Bradley et al. [5] report the distribution of soil carbon in England and Wales. They give data from which the average density of carbon in the subsoil (30–70 cm) can be deduced as 0.5% C by weight. A typical C : N ratio for soil is 10 or less and so this implies very crudely that the average density of N in the subsoil is 0.05%, or half the threshold of (0.1% N) that Sylvester-Bradley et al. [6] suggest might contribute significant amounts of N through mineralisation to the soil nitrogen supply to crops. Without a reported distribution or other measure of the variation it is difficult to state the prevalence of soils containing more than 0.1% N in the UK, but it is unlikely to be greater than 25% of all soils. On the other hand a soil containing 0.05% N in the soil volume from 30 to 100 cm may supply up to a third of the N contained in the topsoil and so about 20 or 30 kg N ha^-1 yr^-1, which would probably be leached if not captured by a crop. Such N is not currently accounted for in advisory systems. Where such N is available to crops, nitrogen application could be reduced by the equivalent amount.

Recent research in New Zealand and Australia has shown that nitrification inhibitors (NIs) can be extremely effective at reducing N₂O emissions from intensively grazed pasture [7]. If the nitrification of ammonium N is prevented, less leaching results and less N₂O is emitted during both nitrification and the subsequent denitrification of the resultant nitrate (NO^-₃). NIs applied to urine patches have been found to reduce N₂O emissions by 61–91%, to reduce NO^-₃ leaching by 30–79% and increase annual pasture yield by 0–36% [8,9,10,11,12,13]. Rates of application were of the order of 10 kg DCD (Dicyandiamide) ha^-1 at a cost of about £26 ha^-1. Research is currently underway in the UK and the Republic of Ireland to attempt to verify these findings for conditions in the British Isles.

Ammonia emissions can be reduced by the use of urease inhibitors [13] and as a consequence indirect emissions of N₂O can be reduced. There is some evidence to suggest that urea is more suited for use on soils prone to waterlogging (because it is primarily lost in dry conditions), whilst ammonium nitrate is more suited to less wet soils (because it is primarily lost under wet conditions) [14]. Although NIs were effective at field capacity in this study, they were ineffective under waterlogged conditions.

Delays and manipulation of delays in mineralization of N from crop residues by means of added material having a wide C : N have been reported [15] as has the effect of clay in soil on temporarily stabilizing N [16]. Better understanding of the mechanisms underlying these phenomena might lead to interventions to retain nitrogen in soil in organic forms until such time as a growing crop could take it up.

There is evidence that N₂O production increases at a non-linear rate when soil NO₃^- content exceeds crop demand. This has been reported under maize [17,18] and spring barley [19]. Under these conditions, and when conditions are also conducive to denitrification, splitting fertilizer N applications could reduce N₂O losses. Data presented by Zebarth et al. [19] suggest that lack of crop N uptake accounts for the relative increase in the proportion of mineral N that is emitted as N₂O at high rates of N application. Dalal et al. [20] state that NO₃^- usually inhibits full denitrification of N₂O to N₂, increasing the N₂O:N₂ ratio. This lends weight to the view that splitting applications should reduce losses and that correct N fertiliser guidance is imperative.

If efficiencies can be found throughout the system, less land need be used, which reduces the environmental footprint of farming as a whole. Alternatively, farmed land might be made more productive. Bulson et al. [21] found a 29% increase in yields from barley and beans by intercropping at 150% total density compared with sole crops. Not only does this increase diversity (see F below) but less land is needed to produce the same amount of food, thus potentially increasing the supply of valuable ES on land no longer in arable production [4].

Removing constraints on productivity is likely to increase yield profitability; it may well also affect the shape of the response curve, shifting the optimum in the direction of either less or more inputs. Careful use of growth regulators and fungicides with growth regulation properties, and maintaining the correct amount of N in soil for sufficient but not excessive growth, have been found to increase oilseed rape yields by 0.36 t ha^-1 [22]. Lynch [23] asserts that crops tend to develop roots in the surface soil for nutrient acquisition (particularly P) and in the subsoil for water. Ho et al. [24] suggest that plants which increase the density of their roots in the topsoil are able to acquire P more effectively than those which do not. They argue further that genotypes with roots that proliferate deep in the soil are better able to withstand drought stress but also found dimorphic varieties that are able to adapt to either nutrient or water stress as appropriate. Waines and Ehdaie [25] conclude that the root systems of modern wheat cultivars are small, having perhaps two thirds of the root mass of the landraces from which they derive. Whitmore et al. [26] argue that the physical impediment to root growth is a more significant stress than the lack of accessibility of water. There thus seems scope for breeding to improve nutrient acquisition and to reduce other constraints to maximum yield and nutrient uptake such as water stress.

If water or root condition is limiting yield [27], removing the constraint(s) and allowing the crop to yield to its full potential should improve nutrient use and benefit environmental quality by reducing losses and emissions. A further benefit of removing constraints to yield is that the farmer gains profit and so there is a strong economic pull from this direction, provided the action needed to remove the constraint is not excessively costly compared with the gain in yield. Because the costs of environmental pollution generally fall over the wider environment, measures to reduce nutrient loss have often been delivered through support (e.g. the EU single farm payment).

The benefit of improved extension can also be seen to act by moving the system towards the optimum production level (Figure 1). Removing constraints to crop growth increases profit and increases the ratio of profit to environmental cost. If the economic and environmental optima are not precisely aligned [1], advice to minimise deterioration of environmental quality will need to be focused on the environmental rather than economic optimum. This can save the cost of fertilizer and as well as environmental pollution. Dailey et al. [28] suggest that improved weather forecasts will make small (6 or 7 kg N ha^-1) but systematic improvements to N offtake and yield where N applications are constrained by regulation such as in nitrate vulnerable zones. Recent research in NW Europe on the effectiveness of extension has tended to focus on meeting air and water quality standards (e.g. [29]) but the advice itself is often based on achieving the economic optimum [30]. Goulding et al. [31] report that nitrogen use efficiency can be 60–90% on experimental plots but sometimes as little as 20–50% on cereal farms; this gap might be bridged by improved extension or outreach from researchers.

If N is uniformly applied at a rate necessary to obtain the maximum or optimum yield, given the existence of spatial variability [32] reduction of N use should reduce losses, but yields cannot be expected to improve detectably. There may be factors other than N supply that limit yield. If so, the simplistic approach of applying more N where growth is poor may be ineffective in increasing yield and risk environmental losses of N. Thus where nutrients cannot be captured by a crop, it makes environmental sense to apply less. It does seem likely, however, that N₂O emission and NO^-₃ leaching can be reduced by applying N in a spatially variable way, since several studies have found no loss of yield with a reduction in total N applied (e.g. [33,34,35]). Savings in these terms have rarely been identified in the literature but could be substantial in global or national terms if most emissions derive from the under-use of applied N by crops (perhaps 1–2 kg N₂O-N ha^-1 or 50 kg NO^-₃-N ha^-1 y^-1.)

Even arable land provides ecosystem services. If a way could be found to increase the services or value of the services provided by arable land then the net environmental cost of farming would be less. For example, Whitmore and Schröder [36] showed that intercropping could reduce nutrient use and nitrate leaching by 20 – 40 kg N ha^-1, without loss of yield or profit. Some measures have multiple effects. For example, increasing organic matter levels can increase soil fertility and, provided increased N supply from soil is taken into account, applications of fertiliser N can be reduced [37]. In addition, Watts et al. [38] found that increasing organic carbon in soil by 0.1% could reduce the specific plough draught by 5 kPa, which represents a fuel saving of more than 5%. Arable land that helped control flooding could make a large and valuable contribution to both the full economy of the UK [39] and to environmental quality downstream. Clearly not all land can perform all services simultaneously, but UK arable production managed as a whole might, even if not all in the same place. Multi-functional land-use at the landscape scale is therefore a means to improve environmental quality as a whole while continuing to farm the food we need to feed ourselves. However, it is important to consider the results of measures to ensure that increasing an ES such as carbon sequestration does not inadvertently lead to unexpected increases in a pollutant such as N₂O emissions [40].