IMAGE - User contributions [en]

Land degradation/Description

2023-03-14T14:31:54Z

Stehfeste: nice that the GLO method is also described here. A few edits, also adjusted the title to describe methodology

{{ComponentDescriptionTemplate
|Reference=Oldeman et al., 1991;Batjes, 1997;Harris et al., 2013;Batjes, 2009;FAO et al., 2009
}}
<div class="page_standard">

Water erosion is the main cause of land degradation (1049 million hectares (Mha), followed by wind erosion (548 Mha), chemical degradation (239 Mha) and physical degradation (83 Mha) (GLASOD; [[Oldeman et al., 1991|Oldeman et al. (1991)]]). IMAGE assesses soil erosion by water ([[Hootsmans et al., 2001]]) by calculating a water erosion sensitivity index, ranging from zero (no erosion risk) to one (extremely high erosion risk). This risk is calculated for each grid cell as the compounded result from the following indices:
==A. Risk of soil erosion caused by water==
Water erosion is the main cause of land degradation (1049 million hectares (Mha), followed by wind erosion (548 Mha), chemical degradation (239 Mha) and physical degradation (83 Mha) (GLASOD; [[Oldeman et al., 1991|Oldeman et al. (1991)]]). IMAGE assesses soil erosion by water ([[Hootsmans et al., 2001]]) by calculating a water erosion sensitivity index, ranging from zero (no erosion risk) to one (extremely high erosion risk). This risk is calculated for each grid cell as the compounded result from the following indices:

* ''terrain erodibility index'': terrain erodibility represents the water erosion characteristics of the terrain in an index that combines surface relief and soil properties, expressed as index numbers. The relief index is a landform characteristic derived from a digital elevation model, calculated from the difference between minimum and maximum altitude in a 10 minute grid cell. The index is 1 for a difference of 300 m or more and zero for no altitude differences, with a linear relationship assumed between the two extremes. The soil erodibility index is derived from indices on soil texture, bulk density and soil depth. Soil characteristics were deduced from the 0.5x0.5 degree resolution in the [[WISE database]] ([[Batjes, 1997]]).

* ''rainfall erosivity index'': this index represents exposure to heavy rainfall, derived from the month of the year with the highest precipitation and number of wet (rainy) days in each month. Rainfall erosivity is largely determined by the intensity of rainfall events, because soil loss only occurs during periods of intense rainfall. Monthly rainfall intensities of between 0 and 2 mm per day are assigned an index value of zero, and days exceeding 20 mm receive a value of one, with a linear relationship assumed between these two end points. Climate data are used for the historical period ([[Harris et al., 2013]]). For future years, predictions are based on changes in precipitation according to scenarios generated by the climate model, see Component [[Atmospheric composition and climate]]. The number of wet days per month is assumed to be constant over time.

* ''land-use/land-cover index'': this index presents the level of protection against water erosion offered by various types of natural vegetation and crops. The basis for this index is the geographic distribution of land-cover types generated by the land-cover model. Most types of natural vegetation provide a high degree of protection against water erosion, while agriculture, and arable agriculture in particular, increases the vulnerability of the soil surface. A composite value is used for grid cells that contain agriculture, based on the distribution of agricultural crops in that world region.

All intermediate and resulting factors are expressed as dimensionless indices from zero to one, and so too is the end indicator, Water Erosion Sensitivity Index.

The susceptibility and sensitivity indices are calculated according to:

:<math>T = (Ia+ SE)/2 </math>
:<math>Ep = (T+R)/2</math>
:<math>WES = Ep*V</math>

with:
: <math>Ia</math> = relief index (-)
: <math>SE</math> = soil erodibility index (-)
: <math>T</math> = terrain erodibility index (-)
: <math>R</math> = rainfall erosivity index (-)
: <math>Ep</math> = water erosion susceptibility index (-)
: <math>V</math> = land-use/land-cover index (-)
: <math>WES</math> = Water Erosion Sensitivity Index (-)

Management systems are in use around the world to reduce the risk of erosion, such as building terraces, zero tillage, planting or conserving protective vegetation zones around fields, and high capacity drainage systems. The Water Erosion Sensitivity Index cannot capture all these and other interventions for the current situation, let alone into the future. The index only indicates areas potentially under threat. Impacts on crop production and soil quality cannot be derived directly from the indicator.

Comparison of the calculation above and the GLASOD degradation status maps by [[Oldeman et al., 1991|Oldeman et al. (1991)]] shows maximum correspondence with use of the classification in the table below. This classification can be used as a guide in analysing the water erosion sensitivity indicator.

<div class="thumbcaption dark">Classification of the Water Erosion Sensitivity Index</div>
<table class="pbltable">
<tr>
<th>Water Erosion Sensitivity Index</th>
<th>GLASOD soil degradation caused by water erosion</th>
</tr>
<tr>
<td>< 0.15</td>
<td>no/low</td>
</tr>
<tr>
<td>0.15 - 0.30</td>
<td>moderate</td>
</tr>
<tr>
<td>0.30 - 0.45</td>
<td>high</td>
</tr>
<tr>
<td>> 0.45</td>
<td>very high</td>
</tr>
</table>

==B. Human-induced soil changes==
Soil degradation is mostly reflected in changes in soil properties, such as soil depth, soil organic matter ({{abbrTemplate|SOM}}) content, and texture. Land cover and land use drive changes in soil properties. Land cover protects the soil against wind and water erosion, and provides organic matter to the soil. Land use tends to remove part of the biomass with harvested crops and residues and may increase mineralisation of SOM through tillage.

An empirical model denominated S-World has been developed that relates change in soil properties to topography, climate (average annual temperature and total annual precipitation), land management and land use, and land cover (as vegetation cover) ([[Stoorvogel, 2014]]; [[Stoorvogel et al., 2017]]). The following soil properties are considered:
* topsoil depth,
* soil depth,
* soil organic matter in the topsoil and subsoil , and
* soil texture (sand and clay content).

S-World is based on the global Harmonised World Soil Database ([[HWSD database|HWSD]]; ([[FAO et al., 2009]]) and the [[WISE database|WISE soil profile database]] ([[Batjes, 2009]]). The compound mapping units in HWSD were disaggregated using detailed terrain information, so that each grid cell could be linked to a unique soil type described in the WISE database. For each soil type, ranges for the main soil characteristics described above were assessed on the basis of the WISE soil profiles. The range of variable, i.e., soil property v for every soil type s is subsequently defined as [v<sub>ls</sub>..v<sub>hs</sub>] in which v<sub>ls</sub> corresponds to the 1<sup>st</sup> decile and v<sub>hs</sub> to the 9<sup>th</sup> decile. S-World downscales each soil property v based on 5 landscape properties or explanatory factors [''p<sub>1</sub>,p<sub>2</sub>… p<sub>5</sub>'']. These explanatory factors are: temperature, precipitation, slope, land management, and land cover. The land management is set to:
* 1.0 for cropland,
* 0.5 for mosaics of cropland and pasture or natural vegetation,
* 0.3 for pasture, and
* 0.0 for natural vegetation;
Land cover is characterised by a remotely sensed {{abbrTemplate|NDVI}} map.

The soil property v at location x with soil s is estimated as: {{FormulaAndTableTemplate|Formula1 Land degradation}}
with w<sub>x </sub>being a weight w∈ [0..1] that determines where v is in the range [v<sub>ls</sub>..v<sub>hs</sub> ]. Different explanatory factors represented by the landscape properties determine w. The weight at location x is calculated as:
{{FormulaAndTableTemplate|Formula4 Land degradation}}
The weight w<sub>px</sub> for landscape property p is calculated as:
{{FormulaAndTableTemplate|Formula2 Land degradation}}
In which c<sub>pv</sub> is a constant that indicates the relative importance of the landscape property p for a soil property v. The sign of c<sub>pv</sub> indicates whether there is a positive or negative relationship between the landscape property and the soil property.

When: {{FormulaAndTableTemplate|Formula3 Land degradation}}
and all the w∈ [0..1] then all values in the range [v<sub>ls</sub>..v<sub>hs</sub> ] are possible based on the landscape properties. Although in practice c is specific for each landscape property, soil type, and soil property, data are lacking to estimate c at that level of specificity. Therefore the model assumes that c is constant per soil and landscape property, or, in other words, the relative impact of landscape properties on a specific soil property is assumed to be constant over the different soil types.

The soil properties are estimated based on land management and land use. This allows for the estimation of soil properties under pristine conditions. For future years, the NDVI map is changed as a function of land use, forest management and assumptions on degradation. To assess pristine conditions, soil properties are calculated with land use set at natural, and land cover represented by the NDVI under pristine conditions.

With this procedure, a change in soil properties (topsoil depth, soil depth, SOM in topsoil and subsoil, and soil texture) can be calculated as a result of land use and land cover. Subsequently, additional soil characteristics, such as water holding capacity and water infiltration rate, can be derived from these soil property values by using pedo-transfer functions ([[Van Beek, 2012]]). These soil characteristics can be used in other models in the IMAGE framework, such as [[LPJmL model|LPJmL]] (Component [[Carbon cycle and natural vegetation]] ) and [[GLOFRIS model|GLOFRIS]] (Component [[Flood risks]]), as alternative input to assess the consequences of historical or future land degradation.
|=1〗, the w∈ [0..1] and all values in the range [v_ls..v_hs ] are possible based on the landscape properties. Although in practice c is specific for each landscape property, soil type, and soil property, data are lacking to estimate c at that level of specificity. Therefore the model assumes that c is constant per soil and landscape property, or, in other words, the relative impact of landscape properties on a specific soil property is assumed to be constant over the different soil types.
The soil properties are estimated based on land management and land use. This allows for the estimation of soil properties under pristine conditions. For future years, the NDVI map is changed as a function of land use, forest management and assumptions on degradation. To assess pristine conditions, soil properties are calculated with land use set at natural, and land cover represented by the NDVI under pristine conditions.
With this procedure, a change in soil properties (topsoil depth, soil depth, SOM in topsoil and subsoil, and soil texture) can be calculated as a result of land use and land cover. Subsequently, additional soil characteristics, such as water holding capacity and water infiltration rate, can be derived from these soil property values by using pedo-transfer functions (Van Beek, 2012). These soil characteristics can be used in other models in the IMAGE framework, such as LPJmL (Section 6.1) and GLOFRIS (Section 7.4), as alternative input to assess the consequences of historical or future land degradation.
<div>

== C. Agricultural degradation based on NDVI data ==
For the Global Land Outlook, IMAGE aimed to explore historical and future degradation effect on crop yields and, their repercussion on the agricultural economy ([[Van der Esch et al., 2021]]). To quantify this effect in relation to other drivers we use the satellite-observed NDVI trends that are shown to correlate with NPP and crop yields. it is assumed that in the 2000-2018 period crop yield reductions in line with NDVI trends take place. In a baseline approach, this trend is assumed to continue at a degressive rate. In a restoration scenario it is assumed that yields are gradually restored to the pre-2000 level by the year 2040 ([[Van der Esch et al., 2021]]). This allows to compare degradation and restoration trends to for example climate change impacts or agronomic improvements (see Figure below).

{{DisplayFigureLeftOptimalTemplate|Other Figure Land degradation 2}}

Terrestrial biodiversity/Description

2023-03-14T14:24:29Z

Stehfeste: just English correction while reading

{{ComponentDescriptionTemplate
|Reference=Bartholome et al., 2004;UNEP-WCMC, 2005;Dixon et al., 2001;Visconti et al., 2011;Alkemade et al., 2012;Alkemade et al., 2011a;Bouwman et al., 2002b;Bobbink et al., 2010;DMA, 1992;Meijer and Klein Goldewijk, 2009;Benitez-Lopez et al., 2010;UNEP, 2001;Verboom et al., 2014;Faith et al., 2008;Musters et al., submitted;Palma et al., 2021;Schipper et al., 2020;Leclère et al., 2020
}}
<div class="page_standard">

The GLOBIO model calculates changes in biodiversity in terrestrial ecosystems, based on seven drivers of biodiversity change: land-use change, land-use intensity, climate change, atmospheric nitrogen deposition, infrastructural development, encroachment and fragmentation.
Four steps in the model are distinguished:
# Drivers of biodiversity change derived from IMAGE results are combined with additional data;
# Mean Species Abundance (MSA) is calculated for each driver and year, using empirical relationships between driver and change in MSA ([[Alkemade et al., 2009]]);
# MSA values for each driver are aggregated to obtain one MSA values;
# Two additional indicators are calculated: Wilderness area, and Species Richness Index (Figure Flowchart).

MSA expresses the relationship of mean species abundance between a disturbed or managed ecosystem and an undisturbed ecosystem, on a scale from 1 (undisturbed or pristine) to 0 (complete loss). This concept is applicable for most ecosystems and dynamics of biodiversity loss, and allows to compare and aggregate across ecosystems and drivers. However, it ignores possible increase in species abundance due to natural processes or in certain agricultural systems, such as European high nature value farmland.

===Land use and land-use intensity ===
Changes in land use and land cover are major drivers of biodiversity change. Land use includes all human activities with a spatial component, such as forestry, agriculture, infrastructure and urban development. The impact of land use on biodiversity ranges from small (where the habitat quality is too poor for a limited subset of species) to large (where complete conversion of ecosystems results in habitat loss for a large number of species).

GLOBIO calculates maps of land-use categories and intensities for the year 2000. The starting point is land-cover data from GLC2000 ([[Bartholome et al., 2004]]) on the major types of forests, rangelands and agricultural land areas, at around 30 arc seconds resolution (1x1km near the equator). These data are combined with the World Database on Protected Areas ([[WDPA database]]; [[UNEP-WCMC, 2005]]) that distinguishes protected and non-protected areas. The land-cover classes obtained are summarised as proportions of cropland, forest and pasture for IMAGE grid cells of 5x5 minutes resolution.

For the period after 2000, changes in land use and land-use intensity from IMAGE are used as regional totals and allocated to the starting map. Data on cropland areas derived from the [[land-use allocation]] model are used as a total claim for each region. Three intensity classes are distinguished on the basis of management intensity ([[land-use allocation]]) for each region, calibrated with areas of irrigated, extensive and intensive croplands from the FAO's farming system typology ([[Dixon et al., 2001]]). Data on three forestry management types are derived from the [[forest management]] module, and data on two grazing intensities from the [[livestock systems]] module.

The pastoral grassland areas are allocated in natural rangelands. Grazing in mixed systems is assumed on managed pastures, where the natural vegetation would be densely forested biomes. The remaining grassland areas (e.g., semi-arid and arid grasslands, tundra) are considered natural areas. All regional cropland, forests and grazing areas are geographically distributed per land-use intensity class by adjusting the proportion per grid cell, avoiding protected areas ([[Visconti et al., 2011]]).

MSA values for all land-use types are derived from the literature (Alkemade et al., 2009; Alkemade et al., 2012) and applied to the land-use map, with proportions of each land-use intensity class to yield the MSA land-use map for the year considered.

===Climate ===
Climate is a key determinant of ecosystems and biodiversity. Climate change causes shifts in species occurrence and abundance, and ultimately may lead to local species extinction. Species distribution models ({{abbrTemplate|SDM}}) are used to describe relationships between climate variables and species distribution.

Regression equations are derived for each biome by applying a large number of SDMs to a series of climate scenarios, and calculating the proportion of remaining species per grid cell (0.5x0.5 degrees). The average proportion of remaining species per grid cell is related to the global mean temperature increase (GMTI) from IMAGE for the scenario considered ([[Alkemade et al., 2011a]]). The regression equation between GMTI and the proportion of remaining species is used to derive the map of MSA levels related to climate change for a given year.

===Nitrogen===
Nitrogen is a plant nutrient that stimulates growth, but some species benefit more than others and become more dominant with higher nitrogen availability. Thus, nitrogen deposition affects the species composition, mainly of plant and invertebrate species. Ecosystems can take up nitrogen without observable effects up to the level at which the assimilative capacity of the ecosystem is exceeded. This level of N input is defined as the critical load ({{abbrTemplate|CL}}).

Deposition rates of atmospheric nitrogen for current and future years are derived from IMAGE (Components [[Emissions]] and [[Nutrients]]), and the map of critical loads is based on Bouwman et al. ([[Bouwman et al., 2002b|2002b]]). The nitrogen exceedance is calculated by subtracting the critical load from the estimated deposition. For forested and grassland ecosystems, the {{abbrTemplate|MSA}} map for nitrogen is derived from the regression equation between nitrogen exceedance and the proportion of remaining species. Regression equations are derived from published impact studies on the effects of a nitrogen surplus on species composition ([[Bobbink et al., 2010]]).

===Infrastructure and Encroachment===
The construction and use of infrastructure, such as roads, railroads and built-up area, may have multiple impacts on biodiversity. Roads have a direct impact on species, for example as the result of traffic disturbance, road kills and habitat fragmentation (see below). There are also indirect impacts, such as increased human access to natural areas, increased hunting, gathering and tourism. Traffic disturbance reduces the breeding success of bird and mammal species, reducing their abundance close to infrastructure. Hunting and gathering reduce populations when intensity exceeds threshold values.

Data on infrastructure are derived from globally available road maps, such as the Digital Chart of the World ([[DMA, 1992]]) and the GRIP, Global Roads Inventory Project ([[Meijer and Klein Goldewijk, 2009]]). Direct impacts occur in a 500 m zone on both sides of roads and an MSA value is derived from a meta-analysis on disturbance effects ([[Benitez-Lopez et al., 2010]]).

Human settlements are the major access points to natural areas, and are likely to correlate with agricultural areas. Thus, 20 km impact zones are calculated around cropland areas and assigned as encroachment areas. Based on literature review of hunting activities, an MSA value of 0.7 is attributed to such zones. The MSA map for infrastructure and encroachment is obtained by combining the MSA map for direct (infrastructure) and indirect (encroachment) effects. In projections, the impact zone of direct effects is broadened according to the GLOBIO2.0 procedure ([[UNEP, 2001]]). Future impact zones for indirect effects are determined by the projections for agricultural areas.

===Ecosystem fragmentation ===
Conversion of natural land to intensive cropping and road construction change vast areas of contiguous wilderness into a fragmented landscape with remnants of natural areas remaining as isolated islands. These relatively small patches are likely to house fewer species than could be expected from their habitat quality, because the individual patches may be too small to sustain viable populations of some species. Based on literature data on minimum area requirements of species, a relationship is constructed between patch size and relative number of species compared to a non-fragmented situation, known as the minimum area requirement (MAR) curve ([[Verboom et al., 2014]]). The relative number of species in a certain patch according to this MAR curve is used as a proxy for mean species abundance (MSA).

The area of natural vegetation patches is calculated by reclassifying the GLC2000 Global Land Cover data into two classes: human-dominated land (including croplands and urban areas) and natural land. Contiguous cells of natural land are grouped together and with an overlay of main roads (see above) are used to produce a map of natural land patches.

In scenario projections, patch sizes change as agricultural land use expands and as new roads emerge ([[Verboom et al., 2014]]). Changes in patch sizes also change the relative number of species and the MSA biodiversity indicator.

===Aggregation===
Total MSA values per area unit are calculated by multiplying the individual MSA values related to the separate drivers of biodiversity change (Figure Flowchart) to arrive at the total effect of all drivers. The contribution of individual drivers to biodiversity loss is also calculated.
Wilderness areas are defined as natural areas with high (>0.8) MSA values. The Species Richness Index (SRI) is calculated by applying species–area relationships according to Faith et al. ([[Faith et al., 2008|2008]]), and using MSA values as a proxy for their intactness parameter. Aggregation from regional to global species richness is based on species lists in the Wildfinder database to avoid double counting ([[Musters et al., submitted]]).

===IMAGE-Land Emulator of Terrestrial Biodiversity===
IMAGE-Land can provide two biodiversity indicators as outputs without running GLOBIO: the MSA and the Biodiversity Intactness Index (BII). This alternative is suitable for time-constraint evaluations (as there is no need for data exchange and GLOBIO runs). Still, it comes with the downside of being a less accurate measure yet reasonably similar to GLOBIO's.

The IMAGE-Land MSA is based on the GLOBIO implementation ([[Schipper et al., 2020]]), but it only accounts for three pressures: Land use, Nitrogen Deposition and Climate Change. The validation of the results between IMAGE-LAnd and GLOBIO indicates a good agreement between the data regarding MSA-Plants (R2 = 0.92) and a lower agreement for MSA-Vertebrates (R2 = 0.72), suggesting that IMAGE-Land MSA Vertebrates should be used parsimoniously.

The BII evaluates terrestrial biodiversity based on land-use changes only. As this index is not part of the GLOBIO outputs, our approach derives from the BII calculations from the PREDICTS Biodiversity model ([[Palma et al., 2021]]), which is used in the Bending the Curve project ([[Leclère et al., 2020]]). A procedure like the one used for the MSA is implemented to calculate BII as part of the IMAGE-land module directly. The BII is the average abundance across a set of species in each area relative to their reference population, which would be populations before any human impacts but are usually populations in the least impacted settings available. Unlike the MSA, the BII accounts only for land use as a driver of species populations. Even though it falls short in including other pressures driving species losses, the level of detail in land-use classes for defining the index is larger than in the MSA, leading to a more sensitive land-use-specific index.
</div>

Agricultural economy/Description

2023-03-14T14:10:05Z

Stehfeste: edited the description of food waste and diets, also adding references to relevant publications.

{{ComponentDescriptionTemplate
|Reference=Hertel, 1997;Britz, 2003;Armington, 1969;Huang et al., 2004;Helming et al., 2010;Banse et al., 2008;Bruinsma, 2003;Woltjer et al., 2011;Van Meijl et al., 2006;Eickhout et al., 2009;Overmars et al., 2014;Alexandratos and Bruinsma, 2012;Gustavsson et al., 2011;Gustavsson et al., 2013
}}
<div class="page_standard">

The MAGNET model ([[Woltjer et al., 2014]]) is based on the standard GTAP model ([[Hertel, 1997]]), which is a multi-regional, static, applied computable general equilibrium ({{abbrTemplate|CGE}}) model based on neoclassical microeconomic theory. Although the model covers the entire economy, there is a special focus on agricultural sectors. It is a further development of GTAP regarding land use, household consumption, livestock, food, feed and energy crop production, and emission reduction from deforestation or afforestation.

===Demand and supply===
Household demand for agricultural products is calculated based on changes in income, income elasticities, preference shift, price elasticities, cross-price elasticities, and the commodity prices arising from changes in the supply side. Demand and supply are balanced via prices to reach equilibrium. Income elasticities for agricultural commodities are consistent with FAO estimates ([[Britz, 2003]]), and dynamically depend on purchasing power parity ({{abbrTemplate|PPP}}) corrected GDP per capita. The supply of all commodities is modelled by an input–output structure that explicitly links the production of goods and services for final consumption via different processing stages back to primary products (crops and livestock products) and resources. At each production level, input of labour, capital, and intermediate input or resources (e.g., land) can be substituted for one another. For example, labour, capital and land are input factors in crop production, and substitution of these production factors is driven by changes in their relative prices. If the price of one input factor increases, it is substituted by other factors, following the price elasticity of substitution.

===Regional aggregation and trade===
MAGNET is flexible in its regional aggregation (140 regions). In linking with IMAGE, MAGNET distinguishes 28 individual large world regions, closely matching the regions in IMAGE (Figure [[Region classification map|IMAGE regions]]). Slightly more detail is provided the European regions in order to properly model the EU single market. Similar to most other {{abbrTemplate|CGE}} models, MAGNET assumes that products traded internationally are differentiated according to country of origin. Thus, domestic and foreign products are not identical, but are imperfect substitutes (Armington assumption; [[Armington, 1969]]).

===Land use===
In addition to the standard [[GTAP database|GTAP model]], MAGNET includes a dynamic land-supply function ([[Van Meijl et al., 2006]]) that accounts for the availability and suitability of land for agricultural use, based on information from IMAGE (see below). A nested land-use structure accounts for the differences in substitutability of the various types of land use ([[Huang et al., 2004]]; [[Van Meijl et al., 2006]]). In addition, MAGNET includes international and EU agricultural policies, such as production quota and export/import tariffs ([[Helming et al., 2010]]).

===Biofuel crops===
MAGNET includes ethanol and biodiesel as first-generation biofuels made from wheat, sugar cane, maize, and oilseeds ([[Banse et al., 2008]]) and the use of by-products ({{abbrTemplate|DDGS}}, oilcakes) from biofuel production in the livestock sector. Second-generation biofuels are also included, with the potential amount of residues available from IMAGE/TIMER ([[Daioglou et al., 2016]]).

===Livestock===
MAGNET distinguishes the livestock commodities of beef cattle, dairy cattle, other cattle (sheep & goats), dairy cattle, poultry, and pig and other animal products. The first three are the ruminant sectors which are grass and crop fed, while the poultry and pigs sectors are crop fed. Modelling the livestock sector includes different feedstuffs, such as feed crops, co-products from biofuels (oil cakes from rapeseed-based biofuel, or distillers grain from wheat-based biofuels), and grass ([[Woltjer, 2011]]). Grass may be substituted by feed from crops for ruminants.

===Land supply===
In MAGNET, land supply is calculated using a land-supply curve that relates the area in use for agriculture to the land price. Total land supply includes all land that is potentially available for agriculture, where crop production is possible under soil and climatic conditions, and where no other restrictions apply such as urban or protected area designations (see also Component Land-use allocation). In the IMAGE model, total land supply for each region is obtained from potential crop productivity and land availability on a resolution of 5x5 arcminutes ([[Mandryk et al., 2015]]). The supply curve depends on total land supply, current agricultural area, current land price, and estimated price elasticity of land supply in the starting year. Regions differ with regard to the proportion of land in use, and with regard to change in land prices in relation to changes in agricultural land use.

===Reduced land availability===
By restricting land supply in IMAGE and MAGNET, the models can assess scenarios with additional protected areas, or reduced emissions from deforestation and forest degradation ({{abbrTemplate|REDD}}). These areas are excluded from the land supply curve in MAGNET, leading to lower elasticities, less land-use change and higher prices, and are also excluded from the allocation of agricultural land in IMAGE (e.g., [[Overmars et al., 2014]], [[Doelman et al., 2018]]).

===Intensification of crop and pasture production===
Crop and pasture yields in MAGNET may change as a result of the following four processes:
# autonomous technological change (external scenario assumption);
# intensification due to the substitution of production factors (endogenous);
# climate change (from IMAGE);
# change in agricultural area affecting crop yields (such as, decreasing average yields due to expansion into less suitable regions; from IMAGE).

Biophysical yield effects due to climate and area changes are calculated by the IMAGE crop model and communicated to MAGNET. Likewise, also the potential yields and thus the yield gap can be assessed with the crop model in IMAGE. External assumptions on autonomous technological changes are mostly based on FAO projections ([[Alexandratos and Bruinsma, 2012]]), which describe per region and commodity, the assumed future changes in yields for a wide range of crop types. In MAGNET, the biophysical yield changes are combined with the autonomous technological change to give the total exogenous yield change. In addition, during the simulation period, MAGNET calculates an endogenous intensification as a result of price-driven substitution between labour, land and capital. Projections of crop yield increase in IMAGE-MAGNET and other global agricultural models were evaluated recently ([[Van Zeist et al., 2020]]). In IMAGE, regional yield changes due to autonomous technological change and endogenous intensification according to MAGNET are used in the spatially explicit allocation of land use (Component [[Land-use allocation]]).

===Food Waste Reduction===
In earlier work like for the SSP scenarios, we applied a uniform reduction of food waste (Stehfest et al. 2019, [[Doelman et al., 2018]]). From IMAGE 3.2 onwards, high- and medium-income regions can reduce food waste to achieve the lowest level among them (we typically do not adress food waste in low-income regions, where food waste tends to be low and largely occurring on the farm level). This implementation is based on Gustavsson et al. ([[Gustavsson et al., 2011|2011]]; [[Gustavsson et al., 2013|2013]]), from where we derive the levels of food waste for five commodity types (cereals, other_plant_based, meat, fish_seafood, melk_eggs) and three food supply chain steps (primary, processing, consumption). Then, we find the lowest food waste level among high- and medium-income regions for each commodity type and food supply chain step and define them as food waste targets. Then, we calculate the change in the production efficiency that resembles the food waste reduction necessary for the regions to achieve their food waste targets. Finally, we calculate the MAGNET shock required to meet the targets and run MAGNET with the new production efficiencies, which resemble the reduction in food waste. The reduction in food waste reduces the pressure on the food system resulting in less agricultural land use, lower GHG emissions and reduced food prices.

===Diet Changes===
Another way to reduce the impact of food consumption on the environment is by adopting healthier diets, as also indicated in earlier work from the IMAGE team (e.g. Stehfest et al. 2009, 2013). These diets could be entirely plant-based or have reduced in meat and dairy consumption, or follow prescribed healthy diets addressing all food groups (Willet et al. 2019). In all cases, the changes especially in livestock-based commodities lead to reduced agricultural land use, and modifications in trade and production systems. Furthermore, in some scenarios we also explore the substitution of meat through artificial meat (Van Vuuren et al., 2018). These changes in consumption are prescribed for MAGNET, where also substitutions with other food commodities are specified.
</div>

Agricultural economy/Description

2021-10-29T13:30:08Z

Stehfeste:

{{ComponentDescriptionTemplate
|Reference=Hertel, 1997;Britz, 2003;Armington, 1969;Huang et al., 2004;Helming et al., 2010;Banse et al., 2008;Bruinsma, 2003;Woltjer et al., 2011;Van Meijl et al., 2006;Eickhout et al., 2009;Overmars et al., 2014;Alexandratos and Bruinsma, 2012
}}
<div class="page_standard">

The MAGNET model ([[Woltjer et al., 2014]]) is based on the standard GTAP model ([[Hertel, 1997]]), which is a multi-regional, static, applied computable general equilibrium ({{abbrTemplate|CGE}}) model based on neoclassical microeconomic theory. Although the model covers the entire economy, there is a special focus on agricultural sectors. It is a further development of GTAP regarding land use, household consumption, livestock, food, feed and energy crop production, and emission reduction from deforestation or afforestation.

===Demand and supply===
Household demand for agricultural products is calculated based on changes in income, income elasticities, preference shift, price elasticities, cross-price elasticities, and the commodity prices arising from changes in the supply side. Demand and supply are balanced via prices to reach equilibrium. Income elasticities for agricultural commodities are consistent with FAO estimates ([[Britz, 2003]]), and dynamically depend on purchasing power parity ({{abbrTemplate|PPP}}) corrected GDP per capita. The supply of all commodities is modelled by an input–output structure that explicitly links the production of goods and services for final consumption via different processing stages back to primary products (crops and livestock products) and resources. At each production level, input of labour, capital, and intermediate input or resources (e.g., land) can be substituted for one another. For example, labour, capital and land are input factors in crop production, and substitution of these production factors is driven by changes in their relative prices. If the price of one input factor increases, it is substituted by other factors, following the price elasticity of substitution.

===Regional aggregation and trade===
MAGNET is flexible in its regional aggregation (140 regions). In linking with IMAGE, MAGNET distinguishes 28 individual large world regions, closely matching the regions in IMAGE (Figure [[Region classification map|IMAGE regions]]). Slightly more detail is provided the European regions in order to properly model the EU single market. Similar to most other {{abbrTemplate|CGE}} models, MAGNET assumes that products traded internationally are differentiated according to country of origin. Thus, domestic and foreign products are not identical, but are imperfect substitutes (Armington assumption; [[Armington, 1969]]).

===Land use===
In addition to the standard [[GTAP database|GTAP model]], MAGNET includes a dynamic land-supply function ([[Van Meijl et al., 2006]]) that accounts for the availability and suitability of land for agricultural use, based on information from IMAGE (see below). A nested land-use structure accounts for the differences in substitutability of the various types of land use ([[Huang et al., 2004]]; [[Van Meijl et al., 2006]]). In addition, MAGNET includes international and EU agricultural policies, such as production quota and export/import tariffs ([[Helming et al., 2010]]).

===Biofuel crops===
MAGNET includes ethanol and biodiesel as first-generation biofuels made from wheat, sugar cane, maize, and oilseeds ([[Banse et al., 2008]]) and the use of by-products ({{abbrTemplate|DDGS}}, oilcakes) from biofuel production in the livestock sector. Second-generation biofuels are also included, with the potential amount of residues available from IMAGE/TIMER (TO ADD: Daioglou et al., 2016).

===Livestock===
MAGNET distinguishes the livestock commodities of beef cattle, dairy cattle, other cattle (sheep & goats), dairy cattle, poultry, and pig and other animal products. The first three are the ruminant sectors which are grass and crop fed, while the poultry and pigs sectors are crop fed. Modelling the livestock sector includes different feedstuffs, such as feed crops, co-products from biofuels (oil cakes from rapeseed-based biofuel, or distillers grain from wheat-based biofuels), and grass ([[Woltjer, 2011]]). Grass may be substituted by feed from crops for ruminants.

===Land supply===
In MAGNET, land supply is calculated using a land-supply curve that relates the area in use for agriculture to the land price. Total land supply includes all land that is potentially available for agriculture, where crop production is possible under soil and climatic conditions, and where no other restrictions apply such as urban or protected area designations (see also Component Land-use allocation). In the IMAGE model, total land supply for each region is obtained from potential crop productivity and land availability on a resolution of 5x5 arcminutes ([[Mandryk et al., 2015]]). The supply curve depends on total land supply, current agricultural area, current land price, and estimated price elasticity of land supply in the starting year. Regions differ with regard to the proportion of land in use, and with regard to change in land prices in relation to changes in agricultural land use.

===Reduced land availability===
By restricting land supply in IMAGE and MAGNET, the models can assess scenarios with additional protected areas, or reduced emissions from deforestation and forest degradation ({{abbrTemplate|REDD}}). These areas are excluded from the land supply curve in MAGNET, leading to lower elasticities, less land-use change and higher prices, and are also excluded from the allocation of agricultural land in IMAGE (e.g., [[Overmars et al., 2014]])(TO ADD: Doelman et al., 2018).

===Intensification of crop and pasture production===
Crop and pasture yields in MAGNET may change as a result of the following four processes:
# autonomous technological change (external scenario assumption);
# intensification due to the substitution of production factors (endogenous);
# climate change (from IMAGE);
# change in agricultural area affecting crop yields (such as, decreasing average yields due to expansion into less suitable regions; from IMAGE).

Biophysical yield effects due to climate and area changes are calculated by the IMAGE crop model and communicated to MAGNET. Likewise, also the potential yields and thus the yield gap can be assessed with the crop model in IMAGE. External assumptions on autonomous technological changes are mostly based on FAO projections ([[Alexandratos and Bruinsma, 2012]]), which describe per region and commodity, the assumed future changes in yields for a wide range of crop types. In MAGNET, the biophysical yield changes are combined with the autonomous technological change to give the total exogenous yield change. In addition, during the simulation period, MAGNET calculates an endogenous intensification as a result of price-driven substitution between labour, land and capital. Projections of crop yield increase in IMAGE-MAGNET and other global agricultural models were evaluated recently [TO ADD van Zeist et al. 2020]. In IMAGE, regional yield changes due to autonomous technological change and endogenous intensification according to MAGNET are used in the spatially explicit allocation of land use (Component [[Land-use allocation]]).
</div>