beginning of the season, the contribution of N fertiliser to N2O emissions was greater in Mackay, reflecting a single fertiliser band as compared to two-sided banding at the Burdekin site (Fig. 1). After the initial peak, fertiliser N contribution to N2O emissions decreased, denoting decreasing fertiliser N availability due to N immobilisation, plant N uptake and loss. Spikes in Ndff N2O later in the season coincided with rainfall events and irrigation (Fig. 1), indicating remineralisation of N fertiliser N contributing to N2O production. Cumulative fertiliser-derived N2O emissions increased with increasing N rates but accounted for less than half of total N2O emitted across N rates (Fig. 2). Comparable contributions of soil N to cumulative total N2O emissions > 60% were reported in field studies investigating the effects of fertiliser types (Linzmeier et al. 2001; Phillips et al. 2013), soil types (Friedl et al. 2017), slurry application (Lampe et al. 2006) and biochar application (Liao et al. 2021). Our study however demonstrates the dominance of soil-derived N2O emissions across a wide range of N rates for the first time over the whole season under field conditions, combining high temporal resolution N2O flux data with 15N2O analysis.

At both sites, N fertilisation increased the soil-derived N2O emissions compared to the non-fertilised control (Fig. 2). At the Burdekin site, soil-derived N2O emissions also increased with increasing N rates. This increase in the soil-derived N2O has been attributed to priming of organic N mineralisation by added N (Schleusner et al. 2018; Xu et al. 2021), evidenced by increased CO2 emissions in response to added N (Rosa et al. 2018). In our study, cumulative CO2 emissions did not respond to N fertiliser application at both sites (Fig. S3), providing no evidence of increased N mineralisation. Added N fertiliser can also undergo microbial immobilisation instead of the native inorganic N, a process termed pool substitution (Jenkinson et al. 1985; Kuzyakov et al. 2000). This increases native soil N availability for N2O producing microorganisms, explaining the increase in soil-derived N2O emissions. Increased labile C availability at the Mackay site due to GCTB management likely led to greater immobilisation and thus pool substitution, which may explain the larger soil-derived N2O emissions at lower N rates at the Mackay site compared to the Burdekin site (Fig. 2). This assumption is further supported by the greater amount of fertiliser 15N remaining in the soil at the Mackay site (Table 2). Regardless of the underlying mechanisms, our results demonstrate that the response of total N2O emissions to N rates was not driven by fertiliser N availability itself, as shown by the lack of correlation between total cumulative N2O emissions and fertiliser 15N uptake or 15N loss. Instead, the response of soil-derived N2O emissions to N rates largely determined the magnitude of total N2O emissions, highlighting the importance of site and soil specific interaction between soil N and N fertiliser addition providing N substrate for N2O production.

The combination of fertiliser 15N recovery in N2O emissions and high temporal resolution N2O measurements enabled calculation of the EF directly from fertiliser-derived N2O emissions in the N rate treatment. The conventional “apparent” EF method calculates fertiliser-derived N2O emissions as the difference in N2O emissions between fertilised treatments and control and thus includes increases in soil-derived N2O emissions due to interaction with N fertiliser addition as the fertiliser-derived N2O fraction. The comparison between apparent and 15N N2O EFs demonstrates that apparent EFs consistently overestimated the contribution of fertiliser N to N2O emissions, showing an increasing discrepancy with increasing EFs (Fig. 3). This discrepancy emphasises a potential bias in evaluating fertiliser N loss via N2O emissions using the conventional EF method, which in the presence of N fertiliser interaction shows fertiliser induced, but not fertiliser-derived N2O emissions. Fertiliser-induced N2O emissions and thus the apparent EF matter for overall greenhouse gas (GHG) accounting, while the use of 15N N2O analysis can inform mitigation strategies about the N sources of N2O emissions to target.

Exponential increase in fertiliser 15N loss and the potential of soil-derived N losses

Fertiliser 15N loss increased exponentially with increasing N rates and accounted for up to 60% of the N applied (Fig. 4), implying that these losses are driven by an increasing mismatch of N rates and plant N demand (McLellan et al. 2018). However, excess N can also be caused by the spatial and temporal allocation of N fertiliser relative to plant N uptake. Compared to the Burdekin site, more fertiliser 15N remained in the soil and less 15N fertiliser was taken up by the plant at the Mackay site (Table 2). Considering that the majority of residual fertiliser 15N was found near the fertiliser band at harvest at the Burdekin (Takeda et al. 2021b) and Mackay (Fig. S2) site, two-sided fertiliser banding at the Burdekin site potentially led to better spatial alignment between fertiliser N and the root distribution compared to the stool splitting application at the Mackay site. On a temporal scale, the high initial contribution of fertiliser N to N2O emissions observed within two months after fertilisation (Fig. 1) indicates a large contribution of fertiliser N to other loss pathways during this period. This further implies that the majority of fertiliser 15N losses occurred earlier than the active plant N uptake period, which is typically at 2–4 months after planting or ratooning in sugarcane (Zhao et al. 2017). The mismatch between fertiliser N availability and plant N demand in space and time resulted in fertiliser 15N losses of 50% of the applied N even at the recommended N rate (Fig. 4). These findings stress the need to adapt N rate recommendations, and to improve the spatial and temporal allocation of N fertiliser to plant uptake to increase both agronomic and environmental efficiency in sugarcane systems.

Fertiliser 15N uptake increased with increasing N rates but > 60% of plant N uptake was derived from the soil N pool even at the highest N rate (Table 2). Fertiliser 15N remaining in the soil at harvest can be utilised by sugarcane crops in subsequent seasons but is also susceptible to be lost, often resulting in little carryover to subsequent crops (Basanta et al. 2003; Dourado-Neto et al. 2010). Together with the observed large soil N contribution to N2O emissions (Fig. 2), these results denote that N turnover is dominated by N derived from the organic N pool. Assuming N losses mostly occur from the NO3− pool via runoff, leaching and denitrification, the soil N contribution to N2O emissions could be used as a proxy for soil-derived N losses via these pathways relative to fertiliser N loss. The soil N contribution to both plant N uptake and N2O emissions decreased with N rates but stayed at > 60% (Table 2; Fig. 2), indicating that soil-derived N loss possibly exceeds fertiliser 15N loss by far. The exponential response of fertiliser 15N loss with increasing N rates (Fig. 4) as well as the interaction between soil N with N fertiliser addition observed in N2O emissions (Fig. 2) further imply soil-derived N loss can increase with increasing N rates. The findings of this study highlight the potential use of the fraction of soil-derived N2O emissions as an indicator for soil-derived N loss relative to fertiliser 15N loss.

Conclusions

This study monitored N2O emissions at high temporal resolution together with the fate of 15N fertiliser in soil, plant and N2O emissions in response to N rates across two intensively managed sugarcane systems with different farming practices. The majority of fertiliser-derived N2O emissions occurred within two months after fertilisation during a growth stage of low plant N uptake, showing a temporal mismatch between plant N demand and supply. The spatial distribution of fertiliser N may further create a spatial mismatch between banded fertiliser and root N uptake, if fertiliser placement together with limited rainfall and irrigation hinder the diffusion of fertiliser N from the band across the soil profile. Future N rate recommendations should therefore not only consider the non-linear increase of N losses with increasing N rates but also integrate management practices that align N availability with plant demand at temporal and spatial scales. Soil-derived N2O emissions accounted for > 60% of total N2O emissions and thus largely determined the magnitude of N2O emissions, highlighting that N2O abatement strategies need to account for the interaction between soil and fertiliser N inputs. The comparison between apparent and 15N fertiliser recoveries shows that apparent recoveries likely underestimate the contribution of soil-derived N to both plant N uptake and N2O emissions if N fertiliser interacts with the native soil N pool. The dominance of soil N observed in plant N uptake emphasises the importance of mineralisation of organic N for N turnover in sugarcane soils, highlighting the need to maintain soil fertility by measures such as retaining crop residue and crop rotation to ensure crop productivity while controlling N losses. Assuming a similar ratio between soil-derived and fertiliser-derived N in overall N loss from the system as observed in plant N and N2O emissions suggests substantial soil-derived N losses > 200 kg N ha− 1 when applying this ratio to the 15N fertiliser loss. This further implies that the fertiliser N remaining in the soil over the season and the plant residue N returned to the soil cannot compensate the soil N exports via plant N uptake and N losses, highlighting the need to investigate in-field N budgets to avoid long-term, unsustainable mining of soil N. The dataset of this study shows the potential to combine 15N2O analysis with 15N recoveries in soil and plant to derive overall N losses from sugarcane systems, aiding modelling approaches to simulate N budgets in future research. This will help to account for soil mineralisation in N fertiliser management and thus mitigate N loss while maintaining crop productivity in intensive sugarcane systems.