Global economic–biophysical assessment of midterm scenarios for agricultural markets—biofuel policies, dietary patterns, cropland expansion, and productivity growth

Land-use decisions are made at the local level. They are influenced both by local factors and by global drivers and trends. These will most likely change over time e.g. due to political shocks, market developments or climate change. Hence, their influence should be taken into account when analysing and projecting local land-use decisions. We provide a set of mid-term scenarios of global drivers (until 2030) for use in regional and local studies on agriculture and land-use. In a participatory process, four important drivers are identified by experts from globally distributed regional studies: biofuel policies, increase in preferences for meat and dairy products in Asia, cropland expansion into uncultivated areas, and changes in agricultural productivity growth. Their impact on possible future developments of global and regional agricultural markets are analysed with a modelling framework consisting of a global computable general equilibrium model and a crop growth model. The business as usual (BAU) scenario causes production and prices of crops to rise over time. It also leads to a conversion of pasture land to cropland. Under different scenarios, global price changes range between −42 and +4% in 2030 compared to the BAU. An abolishment of biofuel targets does not significantly improve food security while an increased agricultural productivity and cropland expansion have a stronger impact on changes in food production and prices.


Introduction
Land is a limited resource that can be used in different ways for the production of food, animal feed, fibre, and bioenergy, or for the protection and maintenance of ecosystems and its services (Foley et al 2011). Land-use decisions are made at the local level. Together with local environmental constraints, they also depend on a local cultural and legal framework that sets socio-economic boundaries determined by e.g. available capital resources and management practices (Tilman et al 2011, Václavík et al 2013, Mauser et al 2015.
A significant share of agricultural commodities is traded internationally (OECD/FAO 2016). The larger this share, the more local production conditions are codetermined by global socio-economic and environmental conditions and political interventions mainly through price signals from world markets (Porkka et al 2017, Golub et al 2012.
Global socio-economic changes are driven mainly by an increase of world population and expected higher incomes (Dellink et al 2017). As a result, global demand for agricultural commodities is estimated to increase, not only for food, but also for bioenergy and livestockbased products (Calzadilla et al 2016, Foley et al 2011, Bruinsma 2011. Thus, pressure on land for intensification and expansion will increase with far-reaching environmental consequences (Foley et al 2011, Gerstner et al 2014, Newbold et al 2015.
Political decisions e.g. on agricultural policies, the demographic and technological development or environmental changes trigger feedbacks between the world market for agricultural commodities and local landuse decisions which again feed back on the world market and the other factors mentioned. These multifacetted feedbacks have been studied by developing and analysing scenarios (Popp et al 2017, Dellink et al 2017, Valin et al 2014. Scenarios help to better understand complex cause-impact relations and identify alternative pathways into the future and their respective trade-offs.
Most scenarios cover the time frame from today towards 2050 or 2100. While this is the established timeframe for climate change studies it is largely impossible to agree on reasonable scenario assumptions for the socio-economic development until the middle, not to speak of the end of the 21st century. The mid-term integrated scenarios, which reach out until 2030, allow to establish a reasonable set of assumptions, which can form a solid basis to carefully analyse feedbacks between economic and biophysical factors in the development of land-use.
Appropriate modelling tools to study the feedbacks between economic and biophysical factors and land-use with integrated scenarios of global change are emerging. The studies combe partial/general computable equilibrium models (e.g. Baldos and Hertel 2014) or optimisation models (e.g. Popp et al 2017) with crop growth models. These studies feed changes in agricultural productivity into economic models. Usually no feedback exists from economic models to the simulation of crop yields and land-use changes. An approach that studies the feedback effects by iteratively coupling the socio-economic results of a computable general equilibrium (CGE) simulation of the global agricultural market and the biophysical simulations of global potential crop yields is followed by Mauser et al (2015). It shows that considering the feedback of economic and biophysical factors changes the spatial land-use patterns and results in larger potential yields compared to other studies.
Several scenario-based studies have looked into potential future developments of the agricultural sector as a whole (Odegard andvan der Veat 2014, Schmitz et al 2014). A model exercise showed that most models agree on an increase of global cropland area between 10%-25% in 2050 compared to 2005 (Schmitz et al 2014). Nevertheless, after harmonising key assumptions significant differences in projections remain. The differences in key assumptions partly represent the uncertainty about major drivers of agricultural markets (von Lampe et al 2014). The diverging results of these studies suggest that in order to develop mid-term scenarios an expert based process should to be used which develops clearly defined storylines and scenario assumptions.
The objective of this paper is to develop and analyse a set of global integrated mid-term scenarios of the likely development of socio-economic and environmental conditions until the year 2030. They are used to explore the range at which the world agricultural market and regional land-use decisions influence each other.

Co-design of storylines and scenarios 4
The global scenarios were developed using a threestep participatory co-design approach according to Mauser et al (2013) in the course of several workshops (figure 1). First, we proposed a preliminary set of storylines (supplementary figure A.1.1, available at stacks.iop.org/ERL/13/025003/mmedia) to experts from globally distributed regional studies of the research program 'Sustainable Land Management' by the German Ministry for Education and Research.
The workshop participants discussed the preliminary storylines and suggested additional storylines. The storylines were then consolidated and further discussions concluded with a final set of six storylines (left part in figure 2). The storylines can be analysed with and without future climate change 5 .
Data for quantifying the storylines with regard to sectoral developments and their aggregation in terms of the temporal and spatial resolution were then identified (column 2 in figure 1). The main outputs are changes in regional and global commodity prices and production, agricultural yield potentials, and changes in land-use.
Finally, each storyline was transformed into a scenario by formulating assumptions for model parameters representing the drivers and by assigning appropriate values (figure 1, right column, figure 2 and section 3). Under the two extreme scenarios, certain parameters sets of the other scenarios are combined. This was presented for discussion in a third workshop. The scenarios were then implemented in a modelling framework (see figure 1 and section 2.2).

Modelling framework
We apply a modelling framework that combines the global CGE model DART-BIO and the dynamic crop growth model PROMET. Results are simulated by either using the DART-BIO model alone or by linking it with PROMET.

DART-BIO
The Dynamic Applied Regional Trade (DART) model is a global multi-sectoral, multi-regional recursivedynamic CGE model (e.g. Klepper and Peterson 2006).  Figure 1. From the co-design process to scenario analysis. Green colour denotes stakeholder-involvement, blue colour scientist's contributions.

Business as might be usual
Baseline assumptions of DART-BIO: biofuel quotas, fixed land endowment, demand preferences according to GTAP-database, productivity growth rates driven by total factor productivity Baseline assumptions of DART-BIO but no implementation of biofuel targets Baseline assumptions of DART-BIO but change in agricultural land endowment according to FAO (2012) Baseline assumptions of DART-BIO but higher elasticities for dairy and meat products in selected Asian countries Baseline assumptions of DART-BIO but decreased productivity growth rates in the agricultural sector and changed elasticities for dairy and meat products in selected Asian countries Baseline assumptions of DART-BIO but biofuel policies, agricultural land endowment according to FAO (2012) DART-BIO has 23 regions (figure S1) subdivided into 18 so-called agro-ecological zones (AEZs) (figure S2), 38 sectors, 45 products (table S1) and 21 production factors (table S2). DART-BIO takes into account 6 Data on AEZs were not yet available for the new GTAP9 database when we did the analysis. economy wide repercussions of policies or changes in assumptions 7 .

PROMET
PROMET is an integrated hydrological land surface process model, which includes a detailed dynamic simulation of crop growth for 18 different staple and bioenergy crops (  development, leaf temperature, water availability and environmental factors, including meteorology and CO 2 concentration for C3 and C4 pathways. For global application, 250 000 statistically representative samples are chosen randomly on the global agriculturally suitable area according to Zabel et al (2014). Management measures, such as irrigation and fertilisation are considered. This allows the simulation of potential yields presuming optimal crop management (no nutrient deficits, pests, diseases and yield losses), denoted as 'potential agro-ecological yields'.

The coupling approach
While in former assessments yield potentials are calculated on sample locations by taking the crop that is currently grown (Pugh et al 2016, Mueller et al 2012, IIASA and FAO 2012, our approach allows farmers to change the allocation of crops by choosing the crops with the highest profitability (Mauser et al 2015).
The profitability (profit per hectare) is influenced by biophysical drivers that determine the agro-ecological yield, and by socio-economic drivers (figure 3) that influence the relative profitability of crops. In a coupling procedure (figure 3), the agroecological potential yields of 18 crops as an output of PROMET are spatially aggregated to the 23 regions applied by DART-BIO and grouped into 10 crop categories (table S3) to match with the crop aggregates used in DART-BIO. Agro-ecological potential yields, harvested areas, and crops' marginal profitability to land determine an agro-economic potential yield in an iterative approach. The approach is performed within five steps (supplement A.2.2). It takes into account that land allocation to crops changes over time due to changing cropping decision of farmers driven by e.g. dietary preferences or climate change.

Business as usual scenario (BAU)
The BAU scenario carries forward the present situation until 2030. Assumptions for important drivers such as population growth, trade policies, and information about the calibration of the BAU scenario are explained in the supplement A.3.1. Biofuel policies are taken from OECD/FAO (2016) and Beurskens et al (2011) (figure 4). We assume that these shares will not change between 2020 and 2030.

No biofuel policies scenario (NBQ)
Some countries have already decided or are discussing to reduce or abolish biofuel targets (e.g. the EU (EC 2016)). Therefore, we implement a scenario in which there are no biofuel policies while the other parameters remain at the BAU levels.

Meat scenario (MS)
It is expected that demand for meat and dairy products (MDP) will rise particularly in Asian regions (Chemnitz 2016, Alexandratos andBruinsma 2012). In the MS scenario, we increase the preference of consumers for MDP by replacing the estimated income elasticities for meat consumption in Asian regions with the highest elasticities found in the respective demand functions (table 1). Thus, with rising incomes, a higher share of a consumers' budget is spent for MDPs 9 .  Figure 4. Share of biofuels on transport fuel consumption in 2020.

Land expansion/contraction scenario (LE)
The

Higher prices (HP)
This scenario combines aspects of the BAU with multiple supply and demand drivers leading to higher prices for agricultural products. Under this scenario biofuel policies and the other policies of the BAU scenario remain in place, and expansion of agriculture into uncultivated land is not possible. In addition, the MS assumptions are used. At the same time, lower productivity growth in the agricultural sector is implemented by reducing yield increases 0.2 percentage points below the BAU rates.

Lower prices (LP)
This scenario represents a combination of changes, now with a tendency towards larger agricultural supply and less demand to lowering future agricultural product prices. We combine the abolition of biofuel targets with keeping MDP consumption preferences as in the BAU scenario. We use assumption of the LE scenario and introduce a partial closing of regional yield gaps. Specifically, the regional yield gap ratio (figure 6), which is the potential agro-economic yield divided by the statistical yield 10 is reduced by 5%. This reduction of yield gaps causes productivities to increase more strongly in regions with high yield gaps.

Climate change (CC)
To address climate change impacts on agriculture, we take the assumption of the BAU and add changes in yields under climate change. This is modelled by using results from the linking of PROMET and DART-BIO, where PROMET is driven by ECHAM5 climate data for A1B conditions 11 . Percentage changes in agroeconomic yield potentials are gathered by comparing the agro-economic yields for the period 1981-2010 with those of 2011-2040.

Trends under the BAU scenario
The global production of agricultural primary products increases by between 30% and 83% between 2007 and 2030 ( figure 7). The strong increase of soybeans and wheat is mainly caused by their use as feed in the livestock industry as increasing incomes lead to higher 10 The statistical yield is derived from the underlying GTAP database (Baldos and Hertel 2014   meat consumption. In addition, soybeans are used for biodiesel production for which demand increases threefold by 2030. Global average prices increase more strongly for primary agricultural products than for processed agricultural goods since land price increase relative to the other factors of production, labour and capital. Prices of livestock bred on pasture land rise by 111% while production only increases by 23%. This result is driven by the increased competition for land leading to a reduction in pasture land, a rise in land prices and consequently in the prices of land intensive meat products. The harvested crop area expands by +5% until 2030 at the expense of pasture land (−9%). Within cropland, oil seeds and wheat show the largest increase in harvested area. Additional results are discussed in supplement A.4.1.

Comparison of scenarios with BAU
Results are presented in percentage differences in 2030 between BAU and a scenario to illustrate the relative impact of the scenarios on production, prices and landuse 13 . At the global level, most scenarios do not show a large deviation from BAU. As illustrated in figure 8 for the case of wheat, the dominating overall trend is an increase in global crop production. The exception is the LP scenario where strong productivity growth strongly increases the global output of crops. Figure 9 summarises the impact of the six scenarios on agricultural products on a global scale.
The production and price changes in the different scenarios vary across crops and across regions. The results are discussed in detail in the following sections.

Impact of the no biofuel policy scenario (NBQ)
Less demand for oil seeds for biodiesel production and less demand for sugar cane, maize, and wheat for bioethanol production lowers their global production, trade flows and prices. The global average prices of annual crops drop by about 1%, only sugar cane shows a strong decline in prices by 8.7% (figure S3 and supplement A.4.2).

Impacts of the meat scenario (MS)
On the global level, additional demand for MDP in Asian countries causes only small changes in production, harvested area and prices 14 . To satisfy the increased MDP demand, the production of processed meat rises by 0.4%, and consequently the production 13 Additional information for each scenario on the changes in trade flows, regional prices and production for each scenario in the supplement A.4. 14 Results are discussed in supplement A.4.3. of the major feedstuff soy meal (and beans) increases by 0.3%. The area used for soy beans production increases at the costs of the area used for rapeseed ( figure S4).
Most Asian countries are net-importers of crops. The higher demand for MDP causes imports to these countries and production of these products and their intermediates to rise.

Impact of the land expansion/contraction scenario (LE)
With changes in harvested area under the LE scenario (figure 5), global average crop production increases while prices decline. The impact on the global production of the different crops varies since they are produced in different regions with different growth rates in land endowment. Consequently, the global palm fruit, soy beans, and sugar cane production increases by up to 8.4% (figure 9).
Global average prices of crops fall by 6% to 20% (see figure S5). Regional prices of crops decrease most in regions with high land expansion (e.g. Latin American regions), while prices increase in regions where cropland area is reduced and trade restricted (Japan) (figure S6). Not only regions with high growth rates in land expansion experience decreasing food prices, but also net importers of crops benefit from lower world market prices for imported products 15 .

Impacts of the higher prices scenario (HP)
The combined effects of smaller productivity increases and rising MDP demand cause the strongest increase in global average prices among all scenarios (2.1% to 4.4% across crops), while the global production of crops decreases by 0.9% to 1.8% (figure 10). The reduced productivity gains obviously constrain the effect of higher MDP demand. Even the production of the main feedstocks does not increase relative to the BAU scenario, even though there is a shift of land towards these crops. One might expect an increase in crop production to satisfy higher demand for animal feed, but the lower productivity on the supply side has a stronger impact on production than the increase in MDP demand. Crop prices increase most in the EU, Africa and Middle East ( figure S7). Surprisingly, price increases are lower in Asian regions, where more MDP is demanded under this scenario. The decrease in productivity growth is smaller in these regions compared to other regions (figure S8), such that regional prices increase less relative to the global average 16 .

Impacts of the lower prices scenario (LP)
The abolishment of biofuel quotas causes less demand for crops, while cropland expansion and higher yields cause a higher production of crops. The results are strong market responses with lower global average prices of agricultural products.
Globally, the largest increases in production occur for palm fruit and 'other oil seeds' (figure 11). These are predominantly produced in countries with a large potential for increasing yields and a large predicted land expansion (Malaysia/Indonesia, Sub-Saharan Africa, Latin America). The production of rapeseed and wheat increases much less due to two competing factors: (1) the reduced demand for biofuels causes rapeseed and wheat production to decrease; (2) yield gaps of wheat are below the average (figure S9) such that wheat becomes less competitive relative to other crops.
Regions are affected very differently. Summed up over all crops, crop production increases by up to 42% in the South American regions, but also Malaysia/Indonesia and South-East Asia show strong increases in production (figure S10). This is caused by a combined effect of cropland expansion and increases in land productivity. Production increases less in e.g. India, since a high increase in productivity coincides with limited land expansion potentials 17 .

Impact of the climate change scenario (CC)
Climate change has globally a small positive impact on potential yields until 2030. The large productive agricultural areas in the higher latitudes experience a moderate temperature increase until 2030 resulting in longer growing seasons. In sum, global production increases for all crops between 1.7% and 6% until 2030, while crop prices decline (figure S11). The regional effects are quite diverse. Crop production increases by 3% in the EU, by 8% in the USA, and 10% in Russia (figure S12). Prices of crops drop in all regions except Brazil, South-East Asia and Malaysia/Indonesia. Tropical regions mainly show a decline in potential yields and crop production, since temperature increase results in higher heat stress for plants 18 . 18 Details in supplement A.4.7.

Discussion
The results clearly show that that accounting for world market responses to regional economic developments such as demand, supply, and prices on world markets is highly important. Global trends as shown in the BAU scenario lead to an increasing scarcity of fertile land with rising prices for primary agricultural products. The trends are mainly driven by higher incomes combined with higher demand for meat products and to a lesser degree by the rising demand for biofuels. These trends cause production and prices of crops to rise and lead to a conversion of 228 Mha of pasture land to cropland. Compared to other studies (Schmitz et al 2014) that predict an expansion between 10% to 25% by 2050, we are at the lower range of cropland expansion.
The comparison of the scenario simulations with the BAU highlights the impact of specific policy as well as supply-and demand-side changes. It turns out that supply-side effects have a stronger impact than changes on the demand side. Figure 12 illustrates in an aggregated way how global production remains relatively stable. Exceptions are the scenarios with land expansion (LE scenario) and higher productivity (LP scenario).
Demand side effects in the MS and the NBQ scenario are small compared to the impacts on global quantities and prices in the other scenarios which indicates that the changes in dietary preferences do not have a large impact on global agricultural markets and land-use. However, the product-and region-specific adjustments show that substitution is taking place to accommodate the increased demand for meat. Adjustments show up in large changes in international trade and in regional adjustments 19 .
Biofuel support is often made responsible for the increasing scarcity of food and thus a threat to food security (e.g. OECD 2008). This cannot be supported by the scenario analysis, since global food consumption rises by only 0.3% compared to the BAU when abolishing biofuel policies. Hence, a reduction in biofuel support does not significantly improve food security 20 . The international adjustments through trade and the integration of the joint production of oils or grains and feed are responsible for our result.
Under the HP scenario with all adjustments, global consumption of food is only 1% lower than in the BAU. But prices do not increase as much as expected which is mostly due to substitution effects towards other nonfood consumption goods.
Much stronger effects occur in the LP scenario. Taking prices of agricultural goods and production quantities as indicators for food security, our results show that the LP scenario leads to an improvement of food security. This holds particularly in regions that face 19 Detailed results on the substitution effects in supplement A.4.2 and A.4.3. 20 Detailed information in supplement A.4.1. problems in improving food security (Latin America, South-East Asia, Sub-Saharan Africa).
Land expansion and climate change have a stronger impact on prices than on production indicating that demand is more inelastic than the supply side.

Conclusions
The set of scenarios, developed in a co-design process, provides consistent information on the impact of different global developments. It also highlights by how much different supply and demand developments as well as policies influence the allocation of resources in future markets. Since local land-use decisions strongly depend on global market conditions using the same scenarios in regional studies increases the comparability and transferability of such studies.
The results indicate that world market effects are themselves the result of very diverse regional developments and adjustment processes to the drivers of the different scenarios. This diversity of responses underscores the advantage of using globally consistent scenarios when analysing the development in a region in detail as the regional development may have an impact on world markets but world markets also transmit the reactions of markets in other regions. These effects cannot be accounted for by regional studies alone.
Global markets attenuate the potentially large regional impacts of changes in policies or supply and demand conditions substantially. Therefore, issues with regional food security become less pressing, too. Contrary to concerns about the impact of biofuel policies or behavioural changes towards higher meat consumption, food security seems to be less of a concern because quantity and price effects remain small. The biggest opportunity for improving food security is productivity increases. The strong price decreases are particularly important since food security is to a large extent connected to the ability to afford food.