Land Use Policy 86 (2019) 299–312

Contents lists available at ScienceDirect

Land Use Policy

journal homepage: www.elsevier.com/locate/landusepol

A linkage between the biophysical and the economic: Assessing the global T
market impacts of soil erosion
Martina Sartoria, , George Philippidisa,b, Emanuele Ferraria, Pasquale Borrellic,
⁎

Emanuele Lugatod, Luca Montanarellad, Panos Panagosd

a
European Commission, Joint Research Centre (JRC), Seville, Spain
b
Aragonese Agency for Research and Development (ARAID), Centre for Agro-Food Research and Technology (CITA) – Agrifood Institute of Aragón (IA2), Government of
Aragón, Zaragoza, Spain
c
University of Basel, Environmental Geosciences, Basel, Switzerland
d
European Commission, Joint Research Centre (JRC), Ispra, Italy

ARTICLE INFO ABSTRACT

JEL classification: Employing a linkage between a biophysical and an economic model, this study estimates the economic impact of
C68 soil erosion by water on the world economy. The global biophysical model estimates soil erosion rates, which are
Q24 converted into land productivity losses and subsequently inserted into a global market simulation model. The
Q10 headline result is that soil erosion by water is estimated to incur a global annual cost of eight billion US dollars to
global GDP. The concomitant impact on food security is to reduce global agri-food production by 33.7 million
Keywords:
tonnes with accompanying rises in agri-food world prices of 0.4%–3.5%, depending on the food product cate-
Soil erosion
gory. Under pressure to use more marginal land, abstracted water volumes are driven upwards by an estimated
Land productivity loss
Computable general equilibrium 48 billion cubic meters. Finally, there is tentative evidence that soil erosion is accelerating the competitive shifts
Model integration in comparative advantage on world agri-food markets.
Global economy
Agriculture

1. Introduction risk (rainfall erosivity, increased number of dry days combined with
strong thunderstorms) and human activities (e.g. land use change, de-
In a changing world of eight billion people facing the critical threats forestation, overgrazing, agricultural intensification) (Panagos et al.,
of climate change, water scarcity and depletion of soil fertility, the 2016). Soil erosion is a major threat to agricultural soil productivity
agricultural economy is faced with the challenge of maintaining food (losses in yields, nutrients and plantations) and may also generate off-
security whilst respecting environmental and ecological boundaries site impacts such as sedimentation, flooding, damage to properties,
(Altieri and Nicholls, 2017). A key element for ensuring a sustainable landslides, and water eutrophication (Boardman and Poesen, 2006).
system of food production is linked to effective soil management, which The best techniques to prevent or reduce soil erosion rates are reduced
requires a reduction in soil erosion rates (Poesen, 2018). Among various tillage, contour farming, terraces, afforestation of slopes, plant residues,
land degradation processes, soil erosion is recognized as a major en- cover crops, grass margins and brush layers (Poesen, 2018; Panagos
vironmental problem causing a loss of topsoil and nutrients, reduced et al., 2016).
soil fertility (Zhao et al., 2013) and, as a consequence, reduced crop A recent estimation of land degradation costs shows that the global
yields (Telles et al., 2011). Furthermore, soil erosion may unlock and economic impact is highly uncertain, from 40 to 490 billion US$, and
thereby increase emissions of CO2, exacerbating global warming varies from country to country (Nkonya et al., 2016). More than two
(Lugato et al., 2018). decades ago, Pimentel et al. (1995) estimated the on-site costs of water
The main causes for soil erosion by water are geomorphological erosion in the United States of America to be about 16 billion US$ per
factors (heterogeneous surfaces, steep slopes) combined with climatic year based on expert knowledge. Similarly, the agricultural

Corresponding author.
⁎

E-mail addresses: [email protected] (M. Sartori), [email protected] (G. Philippidis), [email protected] (E. Ferrari),
[email protected] (P. Borrelli), [email protected] (E. Lugato), [email protected] (L. Montanarella),
[email protected] (P. Panagos).

https://doi.org/10.1016/j.landusepol.2019.05.014
Received 22 December 2018; Received in revised form 19 March 2019; Accepted 8 May 2019
Available online 17 May 2019
0264-8377/ © 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/BY/4.0/).
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

productivity loss due to soil erosion in the European Union is estimated proposed by Wischmeier and Smith (1978) which relies on some in-
to be around 300 million € (Panagos et al., 2018) using a combination trinsic soil properties (e.g. texture, organic matter, structure and per-
of the recent soil loss assessment and the well-known Global Trade meability) currently available at the ISRIC SoilGrids database at 1 km
Analysis Project (GTAP) computable general equilibrium (CGE) simu- spatial resolution (Hengl et al., 2014). The topographic parameters,
lation model. A recent application to the African continent estimates slope and upslope contributing area, needed to compute the LSg , factor
the annual loss of crop yield to be about 280 million tonnes (Wolka are derived from the hole-filled SRTM 3 arc-seconds (ca. 90 m) Digital
et al., 2018), compared with a corresponding figure of only six million Elevation Model (Reuter et al., 2007) for the land surface between
tonnes in the European Union (Panagos et al., 2018). 60 °North and 56 °South and ASTER GDEM v2 data products for the
With one notable exception (Panagos et al., 2018), a typical feature extreme North latitudes (Robinson et al., 2014). The global land cover
of the aforementioned studies is that they carry out a 'first-order' cost and management factor Cg is computed for the year 2001 and 2012,
evaluation exercise focusing on agricultural production losses taking into consideration the individual land cover type, vegetation
(Martínez-Casasnovas and Ramos, 2006; Erkossa et al., 2015; Hein, cover dynamics and farming systems of each cell. Two different ap-
2007). More specifically, the economic value of land productivity loss is proaches are undertaken to estimate the Cg factor values for agricultural
calculated by the direct loss in production of the affected crops (tonnes) and non-agricultural land. For agricultural land, data of 170 different
multiplied by their respective average market prices ($/tonnes). Thus, crops (averaged over a period of twelve years) obtained from the
the vast majority of these studies do not capture the resulting 'second- FAOSTAT database (http://www.fao.org/faostat/en/#data) of the
round' effects of structural economic change that arise owing to shifts in Food and Agriculture Organization's (FAO) are used (more detail in
primary resources, particularly the land factor. Moreover, to the best of Borrelli et al., 2017). To assess the final modelling factor, i.e., Pg , the
our knowledge, there is no study that fully captures these structural information about the proportion of cropland area under conservation
impacts from land productivity losses due to soil erosion at the global agriculture provided by the countries to FAO are used. To evaluate
scale. whether the model outcomes comply with the regional findings of
To close this gap in the literature, an approach akin to Panagos et al. former studies, the global soil erosion maps of 2001 and 2012 are
(2018) is followed. Thus, a sequential modelling framework is em- compared with a set of representative and highly advanced regional soil
ployed, where national and regional soil erosion rates provided by the erosion assessments. More detailed information on Eq. (1) is provided
recent global soil erosion assessment (Borrelli et al., 2017) are first in Appendix A1.
converted into land productivity losses, and then implemented into the The study focuses on 14 million km2, which is considered to be the
Modular Applied GeNeral Equilibrium Tool (MAGNET) (Woltjer and global arable land area where crops are cultivated. This area corre-
Kuiper, 2014). At its core, MAGNET is essentially the GTAP model sponds to approximately 11% of the total modelled area of 125 million
(Corong et al., 2017), although it is preferred largely because it contains Km2, which coincides with the statistics provided by the World Bank1
a superior modelling treatment of agricultural factor markets. The and FAO2 .
counterfactual thus captures the resulting marginal market impacts in It should be recognised that the crop productivity loss due to erosion
agricultural (and non-agricultural) activities, which arise in each region includes high uncertainty and depends on many factors such as erosion
due to soil erosion. rate, crop type, crop yields, seasonality, etc. To estimate the associated
The rest of this paper is structured as follows. Section two explains land productivity losses by region (LPLr) arising from soil erosion, this
how soil erosion rates and land productivity losses are obtained. Section study follows the same approach as Panagos et al. (2018):
three shows how the economic impact of soil erosion is measured, (2)LPLr = SEAr / TAAr *0.08Where SEAr is the area of severe erosion per
whilst the results are presented in Section four. A final section discusses region/country 'r' in hectares and TAAr is the agricultural area in each
how these findings can benefit the formulation of relevant land use region/country 'r'. This study assumes a mean crop productivity loss of
policy, presents some of the caveats and adds some concluding remarks. 8% in arable lands threated by severe erosion (> 11 t ha−1 yr−1). This
assumption is based on a thorough literature review (see Panagos et al.,
2. Estimating global soil erosion rates and land productivity losses 2018 and Table S1 of the Supplementary material) taking into account
experimental results on crop losses in cases of severe erosion in dif-
Long-term annual soil erosion rates are obtained from Borrelli et al. ferent areas all over the world (Panagos et al., 2018).
(2017), who use a combination of remote sensing, spatial analysis
techniques and statistical data in the framework of the Revised Uni-
3. Measuring the economic impact of soil erosion
versal Soil Loss Equation (RUSLE) model. The model provides erosion
rates at a ˜250 × 250 m cell bases for the land surface of 202 countries
3.1. Model framework and data
(around 2.89 billion cells; ˜125 million km2), covering about 84.1% of
the Earth’s land area. The soil erosion (Mg ha−1 yr−1) resulting from
Neoclassical multi-region CGE models enumerate the theoretical
interrill and rill erosion (Fig. 1) processes is based on the following
economic tenets of constrained optimisation, to govern the behaviour of
multiplicative equation:
agents (i.e., households, producers, government, investors – see Fig. 1)
A g = R g*K g *LSg*Cg *Pg (1) across the global economy. The behavioural equations are supported by
market clearing equations and accounting identity conventions to en-
where: Ag[Mg ha−1 yr−1] is the annual average soil loss, Rg [MJ mm
sure a stable equilibrium within the closed system of the model (Fig. 1).
h−1 ha−1 yr−1] is the rainfall-runoff erosivity factor, Kg [Mg h
To underpin the model framework, a ‘benchmark’ equilibrium year
MJ−1 mm−1] is the soil erodibility factor, LSg [dimensionless] is the
of data representing a balanced system of national economic accounts,
joined slope length and slope steepness factor, Cg [dimensionless] is the
gross bilateral trade flows and protection and international transport
land cover and management factor, Pg [dimensionless] is the soil con-
margins is required. To ensure the model replicates the equilibrium
servation or prevention practices factor.
conditions of the benchmark year, the mathematical parameters of the
According to Eq. (1), RUSLE consists of a multiplicative equation
behavioural equations are ‘calibrated’ to the database. Ensuring that the
including five environmental parameters (Fig. 1). The global rainfall
number of endogenous variables and model equations are equal (clo-
erosivity factor (Rg) is computed according to Renard et al. (1997),
sure), powerful computer algorithms are employed to reach an
using a combination of sub-hourly and hourly pluviometry data of 3625
meteorological stations (collected across 63 nations) interpolated using
the Gaussian Process Regression (GPR) (Panagos et al., 2017). The 1
https://data.worldbank.org/indicator/ag.lnd.arbl.zs
global soil erodibility factor Kg is measured based on the equation 2
http://www.fao.org/docrep/005/y4252e/y4252e06.htm

300
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

Fig. 1. Sequential modelling approach adopted in this study and model description.

‘equilibrium’ solution. More specifically, in response to a policy or Smeets et al., 2014; Philippidis et al., 2018a); food security (Rutten
structural shock, the economic system moves to a new ‘counterfactual’ et al., 2013), climate change (van Meijl et al., 2018) and international
solution characterised by a unique set of prices such that demand trade (Philippidis et al., 2018b 2018b). In common with GTAP,
matches supply in ‘n’ markets; income, output and expenditure flows MAGNET is calibrated to the GTAP database (Aguiar et al., 2016),
are equal, and the balance of payments between the current and capital which in this study employs version 9 with a benchmark year of 2011.
accounts nets to zero. Comparing the counterfactual with the bench- The GTAP data encompasses 141 regions and countries, 57 tradable
mark gives an indication of the marginal impact of the shock on market sectors and eight factors of production (including agricultural land).
indicators (i.e., typically in terms of prices, outputs, trade flows and real An important modelling advance over the standard GTAP model is
incomes). that MAGNET explicitly characterises the rigidity in agricultural factor
In this study, a state-of-the-art recursive dynamic, multi-region, markets, both in terms of land transfer between different agricultural
multi-sector neoclassical CGE model, known as the Modular Applied activities; and in the labour and capital markets to characterise the
GeNeral Equilibrium Tool (MAGNET) (Woltjer and Kuiper, 2014), is wage and rent differentials that exist between agricultural and non-
used. A key advantage of MAGNET is its modular structure that allows agricultural labour and capital markets.3 As a result, agricultural sector
the user to easily switch on/off non-standard modelling extensions supply responsiveness in MAGNET is relatively inelastic compared with
which are pertinent to the research question at hand. Given this flex- GTAP. In addition, in contrast with the assumption of fixed agricultural
ibility, the model has been used in numerous contexts including land-
use change (e.g., Schmitz et al., 2014); EU domestic support (e.g.,
Boulanger and Philippidis, 2015); biofuels and bioeconomy (e.g., 3
See Appendix A2 for further discussion.

301
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

land supply in GTAP, the sustainability of land availability is measured dairy, processed sugar, processed rice, vegetable oils and fats, other
more precisely in MAGNET through the use of biophysical data on food). Fertilizers, non-food manufacturing, services and energy and
available agricultural land areas. More specifically, a region specific natural resources activities are aggregated into four sectors (see Table
asymptotic endogenous agricultural land supply function signals S4 of the Supplementary material).
available land areas corresponding to changes in the real rental rate of
land (Eickhout et al., 2009).4 The potential for bringing additional land 4.2. Macroeconomic impact
into agricultural production is limited to the maximum potentially
available land, estimated by the IMAGE land management model (van The CGE model captures the 'first-round' impacts from relative soil
Meijl et al., 2006; Doelman et al., 2018). The default IMAGE asymptote productivity changes across regions. Thus, whilst the magnitude of the
is defined as the total land available for agriculture, which excludes reverse productivity shocks provided by RUSLE is consistent in sign
areas with prohibitively high land conversion costs (mainly ice, desert across all regions, the strength of this effect is highly heterogeneous.
and wetlands), urban and non-productive protected areas (Woltjer and Those regions with larger (smaller) crop productivity deterioration will
Kuiper, 2014).5 exhibit marginal relative deteriorations (improvements) in competi-
tiveness, resulting in a marginal negative (positive) crop production
3.2. Model integration trend. In addition, the model also accounts for 'second-round' economy-
wide ripple effects which are both 'local' and 'broader' in nature. The
The soil erosion rates estimated by RUSLE are long-term averages former are felt through the re-allocation of agricultural land between
based on time-invariant environmental and topographic parameters, competing uses and the vertical transmission from upstream agriculture
and crop management and land cover change (Fig. 1), which change at to downstream food activities (i.e., supply of inputs). The latter reflects
a very slow pace over time. In the CGE model, the resulting equivalent the impacts on the returns to labour and capital (i.e., wages and rents)
regional land productivity change is typically modelled as an exogenous from their redistribution from agricultural to non-agricultural uses, and
technical change parameter in the land demand function, detailing the the resulting economy-wide repercussions on household incomes, pro-
ratio of output per unit of land input. duction and macroeconomic growth. Results show that global losses in
It is assumed that the productivity impacts of soil erosion rate re- crop production are clearly overestimated by a direct-impact compu-
ported for 2010 by the RUSLE model are already embedded within the tation.7 Unless otherwise stated, all marginal impacts reported are ei-
2011 GTAP benchmark data equilibrium. Thus, to assess these marginal ther in percentage terms, volumes or dollar values.
impacts, an exogenous reverse (positive) shock is applied to the land As expected, the macro impacts are fairly muted, given that the
productivity parameters to capture the soil erosion event that led up to annual land productivity shock is relatively moderate and concentrated
the 2011 benchmark year. The difference between this counterfactual in the agriculture sector. The general pattern is that soil erosion is not
and the benchmark data gives us a marginal estimate of the resulting beneficial to real gross domestic product (GDP) growth (Fig. 3 and
market impacts. Table S5 of the Supplementary material): the declining productivity in
agriculture arising from the deterioration of the land factor has an al-
4. Simulation results most unambiguous negative economic impact. In monetary terms, this
amounts to a loss of approximately 8 billion US dollars of GDP.
4.1. Land productivity losses due to soil erosion In all regions, a decrease in the production possibilities with the
same input availability should bestow negative macroeconomic impacts
Fig. 2 shows that the highest productivity losses are observed where to the region under consideration. This is particularly the case where
the highest erosion rates (mean erosion in arable lands > 11 t ha−1 estimated regional land productivity deteriorations due to soil erosion
yr−1) occur in countries with high share of agricultural land. In the are larger (i.e., Indonesia, 'Central America and the Caribbean').
majority of Caribbean countries (Nicaragua, Guatemala, Haiti, El Sal- Equally, regions which have a larger agricultural base and a relatively
vador, Honduras, Panama), Brazil, Central African countries (Congo, larger share of value added accruing from the land factor (i.e., India)
Liberia, Democratic Republic of Congo, Ivory Coast, Malawi and also show greater relative decreases in their GDP, despite more mod-
Ethiopia) and some parts of South-East Asia (Vietnam, Philippines, In- erate changes in land productivity. In relative terms, the biggest losers
donesia, Laos, South Korea) more than 70% of the arable land is ex- due to soil erosion are Indonesia and India, with recorded losses ap-
periencing severe erosion (> 11 t ha−1 yr−1). On the contrary, Aus- proximating 0.1% of GDP, whilst in Nigeria and 'Central America and
tralia, Canada, Saharan countries, the Russian Federation, Kazakhstan, the Caribbean', the reported loss is closer to 0.04% of GDP.
Uzbekistan, Ukraine and most of the European Union have less than 3% In other regions (i.e., Europe, USA and Canada, Oceania, MENA)
of their arable land under severe erosion. On average, more than 3.4 agriculture's share of GDP is relatively small, in some cases heavily
million km2 of arable land worldwide (24%) is suffering from severe subsidized, and land productivity losses are less pronounced. As a re-
erosion. sult, macroeconomic losses are negative and in some cases (USA and
To maximise the richness of available regional land productivity Canada, Oceania) even marginal gains are observed as these regions
estimates generated by the RUSLE model as input for the MAGNET find themselves in a relatively more favourable production and trade
model, outputs are aggregated to 115 countries (see Table S3 of the position (Tables 2A, 2Band Fig. 7).
Supplementary material), the results of which are presented as 18
macro-regions (8′large' countries plus 7 macro-regions grouping 4.3. Agriculture and food security
neighbouring countries and the rest of the world, Table 1)6 .
The sector aggregation in MAGNET includes the seven main GTAP 4.3.1. Production
agricultural cropping activities (i.e., rice, wheat, other cereals, horti- As a measure of global food security, Tables 2A, 2B show that food
culture, oilseeds, raw sugar, and a residual 'other cropping' activity) and production has decreased by approximately 33.7 million tonnes8
seven non-arable and food processing activities (i.e., livestock, meat,
7
Table S11 of the Supplementary material compares the marginal absolute
4
See Appendix A3 for further discussion. change in crop production by country as obtained from the CGE analysis and
5
For further details see Woltjer and Kuiper (2014), pp. 71-77. from a back-on-the-envelope direct-impact estimation. Additional comments
6
Results are also available for all 115 aggregated regions in the are provided in the Supplementary material.
8
Supplementary material document. Physical quantities are updated as ex-post calculations using endogenous

302
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

Fig. 2. Estimated annual absolute land productivity losses (%) from the Global RUSLE model. Country values are reported in Table S2 of the Supplementary material.

Table 1
. Regional aggregation for result visualization. In 2011, the ten largest producers of agricultural goods were: China, Brazil, India, USA, Indonesia, Russia, Thailand,
Nigeria, Argentina and France. We have kept disaggregated the first five countries and the only African country of the list. USA and Canada are aggregated together as
Canada's geographical characteristics a land productivity shock are more similar to those of the USA than Mexico's.
Macro-Regions MAGNET regions Arable land (million
ha)

Brazil Brazil 71
China China 112
India India 160
Indonesia Indonesia 20
Nigeria Nigeria 34
Russia Russian Federation 122
USA&Canada USA and Canada 196
Central Amer. and the Caribbean (CAmer& Costa Rica, Guatemala, Honduras, Nicaragua, Panama, Caribbean 5
Caribb)
Central Asia Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyzstan, Tajikistan 33
West, Central, East and South Africa Benin, Burkina-Faso, Cameroon, Cote d'Ivoire, Ghana, Guinea, Senegal, Togo, Rest of West-Central Africa, 160
(WCES Africa) Rest of South-Central Africa, Ethiopia, Kenya, Madagascar, Malawi, Mozambique, Rwanda, Tanzania,
Uganda, Zambia, Zimbabwe, Rest of East Africa, Botswana, Namibia, South Africa
Europe Austria, Belgium, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, 171
Ireland, Italy, Latvia, Lithuania, Netherlands, Poland, Portugal, Slovakia, Slovenia, Spain, Sweden, UK,
Switzerland, Norway, Albania, Bulgaria, Belarus, Croatia, Romania, Ukraine, Turkey
Middle East-North Africa (MENA) Bahrain, Iran, Israel, Jordan, Kuwait, Oman, Qatar, Saudi Arabia, UAE, Egypt, Morocco, Tunisia, Rest of 60
North-Africa
Mexico Mexico (Rest of North America) 24
Oceania Australia, New Zealand 48
South America Argentina, Bolivia, Chile, Colombia, Ecuador, Paraguay, Perú, Uruguay, Venezuela 62
South-East Asia (SE Asia) Japan, Republic of Korea, Mongolia, Cambodia, Laos, Malaysia, Philippines, Thailand, Vietnam, 80
Bangladesh, Nepal, Sri Lanka, Pakistan
Rest of the World Rest of the World (ROW) 48

(Table 2A), of which 22.5 million tonnes are crops (Table 2B), due to available countries (corresponding numbers are reported in Table S6.1
severe erosion. This is equivalent to 0.41% (0.27% for crops only) of and S6.2 of the Supplementary material). Due to the lower amount of
global agricultural production. Results are also illustrated in Fig. 4 agri-food products available in the international markets and the con-
(absolute variation in agri-food production in million tonnes) for all sequent price increase, the total value of these goods has increased by
24.9 billion US$.
Given the description of the 'first-round' model driver discussed
(footnote continued)
above, the contribution to the total impact on crop output varies sub-
changes in sector specific agricultural production volumes from the model over stantially across macro-regions. According to the output of the RUSLE
each period based on Ramankutty (2005). The original data source is FAOSTAT model, the areas of China and South-East Asia have larger land pro-
data on harvested areas and yields to derive the production quantities which are ductivity losses reflecting the larger soil erosion effects. As a result,
provided as a satellite account for MAGNET. these regions are major drivers in the global crop output deterioration.

303
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

Fig. 3. Marginal % and absolute change in GDP (US $, 2011 prices) due to severe soil erosion.

Table 2A
Marginal absolute change in selected crop activities due to severe soil erosion ('000 tonnes). Negative (positive) numbers mean output losses (gain).
Rice Wheat OthCereals Horticulture OilSeeds Sugar TOT Crops

Brazil −15 −279 175 −974 −1,030 −4,071 −6,330

China −522 1390 −53 −3,227 −716 −700 −3,865
India −888 −748 −143 −2,522 −510 −2,599 −7,739
Indonesia −770 0 −201 −695 −1,675 −334 −4,102
Nigeria −3 1 −20 −255 −10 0 −283
Russia 9 395 96 84 100 45 729
USA&Can 126 −95 627 431 758 −58 1,856
CAmer&Caribb −22 0 −34 −245 −47 −304 −690
Central Asia 2 221 10 38 85 −1 473
WCES Africa 4 17 57 −93 81 49 230
Europe 32 467 755 533 447 1,005 3,320
MENA 8 156 39 437 7 118 797
Mexico −2 −99 −165 −259 −32 −69 −654
Oceania 9 148 162 110 34 129 657
South America 48 115 178 −108 481 23 750
SE Asia −1,798 −270 −191 −2,450 −1,595 −226 −7,136
RestWorld −195 −89 −33 −305 −43 −7 −700
Tot World −3,976 1,343 1,278 −9,381 −3,662 −6,991 −22,526

Table 2B
Marginal absolute change in livestock and food activities due to severe soil erosion ('000 tonnes). Negative (positive) numbers mean output losses (gain).
Livestock Meat ProcSugar ProcRice VegOilFat Dairy TOT Agri-Food (2A + 2.2)

Brazil −385 −182 −1,089 19 −62 −9 −8,170

China −786 −498 −250 −418 −502 −22 −6,737
India −1,114 −207 −760 −451 −490 −268 −11,259
Indonesia −49 −5 −106 −512 −247 0 −5,031
Nigeria 0 0 2 81 −7 6 −228
Russia 3 14 17 4 35 11 817
USA&Can −112 −39 −55 93 84 −110 1,660
CAmer&Caribb −69 −17 −106 −13 −3 −11 −924
Central Asia −18 4 1 3 35 −4 495
WCES Africa 32 6 64 71 94 1 482
Europe 42 −3 602 27 250 35 4,232
MENA −28 35 66 18 44 −15 903
Mexico −98 −30 −18 0 −30 −1 −848
Oceania 72 19 61 7 23 −52 786
South America −128 −49 9 −11 144 −19 667
SE Asia −518 −256 −42 −1,106 −511 −43 −9,839
RestWorld −94 −56 5 −69 44 −28 −921
Tot World −3,242 −1,258 −1,594 −2,257 −1,097 −527 −33,725

More specifically, Indonesia, China, India and the rest of South East regions exhibit the largest crop output increases. More specifically,
Asia's crop production has decreased by approximately 4.1, 3.9, 7.7 and these two regions show crop production rises of 3.3 million tonnes and
7.1 million tonnes, respectively (Table 2A). A similar observation can 1.9 million tonnes, respectively.
be made for Brazil, where the result is a decrease in crop output of 6.3 Despite the negative impacts on crop production in Nigeria and
million tonnes. other big countries of Central-South Africa (e.g., Kenya, Ghana,
In contrast, for the 'USA and Canada' region and Europe, which had Ethiopia, see Table S6.1 and S6.2), overall African crop production rises
smaller crop productivity impacts from the RUSLE model (Fig. 2), both slightly as a result of severe soil erosion (around 375 thousand tonnes),

304
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

Fig. 4. Marginal absolute change in agri-food production (million -M- tonnes) due to severe soil erosion. Results are illustrated for 109 single countries, five macro-
regions and one residual region, the latter for clarity reasons is not shown in the map. Macro-regions and the residual region are illustrated in Figure S1 of the
Supplementary material.

in large part driven by the production gains recorded in South Africa heterogeneous. This observation occurs due to the combination of re-
and Northern African countries (i.e., Egypt)9 . This result is due to two gional patterns of soil erosion across the regions reported by RUSLE and
main factors. Firstly, whilst the demand for more marginal land in- the relative trade competitiveness of individual crops across regions,
creases in all regions (Table S7 and Fig. 5) to compensate for the lower captured in the MAGNET model. For example, rice production is found
land productivity, the availability of unused agricultural land is esti- to be acutely affected by the pattern of soil erosion. The average pro-
mated to be relatively more abundant in the African continent (ap- ductivity shock hitting the top 75% of world rice producers (principally
proximately 4 million against 2.3 million km2 of China, Brazil and in South East Asia and China) is 3.7%, compared with 2.1% for the
India). In this region, land demand expands by about 58,250 km² (26% remaining countries (not shown). As a result, examining the collective
of global rise), whilst in China, Brazil and India the increase is smaller, impact on paddy- and processed rice activities (Tables 2A, 2B), this
although significant (approximately 17,000, 29,500 and 7,400 km² single supply chain accounts for 19% of the global agri-food volume
respectively). Secondly, the countries located in the North and in the decrease. Similarly, horticultural products account for 28% of the agri-
South of the African continent account for a big share of the agricultural food volume decrease, which is also driven by South-East Asia and
production in Africa and compensate the substantial productivity losses China. Given Brazil's comparative advantage in soybean and sugar
occurred in the Central region. As a result, Africa as a continent ex- cane, the same observation can be made for these two crops. More
periences a slight improvement in its comparative advantage and po- precisely, oilseed makes up 11% of the overall agri-food production
sitive production trends. The same result holds for other regions (e.g. decrease, whilst the entire sugar production chain makes up 26% of the
'USA and Canada', Europe and Oceania), where one or the other reason total.
mentioned above may prevail in driving the positive production output. In the case of wheat and other cereals, global production increases
Globally, land demand increases by approximately 223,000 km², by 2.6 million tonnes (which in relative terms is 0.1% for both, see
equivalent to a 0.5% increase in global land use in agriculture. The Fig. 6), and reflects the fact that calculated region wide land pro-
largest contributions arise from cereals (27%), driven by the positive ductivity impacts from erosion effects for the key producers of these
change in production, horticulture (19%) and oil seeds (19%) activities crops, are relatively lower. For example, the largest wheat producers
(Table 3). (e.g., Canada and Russia) are hit by an average productivity shock of
Decomposing the result on production further (Supplementary ma- 1.3%, compared with 3% for the remaining countries.
terial), it emerges that the positive result is driven by cereals and In livestock and food processing activities, the local 'second-round'
horticulture production increases in South Africa and in north African model drivers discussed at the beginning of this section come to the
countries, while some West (Cameroon, Cote d'Ivoire, Ghana) and East fore. With decreased global production in many cropping activities,
African countries (Ethiopia, Kenya, Madagascar and Rwanda) suffer in feed costs are also higher because of soil erosion impacts. As a result,
terms of horticulture production loss, which are typically high value livestock, meat and dairy production is also lower (3.2 million tonnes,
added cash crops for these countries. 1.3 million tonnes and 527 thousand tonnes, see Table 2B). Similarly,
Importantly, it should be noted that whilst land productivity losses the upstream-downstream links between crops and food processing
are assumed to be uniform for all cropping activities within the same sectors show the implications of the net decrease in crop output on food
country, the market impacts on crop activities within a region is processing sectors.

9
This result cannot be directly observed from Table 2A as the North-Africa 4.3.2. Trade
region is aggregated with the countries of the Middle East. The results in Fig. 7 show the marginal impacts on the agri-food

305
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

Fig. 5. Marginal absolute change in land demand by country (km²) due to severe soil erosion. Results are illustrated for 109 single countries, five macro-regions and
one residual region, the latter for clarity reasons is not shown in the map. Macro-regions and the residual region are illustrated in Figure S1 of the Supplementary
material.

Table 3
. Marginal absolute change in land demand (km²) due to severe soil erosion. Percentage value in the last column indicate the percentage change relative to the
amount of land used in agriculture in 2011.
Rice Wheat OthCereals Hortic. OilSeeds Sugar OthCrops Total % of Tot

Brazil 1,393 41 8,747 3,786 9,660 4,020 1,909 29,556 1.1%

China 2,599 4,821 4,136 3,764 702 152 893 17,067 0.3%
India 2,024 1,151 1,980 972 277 203 797 7,404 0.5%
Indonesia 6,099 0 1,805 2,048 3,198 194 4,237 17,581 3.1%
Nigeria 312 20 1,774 2,406 976 6 344 5,838 0.8%
Russia 25 2,789 474 43 1,199 −1 12 4,541 0.2%
USA&Can 379 2,401 7,814 1,085 8,820 119 811 21,429 0.5%
CAmer&Caribb 357 2 1,247 1,495 175 657 670 4,603 1.7%
Central Asia 7 1,534 171 132 503 3 435 2,785 0.1%
WCES Africa 2,096 1,266 17,809 15,459 6,581 488 5,885 49,584 0.5%
Europe 48 2,365 3,236 1,529 2,611 206 272 10,267 0.4%
MENA 46 2,157 1,192 1,717 62 36 120 5,330 0.1%
Mexico 12 77 3,475 1,164 61 358 435 5,582 0.5%
Oceania 14 1,227 1,397 325 330 26 186 3,505 0.1%
South America 786 784 2,159 2,049 3,808 361 1,174 11,121 0.4%
SE Asia 10,338 217 2,114 2,871 1,547 788 2,714 20,589 1.0%
RestWorld 1,450 590 1,196 1,512 835 141 391 6,115 1.1%
Tot WORLD 27,992 21,540 60,820 42,571 41,358 7,765 21,292 223,338 0.5%

trade balance (i.e., exports minus imports) measured in millions of US $US in primary agricultural (crops and livestock) trade, soil erosion is
dollars. On the one hand, the 'production effect' determines the internal found to reduce this by approximately 8.5 billion $US. Similarly, of the
market balance and consequently available exports from each country/ total food trade of 900 billion $US, the corresponding soil erosion im-
region. On the other hand, with increases in real growth, rising real pact is recorded as 3.5 $US billion.
incomes drive additional demand for agri-food products. In developing
countries typified by lower per capita incomes, the marginal demand
increases are expected to be larger given the higher income elasticity of 4.3.3. Food prices
demand. Examining the net impact of these drivers on the trade bal- Examining the affordability of food, Fig. 8 clearly shows soil erosion
ances, large agri-food importers such as China and the rest of South East has inflated food prices due to the productivity effects on producer
Asia have further increased their trade deficits. In contrast, the 'USA prices in all countries. The most impacted commodity is paddy rice,
and Canada', Europe and the 'rest of South America', all of which are net whose world price has risen by 3.5%, followed by world prices in
exporters of agri-food commodities, gain a further relative competitive wheat, other cereals and other relevant staple foods (around 1.5%
edge from soil erosion, resulting in improvements in their agri-food larger)10 . The effects on primary agriculture are then transmitted to
trade balances.
Examining the impacts on total agri-food trade, of the 450 billion 10
The productivity driven effect drives a world price increase also in

306
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

Fig. 6. Marginal percentage change in global agri-food production due to severe soil erosion by country shocks.

Fig. 7. Marginal change in trade balance (exports

minus imports, millions of US$, 2011 prices) of agri-
food commodities. Initial trade balance values (in
millions of US$) are reported as bar labels. Negative
(positive) values indicate that the region is a net im-
porter (exporter) of agricultural products, while ne-
gative (positive) bars indicate that the region deterio-
rates (improves) its initial trade balance.

processed commodities. impacts on the food price index, for the more vulnerable members of
The effects of world prices are again mainly driven by shocks in the population whose food budget shares are particularly high, even
Asian countries, e.g. 37%, 25% and 15% of the change is due to the land marginal price changes could have important implications on the fa-
productivity losses in South Asia, China and Indonesia, respectively. mily food bill.
The same holds for processed rice as well. In terms of global price Decomposing the food price index changes, Fig. 9 also shows the
changes, it is interesting to note that China has the largest impact on extent to which the food price index within each region is mainly af-
most agri-food commodity prices, driving on average one-third of the fected by land productivity shocks from within that same region vis-à-
global price changes (Fig. 8). vis relative cost changes from imports from trading partners. For ex-
Fig. 9 shows the degree to which the affordability of food in each ample, although it is a large agricultural producer, India is only on a par
region has changed due to higher food prices. With food price index with self-sufficiency in most agricultural commodity categories. As a
increases of over two per cent, Indonesia and India are the countries result, India's food price index is almost dominated by the changing cost
whose food prices are negatively affected the most. Despite the muted structure in its domestic market. On the other hand, outliers are the

Fig. 8. Regional land productivity drivers of the marginal percentage change in world prices by commodity.

307
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

Fig. 9. Marginal percentage change in agri-food national price index due to severe soil erosion by country shocks.

Middle East-North Africa (MENA) and Central and South-African (CSA) Even under the (strong) assumption of existing compounded rates of soil
regions, where self-sufficiency levels are well below unity and heavy erosion over time, coupled with projected rising rates of population, the
trade dependence is more characteristic of their domestic markets. implications for food security, natural resource management practises
Thus, non-MENA and non-CSA region land productivity shocks make up (i.e., land, water) and stable societies, particularly in the poorest parts of
respectively 80% and 66% respectively of the impact on the food price the world, are concerning. This reinforces the need for greater engage-
index in each region. ment by stakeholders to raise awareness regarding the central function of
soil preservation in our society (Keesstra et al., 2016).
4.4. Water 11 However, a further look at the results reveals that, compared with
previous ‘first-order’ estimates of soil erosion costs, these findings draw
The land productivity loss due to severe soil erosion requires addi- markedly different conclusions. For example, in contrast to 'first order'
tional marginal land in production (see Fig. 5). Following the MAGNET estimates from Wolka et al. (2018), who measure a soil erosion driven
model assumption that the share of irrigated land in each crop activity production loss of 280 million tonnes in Africa, our study reveals a
is exogenously fixed, an increase in land use increases water abstraction surprisingly diverse picture. Crop production in the African continent
(Table 4 and Fig. 10). Globally, soil erosion has brought about a 1.6% increases marginally by 0.35 million tonnes (due to the positive pro-
increase of the water withdrawn for agricultural purposes (which is duction changes in South Africa and North African countries), since
equal to more than 48 billion cubic meters). In absolute terms, China, marginal land productivity losses for this continent as a whole are esti-
Indonesia and South-East Asia represent approximately 14%, 12% and mated to be lower than in other regions (e.g., China, Brazil, Indonesia).
23% of the global increase, due to the irrigation intensive system of rice Nonetheless, within the Sub-Saharan African region, the prospects for a
production. In proportional terms, Brazil, the 'USA and Canada' region number of African countries are more concerning. For example, some
and South America witness water abstraction increases of up to 5%. West African (Cameroon, Cote d'Ivoire, Ghana and Nigeria) and East
On a commodity basis, just under half of the water abstraction increase African countries (Ethiopia, Kenya, Madagascar and Rwanda) suffer
is due to the impacts of soil erosion in the paddy rice sector. As expected, losses in horticultural and cereals production, which are typically high
this figure is almost exclusively driven by the regions of Asia, due to im- value added cash crops for these countries. In recognition of soil de-
portance of this staple product in the diet (see also Table S8). To com- gradation, a number of soil conservation measures are already im-
pensate for the lower productivity of land, in these countries land demand plemented at regional scale and in many countries12 . For example, in
for rice production increases by about 21,000 km², corresponding to 75% Kenya small scale conservation tillage and terraces are used to improve
of global increase in land demand for this crop (Table S7). water storage capacity and crop land productivity. In Ethiopia, degraded
land areas have been enclosed from human and animal use and enhanced
by additional vegetative and structural conservation measures, to permit
5. Discussion and concluding remarks
natural rehabilitation (WOCAT, 2007; Giger et al., 2018).
Furthermore, comparing with the CGE study of Panagos et al.
Employing an interdisciplinary approach that links a global bio-
(2018), these results present a markedly different picture for the EU
physical model to a global economic model, this study takes a forward
since, unlike their study which only examines erosion in the EU, the
step in understanding the global economic costs of soil erosion. In the
current scenario design models simultaneous erosion effects throughout
context of the broader debate, it provides a direct input into recent
the globe. With its relatively milder erosion rates, the EU now is in a
strategies such as the Economics of Land Degradation initiative (ELD
relatively more favourable production and trade position, which con-
Initiative, 2015; Nkonya et al., 2016) and the Global Land Outlook
trasts sharply with the negative EU production impacts reported in
(GLO) currently proposed by United Nations Convention to Combat
Panagos et al. (2018).
Desertification (UNCCD, 2017).
Drilling down into the results, one also observes that even with an
As a headline figure, the results show that soil erosion is un-
erosion shock corresponding to a single year, there are noticeable
ambiguously detrimental to global food production, resulting in a non-
global shifts in agricultural production in China, India and Brazil. These
trivial decline in agricultural and food production of 33.7 million tonnes.
changes are particularly prevalent in the production of rice (and oil-
seeds on a lesser degree), which decreases by almost 0.5% globally.
(footnote continued)
commodities whose global production is increasing, like wheat and cereals, as
12
domestic price are rising globally. See for example the African Soil Partnership (http://www.fao.org/global-
11
For further discussion of the water computation in MAGNET, see Appendix soil-partnership/regional-partnerships/africa) or the Africa Soil Information
A4. Service (http://africasoils.net/).

308
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

Table 4
.Marginal change in water abstraction due to severe soil erosion (million m3). Percentage values in the last column indicate the percentage change relative to the
amount of water abstraction in 2011.
Rice Wheat OthCereals Hort. OilSeeds Sugar OthCrops Total Marg. % chg over tot withdr.

Brazil 974 0 40 191 31 494 68 1,799 3.9

China 1,870 2,248 1,682 475 118 25 282 6,699 1.6
India 1,329 1,057 177 281 50 391 464 3,748 0.5
Indonesia 4731 0 376 32 703 55 11 5,908 4.0
Nigeria 6 5 1 16 7 66 5 105 1.5
Russia 8 15 4 1 1 0 45 73 0.5
USA&Can 248 44 462 198 218 19 1,166 2,355 1.3
CAmer&Caribb 149 0 12 177 19 262 9 628 4.0
Central Asia 5 20 34 36 7 1 198 301 0.8
WCES Africa 916 114 164 333 133 278 402 2,339 2.6
Europe 56 34 144 217 306 45 427 1,230 1.3
MENA 48 177 63 1,090 111 44 657 2,189 0.9
Mexico 2 63 133 426 6 93 313 1,036 1.8
Oceania 0 0 2 1 0 0 188 192 1.8
South America 665 45 264 871 83 309 372 2,609 2.6
SE Asia 9,460 266 245 257 13 892 280 11,412 2.2
RestWorld 1,551 687 561 1129 258 81 1,842 6,110 2.2
TOTAL 22,018 4776 4,363 5,730 2,064 3,056 6,725 48,733 1.6

Fig. 10. Marginal change in water abstraction due to severe soil erosion (billion -G- m3). Results are illustrated for 109 single countries, five macro-regions and one
residual region, the latter for clarity reasons is not shown in the map. Macro-regions and the residual region are illustrated in Figure S1 of the Supplementary material.

Indeed, our study reveals that falling land productivity, particularly for erosion and increase soil organic carbon (Panagos and Katsoyiannis,
rice production, is a major driver of increased water abstraction in Asia. 2019). In the USA, the Farm Bill extends soil conservation compliance
From a trade perspective, the heterogeneous rates of erosion across the requirements in order to qualify for the crop insurance subsidy (Islam
planet give rise to accelerating current trends where net agri-food ex- and Reeder, 2014). At global scale, the FAO and its Global Soil Part-
porters such as USA, Canada, Europe and Oceanian countries continue nership launched in June 2018 a new programme to reduce soil de-
to improve their net trade balances at the cost of net food importers gradation for greater food and nutrition security in Africa. Other
such as China and South East Asian countries. countries are implementing local measures (WOCAT, 2007; Giger et al.,
These effects call for the prioritization of soil governance and con- 2018), yet a global multilateral environmental agreement on soil pro-
servation strategy in all countries and international policy agenda. In tection is missing (Montanarella and Alva, 2015).
this regard, the European Commission launched the Seventh Measures aimed at reinforcing ecosystem services, ad hoc regulation
Environment Action Programme, which requires that by 2020 land is of human interventions and active farmers' participation contribute to
managed sustainably and soil is adequately protected (Paleari, 2017). minimize soil erosion. To this aim, protection and restoration of diverse
Focusing on agricultural land, the EU's Common Agricultural Policy plant communities on slopes are essential, as trees and diversified ve-
(CAP) links support directly to the need to maintain agricultural land in getation increase soil resistance to rain erosivity (Berendse et al., 2015).
good condition, whilst the post-2020 CAP includes as one of its main Other measures such as reduced tillage, buffer strips, agroforestry, plant
objectives, efficient soil management linked to actions to reduce soil residues and cover crops enhance soil fertility and control water runoff

309
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

(Fageria et al., 2005; Triplett and Dick, 2008). whilst requiring further research, are beyond the scope of this paper.
As in all modelling endeavours, there are caveats to the study. Connected to this last point, future analysis could therefore seek to
Firstly, as discussed in Section 2, there is uncertainty surrounding the broaden the list of indicators beyond recognised metrics such as prices,
soil erosion estimates from the global biophysical model and the as- production, trade and GDP, where the latter has been criticised as a
sumption that land productivity losses occur only in severely eroded misleading measure of success or failure (Robert et al., 2014). Indeed,
land. Secondly, the assumption of average crop productivity losses due in the context of soil erosion, a broader set of indicators is very much
to soil erosion is based on a literature review but in the real world it can inspired by the realisation of the Sustainable Development Goals
vary from region to region. Further, physical and economic models (SDGs), particularly SDG 15, which targets indicators relating to land
typically work at different temporal and spatial scales. The need to degradation and protection of ecosystems (i.e. sedimentation, flooding,
interface RUSLE with MAGNET implies that the site-specific soil erosion landslides, water eutrophication, biodiversity loss, land abandonment,
data have to be adapted at the larger (national) spatial scale of the CGE destruction of infrastructures). The extension of soil erosion to en-
model. Finally, whilst the economic framework provides some insights capsulate these cost concepts may likely reveal even greater costs than
on the biophysical implications of soil erosion (e.g., land usage, water the loss of crop productivity (Telles et al., 2011). The views expressed
abstraction), a fuller treatment of the off-site costs (paid by the society) are purely those of the authors and may not in any circumstances be
such as destruction of infrastructures, sedimentation, flooding, biodi- regarded as stating an official position of the European Commission.
versity and soil carbon losses, landslides, and water eutrophication,

Appendix

Appendix A1 Estimation of the global soil erosion equation

The evaluation of the long-term annual soil loss is carried out using the RUSLE model in a Geographical Information Systems (GIS) environment.
RUSLE belongs to the class of detachment-limited models. Accordingly, the flow can theoretically transport an infinite quantity of sediment, but the
amount of sediment actually available to be transported is limited by the soil detachment capacity given by the rainfall erosivity factor of the model.
The soil loss in megagrams per hectare and year due to inter-rill and rill erosion processes are calculated according to Renard et al. (1997) by the
following multiplicative equation:
A g = R g*K g *LSg*Cg *Pg ((1))
where: Ag [Mg ha-1 yr-1] is the annual average soil loss, Rg [MJ mm hˉ¹ haˉ¹ yrˉ¹] is the rainfall erosivity factor, Kg[Mg h MJ-1 mm-1] is the soil
erodibility factor, LSg [dimensionless] is the slope length-slope steepness factor, Cg [dimensionless] is the land cover and management factor,Pg
[dimensionless] is the soil conservation or prevention practices factor. The subscript g stands for global.
According to Eq. (1), RUSLE consists of a multiplicative equation including six environmental parameters:

• Rainfall Erosivity: The rainfall erosivity factor R, or rainfall erosivity index (EI30), is a numerical descriptor of the rainfall’s ability to erode soil
(Wischmeier, 1958). It expresses the kinetic energy of raindrop's impact and the rate of associated runoff.
• Soil Erodibility: The soil erodibility K-factor [Mg ha MJ-1 mm-1] is an empirical parameter based on the measurements of specific soil erodibility
(Wischmeier and Smith, 1978). This parameter is generally measured based on some intrinsic soil properties such as texture, organic matter,
structure and permeability of the topsoil profile.
• Slope Length and Steepness Factor: The LS-factor, also called the topographic parameter, in the RUSLE model represents the influence of the
terrain topography on the sediment transport capacity of the overland flow (Wischmeier and Smith, 1978). To incorporate the impact of flow
convergence in the estimation of the slope-length factor (LS), the RUSLE equation proposed by Renard et al. (1997) replaced by the ones proposed
by Desmet and Govers (1996).
• Land Cover and Management Factor: The C-factor describes the land cover and management factor that measures the combined effect of all the
interrelated cover and management variables (Wischmeier and Smith, 1978). It may range from 0 to 1 depending on the ground cover. Generally,
values close to zero are typical of forested areas where the ground cover can reach up to 100%, whereas values close to one are typical of bare land.
• Support Practice Factor: The conversation support practice factor, P, is the ratio of soil loss with a conservation support practice like contouring,
strip cropping, terracing and subsurface drainage (Wischmeier and Smith, 1978). Values for the support practice P-factor are generally the most
uncertain and the most difficult to assess above the field-scale. Often, these are not taken into account in the vast majority of basin- and regional-
scale assessments.

Appendix A2 Labour and capital transfer in MAGNET

In the standard GTAP model, capital and labour are treated as perfectly mobile across different industrial uses. This implies that the return to
capital (i.e., rent) and labour (i.e., wage) is equal for each industry ‘i’. MAGNET follows the work on the agricultural variant of the GTAP model
(‘GTAP-AGR’) by Keeney and Hertel (2015). Thus, labour and capital transfer between the primary agricultural and non-primary agricultural sub-
sectors is made ‘sluggish’ via the usage of a Constant Elasticity of Transformation (CET) function (see Fig. A1). The policy implication is that in the
real world, there are observed differences in the return to capital and labour between the two sub-sectors. For both labour and capital, the elasticity
of transformation in each nest is the same as that employed in the GTAP-AGR model, and takes a value of 1.

Fig. A1. The CET Labour/Capital Allocation between agricultural and non-agricultural sub sectors.

310
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

Fig. A2. Theoretical agricultural land supply curve.

Appendix A3 Agricultural land supply in MAGNET

In the standard CGE model treatments, land supply is exogenous in each region. However, in reality, agricultural land supply can adjust due to
the idling of agricultural land or the conversion of land to agricultural uses. The supply of agricultural land depends on its biophysical suitability,
institutional factors (agricultural, urban and nature protection policies) and land price (Tabeau et al., 2006, p.3). Biophysical suitability refers to
climate, soil and water conditions that make a plot of land suitable for cultivation. Accordingly, biophysical parameters will define the maximum
potentially available land surface that can be used for agricultural purposes (the asymptote in Fig. A2). At the outset, the most productive land is used
first. With increases in land usage, farmers must employ less productive land implying that the marginal cost of conversion rises, which is reflected in
a higher land price. This relationship between land usage and prices gives an upward sloping supply curve (see Fig. A2).
Any point along the supply curve is feasible from an agronomic point of view, however, every country/region will be positioned on a specific
point, representing the current relative use of land in the agricultural sector. When the region is currently using a low proportion of all the potentially
available land, any increase in demand for agricultural land will lead to conversion towards agricultural uses at a modest increase in price (e.g. point
A in Fig. A2). In this zone of the supply curve, the supply elasticity is relatively higher, and the marginal cost of converting non-agricultural land into
agricultural land is relatively lower. However, when a region is currently cultivating most of the available land (e.g. point B in Fig. A2), any increase
in demand that requires the conversion of the scarce non-used land to agriculture, will lead to the conversion of the least productive land and at a
relatively higher marginal cost (land supply elasticity is low).
The assumed land supply function for each of the regions in the MAGNET model is:
B
L=A
P (A1)
where L is land supply, P is the real rental value of land, A is the maximum available agricultural land area (the land asymptote), and B is a positive
parameter. The resulting land supply elasticity Es in respect of land price is defined as:
A
Es = 1
L (A2)
In Tabeau et al. (2017), a full list of the land supply elasticities used in MAGNET can be consulted.
A4 Water abstraction in MAGNET
The MAGNET model includes a water module based on satellite data (Haqiqi et al., 2016) for irrigated and rainfed land areas and irrigated water
withdrawals (in cubic meters) for the 140 regions of the MAGNET model for the year 2011. The modelling is tops-down, where changes in water
withdrawals, calculated as an ex-post computation, are driven by proportional endogenous changes in irrigated land usage. Irrigated land use
changes are calculated by assuming that the share of irrigated land in all crop activities in the one-year period contemplated within this study
remains exogenously fixed. By linking the water withdrawals directly to land use instead of crop production implies that intensifying non-land inputs
(e.g. capital, labour and fertilizer) can increase crop production without leading to more water withdrawals.
References
Desmet, P. J. J., Govers, G. 1996. A GIS procedure for automatically calculating the USLE LS factor on topographically complex landscape units.
Journal of soil and water conservation 51, 427–433.
Renard, G., Foster, G.R., Weesies, G.A., Porter, J.P. 1991. RUSLE-Revised universal soil loss equation. Journal of Soil and Water Conservation 46,
30–33.
Wischmeier, W.H. 1958. A rainfall erosion index for a universal soil-loss equation. Soil Science Society of America, Proceedings 23, 246–249.
Wischmeier, W.H., Smith, D.D., 1978. Predicting rainfall erosion losses. Agriculture Handbook, n. 5.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.landusepol.2019.05.014.

References Altieri, M.A., Nicholls, C.I., 2017. The adaptation and mitigation potential of traditional
agriculture in a changing climate. Clim. Change 140 (1), 33–45. https://doi.org/10.
1007/s10584-013-0909-y.
Aguiar, A., Narayanan, B., McDougall, R., 2016. An overview of the GTAP 9 data base. J. Berendse, F., van Ruijven, J., Jongejans, E., Keesstra, S., 2015. Loss of plants pecies di-
Glob. Econ. Anal. 1 (1), 181–208. https://doi.org/10.21642/JGEA.010103AF. versity reduces soil erosion resistance. Ecosystems 18, 881–888. https://doi.org/10.

311
M. Sartori, et al. Land Use Policy 86 (2019) 299–312

1007/s10021-015-9869-6. Van Oost, K., Nearing, M., Ballabio, C., 2017. Global rainfall erosivity assessment
Boardman, J., Poesen, J., 2006. Soil Erosion in Europe: Major Processes, Causes and based on high-temporal resolution rainfall records. Sci. Rep. 7, 4175. https://doi.org/
Consequences. Soil Erosion in Europe. John Wiley & Sons, pp. 477–487 ISBN-13 978 10.1038/s41598-017-04282-8.
0-470-85910-0. Panagos, P., Standardi, G., Borrelli, P., Lugato, E., Montanarella, L., Bosello, F., 2018. Cost
Borrelli, P., Robinson, D.A., Fleischer, L.R., Lugato, E., Ballabio, C., Alewell, C., of agricultural productivity loss due to soil erosion in the European Union: from di-
Meusburger, K., Modugno, S., Schütt, B., Ferro, V., 2017. An assessment of the global rect cost evaluation approaches to the use of macroeconomic models. Land Degrad.
impact of 21st century land use change on soil erosion. Nat. Commun. 8, 2013. Dev. 29, 471–484. https://doi.org/10.1002/ldr.2879.
https://doi.org/10.1038/s41467-017-02142-7. Philippidis, G., Bartelings, H., Smeets, E., 2018a. Sailing into unchartered waters: plotting
Boulanger, P., Philippidis, G., 2015. The EU budget battle: assessing the trade and welfare a course for EU bio-based sectors. Ecol. Econ. 147, 410–421. https://doi.org/10.
impacts of CAP budgetary reform. Food Policy 51, 119–130. https://doi.org/10. 1016/j.ecolecon.2018.01.026.
1016/j.foodpol.2015.01.004. Philippidis, G., Sanjuán, A., Tabeau, A., van Berkum, S., Verma, M., 2018b. A Foresight
Corong, E., Hertel, T.W., McDougall, R.A., Tsigas, M.E., van der Mensbrugghe, D., 2017. Study of European East-West Agrifood Trade Options. German J. Agric. Econ. 67 (3),
The standard GTAP model, version 7. J. Glob. Econ. Anal. 2 (1), 1–119. https://doi. 160–175.
org/10.21642/JGEA.020101AF. Pimentel, D., Harvey, C., Resosudarmo, P., Sinclair, K., Kurz, D., McNair, M., Crist, S.,
Doelman, J.C., Stehfest, E., Tabeau, A., van Meijl, H., Lassaletta, L., Gernaat, D.E.H.J., Shpritz, L., Fitton, L., Saffouri, R., 1995. Environmental and economic costs of soil
Hermans, K., Harmsen, M., Vassilis, D., Biemans, H., van der Sluis, S., van Vuuren, erosion and conservation benefits. Science 267, 1117–1123. https://doi.org/10.
D.P., 2018. Exploring SSP land-use dynamics using the IMAGE model: regional and 1126/science.267.5201.1117.
gridded scenarios of land-use change and land based climate change mitigation. Glob. Poesen, J., 2018. Soil erosion in the Anthropocene: research needs. Earth Surf. Process.
Environ. Chang. Part A 48, 119–135. https://doi.org/10.1016/j.gloenvcha.2017.11. Landf. 43, 64–84. https://doi.org/10.1002/esp.4250.
014. Ramankutty, N., 2005. Global Land Use Data for Integrated Assessment Modelling. Report
Eickhout, B., Van Meijl, H., Tabeau, A., Stehfest, E., 2009. The impact of environmental Prepared for the U.S. Department of Energy: Award N. DE-FG02-01ER63215. .
and climate constraints on global food supply. In: Hertel, T., Rose, S., Tol, R. (Eds.), https://digital.library.unt.edu/ark:/67531/metadc794101/m2/1/high_res_d/
Economic Analysis of Land Use in Global Climate Change Policy. Routledge. 861318.pdf.
ELD Initiative, 2015. Report for Policy and Decision Makers: Reaping Economic and Renard, K., Foster, G., Weesies, G., McCool, D., Yoder, D., 1997. Predicting Soil Erosion by
Environmental Benefits From Sustainable Land Management. Available from Water: a Guide to Conservation Planning With the Revised Universal Soil Loss
www.eld-initiative.org at the web address:. . http://www.eld-initiative.org/ Equation (RUSLE). Agricultural Handbook. pp. 703.
fileadmin/pdf/ELD-pm-report_05_web_300dpi.pdf. Reuter, H.I., Nelson, A., Jarvis, A., 2007. An evaluation of void-filling interpolation
Erkossa, T., Wudneh, A., Desalegn, B., Taye, G., 2015. Linking soil erosion to on-site methods for SRTM data. Int. J. Geogr. Inf. Sci. 21, 983–1008. https://doi.org/10.
financial cost: lessons from watersheds in the Blue Nile basin. Solid Earth 6 (2), 1080/13658810601169899.
765–774. https://doi.org/10.5194/se-6-765-2015. Robinson, N., Regetz, J., Guralnick, R.P., 2014. EarthEnv-DEM90: a nearly-global, void-
Fageria, N.K., Baligar, V.C., Bailey, B.A., 2005. Role of cover crops in improving soil and free, multi-scale smoothed, 90m digital elevation model from fused ASTER and SRTM
row crop productivity. Commun. Soil Sci. Plant Anal. 36, 2733–2757. https://doi. data. ISPRS J. Photogramm. Remote Sens. 87, 57–67. https://doi.org/10.1016/j.
org/10.1080/00103620500303939. isprsjprs.2013.11.002.
Giger, M., Liniger, H., Sauter, C., Schwilch, G., 2018. Economic benefits and costs of Rutten, M., Shutes, L., Meijerink, G., 2013. Sit down at the ball game: how trade barriers
sustainable land management technologies: an analysis of WOCAT’s global data. Land make the world less food secure. Food Policy 38, 1–10. https://doi.org/10.1016/j.
Degrad. Dev. 29 (4), 962–974. foodpol.2012.09.002.
Haqiqi, I., Taheripour, F., Liu, J., van der Mensbrugghe, D., 2016. Introducing irrigation Schmitz, C., van Meijl, H., Kyle, P., Nelson, G.C., Fujimori, S., Gurgel, A., Havlik, P.,
water into GTAP data base version 9. J. Glob. Econ. Anal. 1 (2), 116–155. data can be Heyhoe, E., Mason d’Croz, D., Popp, A., Sands, R., Tabeau, A., van der Mensbrugghe,
downloaded from. https://www.gtap.agecon.purdue.edu/resources/res_display.asp? D., von Lampe, M., Wise, M., Blanc, E., Hasegawa, T., Kavallari, A., Valin, H., 2014.
RecordID=5168. Land-use change trajectories up to 2050: insights from a global agro-economic model
Hein, L., 2007. Assessing the costs of land degradation: a case study for the Puentes comparison. Agric. Econ. 45, 69–84. https://doi.org/10.1111/agec.12090.
catchment, southeast Spain. Land Degrad. Dev. 18 (6), 631–642. https://doi.org/10. Smeets, E., Tabeau, A., van Berkum, S., van Meijl, H., Woltjer, G., Moorad, J., 2014. The
1002/ldr.802. impact of the rebound effect of first generation biofuels on greenhouse gas emissions
Hengl, T., Mendes de Jesus, J., MacMillan, R.A., Batjes, N., Heuvelink, G.B.M., Ribeiro, E., in the EU. Sustainable and Renewable Energy Reviews 38, 393–403. https://doi.org/
Samuel-Rosa, A., Kempen, B., Leenaars, J.G.B., Walsh, M.G., Gonzalex, 2014. 10.1016/j.rser.2014.05.035.
SoilGrids1km — global soil information based on automated mapping. PLoS One Tabeau, A., Eickhout, B., van Meijl, H., 2006. Endogenous agricultural land supply: es-
9https://doi.org/10.1371/journal.pone.0105992. e105992. timation and implementation in the GTAP model. Paper Presented at the 9th Annual
Islam, R., Reeder, R., 2014. No-till and conservation agriculture in the United States: an Conference on Global Economic Analysis. https://www.gtap.agecon.purdue.edu/
example from the David Brandt farm, Carroll, Ohio. Int. Soil Water Conserv. Res. 2 resources/download/2731.pdf.
(1), 97–107. Tabeau, A., Helming, J., Philippidis, G., 2017. Land Supply Elasticities: Overview of
Keeney, R., Hertel, T., 2015. GTAP-AGR: A Framework for Assessing the Implications of Available Estimates and Recommended Values for MAGNET. ISBN 978-92-79-
Multilateral Changes in Agricultural Policies, GTAP Technical Paper N. 24. https:// 69102-7, doi:10.2760/852141, JRC106592. EUR 28626 EN, Publications Office of
www.gtap.agecon.purdue.edu/resources/res_display.asp?RecordID=1869. the European Union, Luxembourg. http://publications.jrc.ec.europa.eu/repository/
Keesstra, S.D., Bouma, J., Wallinga, J., Tittonell, P., Smith, P., Cerdà, A., Montanarella, L., bitstream/JRC106592/kj-na-28626-en-n_.pdf.
Quinton, J.N., Pachepsky, Y., van der Putten, W.H., Bardgett, R.D., Moolenaar, S., Telles, T.S., Guimarães, M.D.F., Dechen, S.C.F., 2011. The costs of soil erosion. Revista
Mol, G., Jansen, B., Fresco, L.O., 2016. The significance of soils and soil science to- Brasileira de Ciência do Solo 35 (2), 287–298.
wards realization of the United Nations Sustainable Development Goals. Soil 2, Triplett Jr, G.B., Dick, W.A., 2008. No-tillage crop production: a revolution in agri-
111–128. https://doi.org/10.5194/soil-2-111-2016. culture!. Agron. J. 100, S153–S165. https://doi.org/10.2134/agronj2007.0005c.
Lugato, E., Smith, P., Borrelli, P., Panagos, P., Ballabio, C., Orgiazzi, A., Fernandez- UNCCD, 2017. Global Land Outlook. Available at:. . https://knowledge.unccd.int/sites/
Ugalde, O., Montanarella, L., Jones, A., 2018. Soil erosion is unlikely to drive a future default/files/2018-06/GLO%20English_Full_Report_rev1.pdf.
carbon sink in Europe. 201.). Sci. Adv. 4 (11). https://doi.org/10.1126/sciadv. van Meijl, H., van Rheenen, T., Tabeau, Eickhout,B., 2006. The impact of different policy
aau3523. art. no. eaau3523. environments on agricultural land use in Europe. Agric. Ecosyst. Environ. 114 (1),
Martínez-Casasnovas, J.A., Ramos, M.C., 2006. The cost of soil erosion in vineyard fields 21–38. https://doi.org/10.1016/j.agee.2005.11.006.
in the Penedès–Anoia Region (NE Spain). Catena 68 (2-3), 194–199. https://doi.org/ van Meijl, H., Havlik, P., Lotze-Campen, H., Stehfest, E., Witzke, P., Pérez Domínguez, I.,
10.1016/j.catena.2006.04.007. Bodirsky, B., van Dijk, M., Doelman, J., Fellmann, T., Humpenoeder, F., Levin-
Montanarella, L., Lobos Alva, I., 2015. Putting soils on the agenda: the three Rio Koopman, J., Mueller, C., Popp, A., Tabeau, A., Valin, H., van Zeist, W.J., 2018.
Conventions and the post-2015 development agenda. Curr. Opin. Environ. Sustain. Challenges of global agriculture in the climate change context by 2050. Environ. Res.
15, 41–48. https://doi.org/10.1016/j.cosust.2015.07.008. Lett. 13 (2018), 064021.
Nkonya, E., Anderson, W., Kato, E., Koo, J., Mirzabaev, A., von Braun, J., Meyer, S., 2016. Wischmeier, W., Smith, D., 1978. Predicting rainfall erosion losses: a guide to con-
Global cost of Land degradation. In: Nkonya, E., Mirzabaev, A., von Braun, J. (Eds.), servation planning. U.S. Dep. Agric. Handb. 537, 1–69.
Economics of Land Degradation and Improvement – A Global Assessment for WOCAT, 2007. In: Hanspeter, L., Critchley, W. (Eds.), Where the Land Is Greener – Case
Sustainable Development. Springer International Publishing, Cham, pp. 117–165. Studies and Analysis of Soil and Water Conservation Initiatives Worldwide, ISBN
Paleari, S., 2017. Is the European Union protecting soil? A critical analysis of Community 978-92-9081-339-2.
environmental policy and law. Land Use Policy 64, 163–173. Wolka, K., Mulder, J., Biazin, B., 2018. Effects of soil and water conservation techniques
Panagos, P., Katsoyiannis, A., 2019. Soil erosion modelling: the new challenges as the on crop yield, runoff and soil loss in Sub-Saharan Africa: a review. Agric. Water
result of policy developments in Europe. Environ. Res. 172, 470–474. Manag. 207, 67–79. https://doi.org/10.1016/j.agwat.2018.05.016.
Panagos, P., Imeson, A., Meusburger, K., Borrelli, P., Poesen, J., Alewell, C., 2016. Soil Woltjer, G.B., Kuiper, M.H., 2014. The MAGNET Model: Module Description. LEI
conservation in Europe: wish or reality? Land Degrad. Dev. 27 (6), 1547–1551. Wageningen UR (University & Research centre), LEI Report 14-057, Wageningen.
Panagos, P., Borrelli, P., Meusburger, K., Yu, B., Klik, A., Jae Lim, K., Yang, J.E., Ni, J., Zhao, G., Mu, X., Wen, Z., Wang, F., Gao, P., 2013. Soil erosion, conservation, and eco-
Miao, C., Chattopadhyay, N., Sadeghi, Seyed H., Hazbavi, Z., Zabihi, M., Larionov, environment changes in the Loess Plateau of China. Land Degrad. Dev. 24, 499–510.
G.A., Krasnov, S.F., Gorobets, Av., Levi, Y., Erpul, G., Birkel, C., Hoyos, N., Naipal, V., https://doi.org/10.1002/ldr.2246.
Oliveira, P.T.S., Bonilla, C.A., Meddi, M., Nel, W., Al Dashti, H., Boni, M., Diodato, N.,

312