Page 1
Emergence of robust precipitation changes across crop
production areas in the 21st century
Maisa Rojas
a,b,1
, Fabrice Lambert
b,c
, Julian Ramirez-Villegas
d,e,f
, and Andrew J. Challinor
d,e
a
Department of Geophysics, Universidad de Chile, 8370449, Santiago, Chile;
b
Center for Climate and Resilience Research, CR2, Universidad de Chile,
8370449, Santiago, Chile;
c
Department of Physical Geography, Pontificia Universidad Católica de Chile, 7820436, Santiago, Chile;
d
School of Earth and
Environment, University of Leeds, LS2 9JT Leeds, United Kingdom;
e
CGIAR Research Program on Climate Change, Agriculture and Food Security, CIAT,
763537, Cali, Colombia; and
f
International Center for Tropical Agriculture, CIAT, 763537, Cali, Colombia
Edited by Kerry A. Emanuel, Massachusetts Institute of Technology, Cambridge, MA, and approved January 30, 2019 (received for review July 3, 2018)
A warming climate will affect regional precipitation and hence
food supply. However, only a few regions around the world are
currently undergoing precipitation changes that can be attributed
to climate change. Knowing when such changes are projected to
emerge outside natural variability
—
the time of emergence (TOE)
—
is critical for taking effective adaptation measures. Using ensem-
ble climate projections, we determine the TOE of regional precipi-
tation changes globally and in particular for the growing areas of
four major crops. We find relatively early (
<
2040) emergence of
precipitation trends for all four crops. Reduced (increased) precipi-
tation trends encompass 1
–
14% (3
–
31%) of global production of
maize, wheat, rice, and soybean. Comparing results for RCP8.5
and RCP2.6 clearly shows that emissions compatible with the Paris
Agreement result in far less cropped land experiencing novel cli-
mates. However, the existence of a TOE, even under the lowest
emission scenario, and a small probability for early emergence
emphasize the urgent need for adaptation measures. We also show
how both the urgency of adaptation and the extent of mitigation
vary geographically.
climate change
|
natural variability
|
precipitation
|
agriculture
|
CMIP5
B
asic thermodynamics principles imply that global precipita-
tion will increase in a warmer world, but with significant re-
gional variations (1). The intensification of the hydrological cycle
will result in a precipitation minus evaporation change pattern that
has been termed the
“
rich-get-richer and poor-get-poorer
”
mechanism applicable over oceans (2). Due to large-scale at-
mospheric circulation and land/sea contrasts, regions that are
already wet, such as the tropics and high latitudes, will become
wetter, while dryer subtropical regions will become dryer (3).
Consistent precipitation changes are found in several regions
in the two most recent simulation datasets of the Coupled Model
Intercomparison Project (CMIP3 and CMIP5), which follow, to
a large degree, the rich-get-richer expected pattern over oceans
(4). Observational datasets indicate robust and statistically sig-
nificant positive precipitation tendencies in parts of the Northern
Hemisphere high latitudes (4), whereas mixed signals have been
found in the tropics (5, 6).
Although the spatial patterns of changes have been widely in-
vestigated, few studies have addressed the question of timing [i.e.,
when the precipitation changes during the 21th century will con-
sistently emerge outside the range of natural variability (7
–
9)].
Notably, no assessment of the timing of changes in precipitation
with a focus on agricultural activities has been carried out to date
(but refs. 10, 11). The time of emergence (TOE) is defined here as
the moment when the magnitude of the ensemble mean pre-
cipitation change becomes greater than the uncertainty due to
noise and natural variability (7) (
Methods
).
Although not the only relevant variable (12), precipitation
variability plays a significant role in the agriculture sector, par-
ticularly as rainfed agriculture constitutes 60
–
95% of farmed
land across the developing world (13). Crop-climate model
projections indicate that reductions in agricultural productivity
of 0.5%, on average, are likely per every percentage point re-
duction in precipitation, although with large spatial variability
(10, 14). In addition, global food supply is vulnerable because a
large fraction of the global food production is concentrated in
selected regions of the world (15, 16). Knowledge of the future
precipitation trends in these regions is therefore of the utmost
importance for the development and implementation of adequate
adaptation strategies that will ensure future global food demand is
met by production (17, 18). Here, we use a multimodel ensemble
from the 21 CMIP5 (19) climate models that simulated all four
representative concentration pathway (RCP) scenarios (20)
(
Methods
and
SI Appendix
, Table S1
) to identify key regions of the
world where a precipitation TOE exists during the 21st century.
We analyze both the annual mean and specific growing seasons
and areas of the four major crops (wheat, soybean, rice, and
maize), which, together, account for
∼
40% of total calorie intake
globally (18). The TOE assessment by growing seasons helps
identify regions in which 21st century crop production is expected
to be permanently exposed to climate change-induced precipita-
tion changes outside the range of natural variability, and may thus
benefit from timely adaptation measures.
Results
The calculation of the TOE for the annual mean precipitation
(Fig. 1) indicates that coherent patterns of increased (blue shad-
ing, 22
–
30% of global area) and decreased (red shading, 1
–
6% of
global area) precipitation trends emerge in all four RCPs in sev-
eral regions around the globe (
SI Appendix
, Table S2
). Spatially,
the TOE resembles the rich-get-richer pattern. The northern high
latitudes, including Canada, the eastern United States, northern
Europe, and Russia, exhibit a very early TOE of positive pre-
cipitation changes, some of them emerging as early as 2020 or
Significance
Using 21st century climate model projections we show that
for many crop-producing regions, average precipitation will
change by more than the long-term natural variability, even
under a low-emission scenario. However, emissions compatible
with the Paris Agreement can significantly reduce cropped land
affected. We identify the regions where adaptation measures
are most needed.
Author contributions: M.R. designed research; M.R. and F.L. performed research; J.R.-V.
contributed new reagents/analytic tools; M.R., F.L., and J.R.-V. analyzed data; and M.R.,
F.L., J.R.-V., and A.J.C. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Published under the
PNAS license
.
Data deposition: The data reported in this paper have been deposited in the Earth System
Grid Federation database,
https://esgf-node.llnl.gov/projects/esgf-llnl/
.
1
To whom correspondence should be addressed. Email: maisarojas@uchile.cl.
This article contains supporting information online at
www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1811463116/-/DCSupplemental
.
www.pnas.org/cgi/doi/10.1073/pnas.1811463116
PNAS Latest Articles
|
1 of 6
ENVIRONM
ENTAL
SCIENCES
SUSTAINAB
ILITY
SCIENC
E
Page 2
having already emerged (3). Southern high latitudes exhibit a
similarly early TOE mostly over oceans. These changes are mostly
explained by the thermodynamic increase in available humidity
(21) and poleward-shifted storm tracks (22). In the tropics, our
results indicate an increase of precipitation in the Intertropical
Convergence Zone (ITCZ) also relatively early (
<
2030). Changes
in the tropics are thought to be mainly driven by atmospheric cir-
culation changes (ascent regions) (2, 23). Another important region
with a positive precipitation change is India (
<
2030 for RCP8.5). As
the ITCZ induces no TOE over either South America or Africa, this
unique response of the Indian subcontinent may result from a
positive response of the summer monsoon to climate change (ref.
24. and references therein).
Negative precipitation trends emerging outside natural vari-
ability occur in the Mediterranean region in all four RCPs in the
early to midcentury, in western Mexico (only in RCP4.5 and
RCP8.5) after the midcentury, and in three subtropical regions in
the Southern Hemisphere around the midcentury in all four
RCPs (Fig. 1), with all of them being Mediterranean-like climate
regions. The drying trend in the subtropical dry regions is mainly
thought to be a result of reduced water vapor in regions of
subsidence (1). In both hemispheres, but especially in the three
Southern Hemisphere regions, the drying occurs predominantly
on the eastern side of the subtropical highs. This suggests a
stronger advection of cool air masses from polar areas that in-
creases the atmospheric stability in those regions and facilitates
subsidence, as a result of a poleward expansion of the Hadley cell
(22, 25), or strengthening of the circulation driven by thermody-
namics of a warmer surface (26). Note that regions where a drying
trend is expected all have significant natural variability in both
observations and CMIP5 models (
SI Appendix
, Figs. S1
–
S3
).
The northern high-latitude TOE pattern is mostly a result of
the winter precipitation increase (
SI Appendix
, Fig. S4
) and
corresponds to a 10
–
20% precipitation increase at the TOE with
respect to the 1986
–
2005 (historical baseline) level (
SI Appendix
,
Fig. S5
). Positive precipitation changes in the tropical regions
(ITCZ and India) at the TOE correspond to changes of 20
–
30%
and are due to both summer and winter contributions. The early
drying (
<
2040) in the Mediterranean region corresponds to a
10
–
30% reduction. Seasonally, the winter precipitation decrease
contributes to the signal in northern Africa and the Mediterra-
nean Sea, whereas the summer drying is more important in
southern Europe, including France (
SI Appendix
, Fig. S4
). In the
three Southern Hemisphere regions, drying at the TOE is 10
–
20% and the largest contribution to the drying is a decrease in
winter precipitation.
Discussion
Changes in total available precipitation are a major risk for the
global supply of food in the 21st century (10, 27). Recent climatic
changes and climate variability have already put a stress on
global food production (28). To estimate the potential implica-
tions of precipitation signal emergence for agriculture, we
computed the TOE globally for the growing seasons in the
production regions of the four major crops. Fig. 2 shows the
TOE by growing season for the RCP8.5 (other RCPs are shown
in
SI Appendix
, Figs. S6
–
S8
), indicating at what time wheat-,
soybean-, rice-, and maize-based farming systems can expect the
regional climate to permanently leave the previous mean climate
state and either settle around a new mean state or continue
changing for the respective growing seasons. The majority of
cropped areas for these crops do not show a TOE. Note that this
does not mean that these regions do not have a precipitation
trend, but just that the trend stays within the limits of natural
variability during the time range of our analysis. Nevertheless, a
globally significant percentage of current production will be af-
fected by a TOE of precipitation changes (
SI Appendix
, Table
S2
). By doing the calculations for the specific growing season of
[year]
2010
2020
2020
2030
2030
2040
2040
2050
2050
2060
2060
2070
2070
2080
2080
2090
2090
2100
2010
2020
2020
2030
2030
2040
2040
2050
2050
2060
2060
2070
2070
2080
2080
2090
2090
2100
120ºW 60ºW 0º 60ºE 120ºE
120ºW 60ºW 0º 60ºE 120ºE
120ºW 60ºW 0º 60ºE 120ºE
120ºW 60ºW 0º 60ºE 120ºE
A
B
C
D
Fig. 1.
TOE (year) of annual precipitation in RCP2.6 (
A
), RCP4.5 (
B
), RCP6.0 (
C
), and RCP8.5 (
D
). The red color scale is used for regions with a decrease in
precipitation. The blue color scale is used for regions with an increase in precipitation. Stippling indicates climatological dry areas (below 50 mm
·
y
−
1
).
2 of 6
|
www.pnas.org/cgi/doi/10.1073/pnas.1811463116
Rojas et al.
Page 3
the crops, a number of differences from annual mean calculation
are apparent. For example, an earlier decreasing precipitation
TOE emerges in Mexico and southern Africa for wheat than in
the annual mean calculation. In Europe, a decreasing precipitation
TOE will affect a larger area of maize production compared with
the annual mean calculation. Overall, for the RCP2.6 scenario
(compatible with the 2 °C warming goal of the Paris Agreement)
the affected areas by a TOE are greatly reduced.
Several agricultural regions in the Northern Hemisphere may
experience a precipitation increase that will be permanently
outside the current climate variability around midcentury (Fig.
2). Of particular importance are the present high-yielding re-
gions in China, northern India, and the eastern United States. A
key issue that emerges is thus the extent to which yield changes
are expected in these regions. As a first approximation, we use a
simple correlational analysis that shows, on average, agricultural
yield will increase with precipitation (
SI Appendix
, Fig. S9
). We
may therefore expect an average increase in agricultural yield
from regions with a positive TOE. While our trend analysis is not
predictive and lacks the physiological detail necessary to explain
the processes behind the yield
–
precipitation associations, it
should capture the first-order yield
–
precipitation interactions in
areas where robust precipitation changes are projected during
the 21st century (
SI Appendix
,
Supplementary Information Text
).
The generally positive yield outcome in regions with increasing
precipitation trends might be dampened by an important caveat:
The clay, clay loam, and loam soils in these regions make them
susceptible to floods with increased precipitation. By itself, in-
creased precipitation does not necessarily increase the flood risk,
since higher temperature will also enhance surface evapotrans-
piration. Therefore, in Fig. 3, we show the surface runoff changes
and temperature at the TOE for RCP8.5. Runoff can be regarded
as an indicator of topsoil saturation, and hence flooding.
Temperature changes at the TOE are in the range of 0.5
–
2 °C. The
enhanced runoff in China, India, and the eastern United States
suggests a higher frequency of flooding events in all three regions.
Due to the relatively early TOE in these regions, temperature
[year]
2010
2020
2020
2030
2030
2040
2040
2050
2050
2060
2060
2070
2070
2080
2080
2090
2090
2100
2010
2020
2020
2030
2030
2040
2040
2050
2050
2060
2060
2070
2070
2080
2080
2090
2090
2100
A
C
D
B
Fig. 2.
TOE (year) of precipitation changes for RCP8.5 by growing season of wheat (
A
), soybean (
B
), rice (
C
), and maize (
D
). The red color scale is used for
regions with a decrease in precipitation. The blue color scale is used for regions with an increase in precipitation. Gray indicates regions where a crop is grown,
and all grid cells with harvested area by crop are shown. Stippling indicates climatological dry areas (below 50 mm
·
y
−
1
).
B
A
Fig. 3.
Change at TOE compared with 1986
–
2005 for RCP8.5 of runoff (
A
)
and temperature (
B
).
Rojas et al.
PNAS Latest Articles
|
3 of 6
ENVIRONM
ENTAL
SCIENCES
SUSTAINAB
ILITY
SCIENC
E
Page 4
increases will be in the lower range, up to 1 °C. Potentially higher
yields due to enhanced precipitation may therefore be negated by
more flood events without investment in infrastructure and other
adaptation measures. Similarly, currently dry and low-yielding re-
gions in eastern Africa and southern India may experience wetter
conditions in the future (
SI Appendix
, Table S3
), which may also
result in higher flood risk due to the low soil moisture absorption in
these regions. Furthermore, over the past decades, wheat in the
United States has increased, but under a cooling trend in the past
decades (28), whereas for the TOE, important warming is expected
in this region (Fig. 3
B
). On the other hand, yield in India has al-
ready shown decreases with recent positive temperature and pre-
cipitation trends (28).
High-yielding agricultural production in the Northern Hemi-
sphere that may be impacted by reduced precipitation is concen-
trated in southwestern Turkey, Italy, southern France, the Iberian
Peninsula, Morocco, and central Mexico. Areas with a TOE of
decreasing precipitation in the Northern Hemisphere currently grow
75 million tons (11% of global production) of wheat and 53 million
tons (8% of global production) of maize. In those countries, about
62% of wheat production and 69% of maize production will po-
tentially be impacted in RCP8.5 (
SI Appendix
, Table S4
).
In the Southern Hemisphere, some regions in Ecuador, Uru-
guay, Argentina, and Papua-New Guinea have a relatively late
(
>
2060) TOE of increasing precipitation. The major feature in
the Southern Hemisphere is the drying of the subtropical land-
masses, particularly around the tip of South Africa, in central
Chile, and in southern and southwestern Australia. In these last
two regions, this represents approximately 50% and 30% of the
country
’
s wheat production, respectively. For southern Africa,
existing evidence suggests that these changes, coupled with
projected temperature changes, would reduce wheat and, to a
lesser extent, maize production significantly (29). In these areas,
42 million tons of wheat (6% of global production) are currently
produced, of which about 34% will be impacted in RCP8.5.
SI
Appendix
, Tables S3 and S4
give detailed values by country.
Fig. 4 shows the cumulative percentage of production affected
by a TOE of the four crops under RCP8.5. Overall, 20
–
30% of
production is affected by precipitation increase, starting early in
this century and quickly increasing after 2040. Precipitation de-
crease is found to affect mostly wheat- and maize-growing regions
from 2030 onward. Note that although most agricultural regions
do not experience a TOE, as identified by the CMIP5 simulations,
production is not evenly distributed among all regions, and these
regions might still be affected by varying precipitation trends
(10, 16).
The TOEs of precipitation changes (positive and negative)
found in this study allow identifying regions where precipitation
changes are projected to move outside known natural variability
within the 21st century, and hence can help to determine time
scales at which policies and actions to adapt to climate change
should be in place (e.g., refs. 11, 30). The spatial pattern of the
TOE is robust to the exact method used to calculate the emer-
gence (
Methods
), whereas the exact timing is more dependent on
the baseline used, including a small probability of an early
emergence of the signal (also ref. 9). In addition to precipitation,
adaptation to climate change must clearly account for impacts
from changes in temperature, evapotranspiration, and extreme
events (among others); hence, many more regions than those
found in this study are concerned. While only further analyses
using detailed crop modeling approaches will help in the devel-
opment of quantitative yield projections (which could be positive
or negative), our TOE analysis identifies the spatial and tem-
poral ranges where these studies may be most urgently needed.
In all likelihood, the positive and negative precipitation trends
identified in this study will require regional investment in ad-
aptation, although a more complete diagnosis of productivity and
cropped area change at the TOE than we have performed here is
warranted to guide such adaptation investments. Flood and
drought intensification may require governments to invest in
infrastructure resilience, and may require farmers to reconfigure
their cropping systems toward growing more drought-tolerant
crops (e.g., from maize to sorghum or millets). One result of
this analysis we would like to stress in particular is that a TOE
was found for most of the regions even in the low-emission
RCP2.6 scenario, highlighting the need for adaptation even un-
der stringent mitigation scenarios. However, in comparison with
higher emission scenarios, pursuing emission cuts compatible
with the Paris Agreement can significantly reduce the potential
impacts of climate change on crops.
Methods
Calculation of TOE.
A total of 21 CMIP5 models were used in this study, an-
alyzing simulations of the historical and all four RCP scenario experiments.
Most of the models comprise more than one and up to 25 ensemble members
in their experiments (
SI Appendix
, Table S1
). The TOE is the ratio between
the climate change precipitation signal and the estimated natural variability
and uncertainty and noise, calculated as described by Giorgi and Bi (7). This
method accounts for model spread, assuming that the models span the
possible range of climate responses. All models were linearly interpolated
onto a common 1°
×
1° grid. A conservative interpolation leads to qualita-
tively the same result (
SI Appendix
, Fig. S10
). At each grid point, the mean
20th century precipitation is calculated as the mean of the 1986
–
2005 pe-
riod. We consecutively calculate the model mean precipitation change using
a 20-y running window of the difference with respect to the base period.
The total uncertainty due to intermodel spread and internal multidecadal
variability is calculated by adding the variance of each model with respect to
the multimodel ensemble mean (intermodel component) to the variance of
each model with respect to its own ensemble mean (decadal component).
Giorgi and Bi (7) emphasize that in this procedure, the information is most
affected by the models with a larger number of realizations; however, as in
their work, the internal decadal variability in this study is much smaller than
the uncertainty due to intermodel spread. After having defined the mea-
sures of signal and noise, the running temporal average of the 20-y en-
semble mean change of precipitation and associated total variance of the
20-y changes are calculated. For each grid cell, we thus obtain a yearly time
series of mean 20-y changes and uncertainty, where, for each year of the
time series, the running average is taken over the previous 10 y and fol-
lowing 10 y. Once the time series of ensemble mean changes and corre-
sponding total standard deviation (STD) are calculated, the TOE is defined as
the time at which the magnitude of the mean change becomes greater than
that of the STD and remains so thereafter. To test the robustness of results,
we also calculated the TOE as the time when the precipitation in each in-
dividual model moves outside the range of its own variability, measured by
the STD of the baseline period. A model mean TOE is the calculated as the
mean of the individual TOE at grid points where at least 70% of models have
a TOE. The results are qualitatively the same, and
SI Appendix
, Fig. S11
[years]
2000
2020
2040
2060
2080
2100
% production
0
5
10
15
20
25
30
35
Positive precipitation changes
[years]
2000
2020
2040
2060
2080
2100
% production
0
2
4
6
8
10
12
14
Negative precipitation changes
wheat
rice
soybean
maize
B
A
Fig. 4.
Cumulative percentage of global production of four major crops
affected by positive (
A
) and negative (
B
) precipitation changes under RCP8.5.
4 of 6
|
www.pnas.org/cgi/doi/10.1073/pnas.1811463116
Rojas et al.
Page 5
shows the result for the RCP8.5. To evaluate the less probable outcomes,
SI
Appendix
, Fig. S11
also includes the 10th and 90th percentiles of the TOE
calculated by this second method. Note that this shows that there is a small
probability for a very early TOE over the same geographical areas that were
identified with the conventional definition of TOE. Furthermore, using
large-ensemble simulations, Zhang and Delworth (9) find that over the pe-
riod 2000
–
2009, more than a third of the globe already shows a distin-
guishable precipitation shift compared with the 1950
–
1999 period, further
corroborating our results.
Given that there are a number of time lengths important for agriculture,
the sensitivity of our results to the lengths of the baseline period used was
also explored. The 20-y baseline choice is justified from a signal-to-noise
perspective. On the other hand, in a 5-y period, a complete El Niño South-
ern Oscillation (ENSO), for example, or the transition from a positive to
negative (or vice versa) ENSO phase can be captured, whereas in a 10-y
period, two ENSO phases can be captured. However, a 20-y period is likely
to be a better representation of the long-term behavior of the production/
food system, which is our interest.
We have rerun the analysis for different time periods: 5-y and 10-y baseline
lengths. For example, a 5-y filter is used to reflect the importance of yield
fluctuations at short time scales (e.g., consecutive crop failures). The results show
a similar spatial pattern but, overall, a delay in the TOE by 5
–
10 y. This is due to
the larger interannual variability at shorter time scales (
SI Appendix
, Fig. S12
).
Model Validation.
The correct calculation of the TOE depends on the ability of
models to reproduce natural variability. To account for model biases, we com-
pare natural and simulated precipitation variability using the coefficient of
variation. Variability of natural and simulated precipitation was evaluated by
calculating the coefficient of variation over the period 1901
–
2005 in the
Climatic Research Unit (CRU) TS3.3 dataset (31) and full CMIP5 monthly data,
respectively (
SI Appendix
, Fig. S1
). Both datasets show good spatial agreement,
although the CMIP5 ensemble tends to underestimate the variance in some
areas, particularly in agricultural regions. In both the data and models, the
interannual variability represents 60
–
80% of the total variance (
SI Appendix
,
Fig. S2
). The slight underestimation of the total variance in the models seems
to stem mainly from the models
’
lower decadal variability compared with the
CRU data (
SI Appendix
, Fig. S3
). However, as intermodel spread is likely larger
than natural variability, our TOE calculation will represent an upper bound of
the true time at which precipitation changes may emerge above/below natural
variability of the historical climate. Finally, a low-pass filter was constructed to
separate the decadal component of this variance, and a high-pass filter to
separate the interannual component of the variance, for the annually aver-
aged CRU and CMIP5 data.
SI Appendix
, Fig. S13
shows the distribution of the
differences between the coefficients of variation of each model and CRU for
three latitudinal bands, indicating that the models have lower variance com-
pared with the observations, especially at middle to high latitudes.
Agriculture Calculations.
The global analysis of TOE for agriculture was per-
formed using the same procedures as for the annual mean, but for the
growing seasons and global harvested areas of four crops. Our analysis fo-
cused on wheat, soybean, rice, and maize, which, together, contribute to
ca.
40% of daily per capita caloric intake across the globe (32). Crop calendars
were gathered from the study of Elliott et al. (33), in which a global crop
calendar for each crop was derived through harmonization of existing
global cropping calendar datasets (34
–
36). Harvested areas and production
data were gathered from a study by Monfreda et al. (15). Crop calendar,
harvested area, and production data were aggregated to the analysis grid
(1°
×
1°) and then used to define where and when (i.e., the planting-to-
harvest periods) to compute mean precipitation and then conduct the TOE
analysis. Gridded production data were then used, together with the results
of the TOE analysis, to compute the total and proportional amounts of area
affected by detectable positive or negative precipitation trends. In Fig. 2, we
only plotted grid cells that had an area of at least 1% of the harvested crop.
For calculations of
SI Appendix
, Tables S3 and S4
, we used values of all grid
cells, without considering a threshold.
To understand possible yield implications of projected seasonal pre-
cipitation changes at the TOE, the following analysis was conducted. The
meta-analysis database from Challinor et al. (10) was used to identify the
extent of variation in the yield response, given precipitation change sce-
narios. We refrain from a more complex analysis since the primary aim of
this study is to detect robust precipitation changes in areas and times at
which crops are grown, rather than to estimate potential effects on crop
productivity, growing areas, or production; to assess the climate change
signal on crop productivity projections; or to understand with sufficient
detail the processes involved in potential yield changes. Future studies could
build upon our analysis to develop a more detailed understanding of yield
implications at the TOE.
1. Allan RP, et al. (2014) Physically consistent responses of the global atmospheric hy-
drological cycle in models and observations.
Surv Geophys
35:533
–
552.
2. Held IM, Soden BJ (2006) Robust responses of the hydrological cycle to global
warming.
J Clim
19:5686
–
5699.
3. Collins M, et al. (2013) Long-term climate change: Projections, commitments and
irreversibility.
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, United
Kingdom).
4. Hartmann DL, et al. (2013) Observations: Atmosphere and surface.
Climate Change
2013: The Physical Science Basis
, Contribution of Working Group I to the Fifth Assess-
ment Report of the Intergovernmental Panel on Climate Change, eds Stocker TF, et al.
(Cambridge Univ Press, Cambridge, United Kingdom).
5. Allan RP, Soden BJ, John VO, Ingram W, Good P (2010) Current changes in tropical
precipitation.
Environ Res Lett
5:25205
–
25208.
6. Byrne MP, Pendergrass AG, Rapp AD, Wodzicki KR (2018) Response of the in-
tertropical convergence zone to climate change: Location, width, and strength.
Curr
Clim Change Rep
4:355
–
370.
7. Giorgi F, Bi X (2009) Time of emergence (TOE) of GHG-forced precipitation change
hot-spots.
Geophys Res Lett
36:653
–
656.
8. Mahlstein I, Portmann RW, Daniel JS, Solomon S, Knutti R (2012) Perceptible changes
in regional precipitation in a future climate.
Geophys Res Lett
39:L05701.
9. Zhang H, Delworth TL (2018) Robustness of anthropogenically forced decadal pre-
cipitation changes projected for the 21st century.
Nat Commun
9:1150.
10. Challinor AJ, et al. (2014) A meta-analysis of crop yield under climate change and
adaptation.
Nat Clim Chang
4:287
–
291.
11. Rippke U, et al. (2016) Timescales of transformational climate change adaptation in
sub-Saharan African agriculture.
Nat Clim Chang
6:605
–
609.
12. 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.
13. Wani SP, Rockstrom J, Oweis T, eds (2009)
Rainfed Agriculture: Unlocking the
Potential
(CAB International, Cambridge, MA).
14. Ray DK, Gerber JS, MacDonald GK, West PC (2015) Climate variation explains a third
of global crop yield variability.
Nat Commun
6:5989.
15. Monfreda C, Ramankutty N, Foley JA (2008) Farming the planet: 2. Geographic dis-
tribution of crop areas, yields, physiological types, and net primary production in
the year 2000.
Global Biogeochem Cycles
22:GB1022.
16. Porter JR, et al. (2014) Food security and food production systems.
Climate Change
2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects
,
Contribution of Working Group II to the Fifth Assessment Report of the Intergov-
ernmental Panel on Climate Change, eds Field CB, et al. (Cambridge Univ Press,
Cambridge, United Kingdom), pp 485
–
533.
17. Springmann M, et al. (2016) Global and regional health effects of future food pro-
duction under climate change: A modelling study.
Lancet
387:1937
–
1946.
18. FAOSTAT (2013) Agriculture data. Available at http://www.fao.org/faostat/en/#data.
Accessed February 1, 2019.
19. Taylor KE, Stouffer RJ, Meehl GA (2012) An overview of CMIP5 and the experiment
design.
Bull Am Meteorol Soc
93:485
–
498.
20. van Vuuren DP, Edmonds JA, Kainuma M, Riahi K, Weyant J (2011) A special issue on
the RCPs.
Clim Change
109:1
–
4.
21. Seager R, Naik N, Vecchi GA (2010) Thermodynamic and dynamic mechanisms for
large-scale changes in the hydrological cycle in response to global warming*.
J Clim
23:4651
–
4668.
22. Wu Y, Ting M, Seager R, Huang H-P, Cane MA (2010) Changes in storm tracks and energy
transports in a warmer climate simulated by the GFDL CM2.1 model.
Clim Dyn
37:53
–
72.
23. Sobel AH, Camargo SJ (2011) Projected future seasonal changes in tropical summer
climate.
J Clim
24:473
–
487.
24. Chaturvedi RK, Joshi J, Jayaraman M, Bala G (2012) Multi-model climate change
projections for India under representative concentration pathways.
Curr Sci
103:1
–
12.
25. Scheff J, Frierson DMW (2012) Robust future precipitation declines in CMIP5 largely
reflect the poleward expansion of model subtropical dry zones.
Geophys Res Lett
39:
143
–
146.
26. Quan X-W, Diaz H, Hoerling MP (2004) Change of the tropical Hadley cell since 1950.
The Hadley Circulation: Present, Past, and Future
. Advances in Global Change Re-
search 21, eds Diaz HF, Bradley RS (Springer, Dordrecht, The Netherlands) pp 85
–
120.
27. Lobell DB, Tebaldi C (2014) Getting caught with our plants down: The risks of a global
crop yield slowdown from climate trends in the next two decades.
Environ Res Lett
9:
74003
–
74009.
28. Lobell DB, Schlenker W, Costa-Roberts J (2011) Climate trends and global crop pro-
duction since 1980.
Science
333:616
–
620.
29. Estes LD, et al. (2013) Projected climate impacts to South African maize and wheat
production in 2055: A comparison of empirical and mechanistic modeling approaches.
Glob Change Biol
19:3762
–
3774.
30. Challinor AJ, Koehler AK, Ramirez-Villegas J, Whitfield S, Das B (2016) Current
warming will reduce yields unless maize breeding and seed systems adapt immedi-
ately.
Nat Clim Chang
6:954
–
958.
Rojas et al.
PNAS Latest Articles
|
5 of 6
ENVIRONM
ENTAL
SCIENCES
SUSTAINAB
ILITY
SCIENC
E
Page 6
31. Harris I, Jones PD, Osborn TJ, Lister DH (2014) Updated high-resolution grids of
monthly climatic observations
–
The CRU TS3.10 dataset.
Int J Climatol
34:623
–
642.
32. Khoury CK, et al. (2014) Increasing homogeneity in global food supplies and the
implications for food security.
Proc Natl Acad Sci USA
111:4001
–
4006.
33. Elliott J, et al. (2015) The global gridded crop model intercomparison: Data and
modeling protocols for phase 1 (v1.0).
Geosci Model Dev
8:261
–
277.
34. Sacks WJ, Deryng D, Foley JA, Ramankutty N (2010) Crop planting dates: An analysis
of global patterns.
Glob Ecol Biogeogr
19:607
–
620.
35. Portmann FT, Siebert S, Döll P (2010) MIRCA2000
—
Global monthly irrigated and
rainfed crop areas around the year 2000: A new high-resolution data set for agri-
cultural and hydrological modeling.
Global Biogeochem Cycles
24:GB1011.
36. Waha K, van Bussel LGJ, Müller C, Bondeau A (2011) Climate-driven simulation of
global crop sowing dates.
Glob Ecol Biogeogr
21:247
–
259.
6 of 6
|
www.pnas.org/cgi/doi/10.1073/pnas.1811463116
Rojas et al.