Fate and cost effectiveness of soil carbon sequestered using supplementary nutrients applied to crop residues under field conditions

Changes in carbon (C) and nitrogen (N) associated with fine fraction (FF) soil organic matter (SOM) were determined to a depth of 1.80 m at a field site that was managed with supplementary nutrients (2007–2012) to sequester soil C from crop residues. Soil C and N stocks were monitored for a further 3 years (to 2015) with supplementary nutrients maintained, and then for 5 years after they were ceased (2015–2020). The increase in FF-C established in the first phase of nutrient addition (9.6 Mg C ha−1) was maintained during the ongoing 3 year period of nutrient addition (9.8 Mg C ha−1) but declined to 3.0 Mg C ha−1 during the subsequent 5 years when nutrient addition ceased. The decline in FF-C stocks was attributed to both less formation of FF-C without the supplementary nutrients applied, and to nutrient mining as insufficient N was being applied to both treatments to fully satisfy the N removed in crops. The addition of supplementary nutrients to the residue influenced crop yield in some years, and seed protein in most years, but the responses were crop and season dependant and there was no overall difference in the amount of C in crop biomass returned to the soil. An economic analysis using long-term fertiliser and grain prices suggested that the strategy to sequester more C in FF-SOM through nutrient addition to residue was essentially cost neutral. Better tailoring of nutrient inputs to achieve the required stoichiometric ratios in SOM, and an assumed payment for the sequestered C (AU$40 Mg−1 CO2 equivalent) provided a reliable economic benefit during the period of nutrient addition. However the economic benefit was lost when nutrient addition ceased, and would reduce if fertiliser prices increased or the C price declined. Our results confirm that addition of supplementary nutrients to C-rich crop residues is an effective means to sequester soil C that can persist for several years even after nutrient addition ceases. However, ongoing nutrient inputs in excess of crop removal are required to ensure long-term sequestration of the C from crop residue.


Introduction
Soil organic matter (SOM) is a key component of soil fertility and has an important role in soil functionality including structure, water retention, nutrient cycling and wider environmental benefits of carbon (C) sequestration for climate change mitigation (Lal 2004;Minasny et al. 2017). It is estimated that anthropological land-use and agricultural intensification has led to a loss of 133 Pg C from the top 2 m of the soil globally, with the highest rate of loss occurring in the past 200 years (Davidson and Ackerman 1993;Sanderman et al. 2017;Swift 2001). For example, in Australian soils some 50% of the original stocks of soil organic C (SOC) have been lost from SOM in recent decades in intensively managed cropping systems (Chan and McCoy 2010). This loss of SOC is associated with similar declines in levels of soil N, where more than half of the organic N associated with SOM has been mineralized and essentially 'mined' from the soil (Angus and Grace 2017). Declining areas of legume-based pastures and limited growth in the total area of grain legumes has exacerbated the loss of soil C and N, and the associated 'soil fertility decline' (i.e., loss of SOM). Inadequate fertiliser N is currently applied to replace the previous contribution of legumes (Angus and Grace 2017).
Rebuilding 'lost' soil C in SOM to mitigate climate change and provide other ecosystem benefits is a high priority for agricultural policy makers and practitioners both in Australia and across the globe (Amudson and Biadeau 2018;Richardson et al. 2019). For example, the Intergovernmental Panel on Climate Change identified 'Agriculture' as having one of the most significant near-term (by 2030) greenhouse gas mitigation potentials, with 90% of the opportunity arising from increased soil C sequestration (Smith et al. 2007(Smith et al. , 2014. While soils are the largest sink for terrestrial C, their capacity to maintain C stocks is mediated by management practices that can accelerate the loss of SOM or limit its formation and retention (Hunt et al. 2020). Various strategies to restore or maintain soil C have been proposed which include; adoption of conservation practices that aim to increase the return of crop residues (and other biosolids such as manures, composts and other wastes) to soil, reducing soil disturbance and maintaining continuous ground cover, enhanced biodiversity through use of more diverse crop rotations especially with legumes, deep ploughing whereby topsoil is transferred into subsoil for SOM conservation (Alcantara et al 2016;Nieder and Richter 2000) and a strengthening of nutrient recycling with a focus on positive nutrient balance (Lal et al. 2007;Giller et al. 2015).
Over the past few decades there has been a considerable increase in the adoption of conservation farming practices across Australia and globally which has resulted in an increase in return of crop residues to soil (primarily on the soil surface) along with reduced cultivation (Llewellyn et al. 2019). Despite this, there is little evidence to show that such practice leads to an increase in C sequestration, and many studies report no effect on C stocks, even when practiced for a decade or longer (Campbell et al. 1991;Soon 1998;Rumpel 2008). In a review of long-term experiments across the globe, Powlson et al. (2011) concluded that whilst many studies 'suggest' that greater SOC accumulation occurs with residue retention, differences were most often not statistically significant. Furthermore, while retention of crop residues can lead to greater SOM levels in surface soil layers, there may be no effect on total soil C stocks with depth due to stratification within the soil profile and often lower levels of SOC at depth (Baker et al. 2007).
Incorporation of residues into soil through cultivation has been shown to increase SOC more consistently throughout the entire soil profile, as the conversion of stubble-C into microbial biomass and subsequently into SOM is facilitated by soil mixing as compared to surface retention (Alcantara 2016;Helgason et al. 2014). In particular, direct application of nutrients to crop residues has been shown to enhance the formation of SOC in both laboratory and fieldbased experiments (Moran et al. 2005;van Groenigen et al. 2006;Poeplau et al. 2017;Kirkby et al. 2013). Jacinthe et al. (2002) showed that humification efficiency (i.e., conversion of wheat residue-C to SOC) was increased from 14 to 32% in field soil over a 4 year period with nutrient addition. In our previous work, Kirkby et al. (2016) similarly demonstrated an increase in soil C stocks that were associated with fine fraction SOM (FF-SOM) to a depth of 1.60 m over a 5 year period. The fine fraction C (FF-C) in the soil increased by 5.5 Mg C ha −1 (equivalent to a 9.5% increase in soil C) when supplementary nutrients (as NPS) were added with the residue (primarily as wheat stubble with average input of 9 Mg ha −1 year −1 ), as compared to a loss of 3.2 Mg ha −1 where residue with no nutrients was incorporated (Kirkby et al. 2016). On the contrary, a net loss of SOC following incorporation of C-rich (but nutrient poor) crop residues has been reported, where a loss of C from 'old' SOM through mineralisation can exceed the formation of 'new' SOM-C generated from freshly added plant materials (Fontaine et al. 2004;Kirkby et al. 2013). In such scenarios, soil microorganisms utilize C from the fresh inputs for energy and mineralise pre-existing SOM to obtain nutrients (particularly N, P and S) required for microbial growth (i.e., priming effect). This is especially evident when C-rich amendments are returned to soil resulting in net mineralization or 'mining' of soil nutrients from SOM Blagodatskaya and Kuzyakov 2008;Guenet et al. 2010a;2010b). In our previous study (Kirkby et al. 2016) supplementary nutrients (NPS) were added to crop residues prior to incorporation into the soil at levels to balance the stoichiometry of the crop residues to more closely match the CNPS stoichiometry of the FF-SOM, which is relatively constant and more aligned to that in the microbial biomass . Significantly, and despite cultivation to incorporate the residues and nutrients to a depth of ~ 0.15 m annually, some 50% of the increase in SOC in response to nutrients occurred below 0.3 m. This suggested either the leaching of C from the surface layers as either soluble C or colloidal material including microbial detritus, or as a direct result of increased root growth and exudation at depth (Kirkby et al. 2016).
Despite evidence for the opportunity to build C from crop residues using targeted nutrient supplementation, there remains considerable uncertainty about the longer-term viability of such an approach under field conditions. We extended the long-term annually cropped field experiment reported in the study by Kirkby et al. (2016) where a 8.7 Mg ha −1 difference in FF-C to 1.6 m was reported over 5 years by adding supplementary nutrients to incorporated crop residue after harvest. In the current study, supplementary nutrient addition to crop resides was maintained for a further 3 years (2012-2014) using the same principles of nutrient supplementation and were then ceased for the subsequent 5 years (2015-2020) under an ongoing annual cropping cycle. Our aims were to: (i) determine the stability (longevity) and fate of the 'new' FF-SOM produced in this way, (ii) determine the cost effectiveness of the strategy by comparing the nutrient costs with the value from the increased productivity or through direct payments for sequestered C in SOM. By extending and then ceasing supplementary nutrient addition to incorporated crop residue and monitoring soil C and N profiles and crop growth and yield over the 8-year period, the fate and cost effectivness of the supplementary nutrient strategy could be assessed.

Site and experimental design
The study was conducted at the CSIRO long-term experimental site on a commercial farm ('Oxton Park') near Harden, in southern NSW, Australia (34° 30′ S, 148° 17′ E). The soil was a Red Chromosol (Isbell 2002) as described in Table S1 and considered as an Alfisol based on USDA Soil Taxonomy (Soil Survey Staff 1999). The site was established in 1990 and managed as a long-term field experiment to investigate the effect of different tillage and crop residue treatments (seven management practices) on soil conditions and crop productivity in an annual, wintercropped farming system (reported in Kirkegaard et al. 1994Kirkegaard et al. , 2020. The treatments were replicated four times in a randomized block design. Individual treatment plots (30 × 6 m) comprised two paired subplots (30 × 2 m), side by side, separated by a central 1 m buffer to allow controlled-traffic management (no wheel traffic on plots).
The experiment reported here was conducted as part of a series of different phases based on an annual cycle of cropping from 1990 to 2020 (Table 1). The experiment utilised only one of the seven long-term treatments, the residue incorporation treatment (RI), in which crop residues were incorporated into the soil with an offset disc harrow to a depth of 0.15 m after the first rain following the annual summer harvest in December/January in each season (Phase 1;1990-2006. The study follows a previous study on the same treatment plots conducted from 2007 to 2012 (Phase 2; Kirkby et al. 2016) which investigated the effects of adding supplementary nutrients to the crop residue at the time of post-harvest incorporation, on the amount of soil C sequestered from the residue ( Table 1). For that study, one of each of the paired sub-plots (30 × 2 m) in the incorporated treatment received supplementary nutrient addition at the time of residue incorporation each year from 2007 to 2012, while the other sub-plot received no supplementary nutrient addition (Kirkby et al. 2016). In the current experiment we maintained the same experimental protocol of nutrient addition for a further 3 cropping years (i.e. added nutrients to the residues of the harvested 2012, 2013 and 2014 crops) (

Seasonal conditions
The period encompassing the application and then cessation of supplementary nutrients onto residues (Phase 2-4, 2006-2019) was generally characterised by below-average growing season rainfall (mid-May to November) which equalled or exceeded the longterm mean (364 mm) in only 3 of the 14 years (2010, 2015 and 2016) (Fig. S1). Spring rainfall (September and October) which has a significant impact on crop yield potential due to its coincidence with crop reproductive phases, was especially low in many seasons. Crops producing high early biomass early in the season may suffer significant yield penalties when spring conditions are dry in the process known as "hayingoff" whereby soil water is exhausted by the vigorous early growth and crops suffer significant stress during the grain filling stage. Risks of "haying-off" are often cited by growers as a reason for conservative application of N fertiliser. In contrast, the rainfall in the periods following residue incorporation until the time of sowing (approximately February to mid-May) was generally near to, or above average, except for two years (2013 and 2018). Together with the warm summer temperatures (mean minimum 10 °C, mean maximum 28 °C) this provided good conditions for the decomposition of the incorporated residues by soil microorganisms prior to sowing in most years.

Residue and nutrient addition treatments
From 2007 to 2014 (Phases 2 and 3), the amount of residue was carefully balanced across all plots after harvest, and prior to incorporation, to ensure that during the period of nutrient addition the amount of residue C returned was identical (~ 9 Mg of stubble residue ha −1 y −1 ), despite the potential differences in crop biomass generated by the treatments. If necessary to balance the treatments, additional residue was sourced from buffers adjacent to the experimental plots. As soon as practicable after harvest (usually late January), the residue was mulched to produce pieces < 0.15 m long using a mechanical plot flail mulcher. Supplementary nutrients where then broadcast onto the surface of the mulched residue for the 'plus nutrient' sub-plots. Following the first significant rains after residue mulching and supplementary nutrient addition (approximately 10 mm or more in one day to soften the soil, which usually occurred in mid-February), the residue was incorporated to a depth of ~ 0.15 m using two passes of a rotary cultivator to ensure that the soil and residue were thoroughly mixed with little residue apparent at the soil surface. Thorough mixing provides maximum contact between the soil, residue and nutrients to enhance microbial activity. In the period from 2015 to 2020 (Phase 4), the amount of residue was not balanced across the plots and the residues were incorporated in both sub-plots as previously, but with no supplementary nutrient addition. For the years where the supplementary nutrient treatments were applied (2007 to 2014), the amount of inorganic nutrients (NPS) added to the residue was calculated as that required to humify 30% of the measured above-ground stubble-C, according to the stoichiometric N requirements of the FF-SOM; C:N:P:S = 10,000:833:200:143 (Kirkby et al. 2013). The residue nutrient analyses from the four replicates varied slightly each year but to ensure there were sufficient nutrients, the analysis of the most nutrient poor residue was used for the calculation and the same amount of nutrients were added to each of the four replicates (Supplementary Information Table S2; Kirkby et al. (2016)). A commercially available granulated fertiliser (Granulok 15, Incitec Pivot Fertilisers; N, P, S = 14.3, 12.0 10.5%, respectively) was used to supply the supplementary nutrients. The granular fertiliser was pulverised (Labtechnics pulverising mill, model LM1, Adelaide, Australia) to produce a powder to enable a more even spread over the mulched residue. The amount of fertiliser added per unit of residue varied slightly each year depending on the N, P and S levels measured in the residue, but was added at an approximate rate of 22 kg per Mg of crop residue. As N was the element needed in the greatest amount for the humification process (833 units per 10,000 units of stubble C), the amount of fertiliser added was calculated according to stoichiometric N requirements. On average 5.3, 1.8 and 0.8 kg of N, P and S respectively was required per Mg of crop residue. Using Granulok 15 to supply 5.3 kg N also supplied 4.4 and 3.9 kg P and S respectively, meaning there was a considerable excess of P and S added to the residue. A consideration of the savings possible in P and S supply with a fertiliser source having more appropriate ratios is included in the economic analysis as outlined below. The full details of the fertiliser applied to the residues post-harvest, and those applied to the subsequent crops at sowing and as top-dressings during the season are provided in Table S3.

Crop management
The plots were cropped annually (May-December) in a sequence involving wheat (Triticum aestivum) canola (Brassica napus) or lupin (Lupinus angustifolius) and managed according to local agronomic recommendations. This generally involved knockdown herbicides to control weeds during the summer fallow period (December-April) and crops sown each year of the experiment in autumn. Fertiliser (NPS) was applied as a seed dressing and N was top-dressed as urea after stem elongation to supply sufficient N to achieve the predicted yield in each season according to local agronomic advice. The full history of fertiliser application to both residue (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014) and to the crops (2007-2019) is shown in Table S3.

Soil sample collection, preparation and analyses
The protocols for soil sampling and analysis reported in Kirkby et al. (2016) for the 2007 and 2012 years were repeated in 2015 (the final year of nutrient addition), and again in 2020 (after 5 years without nutrient addition to the crop residue). Soil cores were taken on each occasion prior to sowing in March 2015 (end of Phase 3) and March 2020 (end of Phase 4). Briefly, at each sampling six soil cores (43 mm diameter), three in each sub-plot were extracted using a tractor-mounted hydraulic corer to a depth of 1.8 m from each replicate, and carefully separated into 0.1 m increments to 0.3 m depth, and 0.3 m increments to 1.8 m depth. The three core segments were bulked for each depth and subplot for analysis and bulk density estimates. To estimate nutrient loads we used the measured bulk density for each treatment in 0-0.1 m and 0.1-0.2 m layers, while below that depth where little change was seen between treatments, a mean bulk density value was used for each of the four treatment blocks from cores taken on all three dates.
The soils were prepared for analysis using a dry sieving/winnowing procedure described in detail in Kirkby et al. (2011). Briefly, the soils were air dried and gently crushed to separate soil, any plant material and gravel and were then passed through a 2 mm sieve to separate the soil from the gravel and any large pieces of plant material. The gravel was weighed and discarded, along with any coarse plant material. Any coarse (> 0.4 mm) or light fractionorganic material that had passed through the 2 mm sieve was subsequently removed from the soil using the dry sieving/winnowing procedure as described in Kirkby et al. (2011). Organic matter remaining in the soil following this procedure (and the associated C) was designated to be fine-fraction SOM (FF-SOM, < 0.4 mm) that was considered to contribute to the more stable, slowly decomposing pool of SOM (Kirkby et al. 2011;Magid and Kjaergaard 2001. Stable SOM is most commonly defined as the 'humic fraction' based on density (i.e., > 1.3 Mg m −3 ) and a size fraction of less than 53 µm (e.g., Buyanovsky et al. 1994;Curtin et al. 2019). Our definition of FF-C is consistent with that reported in our previous fieldbased study (Kirkby et al. 2016) and on the < 0.4 mm fraction as reported by Magid and Kjaergaard (2001). Importantly they compared density fractionation with size-based fractionation and showed a close relationship between the POM > 0.4 mm and < 1.4 Mg m −3 light-fraction POM, indicating the suitability of using 0.4 mm as an appropriate cut-off for the measure of a more stable SOM pool. This was evident with regard to C and N concentrations of the fractions, between changes in the C content of the POM fraction and the easily respired CO 2 and the appearance of the fractions under the microscope (Magid and Kjaergaard 2001). More recently, Coonan et al. (2020a) demonstrated consistency in the measure of the < 0.4 mm fraction defined by Magid and Kjaergaard (2001) with the Kirkby et al. (2011) procedure. A 100 g subsample of the soil was thus subsequently pulverised (Labtechnics pulverising mill, model LM1, Adelaide, Australia) to a soil powder for later chemical analyses. Total C and N concentrations were determined using a dry-combustion analyser (Europa Scientific Model 20-20, Crewe, UK). Stocks of FF-C and FF-N were calculated from the measured FF-C and FF-N concentration values after adjusting for the measured gravel content and bulk density.

Plant sampling and nutrient analysis.
Plant densities were measured by counting established seedlings on both sides of a 0.5 m stick positioned randomly in each plot. Above-ground crop biomass was measured at the start of stem elongation and at flowering by cutting and oven-drying plant material removed at ground level from two 0.4 m 2 bordered quadrats. At harvest, the total crop biomass was measured using two quadrat cuts of plant material removed at ground level and oven dried. The biomass was separated into grain and residue to estimate harvest index and the amount of aboveground residue remaining after harvest. Immediately following the harvest quadrat cuts the whole plot was harvested using a mechanical plot harvester, during which smaller (non-commercial) grain was discarded from the header, so the header yield is consequently lower than the hand-harvested quadrat yield. The C and N content of the oven-dried plant material was determined using the same methodology as described above for the soil (dry combustion analyser). The measurements of final plant biomass and residue and grain N concentrations were used to develop N budgets at the site and to allow mass balance for nutrients to be calculated across the experiment, and to provide yield data for the economic analysis.

Nitrogen mass balance
Although there was no attempt to measure leaching or gaseous N losses, a calculation was made of the N mass balance to assess whether it was likely that there were any large N leaching losses, especially from the (+) nutrient treatment. Fertiliser additions (be it for normal crop requirements or supplementary nutrients added to increase the crop residue humification efficiency), N added as the extra straw to both treatments, and N removed in grain from the two nutrient treatments were measured. These were balanced against the N loads measured in 2012 to estimate what the theoretical loads in 2015 and 2020 should have been assuming no leaching losses. These loads were then compared with the actual loads measured in 2015 and 2020 in both treatments and an assessment made as to whether large amounts of N were unaccounted for over the period of the experiment. This provided an opportunity to understand whether 'N-mining' from the SOM may have contributed to any change in the FF-C levels during and after supplementary nutrient addition occurred. As the additions of P and S were known to be in excess of that removed by the crops due to the supplementary fertiliser used, no detailed mass balance of those nutrients was made.

Statistical analysis
Soil data (%C and %N, C:N ratios, C and N stocks) were analysed and presented separately at each sampling depth using a paired t-test to determine if there was a significant effect of supplementary nutrients. The patterns of the FF-C and FF-N down the soil profile were also compared at the time of measurement in 2012 (end of Phase 2), 2015 (end of Phase 3) and 2020 (end of Phase 4). Fitted curves for C and N concentrations and stocks (Figs. 1a-f and 2a-f) were determined using the "Analysis" package within Sigmaplot 14.5 (Systat Software Inc.), where an exponential decay function best described the response (Supplementary Information; Tables S4 and S5). Total stocks of FF-C and FF-N were calculated from soil concentration and bulk density data and differences compared using one-way ANOVA in GENSTAT 20. Annual crop yield and biomass data were also analysed separately in each season for supplementary nutrient effects and the total biomass and total grain yield across the seasons was also analysed.

Economic analysis
An assessment of the economic viability of using supplementary nutrients applied to crop residues to increase soil C sequestration was conducted by comparing the cost of the additional nutrients applied (based on Starter15 (14:12.7:0:11: NPKS)) costing AUD $500 Mg −1 ) with the changes in crop productivity (grain yield) and changes in soil C (Table S6). As all other crop operations and inputs were identical in both treatments, the difference between the additional expenditure on supplementary nutrients and the changes in income from both crop yield and potential payments for soil C sequestration provided a simple estimate of the economic outcome (profit difference) for the supplementary nutrient strategy. In the central analysis we used the average prices paid for wheat, canola and lupins during the 2007 to 2020 period as reported by SARDI (https:// www. pir. sa. gov. au/ prima ry_ indus try/ grains/ crop_ and_ pastu re_ repor ts), and assumed the price paid for increases in soil C of AUD$40 Mg −1 of CO 2 equivalents (https:// www. renew ablee nergy hub. com. au/ market-prices/).
As a result of significant recent fluctuations in both fertiliser and grain prices as well as different prices for C on global markets, we conducted a sensitivity analysis by varying each of these components by magnitudes that have been observed in the recent years.

Results
Crop growth, yield and quality and residue loads A summary of the crop growth, yield and residue levels for the (+) and (−) supplementary nutrient treatments is provided in Table 2. Crop establishment was unaffected by supplementary nutrient additions to the residue in any year (except for a small reduction in wheat in 2011), and plant populations established were at or above recommended levels for all crops throughout the experimental period. Where measured, there was a tendency towards higher early vegetative biomass when supplementary nutrients were added to residues, which was significant in some years and persisted after supplementary nutrient application ceased in 2014. Despite these intermittent effects on early biomass production there was little impact of supplementary nutrient addition on final crop biomass (2012 only) and no effect on the total biomass grown during Phases 3 or 4. There were few significant effects of supplementary nutrients on grain yield, mostly evident in the machine harvested yield, though the trends were similar in hand-harvested yield. The significant effects on yield were both positive (2014, 2016) and negative (2011,2019) and occurred both during the period of supplementary nutrient application (Phase 2, 3; 2011, 2014) and after it had ceased (Phase 4; 2016, 2019). Notably there was a significant negative impact of (prior) nutrient addition on grain yield and protein in 2019, five years after supplementary nutrient addition had ceased. This occurred under conditions of an extreme spring drought (Fig. S1) providing evidence of 'haying off'. Despite these individual year impacts, there was no significant effect on the total grain yield removed over the 8 years of Phase 3 and 4. In contrast to grain yield, there were consistent significant increases in grain N content in response to the supplementary nutrient addition in Phase 2 and 3 during the period of supplementary nutrient addition, which also persisted in some years during Phase 4 (2017, 2019). This was most extreme during the very dry 2019 season where an increase in grain N content was observed. As grain N can be diluted in higher yielding crops, the amount of N removed in grain combines impacts on grain yield and N content. Supplementary nutrients resulted in higher grain N removal in 2011, 2012, 2013 and 2017 but with less N removal in 2019. The effect of supplementary nutrients on the amount of crop residue produced and remaining increased in only 1 year of 9 (2012), and there was no effect on the total residue returned over the 8 years.   Table S4. The increase in FF-C and FF-N with supplementary nutrient addition evident throughout the soil profile after Phase 2 in 2012 persisted in 2015, after the ongoing period of supplementary nutrient addition in Phase 3 (compare Fig. 1a, b with Fig. 1c, d). Although the effect was no longer significant at all individual depths, it remained so in the surface layers, and the fitted curves down the entire profile remained significantly different. In contrast, at the end of Phase 4 in 2020, 5 years after the supplementary nutrient application to residues was ceased, the differences in FF-C and FF-N were absent at depths below 0.45 m, and the differences in the surface layers had also diminished (compare Fig. 1c, d with Fig. 1e, f). Despite this change in the pattern of differences down the profile, the fitted curves remained significantly different (Table S4), being driven mostly by higher levels of FF-C and FF-N in the plus nutrient treatments persisting in the surface layers (Fig. 1e, f). Thus the increase in FF-C and FF-N established by the addition of supplementary nutrients to residues during Phase 2 and Phase 3 though diminished, persisted for at least 5 years to the end of Phase 4. Throughout the entire period, there was little impact of the supplementary nutrient treatments on the (FF-C):(FF-N) ratio (Table S7). However, consistent with observations by Kirkby et al. (2016), the C:N ratio declined significantly with depth from around 11 to 12 in the surface layers to 5 to 7 in the deepest layers (Table S7).

Change in FF-C and FF-N stocks
The profiles of FF-C and FF-N stocks for the plus and minus supplementary nutrient addition treatments across the 3 phases (2012, 2015 and 2020) are shown in Fig. 2. The parameters for the fitted curves in Fig. 2 are provided in Table S5. Due to the limited impact of the supplementary treatment on bulk density, the pattern of FF-C and FF-N stocks followed the same patterns as those for the concentrations shown in Fig. 1. The cumulative stocks across the entire soil profile for the same sampling times are shown in Table 3.
The cumulative data show that the difference in FF-C and FF-N stocks established at the end of Phase 2 (9.6 Mg ha −1 and 0.82 Mg ha −1 , respectively) were effectively maintained during the ongoing application of nutrients in Phase 3 (to 9.8 Mg ha −1 and 0.99 Mg ha −1 ) but declined to 3.0 Mg ha −1 for FF-C and 0.35 Mg ha −1 for FF-N after 5 years without supplementary nutrient application (i.e., at the end of Phase 4 in 2020). Despite the overall decline in the absolute FF-C and FF-N stocks in both treatments in Phase 3 presumably linked to climatic or overall negative nutrient balance, the differences generated by the supplementary nutrients persisted. and within each panel differences at each depth were analysed separately with significance being indicated (P < 0.01*; P < 0.05**; P < 0.001***). The fit-ted curves for C and N concentrations through the soil profile were calculated using an exponential decay function (see Supplementary Information Table S4), with the r 2 value for each curve shown on the panels. In all cases there were significant differences between the pairs of curves for both C and N for all years with the P value shown on each panel

Nitrogen mass balance
The N mass balance for Phase 2, 3 and 4 is shown in Fig. 3, and was calculated using N inputs (Table S3) and N removal in harvested grain (  and within each panel differences at each depth were analysed separately with significance being indicated (P < 0.01*; P < 0.05**; P < 0.001***). The fitted curves for C and N stocks through the soil profile were calculated using an exponential decay function (see Supplementary Information  Table S5), with the r 2 value for each curve shown on the panels. In all cases there were significant differences between the pairs of curves for both C and N for all years with the P value shown on each panel top-dressed N added to the cereal crops was better matched the removal in grain. The decline in the N balance after 2014 in Fig. 3 does not account for the likely N-fixation by the lupin crop, which if effective, could have been as high as 140 kg N ha −1 (based on 20 kg N per Mg above ground biomass, Peoples and Craswell 1992). Differences in the levels of soil mineral N generated by the supplementary nutrient treatments could have generated differences in the fixed N between the nutrient treatments, so it is difficult to be definitive about the effects on the N balance. In addition to this uncertainty the level of potential Table 3 Cumulative stocks of carbon (C) and nitrogen (N) associated with fine fraction soil organic matter (SOM) to a depth of 1.8 m for Harden soil in 2012, 2015 and 2020 follow-ing supplementary nutrient addition to incorporated residue nutrient across the phases of the experiment Data and the difference are presented as the mean ± 1 standard error of paired replicates (n = 4) across experimental blocks a Two-way analysis of variance for FF-C indicated a significant main effect due to both year (P < 0.001; LSD 0.05 = 3.5) and nutrient addition (P < 0.001; LSD 0.05 = 2.8) with no significant interaction b Two-way analysis of variance for FF-N indicated a significant main effect due to both year (P < 0.001; LSD 0.05 = 0.43) and nutrient addition (P < 0.001; LSD 0.05 = 0.30) with no significant interaction Year ( 47.9 ± 2.7 50.9 ± 3.0 3.0 ± 1.9 5.52 ± 0.29 5.87 ± 0.19 0.35 ± 0.33  Figure S1). Differential N leaching may have occurred between the treatments if higher levels of mineral N were present in the soil at the time of the high rainfall.

Economic analysis
A simple economic analysis comparing the cost of the supplementary fertiliser with the potential change in income derived by yield increases or payments for sequestered soil C (@$40 Mg −1 ) was conducted and is summarised in Table 4. The fertiliser cost was calculated for both the actual costs used (Starter15 @$500 ha −1 ) as set out in Table S6, but also using the cheapest equivalent forms of fertiliser to supply only the levels of N, P and S required to sequester the C (see Table S8). By 2015 supplementary nutrients (as Starter15) amounted to a cost of $1109 ha −1 with an extra 9.8 Mg ha −1 of soil C being sequestered (35.9 Mg CO 2 equivalent @$40 Mg −1 ) worth $1436 ha −1 . Crop yields in the period 2007 to 2015 were changed by a net increase in value up to 2015 of + $837 ha −1 . Consequently, by 2015 the use the Starter15 as a nutrient source generated a return of $2273 ha −1 from additional grain yield and C credit income. This is a profit difference after 8 years of $1164 ha −1 for an investment of $1109 or a 1.05 return on investment. Using the optimised fertiliser strategy reduced the cost to $687, increasing the profit to $1586 ha −1 for a 1.43 return on investment. Notably the payment for soil C @$40 Mg −1 CO 2 equivalent is the larger driver of the profit given the lower overall and variable annual effect of the treatment on crop yield during these seasons.
In the period after nutrient addition ceased (2015-2019, Phase 4) there was no further cost of supplementary nutrients, and additional crop yield income overall up to 2019 was reduced to $369 ha −1 , and the difference in soil FF-C by 2020 had declined from 9.8 Mg ha −1 to 3.0 Mg ha −1 worth only $439. Consequently, the position in 2019 was either a loss of $301 ha −1 (using Starter 15 (14:12.7:0:11, NPKS) as a fertiliser source) or a small profit difference of $121 using optimised fertiliser strategy. In summary, using an optimised fertiliser strategy, supplementary nutrient addition to residue can generate reasonable profits, primarily from payments for sequestered soil C rather than yield benefits, but these may be quickly eroded when nutrient addition to residue ceased for a period of 5 years.
The costs of fertiliser and the prices paid for grain and for soil C are all subject to market fluctuations Table 4 Cumulative profit difference resulting from the application of supplementary nutrients to crop residues after 8 years (2015, end of Phase 3) and in 2020 (end of Phase 4) after a subsequent 5 years without nutrient application to residues Profit outcomes were generated using long-term average (LTA) prices for grain and fertiliser and a carbon (C) price of $40 Mg −1 CO 2 equivalents (as described in Table 4) and are shown in bold text. The effect of changing the C price (± $20 Mg −1 ), changing the LTA grain prices by − 50% and + 50% and increasing the LTA fertiliser price by 50% and 100% on the profit difference are shown which will influence the economic outcomes from these scenarios. For example, in 2021/22 the cost of fertilisers more than doubled and grain prices also increased significantly while the prices paid for C remained relatively static. To examine the sensitivity of the profit difference outcomes to these variations, we adjusted the price paid for fertiliser up by 50% and 100% (which in reality occurred in 2021), and adjusted the price paid for C and for grain by − 50% and + 50% (Table 4). The analysis shows that in 2015 at the end of Phase 3, after 8 years of nutrient addition, the cumulative profit difference of $1586 ha −1 using LTA grain and fertiliser prices and the current C price ($40 Mg −1 ) could vary from as low as $480 ha −1 when fertiliser prices were high and grain prices low, up to $2004 ha −1 with LTA fertiliser prices and higher grain price. Increasing the price paid for C to $60 Mg −1 increased the cumulative profit at average fertiliser and grain prices (to $2304 ha −1 ), with the profit difference ranging from $1198 ha −1 to $2722 ha −1 across different price scenarios. A decrease in C price to $20 Mg −1 halved cumulative profit at average fertiliser and grain prices (to $868 ha −1 ) with a range of − $238 to $1286 ha −1 . In 2015 after 8 years of fertiliser addition, there was little risk of economic loss under any of the price scenarios. However, given the investment in fertiliser ranged from $687 ha −1 to $1374 ha −1 under different scenarios, the return on investment (ROI) ranged from − 0.17 to 1.87 at $20 Mg −1 for C, 0.35 to 2.92 at $40 Mg −1 , and 0.87 to 3.96 at $60 Mg −1 . The risks of relatively small returns on investment (< 1.00) over an 8-year period may discourage growers from adopting the practice especially at low C prices.
In 2020 after 5 years without addition of supplementary fertiliser to the crop residues there were risks of substantial losses (− $970 ha −1 ) at high fertiliser prices and low C grain prices and little chance of profit under any scenario even with a C price of $40 Mg −1 . The risk of losses remained likely even as C prices increased, and especially if fertiliser prices increased above the LTA. Even at the highest C price of $60 Mg −1 and with LTM fertiliser and grain prices, the cumulative profit difference of $340 in 2020 represented a ROI of only 0.49. It is clear that an extended period without ongoing nutrient addition to the residue can lead to a reduction in the original amounts of additional C sequestered by nutrient addition, eroding potential profits from sequestered C which has important implications with respect to when C credits may be paid.

Discussion
The capacity to sequester more soil C from C-rich crop residues, using targeted nutrient supplementation applied to the residues based on the nutrient stoichiometry of FF-SOM has been confirmed in the current experiment. The further 3 years of supplementary nutrient supply during Phase 3 (2013-2015) maintained the difference in sequestered FF-C (9.8 Mg ha −1 ) and N (0.99 Mg ha −1 ) that had been establish established during Phase 2 (2007-2012; Kirkby et al. 2016). This demonstrated the increase in FF-C established during that Phase based on a 0.4 mm size fractionation was durable, at least while the nutrient input to the residue was maintained. The difference in FF-C and FF-N diminished after nutrient supplementation ceased however and, while still significant (+ 3 Mg ha −1 ), raises questions about the longevity of the FF-C sequestered in this soil fraction once supplementation ceases, especially if there is a period of declining nutrient balance. The specific seasonal conditions, soil type, crop sequence choice and management are likely to have influenced the outcomes, and below we discuss these in the broader context of nutrient stoichiometry and how nutrients applied to residues (rather than to the crop) impact FF-C sequestration.
Effects of supplementary nutrients on crop growth, yield and N uptake Supplemental nutrients applied to incorporated crop residues in summer, with warm, moist soils in the absence of a growing crop (or weeds) target FF-SOM build-up by increasing humification efficiency through generation and turnover of microbial biomass Coonan et al. 2020b). In this experiment, the conditions to facilitate such processes were possibly limited by below average rainfall in the December to March period in 2 out of 5 years in Phase 2, no years in Phase 3, and 3 of the 5 years in Phase 4. Net immobilisation of the added nutrients by soil microorganisms into the FF-SOM pool after incorporation (January) would be expected to leave little of the added inorganic nutrients in the soil to directly influence early crop growth. As FF-SOM increased over the years, net mineralisation of N may have also increased to improve the nutrient availability to the crops. The early growth of the crops (sown in May) was increased significantly in 2 of the 8 years only when supplementary nutrients were added to the residue, suggesting the direct effect of residual inorganic nutrients in the soil on early crop growth was minimal. There were also few impacts of the treatment on crop biomass at anthesis or final harvest, while grain yield was significantly affected by the nutrient treatment in 6 of the years. In 10 of the 14 years, spring rainfall was below-average ( Figure S1) limiting the capacity for the expression of any potential growth and yield benefits arising from improved mineralisation of nutrients arising from the higher FF-SOM generated in the plus nutrient treatment.
In fact, small negative impacts (up to ~ 10%) of the plus nutrient treatments on wheat yield due to 'haying-off' in dry springs (van Herwarden et al. 1998;Kirkegaard and Ryan 2014) occurred in the dry years of 2008, 2009, 2011 and 2019, while yield increases occurred in the wetter seasons of 2014 and 2016. Dry spring conditions allowed the slower-growing (−) nutrient treatments to generate similar biomass by anthesis, presumably as the (+) nutrient treatments had exhausted soil water and growth became limited by water availability. Low spring rainfall (September and October) during flowering and grain development also tended to reduce the yield potential of crops in most years. By final harvest, the levels of biomass and yield from the (+) nutrient treatment were significantly higher than the (−) nutrient treatment in only one of the five years of measurement. As a result of this water limitation to growth and yield (not uncommon in rainfed, dryland environments), the outcomes of the supplementary nutrient treatments on soil-C are not confounded by large differences in primary production, returned biomass or nutrients removed in harvested yield.
While the effects of the supplementary nutrient treatments on yield varied across seasons, there was a more consistent increase in grain N concentrations in response to the plus nutrient treatment across most seasons, suggesting access to higher levels of mineralised N later in the season. This is consistent with, and further evidence of mineralisation from the higher levels on FF-SOM sequestered in the plus nutrient treatment. The benefits to increasing grain quality by maintaining or building FF-SOM are of especial significance in the Australian context, where declining grain protein has become an issue in many areas where legume-based pasture and crops have declined, and N-fertiliser applied to crops averages only 40 kg N ha −1 (Angus and Grace 2017).
It is interesting to consider whether more average or above average spring rainfall during the experimental period may have generated more seasons in which nutrients generated both higher yield and protein (as in 2010, 2016, 2017) and how this may have influenced the economic outcomes of the experiment. Dryland farming areas in Australia are predicted to become drier with more variable rainfall under climate change which should provide caution with respect to the reliability and effects of increased SOM on production and partial profit difference (Sun et al. 2020;Hunt et al. 2020).

Effects of nutrient addition and cessation on FF-C and FF-N
The period of ongoing nutrient addition to the retained crop residues during Phase 3 extended the duration of treatments first imposed by Kirkby et al. (2016) and corroborated previous observations from laboratory incubations ), that concluded nutrient supply (NPS) and not C input, was limiting FF-C sequestration in soil. The relatively large difference in both FF-C (9.6 Mg ha −1 ) and FF-N (0.87 Mg ha −1 ) established to 1.8 m in Phase 2 was maintained during Phase 3. Both the total amount, and the pattern of the increase which was evident throughout the soil profile (more than 50% of the increase occurred below 0.3 m) persisted during the extended period ( Figs. 1 and 2). The work confirms previous laboratory (Moran et al. 2005;Kirkby et al. 2013) and the field studies (Jacinthe et al. 2002;van Goenigen et al. 2006;Kirkby et al. 2016;Poeplau et al. 2017) showing direct application of nutrients to crop residues can enhance the formation of SOC by increasing humification efficiency for conversion of wheat residue-C to SOC by soil microorganisms. Though increased fertilisation can also improve C sequestration by increasing net primary productivity, this was not evident during the course of this experiment.
Despite the differences in the FF-C and FF-N concentrations and loads being maintained between the (-) and (+) treatments during Phase 3, the actual C and N concentrations and loads for both treatments decreased to some extent during the period (Table 3). We suspect this may have resulted from microbial nutrient mining due the significant decline in the N balance that occurred in both treatments during that period (Fig. 3). No N fertiliser was applied to the lupin crop in 2014, and only a small amount (7 kg ha −1 ) to following canola crop in 2015 despite some 200 kg ha −1 N being removed in grain during that period. The observed loss in FF-C and FF-N in the (−) nutrient treatment is consistent with the strong negative N balance in this phase where a positive priming effect and net nutrient N mining from preexisting SOM would be expected to occur along with a low humification efficiency (Kirkby et al. 2013). The lower FF-C and FF-N stocks for the (+) nutrient treatment (relative to Phase 2) however is less easy to explain other than that restricted N inputs to the lupin and canola crops occurred which may have also driven a net N mining situation from SOM and the fact that the measure of FF-C and FF-N were not made until March 2015 (just prior to the first crop in Phase 4). Other potential sources of N loss may have been through leaching although the estimated leaching losses during that period were estimated to be low.
Over the 5-year period in Phase 4, after nutrient application was ceased the most notable impact on the FF-C and FF-N profiles was the disappearance of the differences which had existed deeper in the soil profile during Phase 2 and 3, while some differences in the upper layers persisted. Kirkby et al. (2016) speculated about the origin of the differences in FF-C and FF-N at depth, given the incorporation of residues and fertilisers was limited to 0.1 m. They suggested the increases below 0.3 m could be caused by leaching of C from the surface layers as either soluble C, colloidal material including microbial detritus, or as a result of increased root growth at depth. If the increase in microbial biomass and turnover associated with the addition of residue with nutrients generates large amounts of soluble C (or microbial detritus), it is possible that the cessation of this process will restrict additional FF-C moving into the subsoil and that the resident FF-C and N may be more rapidly mineralised. Under such a scenario the high concentrations of FF-C and FF-N established in the surface layers presumably takes a longer period of time to equalise as the FF-C and N are mineralised. The N-balance during Phase 4 was actually reasonably stable after the decline at the end of Phase 3, as the fertiliser applied to the crops better matched that removed in grain. Under those conditions it is likely that the larger microbial biomass associated with the higher levels of FF-C in the + nutrient treatment resulted in a greater rate of residues and soil C being respired. Coonan et al. (2020a) has shown there can be a higher respiration per mass of soil from higher fertility soils compared to lower fertility soils from the same field. This higher rate of respired FF-C in the surface combined with the lack soluble C moving to depth could explain the decline in the difference in FF-C between the treatments after the supplementary nutrient treatment was ceased. Effectively it appears that after nutrient application ceased, the (+) nutrient treatment was slowly returning to the same levels of FF-C and FF-N as the (-) nutrient treatment from the bottom of the profile up. Importantly, it also indicates the transitory nature of the FF-C that was built up with nutrient addition in Phase 2 and Phase 3 of our experiment, and reinforces the need to better understand and differentiate the definition of 'stable C' in soil and its contribution to long-term C stocks at least within the 0.4 mm fraction that we assessed.

Economic outcomes and implications for C sequestration payments
The cost effectiveness and risk of using supplementary fertiliser applied to crop residues to sequester soil C is of interest to farmers and policy makers alike. Richardson et al. (2014) called this a "hidden cost of soil C sequestration" and showed that the inorganic nutrient requirement to generate 1 Mg ha −1 of soil C as FF-SOM from wheat stubble based on stoichiometric composition was 73, 17, and 11 kg ha −1 of N, P and S respectively. At the time, it represented around $150 ha −1 for nutrients that, although not lost from the system, would not be immediately available for plant growth compared to fertiliser applied to growing crops. Based on average long-term fertilizer and grain prices, and an assumed price for soil C of $40 Mg −1 CO 2 equivalents, a grower following our strategy of supplementary nutrient supply over 8 years in Phase 2 and 3 would have made a profit of $1586 ha −1 after investing $687 in nutrients, for a return on investment of 1.43 over 8 years. Although there was a low risk of losing money from the strategy under different price scenarios in that time, at current fertiliser prices (2022, increase of 100%) and C prices closer to the likely $20 Mg −1 the profit declines to just $181 ha −1 over 8 years, and a drop in grain prices could have generated a loss. The economic prospects may have improved without 10 of 14 years having below average spring rainfall generating yield penalties for nutrient addition and limiting the amount of biomass (and therefore FF-C) returned to the soil. Overall the lack of downside risk may encourage some growers to adopt the strategy though other investments of their capital may have greater returns in their business over the same period.
The economic outcome after Phase 4 provides a more cautionary tale, as it is clear that the benefits in higher soil C sequestration will diminish over time after nutrient application ceases. This also questions both the definition of 'stable C' in soil based on size fractionation and further highlights the transient nature of the soil C that we sequestered as FF-C in Phase 2 and Phase 3, especially if the nutrient requirements for FF-C generation cannot be continuously met over time. In this experiment, even after a period of only 5 years without the nutrient addition to the residue, the increase in sequestered soil C had diminished by ~ 75% with a significant effect on the economic outcome. This suggests that nutrient additions to crop residue need to be persistent to maintain the economic benefits offered by increasing C sequestration in soil.

Conclusion
Various studies have evaluated the likely economic benefit to growers from different C farming initiatives focussed on soil C sequestration. Collectively there has been a general consensus that sequestering soil C in cropping soils would be slow, risky and prone to periods of loss due to the impact of climate on both primary productivity (C-input) and the respiration of C. Our strategy of supplemental nutrient addition to crop residue has been shown to generate reasonably rapid increases in soil C, as opposed to adoption of no-till, stubble retention, crop rotations or other farming system interventions, and is relatively easy to adopt with prospects to fit into existing management regimes (Richardson et al. 2019). The value from the strategy at the end of Phase 3 in the current experiment using average pricing generated a value of ~ AU$20 Mg C ha −1 year −1 , which at current average prices for inputs, harvested grain and CO 2 (@ AU$40 Mg −1 CO 2 ) is economically attractive. However the gradual loss of the sequestered C when nutrient supplements ceased demonstrate the challenge in payment systems reliant on permanency while associated benefits to crop yield and quality raise questions about "additionality". Regardless of its climate change mitigation potential, the strategy represents a paradigm shift in agronomy related to nutrient-useefficiency (NUE) which presently favours nutrient application directly to growing crops to deliberately avoid "losses" through immobilisation. Even temporary increases in SOM sequested from crop residues and slowly released to crops could improve overall system NUE. Clearly a wider assessment of the potential value of achieving C sequestration through nutrient supplementation of crop residues across different farming systems and agroecological regions is required.
Author contributions All authors (JAK, AER, CAK) contributed equally to the concept, design and conduct of the experiments reported. Sample processing and preliminary analysis was coordinated by CAK. JAK, AER and CAK contributed equally to the preparation of data for presentation in Tables and Figures, in the writing of the text and in the various stages of the manuscript review.
Funding Open access funding provided by CSIRO Library Services.

Declarations
Competing interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.