Clarifying confusions over carbon conclusions: antecedent soil carbon drives gains realised following intervention

potential


Introduction
The land use sector faces increasing pressure from society to sustainably intensify food production and improve environmental stewardship on the one hand, yet also to reduce greenhouse gas (GHG) emissions and engender carbon removals on the other (Ara et al., 2021;Ibrahim et al., 2019;Sándor et al., 2020). Globally, agriculture accounts for around 18.4 % of the GHG inventory, with 5.5 % from croplands and agricultural soils (Balogh, 2022;Ritchie et al., 2020). Ceteris paribus, GHG emissions from the agricultural sector are expected to rise in parallel with expansion of intensified agricultural land use (Hunter et al., 2017;Taylor et al., 2016). While irrigated croplands offer unique opportunities to modulate both agricultural GHGs and soil carbon sinks (McGill et al., 2018;Sapkota et al., 2020), our knowledge of potential GHG mitigation through SOC sequestration associated with irrigated systems is very much in its infancy (McGill et al., 2018). Irrigated systems conceivably provide necessary impetus for greater production and presumably SOC sequestration, yet greater production is often driven by higher use of nitrogenous fertilisers (Adnan et al., 2022;Rawnsley et al., 2019), which in turn may increase nitrous oxide (N 2 O) emissions, causing net GHG emissions to overcompensate for mitigation realised by enhancement of SOC stocks Christie et al., 2014;Henry et al., 2022). Indeed, recent work has suggested that greater SOC stocks may not ultimately result in lower whole farmnet GHG emissions due to the tight coupling between productivity and GHG emissions (Bilotto et al., 2023;Harrison et al., 2021), since greater soil organic matter and SOC may be conducive to greater crop or pasture production, thereby potentially allowing nitrous oxide and/or methane emissions to rise in parallel with SOC accrual.
Contemporary scholarly research of sustainable intensification via irrigation has primarily focused adopted a productivity and/or economic lens (Ash et al., 2017;Monjardino et al., 2022;Muleke et al., 2022a,b;Muleke et al., 2023;Muleke et al., 2022c). For example, Muleke et al. (2023) concluded that productivity and profitability could be sustainably raised through integration of grain legume crops into existing rotations, while Muleke et al. (2022b) show that the number of profitable crop types fell from 35 to 10 under future climates, reflecting an interplay between commodity price, yield, crop water requirements, water costs and variable costs, wherein water use and costs increased, while crop yields decreased under future climates. In their study, effects of climate change on profit were not related to long-term rainfall, with future climates depressing profit by 11-23 % relative to historical climates, but rather impacts of future climates were more closely related to crop type and maturity duration, with many crop types that were traditionally profitable under historical climates being no longer profitable in future. Such studies demonstrate that productivity (e.g., crop yield) and profitability have often been the motivation of previous work, primarily to the exclusion of associated or interacting effects on soils, including implications of systems interventions on soil health, carbon and greenhouse gas (GHG) emissions.
Confusion among the scientific and practitioner community abounds as to the contexts with which SOC increases following intervention. Previous research has shown that weather conditions (largely precipitation), management and soil texture drive SOC sequestration (Emde et al., 2021;Henry et al., 2022;Singhal et al., 2023), primarily through modulating influence on pasture or crop productivity (Langworthy et al., 2018;Yan et al., 2022). Conversely, other field studies have concluded that tillage and residue management practices manifest greater control over SOC sequestration (Chao, 2019;Karstens et al., 2022;Saurabh et al., 2021). Others argue that changes in SOC sequestration are contingent upon type of crop, genotype, and crop rotation (Antón et al., 2022;Sándor et al., 2020;Yang et al., 2023). Such treatment effects are further confounded by regional differences in global warming, with soil carbon sequestration expected to fall depending on regional specificity and extent of the climate crisis (Bilotto et al., 2023) and frequency of extreme events (Harrison, 2021;Langworthy et al., 2018;Liu et al., 2023). Within such studies, particularised conclusions may well be robust (e.g. , but valid interpretations while bespoke can be voided when extrapolated to new contexts, agroecological regions and/or production systems. We posit that much of this confusion can be attributed to initial (antecedent) SOC stocks: experiments beginning from SOC levels close to their floor (e.g., less than 1 %) could be reasonably expected to improve SOC, whereas antecedent SOC at ceiling levels (e.g., >6 %) might be expected to demonstrate less gain, because stocks are already close to their maximum levels. While scientists and practitioners often focus on within-experiment activities to assess the relative promise associated with an interventionsuch as treatments and climatic conditionsthey tend to downplay or even completely overlook the influence of initial SOC on their results. Experimental results may be further confounded by limited duration of field experiments, which canconfine interpretations drawn to parochial weather conditions realised during the finite window within which the experiment is conducted. Trost et al. (2013) note that it is difficult to estimate the effects of factorial combinations caused by the environment and management on SOC sequestration and, as such, the present study aims to address this knowledge gap.
Here, we build on foundational insights of previous work (Muleke et al., 2023) to investigate how initial SOC stocks influence long-term trajectories of SOC accrual using a real irrigated grains farm as a use case. We invoked three whole-farm adaptations and compared directional change with the existing system (Baseline) to examine how management (crop type, rotation, irrigation, fertilisation) and soil type (sand, clay, sandy loam) impacted on SOC over the long-term. Specifically, our aims were to (1) examine the impacts of antecedent SOC on sequestration for varying whole-farm adaptations and soil types, and (2), quantify effects of these adaptations and soil types on net farm GHG emissions.

Study overview
An irrigated broadacre farm situated in the Riverina region of New South Wales (NSW), 145.8904 E) was used as a case study. A detailed description of the case study, including climate, soils, farm layout, crop rotation, irrigation methods and source of irrigation water supply is provided in Muleke et al. (2023) and Monjardino et al. (2022). In brief, our study was conducted by initialising our biophysical model (APSIM) according to the cropping production system operating on the case study farm. This step was made to ensure that the simulations were realistic and regionally representative. This included initialisation of farm management (fertiliser and irrigation use, crop rotation etc) and soil physical characteristics. Soil and climatic details were derived from publicly available scientific databases (Sections 2.2 and 2.3). Soil data were used to initialise APSIM, while daily climate data drives APSIM (daily computations of crop growth and development are driven by climate inputs, similar to other contemporary cropping systems models).
We next developed three generic adaptations scenarios in consultation with a regional reference group of experts comprising farmers, farmer groups, consultants and extension staff with the aim of improving SOC stocks. These adaptations comprised Simplification (reducing inputs), Diversification (increasing crop type and genotype diversity in rotation) and Intensification through use of greater inputs and crops requiring irrigation (detailed in Section 2.3.3). Because net greenhouse gas (GHG) emissions depend on the balance of nitrous oxide (N 2 O) and SOC sequestration, we computed net emissions asociated with each adaptation. Adaptations were compared with the baseline (status quo operation) to assess direction and extent of change. Each adaptation was made in the model by changing farm management (sowing time, irrigation, fertiliser, crop rotation) as shown in Table 1. Thus, our primary focus in this study was on farming systems adaptation and on change relative to the baseline (fractional and percentage), rather than on absolute values (e.g., actual tonnes SOC/ha).

Historical climate data
We conducted simulations using 110 years of historical climatic data (1 January 1910-31 December 2019). Data for daily maximum and minimum temperature, rainfall and solar insolation were sourced from SILO (Scientific Information for Land Owners) Patched Point Dataset (Jeffrey et al., 2001) (shown in Fig. S1). Historical annual rainfall of the case study location was 403 mm, with 218 mm in winter and 185 mm in summer (Jeffrey et al. (2001). These data have been rigorously screened for anomalies by the data provider using a two-pass interpolation system; erroneous values were excluded from the interpolation, so they did not adversely affect the gridded surface (https://www.longpaddock.qld. gov.au/silo/about/about-data/).

Simulation of crop growth, yield and soil carbon stocks
We adopted the farming systems outlined in Muleke et al. (2023) to examine the impact of initial SOC and whole farm adaptation on SOC accrual and nitrous oxide (S-N₂O) emissions using the Agricultural Production Systems sIMulator (APSIM) v7.10, Keating et al., 2003). APSIM has been comprehensively validated under a wide range of conditions, with such studies showing that the model performs well in simulating crop growth, soil C dynamics and soil nitrogen (N) cycling Harrison et al., 2019;Holzworth et al., 2014;Keating et al., 2003;Luo et al., 2011;Mohanty et al., 2020;O'Leary et al., 2016;Wang et al., 2003a).

Surface and soil organic matter and carbon
Daily climate data are used in APSIM to compute biophysical values, including plant growth, labile soil carbon inputs, mineralisation, plant senescence and soil water infiltration and extraction. Management in the model further influences growth and carbon inputs, with irrigation improving growth during periods of plant water deficit, and nitrogen improving growth during periods of nitrogen stress. Plant growth, soil inputs and soil carbon in the model are derived by rate-state equations: the rate of any given biophysical process depends on the state of that variable on the prior day. For example, growth is greater when leaf area index on the prior day is relatively high. APSIM simulates surface and soil organic matter, of which soil organic carbon is subset. When surface organic matter is added (e.g., from wheat straw at harvest), that residue will be added to the existing pool of surface organic matter. For all simulations, we assumed an initial surface organic matter pool of 1000 kg DM/ha. Each organic matter pool has separate C:N ratios, decomposition rates and specific areas. Leaching of soluble inorganic N (nitrate and ammonium) is determined by rates of rainfall or irrigation. Soil organic matter is partitioned into fresh (FOM), medium (BIOM) and a more stable humus (HUM) decomposition pool, with fractionation between these pools dependent on the type of organic matter and crop residues. A subset of the HUM pool is defined as inert. Residue decomposition is dynamically determined by soil moisture, temperature, C:N ratio of the substrate, whether the residue is standing or prostrate, and the amount of material contacting the soil surface. Decomposition results in some loss of carbon in surface organic matter as CO 2 and transfer of C and N to the soil; material with higher C:N ratios creates an immobilisation demand, which is satisfied from mineral N in surface layers; inadequate mineral N can restrict decomposition of residues in surface layers. For detailed accounts of surface and soil organic matter models in APSIM, readers are referred to Dimes and Revanuru (2004) and Probert et al. (2005).
Given that soil water holding capacity and growth differs across soil types, we adopted three regionally representative soil types: brown vertosol clay, sand and sandy loam, following soil classification nomenclature of Isbell (2021). We used default hydraulic attributes for saturated water flow in APSoil (Table S2; Dalgliesh et al., 2016). To examine effects of initial SOC stocks, we sourced measured data for bulk density and soil organic matter pools (fresh organic matter pool (FOM), microbial biomass pool (BIOM) and humic pool (HUM); Probert et al., 1998) from the APSoil database (Dalgliesh et al., 2012) (Table S1), then initialised APSIM with six initial SOC concentrations (0.5 %, 2 %, 3 %, 4 %, 5 % and 6 %) within the top 100 cm of soil profile, following Dalgliesh et al. (2016). SOM and SOC in the APSoil database were determined by Dalgliesh et al. (2016) using Walkley Black or Leco analytical methods on soil cores taken in the Coleambally region of the case study farm; historical management associated with these data were assumed to be similar to status quo operations on the case study farm (default parameters shown in Table S1). Neither bulk density nor management were altered within any 110-year simulation (except for crop rotation specified in Section 2.3.3). In line with our underpinning aimto examine the effect of initial SOC on subsequent change in the simulation and to account for SOC variability due to historical climate and management types, we specified a range of SOC values that were used to initialise APSIM. Ensuing changes in simulated SOC over the 110 years were an emergent property, driven by climate, and modulated by interactions between antecedent SOC, soil type and management over the course of the simulation.

Management and cropping systems
We used various APSIM crop modules to simulate growth and development of wheat Wang et al., 2003b), canola (Robertson et al., 1999), faba bean (Turpin et al., 2003), maize , rice (Bouman et al., 2001) and cotton (Hearn, 1994). Irrigated crops were sown on fixed dates (mid-May for winter crops, mid-November for summer crops); dryland crops were sown when sufficient autumnal rainfall occurred (25 mm over four days). Sowing dates were set according to information outlined in Muleke et al. (2022a), shown in Table 1. Default soil parameters were adopted from the APSIM soil library following Dalgliesh et al. (2012). Plant available soil water at sowing and application of irrigation and nitrogen (N) fertiliser for each adaptation was set following Muleke et al. (2023) (details provided in Section 2.3.3, Table 2 and Fig. S2). Maximum potential long-term yields were estimated by identifying optimal flowering periods (OFPs), defined as the window that minimises long-term risk of abiotic stress exposure. OFPs for irrigated and dryland crops were computed as the flowering dates corresponding to ⩾95 % of the maximum 15-day running average frost-heat yield following Muleke et al. (2022a),  and Liu et al. (2021).

Whole farm systems adaptations
Three whole farm systems adaptations suggested by regional experts were compared with the existing farm system (the Baseline), which comprised 750 ha of irrigated canola− wheat− wheat in serial rotation, with each winter crop followed by a summer fallow (Table 2, Fig. S1). Summer fallows stored water and nitrogen, incurring weed control costs (0-4 herbicide spray events). Thematic adaptations included Intensification (including higher rates of irrigation and fertiliser, greater farm area under irrigation, and inclusion of rice and maize in the rotation); Diversification (including a grain legume to diversify the crop rotation) and Simplification, being the opposite of Intensification. Further details are provided in Tables 1, 2 and Fig. S1 and previous studies (Monjardino et al., 2022;Muleke et al., 2023).
Diversified scenarios included a canola-wheat− faba bean rotation, with each winter crop followed by a summer fallow (750 ha). Relative to the Baseline, Simplified scenarios were allocated lower irrigation rates and irrigated farm area (only 50 % of wheat area; 375 ha); dryland canola was sown on 750 ha of the farm area (Table 2 and Fig. S1). Intensified scenarios were irrigated at higher rates (ML/ha), assuming 750 ha for the winter crops (canola, wheat) and 25 % of the farm area (188 ha) for summer crops (maize, cotton, rice), with the remaining 75 % of farm area fallowed in summer. Nitrogen fertiliser rates for each crop type in the Baseline scenario were obtained from the case study region. We then applied these fertiliser rates for corresponding crops in the Diversified and Simplified scenarios. We assumed 50-100 kg N/ha greater application rates for each crop in the Intensified scenario, as our aim for this adaptation was to examine the impact of intensification through greater application of N and irrigation (Table 1).

Greenhouse gas (GHG) emissions and global warming potential (GWP)
2.3.4.1. Soil emissions. To quantify direct net farm soil GHG, we converted S-N₂O and the change in SOC over the simulation duration (ΔSOC) to carbon dioxide equivalents (CO₂-e). For S-N₂O, we used a 100 year GWP of 298 (Eq.1; IPCC, 2013): Where Soil N₂O emissions represent tonnes of carbon dioxide equivalents (tonnes CO₂-e) from simulated S-N₂O and N₂O (Layer) represents simulated S-N₂O in uppermost (0-30 cm) and deeper (30-100 cm) soil layers. SOC stocks (0-30 and 30-100 cm) were converted into CO 2 -e by multiplying SOC by 3.67 using Eq. 2 (IPCC, 2013): ΔSOC emissions(tonnes CO₂e) = SOC (Layer) × 3.67 (2) Where ΔSOC emissions represent tonnes of carbon dioxide equivalents from ΔSOC (difference between SOC at start and end of 110-year period), while SOC (Layer) represents simulated SOC in each of the upper (0-30 cm) and deeper (30-100 cm) soil layers in tonnes C per hectare. Net farm GWP across whole-farm adaptations, soil types and initial SOC levels was calculated as the sum of S-N₂O and ΔSOC using Eq. 3, with positive gains in ΔSOC resulting in negative GHG emissions.

Net onfarm GWP(tonnes
Where EI (kg CO₂-e/kg crop yield) represents emissions intensity, annual GHG emissions represents average direct GHG emissions in kilogram carbon dioxide equivalents (kg CO₂-e) and yield represents average simulated harvested crop yield (kg/ha).

Statistical analysis
Regression analyses were performed to test statistical significance of the impact of initial SOC, whole farm adaptation and soil type on relative change in long-term SOC storage. We fitted linear models using R 4.2.2 (R Core Team, 2022) to predict change in SOC, using initial SOC, soil type and adaptation as predictor variables. We adopted confidence intervals (CI) of 95 %, a p-value of < 0.05 to denote significance, and adjusted R-squared (R 2 ) of 0.99 (refer to Table S3). We applied omnibus "analysis of variance" (ANOVA) to the linear model to test the statistical significance between initial SOC, soil type, adaptations and their interaction, while Tukey's post-hoc test was employed to identify disparities between paired level combinations (Tables S4-6). The relative variation contributed by each variable was computed using Eq. 5.

Variation contributed =
SS associated with the variable Total SS × 100 Where SS represents the sum of squares for each factor (initial SOC, soil type, adaptations and interactions therein) from ANOVA (Table 5); Total SS represents total sum of squares for all dependent factors and their interactions in the regression model.

Table 2
Description of the Baseline and whole farm adaptations, after Muleke et al. (2023).
Adaptation Description

Baseline
The current farm system with gravity irrigation. The Baseline represents the current farm situation in terms of agronomy and irrigation.

Diversified
Diversification of the Baseline system using a grain legume (Faba beans; Vicia faba). This adaptation has similar average water usage and irrigated farm area as the Baseline, but a greater variety of winter crops grown within the year − i.e., wheat, canola and a grain legumewith similar inputs and costs.

Intensified
This was designed to be a high-input, high-output adaptation. Relative to the Baseline, this scenario applies higher amounts of water per unit area and per year and assumes a larger portion of the farm area is irrigated (i.e., less unirrigated fallow), and has higher inputs per unit area and year.

Simplified
This adaptation was designed as a low-input, low-output adaptation. Relative to the Baseline, the Simplified scenario was designed to require less irrigation water per unit area and year, has more rainfed crops, and reduced inputs per unit area.

Antecedent SOC stocks determine SOC gain realised following intervention
Across soil types and adaptations, greater initial SOC stocks correlated inversely with long-term SOC change in the surface 30 cm (Table 3 and Fig. 1). In the deeper soil layers (30-100 cm), rate of change in SOC showed little association with initial SOC stocks (Table 3 and Fig. 3). Across all climate years (110 years), soil types and adaptations, a one unit increase in initial SOC within the upper 30 cm decreased temporal SOC accrual by 0.49 % (10 tonnes C/ha) (Fig. 1) or approximately 20 tonnes C/ha in SOC stocks over the whole soil profile (Table 3 and Fig. 1). Sandy loam and sandy soils had more pronounced reductions in SOC with increasing initial SOC stocks (0.07 to − 1.6 % and 0.09 to − 1.4 %, respectively; Table 3 and Fig. 1), whereas clay soils exhibited the lowest decrement in sequestered SOC (1.8 to − 0.1 %), suggesting higher potential to accrue SOC in clay soils.
Simplified scenarios elicited reductions in long-term SOC in the uppermost 30 cm in response to increasing SOC levels (from 0.07 % to − 1.3 %; Table 3 and Fig. 1). In contrast, Diversified and Intensified adaptations showed less decline in SOC accumulation (1.4 to − 0.7 % and 1.2 to − 1 %; Table 3), suggesting superior SOC accrual (Fig. 3) attainable through diversification and intensification. Lower initial SOC (0.5 %) elicited greater gains in SOC accumulation, whereas higher initial SOC concentrations (>2 % for sandy and Sandy loam soils, and >5 % for clay soil; Fig. 2) resulted in considerable SOC loss across adaptations (Table 3 and Fig. 2). This finding suggests that higher antecedent SOC or SOC that exceed ceiling SOC result in lower gains over the duration of a simulation or experimental period.

Table 3
Influence of antecedent SOC on long-term change  in soil organic carbon (ΔSOC) in upper (0-30 cm) and lower soil profiles (30-100 cm) as influenced by soil type (grey rows) and whole-farm adaptation (blue rows).
At lower initial SOC (0.5 %), clay soils had higher rates and absolute SOC change (0.03 t C/ha /yr and 3.4 t C/ha respectively), resulting in larger long-term gains in SOC (50 tonnes C/ha) in the uppermost soil layers. In contrast, sandy soils had lower accrual (0.01 t C/ha/yr and 1.3 t C/ha respectively), as shown in Tables 3, 4, 5 and Fig. 3. Lower SOC accrual with greater antecedent SOC was more evident in sandy soils compared with clay soils (Table 5).
Diversified and Intensified adaptations had higher potential for SOC gains at lower initial SOC content (0.5 %) in soil surface (2.8 and 2.4 t C/ ha respectively; Table 4 and S3) relative to the Baseline and Simplified adaptations (1.8 and 1.5 t C/ha respectively; Tables 4, 5 and Fig. 3). The increased SOC accrual resulted in greater gains in long-term SOC under Diversified and Intensified scenarios (39 and 31 t C/ha respectively) compared with the Baseline and Simplified scenarios (23 and 19 t C/ha respectively; Table 3, Fig. 4 and S6).
Diversified and Intensified scenarios accrued more SOC on clay soils (72 and 50 t C/ha respectively; Fig. 4), compared with Baseline and Simplified scenarios (44 and 35 t C/ha respectively). The greatest losses in SOC (− 17 to − 44 t C/ha) occurred in Simplified scenarios at high initial SOC (6 %), suggesting that simplifying cropping systems (or deintensification) may not be an enabler for long-term SOC accrual.
High SOC accrual for wheat and canola underpinned larger SOC stock gains in Intensified and Diversified scenarios under clay and sandy soils, and Diversified scenarios under sandy loam soil. Enhanced SOC gains for summer crops (maize, cotton and rice) increased SOC stocks for Intensified scenarios relative to Baseline and Simplified scenarios (Fig. 5). Even so, increments in initial SOC elicited greater SOC losses for summer crops (cotton and rice) cultivated on clay and sandy soils under Intensified scenarios (Fig. 5).

Relative impact of initial SOC, adaptation and soil type on SOC stocks
Effects of initial SOC, adaptation and soil type on SOC and interactions therein were significant (Table 5 and S3). Initial SOC accounted for an average of 74 % of the variation in long term SOC stocks, in comparison to 12 % for soil type and 3 % across adaptations (Fig. 6). Clay soils had greater effects on long-term SOC stocks relative to sandy soils (Table S4), whereas Intensified and Diversified scenarios induced greater change in SOC stocks compared with Baseline and Simplified scenarios (Table S5). Low initial SOC (0.5-2 %) were conducive to considerable gains in SOC storage compared with high initial SOC (3-6 %) (Table S6).

Global warming potential (GWP) and GHG emissions intensity
Despite greater mean yield and significant carbon removal by SOC accrual, Intensified scenarios resulted in higher GWP (2.8-5.8 t CO₂-e) and emissions intensities (0.5-0.9 kg CO₂-e/kg crop yield, relative to the Baseline (− 0.1 to 1.9 t CO₂-e and 0.1-0.5 kg CO₂-e/kg crop yield respectively). Diversified scenarios had the lowest net GWP (− 0.7 to 1.5 t CO₂-e) and carbon footprint (− 0.2 to 0.2 kg CO₂-e/kg crop yield). Conversely, Simplified scenarios evidenced lowest yields and greater emissions intensities (0.2-0.6 kg CO₂-e/kg crop yield) in comparison with the Baseline (Figs. 7 and 8). Summer crop rotations (maize, cotton and rice) accounted for a large portion of farm GWP and emissions intensities, associated with increased S-N₂O emissions resulting from higher application rates of N-fertilisers and irrigation on sandy soils (Fig. 9, S4 and S5). High N-fertiliser rates elicited greater N leaching in irrigated summer crops (in particular rice) on clay and sandy soils (Fig. S5). These results highlight a tight coupling between crop productivity, S-N₂O emissions and N-leaching, demonstrating that intensification with synthetic N fertiliser can increase net GHG emissions.
In addition to significant removal of atmospheric CO₂ via SOC accrual, Diversified scenarios were more effective in mitigating GHG emissions compared with Intensification (Figs. 7, 9 and S4), reflecting the potential for Diversification as a prospective climate change adaptation and mitigation intervention through addition of grain legumes (here, faba bean) to the baseline crop rotation, although Diversification had lower yields overall than those modelled for Intensification. Simplification caused relatively little difference to the Baseline (Table 1, Figs. 3,4 and S5).
Sandy and clay soils exhibited greater vulnerability N-leaching compared with sandy loam soils, the former leading to elevated A. Muleke et al. emissions intensities (0.4-0.9 kg CO₂-e/kg crop yield) and GWP (1-4.7 t CO₂-e) from increased S-N₂O emissions ( Fig. 7 and S5). Only sandy loam soils in Intensified systems generated noteworthy GHG emissions from subsoil layers (0.5-0.7 t CO₂-e) primarily due to leaching and S-N₂O (Fig. S5). Higher initial SOC increased GHG emissions and carbon footprint, but had no discernible impact on crop yield (Figs. 7 and 8).

Discussion
Past work has shown that sustainable intensification underpinned by higher rates of irrigation, greater farm area under irrigation and higher rates of N application may represent a financially viable alternative in the face of the climate emergency (Muleke et al., 2023). However, this may not always be the case, as (1) irrigation water may not always be physically available or economically affordable (Monjardino et al.,   Fig. 2. Long-term (1910Long-term ( -2019 change in soil organic carbon (SOC) in the soil surface (0-30 cm; purple and green bars) and subsurface (30-100 cm; red points) at varying antecedent SOC, across soil-types (clay, sandy and sandy loam) and whole-farm adaptations (Baseline, Diversified, Intensified and Simplified).  Table 4 Absolute soil organic carbon (SOC) accrual in the upper (0-30 cm) and deep (30-100 cm) soil layers for varying antecedent SOC, soil-types (grey rows) and whole-farm adaptations (blue rows). Absolute SOC accrual was estimated as the change in SOC stocks over 110-years (1910-2019) relative to base period (1910).
disentangle the impacts of whole-farm intensification/simplification/diversification on (1) long-term SOC accrual and (2) global warming potential (GWP) resulting from adaptations across varying antecedent SOC levels and soil types. While we address our aims via use case, our concepts and systems approach underpinning Simplification, Diversification and Intensification could be generically adapted to any farming system or agroecological region; Simplification being aimed at reducing inputs and resource use, Diversification aimed at increasing diversity of enterprises employed on farm (here, crop types), and Intensification designed to increase irrigation and fertilisation per unit area and per farm area. While numerical results obtained for any production system or agroecological region are contextual, the approach we articulate hereto is amenable to generalisation.
Although Intensification resulted in superior productivity and improved SOC accrual, this adaptation also rsulted in the highest GWP (5.8 t CO₂-e) and emissions intensity (0.9 kg CO₂-e/kg crop yield; Table 4, Figs. 4 and 7). Diversification manifested greater SOC, GWP mitigation (1.5 t CO₂-e) and carbon footprint (0.2 kg CO₂-e/kg crop yield) consistent with past work Zou et al., 2022), but was accompanied by concomitant reductions in yield compared with Intensification ( Figs. 7 and 8). Simplification increased emissions intensities (0.6 kg CO₂-e/kg crop yield), diminished SOC accrual and evoked significant yield penalties relative to the Baseline (Table 4, Figs. 3, 7 and 8). These findings suggest that diversifying irrigated systems would seem to be one solution for both adaptation and mitigation (McGill et al., 2018;Meier et al., 2020;Paustian et al., 2019;Sándor et al., 2016), albeit at the expense of yield and profitability (Monjardino et al., 2022;Muleke et al., 2023). On the other hand, intensification can elevate productivity and profitability (Muleke et al., 2023) but would exacerbate climate change through additional GHG (McGill et al., 2018;Phelan et al., 2015) even though Intensification is often the most resilient form of adaptation (Muleke et al., 2023). These findings suggest that the long-term sustainability and endurance of any purported mitigation or adaptation option will be a function of sustainable improvement in multiple dimensions (economic, environmental, social, institutional).
It is important to note that the conceptualisation of SOC processes in the model we used (APSIM) quantify change in SOC inputs relative to SOC mineralisation; as APSIM does not account for molecular SOC residence times; we thus refer to "SOC accrual" and "SOC stocks" rather than sequestration per se. SOC residence time, including sequestration and formation of mineral-associated organic carbon, are aspects worthy of future investigation. For example, Begill et al. (2023) assessed a wide range of SOC (5-118 g kg -1 ) and clay contents (30-770 g kg -1 ) and found that the proportion of mineral-associated organic carbon (MAOC) was stable across a range of cropland and grassland management types, and across SOC levels. While soil texture influenced the slope of the relationship between SOC and MAOC, no upper limit to MAOC was observed. These insights challenge the notion that MAOC accumulation is limited by the soil fine fraction, and thus may compriseamenable targets for conceptualisation and contrasting within contemporary agro-ecosystemsmodels, such as APSIM.
We showed that long-term SOC stocks in the topsoil (0-30 cm) and subsoil (30-100 cm) in irrigated cropping systems was primarily governed by initial SOC relative to practice changes, soil type and prevailing climate (74 % effect; Table 5 and Fig. 6). We found evidence suggesting that lower initial SOC (0.5-2 %) are likely elicit greater ex poste SOC accrual compared with contexts with higher initial SOC concentrations (3-6 %). Our findings clarify confusions over previous conclusions and are consistent with cross-site empirical studies demonstrating an inverse relationship between changes in SOC and initial SOC (Arndt et al., 2022;Emde et al., 2021;Slessarev et al., 2023).
We showed that clay-rich soils with low antecedent SOC (0.5 %) accrued more SOC (50 t C/ha) in topsoil (30 cm), relative to sandy soils (17 t C/ha; Tables 3-5 and Figs. 2-4). Notwithstanding our discussion above relating to MAOC, this result is consistent with previous research opining that incremental gains in SOC are associated with higher concentrations of silt and clay, wherein fine fractions (<50 µm) tend to protect SOC (Beare et al., 2014;Feng et al., 2013;Matus, 2021), such that C sorbed to mineral surfaces increases with soil specific surface area (Kaiser and Guggenberger, 2003). As such, sand-rich soils are more Fig. 3. Influence of antecedentsoil organic carbon (SOC) on the annual rate and long-term SOC accrual within the upper soil profile layers (0-30 cm) across soil-types and whole-farm adaptations. Green bars represent SOC change (t C/ha) over 110-years (1910-2019). Red points represent mean annual SOC change (t C/ha/yr).
likely to reach their C equilibrium at lower SOC due to lower SOC protective capacity in comparison to clay soils (Angers et al., 2011;Six et al., 2002;West and Six, 2007). We acknowledge that the presence of a theoretical equilibrium or "saturation" point is contextualised to the soil types in question, and indeed, is the subject of ongoing scholarly debate, with some authors suggesting that "mineral-associated organic carbon formation capacity" would be more appropriate nomenclature than use of the term "saturation" (Emde et al., 2022). Within the conceptual and mathematical boundaries of the present study however, our results suggest that sandy soils reached C saturation levels beginning from lower antecedent SOC levels (2 %) than clayey soils (5 %), evidenced by trends shown in Fig. 4. Incremental levels of initial SOC beyond this upper limit induced greater SOC losses in sandy soils compared with clayey soils. Taken together, our results suggest that management practices aimed at increasing carbon inputs may be futile where SOC is relatively high, as C-retention rate asymptotically diminishes as SOC approaches the equilibrium point (Castellano et al., 2015;Guillaume et al., 2022). Ascertaining SOC equilibrium thresholds across sites, climates and soil types would be fertile ground for future research, for knowledge of such thresholds would guide practitioners aspiring to improve SOC as to whether changing management practices would be worth while.
Our findings suggest that winter crops (wheat and canola), cultivated on clayey and sandy soils enhanced SOC accrual for Diversified and Intensified systems (72 and 23 t C/ha respectively; Table 3 and Figs. 4-5) whereas, when grown on sandy loam soils, these crops were only likely to enhance SOC accrual under Intensification (25 t C/ha). Summer crops (maize, cotton, and rice) improved SOC accrual in Intensified scenarios relative to Baseline and Simplified. Nevertheless, these crops were prone to N-leaching on clay and sandy soils, and increased S-N₂O emissions on sandy loam soils, resulting in notable yield losses and increased global warming potential (Figs. 4-8 and S4-5). Broadly, this finding underscores the beneficial impact of winter crops on SOC accrual, and the potential integration into crop rotations aimed at reducing GHG emissions through improved C accrual.
Deeper soil layers (30-100 cm) exhibited minimal variability in SOC accrual, across initial SOC levels, soil types and whole-farm adaptations (as shown in Table 3 and Fig. 2). Notably, only sandy loam soils in intensified systems showed significant GHG emissions in subsoil layers (0.5-0.7 t CO₂-e), primarily due to elevated S-N₂O emissions (Fig. S5).
We found that increasing farm area under irrigation, and/or increasing nitrogen fertilisation rates can and does improve SOC stocks. We note however, that such increase is relative to the status quo farming system in operation and antecedent SOC. In general, we found that over the 110-year simulation, SOC of Simplified systems changed little compared with SOC of the Baseline, suggesting that farm area under irrigation had relatively little impact at the whole farm scale (Simplification irrigated 350 ha of wheat, whereas the Baseline irrigated 750 ha of wheat). When antecedent SOC was low, Diversification and Intensification elicited similarly high SOC increments, however when antecedent SOC was high (e.g., SOC 6 %), Intensified systems lost more SOC than Diversified systems. Inclusion of legumes (faba bean) into canola-wheat-wheat rotations (i.e. Diversified system) improved soil nitrogen fixation and reduced soil C:N ratios, increasing soil organic matter decomposition rates, improving subsequent temporal gains in SOC. Intensified systems generally improved soil organic matter inputs when initial SOC was low, particularly for the highly productive crops, such as maize, which required an average of 15 ML/ha/year compared with 2.8 ML/ha/year for wheat crops (averaged over the simulation). However, when initial SOC was high (6%), high production and residue crops immobilised more mineral nitrogen, subsequently accelerating SOC decrement over the 110-year simulation. In general, changes in total SOC were primarily attributable to changes in the labile pool (microbial biomass and products), and to a lesser extent the humic pool, with inert organic carbon pools remaining relatively stable over the simulation. In general, changes in the proportion of fresh and humic pools of soil organic matter relative to the change in total organic matter were relatively consistent across adaptations, soil types and climate years.

Concluding remarks
This study has underscored existential challenges landholders face in balancing priorities for agri-food production aginst those of environmental stewardship and GHG emissions abatement, with Intensification enabling higher productivity and SOC accrual but also evoking greater global warming potential (5.8 t CO₂-e). Conversely, Diversification (via adding grain legumes to crop rotations) was more conducive to SOC accrual and GHG emissions mitigation, albeit with lower yields compared with Intensification. As well, Intensification (via additional irrigation and nitrogen fertiliser) was more likely to catalyse N 2 O emissions, counterbalancing carbon removals associated with greater SOC stocks. Collectively, these results call for more holistic assessments of GHG emissions (e.g., N 2 O, CH 4 , CO 2 ) to account for pollution swapping associated with intervention. We showed that clay-rich soils have higher asymptotic carbon thresholds (>5 %) and accrue more SOC, while sandy soils have lower SOC ceilings (2 %), truncating SOC plateaus to lower thresholds, as shown in Fig. 4. We found that antecedent SOC exerted the greatest control over long-term SOC accrual relative to practice change, soil type, crop type and climate. We conclude that (1) contextualised identification and benchmarking of soil C-equilibrium thresholds may elicit more probable improvement in carbon removals and agri-food production, (2) improvement in SOC stocks had little impact on productivity compared with addition of N fertiliser, noting that the latter concurrently increased N 2 O emissions, and (3) identification of antecedent SOC relative to SOC ceilings would help guide practitioners as to whether changing practices to improve SOC stocks would likely be a fruitful or futile endeavour.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.