Rise and fall of vegetation annual primary production resilience to climate variability projected by a large ensemble of Earth System Models’ simulations

Abstract

Climate change is affecting natural ecosystems and society. Anticipating its impacts on vegetation
resilience is critical to estimate the ecosystems’ response to global changes and the reliability of
the related ecosystem services, to support mitigation actions, and to define proper adaptation
plans.
Here, we compute the Annual Production Resilience Indicator from gross primary production (GPP)
data simulated by a large ensemble of state-of-the-art Earth System Models (ESMs) involved in
the last Coupled Model Intercomparison Project (CMIP6) of the Intergovernmental Panel on Climate Change (IPCC).
In the Sustainability (Taking the Green Road) and Middle of the Road scenarios (ssp126 and ssp245), the areas where vegetation shows increasing GPP resilience are wider than the areas with decreasing resilience. The situation drastically reverses in the Fossil-fuel Development (Taking the Highway) scenario (ssp585). Among the larger countries, Brazil is exposed to the highest risk of experiencing ears with anomalously low GPP, especially in the Taking the Highway scenario.

Full text

Environ. Res. Lett. 16 (2021) 105001 https://doi.org/10.1088/1748-9326/ac2407
OPEN ACCESS
RECEIVED
17 May 2021
REVISED
31 August 2021
ACC EPT ED FOR PUB LICATI ON
6 September 2021
PUBLISHED
16 September 2021
Original content from
this work may be used
under the terms of the
Creative Commons
Attribution 4.0 licence.
Any further distribution
of this work must
maintain attribution to
the author(s) and the title
of the work, journal
citation and DOI.
LETTER
Rise and fall of vegetation annual primary production resilience to
climate variability projected by a large ensemble of Earth System
Models’ simulations
Matteo Zampieri1,∗, Bruna Grizzetti1, Andrea Toreti1, Pierluca de Palma2and Alessio Collalti3
1European Commission—Joint Research Centre (JRC), Ispra, Italy
2Fincons SPA, Vimercate, Italy
3Institute for Agriculture and Forestry Systems in the Mediterranean, National Research Council of Italy (CNR-ISAFOM), Perugia (PG),
Italy
∗Author to whom any correspondence should be addressed.
E-mail: matteo.zampieri@ec.europa.eu
Keywords: climate change, resilience, agriculture, natural ecosystems, ecosystem services, annual production resilience indicator
Supplementary material for this article is available online
Abstract
Climate change is affecting natural ecosystems and society. Anticipating its impacts on vegetation
resilience is critical to estimate the ecosystems’ response to global changes and the reliability of the
related ecosystem services, to support mitigation actions, and to define proper adaptation plans.
Here, we compute the Annual Production Resilience Indicator from gross primary production
(GPP) data simulated by a large ensemble of state-of-the-art Earth System Models involved in the
last Coupled Model Intercomparison Project (CMIP6) of the Intergovernmental Panel on Climate
Change. In the Sustainability (Taking the Green Road) and Middle of the Road scenarios (ssp126
and ssp245), the areas where vegetation shows increasing GPP resilience are wider than the
areas with decreasing resilience. The situation drastically reverses in the Fossil-fuel Development
(Taking the Highway) scenario (ssp585). Among the larger countries, Brazil is exposed to the
highest risk of experiencing years with anomalously low GPP, especially in the Taking the Highway
scenario.
1. Introduction
The Sustainable Development Goals, formally
embraced by the 2010 Conference of Parties, recog-
nize the importance of ensuring conservation, restor-
ation and sustainable use of terrestrial ecosystems and
their services, and strengthening the resilience and
adaptive capacity to climate-related hazards (SDG-
15 and SDG-13, respectively; United Nations 2016).
Stable ecosystems, characterized by small variations
from their average state despite changes in environ-
mental conditions, are indeed considered healthy and
reliable in terms of the services they provide (MAE
2005, Costanza et al 2014). Ecosystems in good con-
dition are necessary to secure the sustainability of
human activities and human well-being (Maes et al
2020).
The concept of resilience is closely connected to
ecosystem stability. Resilience has been defined either
as the larger disturbance that a system can absorb
without losing its structure, relationships and func-
tionalities (Holling 1973,1996, Walker et al 2004,
Yi and Jackson 2021) or as the time required by an
ecosystem to recover and return back to the equilib-
rium state after a disturbance (Pimm 1984, Yi and
Jackson 2021). These definitions—termed ‘ecological
resilience’ and ‘engineering resilience’, respectively—
are conceptually clear but do not directly provide a
practical way to measure resilience (Morecroft et al
2012, Scheffer et al 2015). In fact, a quantitative
estimation of resilience requires objective methods to
identify and measure the external stresses and shocks
(Meyer 2016). Also as a result of such indeterminacy,
a large number of indicators was proposed to meas-
ure different aspects of resilience (de Keersmaecker
et al 2014, Scheffer et al 2015, Meyer 2016, Yi and
Jackson 2021). Up to date, none of these meth-
ods has been used to evaluate vegetation resilience
© 2021 The Author(s). Published by IOP Publishing Ltd

Environ. Res. Lett. 16 (2021) 105001 M Zampieri et al
at the global level and in future climate scenarios
yet.
Gross primary production (GPP)—the total car-
bon fixation by plants—is a primarily important ter-
restrial ecosystem function, at the point that it was
also considered to be strongly related to resilience
itself (Moore et al 1993, Stone et al 1996, Bryant
et al 2019). Climate change is indeed expected to alter
vegetation GPP resilience by potentially comprom-
ising the availability of water for vegetation in dry
regions (Santini et al 2014, Zampieri et al 2019) and
in general by increasing the frequency, amplitude and
duration of extreme events that are detrimental for
vegetation productivity (Dosio et al 2018, Naumann
et al 2018). At the same time, the increase of atmo-
spheric CO2concentration coming along with global
warming is expected to bring positive effects in terms
of vegetation photosynthetic rate (although acclima-
tion should be also considered) i.e. the so called ‘CO2
fertilization effect’ (Sage and Kubian 2007) and water
use efficiency (Peters et al 2018).
Here, we used the Annual Production Resilience
Indicator (Rp), defined as the squared mean annual
GPP divided by its variance, which was recently pro-
posed for a statistical evaluation of the production
resilience of natural vegetation (Zampieri et al 2019)
and agricultural systems from annual production
time-series (Zampieri et al 2020b). Rpis a simple but
powerful indicator with several interesting properties.
Being inspired by the ecological definition of resili-
ence, Rpis proportional to the amplitude of the largest
disturbances that the system can absorb (measure by
their rareness) and it is potentially consistent with the
engineering definition (i.e. the return timing) as well
(Zampieri 2021). It increases with diversity (number
of species) and it accounts for memory effects, i.e. for
perturbation recovery timings longer than a season
(Zampieri et al 2020b).
We compute the Annual Production Resili-
ence Indicator from an ensemble of 480 Earth
System Models (ESMs) simulations included in
the Sixth Coupled Model Intercomparison Project
(CMIP6, www.wcrp-climate.org/wgcm-cmip/wgcm-
cmip6) and involved in the Sixth Assessment Report
of the Intergovernmental Panel on Climate Change
(IPCC, www.ipcc.ch/report/ar6/wg1/). ESMs are
global climate models with an explicit representa-
tion of carbon processes and cycling over land, atmo-
sphere and the oceans (Dahan 2010, Randall et al
2019), allowing to explore future climate variability
of vegetation GPP according to different greenhouse
gases emission scenarios.
We quantify the relative changes in resilience
of the GPP production with respect to period
1985–2015 for the near and the mid-long terms
(2021–2050 and 2051–2100) under three scenarios of
socio-economical global changes, corresponding to
different levels of greenhouse gases emissions and
land-use (i.e. the Sustainability, the Middle of the
Road, and the Fossil-fuel Development scenarios) and
we compute country level statistics of the larger pro-
jected changes.
2. Data and methods
The Annual Production Resilience Indicator (Rp),
is based on two assumptions (Zampieri 2021). For
annual producing systems such as agriculture or nat-
ural ecosystems that are sufficiently adapted to the
environmental conditions and to the local climate,
it is sensible to assume that the largest disturb-
ances are rarer compared to the ‘normal’ conditions
(assumption 1) and that the largest disturbances res-
ult in larger impacts of the annual production val-
ues (assumption 2). Under such conditions, the size
of the disturbance can be univocally measured by its
rareness e.g. the return period of production anom-
alies (T∗). Focusing on the production function only,
the ecological resilience can be then simply meas-
ured by T∗
MAX, which is the return period of the
largest adverse event that the system can cope with
before losing completely the production ability. This
approach is not sensitive to changes in composition
and structure of the ecosystems, so it may allow for
adaptation according to a more ‘modern’ interpreta-
tion of ecological resilience (Walker et al 2004).
For homogeneous production systems, it can
be demonstrated that T∗
MAX is proportional to the
annual production resilience indicator, defined as:
Rp=µ2/σ2,(1)
where µis the mean and σis the standard deviation
of the annual production (see appendix A available
online at stacks.iop.org/ERL/16/105001/mmedia). In
case the annual production resilience indicator is
evaluated over a region including bare ground, the
indicator is sensitive to the vegetated portion only
(Zampieri et al 2019).
It is interesting and potentially useful to disen-
tangle the effects of changes in the mean and in the
variability of GPP on the annual production resilience
indicator. This can be accomplished by approximat-
ing the RPchange with a first order 2D Taylor expan-
sion of equation (1) as a function of the changes in the
mean and in the standard deviation of GPP as follows:
Rp_s =Rp_h + ∆Rp,(2)
where the s stands for ‘scenario’, the h stands for ‘his-
torical’, and ∆represents the difference between two
periods:
∆Rp≃∂R/∂µ ×∆µ+∂R/∂σ ×∆σ(3)
where ∂is the partial derivative. By computing the
derivatives and dividing both members of equation
(3) by Rpone obtains:
∆Rp/Rp=2∆µ/µ −2∆σ/σ. (4)
2

Environ. Res. Lett. 16 (2021) 105001 M Zampieri et al
Table 1. ESM simulations producing the annual GPP data used in this study; with information on: modules delegated to the
representation of land surface processes and GPP simulations (and associated reference publications); number of simulations available
in the Earth System Grid Federation nodes (ESGF, esgf.llnl.gov) until 31 December 2019 for the historical period and for ssp126, ssp245
and ssp585 future scenarios.
ESM Land model Historical ssp126 ssp245 ssp585
ACCESS-ESM-5 CABLE w/carbon cycle
(de Kauwe et al 2015)
1 1 1 1
CESM2-WACCM CLM5 (Lawrence et al 2019) 3 1 1 1
CESM2 CLM5 (Lawrence et al 2019) 10 2 3 2
CNRM-CM6-1 ISBA with fixed LAI
monthly climatology
(Garrigues et al 2015a,
Garrigues et al 2015b)
30 1 6 6
CNRM-ESM-1 ISBA with interactive LAI
(Garrigues et al 2015b,
Garrigues et al 2015a)
7 1 5 5
CanESM5-CanOE CLASS-CTEM (Arora and
Scinocca 2016)
3 3 3 3
CanESM5 CLASS-CTEM (Arora and
Scinocca 2016)
50 50 50 50
EC-Earth3-Veg LPJ-GUESS v4 (Forrest
et al 2018)
4 3 3 3
INM-CM4-8 No name (Volodin et al
2017)
1 1 1 1
INM-CM5-0 No name (Volodin et al
2017)
10 1 1 1
IPSL-CM6A-LR ORCHIDEE (Chen et al
2016)
30 1 9 6
MIROC-ES2L VISIT-e (Ito and Inatomi
2011)
3 1 1 1
MPI-ESM1-2HR JSBACH (Reick et al 2013) 10 2 2 2
MPI-ESM2-2-LR JSBACH (Reick et al 2013) 10 10 10 10
NorESM2-LM GFDL-LM3.0 (Gerber et al
2010)
3 1 3 1
UKEMS1-0-LL JULES (Harper et al 2016) 19 5 4 5
Thus, an indication on the changes induced by the
mean and the variability on the production resilience
may be obtained by comparing the projected relat-
ive changes of the mean and of the variability, using
the same weight. Equation (4) provides a normalized
indicator of such comparison:
(|∆µ/µ| − |∆σ/σ|)/(|∆µ/µ|+|∆σ/σ|),(5)
which varies from −1 (variability dominates) to +1
(mean dominates), which is useful to assess and
compare the dominant relative changes in different
locations.
In this study, RPand its changes are computed on
a large ensemble composed of all the climate change
simulations for vegetation GPP available from all the
Earth System Grid Federation (ESGF) portals up to
31 December 2019. The full list of simulations is
provided in table 1, along with the detailed reference
to the land surface component of the ESMs.
The ESMs land surface components include a
prognostic representation of the biosphere with spa-
tially distributed vegetation processes such as evapo-
transpiration, photosynthesis, carbon allocation and
growth of leaves, stems and roots interacting with
near surface meteorological variables such as temper-
ature, radiation, wind and CO2concentration, and
soil variables such as moisture, carbon and nitrogen
(see references in table 1). Therefore, GPP variability
in ESMs is the result of both bio-geophysical and bio-
geochemical processes such as soil moisture dynamics
and energy budget, permafrost thawing, atmospheric
CO2fertilization and nitrogen limitation as well as
land use changes defined as a function of the different
future scenarios. Some of the models also include a
simplified representation of abiotic stresses on veget-
ation, such as fires (Lawrence et al 2019, Bastos et al
2020).
Memory effects linked to antecedent drought
conditions are well reproduced since soil mois-
ture dynamics and the related physical feedbacks
were quite well developed already in the GCMs
(Seneviratne et al 2010), which are the predecessors
of the ESMs and from which ESMs inherits the rep-
resentation of abiotic processes. In general, ESMs
provide a reasonable representation of the GPP
response to drought (see citations in table 1), which is,
however, largely variable among models (Knauer et al
3

Environ. Res. Lett. 16 (2021) 105001 M Zampieri et al
Figure 1. Global changes of annual GPP resilience (Rp) computed from an ensemble of 16 ESMs simulations under ssp126
(panels (a) and (b)), ssp245 (panels (c) and (d)), and ssp585 (panels e and f) CMIP6 scenarios in the periods 2021–2050 (panels
(a), (c), and (e)) and 2051–2100 (panels (b), (d), and (f)) compared to the period 1985–2014.
2015, Huang et al 2016, Orth et al 2020). This motiv-
ates the use of a large ensemble for a robust assess-
ment of GPP changes such as the one used here.
The annual GPP is derived by summing up the
monthly GPP outputs for each simulation listed in
table 1consistently with the spatial variability of
vegetation seasonality (Peano et al 2019). The annual
GPP data of each simulation is interpolated on a
common 0.5 degrees regular grid with a second con-
servative remapping method (Jones 1999, Chen and
Knutson 2008). The simulations’ ensemble mean Rp
is computed for each ESMs and for each period and
scenario. Finally, the overall median of the RPchanges
with respect to the historical period is computed
for each future period and scenario. The robustness
of the results is assessed by highlighting the areas
where at least 75% of the models agree on the sign
of changes.
3. Results
Future climate projections display significant changes
of GPP variability resilience (figure 1) compared to
period 1985–2014. The annual vegetation primary
production resilience indicator is anticipated to gen-
erally increase in the lower emission scenarios (ssp126
and ssp245, figures 1(a)–(d), respectively). The lar-
ger positive changes are expected to occur especially
in the snow dominated bioclimatic regions (see table
S2 (available online at stacks.iop.org/ERL/16/105001/
mmedia)). The amplitude and the area covered by
these changes are comparatively larger in the ssp245
scenario than in the ssp126 scenario and increase with
time towards the end of the 21st century (table 2).
Positive changes are also estimated for Central Africa
and the Sahel regions, India and over the Himalayan
Plateau. However, regions with loss of GPP resilience
4

Environ. Res. Lett. 16 (2021) 105001 M Zampieri et al
Table 2. Fraction of global land area where the relative resilience indicator (∆Rp/Rp) exceeds different thresholds (5%; 10%; 15%; 20%)
in the simulation ensemble median. The first estimation (third and fourth columns) considers all areas displaying changes larger than
the thresholds. A second estimate (fifth and sixth columns) is restricted to the areas where at least the 75% of the models agree on the
sign of changes.
Fraction of land area
with changing resilience
Fraction of land area considering
75% models’ agreement
Period Scenario ∆Rp/Rp> 5% ∆Rp/Rp< 5% ∆Rp/Rp> 5% ∆Rp/Rp< 5%
2021–2050 ssp126
ssp245
ssp585
31%
35%
24%
2%
9%
21%
25%
27%
15%
1%
4%
12%
2051–2100 ssp126
ssp245
ssp585
41%
48%
14%
3%
8%
43%
36%
27%
8%
1%
4%
25%
∆Rp/Rp> 10% ∆Rp/Rp< 10% ∆Rp/Rp> 10% ∆Rp/Rp< 10%
2021–2050 ssp126
ssp245
ssp585
15%
21%
12%
0%
2%
8%
13%
19%
9%
0%
1%
6%
2051–2100 ssp126
ssp245
ssp585
27%
35%
8%
0%
2%
27%
26%
30%
6%
0%
1%
20%
∆Rp/Rp> 15% ∆Rp/Rp< 15% ∆Rp/Rp> 15% ∆Rp/Rp< 15%
2021–2050 ssp126
ssp245
ssp585
7%
12%
6%
0%
0%
3%
7%
11%
5%
0%
0%
2%
2051–2100 ssp126
ssp245
ssp585
20%
26%
5%
0%
0%
13%
19%
25%
4%
0%
0%
12%
∆Rp/Rp> 20% ∆Rp/Rp< 20% ∆Rp/Rp> 20% ∆Rp/Rp< 20%
2021–2050 ssp126
ssp245
ssp585
3%
6%
3%
0%
0%
0%
3%
6%
2%
0%
0%
0%
2051–2100 ssp126
ssp245
ssp585
14%
20%
3%
0%
0%
6%
14%
20%
2%
0%
0%
5%
appear as well, especially in Brazil, China and sur-
rounding countries of equatorial America. Under the
ssp245 socio-economic scenario, the CMIP6 ESMs
project resilience losses also in Mexico and the south-
ern part of the US, the Mediterranean region, South-
ern Africa and Australia. This occurs not only in the
mid-long term (2051–2100, figure 1(d)), but also in
the near term (2021–2050; figure 1(c)).
Under moderate emission scenarios (ssp126 and
ssp245), about one third of land area is going to exper-
ience an increase of vegetation annual GPP resilience
over the period 2021–2050 (see table 2). This propor-
tion is slightly lower, about one forth, when consider-
ing only the areas where 75% of the models agree on
the sign of changes. Differently, the area with positive
changes will cover almost half of the global land area
(less than one third when the constraint on models’
agreement is introduced) over the period 2051–2100.
Regions losing resilience cover a smaller percentage of
the global area, about 10% under ssp245 in the near
term. The differences between the plain estimate and
the one constrained on models’ agreement become
negligible for variations of resilience higher than 15%
(table 2).
The results for the ssp585 scenario stand out sig-
nificantly compared to the lower emission scenarios.
Broad areas with negative change (i.e. loss of veget-
ation GPP resilience) appear already in the period
2021–2050 (figure 1(e)) in the Amazon region, the
Unites States, South Canada, Western Europe, the
Mediterranean basin, as well as in the Middle East,
Central, Western and Southern Africa, Southeast Asia
and China, and Oceania. Areas with at least 5% loss
of vegetation GPP resilience are projected to cover
approximately one fifth of the global land (12% con-
sidering models’ agreement); while areas with more
than 15% losses are projected to be limited to 3%.
Positive gains of vegetation GPP resilience in boreal
regions are simulated to be more limited with respect
to the lower emission scenarios. Gains of at least 5%
are expected to cover about one fourth of the global
areas (15% considering models’ agreement); while
5

Environ. Res. Lett. 16 (2021) 105001 M Zampieri et al
Figure 2. Ensemble mean share of the two factors triggering changes in the vegetation annual production resilience indicator (as
from equation (5), see section 2). Positive values (light and dark blue areas) point to changes in the resilience indicator mainly
due to changes in the mean GPP. Negative values (red and yellow areas) are associated with grid points where the changes in the
resilience indicator are mainly driven by changes in the GPP variability.
areas with changes larger than 15% are limited to the
6%, similarly to the ssp126 scenario.
The severity of the projected losses is expected to
further exacerbate in the period 2051–2100. In the
Taking the Highway scenario, less and less regions are
expected to experience gains in vegetation GPP resi-
lience. These regions are: La Plata basin in Argen-
tina, part of the Sahel region, Eastern Africa, Western
India, North-western China and some regions along
the coast of the Arctic Sea. In general, areas gaining at
least 5% resilience are simulated to be limited to 14%
(8% considering model agreement) of global areas,
while regions with more than 15% increase of veget-
ation primary production resilience are limited glob-
ally to 6% of the land area. The areas losing resilience
are expected to outbalance those ones increasing resi-
lience and cover 43% (25% considering models agree-
ment) of global land area with more than 5% resili-
ence losses. Globally, 13% of land areas are predicted
to lose more than 15% vegetation primary production
resilience.
The GPP resilience changes can be determined
either by the change in the GPP mean and by the
changes in the GPP variability due to climate change
(see section 2). Positive resilience changes in the
near term under moderate emission scenarios are
often linked to positive changes in the mean annual
GPP (figures 2(a)–(e) and S4) connected to over-
all higher levels of atmospheric CO2concentration
and to higher mean growing temperature in Boreal
Regions. Negative resilience changes are generally
associated to an increase in the interannual variab-
ility of GPP (see figure S5). The areas affected by
an increase of variability largely change across the
scenarios and reach their maximum extent under the
scenario ssp585 (figures 2(e), (f), S5(e) and (f )).
Gain and losses of resilience are quantified at the
national level in order to provide country-specific
6

Environ. Res. Lett. 16 (2021) 105001 M Zampieri et al
Figure 3. Percentages of area with more than 15% annual GPP resilience change for the ten wider countries, Russia (RUS),
Canada (CAN), the United States of America (USA), China (CHN), Brazil (BRA), Australia (AUS), the European Union (EUR),
India (IND), Kazakhstan (KAZ), and Argentina (ARG). Negative values refer to the percentage of areas with negative GPP
resilience changes.
information for adaptation options, and possibly to
support ambitious mitigation policies. This analysis
is displayed in figure 3for the ten largest coun-
tries (and in table S3 for all World countries). Rus-
sia is characterized by the widest gains of resilience,
which could cover almost 70% of the country area
in period 2051–2100 under the ssp245 scenario. The
spatial extent of areas expected to experience gains
is reduced up to about 15% in the near term under
the ssp585 scenario. This tendency continues towards
the end of the century, under the ssp585 scenario,
when also areas with GPP resilience losses start to
appear. Canada shows a similar picture, but with less
optimistic estimation of predicted losses largely out-
balancing the gains in the 2051–2100 period under
the ssp585 scenario. The US and China display similar
figures, with gains predicted to reach 20% in the low
emission scenarios (ssp126 and ssp245) and losses
ranging from 10% to 15% in the ssp585 over period
2051–2100. Among the largest countries, Brazil is the
one characterized by the most alarming projections,
with the risk of losing resilience in 50% of its total
territory under the ssp585 scenario at the end of the
21st Century. It is worth noting that these changes are
likely to represent an underestimation as the current
trend of land-use change (Freitas et al 2018) is only
partially considered in the ESMs (Hurtt et al 2020).
Australia is estimated to undergo negligible losses .
Nevertheless, Australia will experience comparatively
large losses of resilience towards the end of the cen-
tury under the ssp585 scenario. The European Union
is characterized by a more stable situation with sig-
nificant positive changes only under the ssp245 scen-
ario over the period 2051–2100. India shows projec-
tions similar to the EU, but with significant areas of
vegetation that both gain and lose resilience under the
high emission scenario at the end of the 21st Cen-
tury. Large positive and negative changes in resili-
ence are also estimated for Argentina under the high
emission scenario (2051–2100). Whether or not these
compensating changes in different area are beneficial
for the countries’ adaptive capacity could be subject
of specific follow-up investigations.
Results for the remaining World countries (see
table S3) allow identifying severe cases, such as
losses of resilience higher than 15% over more than
50% of the area under the ssp585 scenario over
the period 2051–2100 in Gabon, Bhutan, Venezuela,
Equatorial Guinea, Malaysia, Peru, Guyana, Lebanon,
Japan, Congo, Bolivia, Honduras, Zambia, South
Korea, Papua Nuova Guinea and other 16 coun-
tries. Under the same scenario, the list of ‘win-
ner’ countries is much shorter, with only Somalia
gaining at least 15% resilience over more than
50% of its territory. Countries having the largest
increase of vegetation production resilience under
the ssp245 scenario in the mid-long term are Rus-
sia and the Northern European countries. Under the
ssp126 scenario, the vegetation production resilience
increases are geographically spread into more con-
tinents. In both ssp126 and ssp245 scenarios, there
are almost no countries experiencing more than 15%
7

Environ. Res. Lett. 16 (2021) 105001 M Zampieri et al
losses of resilience over more than 10% of their
lands.
4. Discussion and conclusions
This study implements a simple and robust statistical
indicator, the Annual Production Resilience Indicator
(Rp), able to provide a bulk estimation of ecological
resilience from annual production time-series. Being
focused on production, Rpis not sensitive to changes
in ecosystem structure and other state variables that
play a fundamental part in Holling’s definition of
ecological resilience (Holling 1973). Conversely, it is
formally consistent with more modern conception
of resilience allowing adaptation and other changes
in ecosystem structure in a resilience framework
(Walker et al 2004). Other indicators based on higher
temporal resolution data could be employed as well
in order to assess more detailed aspects of resilience
in a non-linear dynamics framework (Yi and Jackson
2021). However, Rpalready provides a quick and use-
ful indication of the changes in statistical stability of
the production time series in different periods, and it
is especially suitable for being computed on large data
volumes. Here, the vegetation GPP resilience changes
are estimated with reference to the climate variabil-
ity simulated by a large ensemble of state-of-the-art
ESMs.
While the main properties of Rpare already
demonstrated (Zampieri et al 2019,2020b, Zampieri
2021), as also summarised in the Appendix A
of the Supplementary Material (available online
at stacks.iop.org/ERL/16/105001/mmedia), here we
developed an interesting decomposition of the resili-
ence changes in terms of the mean GPP and its variab-
ility. Such decomposition provides a potentially use-
ful assessment for adaptation planning that should
presumably account differently for the changes in
the mean with respect to the changes in variability
of vegetation GPP in the future climate projections
(Mearns et al 1997, Smit et al 2000, van der Wiel
and Bintanja 2021). However, the implications of the
estimated GPP resilience changes need to be inter-
preted and confirmed through dedicated local-scale
assessments, also identifying the specific drivers of the
changes. For instance, while gains in resilience could
be generally considered a positive feature in dry envir-
onments, their consequences for wet environments
and over high latitudes should be further investig-
ated considering more complex approaches (Drews
and Greatbatch 2017, Laamrani et al 2020).
Our results aim at providing a useful first-
order indication that surely needs to be corrobor-
ated with further dedicated studies also account-
ing for different socio-economic perspectives. How-
ever, the analysis presented here is already poten-
tially very useful to identify the world regions
where there might be serious losses of vegetation
GPP resilience as well as the countries that are
subject to the most urgent necessity of improv-
ing adaptive capacity and resilience to climate-
related hazards under different future emission
scenarios.
Our results show large differences in the changes
of GPP resilience across the globe, depending on
greenhouse gases concentration of the projected scen-
arios. Under low emission scenarios, as found in pre-
vious studies (Hubau et al 2020), the CO2fertiliza-
tion effect often prevail over the negative effect due
global warming (e.g. an increase in metabolic costs
which in turn would lead to a reduction in forest pro-
ductivity and their efficiency, Collalti et al 2020) and
to the increase of climate variability. We find a gen-
eral increase of vegetation GPP resilience over boreals
regions that is in general agreement with the already
observed changes of vegetation distribution in those
areas (Myneni et al 1997). This tendency is expec-
ted to be reinforced in the future climate scenarios,
especially those with higher greenhouse gases emis-
sions (Zampieri and Lionello 2010). However, radi-
ation will always be a limiting factor for the vegeta-
tion development at very high latitudes (Seddon et al
2016).
The main findings point to areas in the mid-
latitudes where vegetation resilience is estimated to
decrease in the higher emission scenarios, such as
the Mediterranean, the mid-West in the US, Cent-
ral America, part of China, Southern Africa and Aus-
tralia. The stability of agricultural production and
the reliability of ecosystem services provided by the
natural vegetation might be compromised in these
regions, unless sensible adaptation actions are taken.
The importance of climate change mitigation is most
evident under the higher emission scenarios, where
vegetation resilience is affected in most land areas and
especially in tropical regions, where society is highly
dependent on ecosystem services and more vulner-
able to climatic changes.
The results of our analysis strongly support the
SDG-13 on taking action to combat climate change
and its impacts. Over areas with a high level of
anthropization, our results are relevant for agricul-
tural production, which is a main source of employ-
ment, livelihood and income for a large portion of
population especially in developing countries (SDG-
1, no poverty) as well as a main food source (SDG-2,
no hunger). Our results are also relevant for the SDG-
15, on the sustainable management of ecosystems and
halting land degradation and biodiversity loss.
Overall, in the scenario with lower climate change
mitigation (i.e. the Fossil-fuel Development scenario),
the areas losing vegetation resilience are larger than
the ones gaining resilience, jeopardizing the stability
of the ecosystems structure (and of the related ser-
vices). Adapting to changes in variability more than
to changes in the mean production of vegetation will
be critical for society and natural ecosystems in areas
experiencing large vegetation GPP resilience losses.
8

Environ. Res. Lett. 16 (2021) 105001 M Zampieri et al
Data availability statement
The data that support the findings of this study are
openly available at the following URL/DOI: https://
esgf.llnl.gov/.
Acknowledgments
For transparency and reproducibility, this paper uses
publicly available data provided by the IPCC through
the Earth System Grid Federation LLNL, CEDA,
DKRZ, GFDL, IPSL, LIU, NCI and NCCS nodes
(ESGF, https://esgf.llnl.gov/nodes.html) and open-
source software to compute the annual production
resilience indicator (Zampieri et al 2020a). This work
was supported by the FOCUS-AFRICA H2020 pro-
ject under Grant No 869575.
Author contributions
MZ conceived the study and drafted the paper, MZ
and BG performed the analysis, PDP downloaded
data and performed initial checking, all authors
reviewed the manuscript and contributed to the final
version.
Conflict of interest
Authors declare no conflicts of interests.
ORCID iDs
Matteo Zampieri https://orcid.org/0000-0002-
7558-1108
Bruna Grizzetti https://orcid.org/0000-0001-
5570-8581
Andrea Toreti https://orcid.org/0000-0002-1983-
2523
Alessio Collalti https://orcid.org/0000-0002-4980-
8487
References
Arora V K and Scinocca J F 2016 Constraining the strength of the
terrestrial CO2fertilization effect in the Canadian Earth
system model version 4.2 (CanESM4.2) Geosci. Model Dev.
92357–76
Bastos A et al 2020 Sources of uncertainty in regional and global
terrestrial CO2exchange estimates Glob. Biogeochem. Cycles
34 e2019GB006393
Bryant T, Waring K, S´
anchez Meador A and Bradford J B 2019 A
framework for quantifying resilience to forest disturbance
Front. For. Glob. Change 256
Chen C-T and Knutson T 2008 On the verification and
comparison of extreme rainfall indices from climate models
J. Clim. 21 1605–21
Chen Y et al 2016 Evaluating the performance of land surface
model ORCHIDEE-CAN v1.0 on water and energy flux
estimation with a single- and multi-layer energy budget
scheme Geosci. Model Dev. 92951–72
Collalti A et al 2020 Forest production efficiency increases with
growth temperature Nat. Commun. 11 5322
Costanza R, de Groot R, Sutton P, van der Ploeg S, Anderson S J,
Kubiszewski I, Farber S and Turner R K 2014 Changes in the
global value of ecosystem services Glob. Environ. Change
26 152–8
Dahan A 2010 Putting the earth system in a numerical box? The
evolution from climate modeling toward global change Stud.
Hist. Phil. Sci. B41 282–92
de Kauwe M G, Kala J, Lin Y-S, Pitman A J, Medlyn B E,
Duursma R A, Abramowitz G, Wang Y-P and Miralles D G
2015 A test of an optimal stomatal conductance scheme
within the CABLE land surface model Geosci. Model Dev.
8431–52
de Keersmaecker W, Lhermitte S, Honnay O, Farifteh J, Somers B
and Coppin P 2014 How to measure ecosystem stability? An
evaluation of the reliability of stability metrics based on
remote sensing time series across the major global
ecosystems Glob. Change Biol. 20 2149–61
Dosio A, Mentaschi L, Fischer E M and Wyser K 2018 Extreme
heat waves under 1.5 and 2 degree global
warming(supplementary data) Environ. Res. Lett. 13 054006
Drews A and Greatbatch R J 2017 Evolution of the Atlantic
multidecadal variability in a model with an improved North
Atlantic current J. Clim. 30 5491–512
Forrest M, Tost H, Lelieveld J and Hickler T 2018 Towards an
advanced atmospheric chemistry-enabled ESM with
dynamic land surface processes: part I—linking LPJ-GUESS
(v4.0) with EMAC modelling system (v2.53) Geosci. Model
Dev. Discuss. 2018 1–21
Freitas F L M, Sparovek G, Berndes G, Persson U M, Englund O,
Barretto A and Mörtberg U 2018 Potential increase of legal
deforestation in Brazilian Amazon after Forest Act revision
Nat. Sustain. 1665–70
Garrigues S et al 2015a Evaluation of land surface model
simulations of evapotranspiration over a 12-year crop
succession: impact of soil hydraulic and vegetation
properties Hydrol. Earth Syst. Sci. 19 3109–31
Garrigues S, Olioso A, Carrer D, Decharme B, Calvet J-C,
Martin E, Moulin S and Marloie O 2015b Impact of climate,
vegetation, soil and crop management variables on
multi-year ISBA-A-gs simulations of evapotranspiration
over a Mediterranean crop site Geosci. Model Dev. 83033–53
Gerber S, Hedin L O, Oppenheimer M, Pacala S W and
Shevliakova E 2010 Nitrogen cycling and feedbacks in a
global dynamic land model Glob. Biogeochem. Cycles
24 GB1001
Harper A B et al 2016 Improved representation of plant functional
types and physiology in the joint UK land environment
simulator (JULES v4.2) using plant trait information Geosci.
Model Dev. 92415–40
Holling C S 1973 Resilience and stability of ecological systems
Annu. Rev. Ecol. Syst. 41–23
Holling C S 1996 Engineering resilience versus ecological
resilience Natl Acad. Sci. 31–43
Huang Y, Gerber S, Huang T and Lichstein J W 2016 Evaluating
the drought response of CMIP5 models using global gross
primary productivity, leaf area, precipitation, and soil
moisture data Glob. Biogeochem. Cycles 30 1827–46
Hubau W et al 2020 Asynchronous carbon sink saturation in
African and Amazonian tropical forests Nature 579 80–87
Hurtt G C et al 2020 Harmonization of global land-use change
and management for the period 850–2100 (LUH2) for
CMIP6 Geosci. Model Dev. Discuss. 2020 1–65
Ito A and Inatomi M 2011 Water-use efficiency of the terrestrial
biosphere: a model analysis focusing on interactions
between the global carbon and water cycles J.
Hydrometeorol. 13 681–94
Jones P W 1999 First- and second-order conservative remapping
schemes for grids in spherical coordinates Mon. Weather
Rev. 127 2204–10
Knauer J, Werner C and Zaehle S 2015 Evaluating stomatal
models and their atmospheric drought response in a land
surface scheme: a multibiome analysis J. Geophys. Res.
Biogeosci. 120 1894–911
9

Environ. Res. Lett. 16 (2021) 105001 M Zampieri et al
Laamrani A, Valeria O, Chehbouni A and Bergeron Y 2020
Analysis of the effect of climate warming on paludification
processes: will soil conditions limit the adaptation of
northern boreal forests to climate change? A synthesis
Forests 11 1176
Lawrence D M et al 2019 The community land model version 5:
description of new features, benchmarking, and impact of
forcing uncertainty J. Adv. Model. Earth Syst. 11 4245–87
MAE 2005 Millennium Ecosystem Assessment. Ecosystems and
Human Well-Being: Synthesis (Washington, DC)
Maes J et al 2020 Mapping and Assessment of Ecosystems and
their Services: An EU ecosystem assessment EUR 30161 EN,
Publications Office of the European Union, Ispra, 2020
ISBN 978-92-76-178330 (https://doi.org/10.2760/757183)
Mearns L O, Rosenzweig C and Goldberg R 1997 Mean and
variance change in climate scenarios: methods, agricultural
applications, and measures of uncertainty Clim. Change
35 367–96
Meyer K 2016 A mathematical review of resilience in ecology Nat.
Resour. Model. 29 339–52
Moore J C, de Ruiter P C and Hunt H W 1993 Influence of
productivity on the stability of real and model ecosystems
Science 261 906–8
Morecroft M D, Crick H Q P, Duffield S J and Macgregor N A
2012 Resilience to climate change: translating principles into
practice J. Appl. Ecol. 49 547–51
Myneni R B, Keeling C D, Tucker C J, Asrar G and Nemani R R
1997 Increased plant growth in the northern high latitudes
from 1981 to 1991 Nature 386 698–702
Naumann G, Alfieri L, Wyser K, Mentaschi L, Betts R A,
Carrao H, Spinoni J, Vogt J and Feyen L 2018 Global
changes in drought conditions under different levels of
warming Geophys. Res. Lett. 45 3285–96
Orth R, Destouni G, Jung M and Reichstein M 2020 Large-scale
biospheric drought response intensifies linearly with
drought duration in arid regions Biogeosciences
17 2647–56
Peano D, Materia S, Collalti A, Alessandri A, Anav A, Bombelli A
and Gualdi S 2019 Global variability of simulated and
observed vegetation growing season J. Geophys. Res.
Biogeosci. 124 3569–87
Peters W et al 2018 Increased water-use efficiency and reduced
CO2uptake by plants during droughts at a continental-scale
Nat. Geosci. 11 744–8
Pimm S L 1984 The complexity and stability of ecosystems Nature
307 321–6
Randall D A et al 2019 100 years of earth system model
development Meteorol. Monogr. 59 12.1–12.66
Reick C H, Raddatz T, Brovkin V and Gayler V 2013
Representation of natural and anthropogenic land cover
change in MPI-ESM J. Adv. Model. Earth Syst.
5459–82
Sage R F and Kubian D S 2007 The temperature response of C3
and C4photosynthesis Plant Cell Environ. 30 1086–106
Santini M, Collalti A and Valentini R 2014 Climate change
impacts on vegetation and water cycle in the
Euro-Mediterranean region, studied by a likelihood
approach Reg. Environ. Change 14 1405–18
Scheffer M, Carpenter S R, Dakos V and van Nes E H 2015
Generic indicators of ecological resilience: inferring the
chance of a critical transition Annu. Rev. Ecol. Evol. Syst.
46 145–67
Seddon A W R, Macias-Fauria M, Long P R, Benz D and Willis K J
2016 Sensitivity of global terrestrial ecosystems to climate
variability Nature 531 229
Seneviratne S I, Corti T, Davin E L, Hirschi M, Jaeger E B,
Lehner I, Orlowsky B and Teuling A J 2010 Investigating soil
moisture–climate interactions in a changing climate: a
review Earth Sci. Rev. 99 125–61
Smit B, Burton I, Klein R J T and Wandel J 2000 An anatomy of
adaptation to climate change and variability Clim. Change
45 223–51
Stone L, Gabric A and Berman T 1996 Ecosystem resilience,
stability, and productivity: seeking a relationship Am. Nat.
148 892–903
United Nations 2016 Report of the Inter-Agency and Expert Group
on Sustainable Development Goal Indicators (New York)
pp 49
van der Wiel K and Bintanja R 2021 Contribution of climatic
changes in mean and variability to monthly temperature
and precipitation extremes Commun. Earth Environ. 21
Volodin E M, Mortikov E V, Kostrykin S V, Galin V Y,
Lykossov V N, Gritsun A S, Diansky N A, Gusev A V and
Iakovlev N G 2017 Simulation of the present-day climate
with the climate model INMCM5 Clim. Dyn. 49 3715–34
Walker B, Holling C S, Carpenter S R and Kinzig A 2004
Resilience, adaptability and transformability in
social-ecological systems Conserv. Ecol. 95
Yi C and Jackson N 2021 A review of measuring ecosystem
resilience to disturbance Environ. Res. Lett. 16 53008
Zampieri M 2021 Reconciling the ecological and engineering
definitions of resilience Ecosphere 12 e03375
Zampieri M, Grizzetti B, Meroni M, Scoccimarro E, Vrieling A,
Naumann G and Toreti A 2019 Annual green water
resources and vegetation resilience indicators: definitions,
mutual relationships, and future climate projections Remote
Sens. 11 2708
Zampieri M and Lionello P 2010 Simple statistical approach for
computing land cover types and potential natural vegetation
Clim. Res. 41 205–20
Zampieri M, Toreti A, Ceglar A, de Palma P and Chatzopoulos T
2020a Analysing the resilience of the European commodity
production system with PyResPro, the Python production
resilience package SoftwareX 15 100738
Zampieri M, Weissteiner C J, Grizzetti B, Toreti A, van den
Berg M and Dentener F 2020b Estimating resilience of crop
production systems: from theory to practice Sci. Total
Environ. 735 139378
10