A future of increasing atmospheric carbon dioxide concentrations, changing climate, growing human populations, and shifting socioeconomic conditions means that the global agricultural system will need to adapt in order to feed the world. These changes will affect not only agricultural land but terrestrial ecosystems in general. Here, we use the coupled land use and vegetation model LandSyMM (Land System Modular Model) to quantify future land use change (LUC) and resulting impacts on ecosystem service indicators relating to carbon sequestration, runoff, biodiversity, and nitrogen pollution. We additionally hold certain variables, such as climate or land use, constant to assess the relative contribution of different drivers to the projected impacts. Some ecosystem services depend critically on land use and management: for example, carbon storage, the gain in which is more than 2.5 times higher in a low-LUC scenario (Shared Socioeconomic Pathway 4 and Representative Concentration Pathway 6.0; SSP4-60) than a high-LUC one with the same carbon dioxide and climate trajectory (SSP3-60). Other trends are mostly dominated by the direct effects of climate change and carbon dioxide increase. For example, in those two scenarios, extreme high monthly runoff increases across 54 % and 53 % of land, respectively, with a mean increase of 23 % in both. Scenarios in which climate change mitigation is more difficult (SSPs 3 and 5) have the strongest impacts on ecosystem service indicators, such as a loss of 13 %–19 % of land in biodiversity hotspots and a 28 % increase in nitrogen pollution. Evaluating a suite of ecosystem service indicators across scenarios enables the identification of tradeoffs and co-benefits associated with different climate change mitigation and adaptation strategies and socioeconomic developments.
Exploring how the agricultural system might shift under different
plausible future climate and socioeconomic changes is critically
important for understanding how the future world – with
a population increase by 2100 ranging from 1.5 billion to nearly
6 billion people
Declining biodiversity due to the loss and degradation of habitat
As global environmental and societal changes continue over the
coming decades, it is critical that we understand not just the
impacts on the natural world, but how those impacts feed back
onto humanity. To explore the possible future evolution of the
Earth system and society, models have been developed that
simulate the global economy, the natural world, and their
interactions. A substantial body of research has been built up
using such models to examine how future land use change will
affect individual ecosystem services such as carbon sequestration
Previously
LPJ-GUESS is
a dynamic global vegetation model that simulates – here, at
a spatial resolution of 0.5
LPJ-GUESS simulates variables that can be used as indicators of
a number of provisioning and regulating ecosystem services (see
also Table 1 in
PLUM is designed to produce
trajectories of land use and management based on socioeconomic
trends and grid-cell-level crop and pasture productivity at
a resolution of 0.5
Demand is satisfied at the country level by either domestic
production or imports, the balance between which is determined
considering commodity prices, management costs (fertilizer,
irrigation, land conversion, and “other management” such as
pesticide use), and changing LPJ-GUESS-simulated productivity due
to climate change and
To solve for land use areas and inputs that satisfy demand, PLUM
uses least-cost optimization, which allows for short-term resource
surpluses and deficits. Such imbalances can be significant in the
real world: global supply of major cereal crops frequently swings
5 % to 10 % out of equilibrium on an annual aggregate
basis, and more extreme imbalances can be seen at the scale of
individual countries
The composition of livestock feed (in terms of which crops are used) is assumed to be flexible, which can result in large interannual fluctuations in demand and production of individual crops as their prices change relative to one another. This is seen, for example, in Fig. S10 in the Supplement, where oil crop demand in the US and Canada triples from one year to the next. This assumption is not expected to materially affect the results in terms of gross decadal trends in total agricultural area and management inputs.
As outputs (feeding into LPJ-GUESS for use in LandSyMM), PLUM
produces half-degree gridded maps of land use area (cropland,
pasture, and non-agricultural land), crop distribution (fraction
of cropland planted with each crop type), irrigation intensity,
and nitrogen fertilizer application rate. Land use areas are
calculated as net change, which neglects certain dynamics – such
as shifting cultivation – that can have significant impacts on
modeled carbon cycling especially in some regions
This section and Fig.
LandSyMM structural overview. Ovals represent external input data and white rectangles represent model runs, with arrows indicating data flow from one model run to the next. Gray rectangles represent model coupling processes whose external inputs have been excluded for simplicity; more information on these can be found in the Supplement.
The first step in running LandSyMM is to perform
“yield-generating” runs in LPJ-GUESS. A simulation of the
historical period generates a model state, which is needed so that
vegetation and soil condition can be fed into subsequent runs
(Fig.
PLUM then combines the future potential yields from LPJ-GUESS
(averaged over 5-year time steps) with its own estimates of
future commodity demand to project land use areas, fertilizer
application, and irrigation intensity
(Fig.
The outputs of land use and management from PLUM for a given
2011–2100 scenario are fed into a final LPJ-GUESS run in order to
produce projections of the ecosystem service indicators analyzed
here (Fig.
Naming convention for LandSyMM runs analyzed in this work, based on land use and management (LU, mgmt.), climate, and
In addition to the LPJ-GUESS runs forced with harmonized
PLUM-output land use and management trajectories, we perform
several experiments to examine the impact of different factors on
the land use and management projections generated by PLUM and
thus the ecosystem service indicators simulated by LPJ-GUESS in
the PLUM-forced runs. By holding either climate, atmospheric
The experiments treated here are based around combined future climate–socioeconomic scenarios. Future population growth and economic development are derived from the Shared Socioeconomic Pathways
Scenarios of future climate change and atmospheric
We use climate input data from the fifth Coupled Model Intercomparison Project
Future socioeconomic data – country-level population and GDP projections – are taken from version 0.93 of the SSP database
Historical land use areas (cropland and pasture fractions), irrigation, and synthetic nitrogen fertilizer application levels were taken from the LUH2 dataset
LPJ-GUESS simulates a number of output variables that here serve as the basis for quantifying ecosystem services.
The carbon sequestration performed by terrestrial ecosystems is
measured as the simulated change in total carbon stored in the land
system, including both vegetation and soil. Ecosystem nitrogen in
LPJ-GUESS is lost in liquid form via leaching (a function of
percolation rate and soil sand fraction), and in gaseous form through
denitrification (1 % of the soil mineral nitrogen pool per day) and fire. Here, we combine these into a value for total N loss. LPJ-GUESS also simulates the emission of isoprene and monoterpenes – the most prevalent BVOCs in the atmosphere
LPJ-GUESS simulates basic hydrological processes such as evaporation, transpiration, and runoff. The latter is calculated as the amount of water by which soil is oversaturated after precipitation, leaf interception, plant uptake, and evaporation. We present change in average annual runoff as a general indicator of trend in water availability. After
Finally, we assess how much land is converted to agriculture within the Conservation International (CI) hotspots, a set of 35 regions covering less than 3 % of the Earth's land area but containing half the world's endemic plant species and over 40 % of the world's endemic vertebrate animal species
LandSyMM simulates net global loss of natural land area over the 21st century in all scenarios (Fig.
Percent change in global socioeconomic, land management, and
atmospheric variables between 2001–2010 and 2091–2100. Ruminant
demand given in units of feed-equivalent weight.
Cropland expansion happens at a more or less constant rate in SSP3 and SSP4, but these scenarios experience very different trajectories of crop commodity demand: SSP4 approximately levels off around mid-century, whereas SSP3 experiences only a brief slowdown in growth followed by constantly increasing demand through 2100 (Fig. S5). The majority of the increased demand in the first half of the century is satisfied by fertilizer application, which increases by more than 75 % from the 2010s to the 2050s while crop area increases by less than 15 %. However, management inputs per hectare in SSP3-60 approximately plateau after mid-century (Fig. S6), while crop demand rises 16 %. Cropland area expands about 10 % between 2050 and 2100, with boosted productivity – thanks to climate change and/or
Although population growth in SSP5-85 is more than twice that of SSP1-45, PLUM simulates very similar trajectories of global crop demand in both: an increase until about 2040 followed by a decrease for the rest of the century, with SSP5-85 crop demand ending slightly higher. SSP5-85 livestock demand increases about 20 % more than in SSP1-45, which explains the rest of the difference in global caloric needs between the two scenarios (Fig. S5). However, because SSP5-85 experiences much stronger climate change and
Figure
Several regional patterns in crop area change stand out in Fig. North America loses cropland in parts of the Great Plains (mainly Although crop demand in South Asia (here, India, Sri Lanka, Pakistan, Afghanistan, Bangladesh, Nepal, and Bhutan) increases by more than 100 % in SSP5-85 and 170 % in SSP3-60 (Fig. S12), after harmonization the cropland area in that region is greatly reduced: approximately 30 % and 20 %, respectively. The raw PLUM outputs showed less loss (8 % and 10 %, respectively) but the same general pattern. Even so, PLUM projects that the region's crop production would approximately double in both scenarios to satisfy most of the increased demand (Fig. S12). While some of this is accomplished through increased management inputs in a region where the yield gap is large in the baseline, it also depends markedly on yield boosts due to increased rainfall (Fig. S4) and rising PLUM expects sub-Saharan Africa to experience crop production increases even larger than South Asia, ranging from China's crop demand peaks by about 2040; by the end of the century, it has either returned to (SSP3-60) or dropped past 2010 levels (by 30 %, 40 %, and 25 % for SSP1-45, SSP4-60, and SSP5-85, respectively; Fig. S15). Crop imports decrease from 14 % of demand to less than 6 %. This fits well with apparent net losses of cropland area in all scenarios, but note that harmonization switched SSP1-45's projection from an 8.5 % gain to a 15 % loss. Moreover, whereas PLUM projected cropland abandonment to occur in the montane shrublands and steppe of the Tibetan Plateau, after harmonization it occurs throughout the eastern temperate and subtropical forests. Slight cropland expansion projected by PLUM in China's subtropical moist forests is increased 300 %–600 % by harmonization in all scenarios except SSP1-45 (
Pasture area is projected to expand significantly in the western Amazon in all scenarios (although in SSP1-45 this is strongly exaggerated by harmonization) and even more so in all scenarios in the African rainforest (Fig.
The African pasture expansion even occurs in SSP1-45, the “sustainability” scenario
First, PLUM makes no assumption about changes in food production needs besides what occurs due to population and GDP changes. The storyline for SSP1, however, with its “low challenges to mitigation”, suggests that people will gradually shift to lower-meat diets
Second, the land use modeling components of most integrated assessment models (IAMs) – for example, all those contributing to the LUH2 trajectories
The spread in land use area projections between the most extreme scenarios is much higher in this work than in
Carbon stored in the land system increases for all SSP–RCP scenarios, primarily due to an increase in vegetation carbon (Fig.
Percent global change in ecosystem service indicators
between 2001–2010 and 2091–2100. CSLF: Congolian swamp and lowland
forests (see Sect.
The contrast between effects of changing climate and atmospheric
Vegetation carbon increases are most pronounced in the tropical and boreal forests (Fig.
Maps showing difference in mean vegetation carbon between 2001–2010 (“2000s”) and 2091–2100 (“2090s”) for
Our results for carbon sequestration fall near the lower end of
previously reported projections.
Another study with LPJ-GUESS,
Global precipitation increases in all scenarios
(Fig.
Such regional patterns in runoff change are arguably more
important than global means, since impacts of low water and
flooding are actually felt at the level of individual river
basins. To evaluate regional impacts, we calculated how much land
area was subjected to intensified wet and/or dry extremes
(Sect.
Between 1971–2000 and 2071–2100 under SSP5-85, basins comprising
48 % of land area showed increasing flood risk, with
a mean P95
Fraction of land with changing drought and/or flood risk between the last three decades of the 20th and 21st centuries in SSP5-85. Numbers in parentheses give each group's mean percent change in runoff. LandSyMM results have been aggregated to basin scale. AK2017:
Most of the changes in SSP5-85 result from climate change, with
some notable exceptions. Land use change alone contributes notably
to increasing drought risk in eastern China, Pakistan, and
northwest India (Fig.
Contribution of land use change in SSP5-85 to
Our results for SSP5-85 are compared with the RCP8.5 ensemble from
Another important difference between
While the evolution of total global nitrogen loss is fairly similar for all scenarios over the first two decades of the simulation, there are notable differences by the end of the century. SSP3-60 and SSP5-85 show large increases in N loss of 28 % and 22 %, respectively. N loss increases about half as much in SSP4-60 (11 %) and only slightly in SSP1-45 (2 %).
Our N loss at the end of the historical period was similar to that
of
One interesting pattern is that climate and management changes can
have similar effects on N losses. SSP3-60 has global fertilizer
application more than double by the end of the century, while
SSP5-85 fertilizer application at end of the run is slightly lower
than in 2011 (Fig.
Global combined BVOC emissions over 2001–2010 totaled
Decreases in isoprene emissions are primarily driven by tropical
deforestation for agriculture, especially the expansion of pasture
in central Africa and South America, and to a lesser extent by the
expansion of cropland in the southeastern US (Fig. S19),
although the latter is exaggerated by harmonization. The
suppressive effect of increasing [
It is important to keep in mind that the implications of changing BVOC emissions depend on complex, regionally varying atmospheric chemistry that governs their effects on existing species (e.g., methane) and the formation of secondary products (e.g., ozone and aerosols). The LandSyMM framework, lacking as it does an atmospheric chemistry model, can thus inform only a surface-level discussion of the possible effects of changing BVOCs. However, we wish to provide context for the benefits and detriments associated with changing BVOC emissions, as well as some limitations related to our model setup.
The globally decreased BVOC emissions in all scenarios could
contribute a cooling effect in the future, due to expected lower
tropospheric ozone concentrations, shorter methane lifetime, and
enhanced photosynthesis thanks to more diffuse radiation. This
could be counteracted somewhat by warming arising from the reduced
formation of secondary aerosols, and it is important to note that
the effects on climate are likely to vary from region to region
The large expansion of agricultural land in SSP3-60 has direct
consequences for habitats in biodiversity hotspots, which lose over
13 % of their non-agricultural land in that scenario
(Fig.
It should be noted that area loss in biodiversity hotspots will
not necessarily correspond to linear decreases in species
richness.
This work is among the first to comprehensively consider the impacts of future land use and land management change on a suite of ecosystem services under different possible futures of climate and socioeconomic change. Using a uniquely spatially detailed, process-based coupled model system, we show that scenarios with high socioeconomic challenges to climate change mitigation – SSP3 and SSP5 – consistently have some of the most severe consequences for the natural world and the benefits it provides humanity via carbon sequestration, biodiversity, and water regulation. These two scenarios also most strongly affect biogeochemical cycling of nitrogen and BVOCs; while increases in nitrogen losses are generally detrimental, the impact of decreased BVOC emissions is likely to vary regionally. However, various elements of uncertainty – related to PLUM parameter values, global climate model selection, and model design – affect these results and remain to be explored.
Policymakers and other stakeholders need options for how we can meet
the needs of a growing and changing society while achieving climate
and sustainable development goals
The code for harmonizing land use and management is available for download on Zenodo
The supplement related to this article is available online at:
All authors contributed to the conceptual design
of LandSyMM. PAl contributed most of the initial text for Sect. 2.2. AA contributed most of the initial text regarding BVOCs in the introduction and text in Sect. 3.2.4 regarding [
The authors declare that they have no conflict of interest.
The authors would like to thank Jonathan Doelman for sharing data about the IMAGE scenarios
This research has been supported by the Helmholtz Association Impulse and Networking fund, the HGF ATMO program, and the UK's Global Food Security Programme project “Resilience of the UK food system to Global Shocks” (RUGS, BB/N020707/1).The article processing charges for this open-access publication were covered by a Research Centre of the Helmholtz Association.
This paper was edited by Stefan Dekker and reviewed by two anonymous referees.