Contribution of the cotton irrigation network to farm nitrous oxide emissions

Additional keywords: furrow irrigation, N₂O emissions, nitrogen use efficiency (NUE), runoff.

In Australia, irrigated cotton is a high-yielding system that requires nitrogen (N) fertiliser inputs to maintain the quality and quantity of yields. Over-fertilisation with N does occur. Comparisons between the internal N-use efficiency (kg lint kg^–1 crop N uptake) from commercial crops and cotton grown under optimum N rates demonstrate that during 2009–12 the industry over-fertilised by ~49 kg N ha^–1 in 2011 (Rochester 2011) and 25 kg N ha^–1 in 2012 (Rochester 2012). Several growers over-fertilised by 80–90 kg N ha^–1 (Rochester et al. 2009; Rochester 2011, 2012). A consequence of the excess N is increased production of the greenhouse gas nitrous oxide (N₂O).

Nitrous oxide is a potent greenhouse gas with a 100-year warming potential 298 times that of carbon dioxide (Butterbach-Bahl et al. 2013). The N₂O molecule is produced as an intermediate compound from two main processes, nitrification and denitrification (Butterbach-Bahl et al. 2013). Rates of N₂O emission are controlled by various environmental factors including soil porosity, temperature, microbial community, pH and availability of mineral N e.g. (Eichner 1990; Bouwman 1994; Butterbach-Bahl et al. 2013). Emissions of N₂O resulting from fertiliser use and manure management comprise 26–35% of total emissions (Syakila and Kroeze 2011), and increased applications of N fertiliser are positively related to direct N₂O emissions (Bouwman 1996; Hinton et al. 2015; Rochester 2003). For cotton, the relationship between N rate and N₂O emissions is exponential, with 1.1–3.5% N applied subsequently emitted as N₂O at N rates of 280–320 kg N ha^–1 (Grace et al. 2016).

Excess N may also be leached (Benjamin et al. 1998) or lost as runoff into the irrigation system (Mchugh et al. 2008). A study of furrow-irrigated cotton in Emerald, Queensland, Australia, with application of 250 kg N ha^–1, showed average N runoff to be 18.8 and 11.3 kg N ha^–1 for 2001–02 and 2002–03, respectively (Mchugh et al. 2008). In furrow-irrigated maize production systems in Iran, nitrate (NO₃^–) runoff ranged from 26 to 70 N ha^–1 after application of 60 kg N ha^–1 (Ebrahimian et al. 2012). Nitrogen species lost via runoff may subsequently undergo denitrification to form N₂O in the water column or drain sediments.

Emissions of N₂O that occur as a result of the transformation of N species lost from the field (e.g. via volatilisation, runoff and leaching) or movement of dissolved N₂O from the field are termed ‘indirect N₂O emissions’ (Reay et al. 2005; IPCC 2006). Indirect emissions are thought to be 29–67% of the magnitude of direct emissions (Reay et al. 2003; Syakila and Kroeze 2011; Outram and Hiscock 2012). Current estimates from the Intergovernmental Panel on Climate Change (IPCC) suggest that for each kg N lost via runoff or leaching, 0.0025 kg N₂O-N may be produced (emission factor, EF5g = 0.0025) but the range of uncertainty is large at 0.0005 to 0.025 (IPCC 2006). Based on the NO₃-N losses reported by Mchugh et al. (2008), we might estimate that ~0.028–0.047 kg N₂O-N ha^–1 would be produced from furrow-irrigated cotton via indirect emissions. Harrison and Matson (2003) have shown with direct measurement that average emissions of 0.04 N₂O-N kg ha^–1 day^–1 can occur within furrow-irrigated wheat production in Mexico.

Indirect N₂O emissions may be a significant component of the total N₂O emissions for Australian cotton systems. However, indirect N₂O fluxes from agriculture have not been measured in Australia. Current estimates for indirect emissions rely on the use of IPCC emission factors with high levels of uncertainty. The aims of this study are to quantify indirect N₂O losses and to compare indirect with direct N₂O emissions, in an Australian furrow-irrigated cotton-farming system.

The research was conducted at the Australian Cotton Research Institute (ACRI) at Narrabri, NSW, Australia (30°19ʹS 149°46ʹE). ACRI is located at the geographic centre of cotton production in Australia. The soil at this site is a high shrink–swell medium grey clay overlying brown clay and is classified as a fine, thermic, montmorillonitic Typic Haplustert (Soil Survey Staff 2010). Cotton is grown at ACRI by using furrow irrigation, and on average, the irrigation network contains water 100 days each year. The irrigation network comprises storage ponds, supply channels, head (supply) and tail ditches for each field, furrows through the field, main tail drains and return channels (which return water to the storage ponds) (Fig. 1). Prior to the irrigation season, water is transferred from the river or groundwater source to supply channels, and then to head ditches. Water is then supplied to the irrigation furrows via siphon from the head ditches. Once the water has transited the field, it empties into the tail ditch and runs off into the main tail drains. The return channel takes the water back to a pump that lifts the water either into the storage ponds or back into the head ditch. The cycling of water around the irrigation network occurs within a 12-h period, and the return water is stored until required for a subsequent irrigation.

Fig. 1.  Location of water samples taken throughout the cotton farm irrigation system of the Australian Cotton Research Institute, Myall Vale, NSW, during the first 3 months of the 2013–14 cotton season. Field water samples were collected at both the tail and head ditches of each field.

Water sampling and measurements were made at ACRI during the 2013–14 Australian cotton season. Samples were taken over 1-week periods in October, November and December 2013. These are the first 3 months of the season and they coincide with land preparation, sowing and fertilising; active growth phase and complete uptake of the fertiliser N by cotton crop; and the final crop irrigation. The blocks, field drains, channels and storage ponds throughout the irrigation network were sampled in triplicate on an ad hoc basis coinciding with the farm irrigation schedule for the sampling period (Fig. 1).

Electrical conductivity (EC), pH and temperature of the water were measured in situ with a WP-81 field meter (TPS, Brendale, Qld). Samples were filtered (0.45 µm) and analysed for NO₃-N, total ammonia N (TAN) and total dissolved N (DTN). Total N (TN) was determined on unfiltered samples; N > 0.45 μm equates to TN – DTN. Nitrate-N and TAN were measured by the cadmium reduction method (Method 4500 Nitrate) and automated phenate method (Method 4500 Ammonia G), respectively (Rice et al. 2012). Samples of TN and DTN were digested by using the persulfate method (Method 4500-N) and the NO₃-N concentration in the digest was measured using the cadmium reduction method (Rice et al. 2012). Dissolved organic N (DON) was determined by subtracting mineral N (NO₃-N and TAN) from TDN. The calculated DON values, except for six samples (mean –0.2, standard error 0.03) were always >0. After sampling, water samples were stored at 4°C, and analysed within 4–7 days of collection. The detection limit of the NO₃-N and TAN analysis was 0.02 mg N L^–1.

Dissolved N₂O (N₂O-N_d) concentrations were determined by using the headspace equilibrium technique (Weiss and Price 1980; Roper et al. 2013). Briefly, during field sample collection, a 6-mL unfiltered water sample was injected into an evacuated 12-mL Exetainer (Labco, Lampeter, UK) and stored at 2−4°C, then returned and analysed in the laboratory within 4–7 days of collection. Harrison and Matson (2003) showed that after 12–48 h there is limited consumption or production of N₂O. However, at the longer storage times used in this study, some consumption or production of N₂O may have occurred but it is expected to be limited because at temperatures <4°C denitrification is limited (Nowicki 1994). Prior to analysis, samples were allowed to warm to room temperature (~25°C) and 10 mL of helium was injected into each Exetainer. The N₂O concentration of the headspace was then measured with a GC-2014 fitted with an electron capture detector (Shimadzu, Kyoto, Japan). The temperature of the laboratory was recorded during sample analysis by an EL-USB-2 data logger (Lascar Electronics, Whiteparish, UK) and used to calculate the N₂O-N_d concentration in the analysed water sample, using the approach of Weiss and Price (1980) and Roper et al. (2013).

To enable the comparison of the direct and indirect emissions, the report fluxes are relative to the source. Thus, the direct emissions are the function of area (ha) of land surface and the indirect emissions are in terms of area of irrigation network.

Direct (or terrestrial) N₂O emissions were calculated from the equation of Macdonald et al. (2015):

where x is the fertiliser rate, and in this case the average rate was 200 kg N ha^–1.

The N₂O flux from the irrigation network surface (13 ha) was calculated using (i) dissolved N₂O concentrations (Cole and Caraco 2001; Clough et al. 2007) and (ii) IPCC emission factors (IPCC 2006).

(i) Dissolved N₂O method. Indirect N₂O fluxes (N₂O-N_df) were estimated from N₂O-N_d concentrations according to Eqn 2:

where N₂O-N_d(water) (μmol m^–3) is the measured concentration of N₂O in the water, N₂O-N_d(eq) (μmol m^–3) is the concentration the water would have if it were in equilibrium with the atmosphere N₂O concentration, and k is the gas transfer coefficient (m s^–1) (Cole and Caraco 2001; Clough et al. 2007).

The gas transfer coefficient, k_total, was calculated as the sum of the transfer velocities attributed to wind (k_wind m s^–1) and water (k_water m s^–1) speed, and these were calculated using Eqns 3 and 4 (Clough et al. 2007; Wanninkhof 1992):

where u₁₀ (m s^–1) is the wind speed at 10 m above the height of the water body; Sc (dimensionless) is the Schmidt number for N₂O; D (m² s^–1) is the temperature- and salinity-dependent diffusion coefficient of N₂O in water; U (m s^–1) is the velocity of water, which was measured using an OTT Flow Meter (OTT, Kempton, Germany); and h (m) is the average depth of the water body. Where water speed was unavailable, k_wind was used instead of k_total.

The wind speed at 10 m height was calculated from measured wind speeds (ACRI weather station) by using the logarithmic wind profile law (Eqn 5):

where Z₀ is the ‘effective roughness height’, here assumed to be 0.001 m; and U₁ and U₂ (m s^–1) are the wind speeds at heights Z₁ and Z₂, respectively (Kubik et al. 2011). Sc and D were calculated in R, using the package ‘marelac’, from measured water salinity and temperature and atmospheric pressure (R Development Core Team 2015; Soetaert et al. 2014).

The average daily N₂O-N_df flux (kg N₂O-N m^–1 ha^–1) was calculated by using Eqn 2. During the irrigation season, the period for which the irrigation network contains tail water is ~15 days.

(ii) IPCC EF₅ method. The default IPCC emission factor for leaching and runoff (EF₅) of 0.0075 has three components: emission factors for groundwater and surface drainage (EF_5g = 0.0025), rivers (EF_5r = 0.0025) and estuaries (EF_5e = 0.0025) (IPCC 2006). Given that water for cotton irrigation usually remains on site, the EF_5g was used to calculate the indirect emissions using Eqn 6:

where (mg L^–1) is the average concentration of TN in the tail and main tail drain water during each of the first four irrigations; v is the volume of water discharging into the field per irrigation (assumed to be 250 000 L ha^–1, which is 25% efficiency of a 100-mm application); n is the number of irrigations within the month; and A is the surface area of the irrigation network (here 13 ha) and B the irrigation area (here 188 ha).

All indirect losses are reported on the water surface area and the direct emissions on the land surface area.

All analyses were performed in R (R Development Core Team 2015). Analyses of variance were used to examine influenced by location and sampling time on the measured parameters (EC, pH, TN, NO₃^–, and N₂O-N_d) using the model: water chemistry = location + sampling time.

Linear regression was used to determine (i) the relationship between N₂O concentration and the other water chemistry parameters, and (ii) the relationship between the two different methods used to calculate indirect N₂O emissions. Where data did not meet assumptions of equal variance, generalised least-squares procedures (in the ‘nlme’ package) were used as an alternative (Hay-Jahans 2011; Pinheiro et al. 2015).

There was a significant effect of location and month on the distribution of EC, pH, TN and NO₃-N of the water sampled (Tables 1–4) Values of EC, TN, NO₃-N, DTN and N > 0.45 μm all increased after the irrigation water transited the field. Conversely, the pH of the water decreased during the transit (Table 1). Throughout the season, the EC and concentrations of the different N species followed a similar pattern, with EC, TN, NO₃-N, DTN and DON peaking during December (Table 2). The water chemistry of the tail water shows that the concentration of N in the DON fractions was often as large as the NO₃-N fraction (Tables 1 and 2). There was a positive correlation between EC and NO₃^– concentration of the discharge water (P < 0.001, r² = 0.51).

Table 1.  Electrical conductivity (EC), pH, and concentrations of various nitrogen species and dissolved N₂O (mean ± standard error) in water samples collected at different locations throughout a cotton farm irrigation system at the Australian Cotton Research Institute, Myall Vale, NSW, during the 2013–14 cotton season
Parameters EC, pH, total N, NO₃-N, dissolved total N (DTN) and dissolved organic N (DON) showed significant differences between locations of sampling. No analysis was undertaken for NH₃

Table 2.  Electrical conductivity (EC), pH, and concentration of various nitrogen species and dissolved N₂O (mean ± standard error) in water samples collected at different sampling times across the cotton farm irrigation system at the Australian Cotton Research Institute, Myall Vale, NSW, during the 2013–14 cotton season
DTN, Dissolved total N; DON, dissolved organic N. All parameters showed significant differences between times of sampling, except NH₃-N for which no analysis was undertaken

Table 3.  Pair-wise comparisons, using Tukey’s HSD, for water chemistry components between each of the irrigation network locations sampled at Australian Cotton Research Institute, Myall Vale, NSW
EC, Electrical conductivity; TN, total nitrogen; DTN, dissolved total N; DON, dissolved organic N; N45, N > 0.45 µm. Components shown are significantly different between each pair of locations. DTN measurements were not available for the main tail drain

Table 4.  Pair-wise comparisons, using Tukey HSD, for water chemistry components between each of the different sampling times at Australian Cotton Research Institute, Myall Vale, NSW
EC, Electrical conductivity; TN, total nitrogen; DTN, dissolved total N; DON, dissolved organic N; N45, N > 0.45 µm. Components shown are significantly different between each pair of sampling times. DTN measurements were available only for November 2013 and December 2013 sampling events

The N₂O-N_d concentrations followed a similar pattern to that of the other N species with time of different sampling and locations, but concentrations were highly variable. Differences in N₂O-N_d concentration due to location were not significant (Tables 1 and 3) despite N₂O-N_d concentrations tending to increase in the tail ditch and the main tail drain. Average concentrations of N₂O-N_d ranged from 0.395 ± 0.045 µg L^–1 (supply channel) to 2.15 ± 1.34 µg L^–1 (tail drain) in the irrigation network for the 3 months of measurement.

Dissolved N₂O method. The cumulative N₂O-N loss from the irrigation water surface during the first four irrigations between October and December 2013 was 0.503 ± 0.338 kg ha^–1 (Table 5).

Table 5.  Total indirect N₂O-N emissions (kg ha^–1 irrigation water surface, 13 ha) from Australian Cotton Research Institute, Myall Vale, NSW
Estimates based on average concentrations of total nitrogen and dissolved N₂O in runoff water in tail and main tail drains only (mean ± standard error)

IPCC EF₅ method. Average total N concentrations for water sourced from tail ditches and main tail drains over the four irrigations was 28.96 ± 6.903 mg L^–1. This corresponded to a cumulative N₂O-N emission of 0.843 ± 0.022 kg ha^–1 from the irrigation water surface, representing a field leaching loss of 23.31 ± 0.61 kg N ha^–1 during the first four irrigations between October and December 2013 (Table 5).

There was a strong, positive linear relationship between monthly N₂O fluxes calculated by using the IPCC EF_5g and dissolved N₂O methods (P < 0.05, R² = 0.99). However, the disparity between the two methods increased with higher N₂O emissions, and total N₂O emissions estimated using the IPCC method were 46% higher than under the dissolved method.

During the cotton season, direct emissions of N₂O-N from the land surface were, on average, 16 g N₂O-N ha^–1 day^–1 (Macdonald et al. 2015). The cumulative direct N₂O-N emission off the entire cotton farm over the season (150 days) was 2.42 kg N₂O-N ha^–1. During the period of indirect measurements (90 days), the direct N₂O-N emissions off the cotton farm were 1.45 kg ha^–1.

The water chemistry of the irrigation water was modified during its transit through the cotton field (Table 1). Nitrate and DON were the main components of TN present in the irrigation water, and both N species were lost from the cotton field (Table 1). The measured NO₃-N concentrations are similar to those from studies within the Australian cotton industry (Mchugh et al. 2008; Weaver et al. 2013) and other irrigated cropping systems (Harrison et al. 2005). Salt and other nutrients accumulate as a result of evaporation from the furrow surface (Noborio et al. 1996) and are remobilised during the first flush at the beginning of an irrigation. Irrigated furrows were less saline than non-irrigated furrows, suggesting that movement of water from irrigated to skip furrows transits through the adjacent hill, removing salts and N, which are then lost via runoff (Fig. 2). The differences in the N concentrations between the sampling events (Table 2) are likely due to the mineralisation or organic N within the hill releasing ammonium and NO₃-N, which can be mobilised by the irrigation water.

Fig. 2.  A schematic showing the movement of water without (white fill arrows) and with (black fill arrows) fertiliser nitrogen compounds from the hill to the skip furrow. TAN, Total ammonia N; DON, dissolved organic N.

Further, we observed significant variation in the water N concentration during irrigation and between irrigations. The soil physical and moisture characteristics also vary within each furrow and mound, and as a result, the irrigation water and dissolved N compounds will transit through the soil at different rates. It is evident from the measured concentrations that the flux of the DON pool must be as important as the NO₃-N in the measured furrow-irrigated system (Tables 1 and 2). The DON is being sourced from the mound as the water passes through from the irrigated furrow to the skip furrow. DON, like NO₃-N, can undergo transformation and conversion into N₂O-N_d in the water column and on the sediment surfaces (Nevison 2000; Tiedje et al. 1982). All N species lost into the irrigation system can potentially undergo subsequent transformations to form N₂O-N_d within the water column and drain sediments (Nevison 2000; Harrison and Matson 2003).

The N₂O-N_d in irrigation water may be sourced from N₂O produced within the field, or from subsequent denitrification or nitrification reactions in the water column. Irrigated cotton fields provide optimal conditions for denitrification, including microbial available carbon, nitrate and anaerobic environment. The much lower N₂O-N_d concentrations in our study than in other studies (Harrison et al. 2005; Outram and Hiscock 2012) are likely due to fields being irrigated when there is a 75-mm water deficit in the soil profile. At such water contents, the formation of N₂O in the surface soil would be negligible (Weier et al. 1993; Davidson et al. 2000), and the measured terrestrial atmospheric flux rates in cotton systems at these soil moistures are small relative to those after emissions that occur when irrigation has ceased (Mahmood et al. 2008; Scheer et al. 2013). There is no pool of N₂O-N to move from the soil during the irrigation, and typically denitrification and N₂O-N emissions occur 1–2 days after the irrigation has ceased. Further, our site is in a semi-arid, irrigated cropping region of Australia, whereas many of the other indirect emissions studies were conducted in areas of higher rainfall (Outram and Hiscock 2012; Risk et al. 2013; Kaushal et al. 2014), which are more conducive to shallow groundwater fluxes of N₂O-N_d.

There was no relationship between N₂O-N_d and the other N components. This is in contrast to several studies that have demonstrated a relationship between N₂O-N_d and NO₃-N concentrations (Harrison and Matson 2003; Reay et al. 2005; Beaulieu et al. 2009, 2011; Warneke et al. 2011) or NH₄-N concentrations (Xia et al. 2013). Water in the cotton irrigation system is transient, only retained for a short period, due to the cessation of irrigation once mounds are ‘wet up’ and short field lengths (<500 m). There is also no lateral groundwater discharge into the canals.

Runoff from the cotton field at the ACRI experimental farm is negligible within 12 h of the start of the irrigation; however, on commercial farms, field irrigations occur over longer periods, due to the field length exceeding 1000 m. An increase in contact time between water and the soil surface could maximise N₂O-N_d production from TN in the irrigation tail water. Further indirect N₂O emissions, resulting from N loading in the irrigation water, may continue downstream (e.g. in storage ponds) and as the irrigation networks dry down, neither of which were measured in this study.

There was a strong positive relationship between the two methods used to calculated monthly N₂O flux. Although both methods gave estimates of N₂O emissions within the same order of magnitude, the IPCC method returned an emission rate 65% higher than that calculated by the dissolved N₂O method. Differences between the dissolved N₂O and IPCC methods may have occurred through the dissolved N₂O method underestimating amounts of N₂O-N produced from the water surface. Alternatively, there are uncertainties associated with the current IPCC EFs for indirect emissions. Although the current EFs have been reduced from previous estimates owing to large discrepancies between measured and IPCC estimated fluxes (Nevison 2000; Reay et al. 2005; Clough et al. 2007), the range of uncertainty for EF₅ is still large, from 0.0005 to 0.025 (IPCC 2006).

European measurements form much of the basis for the IPCC EFs (Reay et al. 2012). Use of local emission factors, or models that account for local climatic conditions, soil characteristics and land management, will then reduce the uncertainty in flux estimates (Reay et al. 2012). A definite need exists to quantify and understand better the processes controlling indirect N₂O emissions within the Australian cotton industry. This, in turn, will provide a better platform for policy decisions and discussions of potential mitigation strategies.

The irrigation network area on a typical irrigated cotton farm may represent only 6.5% of the farm area. Despite having a flux rate per hectare comparable to that of the direct land-surface emissions, at the farm scale, the indirect emissions are a minor component of the N₂O inventory. The indirect emissions estimated by both methods, from the whole farm, were ~2.4–4% of the magnitude of direct land-surface emissions and <0.02% of the fertiliser applied (260 kg ha^–1) to the farm. These are similar to the values reported by Harrison et al. (2005) for furrow-irrigated wheat production in Mexico.

A key issue in the determination of indirect emissions on a per-hectare basis is the accurate quantification of the fate of the water within the irrigation network. The tail water in semi-arid, irrigated cotton systems is typically recirculated and re-used on-farm. Tail water is returned to the farm storage, stored briefly (24 h), and mixed with river or groundwater and returned to the fields for the next field irrigation. The duration and location of the tail water storage will change during each day of the irrigation period depending on farm watering requirements. We have assumed that the indirect emission is mainly sourced from the tail water because of the N loading from the field and is equally spread across the irrigation network.

In this study, point measurements of N components and N₂O occurred only during the period of an irrigation (<12 h), and they were concentrated at the cotton field. During this time, N that has been leached from the fields may be transformed into N₂O, NO_x or N₂ in the storage ponds and either emitted or assimilated in the water column or in the drain sediments. In addition, although many studies of indirect emissions have focused on the emissions of N₂O from the water surface, there may be significant N₂O emissions from the sediments once the canals are drained. Sediments can sequester NO₃-N from the water column (García-García and Gómez 2009), which can lead to significantly higher indirect emissions if irrigation water is allowed to pond and is not re-used. The complexity of these biogeochemical pathways would explain the large uncertainties associated with estimating N₂O emissions by using N₂O-N_d concentrations.

The key to reducing indirect N₂O-N emissions from cotton irrigation networks is to control the N supply to the irrigation water. Improvements in the efficiency of water and N use would reduce the export of total N and, hence, lower the potential for indirect and N₂ emissions.

Indirect N₂O-N emissions may be reduced by maximising the use of plant-available N already present in the water. The tail water contains large amounts of dissolved N, which could be used to fertilise adjacent fields. Reducing water return-times to the field is likely to increase the amounts of N that can be re-used; however, a better understanding is required of the rates of transformation for optimisation of N recycling in the cotton irrigation network.

Estimates of N₂O emissions from the surface waters of a cotton irrigation network are now possible. The concentrations of N₂O-N_d and N₂O-N_EF5g are 0.503 ± 0.338 and 0.843 ± 0.022 kg ha^–1 irrigation surface, respectively, over 90 days. Overall, the indirect emissions from the surface of the irrigation network are not a significant component of the N₂O inventory for Australian cotton systems, because the irrigation network covers only a small area relative to the entire land surface of the farm. The measurement of indirect emissions from irrigated cotton production is not straightforward, owing to the ad hoc re-use and storage of water. Additional N₂O emissions are likely to occur downstream of the field within storages and main tail drains during the irrigation season and as the channels dry down. Nitrogen fertilisation due to the re-use of drainage water and subsequent field N₂O-N emission could also contribute to indirect emissions. Overall, the IPCC EF_5g and the dissolved N₂O indirect flux estimation methods were in agreement, and the EF_5g could be used to estimate indirect fluxes provided local calibration was undertaken. The irrigation network is a prime mitigation target for minimising losses of dissolved N components via denitrification. Rapid re-use of N-enriched tail water, reducing N loss via runoff, and improving efficiency of water and N use are potential methods to reduce N losses.