Page 1
Originally published as:
Zhao, C., Liu, B., Piao, S., Wang, X., Lobell, D. B., Huang, Y., Huang, M., Yao, Y., Bassu,
S., Ciais, P., Durand, J.-L., Elliott, J., Ewert, F., Janssens, I. A., Li, T., Lin, E., Liu, Q.,
Martre, P., Müller, C., Peng, S., Peñuelas, J., Ruane, A. C., Wallach, D., Wang, T., Wu,
D., Liu, Z., Zhu, Y., Zhu, Z., Asseng, S. (2017):
Temperature increase reduces global
yields of major crops in four independent estimates. - Proceedings of the National
Academy of Sciences of the United States of America (PNAS), 114, 35, 9326-9331
DOI:
10.1073/pnas.1701762114
Page 2
Temperature increase reduces global yields of major
crops in four independent estimates
Chuang Zhao
a,1
, Bing Liu
b,c,d,e,f,1
, Shilong Piao
a,g,h,2
, Xuhui Wang
a
, David B. Lobell
i
, Yao Huang
j
, Mengtian Huang
a
,
Yitong Yao
a
, Simona Bassu
k
, Philippe Ciais
l
, Jean-Louis Durand
m
, Joshua Elliott
n,o
, Frank Ewert
p,q
, Ivan A. Janssens
r
,
Tao Li
s
, Erda Lin
t
, Qiang Liu
a
, Pierre Martre
u
, Christoph Müller
v
, Shushi Peng
a
, Josep Peñuelas
w,x
, Alex C. Ruane
y,o
,
Daniel Wallach
z
, Tao Wang
g,h
, Donghai Wu
a
, Zhuo Liu
a
, Yan Zhu
b,c,d,e
, Zaichun Zhu
a
, and Senthold Asseng
f,2
a
Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China;
b
National
Engineering and Technology Center for Information Agriculture, Nanjing Agricultural University, Nanjing, Jiangsu 210095, China;
c
Key Laboratory for Crop
System Analysis and Decision Making, Ministry of Agriculture, Nanjing Agricultural University, Nanjing, Jiangsu 210095;
d
Jiangsu Key Laboratory for
Information Agriculture, Nanjing Agricultural University, Nanjing, Jiangsu 210095;
e
Jiangsu Collaborative Innovation Center for Modern Crop Production,
Nanjing Agricultural University, Nanjing, Jiangsu 210095;
f
Agricultural and Biological Engineering Department, University of Florida, Gainesville, FL 32611;
g
Key Laboratory of Alpine Ecology and Biodiversity, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100085, China;
h
Center
for Excellence in Tibetan Earth Science, Chinese Academy of Sciences, Beijing 100085, China;
i
Department of Earth System Science Center on Food
Security and the Environment, Stanford University, Stanford, CA 94305;
j
State Key Laboratory of Vegetation and Environmental Change, Institute of
Botany, Chinese Academy of Sciences, Beijing 100093, China;
k
Desertification Research Centre, University of Sassari, 07100 Sassari, Italy;
l
Laboratoire
des Sciences du Climat et de l
’
Environnement, Le Commissariat à l
’
Énergie Atomique et aux Énergies Alternatives, CNRS, Université de Versailles Saint-
Quentin, Gif-sur-Yvette 91191, France;
m
Unité de Recherche Pluridisciplinaire Prairies et Plantes Fourragères, Institut National de la Recherche
Agronomique, CS 80006, 86600 Lusignan, France;
n
University of Chicago Computation Institute, University of Chicago, Chicago, IL 60637;
o
Columbia
University Center for Climate Systems Research, Columbia University, New York, NY 10025;
p
Institute of Crop Science and Resource Conservation, University of
Bonn, Bonn 53115, Germany;
q
Leibniz Centre for Agricultural Landscape Research, 15374 Müncheberg, Germany;
r
Department of Biology, University of
Antwerp, 2610 Wilrijk, Belgium;
s
International Rice Research Institute, Los Baños, 4031 Laguna, Philippines;
t
Agro-Environment and Sustainable Development
Institute, Chinese Academy of Agricultural Sciences, Beijing 100081, China;
u
UMR Laboratoire d
’
Ecophysiologie des Plantes sous Stress Environementaux,
Institut National de la Recherche Agronomique, Montpellier SupAgro, 34060 Montpellier, France;
v
Climate Impacts and Vulnerabilities, Potsdam
Institute for Climate Impact Research, 14473 Potsdam, Germany;
w
Centre de Recerca Ecològica i Aplicacions Forestals, Cerdanyola del Valles, Barcelona
08193, Catalonia, Spain;
x
Global Ecology Unit CREAF-CSIC-UAB, Consejo Superior de Investigaciones Científicas, Bellaterra, Barcelona 08193, Catalonia,
Spain;
y
National Aeronautics and Space Administration Goddard Institute for Space Studies, New York, NY 10025; and
z
UMR 1248 Agrosystèmes et
Développement Territorial, Institut National de la Recherche Agronomique, 31326 Castanet-Tolosan Cedex, France
Edited by B. L. Turner, Arizona State University, Tempe, AZ, and approved July 10, 2017 (received for review January 31, 2017)
Wheat, rice, maize, and soybean provide two-thirds of human caloric
intake. Assessing the impact of global temperature increase on
production of these crops is therefore critical to maintaining global
food supply, but different studies have yielded different results.
Here, we investigated the impacts of temperature on yields of the
four crops by compiling extensive published results from four
analytical methods: global grid-based and local point-based models,
statistical regressions, and field-warming experiments. Results from
the different methods consistently showed negative temperature
impacts on crop yield at the global scale, generally underpinned by
similar impacts at country and site scales. Without CO
2
fertilization,
effective adaptation, and genetic improvement, each degree-Celsius
increase in global mean temperature would, on average, reduce
global yields of wheat by 6.0%, rice by 3.2%, maize by 7.4%, and
soybean by 3.1%. Results are highly heterogeneous across crops
and geographical areas, with some positive impact estimates. Multi-
method analyses improved the confidence in assessments of future
climate impacts on global major crops and suggest crop- and region-
specific adaptation strategies to ensure food security for an increas-
ing world population.
climate change impact
|
global food security
|
major food crops
|
temperature increase
|
yield
C
rops are sensitive to climate change, including changes in tem-
perature and precipitation, and to rising atmospheric CO
2
concentration (1, 2). Among the changes, temperature increase has
the most likely negative impact on crop yields (3, 4), and regional
temperature changes can be projected from climate models with
more certainty than precipitation. Meteorological records show that
mean annual temperatures over areas where wheat, rice, maize, and
soybean are grown have increased by
∼
1 °C during the last century
(Fig. 1
A
) and are expected to continue to increase over the next
century (Fig. 1
B
)
—
more so if greenhouse gas emissions continue to
increase. It is thus necessary to quantify the impact of temperature
increase on global crop yields, including any spatial variations, to first
assess the risk to world food security, and then to develop targeted
adaptive strategies to feed a burgeoning world population (5).
Several methods have been developed to assess the impact of
temperature increase on crop yields (6). Process-based crop
models characterize crop growth and development in daily time
steps and can be used to simulate the temperature response of
yield either in areas around the globe defined by grids or at se-
lected field sites or points (1, 7). A third method, statistical
modeling, uses observed regional yields and historical weather
records to fit regression functions to predict crop responses
(8, 9). A fourth method is to artificially warm crops under near-
natural field conditions to directly measure the impact of increased
Significance
Agricultural production is vulnerable to climate change. Un-
derstanding climate change, especially the temperature impacts,
is critical if policymakers, agriculturalists, and crop breeders are
to ensure global food security. Our study, by compiling exten-
sive published results from four analytical methods, shows that
independent methods consistently estimated negative temper-
ature impacts on yields of four major crops at the global scale,
generally underpinned by similar impacts at country and site
scales. Multimethod analyses improved the confidence in as-
sessments of future climate impacts on global major crops, with
important implications for developing crop- and region-specific
adaptation strategies to ensure future food supply of an increas-
ing world population.
Author contributions: C.Z., B.L., S. Piao, D.B.L., F.E., and S.A. designed research; C.Z. and
B.L. performed research and analyzed data; C.Z., B.L., S. Piao, D.B.L., and S.A. wrote
the paper; and X.W., D.B.L., Y.H., M.H., Y.Y., S.B., P.C., J.-L.D., J.E., F.E., I.A.J., T.L., E.L.,
Q.L., P.M., C.M., S. Peng, J.P., A.C.R., D. Wallach, T.W., D. Wu, Z.L., Y.Z., and Z.Z.
contributed to the interpretation of the results and to the text.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
1
C.Z. and B.L. contributed equally to this work.
2
To whom correspondence may be addressed. Email: sasseng@ufl.edu or slpiao@pku.
edu.cn.
This article contains supporting information online at
www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1701762114/-/DCSupplemental
.
9326
–
9331
|
PNAS
|
August 29, 2017
|
vol. 114
|
no. 35
www.pnas.org/cgi/doi/10.1073/pnas.1701762114
Page 3
temperatures (4). Here, we combine these four methods, which
use disparate data sources, time spans, and upscaling approaches
(10), to assess the impact of increasing temperatures on yields of
wheat, rice, maize, and soybean. Grid- and point-based simula-
tions from recent international model intercomparison exercises
(2, 7, 11, 12) and published results of 13 statistical regression
studies and 54 field-warming experiments (
SI Appendix
, Fig. S1
)
are synthesized (
Materials and Methods
).
Results and Discussion
Fig. 2
A
illustrates the impact of temperature on yields of the four
crops at the global scale. The loss in yield for each degree Celsius
increase in global mean temperature is largest for maize (with
multimethod average
±
2 SE) of
−
7.4
±
4.5% per degree Celsius.
All four methods predict a negative impact for maize, but with
varying magnitudes. Mostly the different methods generated
similar results at the country scale (Fig. 3
C
and
SI Appendix
, Figs.
S2 and S3
), but estimates varied between countries. The impact
estimates are consistently negative for four major maize pro-
ducers, together responsible for two-thirds of global maize pro-
duction
—
namely, the United States (
−
10.3
±
5.4% per degree
Celsius), China (
−
8.0
±
6.1% per degree Celsius), Brazil (
−
5.5
±
4.5% per degree Celsius), and India (
−
5.2
±
4.5% per degree
Celsius). The estimated impact on maize crops in France, how-
ever, is smaller (
−
2.6
±
6.9% per degree Celsius), including a
small positive estimate (3.8
±
5.2% per degree Celsius) from
statistical modeling (13).
For wheat, the average estimate from all four methods is a
6.0
±
2.9% loss in global yield with each degree-Celsius in-
crease in temperature (Fig. 2
A
). Results from the four methods
agree more closely on the impact on wheat (
−
7.8 to
−
4.1% per
degree Celsius) than on maize yields (Fig. 2
A
). The results from
different methods are also generally consistent for the top
five wheat-producing countries (Fig. 3
A
) that harvest
>
50% of
the world
’
s wheat. Spatially, however, the impacts are highly
heterogeneous. Estimated wheat yield losses for the United States
(
−
5.5
±
4.4% per degree Celsius) and France (
−
6.0
±
4.2% per
degree Celsius) are similar to the global average, while those for
India (
−
9.1
±
5.4% per degree Celsius) and Russia (
−
7.8
±
6.3%
per degree Celsius) are more vulnerable to temperature increase.
The large yield reductions for Russia are mainly due to the con-
tribution of a markedly higher negative result from the statistical
method (
−
14.7
±
3.8% per degree Celsius; Fig. 3
A
), which did
not account for in-season variations in temperature impact (10).
By contrast, for China, the largest wheat producer in the world,
the multimethod estimate indicates that only 2.6
±
3.1% of yield
would be lost for each degree-Celsius increase in global mean
temperature.
Rice is a main source of calories in developing countries. The
analysis from the multimethod ensemble indicates that a global
increase in temperature of 1 °C will reduce global rice yield by
an average of 3.2
±
3.7%, much less than for maize and wheat
(Fig. 2
A
). Grid- and point-based simulations and field-warming
experiments indicate a negative impact of temperature of ap-
proximately
−
6.0% per degree Celsius, but some statistical
regressions suggest almost no impact. Similar disparities in
estimates between the statistical regressions and other methods
are found for several major rice-producing countries (Fig. 3
B
),
Fig. 1.
Mean annual temperature changes over time. (
A
) Historically ob-
served temperature anomalies relative to 1961
–
1990 for global growing
areas of four individual crops. (
B
) Future projected temperature changes
(2071
–
2100 in comparison with 1981
–
2010 baseline) of four crop-growing
areas and the globe (land and sea surface) under four representative con-
centration pathway (RCP) scenarios of increasing greenhouse gas concen-
trations. Error bars represent SDs in the climate model results.
Fig. 2.
Multimethod estimates of global crop yield changes in response to tem-
perature increase. (
A
) Impacts on crop yields of a 1 °C increase in global temper-
ature in grid-based simulations (Grid-Sim), point-based simulations (Point-Sim),
field-warming experiments (Point-Obs), and statistical regressions at the country
level (Regres_A) (9) and the global level (Regres_B) (8). Circles, means of estimates
from each method or medians for Grid- and Point-Sim. Filled bars, means of the
multimethod ensemble. Error bars show 95% CIs for individual methods (gray
lines) and the ensemble of methods (black lines). (
B
) Projected changes in yield due
to temperature changes by the end of the 21st century. CIs of 95% are given in
square brackets.
Zhao et al.
PNAS
|
August 29, 2017
|
vol. 114
|
no. 35
|
9327
AGRIC
ULTURAL
SCIENC
ES
SUSTAIN
ABILITY
SCIENCE
Page 4
including China, which produces
∼
30% of the world
’
s rice (14).
Similar regression methods produce quite different estimates for
Indonesia, Bangladesh, and Vietnam, which, when averaged
across all methods, lead to small estimated impacts on rice
production for each country. For India, however, estimates from
all methods predict large temperature impacts, with a multi-
method average of
−
6.6
±
3.8% per degree Celsius.
Soybean is the fourth most important commodity crop (14).
Results of just three studies using only two methods are available
for global-scale estimates of the impacts of temperature on soy-
bean yield. The global average reduction in soybean yield is 3.1%
per degree-Celsius rise (Fig. 2
A
), but the estimates are not statisti-
cally significant due to large uncertainties in each method (the 95%
CIs go through zero). Similar effects are estimated with both
methods for the United States, Brazil, Argentina, and Paraguay (Fig.
3
D
), which produce 84% of the global soybean harvest (14). The
largest expected reduction is
−
6.8
±
7.1% per degree Celsius for
the United States, the largest soybean producer. The overall results
for China, the fourth largest producer, however, do not indicate
statistically significant effects of temperature on soybean yield.
We compared different methods for a total of 10 sites and found
that method estimates are similar for most site
–
crop combinations
Fig. 3.
Multimethod estimates of grain yield
changes with a 1 °C increase in global temperature
for the five major countries producing each crop.
(
A
) Wheat. (
B
) Rice. (
C
) Maize. (
D
) Soybean. Grid-Sim,
Point-Sim, Point-Obs, and Regres_A are grid-based
simulations, point-based simulations, field-warming
experiments, and statistical regressions at the coun-
try level (Regres_A) (9), respectively. Regres_C is an-
other regression method used at the country scale
(13). Regres_D
–
K represents various country-level
regression analyses used for specific crops or coun-
tries shown by individual labels D
–
K above the bars.
Vertical axes show the temperature impact on crop
yield in percent per degree Celsius increase. Error
bars are 95% CIs. Values for error margins are not
available for point-based observations for maize
in China.
9328
|
www.pnas.org/cgi/doi/10.1073/pnas.1701762114
Zhao et al.
Page 5
(Fig. 4). Estimates from grid- and point-based simulations are
more similar to each other than to field-warming observations
(Fig. 4 and
SI Appendix
, Fig. S4
). This is not unexpected, as the
two types of simulation have some methodological similarities,
such as model structure, assumptions, and parameters. The grid-
and point-based models both tend to project greater yield loss with
increasing temperature at warmer locations and less yield loss at
cooler locations, a distinction not identified in the field experi-
ments (
SI Appendix
, Fig. S4
).
Some of the impact differences between simulations and field
experiments could be due to the fact that field experiments were
only carried out over a few years and might not represent the
entire variability of climate at this location, while the simulations
represent 30 y. Simulation parameters are also based on the
properties of cultivars that differ from those grown in field ex-
periments. For example, the field experiment in Wageningen (The
Netherlands) indicated a large negative impact of temperature rise
on wheat yield (
−
11.6% per degree Celsius), but used a spring
wheat that is not representative of the region (15). Positive im-
pacts (11.2
±
1.2% per degree Celsius) were observed in wheat-
warming experiments in Nanjing, China, where rising temper-
atures reduce damage from frost and heat stress during the
early and late experimental wheat growing seasons, respectively
(16)
—
factors that are captured less well in crop models (17). For
maize grown in Jinzhou (China), a field experiment and a regression
analysis produced very large negative estimates of impact, but were
not accompanied by margins of error to aid interpretation.
We assumed the temperature response of impact on yield
would be linear and multiplied projected temperature changes
(Fig. 1
B
) with our multimethod impact estimates to give an av-
erage projected decrease in the global crop yields of 5.6% (95%
CI, 0.1
–
14.4%) due to temperature change alone under the
scenario of lowest emissions (RCP2.6) going up to 18.2% (95%
CI, 0.7
–
38.6%) under the scenario of highest emissions (RCP8.5)
(Fig. 2
B
). The estimated responses in yield are primarily from
approximately
+
2 °C warming simulations, regressions, and ex-
periments (
Materials and Methods
), so the estimates of impact for
a global warming scenario near
+
4 °C (RCP8.5) are likely to be
conservative due to the nonlinear impact of rising temperatures
in the real world (4, 18). A nonlinear response to temperature
has also been suggested in simulations (1, 7, 10).
To prepare for adaptation to climate change, it is necessary to
isolate the effects of individual factor for possible impacts on yield,
as changes in different factors usually require different adaptation
strategies. While elevated atmospheric CO
2
concentration can
stimulate growth when nutrients are not limited, it will also in-
crease canopy temperature from more closed stomata (19). Also,
changes in precipitation can have an effect on crops, but projec-
tions on precipitation change are often uncertain. The focus of our
study is on temperature change, one of the most direct negative
impacts from climate change on crops, and does not include other
possible climate change effects from elevated atmospheric CO
2
concentration or changes in rainfall, and possible deliberate ad-
aptation taken by farmers. Farmers have increased yields through
adapting new technologies during the last half-century, but yield
has been also lost through increases in temperatures already (9).
Yield increase has slowed down or even stagnated during the last
years in some parts of the world (20, 21), and further increases in
temperature will continue to suppress yields, despite farmers
’
adaptation efforts.
The direct negative temperature impact on yield could be ad-
ditionally affected via indirect temperature impacts. For instance,
increasing temperature will increase atmospheric water demand,
which could lead to additional water stress from increased water
pressure deficits, subsequently reducing soil moisture and de-
creasing yield (22, 23). However, an accelerated phenology from
increased temperatures leads to a shorter growing period and less
days of crop water use within a cropping season. Such indirect
temperature effects are taken into account in each of the methods,
but are not explicitly quantified. Other indirect temperature im-
pacts include more frequent heat waves and possible temperature
impact on weeds, pests, and diseases (18, 24
–
26). Increases in
management intensity and yield potential could also uninten-
tionally increase yield sensitivity to weather (27).
By combining four different methods, our comprehensive as-
sessment of the impacts of increasing temperatures on major
global crops shows substantial risks for agricultural production,
already stagnating in some parts of the world (20, 21). However,
differences in temperature responses of crops around the world
suggest that some mitigation could be possible to substantially
affect the magnitude (or even direction) of climate change impacts
on agriculture. These impacts will also vary substantially for crops
and regions, and may interact with changes in precipitation and
atmospheric CO
2
, so a reinvigoration of national research and
extension programs is urgently needed to offset future impacts of
climate change, including temperature increase on agriculture by
using crop- and region-specific adaptation strategies.
Materials and Methods
Temperature Data.
Historical observed gridded monthly temperature data are
from the Climate Research Unit (0.5°
×
0.5° grid, CRU TS 3.23;
https://crudata.
uea.ac.uk/cru/data/temperature/
). Future predicted temperature data are from
the Coupled Model Intercomparison Project Phase 5 (CMIP5) Earth System
Models (ESMs) outputs (1.0°
×
1.0° grids;
cmip-pcmdi.llnl.gov/cmip5
) used in
the Intergovernmental Panel on Climate Change (IPCC) Assessment Report 5
(28). According to data availability, the outputs from 15, 20, 11, and 22 ESMs
were included in this study for RCP2.6, RCP4.5, RCP6.0, and RCP8.5 scenarios,
respectively. However, the calculated temperature changes are very similar to
those calculated by using all of the ESMs (IPCC 5). The annual mean temper-
ature over the global growing area of an individual crop was calculated by
weighting each grid cell average (0.5°
×
0.5° grids) according to the crop
growing area within the grid cell (29).
Global Gridded Crop Model Simulations.
The Agricultural Model Intercomparison
and Improvement Project (AgMIP) (30) and Inter-Sectoral Impact Model In-
tercomparison Project 1 (31) initiated a fast-track global climate impact as-
sessment for the main global crops in 2012, including wheat, rice, maize, and
soybean. Seven global gridded crop models were used to simulate crop yield in
0.5°
×
0.5° grid cells over the globe, forced with climate reconstruction for
Fig. 4.
Site-based multimethod ensemble of crop yield changes with 1 °C of
global temperature increase. Site estimates from more than three methods
are shown for wheat (
A
), rice (
B
), and maize (
C
) or from two methods for
soybean (
D
). Grid-Sim, Point-Sim, and Point-Obs are grid-based simulations,
point-based simulations, and field-warming experiments, respectively.
Regres_L
–
N are site-, county- or city-scale regression analyses for specific
crops shown by labels L
–
N next to the mean of the plotted dataset. Error
bars are 95% CIs. Error bars for the Jinzhou (China) results for regression L
and N were not available.
Zhao et al.
PNAS
|
August 29, 2017
|
vol. 114
|
no. 35
|
9329
AGRIC
ULTURAL
SCIENC
ES
SUSTAIN
ABILITY
SCIENCE
Page 6
1980
–
2099 based on HadGEM2-ES (32) derived from CMIP5. The simulations
were carried out under a scenario of constant CO
2
concentration (380 ppm in
2000) and full irrigation, to exclude the possibility of covariance with CO
2
and
precipitation. More detailed information about the simulations can be found in
refs. 1 and 33. Temperature impact values were calculated from yield changes
between 2029 and 2058 (
+
2 °C of global mean temperature) and 1981
–
2010
(baseline), which were then halved to give
+
1 °C of global temperature impact.
For global or country results, all of the grids were averaged by weighting the
corresponding growing area of each crop (29).
Point-Based Ensemble Simulations.
The AgMIP (30) also conducted crop yield
simulations at 30, 4, and 4 representative sites around the world (
SI Ap-
pendix
, Fig. S1
) by using 30 wheat, 13 rice, and 19 maize models, re-
spectively. For wheat, a scenario of
+
2 °C was created by adjusting each
day
’
s temperature by
+
2 °C relative to the baseline (1981
–
2010), with
other factors being constant. For rice and maize, the
+
3 °C scenarios were
used. Model details about simulations for each crop can be found in refs. 7,
11, and 12. The temperature impact was calculated as the yield change
during the warming period relative to the yield during the baseline period
normalized to
+
1 °C impact, assuming the impact showed a linear tem-
perature response. To obtain values for impacts at the country scale, each
country was deemed to be similar to one or more representative sites lo-
cated in said (or nearby) country. As local temperature change can be
different from the country mean, the local point-based estimates were
scaled up by multiplying each country
’
s temperature factor produced by
HagGEM2-ES (28), as in ref. 7. The weighted average temperature impacts
over all of the countries were used to estimate the global scale impact,
weighted by country-level production (14). It should be noted that the
results from only four sites were used to represent all of the rice-/maize-
producing countries, which might not encompass all of the uncertainties
from diverse production systems and is also one limitation in our analysis.
No point-based model-ensemble simulations for soybean were conducted
in AgMIP.
Field-Warming Experiments.
We started with all published peer-reviewed
studies that applied artificial warming treatments on field crops. To
avoid short-term noise, we only selected studies of crops that received all-
day warming treatments for
>
2 mo. Results from laboratory incubators or
controlled environments with constant day
–
night temperature treatment
(e.g., 37/29 °C vs. 29/21 °C) were excluded. The studies with temperature
change (
Δ
T
) unequal to
+
1 °C were adjusted to
+
1 °C impact by dividing
the impact value by
Δ
T
, which assumed a linear relationship between
impacts and
Δ
T.
The studies that produced temperature impacts of
>
50%
per °C were deemed as outliers and excluded. A total of 46 published
studies (available from the corresponding author upon request) and
48 sites (
SI Appendix
, Fig. S1
) were therefore included in the following
analysis. Most of the sites (41 of 48) had a warming magnitude of 1.5
–
3.0 °C, similar to the grid- and point-based simulations. The upscaling
methods from site to country to global scale are the same as for the point-
based model simulations.
Statistical Regressions.
Statistical models used regression equations to link
historical year-to-year variations in yield to variations in selected climate
variables. Different detrending methods were applied in the model to
remove the influence of adaptation measures, such as crop management. In
the statistical regression studies used here, the global-level results of re-
gressions A and B (Fig. 2
A
) used detrending methods with the inclusion of a
quadratic time trend and first differences, respectively, and resulted in more
similar temperature impacts than grid- or point-based simulations. A similar
result was found for the country-level regressions A and C (the country-level
results are in Fig. 3), which used detrending methods with inclusion of a
quadratic time trend and first-differences method, respectively. The results
from statistical models were from 13 published studies (available from the
corresponding author upon request). The interannual fluctuation in tem-
perature over the globe is
∼
2 °C (8), similar to the warming magnitude used
in other methods. To ensure comparability of results, reported values under
local temperature changes were normalized to global surface temperature
changes by multiplying the corresponding temperature factor produced by
HagGEM2-ES (28).
Multimethod Ensemble.
The above four methods constituted the method
ensemble that we used to estimate multimethod means and uncertainties. In
this study, values from the method ensembles were synthesized at site,
country, and global scales. At the country scale, the temperature impacts from
regression methods were only reported for the five countries producing
each crop; thus, the results mainly focus on the relevant top five countries.
The uncertainty for the method ensemble was calculated by using a formula:
var(
Y
)
=
var(
E
(
Y
j
method))
+
E
(var(
Y
j
method)), where the term var(
E
(
Y
j
method))
is a measure of the variability between methods, and
E
(var(
Y
j
method)) is a
measure of the average variability within methods, assuming that this is
random sample of approaches from a population of approaches. Confidence
intervals (CIs) at 95% were calculated for the multimethod mean as: 95%
CI
=
mean of methods
±
1.96
×
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
var
ð
Y
Þ
p
.
Comparisons Between Methods.
A recent study by Liu et al. (2016) (10) com-
pared the temperature impacts on wheat yield estimated by three different
methods. We extended the analyses by including a large number of datasets
from site-based observations (field-warming experiments) and comparing es-
timated impacts on yields of wheat, rice, maize, and soybean
—
the four most
important staple crops for humans. At the country scale, different methods
were compared across countries. For the regression method, the results were
only reported for the five major countries producing each crop, and thus the
comparisons only focused on the relevant five countries. At the site scale, grid-
based simulations were compared with site-based simulations and field-
warming experiments. Grids containing sites of point-based simulations or
warming experiments were selected. The comparisons include absolute yield
under different temperature scenarios and relative temperature impacts. The
baseline and temperature period for each grid was determined when the
rolling 30-y annual mean temperature was equal to the baseline and increased
temperatures used for point-based simulations and experiments. The tem-
perature impact was calculated as the yield changes relative to the baseline
and then adjusted to a
+
1 °C global temperature impact.
Prediction of Yield Changes by the End of Century.
The yield change by the end
of century was calculated as the products of the ensemble estimated yield
response and projections of global temperature rise from CMIP5. As the yield
response (Fig. 2
A
) and predicted temperature change (Fig. 1
B
) both have
uncertainties, a bootstrap resampling approach was used to obtain the
predicted yield change and its uncertainty. At each instance of bootstrap
resampling, one pair of values for yield response and temperature change
was sampled, respectively, from their original data to calculate the predicted
yield change; this procedure assumes the chosen value is a random sample
from a population of values. Repeating the above process 5,000 times gave
5,000 values of predicted yield change, which constitute a new distribution
of the predicted yield change. The 2.5th
–
97.5th percentile was deemed as
the boundaries of uncertainty for the predicted yield change.
ACKNOWLEDGMENTS.
This study was supported by National Natural
Science Foundation of China Grants 41530528 and 41561134016; 111
Project Grant B14001; and the National Youth Top-Notch Talent Support
Program in China. We thank the Agricultural Model Intercomparison and
Improvement Project (AgMIP). S.A. was supported by the Consultative
Group on International Agricultural Research (CGIAR) Research Program
on Climate Change, Agriculture and Food Security, and the CGIAR Re-
search Program on Wheat and the Wheat Initiative. B.L. and Y.Z. were
supported by National High-Tech Research and Development Program of
China Grant 2013AA102404, the 111 Project Grant B16026 and the Priority
Academic Program Development of Jiangsu Higher Education Institutions.
C.M. was supported by MACMIT (climate mitigation in agricultural pro-
duction systems through sustainable resource management) Project Grant
01LN1317A, funded through the German Federal Ministry of Education
and Research. P.C., I.A.J., and J.P. were supported by European Research
Council Synergy Grant ERC-SyG-2013-610028 IMBALANCE-P. P.C. was also
supported from the French Agence Nationale de la Recherche (ANR) Con-
vergence Lab Changement climatique et usage des terres (CLAND). F.E. was
supported by German Science Foundation Project EW 119/5-1 and by the
Joint Programming Initiative for Agriculture, Climate Change, and Food Se-
curity Modelling European Agriculture with Climate Change for Food Secu-
rity Project through German Federal Ministry of Food and Agriculture
Grant 2815ERA01J.
1. Rosenzweig C, et al. (2014) Assessing agricultural risks of climate change in the 21st
century in a global gridded crop model intercomparison.
Proc Natl Acad Sci USA
111:
3268
–
3273.
2. Wheeler T, von Braun J (2013) Climate change impacts on global food security.
Science
341:508
–
513.
3. Porter JR, Gawith M (1999) Temperatures and the growth and development of wheat:
A review.
Eur J Agron
10:23
–
36.
4. Ottman MJ, Kimball BA, White JW, Wall GW (2012) Wheat growth response to in-
creased temperature from varied planting dates and supplemental infrared heating.
Agron J
104:7
–
16.
9330
|
www.pnas.org/cgi/doi/10.1073/pnas.1701762114
Zhao et al.
Page 7
5. Nelson GC, et al. (2010)
Food Security, Farming, and Climate Change to 2050: Sce-
narios, Results, Policy Options
, IFPRI Research Monograph (International Food Policy
Research Institute, Washington, DC).
6. Zhao C, et al. (2016) Plausible rice yield losses under future climate warming.
Nat
Plants
3:16202.
7. Asseng S, et al. (2015) Rising temperatures reduce global wheat production.
Nat Clim
Chang
5:143
–
147.
8. Lobell DB, Field CB (2007) Global scale climate-crop yield relationships and the impacts
of recent warming.
Environ Res Lett
2:1
–
7.
9. Lobell DB, Schlenker W, Costa-Roberts J (2011) Climate trends and global crop pro-
duction since 1980.
Science
333:616
–
620.
10. Liu B, et al. (2016) Similar estimates of temperature impacts on global wheat yield by
three independent methods.
Nat Clim Chang
6:1130
–
1136.
11. Li T, et al. (2015) Uncertainties in predicting rice yield by current crop models under a
wide range of climatic conditions.
Glob Change Biol
21:1328
–
1341.
12. Bassu S, et al. (2014) How do various maize crop models vary in their responses to
climate change factors?
Glob Change Biol
20:2301
–
2320.
13. Lobell DB (2007) Changes in diurnal temperature range and national cereal yields.
Agric For Meteorol
145:229
–
238.
14. Food and Agriculture Organization of the United Nations (2014) FAOSTAT. Available
at www.fao.org/faostat/en/. Accessed August 16, 2016.
15. Van OM, Schapendonk AHCM, Jansen MJH, Pot CS, Maciorowski R (1999) Do open-
top chambers overestimate the effects of rising CO
2
on plants? An analysis using
spring wheat.
Glob Change Biol
5:411
–
421.
16. Tian Y, et al. (2012) Warming impacts on winter wheat phenophase and grain yield
under field conditions in Yangtze Delta Plain, China.
Field Crops Res
134:193
–
199.
17. Ewert F, et al. (2015) Crop modelling for integrated assessment of risk to food pro-
duction from climate change.
Environ Model Softw
72:287
–
303.
18. Schlenker W, Roberts MJ (2009) Nonlinear temperature effects indicate severe dam-
ages to U.S. crop yields under climate change.
Proc Natl Acad Sci USA
106:
15594
–
15598.
19. Long SP, Ainsworth EA, Leakey AD, Nösberger J, Ort DR (2006) Food for thought:
Lower-than-expected crop yield stimulation with rising CO
2
concentrations.
Science
312:1918
–
1921.
20. Brisson N, et al. (2010) Why are wheat yields stagnating in Europe? A comprehensive
data analysis for France.
Field Crops Res
119:201
–
212.
21. Ray DK, Ramankutty N, Mueller ND, West PC, Foley JA (2012) Recent patterns of crop
yield growth and stagnation.
Nat Commun
3:1293.
22. Asseng S, Foster IAN, Turner NC (2011) The impact of temperature variability on
wheat yield.
Glob Change Biol
17:997
–
1012.
23. Zhao C, et al. (2016) Field warming experiments shed light on the wheat yield re-
sponse to temperature in China.
Nat Commun
7:13530.
24. Porter JR, et al. (2014)
Climate Change 2014: Impacts, Adaptation and Vulnerability:
Contribution of Working Group II to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change
, eds Field CB, et al. (Cambridge Univ.
Press, Cambridge, UK).
25. Tack J, Barkley A, Nalley LL (2015) Effect of warming temperatures on US wheat
yields.
Proc Natl Acad Sci USA
112:6931
–
6936.
26. Lesk C, Rowhani P, Ramankutty N (2016) Influence of extreme weather disasters on
global crop production.
Nature
529:84
–
87.
27. Lobell DB, et al. (2014) Greater sensitivity to drought accompanies maize yield in-
crease in the U.S. Midwest.
Science
344:516
–
519.
28. Flato G, et al. (2013)
Climate Change 2013: The Physical Science Basis. Contribution of
Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on
Climate Change
, eds Stocker TF, et al. (Cambridge Univ Press, Cambridge, UK), pp
741
–
882.
29. Chad M, Navin R, Foley JA (2008) Farming the planet: 2. Geographic distribution of
crop areas, yields, physiological types, and net primary production in the year 2000.
Global Biogeochem Cycles
22:89
–
102.
30. Rosenzweig C, et al. (2013) The agricultural model intercomparison and improvement
project (AgMIP): Protocols and pilot studies.
Agric For Meteorol
170:166
–
182.
31. Warszawski L, et al. (2014) The Inter-Sectoral Impact Model Intercomparison Project
(ISI-MIP): Project framework.
Proc Natl Acad Sci USA
111:3228
–
3232.
32. Hempel S, Frieler K, Warszawski L, Schewe J, Piontek F (2013) A trend-preserving bias
correction
–
the ISI-MIP approach.
Earth Syst. Dynam.
4:219
–
236.
33. Elliott J, et al. (2014) The Global Gridded Crop Model intercomparison: Data
and modeling protocols for phase 1 (v1.0).
Geosci. Model Dev. Discuss.
7:
4383
–
4427.
Zhao et al.
PNAS
|
August 29, 2017
|
vol. 114
|
no. 35
|
9331
AGRIC
ULTURAL
SCIENC
ES
SUSTAIN
ABILITY
SCIENCE