Content uploaded by Muhammad Arshad
Author content
All content in this area was uploaded by Muhammad Arshad on Sep 14, 2019
Content may be subject to copyright.
ARTICLE
Climate variability and yield risk in South Asia’s rice–wheat
systems: emerging evidence from Pakistan
Muhammad Arshad
1
•T. S. Amjath-Babu
1
•Timothy J. Krupnik
2
•
Sreejith Aravindakshan
2,3
•Azhar Abbas
1,4
•Harald Ka
¨chele
1
•Klaus Mu
¨ller
1
Received: 28 September 2015 / Revised: 8 June 2016 / Accepted: 21 July 2016
ÓThe International Society of Paddy and Water Environment Engineering and Springer Japan 2016
Abstract Rice and wheat are the principal calorie sources for
over a billion people in South Asia, although each crop is
particularly sensitive to the climatic and agronomic man-
agement conditions under which they are grown. Season-long
heat stress can reduce photosynthesis and accelerate senes-
cence; if extreme heat stress is experienced during flowering,
both rice and wheat may also experience decreased pollen
viability and stigma deposition, leading to increased grain
sterility. Where farmers are unable to implement within-
season management adaptations, significant deviations from
expected climaticconditions wouldaffect crop growth, yield,
and therefore have important implications for food security.
The influence of climatic conditions on crop growth have
been widely studied in growth chamber, greenhouse, and
research station trials, althoughempirical evidence of the link
between climatic variability and yield risk in farmers’ fields is
comparatively scarce. Using data from 240 farm households,
this paper responds to this gap and isolates the effects of
agronomic management from climatic variability on rice and
wheat yield risks in eight of Pakistan’s twelve agroecological
zones. Using Just and Pope production functions, we tested
for the effects of crop management practices and climatic
conditions on yield and yield variability for each crop. Our
results highlight important risks to farmers’ ability to obtain
reliable yield levels for both crops. Despite variability in input
use and crop management, we found evidence for the nega-
tive effect of both season-long and terminal heat stress,
measured as the cumulative number of days during which
crop growth occurred above critical thresholds, though wheat
was considerably more sensitive than rice. Comparing vari-
ation in observed climatic parameters in the year of study to
medium-term patterns, rice, and wheat yields were both
negatively affected, indicative of production risk and of
farmers’ limited capacity for within-season adaptation. Our
findings suggest the importance of reviewing existing climate
change adaptation policies that aim to increase cereal farm-
ers’ resilience in Pakistan, and more broadly in South Asia.
Potential agronomic and extension strategies are proposed for
further investigation.
Keywords Climate change adaptation Adaptive
capacity Pakistan Heat stress Rice–wheat system
Yield risk
Introduction
South Asia’s rice (Oryza sativa)–wheat (Triticum aestivum)
rotational cropping systems occupy over 13 million hectares
of the Indo-Gangetic Plains that span from Pakistan to
Bangladesh. Together, these cereals provide for millions of
farmers’ livelihoods and feed over a billion people (Timsina
and Connor 2001). Rice is typically grown during the
summer monsoon season, while wheat is grown during the
dryer winter months following rice harvest. While South
Asia’s rice–wheat systems are relatively stable, productivity
&Muhammad Arshad
Muhammad.Arshad@zalf.de
1
Institute of Socio-Economics, Leibniz Centre for Agricultural
Landscape Research (ZALF), Eberswalder Str. 84,
15374 Mu
¨ncheberg, Germany
2
International Maize and Wheat Improvement Center
(CIMMYT), House 10/B. Road 53. Gulshan-2, Dhaka 1213,
Bangladesh
3
Farming Systems Ecology, Wageningen University,
Droevendaalsesteeg 1 - 6708 PB, Wageningen, The
Netherlands
4
University of Agriculture, Faisalabad 38040, Pakistan
123
Paddy Water Environ
DOI 10.1007/s10333-016-0544-0
losses from unexpected droughts, flooding, high tempera-
tures, or other unusual and extreme weather events can have
cascading negative effects, adversely impacting livelihoods
on the subcontinent (Lobell et al. 2012; Wassmann et al.
2009). Climatic variability and extremes therefore consti-
tute a major challenge to reliable and consistent crop pro-
ductivity (Gourdji et al. 2013; Knox et al. 2012; Lobell et al.
2008), and hence to farmers’ incomes and food security
(Brown and Funk 2008; Battisti and Naylor 2009). Ray et al.
(2015) showed that one-third of global crop yield variability
can be explained by climatic parameters, although more
research is still needed to understand the ramifications
of these effects and plan appropriate and actionable adap-
tation responses in South Asia.
Roughly thirty percent of the globe’s malnourished
people live in South Asia, where climate change is expected
to lower yields for maize (Zea mays), millet (Pennisetum
glaucum), rapeseed (Brassica napus L.), and wheat (Lobell
et al. 2008). Wheat in particular is sensitive to heat stress.
Exposure to temperatures above 30 °C can damage leaf
photosynthetic apparatus and accelerate senescence,
resulting in reduced grain filling (Asseng et al. 2011; Rey-
nolds et al. 1994; Lobell et al. 2012). At critically high
temperatures [32 °C prior to and during anthesis, photo-
synthetic decline is more pronounced. This can cause pollen
sterility, stigma desiccation, and reduced seed size and
setting (Farooq et al. 2011; Lobell et al. 2012; Reynolds
et al. 2016; Shah et al. 2011). Where farmers are unable to
establish wheat within recommended planting windows or
cannot access stress-tolerant or early-maturing cultivars,
considerable yield declines have been observed (Krupnik
et al. 2015a,b; Mondal et al. 2013), loosely agreeing with
Lobell et al.’s (2008) modeling estimates of a 3–17 % yield
loss for each 1°C of temperature increase in South Asia.
Rice is conversely less sensitive to high temperatures
during tillering and prior to microsporogenesis (cf. Shah
et al. 2011), with an optimal thermal range from 27–32 °C.
Heat stress encountered above this upper threshold
between 7–10 days before anthesis can however result in
pollen unviability, reduced pollen deposition, embryo
abortion and spikelet sterility, thereby lowering yield
(Reynolds et al. 2016; Krishnan et al. 2011; Shah et al.
2011). Rice may also be sensitive to high night-time tem-
peratures. Peng et al. (2004), for example, showed that
yield declines of *10 % can be expected with each 1°C
increase in minimum temperatures in the tropics, although
relative humidity plays a role in mediating these effects,
though with less influence in the subtropics (Krishnan et al.
2011). Under a changing climate, rice may also suffer from
increasing soil salinity, flooding, and heightened pest and
weed competition, though these stresses are geographically
heterogeneous and their effects will vary depending on
farmers’ crop management strategies (Kunimitsu et al.
2014,2015a; Sarker et al. 2012; Wassmann et al. 2009).
Breeding programs focused on wide adaptability and
stacking both biotic and abiotic stress-tolerant traits have
therefore been advised (Shah et al. 2011).
Resource-poor farmers in South Asia may conversely
have limited capacity to invest in technologies and man-
agement practices that could improve their adaptive capacity
(Jain et al. 2015). When climatic variability is combined with
other critical socioeconomic stresses faced by rice and wheat
farmers, for example inability to purchase and apply inputs at
recommended rates or within optimal timeframes, or where
crops are poorly managed due to labor deficits or competing
off-farm livelihood activities, farm households may find
themselves trapped in a cycle of low adaptive capacity and
hence climate vulnerability (Amjath-Babu et al.
2016; Morton 2007). Conversely, for South Asian farmers
who perceive of and act to counter intra-annual climatic
variability, adaptation has been observed to take the form of
varying crop species choice, sowing dates and management
practices, and/or the use of irrigation to hedge against
drought and heat stress (Jain et al. 2015). Not all farmers are
however equally able to respond to climatic variability,
especially if temperature or precipitation shifts are signifi-
cant. This underscores the uncertainty and risks associated
with smallholder crop production in tropical and subtropical
environments, a situation that requires mitigation if
equitable development is to be achieved (Morton 2007).
A number of studies have estimated the impacts of cli-
matic shocks and variation on the yield of major food crops
in the Indo-Gangetic plains, with considerable emphasis on
Bangladesh and India (cf. Lobell et al. 2012; Krupnik et al.
2015a,b; Mondal et al. 2013; Sarker et al. 2012). Studies in
Pakistan, where 70 % of the rural population derive a living
from agriculture, are comparatively rare. Studies that
account for the mediating effects that crop management
practices may have on yield risks also tend to be lacking. The
productivity of the rice–wheat cropping sequence is also
crucial, as each crop contributes 20 and 75 %, respectively,
to the average Pakistani’s caloric intake (Timsina and Con-
nor 2001). In the first decade of the millennium, annual rice
and wheat yields in Pakistan fluctuated by up to 1.31 and 0.57
tha
-1
at the national level (FAOSTAT 2016). Both rice and
wheat are also likely to require different climate variability
adaptation strategies, and like other staples, tailored
research, extension, and supporting services may conse-
quently be necessary (Morton 2007). Yet before developing
comprehensive adaptation plans, empirical evidence of the
impact of climatic variation on yield and yield risks under
farmers’ own crop management practices is required. This
paper responds to this information gap, by studying how
farmers’ crop management practices and climatic deviation
from medium-term trends affect yield and yield risks for 240
wheat and rice farming households in eight of Pakistan’s
Paddy Water Environ
123
twelve agroecological zones (AEZs). We conclude by sug-
gesting avenues for investment in improved crop research
and development programs to address these issues.
Materials and methods
Analytical background
Studies on the effects of climatic variability on yield and
yield risk frequently employ one of three methodologies.
These include agronomic production functions and crop
models (Rosenzweig et al. 2014), the latter increasingly tied
to the use of remote sensing data (e.g., Lobell et al. 2012),
Ricardian approaches (Mendelsohn et al. 1994), or panel
dataset analyses (Deschenes and Greenstone 2007). Each
approach is useful for answering particular research ques-
tions, though each also has weaknesses. The first approach,
while providing potentially precise information on crop
productivity, has been criticized because crop models and
static production functions inadequately consider farmers’
adaptation behavior in response to climatic variability
(Mendelsohn et al. 1994; Deschenes and Greenstone 2007).
Ricardian methods conversely aim to quantify the impact of
climate variability and additional socioeconomic and agro-
nomic variables on the value of agricultural land and/or crop
revenues using cross-sectional regression analyses. Such
approaches are based on the assumption that farmers are
sufficiently economically rational and are able to adapt to
their local environmental settings to produce efficiently. This
assumption could however be problematic where quick
changes in within-season crop management practices may be
required to cope with extreme heat or drought, for example.
Ricardian methods also do not explicitly account for or
model the outcomes of different forms of adaptation
behavior undertaken by individuals; rather, adaptation
activities are aggregated and implicitly assumed as captured
in the model itself.
Deschenes and Greenstone (2007) more recently pro-
posed use of panel datasets to measure the economic
impacts of climate variability on agriculture in the United
States. In this approach, climatic variables such as minimum
and maximum temperature, or precipitation and relative
humidity, are identified as year-to-year weather fluctuations
across geographical units considered in a study. Panel
datasets are now widely employed to assess the impact of
climate on agricultural productivity in developing countries
(cf. Seo et al. 2005; Poudel and Kotani 2013), although there
is a dearth of literature using this approach in Pakistan,
primarily due to a lack of reliable crop input–output and
management data across geographies in panel format. As in
other subtropical contexts, farmers in Pakistan grapple with
uncertainty and face stochastic threats to agricultural
productivity; such uncertainty should be considered in
production function data distributions, especially where the
goal of research is to advise on adaptation.
In order to incorporate estimation of the stochastic effect
of climatic variability on yield distributions, methods
described by Just and Pope (J–P) (1978) have been widely
applied (Isik and Devodas 2006; Holst et al. 2013; Poudel
and Kotani 2013). As demonstrated by recent studies (e.g.,
Guttormsen and Roll 2014; Koundouri and Nauges 2005),
one of the advantages of the generalized J–P model is its
flexibility with cross-sectional farm household input and
output data in the absence of time-series or panel infor-
mation. Under these circumstances, the J–P approach can
be used to assess the influence of climatic parameters on
both yield and yield risk (e.g., yield variance) simultane-
ously, while also accounting for crop management behav-
ior (Guttormsen and Roll 2014). We therefore employed
the J–P production function to examine the effect of cli-
mate variability on mean rice and wheat yields and yield
variance in Pakistan. We estimated the impact of climate
variability, measured as deviation in the current year of
study from the medium-term (32 year) historical average,
and the observed year’s weather parameters (specifically
cumulative rainfall and the number of days above which
species-specific thresholds for optimal growth and extreme
developmental stress were observed) on both the mean and
the variance (indicative of risk) of wheat and rice yields
reported by 240 farm households in eight AEZs in
Pakistan.
Data collection
Primary data
Primary data were collected employing a stratified random
sampling framework to achieve a representative sample of
farmers in each of the eight AEZs studied. Pakistan’s AEZ
classification system groups areas of similar climate and
physiography into units, providing a firm basis for stratifica-
tion (Arshad et al. 2015). Each AEZ was considered as one
stratum, with data collected from one district randomly
selected from within the eight AEZs (Fig. 1). In Pakistan, rice
is grown mainly in the monsoon ‘kharif’ season, while wheat
is grown in the drier winter ‘rabi’ season. Pakistan also con-
tains additional AEZs not considered in this study, although
those zones are primarily rugged mountain or barren and
marginal lands with negligible cereals production.
Two tehsils (administrative units) were selected from
each district chosen within the eight AEZs, and from each
tehsil, one village was randomly selected. Fifteen farm
households were subsequently randomly selected within
each village, creating a unique dataset with 30 rice–wheat
farmers from each AEZ, totaling 240. Sociopolitical
Paddy Water Environ
123
constraints including security risks, weak policing mea-
sures, and cultural constraints in rural areas limited further
data collection in several AEZs.
Each of the 240 farm households were surveyed by well-
trained professional enumerators specialized in agricultural
subjects. The same enumerators were employed for both
questionnaire pretests and actual data collection. Surveys
requested demographic, socioeconomic, and agronomic
crop management information (Table 1). Yields were
estimated by dividing farmers’ reports of harvested grain
weight after air-drying by rice or wheat cultivated area.
Security risks prevented enumerators from prolonged
interviews and stays in several of the studied AEZs. This
rendered collection of detailed panel impossible. Following
Kabubo-Mariara and Karanja (2007), our study therefore
makes use of cross-sectional field survey data and observed
weather parameters during the study seasons and years
(late 2011–late 2012, spanning rabi and kharif), and
medium-term historical climatic data (detailed below).
Secondary data
We utilized climate data from 32 years of weather obser-
vations (November 1980–October 2011), converting them
into monthly mean temperature and cumulative precipita-
tion for each crop’s growing season. Daily temperature and
precipitation observations for the 2011–2012 cropping
seasons during which farmers were surveyed were also
employed. Data were collected from the Pakistan Meteo-
rological Department, for all meteorological stations loca-
ted in each district and AEZ under study. Data were then
categorized according to the seasons during which wheat
and rice are produced, i.e., rabi and kharif, from mid-
November to early April and early July to mid-October,
respectively (Siddiqui et al. 2012).
Farmers’ reports of crop sowing, flowering date ranges,
and maturity were processed to calculate the cumulative
precipitation received during each cropping season and the
number of days that temperatures exceeded species-speci-
fic stress thresholds. For wheat, the cumulative number of
days during which temperatures exceeded 30 °C, indicative
of accelerated senescence (Asseng et al. 2011; Reynolds
et al. 1994; Lobell et al. 2012), were computed for the
entirety of the season. Following Lobell et al. (2012),
Gourdji et al. (2013) and Hatfield et al. (2011), we also
included a critically high threshold for extreme heat
[34 °C observed for the period during which farmers
reported flowering. 34 °C is also the threshold above which
the crop model Agricultural Production Systems sIMulator
(APSIM) is trained to speed senescence and increase grain
sterility during simulations (APSIM 2014).
Considering rice, Krishnan et al. (2011) adapted data
reported by Yoshida (1978) indicating that the upper
threshold for non-stressed growth from sowing to maturity
is 32 °C. The cumulative number of days during which
temperatures exceeded this mark were therefore tabulated
Fig. 1 Eight of Pakistan’s prominent agroecological zones for rice and wheat crop sequences considered in this study
Paddy Water Environ
123
for this species. To evaluate the effect of extreme heat on
rice during anthesis, we utilized a value of 35.5 °C, cal-
culated as the average from reports by Krishnan et al.
(2011), Hatfield et al. (2011), Gourdji et al. (2013) and
Shah et al. (2011).
Estimation procedure and model specification
To examine the effect of climatic variability on yield and
yield risks, we employed the J–P stochastic production
function (Just and Pope 1978). The J–P function accounts
for both stochastic effects and the effects of mea-
sured predictor variables on the probability distribution of
crop yields, and is therefore useful for the measurement of
yield risk. The first step in our analysis was to investigate
whether any significant yield risk (as measured by yield
variance) was observable. The Breusch–Pagan (1980) test
was therefore applied to the null hypotheses that our
models were homoscedastic such that no yield risk is
present. This was followed by estimation of two production
Table 1 Summary statistics of the variables used in all models
Variables Unit Wheat Rice
Mean Std. dev. Mean Std. dev.
Response variable
Yield kg ha
-1
3039.08 719.23 3831.46 960.13
Predictor variables
Area cultivated Hectare 2.93 2.72 3.28 2.24
Tillage Passes season
-1
4.60 0.95 5.23 1.13
Seed rate kg ha
-1
137.75 16.49 7.19 1.96
Farmyard manure kg ha
-1
1420 312.45 930 183.36
Nitrogen
a
kg ha
-1
92.52 27.39 50.40 13.51
Phosphorus
a
kg ha
-1
22.14 7.58 18.99 5.36
Potassium
a
kg ha
-1
18.98 6.93 14.35 4.48
Zinc
a
kg ha
-1
– – 2.83 1.13
Herbicide application Applications
season
-1
1.28 0.77 0.85 0.54
Insecticide application Applications
season
-1
– – 0.87 0.51
Irrigation application Applications
season
-1
4.91 4.03 9.82 5.36
Labor
b
Person-days ha
-1
34.42 7.26 39.78 6.07
Cumulative precipitation during the wheat cropping season mm 387.95 173.06 – –
Days with temperatures [30 °C during wheat growing
season (sowing to maturity)
n20.00 8.01 – –
Days with daytime temperatures [34 °C during flowering
stage in wheat
n0.65 0.48 – –
Deviation of the wheat season’s mean precipitation from the
historical mean
mm 16.28 8.51 – –
Deviation of the wheat season’s mean temperature from the
historical mean
°C 0.54 0.16 – –
Cumulative precipitation during the rice season mm – – 421.40 193.41
Days with temperatures [32 °C during rice growing season
(sowing to maturity)
n– – 88.62 11.88
Days with daytime temperatures [35.5 °C during flowering
stage in rice
n– – 14.12 7.53
Deviation of the rice season’s mean precipitation from the
historical mean
mm – – 26.19 14.43
Deviation of the rice season’s mean temperature from the
historical mean
°C – – 0.38 0.11
a
Applied as urea, diammonium phosphate, muriate of potash, and zinc sulfate
b
Entails an 8 h person day of labor
Paddy Water Environ
123
functions, including both mean and variance functions that
employed a maximum likelihood estimator to rice and
wheat separately. The linear form of J–P function was
estimated for both crops as
Yi¼fX
i;bðÞþei;ð1Þ
where i¼1;2;... nand
ei¼hZ;aðÞUið2Þ
with Uiexpressed as
UiN0;r2
v
:ð3Þ
In Eq. 1,f:ðÞis the deterministic component of the mean
yield function and Yiis the average yield for wheat or rice,
respectively, while Xiand Zare vectors of explanatory
variables, eiis a heteroscedastic noise term, and band aare
the corresponding parameters. In Eq. 2,hð:Þis the variance
function (yield risk component), which captures the effects of
the explanatory variables denoted by Zon the variance of the
outputs, while Uiis a random error term. Both equations
allow agronomic management and climatic variables to affect
both the mean yield levels of both crops ½EðYiÞ¼fX
i;bðÞ,as
well as their yield variance ½Var yðÞ¼r2
vhZ;aðÞ,onan
independent basis. All analyses utilized STATA V.13.0
(Stata Corporation, College Station, Texas).
We assumed that variation in weather conditions affects
crop development and yield formation directly, while the
deviation of the weather in the rice and wheat seasons of
2011–12 from the historical climatic mean may influence
farmers’ crop management decisions (e.g., planting dates,
input application timing, and frequency) as they adaptively
respond to deviation from their historical experience and
hence affect crop growth indirectly. To capture these direct
and indirect effects, we used two models. The first ‘‘re-
stricted model’’ includes the weather-related variables for
rice and wheat described in the ‘‘Secondary data’’ section to
capture the effects of the observed year’s crop management
parameters and weather on yield and yield variance. In the
second ‘‘full model,’’ deviation from historical averages
(1980–2011) for temperature and cumulative precipitation
were also added as separate variables for both crops. The
logic behind the restricted and full models, respectively, is
that yield levels are determined by the crop’s response to
management interventions and observed weather parame-
ters, with an emphasis on cumulative precipitation and heat
stress thresholds (restricted model), and also partly by
farmers’ crop management decisions based on their histor-
ical perceptions of past variability (full model). Evidence
from India indicates that farmers are often cognizant of
historical climate variability, and consequently adjust crop
management to adapt to perceived changes (cf. Jain et al.
2015). For example, farmers may vary the amount of
organic matter inputs, fertilizer and irrigation, shift planting
dates, or choose to use drought-tolerant and early-maturing
varieties, or even abandon particular crop species, to cope
with climatic variation or heat stress (Jain et al. 2015;
Krupnik et al. 2015a; Mondal et al. 2013).
Correct specification of the climate variables is very
important for the reliable estimation of yield functions,
with some suggesting that the effects of changes in climatic
variables on crop yield exhibit nonlinear trends (e.g.,
Schlenker and Roberts 2009). Holst et al. (2013) however
observed that when quadratic terms of the climate variables
are included in J–P models, none of the first-order terms
result in statistical significance. Specification of the pro-
duction function without inclusion of quadratic terms is
therefore superior when investigating the relationship
between climatic variables and crop yields; we conse-
quently used the linear form of the explanatory climate and
agronomic management variables in both models.
Results and discussion
Assessment of homoscedasticity
We performed Breusch–Pagan (1980) tests for the rice and
wheat functions separately. The tests revealed the presence
of heteroscedasticity (indicative of yield variation and
hence risk); the null hypotheses of homoscedasticity was
therefore rejected for both rice and wheat (Table 2). The
presence of heteroscedasticity (Chi
2
=21.41 Chi
2
=48.91
for wheat and rice models, respectively, both with
P\0.000) led us to proceed with estimation of the mean
production and variance functions (representative of yield
risk) by employing a maximum likelihood estimators for
each crop separately.
Wheat yields and risk in the 2011–12 rabi season
More than twenty days during the wheat cropping season in
2011–12 reached temperatures [30 °C. This had signifi-
cant (P\0.01) and negative influence (coefficient -0.35)
on yield measured across the eight AEZs considered in this
study (Table 3), providing backing for similar observations
in India’s nearby Punjab (cf. Lobell et al. 2012), and in
other South Asian locations (cf. Krupnik et al. 2015a,b;
Mondal et al. 2013). High temperatures encountered during
the rabi season can cause a reduction in crop duration,
which can in turn decrease the interception of solar radia-
tion, ultimately reducing growth and yield (Mondal et al.
2013). More pronounced, however, were the effects of
extreme heat measured as the cumulative number of days
that temperature exceeded 34 °C during flowering.
Paddy Water Environ
123
Although less than one day of extreme heat (mean level
over all AEZs) accumulated (Table 1), the negative influ-
ence on yield was significant (P\0.01) and strongly
negative (coefficient -2.36). Our threshold of 34 °Cat
flowering, as suggested by Lobell et al. (2012), Gourdji
et al. (2013) and Hatfield et al. (2011), is more conservative
than the critical stress limit of 32 °C proposed by Farooq
et al. (2011) and Reynolds et al. (2016). The strong nega-
tive influence of terminal heat stress on yield therefore
underscores the urgent need for adaptation measures that
permit wheat to escape from high temperatures that inter-
fere with pollination and grain formation. Both agronomic
and breeding solutions to this problem have been proposed,
including earlier sowing (Krupnik et al. 2015a,b), which
can be aided by specialized machinery and the use of zero-
and reduced-tillage methods that permit farmers to forgo
repetitive and time-consuming land preparation (Krupnik
et al. 2013; Aravindakshan et al. 2015). In Pakistan, these
methods have been observed to advance wheat sowing
by up to 2 weeks (Chaudhry 2000). Other potential solu-
tions include breeding for heat-tolerant and early-maturing
genotypes with improved pollen viability and accelerated
grain-filling rates (Mondal et al. 2013; Reynolds et al.
2016). Earlier establishment of the preceding rice crop can
also result in timely harvests and thus earlier wheat sowing
(Krupnik et al. 2015a), signifying that adaptation should be
tackled from a cropping systems rather than individual crop
perspective.
Considering the subsequent variance function, both
temperature threshold variables showed a positive and
significant correlation with yield variability, further
implying that heat stress increases wheat yield risks in
Pakistan (Table 3). Conversely, cumulative precipitation
coefficients showed no significant influence in either
function, likely because irrigation, which averaged nearly
five applications per season, offset most moisture stress
resulting in a significant (P\0.05) yet small (coefficient
0.001) influence on yield. Irrigation however also signifi-
cantly influenced yield variance (P\0.01), likely result-
ing from the wide range in irrigations (SD ±4.03) applied
by farmers.
Many of the input and agronomic management variables
used in our models, for example the quantity of nutrients
applied, exhibited positive and highly significant
(P\0.01) relationships with mean wheat yield. These
results are plausible and provide evidence for the yield
Table 2 Breusch–Pagan/
Cook–Weisberg test details Breusch–Pagan/Cook–Weisberg test Null hypothesis v
2
P[v
2
Wheat model Constant variance 21.41 0.0000
Rice model Constant variance 48.91 0.0000
Table 3 Estimates of the effects of the studied climatic parameters on wheat mean yield and variance (kg ha
-1
), the latter indicative of
production risk, using the Just–Pope (J–P) restricted model
J–P mean function J–P variance function
Coefficient Std. error Coefficient Std. error
Wheat area (ha) -0.0032 0.0486 0.0168 0.0503
Tillage passes season
-1
0.0009* 0.0004 0.001* 0.0006
Seed (kg ha
-1
) 0.0004 0.0009 0.0004 0.0011
Farmyard manure (kg ha
-1
) 0.0005 0.0004 0.0004 0.0005
DAP fertilizer (kg ha
-1
) 0.0009*** 0.0001 0.0008*** 0.0002
Potash fertilizer (kg ha
-1
) 0.0008*** 0.0002 0.0013*** 0.0004
Urea fertilizer (kg ha
-1
) 0.0022*** 0.0005 0.0022*** 0.0005
Herbicides (applications season
-1
) 0.0008 0.0007 0.0012 0.0008
Irrigation (applications season
-1
) 0.001** 0.0005 0.0016*** 0.0006
Labor (person-days ha
-1
) 0.0011** 0.0004 0.0006 0.0004
Total precipitation during wheat cropping season (mm) 0.0049 0.0033 -0.0023 0.0036
Days with temperatures [30 °C during wheat growing season (sowing to
maturity) (n)
-0.354*** 0.0548 0.352*** 0.066
Days with daytime temperatures [34 °C during flowering stage in wheat (n)-2.365*** 0.673 2.888*** 0.822
Constant 19.74*** 3.786 16.73*** 4.269
Total observations (n) 240 – 240 –
R
2
0.821 0.775
*p\0.1, ** p\0.05, *** p\0.01
Paddy Water Environ
123
enhancing effects of fertility management on yield,
although there may be scope for further improvement given
that farmers’ nutrient application rates were on average 34,
51, and 35 kg ha
-1
less than recommended, respectively,
for nitrogen, phosphorous, and potassium in irrigated wheat
(calculated from PARC 2015). Fertilizer application how-
ever also increased yield variance at the same significance
level. These results are consistent with a recent work in
Tanzania, in which Guttormsen and Roll (2014) used J–P
models and found that fertilizers and irrigation increased
both yield and yield variability, signifying that while
farmers who attempt to grow near-attainable yield levels
benefit from improved nutrient and water management,
they also must deal with increased risk, or take preventa-
tive and/or risk-reducing measures.
The number of tillage passes, which averaged 4.6, also
marginally influenced (P\0.1) yield and yield variance.
Land preparation for wheat is usually accomplished using
mouldboard ploughs and disk harrows, which can create
large soil clods, requiring repetitive planking and leveling
passes to prepare a fine seedbed, providing support for this
result. These observations however require further study and
disaggregation in other farming communities where zero-
and reduced-tillage techniques have been adopted and used
to advance sowing date, as discussed above. The number of
person-days ha
-1
that farmers dedicated to wheat produc-
tion also positively and significantly (P\0.05) influenced
yield, but not yield variance, indicating that increased crop
care logically has a beneficial effect on land productivity.
Influence of medium-term climate variability
on wheat yields and risk
Compared to the wheat cropping season’s mean tempera-
ture calculated from the medium-term (1980–2011) aver-
age, the positive yet slight (?0.54 °C) observed in
deviation in mean temperature during the 2011–12 rabi
season also negatively yet marginally (P\0.1) affected
mean wheat yield. This deviation also positively and sig-
nificantly (P\0.01) affected yield variability, indicating
increased production risk (Table 4). These results imply
that if and where farmers make management decisions
based on their historical perceptions of temperature
regimes, for example by moving sowing forward or
choosing varieties based on their anticipated duration, even
slight positive increases in temperature may introduce
conditions which may affect yield. Farmers—especially
those who are resoure poor—may however be unprepared
to rapidly adapt within the season following crop estab-
lishment. The number of cumulative days above critical
temperature thresholds also exhibited similar and expected
trends in the full model. Our observations of the negative
and relatively strong coefficients for temperature deviation
on wheat yield and variance are consistent with Asseng
et al. (2011), who indicated that global wheat production
may fall by up to 6 % for each 1 °C of temperature
increase. The effects of rising temperatures on yield vari-
ability have conversely been inadequately characterized
under on-farm conditions. Some studies for example sup-
port the hypothesis that variance will increase (e.g., Holst
et al. 2013), while others indicate the opposite (e.g., Isik
and Devadoss 2006), although the latter may have resulted
from the temperate zone in which the study was conducted.
Our results provide evidence for the former hypothesis, but
more research is needed in a variety of crops and agroe-
cologies to provide general support for this thesis.
Conversely, the deviation of rainfall from historical
means exhibited no significant effect on either mean wheat
yield or variability. Cumulative rainfall may have
nonetheless exhibited influence in the full model because
of improved model specification and lack of omitted vari-
able bias. Although both the restricted and full model show
the positive and significant influence of irrigation on both
yield and yield variability, the negative effect of cumula-
tive rainfall (P\0.01) on yield is likely to have resulted
from sudden and large late-season storms that brought up
to 61 mm of rain during harvests in April of 2012. These
conditions reduced national wheat yields by 4 % relative to
the previous year (PAR 2016). With the exception of til-
lage, which was not significant, other agronomic manage-
ment variables generally had a similar effect on wheat
grain yield as in the restricted model, but with slight
variation in significance. The lack of significance of tillage
in full model and negative effect of temperature deviation
suggests the potential for earlier sowing with reduced til-
lage as a potential adaptation strategy to overcome heat
stress (Krishna et al. 2012; Keil et al. 2015).
Rice yields and risk in the 2012 kharif season
Although the number of days during the kharif season with
temperatures [32 °C (indicative of season-long heat stress,
cf. Krishnan et al. 2011) was 88.6 (±11.9 SD), this variable
showed no significant effect on either rice yield or variance
(Table 5). The number of days during which extreme heat
was detected in excess of 35.5 °Cconverselysummedto
14.1 (±7.5 SD), resulting in a significant (P\0.01) and
negative influence on rice yield (coefficient -8.4). This
variable showed the same level of significance on yield
variance, though with positive influence, suggesting
increased yield risk as was observed with extreme heat. The
physiological basis for lower yields and higher variability is
similar in rice as in wheat, including pollen unviability,
reduced success of fertilization, and spikelet abortion
(Reynolds et al. 2016). The effect of high-temperature stress
may however also interact with increasing relative humidity
Paddy Water Environ
123
and solar radiation, though high humidity conditions also
generally lowers solar interception. Very high relative
humidity in excess of 88 % with concurrent tempera-
tures [35.0 °C can increase the percentage of spikelets
lacking sufficient pollen production and stigma interception,
thereby lowering yield, though this effect varies among
cultivars (Krishnan et al. 2011). Our data indicated a sig-
nificant (P\0.05) and positive effect of cumulative rainfall
on rice yields. Precipitation is generally correlated with
relative humidity, although we were unable to explicitly
account for this variable due to a lack of consistent data
availability. Further research may therefore benefit from
examining how relative humidity and extreme heat may
interact.
For years, rice breeders have selected for cultivars with
erect flag leaves to increase the interception of late-season
solar radiation, as this can benefit yields (Chang and
Tagumpay 1970). This plant architecture can however also
increase the potentially damaging effect of high relative
humidity and heat stress by increasing canopy and repro-
ductive organ temperatures relative to ambient air condi-
tions (Krishnan et al. 2011). Traits to increase heat
tolerance and minimize these trade-offs are reviewed by
Reynolds et al. (2016), Krishnan et al. (2011) and Shah
et al. (2011). Agronomic management can also moderate
heat stress. Options include irrigation, use of site- and
climate-adapted cultivars, zero-tillage and direct seeding to
advance establishment dates. Plant growth regulators have
also been proposed to increase pollen production and to
advance flowering to the early morning rather than mid-day
when the sun is at its zenith and heat is most extreme (Shah
et al. 2011).
Most other explanatory variables had the expected effect
on rice yield and variance. Use of phosphorous and zinc,
which averaged 19.0 and 2.8 kg ha
-1
, respectively, both
had significant and positive effects on mean rice yield
(Table 5), indicative of soil deficiencies of each nutrient.
Conversely, even after accounting for nitrogen applied in
DAP, neither nitrogen nor potassium conversely affected
mean yields, likely because of very low application rates
*86.1 and *42.3 kg ha
-1
below recommendations,
respectively (calculated from PARC 2015). Even while
showing significant yield effects, phosphorous and zinc
application could also likely be increased by at least 33 %,
although it should be noted that these variables also sig-
nificantly increased yield variance. Both insecticide appli-
cation and irrigation had similar effects. Repetitive dry
followed by wet tillage, which is widely practiced for rice
in Pakistan, had a positive and significant (P\0.05) effect
on both mean yield and variance. Repetitive wet tillage in
the form of puddling can reduce percolation and floodwa-
ter loss in rice paddies, thereby helping to reduce the
potential for drought stress, though this practice also typ-
ically increases methane production (Shah et al. 2011),
Table 4 Estimates of the impacts of the climatic variability expressed as the deviation from the 32 year average on mean yield and variance
(kg ha
-1
) of wheat, the latter indicative of production risk, using the Just–Pope (J–P) full model
J–P mean function J–P variance function
Coefficient Std. error Coefficient Std. error
Wheat area (ha) 0.0092 0.0451 0.0171 0.0478
Tillage passes season
-1
0.0004 0.0004 0.0004 0.0006
Seed (kg ha
-1
) 0.0009 0.0008 0.0011 0.001
Farmyard manure (kg ha
-1
) 0.0006* 0.0003 0.0006 0.0004
DAP fertilizer (kg ha
-1
) 0.0004** 0.000 0.0002 0.0002
Potash fertilizer (kg ha
-1
) 0.0005* 0.0002 0.0009** 0.0004
Urea fertilizer (kg ha
-1
) 0.0022*** 0.0004 0.0021*** 0.0005
Herbicides (applications season
-1
) 0.0008 0.0006 0.0012 0.0007
Irrigation (applications season
-1
) 0.0011** 0.0004 0.0017*** 0.0005
Labor (person-days ha
-1
) 0.0011*** 0.0004 0.0008* 0.0004
Total precipitation during wheat cropping season (mm) -0.0104*** 0.0037 0.0097** 0.0044
Days with temperatures [30 °C during wheat growing season (sowing to maturity) (n)-0.0404** 0.0162 0.0487** 0.0212
Days with daytime temperatures [34 °C during flowering stage in wheat (n)-0.364*** 0.0523 0.395*** 0.0640
Deviation of the wheat season’s mean precipitation from the historical mean (mm) -0.146 0.885 0.109 1.138
Deviation of the wheat season’s mean temperature from the historical mean (°C) -3.439* 1.758 0.151*** 0.0331
Constant 29.25*** 4.032 28.18*** 5.118
Total observations (n) 240 – 240 –
R
2
0.854 0.807
*p\0.1, ** p\0.05, *** p\0.01
Paddy Water Environ
123
thereby creating negative feedback in the form of increased
global warming potential.
Influence of medium-term climate variability on rice
yields and risk
Comparing the medium-term climatic mean with the
observed 2012 kharif season, mean precipitation positively
(?26.2 mm) deviated from historical observations, result-
ing in a significant (P\0.01) and positive effect on the
yield, but not variance (Table 6). The results for cumula-
tive precipitation followed a similar trend, also mirroring
the restricted model. Deviation of the kharif mean tem-
perature from the medium-term historical mean was lim-
ited (?0.4 °C). This variable nonetheless significantly
affected both mean yields and variance (P\0.01 for
each), respectively, the former negatively and the latter
positively, indicative of decreased yield with increased
yield risk. As with the restricted model, extreme heat over
35.5 °C significantly (P\0.01) and negatively influenced
yield, but less so for yield variance (P\0.1). These results
suggest that the sampled rice farmers have not been able to
adequately adapt to climate variation with practices that
stabilize yield. Conversely, as with the restricted model,
many of the input management variables showed signifi-
cant and positive effects on mean rice yield, though often
with associated increasing risk in the form of heightened
variance.
Methodological challenges and limitations
Given the absence of panel data, we could not include AEZ
dummy variables in the models due to colinearity with
climate variables within our cross-sectional dataset. Neither
relative humidity nor night-time temperatures were inclu-
ded for the same reason. While we relied on weather sta-
tion data, more precise estimates of canopy temperatures
that affect plant growth may improve analyses; future
research could therefore confirm or reject our results by
making use of remotely sensed thermal information for
specific fields. In addition, soil type, slope, measurements
of pest infestation, and market information could also be
used to fine-tune models, though such sampling was not
possible in our case due to security risks in many of the
AEZs studied. Our data also showed little diversity in
farmers’ varietal choice, although stress-tolerant varieties
are increasingly being released in South Asia. Where
farmers adopt such cultivars, future research should
investigate if heat- or drought-tolerant traits confer yield
enhancing and/or variance reducing effects under on-farm
conditions in the presence of farmers’ variable crop man-
agement strategies.
Table 5 Estimates of the effects of the studied climatic parameters on rice mean yield and variance (kg ha
-1
), the latter indicative of production
risk, using the Just–Pope (J–P) restricted model
J–P mean function J–P variance function
Coefficient Std. error Coefficient Std. error
Rice area (ha) -0.0825 0.0719 0.389 0.308
Tillage passes season
-1
0.0014** 0.0007 0.0031** 0.0013
Seed (kg ha
-1
) 0.0012 0.0045 0.0039 0.0061
Farmyard manure (kg ha
-1
) 0.0015 0.0009 -0.0004 0.0018
DAP fertilizer (kg ha
-1
) 0.0017*** 0.0004 0.0019** 0.0008
Potash fertilizer (kg ha
-1
) 7.10e-07 0.0003 -0.0007 0.0006
Urea fertilizer (kg ha
-1
)-5.67e-05 0.001 0.0007 0.0016
Zinc fertilizer (kg ha
-1
) 0.0032** 0.0016 0.0148*** 0.0046
Herbicides (applications season
-1
) 0.0014 0.0019 0.0016 0.0022
Pesticide (applications season
-1
) 0.0027** 0.0011 0.0038* 0.0019
Irrigation (applications season
-1
) 0.0011** 0.0004 0.0016*** 0.0005
Labor (person-days ha
-1
) 0.0015* 0.0007 0.0021 0.0018
Total precipitation during rice cropping season (mm) 0.0232** 0.0105 0.0343 0.0466
Days with temperatures [32 °C during rice growing season (sowing to
maturity) (n)
0.151 0.208 0.676 0.528
Days with daytime temperatures[35.5 °C during flowering stage in rice (n)-8.412*** 2.991 12.00*** 3.581
Constant 3.406 16.98 -62.73 56.58
Total observations (n) 240 – 240 –
R
2
0.590 0.743
*p\0.1, ** p\0.05, *** p\0.01
Paddy Water Environ
123
Conclusions and policy recommendations
Using a modified form of the Just and Pope production
function, we examined the effects of weather parameters
and their deviation from medium-term average climatic
conditions on farmers’ mean rice and wheat yields and
yield variance across eight of Pakistan’s AEZs. Our results
confirm the risks of extreme weather—particularly tem-
perature—and climatic variability on rice and wheat yields
in Pakistan, using data collected under on-farm circum-
stances, while also accounting for farmers’ variable agro-
nomic management practices. The number of days that
wheat exceeded season-long and flowering period thresh-
olds for temperature stress had a significant negative effect
on wheat yield, while also increasing deviation from the
mean, indicative of production risk. In the case of rice,
extreme heat had similar influence during flowering only;
supporting evidence that wheat is significantly more tem-
perature sensitive, while indicating the importance of
adaptation measures to overcome this constraint. Alongside
the coefficient estimates for the number of days in which
these crops exceeded temperature thresholds, the deviation
of temperature from the medium term was in most cases
positive, supporting our hypothesis that farmers may not
typically be able to adapt in real-time to large deviations
from historical climatic trends, resulting in reductions in
yield and increased yield risk. These observations were
however to some extent mediated by observed precipitation
patterns, although where drought to increase, yield vari-
ability would be expected to similarly follow.
These results have important implications for future
research and development strategies to adapt agriculture to
climate change in Pakistan, and also more generally in
South Asia. Agronomic management to escape the period
of critically high heat during flowering in wheat, for
example by using zero-tillage techniques with residue
retention, which can advance sowing dates while con-
serving soil moisture. Using heat-tolerant cultivars, and/or
using short-duration rice varieties and rapid harvesting to
establish the subsequent wheat crop as early as possible,
may also lend advantages. These measures therefore war-
rant policy initiatives and international donor support.
Most importantly, sustained extension efforts are needed to
raise farmers’ awareness of these adaptation strategies,
likely through a combination of mass media communica-
tion and experiential educational and training programs.
Where irrigation is available, timely application can also
help to lower canopy temperatures, especially during
flowering when rice and wheat yields are most sensitive to
heat stress. Our observations of increased wheat yield
Table 6 Estimates of the impacts of the climatic variability expressed as the deviation from the 32 year average on mean yield and variance
(kg ha
-1
) of rice, the latter indicative of production risk, using the Just–Pope (J–P) full model
J–P mean function J–P variance function
Coefficient Std. error Coefficient Std. error
Rice area (ha) -0.0807 0.0618 0.0158 0.354
Tillage passes season
-1
0.0017*** 0.0006 0.0029** 0.0012
Seed (kg ha
-1
) 0.0067 0.0041 0.0034 0.0061
Farmyard manure (kg ha
-1
) 0.0014* 0.0008 -0.0001 0.0019
DAP fertilizer (kg ha
-1
) 0.0013*** 0.0003 0.0012 0.0008
Potash fertilizer (kg ha
-1
) 9.64e-05 0.0003 0.0007 0.0008
Urea fertilizer (kg ha
-1
) 0.0003 0.0009 -0.0004 0.0016
Zinc fertilizer (kg ha
-1
) 0.0029** 0.0014 0.0111** 0.0049
Herbicides (applications season
-1
) 0.0002 0.0016 -0.0025 0.0024
Pesticide (applications season
-1
) 0.0023** 0.001 0.0043** 0.0019
Irrigation (applications season
-1
) 0.0093* 0.0051 0.0128** 0.006
Labor (person-days ha
-1
) 0.0016** 0.0006 0.0035* 0.0018
Total precipitation during rice cropping season (mm) 0.0046** 0.0019 0.0295 0.0457
Days with temperatures [32 °C during rice growing season (sowing to maturity) (n)-0.012 0.305 -0.660 0.653
Days with daytime temperatures [35.5 °C during flowering stage in rice (n)-0.864*** 0.235 1.637* 0.838
Deviation of the rice season’s mean precipitation from the historical mean (mm) 0.0504*** 0.0108 0.0828 0.0568
Deviation of the rice season’s mean temperature from the historical mean (°C) -1.688*** 0.195 1.869*** 0.230
Constant -61.75*** 19.48 -138.0* 75.71
Total observations (n) 240 – 240 –
R
2
0.633 0.783
*p\0.1, ** p\0.05, *** p\0.01
Paddy Water Environ
123
variance with irrigation however indicate that further
educational efforts would benefit farmers by raising
awareness on how to irrigate without destabilizing yields,
for example by avoiding nutrient leaching or gaseous
nitrogen losses that lower nutrient use efficiency and
increase global warming potential. These topics have
however received limited research attention in Pakistan—
especially under on-farm circumstances. Further research
should also consider farmers’ motivations to adapt to per-
ceived changes in the climate—or not to adapt—in order to
understand predominant decision-making processes and
how they influence crop choice, investment in inputs, and
agronomic management practices that influence yield sta-
bility. Such investigations could also assist in identifying
improved extension methods and messages that appeal to
farmers’ preferences and perceptions of changing climatic
conditions, and may ultimately be more impactful. Priori-
tizing investments in these crucial research areas may
therefore assist in developing viable climate change
adaptation practices that have a higher probability of being
adopted by farmers, with important implications for the
maintenance of rice and wheat productivity and food
security in South Asia.
Acknowledgments This study was financed by German Academic
Exchange Service (DAAD) and Higher Education Commission of
Pakistan (HEC) jointly; the field research and data collection com-
ponents of the project were funded by Stiftung Fiat Panis, Germany.
The Leibniz Centre for Agricultural Landscape Research (ZALF),
Germany, provided administrative support throughout the span of this
research work, which is highly appreciated. We thank the Pakistan
Metrological Department for climate data. Peter Crawford and Asad
Sarwar Qureshi assisted with advice on heat stress and tillage in
Pakistan, respectively. The authors also thank the two anonymous
reviewers for useful comments on an earlier version of the
manuscript.
References
Amjath-Babu TS, Krupnik TJ, Aravindakshan S, Arshad M, Kaechele
H (2016) Climate change and indicators of probable shifts in the
consumption portfolios of dryland farmers in Sub-Saharan
Africa: implications for policy. Ecol Indic 67:830–838
APSIM (2014) The APSIM-wheat module (7.5 R3008). http://www.
apsim.info/Documentation/Model,CropandSoil/CropModuleDo
cumentation/Wheat.aspx. Accessed 6 May 2016
Aravindakshan S, Rossi FJ, Krupnik TJ (2015) What does bench-
marking of wheat farmers practicing conservation tillage in the
eastern Indo-Gangetic Plains tell us about energy use efficiency?
An application of slack-based data envelopment analysis. Energy
90:483–493
Arshad M, Amjath-Babu TS, Ka
¨chele H, Mu
¨ller K (2015) What
drives the willingness to pay for crop insurance against extreme
weather events (flood and drought) in Pakistan? A hypothetical
market approach. Clim Dev 8(3):234–244. doi:10.1080/
17565529.2015.1034232
Asseng S, Foster IAN, Turner NC (2011) The impact of temperature
variability on wheat yields. Glob Change Biol 17(2):997–1012
Battisti DS, Naylor RL (2009) Historical warnings of future food
insecurity with unprecedented seasonal heat. Science
323(5911):240–244
Breusch TS, Pagan AR (1980) The Lagrange multiplier test and its
applications to model specification in econometrics. Rev Econ
Stud 47:239–253
Brown ME, Funk CC (2008) Food security under climate change.
Science 319(5863):580–581
Chang TT, Tagumpay O (1970) Genotypic association between grain
yield and six agronomic traits in a cross between rice varieties of
contrasting plant type. Euphytica 19(3):356–363
Chaudhry A (2000) Resource conservation technology and bed and
furrow system for crop production: training manual for training
at TII AME in Uzbekistan and OFWM in Pakistan. NZODA/
ADAF project. p 40
Deschenes O, Greenstone M (2007) The economic impacts of climate
change: evidence from agricultural profits and random fluctua-
tions in weather. Am Econ Rev 97(1):354–385
FAOSTAT (2016) Paddy rice and wheat yields in Pakistan. Food and
Agriculture Organization of the United Nations Statistical
Database. http://faostat.fao.org/. Accessed 8 May 2016
Farooq M, Bramley H, Palta JA, Siddique KHM (2011) Heat stress in
wheat during reproductive and grain filling phases. Crit Rev
Plant Sci 30:1–17
Gourdji SM, Sibley AM, Lobell D (2013) Global crop exposure to
critical high temperatures in the reproductive period: historical
trends and future projections. Environ Res Lett 8(2):024041
Guttormsen AG, Roll KH (2014) Production risk in a subsistence
agriculture. J Agric Educ Ext 20(1):133–145
Hatfield JL, Boote KL, Kimball BA, Ziska LH, Izaurralde RC, Ort D,
Thomson AM, Wolfe D (2011) Climate impacts on agriculture:
implications for crop production. Agron J 103:351–370
Holst R, Yu X, Gru
¨n C (2013) Climate change, risk and grain yields
in China. J Integr Agric 12(7):1279–1291
Isik M, Devadoss S (2006) An analysis of the impact of climate
change on crop yields and yield variability. Appl Econ
38(7):835–844
Jain M, Naeem S, Orlove B, Modi V, deFreis RS (2015) Under-
standing the causes and consequences of differential decision-
making in adaptation research: adapting to a delayed monsoon
onset in Gujarat, India. Glob Environ Change 31:98–109
Just RE, Pope RD (1978) Stochastic specification of production
functions and economic implications. J Econom 7(1):67–86
Kabubo-Mariara J, Karanja FK (2007) The economic impact of
climate change on Kenyan crop agriculture: a Ricardian
approach. Glob Planet Change 57:319–330
Keil A, D’souza A, McDonald A (2015) Zero-tillage as a pathway for
sustainable wheat intensification in the Eastern Indo-Gangetic
Plains: does it work in farmers’ fields? Food Secur
7(5):983–1001
Knox J, Hess T, Daccache A, Wheeler T (2012) Climate change
impacts on crop productivity in Africa and South Asia. Environ
Res Lett 7(3):034032
Koundouri P, Nauges C (2005) On production function estimation
with selectivity and risk. J Agric Resour Econ 30(3):597–608
Krishna V, Aravindakshan S, Chowdhury A, Rudra B (2012) Farmer
access and differential impacts of zero tillage technology in the
subsistence wheat farming systems of West Bengal, India.
CIMMYT-Socio-Economics Working Paper 5. CIMMYT,
Mexico
Krishnan P, Ramakrishnan B, Raja Reddy K, Reddy VR (2011) High-
temperature effects on rice growth, yield, and grain quality. Adv
Agron 111:87–206
Krupnik TJ, Santos Valle S, Hossain I, Gathala MK, Justice S,
Gathala MK, McDonald A (2013) Made in Bangladesh: scale-
appropriate machinery for agricultural resource conservation in
Paddy Water Environ
123
the context of smallholder farming. International Maize and
Wheat Improvement Center, Mexico City
Krupnik TJ, Ahmed ZU, Timsina J, Shahjahan M, Kurishi AA, Miah
AA, Rahman BS, Gathala MK, McDonald AJ (2015a) Forgoing
the fallow in Bangladesh’s stress-prone coastal deltaic environ-
ments: effect of sowing date, nitrogen, and genotype on wheat
yield in farmers’ fields. Field Crops Res 170:7–20
Krupnik TJ, Ahmed ZU, Timsina J, Yasmin S, Hossain F, Al Mamun
A, Mridha AI, McDonald AJ (2015b) Untangling crop manage-
ment and environmental influences on wheat yield variability in
Bangladesh: an application of non-parametric approaches. Agric
Syst 139:166–179
Kunimitsu Y, Iizumi T, Yokozawa M (2014) Is long-term climate
change beneficial or harmful for rice total factor productivity in
Japan: evidence from a panel data analysis. Paddy Water
Environ 12(2):213–225
Kunimitsu Y, Kudo R, Iizumi T, Yokozawa M (2015) Technological
spillover in Japanese rice productivity under long-term climate
change: evidence from the spatial econometric model. Paddy
Water Environ. doi:10.1007/s10333-015-0485-z
Lobell DB, Burke MB, Tebaldi C, Mastrandrea MD, Falcon WP,
Naylor RL (2008) Prioritizing climate change adaptation needs
for food security in 2030. Science 319(5863):607–610
Lobell DB, Sibley A, Ortiz-Monasterio JI (2012) Extreme heat effects
on wheat senescence in India. Nat Clim Change 2:186–189
Mendelsohn R, Nordhaus WD, Shaw D (1994) The impact of global
warming on agriculture: a Ricardian analysis. Am Econ Rev
84(4):753–771
Mondal S, Singh RP, Crossa J, Huerta-Espino J, Sharma I, Chatrath
R, Singh GP, Sohu VS, Mavi GS, Sukuru VSP, Kalappanavar IK
(2013) Earliness in wheat: a key to adaptation under terminal and
continual high temperature stress in South Asia. Field Crops Res
151:19–26
Morton JF (2007) The impact of climate change on smallholder and
subsistence agriculture. Proc Natl Acad Sci USA
104(50):19680–19685
PAR (2016) Wheat crop in Pakistan. Pakistan Agricultural Research.
http://edu.par.com.pk/wiki/wheat/. Accessed 23 May 2016
PARC, Pakistan Agricultural Research Council, Crop Sciences
Research Institute (2015). http://www.parc.gov.pk/index.php/
en/csi. Accessed 12 March 2015
Peng S, Huang J, Sheehy JE, Laza RC, Visperas RM, Zhong X,
Centeno GS, Khush GS, Cassman KG (2004) Rice yields decline
with higher night temperature from global warming. Proc Natl
Acad Sci USA 101(27):9971–9975
Poudel S, Kotani K (2013) Climatic impacts on crop yield and its
variability in Nepal: do they vary across seasons and altitudes?
Clim Change 116(2):327–355
Ray DK, Gerber JS, MacDonald GK, West PC (2015) Climate
variation explains a third of global crop yield variability. Nat
Commun 6:1–9
Reynolds MP, Balota M, Delgado MIB, Amani I, Fischer RA (1994)
Physiological and morphological traits associated with spring
wheat yield under hot, irrigated conditions. Aust J Plant Physiol
21:717–730
Reynolds MP, Quilligan E, Aggarwal PK, Bansal KC, Cavalieri AJ,
Chapman SC, Chapotin SM, Datta SK, Duveiller E, Gill KS,
Jagadish KSV, Joshi AK, Koehler A-K, Kosina P, Krishnan S,
Lafitte R, Mahala RS, Muthurajan R, Paterson AH, Prasanna
BM, Rakshit S, Rosegrant MW, Sharma I, Singh RP, Sivasankar
S, Vadez V, Valluru R, Vara Prasad PV, Yadav OP (2016) An
integrated approach to maintaining cereal productivity under
climate change. Glob Food Secur 8:9–18
Rosenzweig C, Elliott J, Deryngd D, Ruane AC, Mu
¨llere C, Arneth
A, Boote KJ, Folberth C, Glotter M, Khabarov N, Neumann K,
Piontek F, Pugh TAM, Schmid E, Stehfest E, Yang H, Jones JW
(2014) Assessing agricultural risks of climate change in the 21st
century in a global gridded crop model intercomparison. Proc
Natl Acad Sci 111(9):3268–3273
Sarker MAR, Alam K, Gow J (2012) Exploring the relationship
between climate change and rice yield in Bangladesh: an
analysis of time series data. Agric Syst 112:11–16
Schlenker W, Roberts MJ (2009) Nonlinear temperature effects
indicate severe damages to US crop yields under climate change.
Proc Natl Acad Sci USA 106:15594–15598
Seo SN, Mendelsohn R, Munasinghe M (2005) Climate change and
agriculture in Sri Lanka: a Ricardian valuation. Environ Dev
Econ 10:581–596
Shah F, Huang J, Cui K, Nie L, Shah T, Chen C, Wang K (2011)
Impact of high-temperature stress on rice plant and its traits
related to tolerance. J Agric Sci 149:545–556
Siddiqui R, Samad G, Nasir M, Jalil HH (2012) The impact of climate
change on major agricultural crops: evidence from Punjab,
Pakistan. Pak Dev Rev 51(4):261
Timsina J, Connor DJ (2001) Productivity and management of rice-
wheat cropping systems: issues and challenges. Field Crops Res
69:93–132
Wassmann R, Jagadish SVK, Sumfleth K, Pathak H, Howell G, Ismail
A, Serraj R, Redona E, Singh RK, Heuer S (2009) Regional
vulnerability of climate change impacts on Asian rice production
and scope for adaptation. Adv Agron 102:91–133
Yoshida S (1978) Tropical climate and its influence on rice. IRRI
research paper series 20. International Rice Research Institute,
Los Banos
Paddy Water Environ
123
