Introduction

Land is under increasing pressure to deliver a range of competing outputs, and is critical to addressing society’s greatest challenges, such as biodiversity loss, the climate crisis, and food security1,2,3,4. Some land uses, such as the restoration of ecosystems aimed at mitigating or adapting to climate change (a type of nature-based solution, NBS), can reconcile competing land use demands by simultaneously tackling multiple challenges5. NBS can support biodiversity conservation and provide carbon storage and sequestration, as well as producing a range of other beneficial ecosystem services6,7. NBS are widely recognised as one effective and efficient route to address these great challenges8,9,10. However, which combinations of NBS interventions are implemented, where they are deployed, and in what quantity, presents a considerable conundrum, with different choices producing unique outcomes, co-benefits, and trade-offs11,12.

Land use scenario modelling with multi-criteria evaluation plays a crucial role in identifying integrated solutions to delivering multiple public and private good outcomes, by providing a quantitative approach to exploring the complex interactions between different land uses and management strategies13,14,15. Through simulating future scenarios of alternative land use and management practices, a range of potential economic, environmental, and social outcomes (for instance) can be evaluated, with co-benefits and trade-offs identified16. In recent years, indicative scenario modelling has become increasingly popular in evaluating the potential for NBS-focussed land use change to address climate change mitigation, and the consequent impact on food production and other outcomes15,16,17,18,19. The outcomes of scenario modelling have informed policy and been used to advise on land management, by finding where best to implement NBS measures to provide the greatest potential for climate change mitigation at least cost to food production20,21,22.

While scenario modelling has produced valuable conclusions, such exercises have typically explored a limited range of the option space of mitigation measures and management practices that could be implemented e.g13,19,23,24. This could be due to the complexity and resources required to run many (often spatially explicit) pathways15. Furthermore, the limited acceptance, slow uptake, or practical and social implausibility of some combinations of NBS or new land management practices may limit the ambition to explore a wider range of alternative futures25,26. However, this limited option space could lead to prioritisation of sub-optimal solutions or fail to identify the most impactful pathways to achieve specific goals, such as achieving net zero whilst minimising impact to food production or reversing biodiversity loss e.g27,28,29. More recently, scenario discovery, a commonly used method to develop future scenarios with foundations in the fields of computer science and uncertainty analysis has begun to emerge for modelling future land use scenarios by allowing a large number of scenarios to be explored, see Duffy et al.30, Navarro Garcia et al.31, Khan et al.32.

Here, using the United Kingdom (UK) as an example of a densely populated country with limited land and multiple competing interests, we investigate a ‘semi-spatial’ approach to land use scenario modelling based on the principle of scenario discovery. Our computationally efficient method, using land cover functional units (areas of land which share environmental characteristics) does not require the computationally intensive spatial outputs required by spatially explicit scenarios. We generate 10,000 pathways that deploy randomly selected combinations and quantities of eight land-based climate change mitigation measures (upland and lowland peatland restoration, woodland creation, wood pasture creation, semi-natural grassland creation, silvoarable and silvopastoral agroforestry, and organic farming) in addition to fixed quantities of urban expansion, onshore renewables, hedgerow creation and low carbon farming practices. For each of the eight varying measures, we identify which types of land are suitable (i.e. the opportunity area), with each pathway deploying each measure across a randomly sampled fraction of the corresponding opportunity area. Land use change occurs on lower grade farmland first, with the most productive farmland being protected under all but the most ambitious pathways. We then examine the ability of each pathway to deliver the ‘triple challenge’ of nature conservation (for which we use breeding bird habitat availability as a proxy), food production (i.e. domestic production, regardless of whether consumed domestically or abroad, and ignoring imported food), and net zero greenhouse gas (GHG) emissions from Agriculture, Forestry and Other Land Use (AFOLU; excluding emissions from agrochemical manufacture and displacement of fossil fuels or construction materials by bioproducts) in 2050.

To test the efficacy of our approach, we identify pathways which are superior to the best performing scenario in previous work, see Finch et al.15; NBS Extra, which surpassed net zero (102.6% reduction in GHG emissions) whilst providing an increase in breeding bird habitat index of 18.1% but a reduction in food production of 20.8% by 2050, compared to 2015 levels. We also identify the 20 pathways which achieve a net zero land sector at lowest cost to food production, separately for two spatial and governance scales – the United Kingdom as a whole, and it’s four constituent nations (England, Scotland, Wales and Northern Ireland) to which most land use policy decision-making is devolved. In doing so, we seek to understand the consequences for efficiency and potential equity of decision-making at different scales.

Results

Trade-offs and co-benefits

As expected, our modelling reveals a trade-off between GHG reduction and food production, but a synergy between GHG emissions and breeding bird habitat availability (Fig. 1). Of the 10,000 pathways, 281 (2.81%) were discarded as they produced insufficient feed from arable land to supplement the diets of grazing livestock. Hereafter we present the results for the remaining 9,719 pathways. 45.1% of these delivered a net zero UK land sector by 2050, and 21.5% delivered a GHG reduction of > 150%. Whilst there was a strong negative correlation between GHG emissions and food production (r = -0.70, df = 9,717, p = < 0.01; Fig. 1A), this correlation was characterised by a high degree of spread, such that the reduction in food production among net zero pathways (i.e. GHG reduction of 99–101%; 1,125 pathways) ranged from 10 to 80%. Similarly, the impact on the breeding bird habitat index among net zero pathways was wide-ranging, despite the significant correlation indicating a synergy between these outcomes (r = 0.73, df = 9,717, p = < 0.01; Fig. 1C). The trade-off between breeding bird habitat index and food production was weaker (r = -0.39, df = 9,717, p = < 0.01; Fig. 1B). See Figures S1, S2 and S3 for results of independent measure deployment on birds, food, and GHG emissions.

We identified 328 pathways which were superior to NBS Extra (Finch et al., 2023) with respect to both GHG emissions (lower or equal) and food production (greater or equal), highlighting the benefit of our streamlined workflow. Whilst NBS Extra saw a 20.8% reduction in food production, we identified pathways which reduced production by as little as 11.8% whilst still reducing GHG emissions by as much or more than NBS Extra. On the other hand, only 10 pathways were superior to NBS Extra with respect to all three outcomes (i.e. lower or equal GHG emissions, greater or equal food production and greater or equal breeding bird habitat availability), suggesting that NBS Extra was already close to Pareto-optimal across these three outcomes. Constraining pathways to those which were superior to NBS Extra in terms of bird habitat availability also increased the minimum reduction in food production to 18.6%, though this pathway went well beyond net zero, reducing emissions by 127% (cf. 102.6% under NBS Extra).

Fig. 1
Fig. 1
Full size image

Pairwise relationships across pathways between (A) AFOLU GHG emissions and food production, (B) bird habitat index and food production, (C) AFOLU GHG emissions and bird habitat index. Each point represents one pathway, with grey indicating pathways that are inferior to NBS Extra, see Finch et al.15, green indicating pathways that were superior for food production and GHG emissions, and blue indicating the pathways that were superior with respect to all three outcomes (food production, GHG emissions, and breeding bird habitat). Horizontal and vertical dashed lines represent 0% change in net GHG emissions, in food production, and bird habitat index.

Given the trade-offs between food production and both GHG reduction and bird habitat availability, identifying the ‘best’ pathway depends on how the different competing outcomes are valued. 264 pathways are Pareto-optimal (i.e. no outcome can be improved without impairing another), but this falls to 51 when constraints are applied such that net zero is achieved without reducing food production by more than 20%.

Extent of land-use change

High-performing pathways combine multiple mitigation measures across a wide range of ambition levels (Fig. 2), reiterating the fact that there is no single solution to the ‘triple challenge’, but rather a range of potential pathways. Compared to the wider set of pathways which were superior to NBS Extra with respect to GHG emissions and food production, the 10 which were also superior in terms of bird habitat availability tended to involve more woodland creation and lowland peatland restoration, but less upland peatland restoration, wood pasture creation, semi-natural grassland creation, agroforestry, and organic farming. Woodland creation dominated in absolute terms, though did not exceed 40% of the total opportunity area. When comparing NBS Extra to the average change across these 10 superior pathways, NBS Extra had less woodland creation (1.7 Mha with an additional 0.2 Mha of open ground within woodlands for NbS Extra vs. 2.82 Mha), similar amounts of wood pasture (0.5 vs. 0.54 Mha), and more semi-natural grassland creation (0.45 vs. 0.31 Mha) and peatland restoration (100% vs. 87% for lowland and 43% for upland).

Fig. 2
Fig. 2
Full size image

Range in ambition levels across each measure among pathways which are considered efficient. (A) Shows realised ambition expressed in area (Mha); (B) Shows realised ambition as a percentage of the opportunity area. Green points reflect pathways that were superior to NBS Extra see Finch et al.15, for food production and GHG emissions, and blue reflect pathways that were superior with respect to all three outcomes (food production, GHG emissions, and breeding bird habitat). Large, coloured points indicate the mean, per group.

Efficiency versus equity

On average, the top 20 pathways which delivered a net zero land sector at lowest cost to food production saw a 12.5% reduction in food production at the UK scale. However, the aggregate reduction in UK production rose slightly to 14.7% under pathways which delivered a net zero land sector for each individual devolved nation (Fig. 3A). In other words, decision-making is slightly more efficient at the larger spatial scale, because interventions can be prioritised to areas of lowest opportunity cost. However, these ‘UK-efficient’ pathways resulted in considerable inequities in terms of the impact on food production in each devolved nation, which, when averaged across the top 20 pathways, ranged from 9.6% in England to 37.5% in Northern Ireland and 38.9% in Wales (Fig. 3B). In contrast, pathways which delivered net zero for each devolved nation individually resulted in more equitable impacts, with the reduction in food production ranging from 14% in Scotland and 14.3% in England to 26.9% in Northern Ireland (Fig. 3B). Compared to UK-efficient pathways, these ‘devolved-nation-efficient’ pathways saw greater reductions in food production in Scotland, Wales and Northern Ireland, but smaller reductions in England.

Fig. 3
Fig. 3
Full size image

Change in food production at UK and devolved nation level for the 20 pathways which achieve a ≥ 100% reduction in GHG emissions by 2050 at the least cost to food production. (A) Impact on UK wide food production under UK- and devolved-nation-efficient pathways. Large points, values, and coloured lines represent the mean. (B) Impact on devolved nation level food production under UK- and devolved-nation-efficient pathways. Large points represent the mean of the top 20 pathways. Dashed line indicates the median across devolved nations. UK-efficient pathways achieve a ≥ 100% reduction in UK emissions at least cost to UK food production, whereas devolved-nation-efficient pathways achieve a ≥ 100% reduction in an individual devolved nation at least cost to that nations’ food production.

Discussion

Here, we identify multiple pathways to achieving a net zero UK AFOLU sector, with often striking differences in the consequences for predicted bird habitat availability and food production. Pathways which simultaneously achieved multiple potential policy targets required the deployment of multiple mitigation measures, though not in equal quantities.

The ca. 10,000 pathways we modelled revealed trade-offs and co-benefits between birds, food production, and GHG emissions, reinforcing the findings from previous research14,15. We show that reducing GHG emissions from AFOLU can provide a co-benefit to breeding bird habitat availability (Fig. 1C), aligning with the concept of NBS7. This finding supports previous work which has demonstrated a positive relationship between biodiversity and carbon at various scales, such as, Strassburg et al.1, Finch et al.13, Smith et al.14, Finch et al.15. Furthermore, Deng et al.33 highlights that ecosystem benefits are the most studied co-benefit of GHG emissions mitigation.

In contrast, we show a clear trade-off between reducing GHG emissions and maintaining food production (Fig. 1A). Critically though, we show that this trade-off is noisy, with a wide range of impacts on food production for a given reduction in GHG emissions. This implies that some pathways are more ‘efficient’ than others, and our modelling approach identified 328 pathways which delivered a greater or equal reduction in GHG emissions than NBS Extra15 at the same time as a smaller or equal reduction in food production. The most efficient net zero pathway resulted in a 43% smaller reduction in food production than NBS Extra. Unfortunately, most of these 328 pathways deliver a smaller increase in bird habitat availability. We identified only 10 pathways which were superior to NBS Extra across all three outcomes, and the most efficient of these resulted in only a 22.5% smaller reduction in food production than NBS Extra. More targeted sampling of the option space in the vicinity of these superior pathways may, however, lead to even greater efficiencies. To avoid offshoring of environmental impacts via additional food imports, any shortfall in domestic production should be mitigated through a combination of demand- and supply-side measures, including waste reduction, yield gap closure, and shifts towards less land-demanding food products15,34,35.

Achieving UK AFOLU net zero whilst minimising impacts to food production and increasing breeding bird habitat availability requires changes to the way in which land is used. Among our 10 ‘superior’ scenarios (blue points in Figs. 1 and 2), the average area of land-use change was 6.4 Mha, falling to 3.9 Mha when excluding measures which involve limited change such as upland peatland restoration, agroforestry, and organic farming, and to 3.1 Mha when additionally excluding measures which still involve farming such as wood pasture and semi-natural grassland. These quantities equate to 26%, 16% and 12% of the UK’s land area, respectively, meaning that the majority remains unmodified. Not all measures were equally favoured under these scenarios, with woodland creation and lowland peatland restoration achieving the highest relative ambition levels (i.e. as a proportion of their maximum opportunity), due to their per-unit-area efficacy in providing GHG emissions reductions and increases in bird habitat availability (Figures S1 and S3). In contrast, measures such as silvoarable agroforestry, semi-natural grassland creation, and organic farming deliver relatively smaller increases in bird habitat availability (Figure S3) and incur a bigger loss of food production for a given reduction in GHG emissions (Figure S2), so tend to be deployed in smaller quantities. The relative importance of different measures changes depending on how different objectives are prioritised and is also likely to be sensitive to particular groups of birds (e.g. farmland waders) being given greater weighting due to variation in habitat preferences. Despite our models favouring high levels of peatland restoration and woodland creation to achieve AFOLU net zero, maintaining a diverse mix of habitats is essential for biodiversity including habitats not modelled here, such as heathland, scrub and habitat mosaics.

These 10 ‘superior’ pathways combined multiple mitigation measures but were characterised by particularly large areas of woodland creation, averaging 2.9 Mha. This equates to 12% of the UK’s land area, or an 86% increase compared to current woodland cover. Annualised between 2020 and 2050, this implies a woodland creation rate of more than 95 kha/yr, which is three-fold higher than the target the UK government is currently failing to meet, and almost double the 50 kha/yr target required by 2035 in an illustrative Net Zero Strategy36. This quantity of land-use change is therefore extremely ambitious, though we also identify pathways which achieve AFOLU net zero with as little as 22 kha/yr of new woodland. In addition to woodland, superior pathways also involved a combined area of wood pasture and agroforestry totalling almost 1.2 Mha on average. Again, this figure is substantially higher than recommended in, for example, the CCC’s sixth carbon budget pathways (0.14–0.34 Mha new agroforestry). Part of this discrepancy is likely due to the CCC pathways resulting in 0.35–4.60 Mha of ‘surplus’ land, which is released from agriculture (and so avoids agricultural emissions), but which is put to no other use. In contrast, our pathways assign an explicit cover or use to all land. In addition, three of the five CCC pathways result in a net emission of between 8 and 26 Mt CO2e by 2050, so do not actually deliver a net zero land sector. Our ‘superior’ pathways typically involve more lowland peatland restoration but less upland peatland restoration than CCC pathways, though for both measures the range of ambition is very wide. Organic farming occupied 1 Mha on average, or 8% of remaining farmland.

By comparing pathways which achieve net zero ‘efficiently’ (with respect to food production) at either the UK-scale or for the individual nations within the UK, we highlight an additional trade-off, between efficiency and equity. Achieving a UK-wide net zero AFOLU sector at least cost to food production involves relatively small impacts on England compared to the devolved nations of Scotland, Wales, and Northern Ireland. This is primarily explained by England having a disproportionately large area of high-grade farmland which is protected from land-use change where possible by our model constraints. In contrast, achieving AFOLU net zero within each individual nation increases the UK-scale reduction in food production slightly, but results in more equitable impacts among the four devolved nations.

As with any modelling exercise, our approach to scenario discovery is subject to uncertainties and limitations, though we sought to minimise these where possible. These uncertainties, and our approach to validation and calibration, are set out in Finch et al.15. An important limitation is the fact that we only consider a limited range of mitigation measures, though other solutions are available to address the climate and biodiversity crises whilst ensuring food security. There are more nuanced and regional measures to be taken see Jones et al.37, such as changes to grazing management38 or changing cropping patterns, illustrated by Lloyd et al.39, which would likely provide a different set of results. However, without robust evidence for the impacts of such measures on breeding birds, food production, and GHG flux, we are unable to parametrise these measures within our modelling framework. An additional limitation of our semi-spatial approach is that, without spatially-explicit outputs, the bird habitat availability model cannot capture landscape or compositional effects. We discuss the consequences of this for comparison with Finch et al.15 in the methods (see also Table S3 and Figure S4).

There are multiple pathways to achieve a net zero AFOLU sector in the UK and its constituent nations, and our semi-spatial approach to scenario discovery identifies pathways which minimise trade-offs and maximise co-benefits Whilst there is space for multiple solutions to be deployed in tandem, measures such as woodland creation and lowland peatland restoration are particularly important. It is important to reflect that the inevitable social impacts of land use change (from employment and economics to culture and human-nature connectedness) need attention alongside the biophysical impacts considered here. Scenario modelling is only one tool to be used in support of land use decision-making: other approaches, especially those which engage with local communities, as per NbS guidelines7,40 are also necessary.

Methods

Land cover functional units

We developed a ‘semi-spatial’ approach based on the land cover functional unit (LCFU) concept. LCFUs are defined as areas of land with homogenous environmental characteristics such as land cover, climatic conditions, and elevation41, to divide and characterise areas. These LCFUs need not be spatially congruent but are instead defined by shared characteristics. We generated an LCFU table for the UK by cross-tabulating 11 high-resolution spatial data layers (Table S1). Each unique combination of attributes across these 11 layers represents one of 168,864 LCFUs (cf. 211.2 million 25-m pixels). Additional columns were generated in the LCFU table, derived from combinations of the original spatial data values (Table S1), to identify opportunities for each of eight climate change mitigation measures considered in this study. Opportunity for mitigation measure deployment varied and was based on differing constraints: Woodland creation could occur on arable and horticultural land or improved grassland, on non-peat soils, outside of wader hotspots and within the woodland opportunity area identified by Bradfer-Lawrence et al.6; wood pasture was similar, with the exclusion of arable and horticultural land and inclusion of semi-natural grassland land covers; opportunity for agroforestry creation matched that of woodland, but with the land cover constraint defining agroforestry type: silvoarable (arable and horticultural land only) or silvopasture (improved grassland only); semi-natural grassland conversion could only occur on arable and horticulture or improved grassland, as could organic farming; peatland restoration only occurred on peat soils with arable and horticulture and improved grassland land covers for lowland peat, and bog or fen land covers for upland peat. LCFU suitability for each of the measures was designated by truth values (TRUE or FALSE). To deploy each mitigation measure, we modified the land cover of part or all of each LCFU for which the corresponding opportunity score was TRUE. This tabular approach to land cover manipulation allows the exploration of a much larger number of pathways than a fully spatial approach and is detailed in Fig. 4.

Fig. 4
Fig. 4
Full size image

Research workflow. Outline of steps taken, from spatial data to land cover functional unit table creation, followed by pathway creation, modelling and evaluation of outcomes.

Pathways

We generated 10,000 ‘randomised’ land use pathways, each deploying eight climate change mitigation measures (organic farming; peatland restoration, both upland and lowland; semi-natural grassland creation; woodland creation (66% of the total as broadleaved, 33% as coniferous); silvoarable agroforestry; silvopastoral agroforestry; wood pasture) in unique combinations and quantities. Each pathway was produced by drawing a random percentage value for each mitigation measure representing the level of ambition, where 0 indicated no change, and 100 indicated change across the entire corresponding opportunity area. Random ambition values were chosen due to their efficiency in covering option space when compared to using a grid-based approach42. Ambition values of all measures except peatland restoration were drawn from a non-uniform beta distribution (α = 0.7, β = 3) to reduce the number of pathways with unfeasibly high ambition across multiple measures, which arose due to the different measures having overlapping opportunity areas. The target area of land for each measure was then calculated as the product of the level of ambition and the total size of the opportunity area (see Table 1). For simplicity, we deployed low carbon farming practices, such as behavioural and innovative measures, improved nitrogen use efficiency and electrification of machinery and power, as per Finch et al.15 and hedgerow creation at constant ambition across all pathways, corresponding to ‘High’ ambition in Finch et al.15.

Table 1 Examples of how the opportunity area corresponds with ambition level and the area of deployment for three mitigation measures.
Full size table

In addition to the nature-based climate mitigation measures, our pathways also deployed urban expansion and onshore renewables (wind farms and solar parks). The rate of urban expansion was based on the projected 2050 urban area24 equating to an increase of 23.2% on the current urban area. New urban areas replaced farmland adjacent to existing built land. Next, onshore wind and solar were installed at the CCC Widespread Innovation 2050 capacities43 of 35 GW and 90 GW, respectively, based on the opportunity maps created by Copping et al.19 and assuming capacity densities of 5 MW/km2 for wind and 45 MW/km2 for solar43. The quantity of urban expansion and onshore renewables was fixed across all pathways and constrained the area available for new nature-based climate mitigation measures. Combined, urban expansion and onshore renewables occupied 449 kha, ~ 2% of the UK’s land area (24,361 kha).

We updated the opportunity area for each measure to account for the deployment of new urban land and onshore renewables, then deployed nature-based mitigation measures in a pre-defined order, based broadly on their expected GHG mitigation potential per unit area: peatland; woodland; wood pasture; grassland; agroforestry (silvoarable then silvopasture); organic farming. The relative order of different measures is only relevant for pairs of measures with overlapping opportunity areas. To deploy mitigation measures, we modified the land cover attribute of LCFUs within the opportunity area corresponding to each measure. We prioritised change in LCFUs representing low grade farmland first, moving through the Agricultural Land Classification grades (ALC) in order to preserve the best and most versatile land, if possible. Any LCFU undergoing modification was removed from the opportunity area of other measures.

For each pathway we then predicted the impact of land cover change on breeding bird habitat availability, food production, and GHG emissions, following the modelling approach developed by Finch et al.15. We present 10,000 explorative pathways with random levels of ambition covering the option space across the eight mitigation measures, with results representing the year 2050.

Food production

We utilised the food production evaluation model developed by Finch et al.15. Our model estimated total agricultural production (in kcal) from land for each pathway in 2050. We included all crop and livestock production, including those which are fed imported crops. We do not distinguish between products grown for export versus domestic consumption. We derived per-hectare rates of food production per LCFU by allocating reported production of crop and livestock products according to land cover, Agricultural Land Classification, and NUTS1 region (Nomenclature of Territorial Units for Statistics). We converted total production into kcal according to product-specific nutritional information44,45.

For agroforestry and hedge creation we eliminated grass and crop production from the direct footprint of trees, hedges, and scrub, and for silvoarable agroforestry, we applied an additional yield penalty of 29% to the crops grown between trees to account for likely shading effects, based on van der Werf et al.46, Graves et al.47. For organic farming we applied yield penalties specific to different arable crop types and to grassland48,49, and also specified a minimum area of 15% temporary grassland and 10% legumes (beans and peas) on arable land, reducing the fraction of other crops pro rata. Land with wind farms retained its previous land cover class, but with the footprint of turbines removed. This equates to 2% of a given area50,51. Land converted to solar parks was considered improved grassland and the footprint of solar panel infrastructure, equating to 5% of a given area was removed52. We assumed that low-carbon farming practices had no impact on food production, except where changes in crop rotations influenced the ratio of different crop outputs.

GHG emissions

We utilised the GHG emission evaluation model developed by Finch et al.15. The model estimated the total net flux of GHGs from the land sector in 2050 for each pathway. We express emissions in terms of CO2 equivalents (CO2e), using IPCC AR4 100-year global warming potential values (GWP100 = 25 for methane, 298 for nitrous oxide)53.

To estimate changes in net GHG emissions from peatlands, we used emissions factors and condition estimates from Evans et al.54. For woodland, we first converted annual estimates of standing biomass and debris into annual age-specific increments. For biomass removed through rotational clearfell, we calculated decay profiles following Morison et al.55. We assumed that all coniferous woodland was represented by Sitka Spruce plantations, thinned and harvested on a clearfell rotation, with rotation length of 50, 60 or 70 years depending on the yield class of each LCFU. We assumed 20% of broadleaved woodland was managed (thinned and clearfelled on a rotation of 40–50 years depending on yield class, represented by Silver Birch) and 80% unmanaged (neither thinned nor clearfelled, represented by Sycamore/Ash/Birch). For wood pasture, we assumed additional sequestration of -0.37 t CO2/ha/yr from scrub for the first 15 years after establishment, reflecting a final carbon density of 36.7 t CO2/ha56. For new hedges, we assumed that total biomass (above and below ground) reaches 128.3 t CO2 ha− 157, after 20 years, implying an annual flux of -0.95 t CO2 km− 1 (assuming 1.5 m width) for the first 20 years following hedge creation, after which we assumed no additional net sequestration due to regular management. For agricultural emissions, we included sources as defined by IPCC guidance, as well as emissions from agricultural activities (including feed imports, fertiliser and pesticide manufacture, and machinery manufacture and maintenance). We scaled reported 2015 emissions in proportion to changes in the underlying driver of each emission source. Final emissions were scaled down based on our low carbon farming measure (see Finch et al.15 for details).

Finally, for land-use transitions we estimated the changes in soil carbon stocks according to expected changes in equilibrium of soil carbon density following the transition between woodland, cropland, grassland, and urban land57. We assumed instant changes in non-forest biomass as a result of land-use change, deriving equilibrium biomass carbon density for annual cropland, shrubby grassland, and non-shrubby grassland from Brown et al.57. For the deployment of renewable energy, the footprint of wind turbines and solar panels (2% and 5% of a given area, respectively) were considered a transition to urban land.

Breeding birds

Our approach to predict the effect of future land use on bird habitat availability differs slightly from Finch et al.15, though is based on the same underlying UK Breeding Bird Survey (BBS) data. Because our pathways are not spatially explicit, we rely on estimates of habitat-specific abundance which ignore landscape effects. This tends to result in less extreme responses (Figure S4). For the purposes of making comparisons with Finch et al.15, we re-run their pathways through our simplified bird habitat model to produce comparable estimates. For the scenario NBS Extra, which according to Finch et al.15 delivered an 18.1% increase in habitat availability, the comparable figure according to our model is 11.5%. We used BBS data to fit habitat-abundance models for each of 120 native terrestrial species (listed in Table S2) and used these models to predict expected abundance in each LCFU based on land cover. We produced two population estimates: a pessimistic estimate where no birds are present in areas converted to solar or wind farm, and an optimistic estimate, where there is no impact from renewables. The mean of the two estimates was then calculated. We chose this method due to a lack of data showing species’ responses to onshore renewables, and due to the minor differences between estimates demonstrated by Copping et al.19. We then estimated the total population size for each pathway based on 2050 land cover and calculated the geometric mean change in predicted population size from 2015 across all species. Our approach ignores limiting factors such as climate, disease, and persecution, therefore predicted changes should be interpreted as changes in carrying capacity of available habitat rather than population projections.