Competition for land

A key challenge for humanity is how a future global population of 9 billion can all be fed healthily and sustainably. Here, we review how competition for land is influenced by other drivers and pressures, examine land-use change over the past 20 years and consider future changes over the next 40 years. Competition for land, in itself, is not a driver affecting food and farming in the future, but is an emergent property of other drivers and pressures. Modelling studies suggest that future policy decisions in the agriculture, forestry, energy and conservation sectors could have profound effects, with different demands for land to supply multiple ecosystem services usually intensifying competition for land in the future. In addition to policies addressing agriculture and food production, further policies addressing the primary drivers of competition for land (population growth, dietary preference, protected areas, forest policy) could have significant impacts in reducing competition for land. Technologies for increasing per-area productivity of agricultural land will also be necessary. Key uncertainties in our projections of competition for land in the future relate predominantly to uncertainties in the drivers and pressures within the scenarios, in the models and data used in the projections and in the policy interventions assumed to affect the drivers and pressures in the future.


INTRODUCTION The UK Foresight Global Food and Farming Futures
Project is considering how a future global population of 9 billion can all be fed healthily and sustainably (Foresight 2009). The project has identified 19 'drivers' (with subcategories) affecting food and farming in the future, one of which is competition for land. The purpose of this review is to examine competition for land, and to consider the direct and indirect pressures and drivers affecting it. The scope of the review is global and the time scale considered is the past 20 years and the next 40 years (1990 -2050).
In addition to agriculture, use is included for forestry, non-food crops and protected areas for biodiversity, as well as use of land for bioenergy and land degradation/restoration. The impact of policy on influencing each of these factors is discussed in §2c.
We summarize the quantitative information on changes in land use and land quality over the last 20 years, both globally and disaggregated according to the major regions of the world. The most recent synthesis of agricultural land-use change was conducted for the International Assessment of Agricultural Knowledge, Science and Technology for Development (IAASTD), particularly the chapter 'Outlook on agricultural changes and its drivers', dealing with land-use and land-cover change (van Vuuren et al. 2008). That study collated projections from the Land use and cover change synthesis book (Alcamo et al. 2005), the scenarios from the Global Scenarios Group (Raskin et al. 2002), IPCC Special Report on Emissions Scenarios (SRES) (IPCC 2000), the Millennium Ecosystem Assessment (MEA 2005), UNEP's Global Environment Outlook (UNEP 2002) and some models from the EMF-21 study of the Energy Modelling Forum (e.g. Kurosawa 2006;van Vuuren et al. 2006). We expand on that synthesis by adding more recent studies in § §4 and 5. In these sections, we present projections of land-use change to 2050 and examine the impact of changes in non-agricultural policy (e.g. forest and protected land policy) on competition for land. We briefly examine the assumptions upon which the projections are based and identify the main areas of uncertainty.
We conclude by assessing and ranking the most important external factors that may affect the land available for agriculture between now and 2050, and by discussing future needs to reduce uncertainties in these projections.

FACTORS AFFECTING COMPETITION FOR LAND
Although competition for land has been identified as a driver affecting land use, food and farming by the Foresight Global Food and Farming Futures Project, it is actually an emergent property of a range of other drivers and pressures. Figure 1 presents a conceptual framework for analysing drivers and related pressures of competition for land at different geographical scales.
In understanding interrelated causes for competition for land, we distinguish between drivers and pressures. Pressures represent direct causes, the visible motivations for competition for land (right-hand side of figure 1). Drivers (underlying causes) for competition are factors of higher causal order that determine the degree of the actual direct pressures (left-hand side of figure 1), (see Chomitz & Gray 1996;Kaimowitz et al. 1998;Geist & Lambin 2002;Wunder 2003;Niesten et al. 2004;Rudel et al. 2005, on these different drivers and pressures; S. Klappa 1999, unpublished data).
We do not attempt to review the drivers and pressures in detail here, since they are covered by the other driver reviews in this issue. In §2a, however, we discuss a few drivers and pressures to demonstrate how they impact upon land use through their impact on competition for land.
(a) Population growth, agricultural intensification and dietary preference The growth in the human population from about 3 billion in 1960 to 6.8 billion in 2010, coupled with increased income and changes in diet, has been accompanied by substantial increases in crop and animal production (2.7-fold for cereals, 1.6-fold for roots and tubers and fourfold for meat; Foresight 2009). This increase will need to be maintained if the projected population of 9 billion by 2050 is to be sustained. Past increases in crop production have occurred as a result of both extensification (altering natural ecosystems to produce products) and intensification (producing more of the desired products per unit area of land already used for agriculture or forestry). Of the world's 13.4 billion ha land surface, about 3 billion ha is suitable for crop production (Bruinsma 2003) and about one-half of this is already cultivated (1.4 billion ha in 2008). The remaining, potentially cultivatable, land is currently beneath tropical forests, so it would be undesirable to convert this to agricultural land because of the effects on biodiversity conservation, greenhouse gas emissions, regional climate and hydrological changes, and because of the high costs of providing the requisite infrastructure. Therefore, increased yield and a higher cropping intensity will need to be the main driver behind future growth in food production (Bruinsma 2003). Table 1 shows that, according to the projection of Bruinsma, extensification will still contribute significantly to crop production in Sub-Saharan Africa (27%) and Latin America and the Caribbean (33%). There is almost no land available for expansion of agriculture in South and East Asia and the Near East/ North Africa (and there may be loss of agricultural land to urban development) so that intensification is expected here to be the main means of increasing production (Gregory et al. 2002;Bruinsma 2003).
The main means to intensify crop production will be through increased yields per unit area together with a smaller contribution from an increased number of crops grown in a seasonal cycle. As cereal production (wheat, maize and rice) has increased from 877 million tonnes in 1961 to 2342 million tonnes in 2007, the world average cereal yield has increased from 1.35 t ha 21 in 1961 to 3.35 t ha 21 in 2007. Simultaneously, per capita arable land area has decreased from 0.415 ha in 1961 to 0.214 ha in 2007 (Foresight 2009). Put another way, had the increases in yield of the last 40-50 years not been achieved, almost three times more land would have been required to produce crops to sustain the present population; land that, as indicated above, does not exist unless unsuitable for cropping. Without changes in productivity, the growing population would have led to an even greater expansion in agricultural area than observed, and competition for land would have been greatly intensified.
There have also been substantial changes in human food consumption reflected in dietary and nutritional changes over recent decades (Schmidhuber 2003). There is an increasing demand for livestock products, particularly in developing countries (Smith et al. 2007), and given the lower efficiency of livestock products compared with the direct consumption of vegetal matter (Stehfest et al. 2009), an increasing proportion of livestock products in the diet is expected to increase competition for land.
(b) Non-food goods and services While agricultural production for food consumption is one of the predominant land-use activities across the globe, land is also used for the production of timber, fibre, energy and landscape amenities as well as being consumed by urbanization.
(i) Forest products and fibre Historically, the production of forest products has grown rapidly-and again, in the future a further increase is necessary (upto 2030 by 1.4% per annum for sawnwood, and 3% for paper and wood-based panels; FAO 2009a). But worldwide, the area of forest and woodland has decreased over the past decade (FAO 2009a,b;Foresight 2009), mostly at the expense of agricultural expansion. However, regional differences in forest areas and timber production are stark, with declines occurring in developing countries, but forest expansion in developed countries (table 2).
The different trends between developed and developing countries arise from a number of factors that reflect competition with other land uses.
-Wood substitutes. Developed countries have replaced the use of wood as a source of fuel and in construction. -Agricultural expansion. Demand for agricultural products has been growing only modestly in developed countries, but rapidly in developing countries. -Trade patterns. Developing countries tend to export primary products. -Public services. Forests have amenity value in developed countries, and subsequently are often protected from deforestation through policy.
While the area of forest in developed countries is increasing only slightly, demand for wood products has fuelled deforestation in other parts of the world. The global production of fibre crops has almost doubled between 1961 and 2007 (Foresight 2009), but the land area used to produce these crops has declined by about 10 per cent over the same period (FAOSTAT 2010). This reflects the increase in global consumption of fibre goods, but also the increased productivity (yields) of fibre crops. The area declines suggest that the competition between food and fibre production may be decreasing.  (ii) Energy crops The growth of crops for bioenergy has been highlighted as a potential competitor for land with food crops. It is noteworthy, though, that the area occupied by bioenergy and its by-products in 2004 was only 14 Mha compared with 1500 Mha of crops (i.e. about 1% of the total cropped area) and 4500 Mha of pastures worldwide (IEA 2006). While the reasons for growing crops for bioenergy are complex, the use of land for them is likely to increase in the future (FAO 2009b). An important issue for competition for land is the potential clearing of new land for biomass crops. Using biomass for energy is likely to have both positive and negative competitive effects on food production and therefore on land, with national and regional policies beginning to reflect differing components of these inter-linkages. With global oil stocks becoming increasingly threatened (UKERC 2009), fossil fuel prices will inevitably continue to rise and alternative sources of energy will be needed, not least to maintain agricultural yields. Bioenergy is likely to fill a significant part of this emerging energy gap for agriculture, which in turn will require more integrated energy/agriculture/land-use policies to circumvent adverse impacts of competition for land.
(iii) Amenity activities and biological conservation An increasing trend in some parts of the world is the use of land for amenity activities and/or biological conservation. This includes recreational uses such as public parks, golf courses and other sports facilities, as well as the conservation of traditional landscapes for their aesthetic, cultural or natural heritage value. Land competition between amenity and other uses depends strongly on geographical location, with stronger pressures for amenity use occurring on land near to urban centres. However, many cultural landscapes are multi-functional, being used, for example, for food or timber production, as well as offering amenity services. Setting aside land for amenity or conservation potentially increases competition for land on the remaining area, which we return to in § §4 and 5.
(c) Land and soil degradation Degradation of soil and land through inappropriate use or the addition of pollutants has been a topic of concern for many decades, because of the potential impact on biodiversity, and the availability of land for the human population to feed itself. Degradation of land intensifies competition for land, since it reduces the quantity of land suitable for a range of uses such as food production. ISRIC (1991) produced a world map of human-induced soil degradation based on the knowledge of 250 experts from six continents showing that of the 11.5 billion ha of vegetated land, 15 per cent was degraded. Erosion was the main process of degradation, and about 20 per cent of the agricultural land worldwide was moderately degraded and 6 per cent strongly degraded (Oldeman 1994). A more recent global assessment of land degradation (ISRIC 2008) identifies 24 per cent of land as degrading, mainly in Africa (south of the equator), SE Asia and southern China, North and Central Australia, the Pampas and parts of the boreal forest in Siberia and North America. Although cropland occupies only 12 per cent of land area, almost 20 per cent of the degrading land is cropland, with forests also over-represented (28% of area but 42% of degrading land). Some 16 per cent of the land area is improving, including cropland, rangeland and forests. Overall, the assessment shows the importance of natural catastrophic phenomena and human management in driving degradation, with the latter also instrumental in speeding up rehabilitation. Agriculture almost always results in stresses being applied to land (for example, by reducing organic matter returns to soils or the imposition of a physical stress such as tillage), but the properties of some soils allow them to recover naturally and rapidly, while others may require amendments (e.g. inputs of fertilizer) or other physical interventions to regain their productive ability (Greenland & Szabolcs 1994). By reducing degradation rates or increasing rates of land rehabilitation, competition for land in areas containing degraded land could be reduced (Debeljak et al. 2009).
(d) The role of policy Agricultural policy in many developed countries is dominated by protectionism, established through trade tariffs and producer support (subsidies). Subsidies affect land-use decisions by influencing the types of land-use strategies adopted by a land manager. So, for example, farmers will only grow crops for which they receive financial support through direct payment. In this sense, subsidies tend to limit competition for land. Subsidies also distort markets on a global scale and influence the competitiveness of agricultural land use in other regions of the world. Conversely, policy liberalization often leads to land-use diversification as seen, for example, in New Zealand following the 1984 agricultural policy reforms (MacLeod & Moller 2006), which overnight led to the complete removal of production subsidies (Smith & Montgomery 2004). In doing so, however, a liberalized land-use policy is likely to increase competition between land uses.
Pressure from the World Trade Organization, among other drivers, has in part led the governments of the developed world to move away from production-related support to new policy directions based on rural development or environmental protection. Policies such as the Less Favoured Area scheme in Europe, for example, were designed with the objective of protecting agricultural land use in areas with a competitive disadvantage, usually because of physical limitations such as topography or climate. By preserving the status quo of traditional agricultural landscapes, such policies limit or remove entirely the competition between alternative land uses. Other policies such as the European agri-environment schemes compensate farmers for managing their land to high environmental protection standards. The common theme in rural development and environmental protection policies, however, is the support of farmer incomes, and this leads to the maintenance of current land-use practices that limits land competition.
Competition for land is associated with deforestation owing to agricultural expansion while, at the same time, expansion of forests is leading to competition with other land uses. Furthermore, permanent forest clearing is associated with the loss of many other ecosystem services. Thus, deforestation is not only a phenomenon of competition for land per se, but is also important in considering the wider concept of competition for ecosystem services.

OBSERVED GLOBAL TRENDS IN LAND USE, 1990 -2010
Since 1960, agricultural area has increased from just under 4.5 billion ha to just over 4.9 billion ha in 2007 (FAOSTAT 2010). During the last 20 years, there has been an overall increase in agricultural area from 4.86 billion ha in 1990, but showing some fluctuations, with the greatest area of 4.98 billion ha recorded in 2001. Figure 2 shows the absolute and percentage change in agricultural and forest/woodland area for the world, and for each world region, 1990-2007.
As described in §2, the close to tripling of global food production since 1960 has largely been met through increased food production per unit area. For example, Bruinsma (2003) suggests that 78 per cent of the increase in crop production between 1961 and 1999 was attributable to yield increases, and 22 per cent to expansion of harvested area. Land use has therefore changed, despite smaller changes in land cover.
While yield increases have outpaced increases in harvested area in most regions, the proportions vary. For example, 80 per cent of total output growth was derived from yield increases in South Asia, compared with only 34 per cent in Sub-Saharan Africa. In industrial countries, where the amount of cultivated land has   Box 1. Models used for examining land-use change and competition for land in this review.
The IMAGE Integrated Assessment Model (MNP 2006) is a modelling framework often applied in the field of land-use/ land-cover change, because it is able to provide a geographically explicit description of land use. The land-use/land-cover description of IMAGE can be coupled to other models such as the global trade analysis project (GTAP) model or the IMPACT model. The rule-based allocation of IMAGE accounts for crop productivity and other suitability factors, such as proximity to existing agricultural land and water bodies. Bioenergy crops are grown on land other than that required for food production, forests, nature reserves and urban areas (MNP 2006). The macro-economic EPPA model was developed to examine climate and energy policy applications. Future scenarios are driven by economic growth using the GTAP data as base information, simulating the economy recursively at 5-year intervals. Five land types are considered to be interchangeable without restrictions, as long as conversion costs are met (submodel EPPA-PCCR, Pure Conversion Cost Response; EPPA-PCCRN, normalized PCCR; and EPPA-OLSR, Observed Land Supply Response). Reversion to the natural state occurs under no costs, with any prior investment being fully depreciated.
The MiniCAM model is also an environmental integrated assessment model. Land is allocated between different categories according to its expected profitability. This is determined by the productivity of the respective product, product price, the rental rate of the land and the non-land cost of production. Managed and unmanaged ecosystems are interchangeable according to the above.
Quickscan is a spreadsheet-based model constructed to analyse the bioenergy potential under consideration of key drivers and correlations. The study considered here identified the consumption of animal products as a key factor for agriculture land use and examined alternative production systems. The resulting spared land is considered to be available for bioenergy production. Therefore, the study considered only had the land categories 'bioenergy' and 'pasture'.
The GRAPE model is a macro-economic model of climate change. Land is allocated according to food demand of the population and the land-use rent that takes carbon costs (external costs caused by energy systems, land use and land-use change) into account.
For the MESSAGE-MACRO framework, integration of agriculture and forestry sectors has been achieved through linkages to the DIMA and AEZ-BLS models as described in Riahi et al. (2007). While land requirements for bioenergy supply and CO 2 mitigation via forest-sink enhancement are based on the sensitivity analysis of the DIMA model, the AEZ-BLS framework provides inputs with respect to agricultural land expansion.
GLOBIOM integrates the agricultural, bioenergy and forestry sectors. Changes in the demand on the one side and profitability of the different land-based activities on the other side are the major determinants of land-use change. Spatially explicit land-use suitability and respective productivities, as well as environmental effects, are taken into account (Havlík et al. in press).
The IMPACT model projects global food supply, food demand and food security to the year 2020 and beyond. Demand is determined by prices, income and population growth. From cropland and urban land projections, only cropland is considered in this analysis. been stable or declining, increased output was derived predominantly through the development and adoption of agricultural knowledge science and technology, which has served to increase yields and cropping intensity (van Vuuren et al. 2008). The role of land-use change and adoption of agricultural knowledge, science and technology have, therefore, varied greatly between regions. In some regions, particularly in Latin America, the abundance of land has slowed the introduction of new technologies (van Vuuren et al. 2008).

PROJECTED GLOBAL TRENDS IN LAND USE, 2010 -2050 (a) Changes in land use
The previous sections have shown that land-use changes are a result of the interaction of a variety of drivers and pressures. In particular, population growth and a shift towards more meat-intensive diets have in the past contributed to an increasing demand for agricultural land. These factors are expected to continue to be important in the future, although trends will differ in time and across regions. Historically, the demand for more agricultural production has been partly compensated by technological advances, and improving technology will determine whether yields will continue to improve in the future.
The complexity of the interactions between different drivers necessitates the use of scenario studies using models of land resources and land use, to analyse the consequences of particular trends and policies. There is a variety of studies and a range of models for addressing these issues. Box 1 gives an overview of the most commonly used models for such analysis at the global scale. For a review of land-use change scenarios at the regional scale, see Alcamo et al. (2006); Busch (2006) and de Chazal and Rounsevell (2009).
Future land-use trends are described as part of studies that look into long-term agriculture trends (such as the projections published by the Food and Agriculture Organization of the United Nations (FAO) and International Food Production Research Institute (IFPRI). In addition, studies focusing on agricultural trade increasingly tend to describe the relationships between trade flows and land use. Finally, integrated assessment models, used for examining global environmental change and climate change, are increasingly applied to investigate how climate policies might interact with land-use change. The type of models used in these different areas vary greatly, ranging from models derived from the economic tradition (general equilibrium models, e.g. GTAP-type models) to partial agricultural-economy models (like IMPACT), and models that focus mostly on the interaction of economic activity and biophysical indicators (e.g. the IMAGE and GLOBIOM model; box 1). General equilibrium models account for the economic linkages of the land-use sector with the rest of the economy and allow for assessment of income generation owing to land-use activities. Another strength of these models is their consistent description of agriculture trade. Partial equilibrium models allow for detailed study of agricultural production of different crops and within different regions. Moreover, some of these models are also able to represent specific land-use-related policies. Biophysically based models allow the relationship between environmental parameters (production potential based on soils and climate; land cover), land use and agriculture to be studied. Within the scope of this paper, we will not be able to review the complete literature of land-use scenarios; instead we will focus on a few noteworthy projections (table 3), while in table 4 we provide some details on the selected models, emphasizing how these models handle land use. For full details, the reader is advised to consult the references given. the IIAASTD study (van Vuuren et al. 2008) all focused on the relationship between environmental change and the agriculture sector. In these studies, a combination of the IMAGE model and IFPRI's IMPACT model was used to define the scenarios. The scenarios of the other studies look at more specific cases in regard to climate policy and biofuel potential. The general trends common to the scenarios considered here show an increase in land for bioenergy, crops and livestock, with forest and other lands decreasing. The exceptions here are scenarios implementing a carbon tax and a lower meat diet where more land is converted back to unmanaged forest. The scenarios considered by the IMAGE model, and those used in a wider range of studies, are given in tables 5 and 6, respectively. Table 7 shows the different land categories considered by each of the models we compare in this section.
(ii) Changes in consumption Global food production is projected to increase, driven by population growth and changes in diet ( §2a). The increase in production is somewhat slower than in the past, as a result of a slowdown in population growth. Diets are projected to become more meatintensive, with annual per capita meat consumption increasing. The growth in production of cereals over the 2000 -2050 period, based on a range of assessments, varies between 43 and 60 per cent (figure 3). The differences are relatively small since estimates of consumption growth are mostly driven by the increase in the global population (which shows relatively little variation between the different scenarios in 2050). An increasing share of cereals will be used as animal feed to supply the rapidly growing demand for livestock products. As incomes increase, demand for animal products also increases. This trend, which has been empirically established in all regions, is assumed to continue in the scenarios of the three groups of studies considered here. As a result, meat demand is projected to increase at a greater rate than the global population, and diets are projected to become more meat-intensive. For instance, the IFPRI calculations show annual per capita meat consumption increasing, on average, from 90 kg per person per year to over 100 kg between 2000 and 2050 in high-income countries, and from around 25 kg to nearly 45 kg per person per year in low-income countries during the same period. This trend is relevant for land use, since animal products require much more land than crops. On average, the production of beef protein requires several times more amount of land than does the production of vegetable proteins, such as cereals (Stehfest et al. 2009). While meat currently represents only 15 per cent of the total global human diet, approximately 80 per cent of the agricultural land is used for animal grazing or the production of feed and fodder for animals (FAO 2006). It should be noted that this includes extensive grasslands in areas where other forms of agriculture would be extremely challenging. Interestingly, future meat production varies considerably more than future cereal production among the scenarios (figure 3), since different scenarios show much more divergence in per capita meat consumption than for per capita cereal consumption. Some studies have looked into the consequences of reducing consumption of livestock products, with proteins being substituted by additional consumption of pulses (Stehfest et al. 2009), and shown that far less land would be required for agriculture under such extreme scenarios.

(iii) Cropland
The actual demand for cropland in the future depends on the balance between increases in agricultural demand and increases in yield improvement. Historically, yield improvements (approx. 80%) have been more important in increasing production than expansion of agricultural land (approx. 20%; see § §2 and 3 for more details). As a result, agricultural areas have expanded by about 5 per cent since 1970. Scenarios show a very large variation in the expected development in cropland (figures 4 -6). The 2050 projections for cropland increase range from as low as 6 per cent (e.g. the Technogarden scenario of the MEA), to an increase of more than 30 per cent (such as for the SRES A2 scenario, and one of the scenarios of the EPPA model; numbers represent the  Rosegrant et al. (2001Rosegrant et al. ( , 2008 60% interval of the literature). The average increase is around 10-20% (see also van Vuuren et al. 2008). In general, models with a stronger focus on physical parameters tend to project somewhat lower growth rates than models with a more macro-economic orientation (figure 6). The slightly lower contribution (on average) from the expansion of crop area can be attributed to increasing land scarcity and reduced growth of the global population. The decreasing quality of land brought into production, however, may mean that a greater percentage of gains in total production will need to be found from crop area expansion than has historically been the case (as indicated in MEA 2005). Even in the two scenarios with little global expansion of cropland, a considerable expansion of arable land still occurs in Africa, Latin America and partly in Asia, but this is compensated for by a decrease in arable area in temperate zones. Across the assessments, the area in crop production increases from 1.4 billion ha (or 10% of Earth's land surface) to up to 2.3 billion ha. As indicated by FAO, this expansion is within the scope of total land available for crop production (Bruinsma 2003). The fact that the assessments considered here agree on a rather flexible continuous response of the agricultural system to demand increases is interesting, as more sceptical views have also been expressed. An important implication, however, is further loss of the area available for unmanaged ecosystems (figures 4 and 7).
(iv) Animal husbandry and pastures Increases in meat production will occur through a number of means, including changes that lead to intensified production systems, such as more efficient conversion of feed into animal products, and via expansion of land use for livestock (figure 6). Previous scenarios indicate that most of the increases in world livestock production will occur in developing countries (Bouwman et al. 2005). For grazing systems, this means that some intensification is likely to occur. Considerable intensification is likely in mixed systems, with further integration of crop and livestock in many places. Strong growth is implied for confined livestock production systems. In the FAO scenario, for instance, at least 75 per cent of the total growth is in confined systems, although there are likely to be strong regional differences (e.g. less growth of these systems in Africa ;Bruinsma 2003). This is a continuation of historic trends. The major expansion in industrial systems has been in the production of pigs and poultry, as they have short reproductive cycles and are more efficient than ruminants in converting feed concentrates (cereals) into meat. Industrial enterprises now account for 74 per cent of the world's total poultry production, 50 per cent of pig meat and 68 per cent of eggs (FAO 2006(FAO , 2009a. At the same time, a trend to more confined systems for cattle has been observed, with a consequent rapid increase in demand for cereal-and soy-based animal feeds (these trends are included in the projections discussed in the previous section; see Delgado et al. 1999). For grazing land, the range of 2050 scenario projections ranges between a 5 per cent contraction to a 25 per cent increase (60% interval). Most studies show an increase of 10 per cent or less. The IAASTD baseline, for instance, projects an almost constant grazing area (van Vuuren et al. 2008). These numbers are lower than for croplands, representing the general view that croplands are expected to grow faster than the grazing area, driven by a further intensification of livestock production systems (and despite the rapid growth in meat consumption). The vast area of land used for animal husbandry also means that some studies looking into alternative pathways for land use often identify a large potential for reduction here, either by low-meat diets (Stehfest et al. 2009), or intensification (Smeets et al. 2007).
(v) Total land balance Obviously, the total demand for agricultural area arises from trends in cropland and grassland. Studies show diverging trends (figure 6), but there are also some common characteristics. First of all, almost all studies show an expansion in 2020 and 2050 of the area for cropland and grassland (as already noted in the previous sections). Second, in most studies, expansion of grassland area or cropland area represents the most dominant expansion category in 2020; by 2050, in some studies, however, bioenergy also becomes important (especially EPPA, MiniCAM and Quickscan). As indicated earlier, cropland expansion is generally more important than expansion of grassland, but there are some noteworthy exceptions (GEO4, and EPPA in 2020). In nearly all studies, both forest area and other areas (savannah, natural grasslands etc.) decline. The lowest numbers of land-use change are reported for the MEA (2005) scenarios, the IAASTD scenario, the IMAGE representation of the FAO baseline and the MiniCAM reference. Some of these scenarios include high levels of technology change (Global Orchestration, Technogarden and high AKST). High rates of land-use change are reported for several of the EPPA and MiniCAM scenarios. It should be noted that figure 6 represents a global picture. Much more change may happen at the regional level. A considerable expansion of arable land still occurs in Africa, Latin America and partly in Asia, but this is compensated for by a decrease in harvested area in temperate zones.
An important implication, however, is further loss of the area available for unmanaged ecosystems. This is already shown in figure 6; figure 7 shows the remaining as above with implementation of a mitigation policy to achieve atmospheric CO 2 of 550 ppmv by 2095 Quickscan: Smeets et al. (2007) system 1 mixed animal production, rainfed agriculture system 2 mixed animal production, rainfed and irrigated agriculture system 3 landless animal production, rainfed and irrigated agriculture system 4 landless animal production, very high crop production technology, rainfed and irrigated agriculture GRAPE: Kosugi et al. natural areas globally-but again it should be noted that these global figures hide underlying regional trends. In general, across the assessments, total natural areas decline by about 0-20%. This includes so-called baseline projections; but also scenarios that focus more on the projection of ecosystem services such as the MEA's Technogarden scenario; or the Sustainability First scenario of GEO4. There are only a few studies that have looked at incremental switches in management systems, such as those to semi-natural forest management (e.g. Havlík et al. in press) and changes in grassland management. A great impact on land-use change can also come from carbon incentives as demonstrated by Wise et al. (2009a). For example, the scenario examined by Wise et al. (2009a) in which (i) it is assumed that greenhouse gas emissions of the energy system are regulated and (ii) there is no regulation of emission from land-use change, according to MiniCAM work, will result in massive land-use change towards bioenergy and crops. In contrast, a policy that targets all potential greenhouse gas emissions (also from land use) can lead to preservation of woodland. A similar trend can be observed from the GRAPE model, which also takes carbon cost into account. In fact, these studies suggest that carbon taxing could have an impact on changing diet via the induced prices of meat. Ever-increasing competition for land may endanger the integrity of currently protected areas, which are located and classified in the World Database on Protected Areas (UNEP-WCMC 2009). Most model studies discussed above either assume projected areas to be constant, or even ignore this category as a special land category. There is one major exception, which is the Sustainability First scenario as part of UNEP's GEO4. Based on a minimum share of protected land by biome category, this study assumes that projected  (2007). This section serves to illustrate some of these uncertainties inherent in future projections of land use and of competition for land, and how these are critically dependent upon future policies on forest protection and bioenergy supply, and future trends in agricultural product preferences and consumption.
Given that there is substantial uncertainty about how different drivers will evolve and how they will impact upon the competition for land, here we illustrate the impact of uncertainty by presenting results of eight selected changes of drivers between 2020 and 2030. The analysis presented here was carried out using the GLOBIOM model (table 4 and box 1; Havlík et al. in press) over a short timeframe, to reduce the level of uncertainty introduced by the time period considered. Four uncertainty domains were identified for the quantitative modelling analysis on a global scale: biofuel, meat and wood demand and infrastructure development. In total, eight alternative scenarios were modelled under these four uncertainty domains, since the biofuel scenarios included five variants, differentiated by the expected     biofuel mix (table 8). To assess uncertainty in this analysis, the policy shock (table 8) was incorporated in the baseline for each scenario separately, and the model was re-run with the new assumptions. The scenarios were defined in such a way that any expansion of cropland would occur at the cost of forest land in order to have a 'pure' measure of the degree of competition for land. Under this scenario specification, the GLOBIOM model considers only drivers of deforestation coming from agriculture or bioenergy production. We consider that the model operates under the constraint of a fixed total land area, and allocates land use according to the economic competitiveness of different land-use activities. Deforestation is used as a measure of the degree of competition for land and is itself costly. The cost of avoiding deforestation is equal to the difference between the cost of deforestation itself and the income from agricultural production that would occur on that land subsequently if it were deforested and used for agriculture (opportunity cost). Under avoided deforestation, the degree of competition for land is mitigated at the cost of land-use intensification and reduced consumption (Havlík et al. in press). Figure 8 presents the global deforested area, which serves as a proxy for the degree of competition for land, between 2020 and 2030 in Mha. The red line displays the baseline scenario. The biofuel scenarios 1, 3 and 4, and the meat policy shock scenario, cause more deforestation. These scenarios are associated with agricultural land expansion owing to additional production of commodities. Improvement of infrastructure in emerging and developing economies on the one hand leads to higher pressure on natural ecosystems on the frontier, and on the other hand increases global productivity of agricultural production, and will therefore reduce land expansion in the long term. The infrastructure scenario leads to some 3 Mha more deforestation compared with the baseline. The result for the wood scenario is very close to baseline results, causing 0.35 Mha less for additional wood consumption since the relative value of forest increases. The only scenario that leads to less deforestation is the fifth biofuel scenario in which second-generation biofuels are used. This is associated with afforestation activities using high-yielding short-rotation forests. This policy shock scenario leads to a reduction in deforestation of more than 5 Mha over the period 2020-2030, when compared with the baseline.
The scenarios demonstrate the range of impacts a single biofuel production policy shock can exert on deforestation depending on the type of biofuel production technology used. Further sources of uncertainty lie in the resolution and quality of the land category considered. Many studies do not distinguish between managed and unmanaged forest and do not consider conversion to short-rotation coppice as deforestation. Therefore, in terms of net deforestation, natural forest can be converted in such models to short-rotation coppice without showing land-use change. In the scenarios presented here, deforestation is defined as conversion of unmanaged natural forest to cropland. The development of different forest types was tracked separately. For example, shortrotation plantations were only allowed to expand into cropland and grassland and therefore could only indirectly lead to deforestation through cropland expansion elsewhere into unmanaged forest. Increasing forest management intensity does not lead to deforestation. Lower deforestation in the secondgeneration biofuel and WOOD scenario is due to the 8.3% of biofuel in total transport energy in 2030 biofuels-portfolio BIOF1 ¼ 15% share of biofuels in total transport energy in 2030 in the form of a mix of all three types of biofuels (first-generation biodiesel and ethanol and second-generation bioethanol) biofuels-ethanol BIOF2 ¼ 15% share of biofuels in total transport energy in the form of first-generation ethanol only in 2030 biofuels-biodiesel BIOF3 ¼ 15% share of biofuels in total transport energy in the form of first-generation biodiesel only in 2030 biofuels-first generation BIOF4 ¼ 15% share of biofuels in total transport energy in 2030 from first generation (mix of biodiesel and bioethanol) only biofuels-second generation BIOF5 ¼ 15% share of biofuels in total transport energy in 2030 from second generation only. wood WOOD ¼ overall additional increase of 15% in demand for wood in 2030 meat MEAT ¼ overall additional increase of 10% for meat in 2020 and 15% in increased value of managed forest, reducing deforestation as described above. However, the increased value of forest management leads to higher conversion of unmanaged forest to managed forests using seminatural forest management practices. Another source of uncertainty arises from the models themselves. All models provide an imperfect representation of reality and rely on the availability and quality of input data and additional assumptions. For example, in GLO-BIOM, there is no explicit link assumed between increased animal production and grassland demand. Consequently, the MEAT scenario will overestimate the degree of deforestation owing to the restrictive grassland assumptions. It is important to be aware of these inherent uncertainties when dealing with future projections. Improved models, data and more sophisticated scenarios will allow this uncertainty to be reduced in the future, but projections of future policy impact will always contain a degree of uncertainty.

CONCLUDING REMARKS
We have shown that competition for land, in itself, is not a driver affecting food and farming in the future, but is an emergent property of other drivers and pressures. There is considerable uncertainty over projections of intensity of competition for land in the future, and the regional distribution of this competition. Modelling studies show that future policy decisions in the agriculture, forestry, energy and conservation sectors could have profound effects, with different demands for land to supply multiple ecosystem services usually intensifying competition for land in the future. Given the need to feed 9 billion people by the middle of this century, and increasing competition for land to deliver non-food ecosystem services, it is clear that per-area agricultural productivity needs to be maintained where it is already close to optimal, or increased in the large proportion of the world where it is suboptimal. It remains a challenge to deliver these increased levels of production in a way that does not damage the environment and compromise other ecosystem services (Royal Society 2009).
In summary, in addition to policies addressing agriculture and food production, further policies addressing the primary drivers of competition for land (population growth, dietary preference, protected areas, forest policy) could have significant impacts in reducing competition for land. Technologies for increasing per-area productivity of agricultural land will also be necessary. Key uncertainties in our projections of competition for land in the future relate predominantly to uncertainties in the drivers and pressures within the scenarios, uncertainties in the models and data used in the projections and the policy interventions assumed to affect the drivers and pressures in the future. Though price has been used as an indicator of land scarcity and, therefore, competition for land, the development of other indicators to assess the intensity for competition for land is in its infancy, and the development of new metrics will advance our understanding of competition for land in the future.  Review. Competition for land P. Smith et al. 2955 Phil. Trans. R. Soc. B (2010)