Major limitations to achieving “4 per 1000” increases in soil organic carbon stock in temperate regions: Evidence from long‐term experiments at Rothamsted Research, United Kingdom

Abstract We evaluated the “4 per 1000” initiative for increasing soil organic carbon (SOC) by analysing rates of SOC increase in treatments in 16 long‐term experiments in southeast United Kingdom. The initiative sets a goal for SOC stock to increase by 4‰ per year in the 0–40 cm soil depth, continued over 20 years. Our experiments, on three soil types, provided 114 treatment comparisons over 7–157 years. Treatments included organic additions (incorporated by inversion ploughing), N fertilizers, introducing pasture leys into continuous arable systems, and converting arable land to woodland. In 65% of cases, SOC increases occurred at >7‰ per year in the 0–23 cm depth, approximately equivalent to 4‰ per year in the 0–40 cm depth. In the two longest running experiments (>150 years), annual farmyard manure (FYM) applications at 35 t fresh material per hectare (equivalent to approx. 3.2 t organic C/ha/year) gave SOC increases of 18‰ and 43‰ per year in the 23 cm depth during the first 20 years. Increases exceeding 7‰ per year continued for 40–60 years. In other experiments, with FYM applied at lower rates or not every year, there were increases of 3‰–8‰ per year over several decades. Other treatments gave increases between zero and 19‰ per year over various periods. We conclude that there are severe limitations to achieving the “4 per 1000” goal in practical agriculture over large areas. The reasons include (1) farmers not having the necessary resources (e.g. insufficient manure); (2) some, though not all, practices favouring SOC already widely adopted; (3) practices uneconomic for farmers—potentially overcome by changes in regulations or subsidies; (4) practices undesirable for global food security. We suggest it is more realistic to promote practices for increasing SOC based on improving soil quality and functioning as small increases can have disproportionately large beneficial impacts, though not necessarily translating into increased crop yield.


| INTRODUCTION
The "4 per 1000 initiative: Soils for Food Security and Climate", also known as "4 per mille" or "4&", for increasing soil organic carbon (SOC) stocks was launched by the French Ministry of Agriculture in 2015 in preparation for the Paris climate conference of the United Nations Framework Convention on Climate Change (UNFCCC). It is described in several web sites including: 1. https://www.4p1000.org/4-1000-initiative-few-words 2. http://newsroom.unfccc.int/lpaa/agriculture/join-the-41000-initia tive-soils-for-food-security-and-climate It is also described by Chabbi et al. (2017) and . The aim of the initiative is to promote land management practices leading to an increase in the stock (i.e. quantity as opposed to concentration) of SOC at the rate of 4& (0.4%) of the initial value per year for 20 years. It was originally suggested that, if this rate of C sequestration was achieved for all soils globally to a depth of 40 cm, the C removed from the atmosphere would equal annual CO 2 -C emissions from fossil fuels of 8.9 Gt and thus "halt the annual increase in CO 2 in the atmosphere" (http://4p1000.org/under stand). Later enunciations of the initiative have recognized that increases in SOC are only likely in soils that are being actively managed for agriculture (and possibly under managed forestry) and, even in these, such a rate of increase may not be achievable everywhere (Chabbi et al., 2017;Chambers, Lal, & Paustian, 2016;Lal, 2016;. It is suggested that C sequestration in agricultural soils alone, as opposed to all soils, might be limited to c. 1 Gt/ year (Chabbi et al., 2017;Smith, 2016). There is also some uncertainty about the soil depth referred to in the initiative; in some publications, the 0-100 cm depth is mentioned in addition to 0-40 and 0-30 cm (Chabbi et al., 2017).  published a compilation of data from 20 regions of the world showing opportunities and limitations for achieving the 4& rate of SOC increase.
The initiative has been generally welcomed as laudable because any contribution to climate change mitigation is helpful and equally, or perhaps more importantly, any increase in SOC is virtually certain to improve the quality and functioning of many soils. This is especially relevant in the light of the reported widespread incidence of land degradation globally (UNCCD, 2017). However, there have been significant criticisms of the initiative by several authors (van Groeningen, van Kessel, Hungate, & Oenema, 2017;Baveye, Berthelin, Tessier, & Lemaire, 2018;de Vries, 2018;VandenBygaart, 2018;White, Davidson, Lam, & Chen, 2018), suggesting that there are many situations where the 4& rate of increase in SOC is not feasible for land managers in practical situations and that some of the experimental examples given by  are not representative of what is practically achievable in wide-scale agriculture. There also appear to be differences of opinion as to whether the 4& goal is a specific target, an aspiration goal or even "more of a concept (or even a slogan)" as suggested by Minasny et al. (2018).
The aims of this paper were twofold. First, to report the rates of SOC increase in several long-term agricultural field experiments run by Rothamsted Research at three sites in southeast England between 1843 and 2013. In temperate climates, SOC changes slowly in response to changes in management so periods of years or decades are required to reliably detect and quantify rates of change (Macdonald et al., 2015;Storkey et al., 2016). We use data from 16 separate experiments on three different soil types, giving over 110 treatment comparisons. It is recognized that the sources of data are geographically constrained but the cropping systems are diverse, and treatments cover a wide range of management practices relevant to temperate regions, though do not include reduced tillage. The results represent the largest concentration of long-term SOC data globally, and the soil types and climates are broadly representative of temperate regions globally. Second, we attempt to use the long-term data to provide an evidence-based assessment of the likely range of situations where the 4& target might be achieved in practical agriculture, and for how long increases at this rate might continue.

| Experimental sites and soils
Soil samples from 16 experiments with a cool temperate climate in southeast United Kingdom were analysed for their organic carbon content. Much of the data have not been published previously and for several experiments, data have been updated. The soil types are silty clay loam at Rothamsted, sandy loam at Woburn and sandy clay loam at Saxmundham (see Table 1 for soil classifications). Long-term average annual rainfall is c. 700, 650 and 610 mm at Rothamsted, Woburn and Saxmundham respectively. Until the late 1980s, the mean annual temperature was c. 9.0-9.4°C at the three sites, but, at Rothamsted and Woburn, where meteorological recording has continued, it has increased by about 1°C in the last 25-30 years (Scott, Macdonald, & Goulding, 2014;Johnston, Poulton, Coleman, Macdonald, & White, 2017). In all-arable experiments, soils were cultivated by inversion ploughing to a depth of 20-22 cm. Table 1 lists the experiments with references to publications giving details. In subsequent Tables, data are grouped according to treatment, taking data from different experiments where required rather than discussing each experiment separately. Further information appropriate to this paper is in Supporting Information.

| Soil sampling and analysis
Most soils were sampled to 23 or 25 cm, i.e. a little deeper than current plough depth, with a semicylindrical gouge auger (2-3 cm diameter), taking many cores per plot. However, early samples on Broadbalk and Hoosfield were taken from several positions within each large plot with an open-ended iron box sampler (typically 30.5 9 30.5 9 22.9 cm deep). Each "box" sample provided a large mass of soil which made it possible to determine the weight of fine soil (<6 mm, the standard procedure at Rothamsted at the time for samples that were to be archived) per unit volume. Similar box samples have been taken from many long-term experiments at Rothamsted, Woburn and Saxmundham and the measured soil weight is used for these and other experiments on the same soil type. Soils were air-dried and subsamples ground to <355 lm for nitrogen (N) and carbon (C) analysis. Organic carbon was determined by Tinsley (1950) or by modified Walkley-Black (corrected to make the data equivalent to Tinsley;Kalembasa & Jenkinson, 1973). Later soils, and some earlier archived soils, were analysed for total C by combustion (LECO Corp., St Joseph, Michigan, USA) (Jenkinson, Poulton, & Bryant, 2008), on the assumption that most organic matter is held on the mineral particles. Where bulk density has not changed and, therefore, the weight of soil to a defined depth has not changed, the soil is sampled to the same depth. But, where bulk density has changed allowance must be made for this as discussed in detail in Johnston et al., 2017. In some of the experiments discussed here bulk density has declined due to addition of bulky organic material or a period in pasture, and later soil samples should have been taken to a slightly greater depth so that SOC is determined in the same weight of mineral soil. If this is not done, and the same weight of soil is used for the start and end of the measured period, this will overestimate the amount of C being sequestered. If the soil weight has been measured at the start and end of a period, but no allowance is made for the mass of extra soil which should have been sampled, and the amount of C it contained, then the amount of C being sequestered will be underestimated. Where bulk density has declined and where the weight of soil has been determined we have made a simple correction for that change. If, for example, the soil weight to a depth of 23 cm at the beginning and end of a period in ley-arable cropping was 3,770 t/ha and 3,470 t/ha respectively, then the difference, 300 t/ha, represents the amount of "extra" soil that should have been sampled at the end (see Johnston et al., 2017 Where the topsoil contains a higher concentration of OC (e.g. in grassland or woodland sites or where large amounts of manure have been applied) an increase of c. 5& will equate to 4& for the 0-40 cm layer.

| Statistical analysis
All the experiments reported here are, or were, long-term but for some we present changes over relatively short (e.g. 3-7 years), sometimes consecutive, periods (see Tables 5, 6, 9, S1-S5). Many of the experiments were established before the introduction of modern experimental design so that treatments were not necessarily replicated or randomized. Thus, the degree to which conventional statistics can be used to assess whether the observed changes are statistically significant is limited. Where errors have been calculated, and published, these are given in the appropriate tables. In other cases, where possible, the standard error of the mean of the differences between replicate plots for the stock of SOC at the start and end of each period is given together with the standard error of the annual rate of increase.
It is clear from these long-term experiments that measuring SOC differences between treatments with sufficient precision to assign statistical significance is challenging due to soil spatial variability and the large background of SOC against which changes occur. Reliably detecting the relatively small changes likely to result from application of the "4 per 1000" initiative will be even more difficult for the rea- quently. An advantage of this procedure was that it enabled data from the two or three FYM treatments in the experiments that were started at different times to be amalgamated, thus giving some degree of replication. Changes in SOC between each sampling point are given in Table S1, making allowance for the SOC in any soil to a greater depth that should have been sampled due to decreasing soil bulk density (see above). For both experiments the exact starting value has been estimated using the approach described by Jenkinson and Johnston (1977) because no samples were taken at the start of these two experiments. The FYM-treated plots are not replicated but soil from the several large box samples (see above) taken periodically from each large plot have been analysed separately for SOC.
For each experiment, a simple limiting exponential model was fitted (GenStat â , 2016) using the following equation: where org C is SOC, in t/ha, a is an asymptote representing the maximum capacity of the soil for SOC under the specific management, the slope coefficient, b, represents the available capacity of the soil to take up SOC and r is the exponential coefficient, representing the rate at which the soil accumulates SOC. was not applied to the two plots that started in 1843 and 1884 when these plots were fallowed, usually 1 year in five between 1926 and 1967, to control weeds (Johnston & Garner, 1969). Thus, for these two plots there is a break in the annual application of FYM between soil sampling in 1914 and that in 1967. This has been overcome by considering five separate series of data; 1843-1914 and 1967-2000 for the FYM plot since 1843; 1885-1914 and 1967-2000  3 | RESULTS

| Comparing our data with the 4& initiative
The "4 per 1000" initiative relates to SOC in the 0-40 cm depth of soil but data from the long-term experiments described in this paper are mostly derived from sampling to a depth of 0-23 cm, slightly deeper than the usual plough depth for arable soils at these sites. In Support- (1.15% C) in Hoosfield and 28.8 t/ha (1.00% C) in Broadbalk (Table S1). This is because the fields had been in arable cropping for several hundred years before the start of the experiments, probably in 5-course rotations where FYM was applied 1 year in five. On Broadbalk it is known that no manure was applied in the 5 years prior to the start of the experiment in autumn 1843 (Lawes & Gilbert, 1895). In both experiments SOC initially increased rapidly with  c Data to which the exponential models were fitted included any additional C present in the soil to an "equivalent" depth at the end of each period (see Supporting Information and Table S1).
T A B L E 2 Predicted amounts of organic C in topsoil, together with standard errors, from exponential models, org C = a + b 9 r time , fitted to soils given 35 t FYM ha À1 year À1 ; Hoosfield a and Broadbalk b  (Table 3).
Thus, with an annual FYM application of 15.1 t/ha, the "4&" goal (measured as 7& for the 0-23 cm soil depth) was exceeded over an extended period on this soil type, and it is likely that the rate of increase was much greater in the earlier years of each experiment.
In a much sandier soil (c. 13% clay) in the Woburn Green Manuring Experiment, starting at a lower SOC content, applying 25.1 t FYM/ha every 2 years increased SOC at an average rate of 14& per year, during a much shorter period of 18 years (Table 3). and where more root crops were grown in the 5-year rotation there was no increase in SOC from applying FYM. Johnston et al. (2017) considered that these differences were due to the additional number of soil cultivations needed to grow root crops. Changes in SOC in the different rotations are described by Johnston et al. (2017) and summarized here in a later section.
3.3 | Effects of various organic amendments on SOC stocks For L3, the amount of organic C at the end includes additional C present in the soil to an "equivalent" depth; i.e. so that the same mass of mineral soil was being considered both at the start and end of the period. to 32 t C/ha; Table 5), and the SOC content was nearer its equilibrium value in the soils with organic amendments. The effect of N fertilizer on SOC stock is discussed later.
In the Woburn Organic Manuring Experiment, the rate of SOC increase in the first 10 years of compost application (at 40 t/ha annually) was slightly lower than in the Market Garden Experiment (36& per year), possibly because of the higher initial SOC content  (Table 6).
In the Woburn Green Manuring Experiment, one treatment received straw at 3.77 t ha À1 year À1 every 2 years (Table 6). This increased SOC by an average of 6& per year over the next 18 years. In the Woburn Organic Manuring Experiment, straw was incorporated at different times and effects were compared with treatments receiving only inorganic fertilizers (shown in Table 3). In plots receiving only inorganic fertilizers, there were small decreases in SOC. Straw application led to variable increases ranging from zero to 24& per year measured over 6 or 10 years but, due to the variability, only the increase in SOC with largest rate of applied straw was statistically significant (Table 6).
In 1987, experiments comprising three rates of straw incorporation were started at both Rothamsted (18%-27% clay) and Woburn (14% clay). Where the amount of straw applied was equal to that produced, the average rates of SOC increase measured over 22 years were small and not significantly different from zero (Table 7). Where the rates of straw applied were two or four times the amount produced the increases were larger (5&-10& per year) but statistically only the four-times rate increased SOC significantly compared to no straw.
Three experiments, all on sandy soil with an initial SOC concentration of <1%, provide limited sets of data on the impact of green manures or cover crops on SOC (Tables 6 and S2). In one case, green manures led to a small increase (7& per year over 18 years) and in another a combination of green manures and grass leys within a mainly arable rotation increased the rate to 29& per year, measured over 6 years and probably just statistically significant. In another treatment, inclusion of overwinter cover crops in 4 years out of 10 failed to increase SOC (Table 6).   (Table 8).

|
Long leys (6 or 8 years) were also grown in the Woburn Organic Manuring experiment (Mattingly et al., 1974) where SOC increased significantly; increases ranged from 14& to 34& per year, measured at the end of the 6-or 8-year period in the ley (Table 9). As noted for organic additions, SOC increases tended to be greater when the initial starting value was lower.
Ley-arable experiments were started on two sites on silty clay loam soil at Rothamsted in 1949 (Johnston, 1973). One site, Fosters, had been in long-term arable cropping and had an initial SOC content of about 1.5%. In this experiment, some plots were put into permanent grass, some into a rotation of 3-year leys followed by 3years of arable crops and some plots continued to grow arable crops each year. Rotations that included a 3-year grass + N ley or a 3-year lucerne ley followed by 3 years of arable crops caused little increase in SOC compared to continuous arable (Table 9). Rotations that initially included grazed leys, but were replaced by cut grass/clover leys in the early 1960s, gave small increases in SOC (2&-7& per year; much carbon was also accumulated in the above-ground vegetation (Poulton et al., 2003). On these two woodland sites SOC had not reached a new equilibrium, even after >100 years (Figure 3). By contrast, when there was a change from old arable land to permanent

| General considerations
Any increases in SOC will almost certainly improve soil functioning and quality. There is considerable evidence that even small increases can have disproportionately large and beneficial effects on soil biological activities and physical properties such as water infiltration, aggregate stability and ease of tillage (Haynes, Swift, & Stephen, 1991;Snyder & Vazquez, 2005;Blair, Faulkner, Till, & Poulton, 2006;Thierfelder & Wall, 2012;Thierfelder, Cheesman, & Rusinamhodzi, 2013;Verhulst et al., 2010;Watts, Clark, Poulton, Powlson, & Whitmore, 2006). For example, even where zero tillage increases SOC concentration in near-surface soil (e.g. 0-10 cm), but with minimal impact on SOC stock to a greater depth, various soil physical properties can be considerably improved (Powlson, Stirling, Thierfelder, White, & Jat, 2016;Thierfelder & Wall, 2012). As pointed out by Janzen (2015), it is the process of decomposition of organic matter entering soil that delivers improvements in soil structure and functioning and the release of nutrients to crops: it is not necessary that a large increase in SOC stock is attained.
Where large amounts of FYM have been applied over many years much of the FYM-C has been lost even though the increase in SOC has been large. For example, in the silty clay loam soil on Broadbalk only 11% of the 480 t organic carbon added in FYM since 1843 has been retained in the soil to a depth of 46cm (Table 10), yet the soil properties are drastically altered compared to that with no manure addition. In the treatment where FYM had only been applied since 1968, SOC was still far from reaching a new equilibrium level and a much larger proportion (35%) of the added C could be found. In the sandy loam at Woburn, modelling with RothC-26.3, showed that 5%, at most, of the C added in plant residues over a 70-year period is still present in the soil (Johnston et al., 2017) but, again, this caused measurable changes in soil properties. However, improvements in soil physical properties caused by increased SOC do not necessarily translate into consistently increased crop yields (Hijbeek et al., 2017).
In contrast to the impacts of increasing SOC on soil properties and quality, the requirements for SOC increases for mitigating climate change through soil carbon sequestration are more stringent. First, it is necessary that SOC stock (i.e. the quantity of organic C in soil) is increased, not just the concentration of C in the surface layer. Second, it is essential that the additional C sequestered in soil would otherwise have been in atmospheric CO 2 and is not simply being transferred from one terrestrial location to another. Other well-known caveats must also be observed including the fact that the rate of SOC increase slows as the new equilibrium value is approached (Johnston et al., 2009;Powlson et al., 2012;Smith, 2014) and that increases are reversed if the modified management practice is not continued Includes any additional C present in the soil to an "equivalent" depth; i.e. so that the same mass of mineral soil was being considered at both the start and the end of each period.   indefinitely (Powlson, Glendining, et al. 2011;Powlson, Whitmore, & Goulding, 2011;Mackey et al., 2013).
In considering the climate change mitigation potential of any change in land management practice, it is the overall impact on all greenhouse gas fluxes that must be assessed, not only changes in SOC (Smith et al., 2008). Some practices leading to C sequestration may increase emissions of trace greenhouse gases, especially N 2 O.
For example, a recent global meta-analysis of experiments with manure showed that, on average, manure addition increased N 2 O emission by 33% compared to inorganic N fertilizer and this could largely offset the benefit of increased SOC stock (Zhou et al., 2017).
However, the trend was smaller with FYM than with manures containing a larger proportion of readily mineralizable N such as poultry manure.
Of the 114 treatment comparisons within the long-term experiments reported in this paper, almost two-thirds showed SOC stock increases >7& per year (or 5& in soils with larger initial SOC content); these increases, mostly measured in the 0-23 cm soil layer, being roughly equivalent to 4& goal specified for the 0-40 cm layer in the "4 per 1000" initiative. The increases were predominantly from organic inputs (manure, compost or straw) or from inclusion of pasture leys instead of continuous arable cropping. However, in evaluating the practicality of the "4 per 1000" initiative, considerable caution is required in transferring results from experiments such as these, and those reported by , to real world situations. Experimental results need to be evaluated in the light of the following practical and logistical considerations: 1. Is the practice suitable for a wide range of soil types and environmental conditions and possible for farmers to adopt in practical situations?
2. Is the practice profitable for a farmer? If the answer is "not under current conditions" but the practice is highly beneficial, either for climate change mitigation or long-term soil quality improvement or food security, there could be an argument for changes in policy or financial arrangements to promote uptake of the practice.  Between 1926 and 1967, sections of the experiment were fallowed each year to control weeds. The two sections in continuous wheat being considered here (sampled in 2000) were fallowed nine times in this period (eight times on section 9; 10 times on Section 1); FYM was not applied in these years.

| Removal of land from agriculture
Removing land from arable cropping, and allowing natural regeneration to deciduous woodland, led to large accumulations of organic C in soil in addition to that in trees; SOC stock doubled or trebled in a little over a century (Figure 3 and Poulton et al., 2003). Initially the rates of increase in these woodland reversion sites were very large and still exceeded 4& per year (in the 0-23 cm soil layer) during the final 30 years of measurement. Similarly, conversion from arable to permanent grass caused an increase of 55% in SOC in 58 years (Table 9). Piñeiro, Jobb agy, Baker, Murray, and Jackson (2009)  and prone to erosion, conversion from agriculture to forest or grassland may be a logical strategy (Albanito et al., 2016;Smith et al., 2013). A good example is the "grain for green" programme in China that reduced soil erosion and led to considerable increases in SOC (Chadwick et al., 2015;Song, Peng, Zhou, Jiang, & Wang, 2014).

| Addition of manures and other organic materials
The annual application of FYM at 35 t fresh material per hectare on Broadbalk since 1843 and Hoosfield since 1852 led to a high rate of SOC increase for several decades (Figures 1 and 2; Table 2) but such increases are unlikely to be achieved in practical farming situations.
Few farmers would have such large quantities of animal manure available each year for all fields on their farm and, even if they did, they may be prevented from applying such amounts continuously because of the risk of nitrate and phosphate pollution (Goulding, Poulton, Webster, & Howe, 2000;Hesketh & Brookes, 2000) and/or government legislation (e.g. www.gov.uk/guidance/nutrient-manage ment-nitrate-vulnerable-zones). However, on the positive side, even the more practically relevant application regimes (lower rates and/or applied every second or fourth year) in the Saxmundham and Woburn experiments led to rates of increase in SOC of 8&-14& per year even when averaged over 20-70 years ( content, and so benefit the most from organic additions, but is contrary to the current tendency in many countries for specialization, with animal and arable enterprises being spatially separated. Specialization is favoured by a range of practical (soil type and climate) and economic factors. If this trend is to be reversed to achieve environmental benefits, including increased SOC stocks, it is likely that significant changes in policies and financial incentives will be required. For example, if policies facilitated alterations in farm structures such that the estimated 300 Mt of solid manure produced in the European Union (Foged et al., 2011) were distributed more evenly this would be beneficial for soil quality.
While manure applications are very effective at increasing SOC and improving a wide range of soil functions, it should be recognized that these increases will generally not be delivering climate change mitigation but are rather a transfer of C from one location to another (Schlesinger, 2000). Globally, virtually all manure is currently being applied to soil at some location, though often in a suboptimal way. So, while there is certainly scope to make more rational and efficient use of manures both for soil C enhancement and for nutrient supply, almost all manure is already being used to some extent.
Thus, it is incorrect to assume that all SOC increases observed in experiments on manure application can be transferred to practical situations and fully treated as climate change mitigation: at least part of the benefit will already be accruing (Powlson et al., 2012;Powlson, Glendining, et al., 2011;. Where manure is being used inefficiently, or is applied to soil with an already high SOC stock, there is an opportunity to manage it differently by applying instead to low-SOC soils with potential for some degree of climate change mitigation plus numerous other benefits for soil quality, nutrient supply and decreased water pollution (Chadwick et al., 2015). Although, of course, there are major practical barriers to transporting manure, even over moderate distances.
The different organic amendments tested in experiments on a sandy loam soil (sewage sludge, now commonly called biosolids, and various composts) all led to high rates of SOC accumulation (Table 5). As with most of the FYM treatments, the application rates were large so the absolute rates of SOC increase cannot be directly transferred to practical farming but there is a strong indication that, for a given application rate, they deliver larger increases in SOC than FYM and the effect is longer lasting. This is presumably because these materials have already undergone greater decomposition than FYM during composting or sewage treatment so the organic C applied to soil will be somewhat more recalcitrant (Johnston, 1975).
An important factor regarding these materials is that they represent organic resources that are not currently widely utilized. In many countries, a significant proportion of food waste and similar organic materials is currently disposed of in landfill where decomposition returns CO 2 or CH 4 to the atmosphere (Bijaya, Barrington, & Martinez, 2006). Their greater use as a soil amendment can contribute to genuine climate change mitigation in addition to soil improvement (Powlson et al., 2012). However, it is recognized that some of these

| Retention of crop residues
Straw additions had variable, though generally positive, impacts on SOC stocks (Tables 6 and 7). Irrespective of soil type (silty clay loam at Rothamsted or sandy loam at Woburn), any increases tended to be greater where there was less SOC at the start of the experiment. Elsewhere, in temperate climatic regions, straw addition has also given positive but generally small increases in total SOC stock (Powlson et al., 2012;Powlson, Glendining, et al., 2011;van Groenigen et al., 2011). For example, in the 25 straw incorporation experiments of 6-to 56-year duration reviewed by Powlson, Glendining, et al. (2011) and , the increase in SOC was only statistically significant in six cases. However, as mentioned earlier, there is evidence that even small increases in SOC can have disproportionately large and beneficial effects on soil physical and biological properties (Houot, Molina, Chaussod, & Clapp, 1989;Malhi & Lemke, 2007;Ketcheson & Beauchamp, 1978;Thierfelder & Wall, 2012). In many regions, much cereal straw is already returned to soil (e.g. an estimate of 50% in the United Kingdom in 2008; Copeland & Turley, 2008) with much of the remainder being used for animal bedding and eventually returned to soil as FYM, so the scope for additional straw return to soil for climate change mitigation is limited. In smallholder agriculture in tropical regions the use of crop residues for animal feed or bedding is regarded as a significant barrier to direct return of crop residues to soil as part of wider adoption of conservation agriculture (Powlson et al., 2016;Thierfelder et al., 2013;Giller et al., 2011).

| Conversion from continuous arable to leyarable cropping
In the experiments considered here this change of management often led to increases in SOC exceeding 7& per year (equivalent to 4& per year in the 0-40 cm depth), sometimes for several decades (Tables 8 and 9). However, leys of just 3 years generally had only small effects. Increases in SOC from growing leys compared to continuous arable crops is a genuine transfer of additional C from atmosphere to soil due to additional inputs of photosynthate entering the soil during the pasture phases of the rotation, mainly through the root mass. The issues when considering the wider adoption of this system mainly concern profitability at the farm scale. As discussed above in the context of manure use, many farmers in developed countries apparently find that the practical and economic benefits of specialization outweigh the less immediate benefits of improved soil quality that can be gained from a mixed animal/arable farming sys- tem. An expansion of mixed systems would require changes in policy and financial incentives and presupposes that there is a consumer demand for the animal products derived from the grass phase. Such a change would only be logical from consideration of climate change if it was accompanied by a decrease in the number of animals fed on grain in more intensive animal productions systems and probably an overall decrease in consumption of meat and dairy products (Baj zelj et al., 2014). Such changes face major social and behavioural barriers and are unlikely to be achieved rapidly.

| Addition of N fertilizer
In three examples from the Broadbalk Experiment, increasing the annual application of N fertilizer (with adequate supplies of other nutrients) caused increases in SOC >4& per year in the 0-23 cm soil layer that continued for periods of between 13 and 33 years (Table 6). This is consistent with earlier studies on Broadbalk (Glendining et al., 1996) and reviews of numerous experiments worldwide showing increased SOC in soils receiving N fertilizer compared to unfertilized or unbalanced fertilizer applications (Ladha, Reddy, Padre, & van Kessel, 2011;Han, Zhang, Wang, Sun, & Huang, 2016). It is presumed that this is due to increased inputs of organic C into soil from roots, exudates and above-ground crop residues resulting from increased crop growth. It may also be that increased %N in these residues leads to greater conservation of organic C in soil organic matter. For regions of the world where fertilizer applications are currently low or nonexistent, this finding demonstrates a practical opportunity for increasing SOC with simultaneous benefits for crop production, provided other aspects of soil fertility, water availability and crop protection are in place. However, in such regions (e.g. Africa), there are major infrastructure and economic barriers to overcome to achieve rational fertilizer use. Where it does occur, it will be driven by food security goals, but improved soil quality will be a welcome co-benefit. Increases in SOC derived in this way are unlikely to deliver climate change mitigation because of the greenhouse gas emissions associated with N fertilizer, especially direct and indirect N 2 O emissions, though yield-scaled greenhouse gas emissions (i.e. emission per unit of agricultural product) may decrease (Linquist, van Groenigen, Adviento-Borbe, Pittelkow, & van Kessel, 2012). In addition, there are large CO 2 emissions associated with the manufacture of N fertilizer (Schlesinger, 2000), for example 3-8 t CO 2 -equivalent per t N under European manufacturing conditions (Brentrup & Palli ere, 2008). In regions where agriculture is well developed (e.g. Europe, North America), N fertilizer is already widely used, often at near-optimum rates, so there is little opportunity to achieve increased SOC stocks through this mechanism. In regions where N is frequently overused (e.g. China) the overriding aim is to decrease applications to decrease water pollution and greenhouse gas emissions (Zhang et al., 2013): it is clearly not acceptable to justify excessively large N applications on the grounds of increasing SOC.

| Practical limitations to achieving SOC
increases of 4 per 1000 compared to initial SOC stock in agricultural soils Any policies or initiatives, such as "4 per 1000", that lead to increased SOC and remove CO 2 from the atmosphere are to be welcomed. Even small increases in SOC are likely to improve soil quality and functioning for sustainable crop production, ecosystem services or both. But only in some cases will there be a concurrent benefit for climate change mitigation. Results from the long-term experiments reviewed here demonstrate several land management practices that can increase SOC stocks, often at rates well above the 4& per year rate (in the 0-40 cm depth). However, as discussed above, there are many situations where these practices cannot be widely adopted, either because they are impractical for farmers or because of wider global issues. The limitations influencing widespread adoption vary between different regions of the world but mostly fall into one or more of the following categories: 1. Farmers do not have the necessary resources available, e.g. insufficient manure due to lack of animals or insufficient crop residues because they are required for other purposes, e.g. smallholder farmers in Africa (Thierfelder et al., 2013;Giller et al., 2015).
2. The practice is already widely used, giving limited scope for achieving increased SOC accumulation through greater adoption.
An example is the return of crop residues to soil in many situations in Europe and North America. For these reasons, we conclude that promoting the "4 per 1000" initiative as a major contribution to climate change mitigation is unrealistic. Of course, any contribution to mitigating climate change is beneficial but the reasoning we have set out, and that of Baveye et al. (2018), strongly indicate that the magnitude is far smaller than claimed, or implied, by the proposers of the initiative. We suggest that a more logical rationale for promoting practices that increase SOC is the urgent need to preserve and improve the functioning of soils, both for sustainable food security and wider ecosystem services. As discussed earlier, there is significant evidence that even small increases in SOC can have disproportionately large impacts on a range of soil physical properties so even very partial success in meeting the "4 per 1000" target can be highly beneficial in this respect. We recognize that this is a more nuanced argument with less obvious political appeal but it is more firmly based on sound evidence. It is also in line with attempting to meet the UN Sustainable