INTERNATIONAL MAIZE AND WHEAT
Modeling Temperature Response
in Wheat and Maize
Jeffrey W. White, Editor
Natural Resources Group
Geographic Information Systems
INTERNATIONAL MAIZE AND WHEAT
Modeling Temperature Response
in Wheat and Maize
Proceedings of a Workshop, CIMMYT,
El Batin, Mexico, 23-25 April 2001
Jeffrey W. White, Editor
Natural Resources Group
Geographic Information Systems
CIMMYT (www.cimmyt.org) is an internationally funded, ..,i., l.fit scientific research and training
organization. Headquartered in Mexico, CIMMYT works with I i. ,li!. dl. research institutions worldwide to
improve the productivity, I *fi J-Li',t: and sustainability of maize and wheat systems for poor farmers in
developing countries. It is one of 16 food and environmental organizations known as the Future Harvest Centers.
Located around the world, the Future Harvest Centers conduct research in partnership with farmers, scientists,
and policymakers to help alleviate poverty and increase food security while protecting natural resources. The
centers are supported by the Consultative Group on International Agricultural Research, whose members
include nearly 60 countries, private foundations, and regional and international. F :;i.:,;i Financial support
for CIMMYT's research agenda also comes from many other sources, including foundations, development banks,
and public and private agencies.
F U T U R E" Future Harvest builds awareness and support for food and environmental research for a
HARV'/E ST world with less poverty, a healthier human family, well-nourished children, and a better
environment. It supports research, promotes partnerships, and sponsors projects that bring the results of research
to rural communities, farmers, and families in Africa, Asia, and Latin America (www.futureharvest.org).
International Maize and Wheat Improvement Center (CIMMYT) 2003. All rights reserved. The opinions
expressed in this publication are the sole responsibility of the authors. The designations employed in the
presentation of materials in this publication do not imply the expression of any opinion whatsoever on the part
of CIMMYT or its contributory organizations concerning the legal status of any country, territory, city, or area, or
of its authorities, or concerning the delimitation of its frontiers or boundaries. CIMMYT ... 11 .. .fair use of
this material. Proper citation is requested.
Correct citation: White, J.W. (ed.). 2003. Modeling Temperature Response in Wheat and Maize: Proceedings of a
Workshop, CIMMYT, El Batdn, Mexico, 23-25 April 2001. NRG-GIS Series 03-01. M6xico, D.F.: CIMMYT.
AGROVOC Descriptors: Simulation models; Temperature; Maize; Wheat; Soil temperature; Soil water balance;
Nitrogen metabolism; Soil transport processes; Environmental factors; Thermal analysis; P31, ., ;' .i .J;i,
Statistical data; Crop yield
Additional Keywords: CIMMYT
AGRIS Category Codes: FO1 Crop Husbandry
P01 Nature Conservation and Land Resources
Dewey Decimal Classification: 633.1
Design and layout: Marcelo Ortiz S.
Printed in Mexico.
L H arrin g to n ............................................................................................................................................................ iv
J.W W h ite ............ ...................................................................................................... v
J.W W h ite ........................................................ .................................................................................. ............. v i
A Code-Level Analysis for Temperature Effects in the CERES Models
P. -,;. and U. Singh ................................................... ......... 1
A Physiological Perspective on Modeling Temperature Response in Wheat and Maize Crops
J.W W white and M R eyn olds .................................................................................. ........................................... 8
Re-examining Current Questions of Wheat Leaf Appearance and Temperature
G .S. M cM aster and L.A H unt. ................................................................ .................... .............................. 18
Simulating Response to Temperature
L.A. H unt, W Yan and G .S. M cM aste .................................................................................. ............................ 23
Evaluation and Calibration of CERES3-Maize: Tiller Number Simulation
A .S. du Toit and M .A P rinsloo ................................................................................................ ............. ............. 30
A Comparison of Approaches to Modeling Phenology as Applied to Genotype by
Sowing Date Interactions in Wheat
J.W W ite, S.S. Dhillon, L.A. Hunt and P.D. Jamieson ....................................................... ..................... 35
Short Description of the Model Statistical Package and Weather Analogue Program
A.S. du it and D.L. du it ........................................................ ........ 42
Report of the GCTE Tropical Cereals Network Inaugural Workshop
A .S. du it, J. Ingram and J.W W white .................................................................................... .......................... 47
Participants and Contact Inform action ........................................................................................................................ 52
This is the third international ..... 1, 1;,. workshop that CIMMYT's Natural Resource Group has hosted in
Mexico. Workshop themes are moving from broad ("Directions in Modeling Wheat and Maize for Developing
Countries," in 1998) to more specific i. l, .,. 1rI... Extremes of Wheat and Maize Crop Performance in the
Tropics," in 1999) to even more r...:r f.: with this year's focus on modeling temperature responses.
The .w'i- t,. of how best to describe a crop's response to temperature holds broad interest for CIMMYT's
research activities. In order to serve our partners scattered throughout the developing world, CIMMYT deals
with agroclimatic I .. that vary from tropical lowlands, where high temperatures can limit growth and
development, to highlands and temperate regions, where low temperatures and frosts are prevalent. While
models frequently are adjusted to better represent specific local conditions, CIMMYT's needs are best served by
models that show reliable and robust performance across environments, requiring a minimum of local
calibration or adjustment.
CIMMYT further recognizes that temperature stresses often are a component of effects of water deficits, whether
in rainfed or irrigated systems. To fully understand options for crop improvement and crop management where
water is a limiting factor, crop responses to temperature is also a concern.
The prospect of global warming provides further arguments for the need to understand better how crops
respond to temperature. Predictions of the Intergovernmental Panel on Climate (C 1.... that temperatures may
rise 2' to 5C by 2050 imply major shifts in cropping practices, cultivars and even crop species in coming years.
Crop models can offer valuable in ;.Ii into how agriculture will be affected by such. I.,,.. but for any
predictions to have Iw..! Il.. the underlying models must be based on sound physiological principles and be
tested over a realistic range of conditions. The workshop topic thus is also timely given proposals for the CGIAR
to develop a Ci, i i P, ... :, on Global -' N 1 ..,,
CIMMYT was especially pleased that most participants in the modeling workshop were able to stay an
additional two days for the GCTE Tropical Cereals Network Inaugural Workshop, which saw the enthusiastic
launch of this new GCTE network for maize, rice, sorghum and millets.
We thank David Poland, Marcelo Ortiz, Kristian Harrington-Col6n, and others for their assistance in editing and
laying out the final version for printing.
Natural Resources Group
The workshop "Modeling Temperature Response in Wheat and Maize," held from 23 to 25 April 2001 at
CIMMYT's headquarters at El Batan, Mexico, examined various approaches for modeling responses of crops to
temperature. Effects on both growth and development were considered.
One might expect that temperature responses have been examined so il.1. *.I.l.. that little novel or useful could
emerge from such an undertaking. However, in the second modeling workshop, i i, *,.Jlrg Extremes of Wheat
and Maize Crop Performance in the Tropics," the papers highlighted i-. .1 i. deficiencies in the responses of the
CERES Maize and Wheat models, with attempts to improve temperature responses of the models during the
working sessions. One suggestion was that these deficiencies reflected the .;;, I. y i' -,1 n,.,- of tropical and
subtropical conditions among the model development and testing datasets.
Further incentives for the workshop topic derived from two growing and somewhat inter-connected fields of
model application. Concerns over global change have lead to numerous studies on potential impacts of global
temperature and CO2 increases as well as on the potential for reducing atmospheric CO2 through "carbon
-. I- t.'. i"-" in agricultural soils. The second topic is integration of crop models with geographic information
systems. For land areas larger than a single x i0i 1;. or a large farm, spatial variation in temperature is usually a
primary determinant of differences in .., .i yield and environmental impacts.
The first day of the workshop opened with short oral presentations on aspects of modeling temperature response
in wheat and maize. The fi ll. .-; ii-, two days were dedicated to informal -. ., I 1 I:_ groups that examined -p...:.i,.:
topics in detail. In most cases, this involved running test data sets and J...l, i. model code.
Immediately following the workshop, a meeting was held to inaugurate the Global Change and Terrestrial
Ecosystem (GCTE) T ..- iJ Cereals Network, to provide a home for modeling efforts in maize, sorghum and
millet. (The GCTE Wheat Network was founded in 1992.)
During the workshop "Modeling Temperature P>" '*> in Wheat and Maize," held from 23 to 25 April 2001 at
CIMMYT's headquarters at El Batan, Mexico, participants examined various approaches for modeling
temperature effects in crops. Effects on both growth and development were considered.
A review of the FORTRAN code of the CERES models by Wilkens and Singh showed how temperature is
tl. ia. -1l to influence numerous processes, including carbohydrate production, -_ LIl, soil nitrogen dynamics,
root growth and evapotranspiration. The effect of temperature on grain ni.. j.- filling rate showed an
unexpected Jd i,, .,t-i iti which merits further investigation. It was suggested that the software code be
modified to allow greater flexibility in modifying temperature functions. This would facilitate testing of
.,'. .'I .- .1 model improvements.
White and Reynolds reviewed the expected temperature responses that crop models typically consider:
development (including phenology and morphology), ph.+-. lt.,--. ntI.. o-p .-:, t;. partitioning, and to a lesser
extent, nutrient uptake. Emphasis was given to the need to distinguish between immediate responses and
acclimation .-ff.' 1- and to use realistic growing conditions.
In an experiment where wheat was subjected to different soil temperatures, McMaster and Hunt found that
while increased soil temperature accelerated germination, soil temperature had no effect on subsequent
development. The authors. -1 I.. .1 that since meristematic regions in the shoot occur in various locations (i.e.,
apical meristem, internodes, and leaf sheath) and are subject to distinct thermal regimes, air temperature may
have a greater effect on development than was previously expected.
CERES-Maize currently does not simulate tiller production as found at low populations. Du Toit and Prinsloo
compared three approaches for modeling tiller production, ;m, 1,,lin two that consider temperature and
During the working groups, wheat modelers focused on reviewing modeling response to planting dates for
winter-sown spring wheats, ii; the data sets of S.S. Dhillon from Ludhiana, India for validation. The work
was continued following the workshop and showed that both CERES-Wheat 3.5 and Sirius had difficulties
simulating vernalization under the relatively warm conditions prevailing at Ludhiana.
To facilitate model validation and adaptation in the Highveld Ecoregion Project, two software packages were
developed by Du Toit and Du Toit. The Model Statistical P 1 i..,. uses standard outputs of CERES-Maize 3.0 to
calculate linear regression statistics (slope, intercept, and r2), D-index, and the systematic and unsystematic mean
square errors. The '. lI. i Analogue Program allows users to create mid-season projections of crop performance
based on five sets of historical data -h,. .; ii.-, the greatest similarity to the .. ,,- .;:_ season. The Model Statistical
r.,. I .i..:. was demonstrated .Im ii,. the working group and was found to be very L.'..,i; in:, for rapid assessment
of model performance.
During the in 111 .11 I meeting of the Global Ci i,. ., and Terrestrial Ecosystem (GCTE) T ... ii Cereals Network,
participants reviewed models available for maize, sorghum and millet and identified possible sources of data for
model evaluation. Plans for subsequent activities of the Network were outlined.
All soil t,, .r_ transformation processes are affected
by changes in soil temperature. Soil temperature in the
CERES models is simulated using daily air
temperature data, soil moisture status and available
water content (water balance I.Im1 ...' albedo,
solar radiation, and day of the year. The CERES
models use a simple soil temperature relationship to
modify soil N rates (Figure 6). As the soil temperature
increases, the rate of the processes continues to
10 0 10 20 30
Soil Temperature ('C)
Figure 6. Effect of soil temperature in a given layer on
Temperature and temperature derivatives are linked
to nearly every process in the CERES model.
Temperature-based functions can obfuscate true
physiological responses in some cases and need to be
critically evaluated as we continue model
development. Increased flexibility in modifying
temperature functions may be needed to improve the
To help address these concerns, the DSSAT crop models
have been re-designed and programmed to facilitate
more efficient incorporation of advances in
understanding of the underlying science. The basis for
the new DSSAT Cropping System Model (CSM) .I. ;,.
Sis a modular structure in which components are
structured to allow easy replacement or addition of
modules (Jones et al. 2003). This modular approach
makes it easier to isolate temperature effects from one
process (module) to another, in order to avoid spurious
Nitrification or auto-correlated temperature-related effects.
Hoogenboom, G., P.W. Wilkens, P.K. Thornton, J.W. Jones, LA. Hunt, and D.T.
Imamura. 1999. Decision support system for agrotechnology transfer v. 3.5.
40 50 In G. Hoogenboom, P.W. Wilkens, and G.Y. Tsuji, (eds.), DSSAT version 3,
Volume 4. Honolulu: University of Hawaii. Pp. 1-36.
Jones, J.W., G. Hoogenboom, CH. Porter, K.J. Boote, W.D. Batchelor, LA. Hunt, P.W.
nitrogen Wilkens, U. Singh, A.J. Gijsman, and J.T. Ritchie. 2003. The DSSAT Cropping
System Model. European Journal of 4.i,,,,, ,i i 235-265.
Ritchie, J.T. and D.S. NeSmith. 1991. Temperature and crop development. In R.J.
Hanks and J.T. Ritchie (eds.), Modeling plant and soil systems. Agronomy
Monograph 31. Madison: American Society of Agronomy. Pp. 5-29.
Singh, U. 1985. A crop growth model for predicting corn (Zea mays I.) performance
in the tropics. Ph.D. Thesis, University of Hawaii.
Swan, J.B., E.C Schneider, J.F. Moncrief, W.H. Paulson, and A.E. Peterson. 1987.
Estimating corn growth, yield, and grain moisture from air growing degree
days and residue cover. Agronomy Journal79: 53-60.
Tollenaar, M., T.B. Daynard, and R.B. Hunter. 1979. Effect of temperature on rate of
leaf appearance and flowering date in maize. Crop Science 19: 363-366.
Appendix A. Definitions of variables referred to in the example source code for CERES-Wheat version 3.5.
ALBEDO Albedo of crop plus exposed soil
P1V Relative amount that development is slowed for each day of unfulfilled vernalization, assuming that 50 days of vernalization is sufficient for all cultivars
P1D Relative amount that development is slowed when plants are grown in a photoperiod 1 hour shorter than the optimum (considered to be 20 hours)
P5 Relative grain filling duration based on thermal time (degree days above a base temperature of 1 C), where each unit increase above zero adds 20
degree days to an initial value of 430 degree days
G1 Kernel number per unit weight of stem (less leaf blades and sheaths) plus spike at anthesis (per g)
G2 Kernel filling rate under optimum conditions (mg/day)
G3 Non-stressed dry weight of a single stem (excluding leaf blades and sheaths) and spike when elongation ceases (g).
PHINT Phylochron interval, the interval in thermal time (degree days) between successive leaf tip appearances
SUMDIT Cumulative growing degree days (sum of DTT over time)
DTT Increment in thermal time each day
CARBO Realized production of carbohydrate per day after accounting for potential effects of temperature and stress factors
C02 Ambient concentration of CO2
C02X Reference concentration of CO2 associated with a given effect of CO2 on photosynthesis
C02Y Effect of CO2 concentration on photosynthesis for a given level of CO2
EEQ Equilibrium evaporation rate
EO Potential evapotranspiration (mm/day)
LAI Leaf area index
NSTRES Index of nitrogen stress (0 to 1 scalar)
PCARB Potential production of carbohydrate per day as limited only by solar radiation and light interception
PC02 Effect of CO2 on photosynthesis at reported concentration of CO2 (scalar)
PLTPOP Population of plants in field, usually assumed equal to sowing rate (plants/m2)
PRFT Effect of temperature on carbohydrate accumulation
ROWSPC Distancing between rows of plants in field (row spacing, in cm)
RTDEP Depth of root development (cm)
SALB Albedo of bare soil
SHF(L) Soil hospitality factor at depth L (0 to 1 scalar)
SLANG Solar radiation
SRAD Solar radiation as reported in file of daily weather (MJm-2day-1)
SWDF (0 to 1 scalar)
SWFAC (0 to 1 scalar)
TD Average daily temperature
TMIN Minimum temperature as reported in file of daily weather (C)
TMAX Maximum temperature as reported in file of daily weather( C)
XHLAI Leaf area index
A Physiological Perspective on Modeling
Temperature Response in Wheat
and Maize Crops
Jeffrey W. White1 and Matthew P. Reynolds2
1 Natural Resources Group, CIMMYT, Mexico
2 Wheat Program, CIMMYT, Mexico
In process-based models of wheat and maize crops, temperature ..' ..', influences both growth and
development as well as having indirect influences through water and nutrient balances. ?T..'. :' 0 .: crop
response to temperature would thus seem to be a :...'. U.' .. -.... J enterprise, as temperatures are readily
measured and .:*:.. through familiar procedures. `i... ., there is a large and valuable literature on
physiological responses to temperature, we are far from having a comprehensive understanding of how crops
respond to temperatures, especially as models move towards quantitative accuracy. This paper reviews the
physiology of temperature .,'. : on development, photosynthesis and respiration, with emphasis on maize
and wheat. The paper also suggests ways to make research more relevant for application in LT. to model
and predict crop response to the elevated temperatures expected under climate change scenarios. These
include accounting for possible, r of acclimation, ensuring that temperature treatments are relevant to
field conditions, using a wide enough range of temperatures to j,;i. characterize responses, and
characterizing the ..;.'. makeup of the plants under study.
At first glance, modeling the response of a crop to
temperature would seem straightforward.
Temperature is measured more readily than other
environmental variables, and differences in
temperature .;:,' are readily obtained under field
conditions by varying planting dates or using sites
that differ in elevation or latitude. In controlled
environments, temperature is easily modified through
thermostatically controlled ii. .It i:. and cooling
systems. Introductory biology courses teach that
reaction rates of many metabolic processes increase
' p" "."r"t.ill with temperature, approximately
doubling with each increase of 10C. The widespread
and highly repeatable occurrence of heat shock protein
(HSP) responses (Waters et al. 1996; Nakamoto and
Hiyama 1999) would seem to provide a direct window
into the molecular biology of crop response to
In practice, however, .ifi f., in the temperature
response of crops has proven remarkably difficult.
Different parts of a plant experience different
temperature regimes (Monteith and Unsworth 1'" ':
McMaster and Hunt, p. 18, this volume). Responses
vary according to conditions a plant has grown under,
i, t,. !, the important ability of crops to acclimatize.
Finally, attempts to improve adaptation li..... i
selection for specific stress responses, such as the one
to heat shock, have met with limited success.
Notwithstanding these potential complexities,
concerns over global warming-thought to be on the
order of 1.4 to from 1990 to 2100 (IPCC 2001)-
are driving widespread interest in estimating the
impacts of increased temperature on crops (e.g., Amir
and Sinclair 1991; Goudriaan 1 -"-'. Rosenzweig and
Tubiello 1996). Process-based crop models are often
used in assessing impacts of global II ... so there
is justification for reviewing the physiological
processes represented in such models.
This paper reviews basic physiological concepts related
to the effect of temperature on crop growth and
development with emphasis on wheat ( *. .'. -..1.
aestivum) and maize (Zea mays) crops. The ..' -. ~ i. .l. is
that such a review can help identify areas where crop
models can readily be t n ...l iI .1 or, ,li, I i, 1I.
suggest areas where more research is needed.
rl;. i..1. .' d development is usually analyzed in terms
of progress toward stages such as germination, seedling
emergence, initiation of floral primordia, flowering, and
S.11 ,1 1 maturity. T,...., is estimated by
integrating a developmental rate over the interval from
one stage to the next. The rate usually is a function of
temperature and photoperiod but also may vary with
;! '... 1 or water deficits, depending on the
developmental stage and crop species. If the daily mean
air or crown temperature is used, then the integration is
numerically equivalent to a summation. If lower or
upper temperature limits are imposed and the effect of
temperature on development rate is assumed constant,
the approach is equivalent to the widely used "growing
.1. *. day" (GDD) or "thermal time" concept. In this
concept the lower limit is termed the base temperature
(Tbase) and the upper limit is termed the c.1 i.ni.l
From this core simplicity among approaches, however,
numerous variants arise. Mean temperature may be
replaced by a diurnal curve, often interpolated from
daily minimum and maximum temperatures. While
conventional GDD models assume that above Topt the
rate continues at its maximum value, many simulation
models assume a decline in developmental rate at
very high (supra-i ...in il temperatures (Hunt et al.,
p. 23, this volume).
In wheat, vernalization represents a mechanism
independent of the basic temperature effect on
development in which seeds or plants require
exposure to low temperatures in order to initiate floral
primordia. Spring wheats have a low vernalization
requirement, while winter wheats may require 30 to 40
days of vernalization (Flood and Halloran 1986). The
process is generally assumed to occur at temperatures
of 0 to 100C and is unusual in that cooler
temperatures speed up the process, down to
temperatures near o0C. Models that include
vernalization usually represent the process th. ,..IJ a
rate of progress or "vernalization day" accumulator.
Besides direct effects of temperature on development,
increased temperatures are sometimes associated with
greater photoperiod sensitivity (e.g., Wallace et al. 1991;
White et al. 1996). This effect has been noted for time to
tassel initiation in maize (Ellis et al. 1992) and appears to
involve a decrease in the critical -i. i l.t. at higher
temperatures Fd.. ,,-l.. and Bolaiios I2 = =:). Cao and
Moss (1' -' *; found a temperature x photoperiod effect in
wheat and barley. Slafer and Rawson (1995) interpreted
similar results as evidence for an effect of photoperiod
on the basic temperature response of wheat;
mechanistically, however, it seems easier to conceive of
low temperatures slowing synthesis of an inhibitor that
is produced under less inductive photoperiods
(Hoogenboom and White 2003).
In contrast to processes such as photosynthesis and
respiration, there has been little effort devoted to
integrating research on the biochemistry of phenology
with quantitative models. Recent advances in
understanding the control of fl.. .... in Arabidopsis,
where approximately 60 .. 1. are now known to affect
floral development (Koornneef et al. i-'-' .. may offer a
road for merging biochemistry and ln...I.., ,: For
example, the FLF gene of Arabidopsis affects vernalization-
dependent flowering by encoding a protein that represses
the transition to flowering (Sheldon et al. 1999). Activity
of FLF is greater in vegetative rosette leaves of Arabidopsis
i .,, 111 1 i. .)ductive tissue and is sustained at a uniform
level at least until time of bolting. Such information
should allow developmental stages to be defined more
accurately and assist in determining which tissues are the
most relevant in temperature or photoperiod responses.
The effect of temperatures in reducing the length of the
growth cycle, especially the ., ;LC f;I,;:. phase, is usually
identified as the single most important factor in
explaining reduced yields at i ;: Ii. I temperatures.
' .L ..J.l and Cuellar:'-"*'. ) found a 3.1 day shortening of
grain tibnl: per oC temperature increase, vs. 2.8 days per
C in seven previous studies they reviewed. The effect on
yield was mainly through decreased grain weights.
Temperature affects morphology li ., i,, differential
effects on cell division and expansion. Higher
temperatures are associated with I ... I specific leaf
area (e.g., Midmore et al. 1' :, In their classic study
of temperature responses among 22 races of maize,
Duncan and Hesketh (1V- found that highland races
had greater relative leaf area growth rates than
lowland races at temperatures up to 300C (Figure 1).
In contrast, the effect of warmer temperatures on
increasing final leaf number was similar across races.
Of course, due to the dominant effect of temperature
on development, crops grown at higher temperatures
often have a lower Leaf Area Index (LAI) and canopy
size simply because of the shorter growth duration.
Growth is usually defined as an irreversible increase
in size, with "size" ..1 t ii. d as height, volume, or
fresh or dry weight. We focus our discussion on
changes in dry weight t, .. .,11 an increase in biomass
due to photosynthesis, with the increase balanced
against losses il 11....i respiration and senescence.
Species differ markedly in how temperature affects leaf
net photosynthetic rates (An), and these differences
play a major role in ,,3 -fl, *.., (Bj6rkman et al. 1980).
Stomatal limitations do not .I .. .i to be a major factor.
Several studies have shown that stomatal conductance
increases at higher temperatures and vapor pressure
deficits (VPDs) (Idso et al. 1'-. 1 Cornish et al. 1991;
Amani et al. 1996).
E 0.2 Highland
13 18 23 28 33 38
(Day time maximum)
Figure 1. Effect of temperature on relative leaf growth rate for means
of four highland and four lowland races of maize grown in controlled
temperature glasshouses (Duncan and Hesketh 1968).
The quantum efficiency of photosynthesis in C3 species
decreases almost linearly with temperature, due mainly
to increased photorespiration. In contrast, C4 species
show remarkably little variation over a broad range of
temperatures (Ehleringer and Pearcy 1983).
The stability of lipids in the thylakoid membranes is
also it.. i...li to influence temperature response of
photosynthesis (Raison et al. 1980; Carpentier 1999).
Membranes oi- i,. ,, phase transitions at critical low
and high temperatures. The values of the critical
temperatures are related to the degree of saturation
of membrane lipids and I .. i.'i. the types of
carotenoids present (Carpentier 1999). The
composition of membranes varies with growth
temperature, and such variation is presumed to be a
major part of acclimation mechanisms.
Studies of the short-term response of leaf net
photosynthetic rate in wheat suggest that sensitivity to
high temperature is very dependent on acclimation.
Both Blum (1-' *.,. and Sayed et al. (1989) showed that
the carbon .:h ;_,_ rate was relatively stable when
measured in the 23-33C n.... if plants were
acclimated to daytime temperatures in the region of
30C. However, plants grown with cooler day
temperatures (13-20'C) showed a marked decline in
An at temperatures above 25'C. The relative stability
of leaf An for wheat acclimated to warm conditions
was confirmed in field studies of 16 genotypes, where
An was stable when leaf temperatures varied naturally
(with time of day) between approximately 29' and
34C (Reynolds et al. _'* I I' I Nonetheless, genetic
variation for leaf An among cultivars at high
temperatures has been shown by both Sayed et al.
(1989) in controlled environments and by Reynolds et
al. '* '* *. in the field. Furthermore, longer-term
exposure to high temperature reduces An. Al-Khatib
and Paulsen (1990) described an accelerated
development rate at 32/27C versus 22/17C (day/
; :_ ]I temperature regimes, c.. ,.l ltil in decreased
duration of photosynthetic activity and lower kernel
weights, -.C-p.. )i in sensitive cultivars. FI .i I .J
exposure to high temperatures in controlled
environments also resulted in premature loss of
chlorophyll (Al-Khatib and Paulsen 1' -i In warm
field environments, a large proportion of variation in
photosynthetic rate among cultivars during
.,1 ir;flli,. was explained by differential. 1t1.. l 'l. II 1
loss after anthesis, which explains 30-35% of the
variation in final grain yield (Reynolds et al. ',
ji., i. ,m,,:. other photosynthetic tissues in wheat
respond quite differently. Blum (1986) found that
glume photosynthesis declined drastically above
25C whether plants were acclimated or not, while
carbon fixation by awns increased with higher
temperature irrespective of acclimation. Awns may
intercept an appreciable proportion of incident
radiation (e.g., > '-, in :.. '... 1 I with long awns,
such as many durum wheat cultivars). They are
hence likely to contribute '.if .i. .mii, to total
canopy photosynthesis during _,i.]]fiiiI .ll-, ..- ,.Al.
in warm environments.
While photosynthesis may be relatively heat stable in
acclimated plants, studies of how starch synthesis
responds to temperature suggest that grainfilling may
be affected more by the inhibition of conversion of
sucrose to starch at temperatures above 30C (Bhullar
and Jenner 1' -.. soluble starch synthase appears to be
especially sensitive (T ....i n- et al. 1993).
Duncan and Hesketh (1968) found that highland and
lowland maize races showed similar responses of An
to growth temperature, but lowland races had higher
rates above 18C r-'F .. 2).
F. ii'; I;. rates usually increase "i uli.iiL with
temperature up to an inflection point, where various
effects can slow metabolism (Loomis and Connor
1992; Smith et al. 1999). Such responses are mainly
- 1.6- Lowland
1 Duncan and Hesketh
13 18 23 28 33 31
Figure 2. Effect of growth temperature on leaf photosynthetic rate for
highland and lowland races of maize grown in controlled temperature
glass houses. The temperature is the daytime maximum and
corresponds to the measurement temperature. Based on data from
Duncan and Hesketh (1968).
attributed to the maintenance (instead of growth)
component of respiration (Loomis and Connor 1992).
Differences in temperature response of 1.' p ..' '..ni are
often associated with differences in adaptation,
leading Smith et al. (1999) to suggest that these
responses may be better indicators of adaptation than
those found for photosynthesis.
In a study of 16 wheat cultivars grown under warm
field conditions, dark respiration rates measured in
two experiments with different sowing dates were
substantially higher in the warmer environment.
Taken with the fact that final biomass, but not leaf
photosynthetic rate, was reduced in the warmer
environment, this finding supports the idea that
I..- t. ;-,tl;. had a higher metabolic cost under warmer
conditions (Reynolds et al. _-' -' I;. Nonetheless,
temperature was clearly not the only factor
,itl.I,.n, 'J respiration. F', p. 1..n rates in both
environments were relatively stable until anthesis but
declined by approximately 35% during grainfilling,
more or less mirroring loss in total leaf chlorophyll.
Despite these clear main effects for respiration, the
trait was not well correlated with any of the
performance traits when comparing cultivars.
Relatively little is known about how metabolic
responses to temperature interact to mediate
. i.4t.41tii i of crops to stressful environments with
respect to performance. TIr.., ri. ..11', the functional
genomics approach to understanding stress response
would make it possible to reveal the biochemical and
-". ,. basis of any given phenotypic response to the
environment. However, adaptation to stress at the
whole plant level involves the interaction of many
I. which are expressed at multiple levels (i.e.,
tissue, phenological stage, time of day, etc.).
Considerable investment will be needed before such
1.l,., A.ir..-,,- become routine. Moreover, much of the
research at the molecular level considers survival
mechanisms rather than I .Ji i;. '- which depends
more on stress avoidance than tolerance.
;' .' % i i 1 1 L th ese i ... i .... 1 1. ..t .
transformation of wheat has been attempted based on
known metabolic responses to stress. For example, late-
I,. %.l ?, i ..-r i--.l...1 .. t (LEA) proteins .,t .- .., when
drying initiates in developing seeds and disappear after
imbibition ciF', .., i et al. 1993). The genes are similar to
those expressed in drought stressed ;. .1; tissue of
wheat (Curry et al. 1991), and ABA can induce
expression of these proteins. Sugar synthesis also seems
to play a role in drought stress, providing compatible
solutes for osmotic adjustment i F... ILw t et al. 1995) or
through various protective roles, including protection of
membranes (Crowe et al. 1992). Antioxidants such as
superoxide dismutase and ascorbate peroxidase increase
in response to drought stress (Mittler and Zilinskas
1994) and probably play a role in tolerance, since the
excess radiation and increased photorespiration
associated with stress can result in accumulation of
active oxygen species. Other relatively simple
biochemical processes involved in drought stress may
also lend themselves to genetic transformation. These
include osmotic adjustment, repair and degradation of
proteins, and structural adjustment-for example, of the
cell wall (Ingram and Bartels i .*
Fowler et al. (1999) reviewed physiological and
molecular research on cold tolerance in wheat and used
this synthesis to guide modification of the CERES
Wheat model. Their approach included cumulative
effects of hardening, acclimation, and de-hardening, all
as functions of the daily mean crown temperature.
For maize cultivars in Canada, Ying et al. (2000)
found that recently released hybrids showed less
reduction in photosynthesis after exposure to low
night temperatures. This effect was also reflected in
changes in chlorophyll fluorescence, -, _- ic_. a
stress effect on photosystem II; however, subsequent
work indicated that the effect is not il1,h, 1,
photoinhibition (Ying et al. 2002).
Freezing injury is associated with membrane damage
and subsequent tissue dehydration (Thomashow 1' .* ,
Acclimation at the cell level appears to involve
processes that either stabilize membranes or reduce the
fI,, i point of the cytoplasm.
The main effects of temperature on plant nutrition are
presumably indirect and derive from effects on overall
plant growth, including root elongation. However,
temperature also affects availability of nutrients in the
soil solution and uptake at the root surface. At low
temperatures, uptake and metabolism of ammonium is
greater than that of nitrate. This difference can greatly
alter the cation-anion balance and root-induced
.:-:,..:_.. in rhizosphere pH (Marschner 1''' -.,.
Similarly, low temperatures can increase problems of
zinc ..1.fi ; i,, .. i -c.,. .i,. due to decreased solubility
of soil Zn (Lucas and Knezek 1972). For phosphorus,
Mackay and Barber (19 I; concluded that temperature
effects on P movement in the soil and P uptake at the
root surface were much less important than
temperature effects on root growth per se.
Whole Plant or Community
While the previous discussions have focused on
separate processes affected by temperature, the
ultimate question is how the various physiological
processes interact to affect overall 3i' ... iI and yield.
To examine aggregated responses physiologists rely
on various analytic l 1 I with growth
analysis being perhaps the most widely used
approach to analyze the behavior of a crop as a
community of plants.
Although the life cycle of a wheat crop is accelerated
and reduced in its duration at higher temperatures
(Midmore et al. 1982), the reduction in cycle 1 :li only
partially accounts for lower productivity. In field
studies comparing performance of different genotypes
in .. ) I, i, thermal environments, the I; I I
rate was on average over 10% lower at the warmer site.
In addition, above ground biomass was 6% lower for
the three most heat tolerant genotypes and 15% lower
for the three most heat sensitive lines 1', .. n..i. et al.
1998). One factor seems to be related to reduced light
interception. In a parallel experiment, canopy
establishment of a single heat tolerant cultivar was
compared at two sowing dates ..n-; in o. in
temperature. Plant dry weight at the 5-leaf stage was
considerably lower at the warmer sowing date and was
associated with lower ground cover and reduced early
light interception (Badarrudin et al. 1999). In the
comparison by Midmore et al. (1984) of wheat lines in
contrasting thermal environments, green-area index
measured shortly after flowering showed substantially
lower values at the warmer sites. Several values fell
under the critical LAI value of 3, an approximate lower
limit for full l1, l. -., of radiation. F;n ill. loss of
chlorophyll during :_ a;hfi;ll;,-, has been associated
with reduced field performance in warm environments
(Reynolds et al. 2000). Reduced stand establishment,
lower final green-area index, and accelerated chlorophyll
loss are all factors that reduce light interception under
warm conditions, and are likely to have a ,, ,.l, .. impact
on total canopy photosynthetic rate.
The impact of reduced source on yield is confirmed by
numerous reports of reduced kernel weight in response
to elevated temperature, with kernel weight affected
typically by a 2-5% decrease per C increase (Wardlaw
and .i r' ,. 1994). In addition to the direct effect of
temperature on source potential, other studies have
shown that potential kernel number in wheat shows a
negative association with temperature during the spike-
growth stage. Fischer (1985) calculated a kernel number
reduction rate of 4% per "C in the range of 14"-22"C. In a
study .ill. 11qii ,. to explain wheat species by year
interactions in a relatively stable environment
(northwest Mexico, irrigated), one of the factors best
Stl.ii why certain years favored bread wheat over
durum yield was temperature during the spike growth
stage (Reynolds et al. 2002). Durum wheat was found to
be more sensitive to warmer temperatures during this
period. A similar analysis was conducted for 16 wheat
genotypes grown in 20 .:..it 1 i thermal
environments ,. p"' "ti". sites in various countries).
These were a subset of the 40 environments (Reynolds
et al. : ''- for which detailed temperature data were
available. The analysis ,;111 .vealed that temperature
in the -; .. .rth stage was the most significant factor
1 in;,,. G x E effects. '-1I. -i. ii i average minimum
temperature during approximately 30 days before
fi., ,. ,i i. Ii ;-..1 from 10.5 to 16"C, .1. g. i:. on
environment (Vargas et al. 1998). The results would
suggest that high night temperatures reduced grain
number (and therefore yield) either because accelerated
development resulted in a .-1.i i. -i growth stage or
because high respiration costs reduced spike growth itself.
It is also possibly due to a combination of both factors.
Hardacre and Turnbull l '- --, found that,. i i ." .' .i'
rate and net assimilation rate of maize increased with
temperature in controlled environments from 16" to 28"C.
Radiation Use Efficiency
The amount of biomass accumulated per unit solar
radiation intercepted, radiation use efficiency (RUE), is
another widely used indicator of crop performance
(Charles-Edwards 1982; Sinclair and Muchow 1999). In
simplified terms, RUE integrates effects of n, ,,
structure and light interception, photosynthesis within
the I .. and loss of assimilate I11 .I. ;.1, respiration
and senescence. Implicit in this analytic approach are
the assumptions that photosynthesis is !.,. I 1 I. i .1I
to weight loss tlh ..Jci I -p ;t;. ,, and that the
response of canopy photosynthesis to solar radiation
is linear (Charles-Edwards 1982; Kiniry et al. 1989).
In a review of published values of RUE for maize
and sorghum, Kiniry et al. (1989) concluded that
there was no variation in RUE with temperature for a
range of 19" to 27C. Using a range of planting dates
,.i; in;, mean temperatures from 16 to 21C, however,
Andrade et al. (1993) found that maize RUE
increased linearly from 2.2 to 3.2 g/MJ.
Batts et al. (1;"- studying winter-wheat grown at
mean temperatures from 9" to 12"C in temperature
gradient tunnels in the UK, found that harvest
index (HI) of one cultivar was constant while it
declined slightly at higher temperatures in the
other. In the study of Mid more et al. (1', : ; on
spring wheats in Mexico, the highest values of
harvest index i1;: ;4 to I: --': were obtained at the site
with intermediate temperatures. In an extensive
experiment where 16 wheat cultivars were grown in
over 40 environments internationally (Reynolds et
al. 1 -"*'' harvest index was found to be sensitive to
temperature. ,..', .1. across .. ni.t. I. harvest
index ranged from 0.4 in the cooler environments to
0.3 at some warmer sites.
Lafitte and Edmeades (1997),.. ni' ;.'.
performance of diverse maize types in Mexico,
found that the response of harvest index differed
with expected adaptation of cultivars. Two lowland
cultivars showed a slight increase in harvest index
(roughly from 0.3 to 0.4) for mean temperatures
from 17" to 28"C, while highland cultivars
decreased from values of 0.35 to almost 0.
Studies looking directly at temperature effects on
growth and yield at the field level ., ... 11i, use either
multiple locations or sowing dates (e.g., Muchow et al.
: --. Mid more et al. 1982 and I -. I. Lafitte and
Edmeades 1997) or temperature treatments applied to
either open top chambers or temperature gradient
tunnels (e.g., Batts et al. 1' -,, Table 1 summarizes
reported effects of temperature on yield, including both
field studies and simulation exercises. In most studies,
warmer temperatures reduced '- ..II '.l .p 'Anl, which
in turn limited total biomass accumulation and yield.
Table 1. Reported effects of mean temperature on grain yield in wheat and maize. Effects were estimated as declines from maximum reported grain
yields in each study.
Temperature grain yield effect on grain
Source Description range (C) (kg ha') yield (kg ha-' C-')
Midmore et al. 1984 Diverse cultivars grown at various sites and sowing
dates in Mexico 14-25 5,900 450
Amir & Sinclair 1991 Simulated for 10 year period in Israel [8C range] 9,500 760
White et al. 2001 Simulated for cv. Seri M-82 with CERES-Wheat assuming
18 MJ m-2d-1 solar radiation 10-30 10,000 500
Rosenzweig & Tubiello 1996 Simulations with CERES-Wheat at four locations in the USA
(maximum yields and effects shown here correspond to the four sites) [4'C range] 3,000 300
Wheeler et al. 1996 Two seasons data for winter wheat cv. Hereward grown in an
outdoor temperature gradient tunnel in the UK 14- 21 9,000 1040
Batts et al. 1998 Mean of two winter wheat cultivars grown in an outdoor
temperature gradient tunnel in the UK 9-12 8,500 1800
Muchow et al. 1990 Field trials at five locations in Australia and the USA with
various cultivars 18-29 17,000 810
White et al. 2001 Simulated for cv. Suwan 8222 with CERES-Maize assuming
18 MJ m-2d- solar radiation 15-30 14,000 530
The large variation in the magnitude of the response
presumably reflects differences in management and
solar radiation regimes. Based on this limited set of
data, wheat appeared to be more sensitive than
maize, which agrees with the expectation that C4
crops are more heat tolerant than C3 crops. The
responses predicted by simulation models ':.'. ..-I.
were smaller than those observed in the field,
I.. Ii[. indicating the importance of the additional
influences of water deficits, diseases and pests.
Monteith and Unsworth (1990) illustrated the
p. .I.I.ii for large gradients of air temperature
within a crop canopy, both for daytime and nighttime
conditions. In wheat, mid-day canopy temperatures
of well-watered crops are often 5 to 8'C lower than
the air temperature (e.g., Amani et al. 1996). While air
temperature is readily measured, Ehleringer (1991)
emphasized that accurate measurement of different
tissues within a plant requires careful attention to
energy budgets of the tissue and the temperature
probe. A recent study with a set of 16 wheat
. .. I,, -". grown under different thermal regimes in
the field involved measuring different organ
temperatures. While there were *, f. ii. ,i main
effects for leaf, stem and spike temperature, no
interaction with genotype was found (Ayeneh et al.
-'12 -, an important finding for L t.I ..Ii.Lan'_ modeling
approaches to different cultivars.
Due to the effects of the total energy budget, imposing
meaningful temperature treatments is also problematic.
Growth chambers seldom provide radiation regimes
similar to natural environments. Photosynthetically
active radiation is usually lower than in natural
sunlight, and net radiation will never match conditions
in the field. "Cii;,l. while 1in.:' elevation, latitude,
or planting date will result in readily measured changes
in air and soil temperature, these treatments are usually
associated with changes in the radiation regime,
photoperiod, or soil conditions. In their review of
phenology in wheat, Slafer and Rawson (1995) noted
that use of sowing date studies to characterize
temperature sensitivity of phases prior to anthesis is
problematic. Developmental stages just prior to
flowering are usually exposed to a narrower and more
favorable range of conditions than earlier stages.
A further concern is that standard approaches for
i lp.-i;i'-, temperature treatments seldom allow use of
true replication. Replication within a chamber,
glasshouse or sowing date plot is statistically
equivalent to sub-sampling.
Physiological processes vary remarkably in their
response to temperature. While maintenance
respiration increases exponentially with temperature
(up to a point), development and photosynthesis
seem better described with other types of curves.
The widespread use of "broken-stick" models that use
straight lines to represent temperature responses
seems to be viewed with embarrassment by
researchers who favor models rooted in physical
processes. There clearly is a problem in -.. i. i'; a
linear response if the temperature response is dictated
by kinetic energies -,i. ." i:_- a Boltzmann distribution
at the molecular level, which thus should show an
. I ... .?r t,:.,l response (e.g., Nobel 1991). Not
surprisingly, various researchers have attempted to
explain the robustness of linear models as essentially
simplifications of inherently curvilinear responses.
For example, Sinclair (1'-'- ;, argued that a linear
approximation works well because responses "are
usually obtained in a relatively narrow temperature
range, say 288' to 303'K."
Our growing understanding of ..III.i. and
biochemistry of development, however, indicates that
many processes involve complex biochemical
networks, in many cases with surprising levels of
redundancy or alternate pathways (von Dassow et al.
Some regulatory processes may involve only
one or two molecules per cell binding to a single
DNA molecule (Anonymous 1999), ,,_.- .-. n,_. that
models based on i. .1..I'Il. distributions of kinetic
energy will often be* i- j .1 *., ; I, Similarly, in cases
where membranes are involved, responses may
reflect phase changes or abrupt shifts in binding of
proteins to membranes. Note that approximately 20%
of proteins show trans-membrane segments (Boyd et
al. 1' Wallin and von Heijne 1998).
While our qualitative ...1, I ... of temperature
effects appears quite detailed, we are still far from
being able to convert such ii -- 1. .1,.. into the
quantitative predictions of adaptation sought by
breeders, agronomists and modelers. These
predictions are also m.l.1 Ii in .r u. 'i- to assess the
impact of global warming on agriculture. Process-
based models offer an attractive vehicle for
integrating knowledge of effects of temperature in a
quantitative framework, with the dual goals of
,, ;. *;M I:- "best bets" on the effects of temperature
based on current knowledge and of : Ii.in further
In .mt, n'I1ni to improve our ,1 ) i I ,,,i, of the
effects of temperature, we suggest various points that
researchers and modelers should keep in mind:
* Assess effects of conditioning or acclimation.
* Use a large -. i .r- number and broad enough
range of temperature treatments to permit different
shapes of response curves to be distinguished.
* As far as possible, grow plants under field
conditions or demonstrate how results from
controlled environments relate to field conditions.
* Monitor air, soil, leaf, or crown temperatures as
appropriate to ensure that the i., -i. 1, -i lb
relevant temperature is 1i..'-'itl l..d
* Use materials that are genetically well characterized
for major genes that may influence responses (e.g.,
for photoperiod and vernalization).
* DIE -t:;_ ,. -i clearly among models that are
proposed as pragmatic solutions for a well defined
geographic domain and those that represent
analytical tools for understanding 1 .-i, ..I.. -i. i
Al-Khatib, K. and G.M. Paulsen. 1984. Mode of high temperature injury to wheat
during grain development. Physiologia Plantarum 61: 363-368.
Al-Khatib, K. and G.M. Paulsen. 1990. Photosynthesis and productivity during
high-temperature stress of wheat genotypes from major world regions. Crop
Science 30: 1127-1132.
Alexandrov, V.A. and G. Hoogenboom. 2000. The impact of climate variability and
change on crop yield in Bulgaria. Agric. and Forest Meteor. 104: 315-327.
Amani, I., R.A. Fischer, and M.P. Reynolds. 1996. Canopy temperature depression
association with yield of irrigated spring wheat cultivars in a hot climate.
Journal of Agronomy & Crop Science 176:119-129.
Amir, J. and T.R. Sinclair. 1991. A model of the temperature and solar-radiation
effects on spring wheat growth and yield. Field Crops Research 28: 47-58.
Andrade, .H., S.A. Uhart, G.G. Arguissain, and R.A. Ruiz. 1992. Radiation use
efficiency of maize grown in a cool area. Field Crops Research 28: 345-354.
Anonymous. 1999. Advanced computational structural genomics htp://
cbcg.bl.gov/ssi-csb/Meso.html (25 March, 2000).
Ayeneh, A., M. van Ginkel, M.P. Reynolds, and K. Ammar. 2002. Comparison of
leaf, spike, peduncle and canopy temperature depression in wheat under
heat stress. Field Crops Research (in press).
Badaruddin, M., M. Reynolds, and O.A.A. Ageeb. 1999. Wheat management in
warm environments: Effect of organic and inorganic fertilizer, irrigation
frequency, and mulching. Agronomy Journal91: 975-983.
Batts, G.R., R.H. Ellis, J.I.L Morison, P.N. Nkemka, P.J. Gregory, and P. Hadley.
1998. Yield and partitioning in crops of contrasting cultivars of winter wheat in
response to C02 and temperature field studies using temperature gradient
tunnels. Journal of Agricultural Science 130: 17-27.
Bhullar, S.S. and CE Jenner. 1986. Effects of temperature on the conversion of
sucrose to starch in the developing wheat endosperm. Australian Journal of
Plant Physiology 13: 605-615.
Bjiirkman, 0., M.R. Badger, and P.A. Armong. 1980. Response and adaptation of
photosynthesis to high temperatures. In N.C Turner and P.J. Kramer (eds.)
Adaptation of plants to water and high temperature stress. New York: John
Wiley & Sons. Pp. 233-249.
Blum, A. 1986. The effect of heat stress on wheat leaf and ear photosynthesis. J.
Exp. Bot. 37: 111-118.
Bohnert, H.J., D.E. Nelson, and R.G. Jensen. 1995. Adaptations to environmental
stresses. Plant Cel7: 1099-1111.
Bonham-Smith, P.C., M. Kapoor, and J.D. Bewley. 1988. A comparison of the stress
responses of Zea mays seedlings as shown by qualitative changes in protein
synthesis. Canadian Journalof Botany 66: 1883-1890.
Boyd, D., C. Schierle, and J. Beckwith. 1998. How many membrane proteins are
there? Protein Sci. 7: 201-205.
Brain, P. and R. Butler. 1988. Cumulative count data. Genstat Newsletter 22: 38-45.
Cao, W. and D.N. Moss. 1989. Temperature and daylength interaction on
phyllochron in wheat and barley. Crop Science 29: 1046-1048.
Carpentier, R. 1999. Effect of high temperature stress on the photosynthetic
apparatus. In M. Pessarakli (ed.), Handbook of Plant and Crop Stress. New
York: Marcel Dekker, Inc. Pp. 337-348.
Charles-Edwards, D.A. 1982. Physiological determinants of crop growth. New York:
Cornish, K., J.W. Radin, E.L Turcotte, Z.M. Lu, and E. Zeiger. 1991. Enhanced
photosynthesis and gs of pima cotton (Gossypium barbadense L.) bred for
increased yield. Plant Physiology 97: 484-489.
Crowe, J.H., FA. Hoerkstra, and L.M. Crowe. 1992. Anhydrobiosis. Annual Review of
Plant Physiology 54: 579-599.
Curry, J., CF Morris, and M.K. Walker-Simmons. 1991. Sequence analysis of a DNA
encoding Group 3 LEA mRNA inducible by ABA or dehydration stress in wheat.
Plant Molecular Biology. 16:1073-1076.
Donald, CM. and J. Hamblin. 1976. The biological yield and harvest index of
cereals as agronomic and plant breeding criteria. Advances in Agronomy 28:
Duncan, W.G. and J.D. Hesketh. 1968. Net photosynthetic rates, relative leaf growth
rates, and leaf numbers of 22 races of maize grown at eight temperatures.
Crop Science 8: 670-674.
Edmeades, G. and J. Bolafios. 2001. Issues Arising from the use of CERES for
Tropical Maize. In J.W. White and P. Grace (eds.), Directions in Modeling Wheat
and Maize for Developing Countries: Proceedings of a Workshop, CIMMY, El
BatOn, Mexico, 4-6 May 1998 4: i. D.F.: CIMMYT. Pp. 13-17.
Ehleringer, J. and R.W. Pearcy. 1983. Variation in quantum yield for C02 uptake
among C3 and C4 plants. Pl .:v P,. 73: 555-559.
Ehleringer, J. 1991. Temperature and energy budgets.In R.W. Pearcy, J. Ehleringer,
H.A. Mooney and P.W. Rundel (eds.), Plant Physiological Ecology. New York:
Chapman and Hall. Pp. 117-135.
Ellis, R.H., R.J. Summerfield, G.O. Edmeades, and E.H. Roberts. 1992. Photoperiod,
temperature, and the interval from sowing to tassel initiation in diverse
cultivars if maize. Crop Science 32:1225-1232.
Fischer, R.A. 1985. Number of kernels in wheat crops and the influence of solar
radiation and temperature. Journalof. lyAu.f: u ..",,ji e 105: 447-461.
Flood, R.G. and G.M. Halloran. 1986. Genetics and physiology of vernalization
response in wheat. Advances in Agronomy 39: 87-125.
Fowler, D.B., A.E. Limin, and J.T. Ritchie. 1999. Low-temperature tolerance in cereals:
Model and genetic interpretation. Crop Science 39: 626-633.
Garcia-Huidrobo, J., J.L Monteith, and G.R. Squire. 1982. Time, temperature and
germination of pearl millet (Pennisetum typhoides S & H.). I. Constant
temperature. Seed Science and Technology 33: 288-296.
Gausterland, T. 1998. Structural genomics: Bioinformatics in the driver's seat. Nature
Biotechnology 16: 625-627.
Goudriaan, J. and H.H. van Laar. 1994. Modeling Potential Crop Growth Processes.
Dordrecht: Kluwer Academic Publishers.
Goudriaan, J. 1996. Predicting crop yields under global change. In B.H. Walker,and
W.L. Steffen, (eds.), Global change and terrestrial ecosystems. International
Geosphere-Biosphere Programme Book Series Vol. 2. Cambridge, UK:
Cambridge University Press. Pp. 260-274.
Hall, A.E. 1992. Breeding for heat tolerance. 'i.~v: ar.-A.,lq Reviews 10: 129-168.
Hardacre, A.W. and H.L. Turnbull. 1986. The growth and development of maize (Zea
mays L.) at five temperatures. Ann. Bat. 58: 779-787.
van Herwaarden, A.E; R.A. Richards, G.D. Farquhar, and J.F Angus. 1998. 'Haying-
off', the negative grain yield response of dryland wheat to nitrogen fertilizer.
Ill. The influence of water deficit and heat shock. Australian Journal of
Agricultural Research 49: 1095-1110.
Ho, T.H.D. and M.M. Sachs. 1999. Environmental control of gene expression and
stress proteins in plants. In H.G. Jones, T.J. Flowers and M.B. Jones (eds.), Plants
under Stress. Soc. Exper. Biology Seminar Series 39. Cambridge, UK: Cambridge
Univ. Press. Pp. 157-180.
Hoogenboom, G. and J.W. White. 2003. Modification of a crop simulation model that
incorporates gene action. Agronomy Journal(In press.).
Hunter, E.A., CA. Glasbey, and R.E.L Naylor. 1984. The analysis of data from
germination tests. J. Agric. Sci. 102: 207-213.
Idso, S.B., R.J. Reginato, K.L Clairson, and M.G. Anderson. 1984. On the stability of
stress baselines. Agric. & For. Meteor. 32: 177-182.
Ingram, J. and D. Bartels. 1996. The molecular basis of dehydration tolerance in plants.
Annual Review of Plant Physiology and Plant Molecular Biology 47: 377-403.
IPCC (Intergovernmental Panel on Climate Change). 2001. IPCC Third Assessment
Report: Climate Change 2001: Impacts, Adaptation, and Vulnerability. httLp
Jame, Y.W., H.W. Cutforth, and J.T. Ritchie. 1998. Interaction of temperature and
daylength on leaf appearance rate in wheat and barley. Agric. and Forest
Meteor 92: 241-249.
Jamieson, P.D., M.A. Semenov, I.R. Brooking, and G.S. Francis. 1998. Sirius: A
mechanistic model of wheat response to environmental variation. European
Journal of Agronomy 8:161-179.
Keeling, P.L, P. Bacon, and D. Holt. 1993. Elevated temperatures reduces starch
deposition in wheat endosperm by reducing the activity of soluble starch
synthase. Plant 191: 342-348.
Kiniry, J.R., C.A. Jones, J.C. O'Toole, R. Blanchet, M. Cabeiguenne, and D.A. Spanel.
1989. Radiation use efficiency in biomass accumulation prior to grain filling in
five grain-crop species. Field Crops Research 20: 51-64.
Koornneef, M., C. Alonso-Blanco, A.J.M. Peeters, and W. Soppe. 1998. Genetic control
of flowering time in Arabidopsis. Annual Review of Plant Physiology and Plant
Molecular Biology 49: 345-370.
Lafitte, H.R. and G.O. Edmeades. 1997. Temperature effects on radiation use and
biomass partitioning in diverse tropical maize cultivars. Field Crops Research 49:
Langridge, J. 1963. Biochemical aspects of temperature response. Annual Review of
P'.vn Ph ,',),,,! 11: 441-462.
Lewin, B. 2000. Genes VII. New York: Oxford University Press.
Loomis, R.S. and D.J. Connor. 1992. Crop ecology: Productivity and management in
agricultural systems. Cambridge, UK: Cambridge University Press.
Lucas, R.E. and B.D. Knezek. 1972. Climatic and soil conditions promoting
micronutrient deficiencies in plants. In J.J. Mortvedt, P.M. Giordano and W.L.
Lindsay (eds.), Micronutrients in agriculture. Madison: SSSA. Pp. 265-288.
Mackay, A. D. and S. A. Barber. 1984. Soil temperature effects on root growth and
phosphorus uptake by corn. SoilSci. Soc. Am. J 48: 818-823.
Marschner, H. 1986. Mineral nutrition of higher plants. London: Academic Press.
Midmore, D.J., P.M. Cartwright, and R.A. Fischer. 1982. Wheat in tropical
environments. 1. Phasic development and spike size. Field Crops Research 5:
Midmore, D.J., P.M. Cartwright, and R.A. Fischer. 1984. Wheat in tropical
environments. II. Crop growth and grain yield. Field Crops Research 8: 207-227.
Mittler, R. and B.A. Zilinskas. 1994. Regulation of pea cytosolic ascorbate
peroxidase and other antioxidant enzymes during the progression of drought
stress and following recovery from drought. Plant. Journal 5: 397-405.
Monteith, J.L. and M.H. Unsworth. 1990. Principles of Environmental Physics.
London: Edward Arnold.
Muchow, R.C., T.R. Sinclair, and J.M. Bennett. 1990. Temperature and solar radiation
effects on potential maize yield across locations. Agron. J. 82: 338-343.
Nakamoto, H. and T. Hiyama. 1999. Heat-shock proteins and temperature stress. In
M. Pessarakli (ed.), Handbook of Plant and Crop Stress. New York: Marcel
Dekker, Inc. Pp. 399-416.
Nobel, P.S. 1991. Physiochemical and Environmental Plant Physiology. San Diego:
Raison, J.K, J.A. Berry, P.A. Armond, and CS. Poke. 1980. Membrane properties in
relation to the adaptation of plants to temperature stress. In N.C. Turner and
P.J. Kramer (eds.), Adaptation of plants to water and high temperature stress.
New York: John Wiley & Sons. Pp. 261-273.
Rawson, H.M., P.A. Gardner, and M.J. Long. 1987. Sources of variation in specific
leaf area in wheat grown at high temperature. Australian Journal of Plant
Physiology 14: 287-298.
Reynolds, M.P., 1994. Summary of data from the Ist and 2nd International Heat
Stress Genotype Experiment. In D.A. Saunders and G.P. Hettel (eds.), Wheat in
Heat Stressed Environments: Irrigated, Dry Areas and Rice Farming Systems.
Proceedings of the International Conference, Wheat in Hot, Dry, Irrigated
Environments, WadMedani, Sudan, 1-4 February, 1993. MNxico, D.E: CIMMYT.
Reynolds, M.P., M.I. Delgado, M. Gutierrez-Rodriguez, and A. Larque-Saavedra.
2000. Photosynthesis of wheat in a warm, irrigated environment. I. Genetic
diversity and crop productivity. Field Crops Research. 66: 37-50
Reynolds, M.P., R.P. Singh, A. Ibrahim, 0.A. Ageeb, A. Larqub-Saavedra, and J.S.
Quick. 1998. Evaluating physiological traits to compliment empirical selection
for wheat in warm environments. Euphytica 100: 85-94.
Reynolds, M.P., R. Trethowan, K.D. Sayre, and J. Crossa. 2002. Physiological factors
influencing genotype by environment interactions in wheat. Field Crops
Research 75: 139-160.
Roberts, J.K., N.A. DeSimone, W.L. Lingle, and L. Dure Ill. 1993. Cellular
concentrations and uniformity of cell-type accumulation of two Lea proteins in
cotton embryos. Plant Cell : 769-780.
Rosenzweig, C. and EN. Tubiello. 1996. Effects of changes in minimum and
maximum temperature on wheat yields in the central US: A simulation study.
Agric. & For. Meteor. 80: 215-230.
Sayed, H., M.J. Emes, M.J. Earnshaw, and R.D. Butler. 1989. Photosynthetic
responses of different varieties of wheat to high temperature. J. Exp. Bot. 40:
Sheldon, C.C, J.E. Burn,, P.P. Perez, J. Metzger, J.A. Edwards, W.J. Peacock, and E.S.
Dennis. 1999. The FLF MADS box gene: A repressor of flowering in Arabidopsis
regulated by vernalization and methylation. Plant Cell 11: 445-458.
Sinclair, T.R. 1994. Limits to crop yield? In K.J. Boote, J.M. Bennett, T.R. Sinclair and
G.M. Paulsen (eds.), Physiology and Determinations of Crop Yield. Madison:
Am. Soc. Agron. Pp. 509-532.
Sinclair, T.R. and R.C. Muchow. 1999. Radiation use efficiency. Advances in
Agronomy 65: 215-265.
Slafer, G.A. and H.M. Rawson. 1994. Sensitivity if wheat phasic development to major
environmental factors: A re-examination of some assumptions made by
physiologists and modelers. Australian Journal of Plant Physiology 21: 393-426.
Slafer, G.A. and H.M. Rawson. 1995. Photoperiod x temperature interactions in
contrasting wheat genotypes: Time to heading and final leaf number. Field
Crop Research 44: 73-83.
Smith, B.N., A.R. Jones, L.D. Hansen, and R.S. Criddle. 1999. Growth, respiration rate,
and efficiency responses to temperature. In M. Pessarakli (ed.), Handbook of
Plant and Crop Stress. New York: Marcel Dekker, Inc. Pp. 417-440.
Thomashow, M.F 1999. Plant cold-acclimation: freezing tolerance genes and
regulatory mechanisms. Annual Review of Plant Physiology and Plant
Molecular Biology 50: 571-600.
Vargas, M., J. Crossa, K. Sayre, M. Reynolds, M. Ramirez, and M. Talbot. 1998.
Interpreting genotype by environment interaction in wheat by partial least
squares regression. Crop Science 38: 679-689.
Vierling, E. 1991. The role of heat shock proteins in plants. Annual Review of Plant
Physiology and Plant Molecular Biology 42: 579-620.
von Dassow, G., E. Meir, E. Munro, and G.M. Odell. 2000. The segment polarity
network is a robust developmental module. Nature 406: 188-192.
Wang, E. and T. Engel. 1998. Simulation of phonological development of wheat
crops. Agricultural Systems 58: 1-24.
Wallace, D.H., P.A. Gniffke, P.N. Masaya, and R.W. Zobel. 1991. Photoperiod,
temperature and genotype interaction effects on days and nodes required for
flower of bean. J. Amer. Soc. Hart. Sci. 116: 534-543.
Wallin, E., and G. von Heijne. 1998. Genome-wide analysis of integral membrane
proteins from eubacterial, archaean, and eukaryotic organisms. Protein Sci. 7:
Wardlaw, I.E and C.W. Wrigley. 1994. Heat tolerance in temperate cereals: An
overview. Australian Journal of Plant Physiology 21: 695-703.
Waters, E.R., G.J. Lee, and E. Vierling. 1996. Evolution, structure and function of the
small heat shock proteins in plants. J. Exp. Bat. 47: 325-338.
Wheeler, T.R., G.R. Batts, R.H. Ellis, P. Hadley, and J.I.L. Morison. 1996. Growth and
yield of winter wheat (Triticum aestivum) crops in response to C02 and
temperature. Journ. of Agric. Sci. 127: 37-48.
White, J.W., J. Kornegay, and C. Cajiao. 1996. Inheritance of temperature sensitivity
of the photoperiod response in common bean (Phaseolus vulgaris L.).
Euphytica 91: 5-8.
White, J.W., P. Grace, P.N. Fox, A. Rodriguez-Aguilar, and J. Corbett. 2001. Modeling
temperature-driven variation in wheat and maize production at a global scale.
In Proceedings of the SAAD3 Symposium, Universidad NacionalAgraria La
Molina, Nov., 1999. International Potato Center (CIP). http
Wiegand, C.L., and J.A. Cuellar. 1981. Duration of grain filling and kernel weight of
wheat as affected by temperature. Crop Science 21: 95-101.
Ying, J., E.A. Lee, and M. Tollenaar. 2000. Response of maize leaf photosynthesis to
low temperature during the grain-filling period. Field Crops Research 68: 87-96.
Ying, J., E.A. Lee, and M. Tollenaar. 2000. Response of leaf photosynthesis during
grain-filling period to duration of void exposure, acclimation, and incident
PPFD. Crop Science 42: 1164-1172.
Re-examining Current Questions of Wheat
Leaf Appearance and Temperature
Gregory S. McMaster1 and L.A. Hunt2
1 USDA-ARS, Great Plains Systems Research, Fort Collins, Colorado, USA
2 Department of Plant Agriculture, University of Guelph, Guelph, Ontario, Canada
The rate of leaf appearance, or the phyllochron, is critical in simulating canopy development, structure, and
dynamics. Manyfactors can influence the phyllochron of wheat (Triticum aestivum L.), but the most
important factor is temperature. This paper explores some current questions about the complicated
relationship between wheat leaf appearance and temperature. :- .. :... .', the questions of whether the
phyllochron is linearly related to temperature and where the site of temperature perception is located are
addressed. While the temperature response across the entire temperature range is clearly non-linear, the
existence of a linear '. ; has caused some confusion. Another confounding factor is that both during and
across days, temperatures when leaves are produced are normally within the linear but often can
fluctuate beyond the linear phase. In addition, cultivars can vary slightly in their temperature response, and
temperature can interact with other factors. While measuring the i ,, ". ...... response to temperature may be
largely a physics problem, it is complicated by the question of where to measure the temperature since the
meristematic region covers a -. ''/ ; distance and temperatures in other parts of the plant ''' .
secondary factors that influence the phyllochron.
Many factors influence the rate of wheat leaf
appearance, or .l, ll ..11, 1. including light (. t .l1i.,
quantity, and photoperiod), water, nutrient I ,i 1-.1, i'
IJ;,,., and CO2 (McMaster 1997 and Wilhelm and
McMaster 1995 provide reviews). However,
temperature is always cited as the primary factor
controlling the phyllochron. Unfortunately, all these
factors interact in a very complex manner (especially
light, water, and salinity), and it is difficult to isolate
temperature response from other factors.
A great deal of work has examined the relationship
between temperature and the phyllochron, particularly
in the last four decades. Some work raises new
questions, and other times .p, t..11 arise because
earlier work has been "forgotten" while some of us
pursue the "new" questions. Queries such as, Is the
response to temperature linear?" and, "Where should
we measure temperature (e.g., air temperature above
the .o1, vs. soil temperature at crown .,. Itil are
examples of :1'.. I.. ,- receiving much attention and
I:, .'i Li.,; confusion in the last decade.
Leaf appearance is a critical process involved in canopy
development, structure, and dynamics (Wilhelm and
McMaster 1995). This paper explores some aspects of
the complicated relationship between wheat leaf
appearance and temperature.
Although different definitions of plastochron and
1.1,i.. 1.. i .1 /le been used across time (Wilhelm and
McMaster 1995), we define these terms as normally used
today. The plastochron is the interval of time between
appearance of successive leaf primordia on a shoot apex;
the phyllochron is the interval of time between
appearance of successive leaves through the whorl of
the subtending leaf on a shoot. The inverse of the
1pl, 11.. .11.n is also referred to as the leaf emergence or
appearance rate. The phyllochron is therefore the result
of the rate of leaf primordia initiation (i.e., the
plastochron) and subsequent development and growth
of the leaf primordium. One consideration of the
phyllochron is that each leaf must grow 1l .... 1,
successively greater sheath lengths before emerging.
The plastochron is p. -"P11 11.. composed of the normal
-i.. ih process during which cells first divide at the
meristem (in this case the shoot apex) and then enlarge
to form the leaf primordium. The growth process of
the leaf primordium then proceeds by further cell
division and expansion (Dale 1988). The region where
this occurs is commonly referred to as the leaf
extension zone (Skinner and Nelson 1995) and
certainly can be considered a meristem. 1,,l1..n 1i
there is some debate whether the leaf extension zone is
a separate meristem from the shoot apex, for purposes
of this paper it is important to I .. .. ,i only that cell
division and expansion are occurring across a certain
distance within the shoot. The plastochron is therefore
the result of one meristem (the shoot apex) covering a
limited distance, whereas the phyllochron is the result
of two meristems (the shoot apex and the leaf
extension zone) covering a much greater distance.
There are significantly more cells in the mature leaf
(blade plus sheath) than in the leaf primordium.
Therefore, we suggest some reasons that the
plastochron so often seems to be even more influenced
by temperature than the phyllochron:
* The cell life cycle is very much controlled by
temperature (Arkebauer and Norman 1995)
* Cell expansion rates are not only influenced by
temperature, but are also dependent on resources
such as water for turgor pressure, and nutrients and
carbohydrates for 1.I,, ;1,li. the cell
* Given that there are so many more cells f.. 1in4. the
mature leaf, considerably more resources are needed
for cell expansion, allowing other factors to interact
with temperature effects
Resources needed for cell division and expansion
(water, nutrients, and carbohydrates) originate in areas
outside of the shoot apex. Uptake or production of
these resources in the root and canopy systems are
influenced by temperature, but the temperature varies
throughout these systems and may differ substantially
from the shoot apex temperature.
Is the Temperature Relationship
The question, "Is the temperature relationship linear?"
became important in the last few decades once
researchers started ..-i al- t;f ;i -, the phyllochron using
thermal units to represent time rather than days. Based
solely on the many temperature response studies in
growth chambers, there is no debate-the temperature
response is not linear across the entire temperature
range within which wheat can grow. Two studies
strongly ill I It. this.
In a classic study, Friend et al. (1' '.i studied the main
stems of a spring wheat (cv. Marquis) under controlled
environments. Constant day/night temperatures of
100, 150, 200, 250, and 300C were used in combination
with varying light intensities and duration. Some of the
important results were that:
* Temperature interacted with I .I, .), i ; ,l and light
* Within constant conditions, the phyllochron within
the main stem was linear, so that age or p. ,-;I t.. of
leaf does not matter .. i,.. the ontogenetic decline
hypothesis postulated by some)
* The phyllochron was "linear" from 10( to 25C, and
decreased -I:._ -iti at 300C (Figure 1)
Twenty-seven years later, Cao and Moss (1989) revisited
this question for eight winter wheat and barley varieties.
In controlled environments of constant day/night
temperatures of 7.5, 10P, 12.5, 15, 17.50, 20, 22.5, and
LER vs. temperature
5 10 15 20
25 30 35
Figure 1. Leaf emergence rates adapted from two studies.
Note: Actual values are approximations from graphs.
...e... Friend et l. (1062)
-0- Coo and Moss (1989)
250C with varying I1, .!'. ;. the appearance rate of
the first four leaves on main stems were measured. The
important results to note for this paper were that there
were hit variations among cultivars in response, that
temperature and photoperiod interacted, and that the
phyllochron was "linear" from about 100 to 17.50C,
either reaching an asymptote or decreasing slightly after
about 22.50C, F,. ,,, 1).
The third result in both studies is critical because it clearly
shows that the response is not linear across the entire
temperature range, just a certain -*. 1..i of it. Further-
more, we should expect -I_1ltI variation in temperature
responses among cultivars but the pattern of response is
generally the same for most, if not all, cultivars.
Why is there still a debate on whether the relationship
between the phyllochron and temperature is linear? We
previously alluded to part of the reason-how we
represent temperature as thermal time. The most
common form of calculating daily growing degree-days
(GDD) r i'. r 1. t, i and Wilhelm 1997), is adding the
daily maximum and minimum temperature and
dividing by two to get the daily average temperature.
The base temperature, below which the process does
not occur, is typically set to 0C (McMaster and Smika
1988; McMaster 1997) and is subtracted from the daily
average temperature. Certainly, if the base temperature
is set higher, then non-linearity at lower temperatures
will tend to be negated. If an upper temperature
threshold is used, then, mathematically, this would tend
to mimic a slight depression found at higher
Another reason is that when temperatures are
considered either diurnally or across days, the
temperatures are often in the "linear" portion of the
temperature response or "equally" above and below
these temperatures when leaves are being produced.
We have calculated long-term monthly maximum,
minimum, and average daily air temperatures for a
number of wheat production locations throughout the
world, and generally the results are the same. In Figure
2, Fort Collins, Colorado is used to illustrate the results
of expected temperatures during leaf appearance.
September i.- .. October (Oct), -I. ,I (Apr), and May
are the months when most leaves appear. The monthly
maximum, minimum, and average temperatures for
those months are as follow:
* Sep: 23.9 (max), 7.20 (min), and 15.50 (avg) C
* Oct: 17.9 (max), 1.20 (min), and 9.50 (avg)C
* Apr: 15. i 0.40 (min), and 8.00 (avg) C
* May: 20.40 (max), 5.60 (min), and 13.00 (avg) C
These temperatures fall within (or are close) to the
linear region of temperature response reported by
Friend et al. (1962) and Cao and Moss (1989).
Measurement interval can also affect the perceived
influence of temperature. The longer the interval for
measuring the phyllochron, the greater the chance that
temperatures will primarily be in the linear region and
the influence of temperatures in the non-linear region
will be reduced. Unless either the temperatures are in the
non-linear,, 'i : for a I, .,f" i i percent of the
measuring interval or are extreme, or both, the diurnal
and daily fluctuations in temperature will result in the
, 1 1. 11. -. 1..1- .i' ,1 .1 i',. linear in field conditions.
What are the In .1 .It ...- for modeling the phyllochron
response to temperature? First, for most field conditions
the simple GDD model of assuming a linear response to
temperature works W. ; ;,: .1, well in predicting a
complex process. This is not because it is theoretically
correct, but because of some of the reasons listed above.
One caveat is that if the cultivar being simulated is
unusual in its temperature response pattern (e.g.,
shorter linear phase, linear phase not correlated with
field temperatures, response to high temperatures
differs, etc.), the accuracy will then be reduced. If
greater accuracy is desired, non-linear approaches for
calculating the GDD look promising (e.g., Yan and Hunt
Where to Measure Temperature?
When the issue of whether the phyllochron was linear
with temperature as measured using GDD surfaced
around 25 years ago, temperature was measured using
J -- Tavg
"i", -----T- max
20 \ ---- Tini /
7\ -- Firstleaf
S\ ----- Flag leaf /
A S 0
N D J F M
A M J J
Figure 2. Expected temperatures during leaf appearance, Fort Collins,
Colorado (N 40:34, W 105:05, elev. 236 m).
air temperature above the canopy. Yet we know from a
long history of controlled environments and field root/
shoot temperature studies that soil temperature might
be a better approximation of shoot apex temperature
than air temperature, at least when the shoot apex is
below the soil surface. As situations 1., .,I to arise
where the linear GDD model using air temperatures did
not predict the phyllochron as well as desired, attention
focused on shifting measurements to soil temperature at
crown depth. Unfortunately, as with all good
paradigms, lt .%Il.h-,, anomalies began to surface when
using soil, rather than air, temperature. Let us mention a
couple of these anomalies and then propose some
adjustments to our paradigm.
The first anomaly was reported in 1998 (McMaster and
\'., l. n. :Il liugh this examined I-l. 11I. rather than
the phyllochron, the results are applicable because again
it is the shoot apex where many of the events occur.
Examining 23 site-years across seven locations in the
Central Great Plains with varying management practices
and ., i, n it was found that using soil temperature
provided no, or negligible, improvement over using air
temperature in p .1;, 1;'.. .I .. 1 :., 1. l stages.
Similar results were found, but not reported, by Betty
j i '! r and Ron Rickman in Pendleton, Oregon. One of
the main reasons for this is that mean soil temperatures at
3 cm were very similar to mean air temperatures above
the canopy (although the amplitude did vary i,.. ~.i, ,
and clearly the relationship between air and soil
temperature was the same.
The other anomaly was found in an unpublished
experiment completed in 2000. This experiment was an
. .i i ,... ~,1; of the GCTE Wheat Network Meeting in
Maricopa, Arizona in May 1997. At that time, a number
of modelers (John Porter, Wally Wilhelm, Pete
Jamieson, Joe Ritchie, and Michael Kirby, among
others) designed an experiment conducted in the field
at Fort Collins, Colorado. A p;... wheat (cv. Nordic)
was planted at three planting dates (mid-March, mid-
April, and mid-May) for two years. A complete
randomized block design with four : -.1 ..tr ..- was
used for two treatments. One treatment was ambient
soil temperature; the second treatment raised the soil
temperature at 2 cm depth (presumably crown depth)
30C above the ambient soil temperature. This was done
by means of heat tape located 3 cm below the seed,
which was planted at 2 cm. As expected, raising the soil
temperature resulted in earlier seedling emergence by
as much as a week; however, the rate of leaf
appearance was not affected by increasing the soil
temperature as expected (Fig. 3). This was true for all
planting dates in both years. The phenology
experiment anomaly was largely explained because of
the similarity of soil and air temperature, but this could
not be the explanation in this experiment because there
was a ;, .ni; .I difference in soil temperatures that
should have been reflected in differences in the
phyllochron. How can this anomaly be explained and
our p: a:l;;- i J saved?
Two ..-L.L.tr ..- may play an important role. The first is
that if a non-linear beta response function is used to
calculate GDD (Yan and Hunt 1999; Hunt et al., p. 23, this
volume), then differences in the phyllochron 1- i,, ri,.-,
between the two soil temperature treatments are largely
accounted for. The second explanation returns to the issue
of the site of temperature perception and some aspects of
wheat anatomy and development discussed previously.
1998 MS Haun
80 100 120 140 160 180
Day of year
1999 MS Haun
80 100 120 140
Day of year
Figure 3. Main stem Haun values (Haun 1973) for three planting dates
in two years. The two treatments were heating the soil at 2cm depth
to +3C (+3) over ambient soil temperature (+0).
Two meristems result in the phyllochron: the shoot
apex producing the leaf primordium and the leaf
extension zone where the leaf primordium develops
and grows the mature leaf. Until the ,, .nt. Ldc stage of
single ridge, the length of the shoot apex is roughly 1
cm, with a primordium 1, ,.'il, of about 1.5 mm
reported for tall fescue (Festuca arundinacea Schreber)
(Skinner and Nelson 1995). A further complication is
the need to know the depth at which the crown is
located. Cultivars vary in their crown depth, and
within a cultivar the crown depth can vary
considerably I, .. ..- i- i to many factors. A
temperature gradient normally exists with respect to
depth from the surface. Therefore, the first problem is
knowing at which "" "at or depth to measure soil
Another complication in ..1. :;.:1; I:_ the appropriate
depth at which to measure soil temperature is the
importance of the leaf extension zone. Skinner and
Nelson (1995) delineate the leaf extension zone into
different sections, such as cell division, cell expansion,
secondary cell wall growth, carbohydrate and N
deposition, etc. for tall fescue. These sections overlap
to varying degrees, and the total distance can be as
much as 70 mm. Because expansion of the cells pushes
the older leaf tissue up IIr .. 1, the whorl of
subtending leaves, we presume that much of this zone
occurs more or less vertically from the leaf
primordium. Depending on the depth of the crown/
shoot apex and length of the leaf extension zone, some
of the leaf growth could be occurring above the soil
surface, and certainly across a temperature gradient in
the soil. Again, what point is to be used to measure
soil temperature? The final confounding factor
recognizes that the water, carbohydrates, and
nutrients required for cell division and ,;I... i;- are
coming from other parts of the plant, and that these
parts are experiencing different temperatures than the
shoot apex/leaf extension zone.
Simulating the complicated relationship between the
phyllochron and temperature of wheat can be aided
by understanding the location of meristematic tissue
i..'1,1.. ;Iq. leaves, the temperature response pattern
(non-linear across the entire range, but linear within
certain temperatures), temperatures during leaf
production and whether the temperature is in the
linear region or not, factors that interact with
temperature to influence the phyllochron, and certain
limitations of the GDD characterization of the effect of
temperature. The simple linear GDD approach that
uses air temperature above the :-- ,I -,: works well in
most situations. If improvements are desired, the use
of non-linear responses shows promise.
Arkebauer, T.J. and J.M. Norman. 1995. From cell growth to leaf growth: I. Coupling
cell division and cell expansion. Agronomy Journal87: 99-105.
Cao, W. and D.N. Moss. 1989. Temperature effect on leaf emergence and
phyllochron in wheat and barley. Crop Science 29:1018-1021.
Dale, J.E. 1988. The control of leaf expansion. Annual Review of Plant Physiology
and Plant Molecular Biology39: 267-295.
Friend, D.J.C., V.A. Helson, and J.E. Fisher. 1962. Leaf growth in Marquis wheat, as
regulated by temperature, light intensity, and daylength. Canadian Journal of
Haun, J.R. 1973. Visual quantification of wheat development. Agronomy Journal
McMaster, G.S. 1997. Phenology, development, and growth of the wheat (Triticum
aestivum L) shoot apex: A review. Advances in Agronomy 59: 63-118.
McMaster, G.S. and D.E. Smika. 1988. Estimation and evaluation of winter wheat
phenology in the central Great Plains. Agricultural and Forest Meteorology 43:
McMaster, G.S. and W.W. Wilhelm. 1997. Growing degree-days: One equation, two
interpretations. Agricultural and Forest Meteorology 87: 289-298.
McMaster, G.S. and W.W. Wilhelm. 1998. Is soil temperature better than air
temperature for predicting winter wheat phenology? Agronomy Journal 90:
Skinner, R.H. and .J. Nelson. 1995. Elongation of the grass leaf and its relationship
to the phyllochron. Crop Science 35: 4-10.
Wilhelm, W.W. and G.S. McMaster. 1995. The importance of the phyllochron in
studying the development of grasses. Crop Science 35: 1-3.
Yan, W. and L.A. Hunt. 1999. An equation for modelling the temperature response
of plants using only the cardinal temperatures. Annals of Botany 84: 607-614.
Simulating Response to Temperature
L.A. Hunti, W. Yan1, and Gregory S. McMaster2
SDepartment of Plant Agriculture, University of Guelph, Guelph, Ontario, Canada
2 USDA-ARS, Great Plains Systems Research, Fort ( I Colorado, USA
Temperature is one of the most important factors determining plant growth, development, and yield. Accurate
summarization of plant temperature response is thus a prerequisite to successful modeling of crop systems and
application of models to management. This paper reviews various equations that have been used to describe
temperature response for a number of :. !' "'"' The beta function, as used in some recent analyses, has been
shown to summarize data dealing with the overall :. .I. development of maize and wheat in a realistic manner.
However, consideration of r -.. .. *;. ;. photosynthesis indicates that the function may not be appropriate
for all processes, .. .:: ; .... (such as photosynthesis) that involve many sub-processes, each with its own
response characteristics. Further, ,. ... : parameters may be necessary for ,:' ... .: genotypes, i~: times
during a plant's life cycle or geographic, .'. Careful consideration of all such aspects will be necessary
for accurate simulation over '.. r. .: regions or over contrasting climate change scenarios. An approach involving
interpolation between data points describing specific temperature responses, rather than a mathematical function,
may well have the widest utility. Such an interpolation should be non-linear.
Temperature is arguably the most important
environmental factor that affects plant development,
growth, and yield. All biological processes respond to
temperature, and all responses can be summarized in
terms of three cardinal temperatures: a base or
minimum (T,,in), an optimum (Topt), and a maximum
(Tmax). However, the nature of the response to
temperature between these cardinal points, which is
important for calculating the i1 I. .....,, adaptation,
and yield of various crops (Wang 1960; Cross and
Zuber 1972; Undersander and Christiansen 1988;
Shaykewich 1995), is not summarized as easily. There
have been, however, numerous attempts to develop
functions J,, ;l., response to temperature, some
with more general application than others. Some of
these attempts are summarized and illustrated here.
Temperature Response Functions
Within a limited range of temperature, the rate of
plant development or growth is often found to be a
linear function of the temperature. In this range, the
time required to develop to a certain stage is related to
the sum of daily temperatures above a specified base
or minimum temperature. Such a linear model is
convenient and effective when the temperature does
not '*r'. :! h or exceed the optimum, Topt
(Summerfield and Roberts 1987). However,
temperatures frequently approach and exceed the Topt
in natural conditions. To accommodate this situation,
many researchers have adopted a bilinear approach
(e.g., Olsen, McMahon and Hammer 1993) in which
two different linear equations (Fip,, i. a., 1) are used to
describe the responses to sub-optimum and supra-
' 1II ...... temperatures. This approach has been
successfully applied to several crops (OiC, ",'_ et al.
1995 and 1996 for pigeonpea; Craufurd et al. 1998 for
sorghum, among others).
r= al+ biT (T
ar= a+ b2T (11' opt)
There are four parameters in the bilinear approach, a1,
bly a2, and b2, from which the three cardinal
temperatures can be derived. However, the derivations
may not always be meaningful. As Craufurd et al.
(I .-. point out, "The estimation of Tin usually
requires considerable extrapolation and the standard
error (SE) of this Tmn is large in comparison with the SE
of Topt." The same can be said for the estimation of
Tmax. In the work of Craufurd et al. (1998), estimates of
Tmax ranged from 36.80 to 58.90C for leaf appearance
rate; for leaf tip appearance rate a value of 1980C was
obtained for one genotype, obviously an over-
estimation. Further, the maximum rate of any process
at Topt is also likely to be over-estimated since it is
obtained from two linear equations, while the real
response curve is generally curvilinear.
A multilinear model can be constructed from three or
more linear components (e.g., Coelho and Dale 1980)
making it less rigid than the bilinear model. Some
crop system simulation packages (e.g., CROPSIM,
Hunt and Pararajasingham 1995 and later versions)
have adopted this approach, even though five or
more parameters are required to describe the
temperature response of a process. Although closer
to reality than linear or bilinear models, the greater
number of parameters renders this approach subject
to calibration errors. Moreover, the parameters are
usually high empirical.
Rather than rigid combinations of linear equations,
some researchers have argued that it is preferable to
use an approach in which a smooth curve describes the
temperature response of a given process (e.g., Cross
and Zuber 1972, Shaykewich 1995). Exponential,
logistic, and polynomial equations give smooth curves.
Of these, an exponential equation is usually effective in
describing the responses at low to intermediate
temperatures. It does not describe the response to high
temperatures, however, because it does not allow for a
reduced rate of development at high temperatures (e.g.,
the curve in Tollenaar et al. 1979).
By contrast, a three parameter logistic equation allows
for very slow activity at high temperatures, and has been
used by Shakewich (1994) to summarize leaf appearance
rate data obtained by Tollenaar et al. (1979) for maize:
LARmax/(1 + ea+bT) Eq. 2
Leaf appearance rate
Leaf appearance rate, maximum
This equation gave an r2 = 0.94 and an inflection point
A three-parameter quadratic equation (Yan and
Wallace 1996, 1998) goes one step further than the
logistic by allowing for a reduced rate of development
at high temperatures:
r= Rmax b(T- Topt)2
This form of expression has been used in the heat unit
system (Brown 1975) for maize (Zea mays L.) with a
Tmin of 100 and a Topt of 300C for the daytime part.
Because the true temperature response is rarely a
symmetric parabola, however, estimation of Tmin and
Tmax can be difficult. Applications of the quadratic
model at low or very high temperatures can be
inaccurate as a result. Higher order polynomials can
produce a more realistic temperature response curve
(Tollenaar et al. 1979) and improve prediction of crop
development (Stewart et al. 1998). In addition to
requiring more parameters, however, higher order
polynomials also have parameters that are difficult to
interpret in biological terms.
The standard "beta-distribution," described in many
handbooks of mathematics and characterized by a
unimodal curvilinear response to an independent
variable x in the range of [0,1], also has been used to
describe temperature response in plants (Yin et al.
T m Tax
RfT-T T x-T "T cP
Equation 4 fits experimental data to five parameters:
the three cardinal temperatures, the maximum rate
Rmax at Topt, and c, a parameter that determines the
shape of curve. Yin and colleagues (Yin and Kropf
1996; Yin et al. 1996) reported successful simulation of
rice development using this equation.
Compared with previous models, the expression by
Yin and colleagues has the advantage of producing
smooth, and for some processes, realistic curves. All
parameters except c are biologically meaningful. Yan
and Hunt (1999) simplified the equation by
eliminating c and placing Tmin equal to zero, and
reported on the effectiveness of this simplified
equation in summarizing published temperature
response data for the growth and development of a
number of species. The simplified expression,
however, essentially replaces the'shape' parameter of
the basic beta function (i.e., c) with the maximum
temperature. This replacement may produce realistic
curves in some circumstances, but certainly not for all
processes and situations.
The problem of effectively .1. hi..: the initial phase
was overcome by Tollenaar et al. (1979) by making use
of a polynomial (Figure 2). Such an approach was also
found useful by Stewart et al. (1998) for summarizing
the development of field grown maize during the
.i ,kcii to silking period i,. i n 3).
Examples of Temperature
A multilinear function fitted to data from various maize
experiments is shown in Figure 1. This function, which
had four segments and five parameters (60, 210, 280,
320, and 440C), could be used to compute the
effectiveness of any ,.. i, i day for the : i. ti and
development of maize. However, the sharp 'shoulders'
may limit the value of the approach particularly for the
first 'leg' of the response curve. This section of the
curve is part of a generalized logistic type, with activity
increasing slowly with temperatures at values just
above the base and accelerating until an apparently
linear phase is reached.
0 10 20 30
Figure 1. Growth rate of corn at indicated temperatures divided by that
at 28-32C. Figure reworked from Coelho and Dale (1980).
Lines drawn from the following relationships:
GR(N)=0.027T -0.16 6'C T<21C
..:'.li ;:, ,..:, 1.41 21'C T<28C
GR(N)= -0.083T + 3.67 32C
..:l' 6C>T< 44'C (14)
10-cm-depth soil temperature used from planting to the date on which the growing
point emerged from the soil, and air temperature thereafter.
Note: Actual e ,J, p u 'p-- i.. ... .... ,y i
5 10 15 20 25 30 35
Figure 2. Relationship between rate of leaf appearance and temperature
for corn. After Tollenaar et al. (1979).
Y= 0.2834- 0.0639T+ 0.0491T2-::::":::::::';' i
Curve B: alternate polynomial
Y = 0.0097 0.0360T + 0.00362T2- 0.0000639T3
Crosses indicate mean rate of leaf appearance at five constant temperatures
(standard deviation = 0.043 leaves/day).
0 5 10 15 20 25 30 35
Mean daily temperature
Figure 3. Thermal time calculated from field data for four maize hybrid
groups over the planting to silking period. After Stewart et al. (1998). There
were no significant differences in the optimum temperature between groups, which had
growing degree day ratings from 1100 to 1400'Cd.
However, it proved less useful from silking to
maturity (Figure 4).
Angus et al. (1'-' .: ) found that an ..., .tl function
(Figure 5) without an initial 'leg' best summarized
data from wheat experiments over a ,, ..., of widely
contrasting sites. This example emphasizes the need
for a general function that can accommodate
situations where 'legs' of different durations are
clearly present, ". ti, i with those in which there is
no apparent'leg' or only a very limited phase of
.i.,'11:. increasing rate.
The iml'i,,. .i'beta' function of Yan and Hunt (1' ,
well summarized maize data .r- ...... 6), though the
upper temperature serves as a'shape' parameter in
this -- i,,l;fifJ.1 form (Figure 7). The more general
function of Yin et al. (1'- -,1 seems in turn likely to
have more general 1.1.i,, kn However, even for the
general function, the 'shape' parameter affects both
the ascending and descending 1.... of the curve; it thus
may not allow for effective description of the impact
of high temperatures, ,in. Zin m. essential for
.1..l n .drl ,,, of models to climate change scenarios
involving an increase in temperature.
0 5 10 15 20
Mean Daily Temperature ('C)
0 5 10 15
20 25 30
35 40 45 50
Figure 4. Thermal time calculated from field data for four maize hybrid
groups over the silking to maturity period. After Stewart et al. (1998).
Figure 6. Measured relative rates of development or growth of maize,
together with predicted relative rates based on a single modified beta
function with T,, = 41 and Topt 31C. From Yan and Hunt (1999).
0 10 20 30 40 50
Figure 5. Exponential functions used to summarize development of
wheat cultivar UQ189 at a number of widely separated locations.
The temperature factor calculated from the following expression: R,,m [1-e-] (T-mini]
where Rmax is the maximum rate of development, and ] and Tmm are parameters.
Values for 3 were 0.15 and 0.077C-1 for emergence to anthesis and anthesis to
maturity phases respectively, and for Tmin 3.5 AND 8.9'C for the same periods. After
Angus et al. (1981).
0 10 20 30 40 50
Figure 7. Comparison of different temperature functions to illustrate the
effect of changes. Tm. on the slope of the low temperature response arm of
the modified beta curve of Yan and Hunt (1999).
--- ...... .:,t. ,: ,!,, ,.... I:. = 45.3, Tmin= 21.8) for leaf appearance (Yon and Hunt, 1999)
S ,,,..,, ii. ,, i,,,,.1..., ... = 33.4, Tmin= 1" i ... ,. ,, ., I,,,,,i ,,,, 1999)
Discussion and Conclusions
Temperature is among the most important single
factors determining plant growth and development
(and hence agricultural production), and a model that
allows for summarization (and hence simulation) of
the temperature response of plant growth and
development is necessary for several applications.
1 11.. ,. L. .1.. of the optimum and maximum
temperatures for the growth and development of a
genotype, and the nature of the response surface, is
vitally important to the successful 1"' d,, i,. ., of its
- \ i i adaptation, and yield in a particular
environment. A beta function appears to be useful for
summarizing the response surface for a number of
processes, but may not have utility for all of them.
This is especially true for those processes that may
have a wide optimum temperature range, as may be
found when the overall process is the resultant of a
number of sub-processes. Photosynthesis is one such
example. Early work (e.g., Murata and lyama 1963;
Figure 8) showed a wide optimum temperature range
for the process; however, more recent work has shown
that the nature of the response surface depends on the
conditions under which measurements were made
(e.g., Acock 1991; Figure 9) and that it may be ...- rl
Figure 8. Photosynthesis temperature response curves for wheat and
maize as reported by Murata and lyama (1963).
to describe overall temperature response surface by
using simple functions for various sub-processes
.i .... .' and P.1 .- 1, 1994; Bernacci et al. 2001). The
use of an approach in which an interpolation (possibly
non-linear) between specified points is used instead of
a 1" .' 'c mathematical formulation may thus have
the most widespread application. Such an approach,
which has been used widely in simulation iq. 1. 1;, .
can result in good simulation when the specified
points or cardinal temperatures are realistic (Figure 10).
Even with such an approach, however, the selection of
the cardinal points for use with a particular process is
not a simple matter. Many aspects must be considered,
including the following:
* The nature of the response surface (i.e., the presence
of a'leg' or the width of the -.pti;.,ai band)
* Cl, -ill n.. ,.i the life cycle
* Ecotype differences
* Acclimation possibilities (i.e., changes induced by
the conditions experienced d,..-;_. growth)
The problem of effectively defining plant temperature, as
contrasted to air or soil temperature, and the impact that
this may have on any apparent temperature response
(see Jamieson et al. 1995) must also be kept in mind.
Figure 9. Leaf net photosynthetic rates measured at various
temperatures and two light flux densities. After Acock (1991).
Tomato Plants grown at 20C and 380 [mol photon m-2s-' with a 16-h photoperiod,
and rates measured with 300 ul L-1 CO2. Points are measured data and curves are
predicted using a simple model.
* 1880Inol m-2s-1
I 188 inol m2s'
80 100 120 140
80 100 120 140
160 180 200 220
160 180 200 220
120 140 160 180 200 220 240
Cardinals: 0, 20, 30, 40
120 140 160 180 200 220 240
80 100 120 140 160 180 200 220 240 260
Day of year
120 140 160 180 200 220 240 260 280 300
Day of year
* Measured (Mc Master) Simulated
Figure 10. Simulated and measured leaf numbers for wheat cultivar 'Nordic' planted on two dates in the field at Fort Collins, Colorado.
Simulations were made with different temperature responses, as specified with cardinal temperature values (C). Too low an optimum resulted in slow simulated leaf
appearance; too high an optimum resulted in the opposite.
Acock, B. 1991. Modeling canopy photosynthetic response to carbon dioxide, light
interception, temperature, and leaf traits. In K.J. Boote and R.S. Loomis (eds),
Modeling crop photosynthesis From biochemistry to canopy. CSSA Special
Publication Number 19. Madison: American Soc. of Agronomy. Pp. 41-55.
Angus, J.F., D.H. MacKenzie, R. Morton, and C.A. Schafer. 1981. Phasic development
in field crops. II. Thermal and photoperiod responses of spring wheat. Field
Crops Research 4: 269-283.
Bernacchi, C.J., E.L Singsaas, C. Pimentel, A.R. Portis, Jr., and S.P. Long. 2001.
Improved temperature response functions for models of Rubisco-limited
photosynthesis. Plant, Cell & Environment 24: 253-259.
Boote, K.J. and N.B. Pickering. 1994. Modeling photosynthesis of row crop canopies.
oHrt. Science 29:1423-1434.
Brown, D.M. 1975. Heat units for maize in Southern Ontario. Fact sheet AGDEX 11/
31, Order No. 75-077. N.p.: Ontario Min. of Agric. and Food.
Coelho, D.T. and R.E Dale. 1980. An energy-crop growth variable and temperature
function for predicting maize growth and development: Planting to silking.
Agronomy Journal 72: 503-510.
Craufurd, P.Q., A. Qi, R.H. Ellis, R.J. Summerfield, E.H. Roberts, and V. Mahalakshmi.
1998. Effect of temperature on time to panicle initiation and leaf appearance
in sorghum. Crop Science 38: 942-947.
Cross, H.Z. and M.S. Zuber. 1972. Prediction of flowering dates in maize based on
different methods of estimating thermal units. Agronomy Journal 64: 351-355.
Hunt, LA. and S. Parajasinghm. 1995. CROPSIM-WHEAT: A model descibing the growth
and development of wheat. Canadian Journal of Plant Science 75: 619-632.
Jamieson, P.D., I.R. Brooking, J.R. Porter, and D.B. Wilson. 1995. Prediction of leaf
appearance in wheat: A question of temperature. Field Crop Research 41: 35-44.
Murata, Y., and J. lyama. 1963. Studies on the photosynthesis of forage crops. II.
Influence of air temperature upon the photosynthesis of some forage and
grain crops. Proceedings of the Crop 54". i ,le v i~ ) of Japan 31: 315-321.
Olsen, J.K., C.R. McMahon, and G.L. Hammer. 1993. Prediction of sweet maize
phenology in subtropical environments. Agronomy Journal 85: 410-415.
Omanga, P.A., R.J. Summerfield, and A. Qi. 1995. Flowering of pigeon pea (Cajanus
coaan) in Kenya: Responses of early-maturing genotypes to location and date
of sowing. Field Crops Research 41: 25-34.
Omanga, P.A., R.J. Summerfield, and A. Qi. 1996. Flowering of pigeon pea (Cajanus
coajn) in Kenya: Responses of medium and late maturing genotypes to
location and date of sowing. Experimental Agriculture 32: 111-128.
Shaykewich, C.E 1995. An appraisal of cereal crop phenology modeling. Canadian
Journal of Plant Science 75: 329-341.
Stewart, D.W., L.M. Dwyer, and L.L Carrigan. 1998. Phenological temperature
response of maize. Agronomy Journal90: 73-79.
Summerfield, R.J. and E.H. Roberts. 1987. Effects of illuminance on flowering in
long- and short- day grain legumes: A reappraisal and unifying model. In J.G.
Atherton (ed.), Manipulation of flowering. London: Butterworths.
Tollenaar, M., T.B. Daynard, and R.B. Hunter. 1979. Effect of temperature on rate of
leaf appearance and flowering date in maize. Crop Science 19:363-366.
Undersander, D.J. and S. Christiansen. 1986. Interactions of water variables and
growing degree days on heading phase of winter wheat. Agricultural and
Forest Meteorology 38: 169-180.
Yan, W. and LA. Hunt. 1999. An equation modeling the temperature response of
plant growth and development using only the cardinal temperatures. Annals of
Botany 84: 607-614.
Yan, W. and D.H. Wallace. 1996. A model of photoperiod x temperature interaction
effects on plant development. Critical Reviews of Plant Sciences 15: 63-96.
Yan, W. and D.H. Wallace. 1998. Simulation and prediction of plant phenology for
five crops based on photoperiod by temperature interaction. Annals of Botany
Yin, X., M.J. Kropff, G. McLaren, and R.M. Visperas. 1995. A nonlinear model for
crop development as a function of temperature. Agricultural and Forest
Yin, X., M.J. Kropff, and J. Goudriaan. 1996. Differential effects of day and night
temperature on development to flowering in rice. Annals of Botany 77: 203-213.
Yin, X., and M.J. Kropff. 1996. The effect of temperature on leaf appearance in
rice. Annals of Botany 77: 215-221.
Evaluation and Calibration of CERES3-Maize:
Tiller Number Simulation
A.S. Du Toit & M.A. Prinsloo
ARG Grain Crops Institute, Potchefstroom, RSA
The purpose of this study was to adapt CERES3 to simulate .*.:. .. .. in r:! : .. : :.. I among maize cultivars
without increasing the number of genetic ... .. ,- as ..'! as to simulate ::. of management and
environmental conditions on tiller number. The simulation of tiller numbers is of limited interest in maize
production areas where the crop is produced at plant populations greater than 3.0 plants/m2, but the ...:. ...... of
tillers to yield can be significant at lower populations. Cultivars show great variation in tiller potential, and this
potential varies with the planting date. Such variability is confirmed by field data from South Africa showing that
the standard deviation of the measured values of tiller numbers was greater than the mean. An. ,* ':;: for the
simulation of tiller numbers was developed using an average photoperiod and a minimum temperature for the time
period from emergence to the end of the juvenile stage. The results indicated that tiller numbers could be simulated
with some 1 of success, but there is need for further investigation. The period of plant development in which
tiller numbers are ... by environmental conditions is of particular interest. A /.'. ... ;: the influence ; .,:'
numbers on kernel numbers is simulated, the impact of tillering on leaf development is not incorporated into the
modified CERES. This topic also merits further investigation.
Tiller formation is common in the gramineae (Verwey
and Hammes 1-' -' and tillers often make important
contributions to grain and forage yields. The tillering
process is simulated in CERES3 routines for barley
(Otter-Nacke et al. 1991), millet (Singh et al. 1991),
... .r.ii (Alargarswamy and Ritchie 1991), and wheat
(Godwin et al. 1 -. However, maize is simulated as a
single plant in CERES Maize vl.0 (Jones and Kinry
1'- ..., CERES-Maize v2.1 F at bI.- et al. 1992), and
CERES3 (Hoogenboom et al. 1 "'-1 in -i i. of the
maize plant .-. J, in, tillers in plant I ~ li i o,. '
below 4.0 plant/m2 (Tetio-Kagho and Gardner 1 '
Verwey et al. (1994a) determined that more tillering
occurs at night temperatures of 5C than at either 10C
or 18'C during the first 17 days after emergence.
Results by Pretorius (1985) showed that i1,ll.r, 1 varies
with cultivars, I i ..iin, dates, and plant I'i Jl"d -II
Tetio-Kagho and Gardner (1 '- -' reported differences
in the amount of tillers over seasons, and Verwey et al.
(1 '-i. found that tiller numbers differed between
various sites within the same season. Du Toit (1 .;
showed that a contribution of 25% by tillers to total
grain yield is not uncommon, with the number of ears
per plant (main, secondary, and tiller) making a
significant contribution to the simulation error of the
standard CERES3 model.
The purpose of this study was to adapt CERES3 to
simulate differences in till.. r I p. lq. i;al of maize
cultivars without increasing the number of genetic
... .iirl ,.t- as well as to simulate tiller development for
different management and environmental conditions.
Materials and Methods
Two trials were conducted for the development of the
tiller algorithm. One trial involved different planting
dates at Potchefstroom, South Africa (lat. 26.73"S, long.
27.08'E, elevation 1345 m) in 1990-91. The other
pertained to the determination of genetic coefficients,
and was conducted in 1992-93. These two trials were
combined, providing 123 data points with 20 cultivars
and 21 planting dates over two seasons.
Three approaches were attempted to improve the
simulation of tiller numbers:
1. ,.1.'!'J the algorithm for tiller production used in
CERES3-Sorghum. The logic used in this subroutine
is I,, i ..- 1I later in this paper.
2. r1 I.. i i.j the CERES3-Sorghum ,l' d.... to include
effects of environmental factors that might affect tiller
production (also described later in this paper).
3. F.- i, tiller number as a function of a maximum
p ..'t ni q number that is subsequently reduced due
to environmental and management effects. This
approach is explained below.
The period from emergence to the end of the juvenile
stage (ISTAGE 1) was used to estimate the number of
tillers initiated. The maximum number of tillers
(TILG2) per cultivar was calculated from the genetic
coefficient G2 (maximum possible kernel number per
plant) as indicated in Equation 1.
TILG2 = -8.55 + 0.037 G2 + (-3.2E-5 G2**2) Eq. 1
The basis for the functions is described later in this
paper. In Equation 2, tiller number was calculated
from the average minimum temperature (ATMIN)
from emergence to the end of the p ; .-,.- stage:
TILMIN = 3.178 0.71 ATMIN**0.5 Eq. 2
E.l ai'. 3 calculates tiller number from the average
photoperiod (APHOTO) for ISTAGE 1.
TPHOTO = 182.135 25.634 APHOTO
+ 0.903 APHOTO**2 Eq. 3
When TILMIN and TPHOTO are greater than or equal
to one tiller per plant and Equation 2 is greater than or
equal to the theoretical value of 81.55% of F.p, .i..n 3,
then the number of tillers per plant (TILN) is the
maximum value of TPHOTO and TILMIN. However,
if TPHOTO is greater than TILMIN, TILN is equal to
When TPHOTO and TILMIN are both less than one
tiller per plant, and TILMIN is less than 81.55% of
TPHOTO, TILN is the minimum of TILMIN, TPHOTO
and TILG2 TILMIN. However, if TILN is less than or
equal to 0.24 tillers per plant TILN is calculated from
T, 1,, kc- ,,) 4.
TILN = (-6.26 + 0.01 *G2 -. I ,* G2**2)
+ (3.61 + 6.75E + 22 EXP(-4 APHOTO)) Eq. 4
To simulate the influence of row width on the number
of tillers, a function was developed using the data of
Du Toit (1996). Inter-row I ) ;":. (TT) was used in
Equation 5 to calculate tiller number where TT is the
inter-row ;,,.- in meters.
TILN = TILN AMIN1(((-0.4099 + 1.86 *
TT- 1.27*TT**2)/0.23), 1)
Tetio-Kagho and Gardner (1988) calculated tiller
number from plant p 'V .,l,11 ;, This function was
included through Equation 6 in order to prevent the
simulation of TILN for '.,i ,... greater than 3.8
TILN = AMIN1 (TILN, (2.9 0.76 PLTPOP)) Eq. 6
Based on experimental data, kernel number for a tiller
was assumed to be 60% of the kernel number of the
second ear on the main stem.
Results and Discussion
The results in Table 1 indicate greater tiller numbers
for the first planting date than the second and third
I in II( dates, which is consistent with results by
Pretorius (1' -.-': that early planting dates tend to have
more tillers. This is possibly due to the effect of low
minimum temperatures during the early part of plant
development on tiller number (Verwey et al. 1994a).
Lower average minimum temperatures were recorded
for the first planting date than for the second and third
planting dates, while the average minimum
temperatures for the second planting date were lower
than those of the third planting date.
Table 1. Tiller number for 20 cultivars planted on three planting dates
(Potchefstroom, South Africa) as reported by Du Toit (1996).
Cultivar Planting date
07 Sept 92 19 Oct 92 22 Nov 92 Average
SENKUIL 0.30 0.00 0.08 0.13
SR-52 0.63 0.00 0.10 0.24
RS-5206 0.70 0.03 0.05 0.26
PAN-6479 0.70 0.03 0.20 0.31
PAN-6578 0.78 0.05 0.13 0.32
A-210 0.98 0.25 0.15 0.46
CRN-4526 1.13 0.10 0.15 0.46
PAN-473 0.63 0.35 0.50 0.49
PAN-6364 1.00 0.08 0.45 0.51
PAN-6528 1.40 0.05 0.10 0.52
PAN-6363 1.30 0.23 0.23 0.58
A-1257 1.03 0.30 0.60 0.64
SNK-2340 1.98 0.10 0.08 0.72
SNK-2776 1.58 0.38 0.35 0.77
A-1849W 2.03 0.03 0.38 0.81
RO-410 1.75 0.20 0.53 0.83
CRN-3414 1.90 0.18 0.43 0.83
RO-411 2.38 0.15 0.38 0.97
PAN-6552 1.83 1.08 1.65 1.52
TX-24 2.25 0.95 1.58 1.59
Average 1.31 0.23 0.40 0.65
The existing sorghum tiller algorithm in CERES3 was
used as a first step in ..i ... ....p. i:_ a tiller algorithm for
maize in CERES3.
CERES3-Sorghum simulates tiller production from 120
heat units after emergence until the end of the juvenile
stage. The influence of solar radiation (SOLRAD) and
of daily heat units (DTT) on tiller number (TILN) is
accumulated (SUMRTR) over this period as follows:
SUMRTR = SUMRTR + SRAD 23.9 / DTT
The average influence per day (RTR) is calculated by
dividing SUMDTR by the duration of the period in
RTR = SUMRTR / TDUR
If RTR is less than 27, then TILN equals unity. This
rule was not applied to the maize model, so two
temporary constants (TC1 and TC2) are calculated.
TC1 is calculated from RTR, while TC2 is calculated
from plant population (PLTPOP) and TILN as follow:
TC1 = 1.0/25.0*(RTR-27.0)
TC2 = 6.25E-5*(40.0-PLTPOP*TILN)**3
TILN is accumulated over this period from the
fraction of a leaf emerging day1 (TI), the smallest
value (AMIN1) of TC1 and TC2, and the water stress
TLN= TILN+TI*AMIN IT,- i TC2)*TURFAC Eq. 11
However, if TILN ..... 1,. by PLTPOP exceeds 40.0
TILN= i /PLTPOP
The observed standard deviation ,i-, ') was greater
than the mean (Table 2) in the sorghum algorithm, an
indication of the great variation in tiller numbers. R2 and
D-indexes indicated very low levels of accuracy. Root
mean square error (RMSE) and root mean square error -
systematic (RMSEs) values similarly indicated that
almost all the error could be explained by the bias of the
tiller simulation. Thus, .i.i.1.. ;I the CERES3-Sorghum
tillering algorithm to maize showed little promise.
To simulate differences between, nil,, I linear
correlations were calculated between the n'. n.
coefficients used to represent cultivar differences in
Table 2. Quantitative measures of the calculation of maize tiller
numbers by the sorghum algorithm in CERES3 using 123 data points
(Du Toit 1996), as compared to outputs from three routines for
simulating tiller production in (ERES-Maize.
Modeling approach for tiller development
CERES3- CERES3-Sorghum + Proposed
Parameters Observed Sorghum environmental effects method
Minimum 0.01 0.01 0.0064 0.12
Maximum 2.38 0.37 2.48 1.72
Mean 0.49 0.134 0.67 0.45
Std. Deviation 0.53 0.069 0.55 0.39
Slope -0.26 0.62 1.02
Intercept 0.53 0.073 0.033
MAE 0.40 0.39 0.25
RMSE 0.64 0.49 0.36
RMSEs 0.64 0.25 0.25
RMSEu 0.07 0.42 0.26
D-Index 0.40 0.78 0.83
R2 0.00 0.42 0.55
1 MAE = Mean Absolute Error; RMSE = Root Mean Square Error; RMSEs = Root Mean
Square Error- -'..**., RMSEu = Root Mean Square Error unsystematic; D-Index =
index of agreement, Willmott (1982). See also Du Toit et al. (2000) and Du Toit & Du
Toit, p. 42, this volume.
CERES-Maize and tiller production by various
cultivars (Table 3). P5 (growing degree days from
silking to maturity) was excluded because tiller
initiation occurs before the sixth leaf stage (Hanway
1 -v-... The significant correlation of P1 with tiller
number suggested that the initiation of tillers occurred
during the juvenile stage. This was also noted in
previous research where the tiller number was
determined in the first 17 days after n.r. :_ ,. i,:.
(Verwey et al. 1 "' II )
The significant correlation between G2 and tiller
number suggested that G2 could be used to calculate
the genetic potential of a cultivar to produce tillers.
Assuming that the ... ,t : tiller number results from
processes that reduce a potential or optimal number,
F.Ii.h ..11 2 was developed to estimate the optimal
number for a given cultivar.
A second step was to identify environmental factors
that might affect tiller numbers. Measured tiller
numbers were subtracted from the calculated values
to give the residuals of Equation 1. Using backward
i. ;,.. the average photoperiod (AFOTO) during
the juvenile stage appeared to be the environmental
factor showing the strongest relation with the
residuals. A cubic function with AFOTO as the
dependant factor gave the greatest r2 for the residuals.
Tiller numbers were simulated by .I1;in the values of
the cubic function to the values of Equation 1,
j. :;.1-.; the results shown in Table 2. The root mean
square error-unsystematic (RMSEu) was greater than
the RMSEs but closer to the RMSE, indicating that the
error is random. To improve accuracy, changes need to
be made in the .1d i ill i for tiller simulation.
Table 3. Linear correlation between tiller numbers and four
Correlation Slope Y Int.
Y Variable (r) (b) (a) P
PI: Duration of
juvenile phase -0.18 -7.17 247 0.046 *
P2: Effect of photoperiod on
development rate -0.07 -0.029 0.88 0.358 ns
G2: Maximum possible number of
kernels per plant +0.21 30.35 527 0.017 "
G3: Kernel weight coefficient -0.08 -0.06 6.55 0.358 ns
The third approach was the one described in the
section on model modifications above. Of the three
approaches, the last approach showed the greatest
level of accuracy (Table 2). Comparison of the RMSE,
the RMSEu and the RMSEs for the latter two
approaches indicated that the systematic errors of the
two approaches were almost the same in areas where
both the RMSE and the RMSEu decreased. The
RMSEu showed the greatest decrease, i i'.ij due to
the use of both minimum temperature and
photoperiod in the calculation of tiller numbers. The
RMSEu and the RMSEs were similar, indicating that
the error is both random and biased. This suggests
that further factors need to be included in the
calculation of tiller numbers.
The simulation of tiller number is of limited use in
maize production areas where the crop is produced at
plant populations greater than 3.0 plants/m2.
However, the contribution of tillers to yield in the
western T;...1 1.1 is -:., t (Du Toit 1996) and it is
also important in many small-holder systems.
Cultivars show great variation in tiller potential,
which is also affected by planting date. This is
confirmed by the observation in this study that the
standard deviation of the measured values was
greater than the mean.
An algorithm for the simulation of tiller numbers was
developed using average l1, I I. ; .1 and minimum
temperature from emergence to the end of the juvenile
stage. The results indicate that tiller numbers could be
simulated with some degree of success but merits
further ,.' i .-.i ,.;.; The period of plant
development in which tiller numbers is affected by
environmental conditions is of particular interest. The
basis for the simulation of tiller number in maize was
improved for CERES3, without an increase in the
number of genetic coefficients. Although the influence
of tiller number on kernel number is simulated, the
impact of ,II.'i, j on leaf development is not
incorporated in the modified CERES and also merits
further in ; -. ii .. ,
Alargarswamy, G. and J.T. Ritchie. 1991. Phasic development in CERES- Sorghum
model. In T. Hodges (ed.), Predicting crop phenology. Boca Raton: CRC Press
Du Toit, A.S. 1996. Quantification of the compensation ability of the maize plant.
PhD thesis, UOVS, Bloemfontein, South Africa.
Du Toit, A.S., J. Booysen, and J.J. Human. 2000. Use of linear regression and a
correlation matrix to evaluate CERES3 (maize). In J.W. White and P.R. Grace
(eds.), Modeling extremes of wheat and maize crop performance in the
tropics. Proceedings of a workshop, CIMMYT, El Baton, 19-22 April, 1999.
NRG-GIS Series Paper 00-01. Mexico, D.F.: CIMMYT. Pp. 19-31.
Godwin, D.C., Ritchie, J.T., Singh, U., and Hunt, LA. 1989. A users guide to CERES-
Wheat V2.1. Muscle Shoals: International Fertilizer Development Center.
Hoogenboom, G., J.W. Jones, P.W. Wilkens, W.D. Batchelor, W.T. Bowen, L.A. Hunt,
N.B. Pickering, U. Singh, D.C. Godwin, B. Baer, K.J. Boote, J.T. Ritchie, and J.W.
White. 1994. Crop models. In G.Y. Tsuji, G. Uehara, and S. Balas (eds.),
DSSAT3. Vol. 2-2. Honolulu: University of Hawaii.
Jones, CA and J.R. Kiniry. 1986. CERES-Maize: A simulation model for maize
growth and development. College Station: Texas A & M University Press.
Otter-Nacke, S., J.T. Ritchie, D. Godwin, and U. Singh. 1991. A user's guide to CERES
Barley V2.1 Muscle Shoals: International Fertilizer Development Center.
Pretorius, J.P. 1985. Die invloed van c en plantdigtheid op reproduktiewe groei en
opbrengs van mielie cultivars. PhD thesis, UOVS, Bloemfontein, South Africa.
Ritchie, J.T., U. Singh, D.C. Godwin, and LA. Hunt.1992. Genetic coefficients. In A
users guide to CERES Maize V2.10. International Fertilizer Development Center
Simulation Manual IFDC-SM-1, April 1992. Second edition. Muscle Shoals:
International Fertilizer Development Center.
Singh, U., J.T. Ritchie, and PK. Thornton. 1991. CERES-CEREAL model for wheat,
maize sorghum, barley and pearl millet. Agronomy Abstracts 78.
Tetio-Kacho, F and EP. Gardner. 1988. Response of maize to plant population
density. I. Canopy development, light relationships, and vegetative growth.
Agron. 1. 80: 930-935.
Verwey, J.F and P.S. Hammes. 1989. A comparison between single and multi-
stemmed plants. S. Afr. Plant Soil, 6(4): 235-238.
Verwey, J.E, P.S. Hammes, and H. Coetzer. 1994a. Influence of night temperature
on tillering of maize. Appl. Plant Sci 8(2): 61-63.
Verwey, J.F., P.S. Hammes, and H. Coertzer. 1994b. Spruitvorming by mielies en die
invloed daarvan op die duurte van antese. Toeg. Plantwet. 8: 10-13.
Willmott, C.J. 1982. Some comments on the evaluation of model performance. Bu.
Amer. Meteorol. Soc. 63: 1309-1313.
A Comparison of Approaches to
Modeling Phenology as Applied to Genotype
by Sowing Date Interactions in Wheat
Jeffrey W. White', S.S. Dhillon2, L.A. Hunt3, and P.D. Jamieson4
N natural Resources Group, CIMMYT, Mexico
-I u, ,,f:, A:_, I.:;. lltc. al University, Department of Plant Breeding, Ludhiana, Punjab, India
3Department of Plant d i. uli,,. University of Guelph, Ontario, Canada
4NZ Institute for Crop & Food Research, Ltd, New Zealand
Ewha, ,..'.:;' i land preparation for wheat crops -:..; ....' rice allows the wheat to be sown much earlier in the rice-
wheat rotations prevalent in South Asia. The earlier sowing exposes the crop to warmer temperatures and longer
photoperiods during crop establishment. To test how well wheat models can account for .. :, :. cultivar differences
in response to early plantings, simulations fiom the CERES Wheat and Sirius models were compared to observed
responses for two cultivars sown on multiple dates at Ludhiana, Punjab, India. Both models showed problems,
especially in .. .!.''... vernalization. The Ludhiana environment is on the edge of vernalization : 1. i '.. so any
error in assumptions about the upper limiting temperature for vernalization response is likely to have a large ,T ;.
The rice-wheat system of South Asia is one of the
most productive i, i. Ilii il systems in the world.
T1 i ..r-. double cropping of rice and wheat, farmers
make efficient use of the year-round growing season,
matching rice to the warm, wet monsoon season and
irrigated wheat to the cool, dry winter season
(Fujisaka et al. 1'"-- ,' Concerns over both the
biophysical and socioeconomic sustainability of this
system have fostered extensive research on strategies
for protecting yield gains while 111.,... in_ factors
usually ih,.. 111i to be associated with conserving the
resource base. These include crop diversification,
more efficient use of irrigation water, increased soil
organic matter, and reduced use of external inputs.
Various strategies for eliminating tillage between the
rice and wheat crop have shown L'" h i r ii promise
(Hobbs et al. 1 "'-, Mehla et al. 2000; Timsina and
Connor A: .:.: A key aspect is that wheat can be sown
closer to the 1-..inm.l date by ..imI.,i., the period of
waiting for paddies to dry and be tilled. With the
earlier, optimal sowing, the wheat crop matures
under cooler, more favorable conditions, which is
reflected in higher yields. Estimates from the Punjab
are that wheat grain yield declines 1 to 1.5% per day of
delay in sowing (Dhillon and Ortiz-Monasterio 1 "'
Ortiz-Monasterio 1'" -4 Further benefits include that
the sowing date is less subject to adverse events such
as untimely rains that can delay or prevent land
Given the prospect of more frequent early -.. : ;,._ the
.1"1. '1,' has arisen whether certain cultivars are
especially suited for such practices (Mehla et al. ..
Although characteristics such as the ability to emerge
from a rough seedbed with surface residues are also
desirable, an appropriate phenological response seems
essential. So-called ":.1q, revolution" wheats that
predominate in the region are relatively insensitive to
photoperiod and have a low vernalization
requirement, but it is -:. .- !-,:. that these low levels are
still i,. i. i to influence development. Mehla et al.
- = =!:!. noted that two cultivars favored for timely
(early) sowing, WH 542 and PBW' i-' have a longer
.. ,- L -l. t.,: ,- phase and shorter _- .,,m 1, l _. phase than
older varieties such as HD 2009 and HD 2329.
Simulation models are .1. iili ill. valuable analytic
tools for analyzing cultivar differences in the
response of phenology to sowing date. Many wheat
models allow for quantitative differences among
cultivars in inherent earliness earlinesss per se),
vernalization requirement, and photoperiod
sensitivity. These models would thus seem suited to
analyzing the underlying processes of genotype by
sowing date interactions. This paper compares
simulations from two models, CERES-Wheat and
Sirius, for (..,ii.i.-- tE : -- *1, ... .,- described in the
sowing date study reported by Dhillon and Ortiz-
Monasterio (1993) and Ortiz-Monasterio et al. (1994).
The two models have similar objectives but differ in
their approach to InII1 I,,y development.
Materials and Methods
Data for days to heading vs. sowing date for the
spring wheat cultivars HD 2329 and PBW 34 were
obtained from a series of .. ..b I.,- by .- ;,,.. date
experiments conducted from 1985 to 1992 at
Punjab Agricultural U,,;, ;i Ludhiana, India
(lat. 30.93N, long. 75.87E, elev. 247 m.), as
reported by Dhillon and Ortiz-Monasterio (1993)
and Ortiz-Monasterio et al. (1 '** HD 2329 is
reported to be dominant for all four Vrn loci, and
thus has a very low vernalization requirement (J.
van Beem, personal communication, 2001).
To calibrate HD 2329, seven sowing dates from 25
October to 25 December 1- -' were used; for PBW
34, ten sowing dates from 25 October to 25
December 1985 were used. For validation, sowing
dates for both cultivars were 15 October to 25
December 1' '. In our analyses, days to anthesis
was estimated to be seven days after heading, as
reported in the original study. Trials were well
irrigated and fertilized, so phenology data should
show little effects of water or nutrient deficits.
Figure 1 shows long term climatic conditions at
Ludhiana, and Figure 2 shows variation in daily
maximum temperature for three years to illustrate
the year-to-year and within year variability in
Both models use thermal time as their basis for
predicting phenology and include effects of
vernalization and photoperiod. However, the models
use different approaches to account for these effects.
CERES-Wheat uses a well-tested phasic development
approach, where the thermal duration of phases
between observable apical events is modified by
factors related to photoperiod and the accumulation of
vernalization. Variations on the approach are used in
several wheat simulation models, such as
AFRCWHEAT2 (Porter 1993), as well as various
CERES derivatives, such as Cropsim-Wheat (Hunt and
r....i..j. in ,i.i.. 1995). The approach has proved
robust in a variety of environments. Sirius attempts to
move its empirical i .., -in q- -. down one level of
I .. I. .1q, ,1 to simulate the effect of photoperiod and
vernalization on mainstem final leaf number (FLN)
and, through it, on the thermal time to anthesis. This
40- Maximu 200
S20 Minimu00m /
.."."" ....... .. .. / Precipitation
0 0 .. '**.
J F M A M J J A S 0 N D
Figure 1. Long term monthly values for mean maximum and minimum
temperature and total precipitation at Ludhiana, India (FAO, 1998).
Oct. Nov. Dec. Jan. Feb. Mar. Apr.
Figure 2. Variation in daily maximum temperature from October to
April for three years at Ludhiana, India.
approach has also proved successful in a variety of
environments. More detailed descriptions of both
Version 3.5 of CERES-Wheat was used as provided by P.
Wilkens (IFDC). The most complete documentation is
Ritchie and Godwin (N.d.) for version 2.0; additional
information is found in Godwin et al. (1989) and
Hoogenboom et al. (1' '. Key developmental events in
CERES-Wheat are ... .., -~... dil; emergence, and
terminal spikelet initiation. The rate of development
varies with temperature per se (thermal time or growing
degree days), vernalization and photoperiod.
Vernalization is assumed to occur at temperatures
between 0 to 15"C; temperatures above 7C are assumed
to have decreasing effectiveness. The relative
vernalization effectiveness varies from 0 to 1 depending
on the crown temperature. A ul f.,. -" i l(.
vernalization coefficient (P1V) conditions the cultivar
sensitivity, with values of 1 or lower 1., ;i, .,, ,t 11t ,.
of spring wheats. De-vernalization occurs if the number
of vernalization days accumulated is less than 10 units
and the maximum temperature is over 30C.
Photoperiods shorter than 20 h (calculated based on civil
i%, ih,,it I delay development through a rate modifier (DF)
that varies from 1 to 0 as photoperiods shorten. A
cultivar-specific photoperiod sensitivity coefficient (P1D)
conditions the cultivar 0I r. with values of 0 to 3
being representative of day-neutral spring wheats.
Thermal development time (TDU) is accumulated up to
:' degree days, ending with terminal spikelet
formation. Daily thermal time (DTT) is either reduced
due to the influence of vernalization (VF) or photoperiod
(DF), J, I ~ l 11; on which rate-reducing factor is smaller
(and provides the greatest inhibition of development):
TDU = DTT [minimum(VF, DF)] Eq. 1
There is still considerable discussion I- _.I.J n_ the
correct values for the cardinal temperatures of wheat,
but a base temperature (Tbase) of 1C and optimal
temperature (Topt) of 25C were found satisfactory in
preliminary calibrations. The duration from terminal
spikelet formation to anthesis is affected only by
temperature and is assumed to take 5 phyllochrons (3
to end of leaf growth and 2 to more to anthesis).
Sirius 2000 is the latest version of Sirius (Jamieson et al.
: .- 1 -., It is an advance on Sirius 99 (Jamieson and
Semenov 2000), mostly differing in its J.1.-i i i h..? n of soil
water. The version used in this study has essentially the
same l.1 ,1, ..;, 1 descriptions as the, ;i i .11 except
that the temperature response of vernalization was
modified based on the recent controlled environment
work of Brooking and Jamieson (2001).
Whereas the phenological models in CERES-Wheat are
based on phasic development-i.e., the (modified)
thermal time intervals between observable apical
events-in Sirius major variations in timing are
associated with 1, ... in final leaf number (FLN).
Phases as such play a much less important role in
Sirius (Jamieson et al. 1 .-. i, The .1 ... -i. ;.,1 events
for which timing is described are emergence, flag leaf
,ni..' appearance (FL, or full expansion of the flag leaf
lamina), .mi.I- i- and the 1....2 .iiiii and ending of
grain filling. The minimum FLN observed in the field
is usually about seven or eight, but FLN can exceed 20.
The thermal time intervals to emergence and between
events after FL are assumed to be fixed in thermal
time-e.g., FL to anthesis is three phyllochrons, and
anthesis to I. I i .i111 of grain filling is one
phyllochron. The thermal duration of grain 1 lim may
vary with cultivar but is usually about 550C days
(0C base). These aspects of the .1, ..I. .;, 1 model
are similar in concept to CERES-Wheat. The major
difference is that in Sirius the date of FL is calculated
from FLN and the response of leaf appearance rate to
temperature (Jamieson et al. 1995). FLN is itself
calculated from responses to daylength (Brooking et al.
1995) and vernalization it:'.... I 1 1 1996, Robertson et
al. 1996). Thermal time for development is calculated
using a base temperature of 0C. No optimum
developmental temperature is assumed, so the rate of
leaf appearance increases along with temperature with
no limit. Thermal time is not modified in any way by
photoperiod or temperature (i.e., the concepts of
Lhi, 't.. i, i ii or photo-vernal-thermal time are not
used). Rather, the thermal time from emergence to FL
(and therefore to later stages) depends on FLN.
To model the .i i. ;, .-, (photoperiod) response, FLN
is assumed to be fixed according to the ,I ,I: (DL)
that occurs when there are FLN/2 leaves on the plant:
FLN = FLNmin + SLDL (16-DL)
FLNmin is the minimum number of leaves (usually
about 8) produced in long days (i.e., .. ...... 1; r.: 16
hours). The daylength response parameter SLDL is the
number of extra leaves produced per hour of
daylength below 16 hours. The procedure involves
daily estimates of FLN based on current daylength,
and a check to see whether the actual leaf number is
half the estimated FLN. When this latter occurs, FLN
is fixed and timing depends on how long it takes to
produce the remaining leaves.
The vernalization model also predicts FLN as its main
method of influencing time to flowering. It requires
three parameters, which are:
1.-The maximum FLN (FLNmax). This is the number of
leaves that would be produced on the mainstem in
continuous exposure to warm long days.
2.- The vernalization rate at 0C (VBEE). This is reciprocal
of the time taken to saturate the vernalization
requirement at 0C.
3.- The temperature response of the vernalization rate
(VAI), assumed to be the same for all cultivars.
Winter cultivars (i.e., those with a vernalization
requirement) are assumed to have a large I' ,t. ni'.'i i.! i
number at germination (FLNmax). The number
decreases as vernalization proceeds. Vernalization is
completed when FLN is minimized, either because the
number of leaf primordia (produced twice as fast as
leaves .i.. ,i on the apex is equal to the potential FLN,
or because the potential FLN has fallen to 8. The rate of
vernalization increases with temperature from 0C until
8C (Brooking and Jamieson 2001) and then falls linearly
to reach zero at 17C. Vernalization does not proceed at
temperatures outside the range 0 to 17C.
For the 25 December .i ,mt4w. both cultivars behaved
similarly. However, HD 2329 flowered much earlier
than PBW 34 in earlier pi: dnvr, A simple .,.Il..,bi, .
". I.lOt Pr., f.ri this pattern is that flowering of PBW
34 was delayed in early sowings due to a vernalization
response. Later ii, -, %i. i, of HD 2329 with
intermediate sowing dates could reflect either cooler
temperatures or a sensitivity to short days.
In calibration of CERES Wheat for HD 2329, the best fit
between observed and simulated data was obtained
by assuming that there was no vernalization
response (P1V = 0) and a -lhJlt sensitivity to
photoperiod (P1D = 2.0). Allowing for even a small
vernalization requirement resulted in excessive delay
in time to anthesis at early dates. In contrast, the best
fit for PBW 34 required a moderate vernalization
(P1V = 1.8). Determining the correct photoperiod
response for PBW 34 proved problematic because
CERES uses the minimum value of the two rate
modifiers (i.e., for vernalization and photoperiod, Eq. 1).
There was no effect for values of P1D from 0 to 2, but a
value of 2.5 resulted in a slight improvement if combined
with a shorter phyllochron interval (PHINT = 92).
The summary statistics for model performance of
CERES Wheat suggest considerable room for
improvement (Table 1). Figure 4 shows partial results
of a sensitivity analysis where P1V was given values
of 0, 1.0, 1.5 and 2.0, and P1D was varied from 0 to 3.0
in steps of 0.5. With no vernalization effect (P1V),
Modeling Anthesis Date
Data were first assembled in DSSAT 3 format, a
predecessor of the ICASA format for data interchange
(Hunt et al. '" '). Each model was calibrated for the
two cultivars using the data for 1985 (HD 2329) or
1988 (PBW 34) and then validated using the 1989 data.
Standard parameters for goodness of fit were
calculated to facilitate comparisons of model outputs.
Overview of Field Response
The effect of sowing date on time to anthesis indicates
the contrast between HD 2329 and PBW 34, F,. ,,, 3).
S30 ....-....* ............... ..
15-Oct. 25-Ot. 4-Nov. 14-Nov. 24-Nov. 4-Dec. 14-Dec. 24-Dec
Figure 3. Effect of sowing date on days to anthesis in 1989 for cvs. PBW
34 and HD 2329. Based on data of Dhillon and Ortiz-Monasterio 1993.
increasing photoperiod sensitivity delays anthesis
by 20 to 25 days (Figure 4A). Adding a vernalization
effect overrides the possible effect of low
photoperiod sensitivity (e.g., Figure 4B, C and D).
This reflects the use of a minimization function in
Eq. 2, in which only the stronger of the two
processes affects development. Our understanding
of the 1l 1 ;. ,1,: ..- of fi. %, ;. indicates that
vernalization and photoperiod are two separate
processes that affect reproductive development. It is
difficult to envisage a physiological justification for
the minimization function.
An additional concern is that the photothermal
time from emergence to terminal spikelet formation
is fixed at 400 growing degree days. This implies
that there is no difference in inherent earliness
earlinesss per se) of wheat .-: lt; .., which is also
not supported by data for wheat given
vernalization treatments and grown under warm
temperatures and long photoperiods (e.g.,
Midmore et al. 1982). Reducing the phyllochron
interval (PHINT) will reduce time to anthesis but
leads to unrealistic leaf numbers.
Table 1. Results of simulations of the duration sowing to anthesis for
HD 2329 and PBW 34 for the calibration and validation runs of CERES-
Wheat and Sirius.
HD 2329 PBW 34 HD 2329 PBW 34
Vernalization 0 1.8 0 0.22
Photoperiod sensitivity 2.0 2.5 0.2 0.25
Phyllochron 95 92 100 100
r2 0.69 0.92 0.81 0.48
Slope 1.58 1.55 0.62 0.71
Intercept -58.8 -30.6 34.9 30.6
MAD 4.7 4.2 4.7 6.9
RMSE 6.4 4.9 5.3 9.0
No. observations 7 10 7 10
r2 0.94 0.74 0.95 0.40
Slope 1.28 1.44 0.86 0.44
Intercept -32.0 -39.8 11.9 58.6
MAD 6.4 4.3 1.9 5.0
RMSE 6.8 5.2 2.1 7.1
No. observations 8 8 8 8
Note: ': i '.;; :h. .I .l ...! ,i !..,|. i .. ,";I ii. h .l.:: :,...i. i and
il:. .,,;Hi,;,. ;.ii .. ., i,,. n.,.:. .1, .; | ,-,,,;.-....; ,,h; ..- ,; i The calibration and
validation parameters are: r2= proportion of variance in Y (observed) that is explained by
X ...iJ l.h i'l: ..i ... i. .. i !. Hi =- root mean square error.
288 298 309 319 329 339 349 359
Date of sowing
288 298 309 319 329 339 349 359
Date of sowing
288 298 309 319 329 339 349 359
Date of sowing
2 0 20 0
80o 2 ,001S
7 0 -------- j ------
288 298 309 319 329 339 349 359
Date of sowing
Figure 4. Sensitivity analysis for days to anthesis with CERES-Wheat
using varying sensitivity to vernalization (P1V) and photoperiod (PI D).
Simulations are for 1989-1990 weather data at Ludhiana.
Calibration of HD 2329 assumed no vernalization
response. In the absence of other information, the
phyllochron was chosen at a midrange value of 100 C
days. The experiment site is at latitude 30.9N, and
daylength variation over the range of ',.. dates
used was small (10.0 to 12.7 h). This rules out
daylength variation as a major cause of variation in
the duration sowing to flag leaf emergence. With no
.1. 1, ,11,.., i. ponse (i.e., a fixed FLN of 8), anthesis was
systematically predicted early. Setting the .L.r r_.tl
response parameter SLDL at 0.2 leaves/hr resulted in
the addition of nearly one further leaf and gave a close
match between simulated and observed anthesis date
(Table 1). Variation in predicted FLN over the I ..... of
sowing dates in either the calibration or validation
datasets was only about 0.2 leaves, equivalent to about
20'C days or about one day. The predictions provided
a very close match to observed anthesis dates in the
The situation with PBW 34 was quite different. In
contrast to HD 2329, the variation in the interval
.. nl ; implied substantial variation in FLN.
Because of the small variation in daylength, this could
only be associated with a vernalization response. Use
of vernalization parameters typical of European
wheats resulted in systematic, very late prediction of
anthesis. The phyllochron, as with HD 2329, was set at
100 C days. SLDL was chosen so that FLN would be
about nine, with vernalization saturated by
emergence. Insufficient vernalization would then add
further leaves to increase the duration. The best value
for VBEE (0.22/day) provided a good fit for the later
sowings, but the errors in anthesis i I i,. nI for the
early sowings were large and both early and late. The
performance of the model in this situation was
substantially inferior to the less mechanistic but more
conservative approach used by CERES-Wheat. This
bears further investigation.
Sirius simulations were run for the 1989 dataset with
FLN adjusted so that the simulated anthesis date
matched the observed anthesis date for each run. This
FLN was then designated as "observed." If the actual
phyllochron had been larger or smaller than assumed,
these values of FLN woul have been systematically
high or low, but the pattern of change would have been
preserved. These values were then compared with FLN
predictions using the parameter values from the
calibration (Figure 5). In this set, there were five very
close matches of simulated FLN observed, as well as
one substantial overestimate and two substantial
underestimates. These corresponded with the large
errors in estimated anthesis date. Note that the pattern
of variation was .- t.i,. ., with the overestimate of
FLN derive from an early sowing and the
underestimates from later sowings. Both observed and
predicted FLNs for the late sowings were close to nine
leaves, and indicated that the vernalization requirement
was satisfied very early-probably by emergence.
The errors in the predictions of FLN led to some odd
behavior in the simulations. The predicted anthesis
date for the second sowing was substantially later
than those predicted for the next two later sowings.
It is clear that, as far as simulation is concerned, the
Ludhiana environment is on the edge of vernalization
effectiveness. Any error in -- t -. -. about the
upper limiting temperature for vernalization response
is likely to have a large effect. In Sirius this temperature
is set at 17C, but the maximum daily temperature
during the wheat growing season was always greater.
Vernalization then relied on the fact that minimum
temperatures were less than 17C for a portion of the
season. Coupled with the small vernalization
requirement of the cultivar, this meant that the slight
differences in temperature regime associated with
different sowing dates can lead to i -nif. 511i
differences between simulation results and what is
observed in nature. The environment is a severe test of
vernalization models, and the dataset is extremely
valuable for that reason alone.
12 N Observed
6-Oct. 5-Nov. 5-Dec. 4-Ja
Figure 5. Comparison of Sirius simulated final leaf numbers with those
observed for PBW 34 in 1989.
The simulation exercises ir-lh _i-t the potential impact
of vernalization and photoperiod sensitivity with early
plantings in rice-wheat systems of South Asia. They
suggest that once-popular cultivars such as HD 2329
and PBW 34 would differ markedly in their response to
early sowing due to simple genetic differences for
vernalization and .1 .. I..!.. i i -.. ii The data set of
Dhillon and Ortiz-Monasterio contains data for 32
cultivars in series of experiments involving seven to ten
plating dates over seven years with '. ,. ;,, .. numbers of
cultivars each year. This information represents a
valuable source of data for unraveling this problem
further. Furthermore, if the data set can be linked to
more complete information on crop management (e.g.,
probable times of ; i._t *.,', the set could serve as a
foundation for relating the genetic effects to grain yield.
The results also indicate that the development routines
in CERES-Wheat and Sirius require revision. In CERES,
the use of a minimization function for determining
effects of vernalization and photoperiod appears
unrealistic and makes calibration difficult. Furthermore,
in order to predict leaf number accurately and reflect
cultivar differences with respect to inherent ,..,In r1 -
the total photothermal time from emergence to double
, i1 formation should be a variable rather than a
constant value of 400 GDD, as is currently assumed.
Prediction of anthesis dates for spring wheat cultivars
with Sirius is satisfactory. However, a problem exists
simulating vernalization effects for wheat cultivars in
certain environments. These environments are those
where temperatures are in the upper range of
vernalization effectiveness and, for at least some of the
time during early development, daily temperature
ranges either span the upper limit or are entirely
outside it. Solving this problem will require, at least, a
sensitivity analysis of anthesis date prediction to
variation in the upper temperature limit and the shape
of the temperature response.
Both approaches had problems dealing with
vernalization in this environment. We again
emphasize that data from such environments are an
extremely valuable resource because they provide a
severe test of the assumptions assembled in
simulation models. Only through exposure to such
testing situations will advances in our
understanding be made.
Aggarwal, PK., K.K. Talukdar, and R.K. Mall. 2000. Potential yields of rice-wheat system
:,-; i ,;,t ., ;-,,j,;:, .,-:, of India. Rice-Wheat Consortium ,:.. ; 10. New
Delhi: Rice-Wheat Consortium !., i,. I,,I,, C,,:,,j. :. Plains. Pp. 16.
Brooking, I.R. 1996. The temperature response of vernalization in wheat A
developmental analysis. Annals of Botany78: 507-512.
Brooking, I.R., P.D. Jamieson, and J.R. Porter. 1995. The influence of daylength on the
final leaf number in spring wheat. Field Crops Research 41: 155-165.
Brooking, I.R. and Jamieson, P.D. 2001. Temperature and photoperiod response of
vernalization in near-isogenic lines of wheat. Field Crops Research 79: 21-38.
Dhillon, S.S., and J.I. Ortiz-Monasterio R. 19','. I',i-. dateae of sowing on the yield and
yield components of spring wheat and their relationship with solar radiation and
temperature at Ludhiana, Punjab, India. Wheat Special Report No. 23b. CIMMYT,
Mexico D.F., Mexico.
FAO (Food and Agriculture Organization of the United Nations). 1998. CLIMWAT, a
climatic database for CROPWAT. http://www.fao.org/WAICENT/FAOINFO/
AGRICULT/AGL/aglw/climwat.htm (2 December 1999).
Fujisaka, S., L. Harrington, and P. Hobbs. 1994. Rice-wheat in South Asia: Systems and
long-term ,.... i.,Jo b ii I," ...... iJ .,-,.,, i. ,..,h 1,; .,!,... m:, ;,.,
Godwin, D.C., J.T. Ritchie, U. Singh and L.A. Hunt. 1989. A user's guide to CERES Wheat:
V2. 10. Muscle Shoals: International Fertilizer Development Center.
Hoogenboom, G., J.W. Jones, P.W. Wilkens, W.D. Batchelor, W.T. Bowen, L.A. Hunt, N.B.
Pickering, : ,ii. D.C. Godwin, B. Baer, K.J. Boote, J.T. Ritchie, and J.W. White.
1996. Crop models. In G.Y. Tsuji, G. Uehara and S. Balas (eds.), DSSATversion 3.
Volume 2. Honolulu: IBSNAT, University of Hawaii. Pp. 95-281.
Hobbs, P.R., G.S. Giri, and P. Grace. 1997. Reduced andzero tillage options for the
establishment of wheat after rice in South Asia. RCW Paper No. 2. Mexico, D.F.:
-:., wni.l.,, ... I..." m j i.. i..i : !,,I .Plains and CIMMYT.
Hunt, L.A. and S. Pararajasingham. 1995. Cropsim-Wheat: t .-..,:'h ." ... i-l..,, ...i,
and development of wheat. Canadian Journal of Plant Science 75: 619-632.
Hunt, L.A., J.W. White, and G. i.....:; ,. ...:::i ;,.......- .., 1 .-.,:.. in
documentation and protocols for exchange and use. Agricultural Systems 70:
Jamieson, P.D., I.R. Brooking, J.R. Porter, and D.R. Wilson. 1995. Prediction of leaf
appearance in wheat: A question of temperature. Field Crops Research 41: 35-44
Jamieson, P.D., I.R. Brooking, M.A. Semenov, and J.R. Porter. 1998a. Making sense of
wheat development: A critique of methodology. Field Crops Research 55: 117-127.
Jamieson, P.D., M.A. Semenov, I r i-,.-..-.-;; : and G.S. Francis. 1998b. Sirius: A
mechanistic model of wheat response to environmental variation. European Journal
Jamieson, P.D. and M.A. Semenov. 2000. Modeling nitrogen uptake and redistribution in
wheat. Field Crops Research 68: 21-29.
Mehla, R.S., J.K. Verma, R.K. Gupta, and P.R. Hobbs (eds.). 2000. Stagnation in the
productivity of wheat in the Indo-Gangetic plains: zero-till-seed-cum-fertilizer drill
as an integrated solution. Rice-Wheat Consortium Paper Series 8. New Delhi: Rice-
Wheat Consortium for : ..-. J... Cu... ". i'l;., Pp.12.
Midmore, DJ., P.M. Cartwright, and R.A. Fischer. 1982. Wheat in tropical environments. I.
Phasic development and spike size. Field Crops Research 5:185-200.
Ortiz-Monasterio, J.I., S.S. Dhillon, and R.A. Fischer. 1994. Date of sowing effect on grain
yield and yield components of .,;,.,. .;.. ..; -i...i- ; n- ,,.n-l relationship with
radiation and temperature in Ludhiana, India. Field Crops Research 37:169-184.
Porter, J.R. 1993. AFRCWHEAT2: A model of the growth and development of wheat
incorporating responses to water and nitrogen. European Journal of Agronomy
Ritchie, J.T. and D. Godwin. (N.d.). CERES Wheat 2.0 http://nowlin.css.msu.edu/
wheatbook/(25 August 2001).
Robertson, M.J., I.R. Brooking, and J.T. Ritchie. 1996. The temperature response of
vernalization in v Ih- l r' dJ-ilI,,. : .-,.,,,. ei, I ..l number of mainstem
leaves. Annals of Botany 78: 371-381.
Timsina, J. and D.J. Connor. 2001. '-,.J.::.., ..;*, ,ud management of rice-wheat cropping
systems: Issues and challenges. Field Crops Research 69: 93-132.
Short Description of the Model Statistical
Package and Weather Analogue Program
A.S. Du Toit1 & D.L. du Toit2
1 ARC Grain Crops Institute, Potchefstroom, RSA
2 Sustainable Farming System, Kock Park, RSA
Tofacilitate model validation and adaptation in the Highveld Ecoregion Project, two software packages were
developed. The Model Statistical Package I. :') uses standard outputs of CERES-Maize 3.0 to calculate
indicators of model accuracy, including the linear .. ... ... statistics (slope, intercept, and r2), D-index, and
the systematic and unsystematic mean square errors. The Weather Analogue ":. ... (WAP) allows users to
create mid-season projections of crop performance based on the five sets of historical weather that show the
greatest similarity to the .... ',, season.
The Highveld .. i.' -., of southern Africa is an
important producer of rainfed maize but its highly
variable rainfall patterns present a major challenge to
producers. Simulation models offer various
S-.. --.li-. for guiding farmers on managing
production risk. Twoi .p of the Highveld
Ecoregion Project' are "to validate existing ICASA
models and adapt them to be applicable for
agricultural systems in the Highveld Ecoregion." To
facilitate model testing and .q'i ,h..i, i..i i of models in
the Highveld region, two software tools were
developed to analyze outputs from the DSSAT group
of models. These two tools were the Model Statistical
Package (MSP) and the Weather Analogue Program
(WAP). This paper ni. these two tools, using
CERES-Maize v3.0 as part of DSSAT models
(Hoogenboom et al. 1' -' i,: to illustrate the software
features. Such tools would also be of great use in
evaluating proposed improvements in the temperature
response of crop models.
The CERES Model
CERES models for maize, sorghum, wheat, millet, and
barley have been combined, ..- .hll, in a generic
1 Funded 1, il [.....,.... Fund Managed by ISNAR
multi-crop CERES3. This version runs from a single
executable fileset of code, ; ... i 1 .1i, i;,:. the
development and growth sections for each individual
model into a in,.I. module with a single soil
component (Tsuji et al. 1994). According to Ritchie
(1991), generic models should allow users to follow
more uniform procedures for ..,l l .tii models and
for linking them with components not included in
the .. i;i model. The ., ,, ;, CERES3
(Hoogenboom et al. 1994) was used for the
simulations, with modifications as reported by Du
Model Statistical Package
In order to verify and calibrate a model, well-defined
criteria are needed to evaluate model performance. It
is generally accepted that the ultimate test of a
simulation model is the accuracy with which it
describes or mimics the actual system, usually
involving comparisons between simulated and
observed data (Willmott 1982; Jones and Kiniry 1986;
Oreskes et al. 1994). A number of statistical methods
for ", o,1 -;i-, model performance are available. These
include linear regressions techniques as proposed by
Jones and Kiniry (1' "-"., and Flavella (1992) and D-
index, systematic and unsystematic mean square
errors as proposed by Willmott (1982).
Jones and Kiniry (1986) used linear regression
techniques of the form y = a + bx with simulated
results as the independent variable. Good model
performance was obtained when the intercept (b)
approached zero and the slope of the regression
approached unity, indicating a near perfect
relationship between observed and simulated values.
Complementary to this regression, the Pearson
correlation coefficient (R) can also be calculated,
indicating the similarity or inverse similarity of a
response in y for a response in x. The coefficient of
determination (r2) is readily calculated from R,
signifying the percentage of variation that is
accounted for by the model.
The deficiencies of the latter statistical parameters
were pointed out by Willmott (1982) and Harrison
(1990). They indicated that the observed and
simulated data might occur in a narrow band, whereas
this is usually not the case with the coefficients.
Secondly, although significance levels can be
calculated, it is difficult to identify the point when a
model is valid or not valid. Savage (1993) further
warned against the use of a correlation coefficient if
the data are not randomly distributed.
The D-index (index of agreement), RMSEs (root mean
square error systematic), RMSEu (root mean square
error unsystematic) and RMSE (root mean square
error) are four indicators that Willmott (1982)
recommended for model evaluation. Due to
limitations in the use of correlation coefficients as an
agreement index, Savage (1993) stated that the
statistics as defined by Willmott (1982) should be
According to Wilmott (1982), the D-index of a "good"
model should approach unity and the RMSEs
approach zero, whereas the RMSEu should approach
the RMSE. The mean absolute error (MAE) and RMSE
are among the best overall measures of model
performance, as they summarise the mean difference
in the units of observed and predicted (Willmott 1982):
MAE = N-1 Pi-Oi Eq. 1
RMSE = [N-1 (Pi-Oi)2]o Eq. 2
N is the number of observed values and Oi and Pi are
observed and predicted values for the i-th data pair.
The systemic and unsystemic RMSE require
calculating the intercept (a) and slope (b) of the least-
squares regression, ^P = a + bO Pi (Willmott 1982),
RMSEs = N-1 Pi-Oi2]
RMSEu = [N-1 Pi- Pi ]05
where Pi is regarded as the best estimate of the
predicted quantity (Savage 1993). The advantage of
RMSEs is that it indicates the bias (deviation of the
actual slope value from the 1:1 line) in a particular
model, instead of the random variation (RMSEu) that
may occur (Savage 1993). Willmott (1982) proposed an
"index of agreement" (D) of the form:
D = 1- [ (Pi-Oi)2/ P'i + d 2
where P'i = Pi O (average of the observed) and O'i =
Oi O. The index (D) is intended to be a descriptive
measure. It is also both a relative and bounded
measure, which can be widely applied in order to
make cross-comparisons between models
(Willmott 1 1982)
Observed vs. simulated graphs, also known as 1:1
graphs, are widely used in simulation evaluations
(Willmott 1982; Jones and Kiniry 1986). Harrison
(1990) recommended that it is appropriate to combine
whatever statistical method is used with a 1:1 graph
since it may be particularly helpful in identifying the
pattern of differences between the predicted and
The MSP software reports all of the above-mentioned
statistical parameters using the "MAIN GROWTH AND
DEVELOPMENT VARIABLES" provided in the file
Overview.out by comparing the predicted with the
measured values as indicated in Table 1. The
measured values are then subtracted from the
predicted values; the difference is the error or residual.
A correlation matrix is calculated between the error
and each of the 50 predicted environmental and stress
factors as indicated in Table 2. The correlation matrix
is presented as a regression matrix table by MSP as
indicated in Figure 1.
Table 1. Main growth and development variables in the Overview.out file as output from a model run with CERES3.
@ VARIABLE PREDICTED MEASURED
FLOWERING DATE (dap) 81 -99
PHYSIOL. MATURITY (dap) 160 -99
GRAIN YIELD (kg/ha) 5882 5853
WT. PER GRAIN (g) 0.4016
GRAIN NUMBER (GRAIN/m2) 1237
MAXIMUM LAI (m2/m2) 1.20
BIOMASS (kg/ha) AT ANTHESIS 2441 -99
BIOMASS N (kg N/ha) AT ANTHESI 0 -99
BIOMASS (kg/ha) AT HARVEST MAT. 8244
Table 2. ENVIRONMENTAL AND STRESS FACTORS in the Overview. out file as output from a single run with CERES3.
I-DEVELOPMENT PHASE-I I-TIME-I I-- WEATHER-- I I- WATER-I I- NITROGEN-I
DURA TEMP TEMP SOLAR PHOTOP PHOTO LEAF PHOTO LEAF
TION MAX MIN RAD [day] SYNTH EXPAN. SYNTH EXPAN.
days oC oC MJ/m2 hr
Emergence-End Juvenile 15 28.27 11.67 20.86 13.47 0.000 0.000 0.000 0.000
End Juvenil-Floral Init 7 25.90 12.90 22.23 13.64 0.000 0.000 0.000 0.000
Floral Init-End Lf Grow 49 22.62 13.10 22.68 13.72 0.000 0.000 0.000 0.000
End Lf Grth-Beg Grn Fil 16 25.45 12.31 20.63 13.37 0.000 0.000 0.000 0.000
Grain Filling Phase 62 23.05 13.60 21.18 12.38 0.000 0.000 0.000 0.000
rrioc uays i emp max i emp min 0olor nda
End juvenile 0.0195 0.2331 0.0195 0.1884
Floral Init 0.5139 0.0391 0.8345 0.4896
L Growth 0.9982 0.829 0.0314 0.433
Gm Fil 0.7298 0.0412 0.0002 0.028
Filling phase 0.7198 0.6626 0.6807 08619
Minimum 781 555
Si: J I i
I_ .- iI '. -
Si1 ... 1: ii
Figure 1. Output of the MSP to indicate the statistical parameters and relationship between the environmental and stress factors and the difference
between the observed and the simulated.
Weather Analogue Program
The Weather Analogue Program (WAP) was
developed to assist seasonal predictions modeled
for a given site. WAP uses an analogue
methodology whereby the most current weather
conditions of the present season are compared to
the five best fitting historical seasons in the
database. The information is then presented in the
form of a graph and a table. The search algorithm
uses the "index of agreement" (D) as represented in
Eq. 5 (Willmott 1982) but compares daily weather
data instead of observed or simulated crop
performance. Each season is presented as the
simulation output of a model with the current
season as the observed values. Figure 2 shows the
graph indicating the two closest years. Although
they are presented as cumulative values, the search
algorithm uses non-cumulative data.
The data are indicated as cumulative on the graph
to help with interpretation. WAP finds the five
models that best fit the weather data of the current
season as indicated in Figure 3.
This program has a user-friendly spreadsheet
interface for data input and stores data in standard
DSSAT v3 and 3.5 ASCII format. The most recent
data of the current season are appended to the
historical weather datasets as indicated in Figure 4.
The program also exports five files, consisting of
the most recent weather data from the current
season combined with daily data for the five
analogue years that are used to complete the
season. The program can be run via the keyboard
or from an ASCII file used within a GIS framework.
CERES runs from a control file that describes the
experimental simulation trial. For South Africa,
such a trial consisted of six plantings at two-week
intervals in combination with super short, short,
medium and long growing season cultivars. All of
these were in combination with a low, medium,
and high plant population. Each of these
treatments was then run for all five of the analogue
weather files exported by WAP (Figure 5).
oatal Anaous O6a |l aBpdl
'- "-* I :- '
Figure 2. The current season in comparison with the two closest analogue years.
Pael eneou|l 9r0h Gild
Brest Sea1on |DdI m
S I I.- a1
Lo~a( r: rliY_
Figure 3. The five closest analogue years and weather statistics.
Season la alt | |
uI I : I. TM 1 I *': 'J L E'lF '1
J ul _, i',,'1 i: i -, 1 6 6" i- i "I ,i 1,
Jul ___I -I ______* l64 61 '1 ____* _
l u l i I 'I ". I i'l I 11 6 1: i I
Jul l lll UI l jI Ij Ill
Jul 7P I ll II.I I :II.II I
J.lj I1 i' I d I. I
Figure 4. Spreadsheet type interface for data input, indicating both calendar
date and Julian date.
Both programs have been included as tools for
DSSAT4 and will be distributed with the DSSAT4
software. For DSSAT3+ users the two programs can be
obtained via D.L. du Toit.
Figure 5. Flow diagram for the WAP crop growth simulation
Du Toit, A.S. 1996. Quantification of the compensation ability of the maize plant.
PhD thesis, UOVS, Bloemfontein, South Africa.
Flavella, P. 1992. A quantitative measure of model validation and its potential use
for regulatory purpose. Advances in Water Research 15: 5-13.
Harrison, S.R., 1990. Regression of a model on real-system output: An invalid test
of model validity. Agricultural Systems 34(3): 183-190.
Hoogenboom, G., J.W. Jones, P.W. Wilkens, W.D. Batchelor, W.T. Bowen, LA. Hunt,
N.B. Pickering, U. Singh, D.C. Odwin, B. Baer, K.J. Boote, J.T. Ritchie, and J.W.
White. 1994. Crop models. In G.Y. Tsuji, G. Uehara, and S. Balas (eds.),
DSSAT3. Vol. 2-2. Honolulu: University of Hawaii.
Jones, CA. and J.R. Kiniry. 1986. CERES-Maize: A simulation model for maize
growth and development. College Station: Texas A & M University Press.
Oreskes, N., K. Shrader-Frechette, and K. Belitz. 1994. Verification, validation, and
confirmation of numerical models in the earth sciences. Science 263: 641-646.
Ritchie, J.T. 1991. Specifications in the ideal model for predicting crop yields. In R.C.
Muchow and J.A. Bellamy (eds.), Climatic risk in crop production: Models and
management for the semi-arid tropics and sub tropics. Proc. International.
Symposium, St. Lucia, Brisbane, Queensland, Australia, July 2-6, 1990.
Wellingford, UK: CAB International. Pp. 97-122.
Savage, M.J. 1993. Statistical aspects of model validation. At Workshop on the field
water balance in the modelling of cropping systems, University of Pretoria,
Tsuji, G.Y., J.W. Jones, G. Hoogenboom, L.A. Hunt, and P.K. Thornton. 1994.
Introduction. In G.Y. Tsuji, G. Uehara, and S. Balas (eds.), DSSAT3. Vol. 1-1.
Honolulu, University of Hawaii.
Willmott, C.J. 1982. Some comments on the evaluation of model performance.
Bulletin of the American Meteorological Society 63: 1309-1313.
Both programs are distributed in 18 countries
worldwide and over 100 requests for the programs
have been received. WAP has been installed on a
number of farmers' and consultants' computers in
South Africa. It was also included in the yield
estimate methodology of ARC-GCI. MSP and WAP
have been included and released as utility software for
DSSAT 4. Both programs are available from either
author and are distributed free of charge. Please send
an email to either Deon@igg2.agric.za or
ADToit@mail.ifas.ufl.edu for more information.
Report of the GCTE Tropical Cereals
Network Inaugural Workshop1
A.S. Du Toit, John Ingram, and Jeffrey W. White
The inaugural workshop of the GCTE T. J. A Cereal
Network was held at CIMMYT Headquarters, Mexico
on Thursday 26-Friday 27 April 2001. It was convened
in association with the CIMMYT workshop Modeling
Temperature Pi p... I in Wheat and Maize held 23-25
The objectives of the GCTE T. *pi.;;:. Cereals
* To refine and adapt current crop production models
for tropical cereals for use in global change studies in
a wide variety of conditions.
* To design and undertake experiments to provide
improved mechanistic ,,1, i .11;. : of global
change impact on crop production, in order to aid in
The Network is envisioned to initially include studies
on maize, sorghum, and millet. Modelers and
experimentalists will be given equal importance in
achieving the objectives.
The primary objective of the GCTE workshop was to
establish the inaugural Network membership. To this
end interested scientists were invited to present
models and datasets suitable for model development.
Standardized descriptions (metadata) of models and
datasets, which will be included as ;, n., I l
contributions to the Network, will be collated and
published on the GCTE Focus 3 web site. The
.. ...1 1i set will be submitted to the GCTE Scientific
ft.... In_. Committee for endorsement as GCTE Core
The workshop organizers wish to acknowledge and
thank C1MMY'T for its support in convening and
hosting the workshop.
Thursday 25 April
Professor Tim Reeves (then Director General of
CIMMYT) opened the proceedings and indicated the
need to increase food supply in order to meet
projected demands in coming decades. He
emphasized the need to refine agricultural systems in
order to match the high levels of productivity seen in
developed countries without jeopardizing the natural
Tony Hunt presented a flow diagram showing where
this meeting fit within the wider GCTE framework. It
was agreed that the nascent GCTE Tropical Cereals
Network would be led by Andre du Toit, with Bill
Batchelor, Richard Vanderlip, and Upendra Singh
acting as coordinators for maize, sorghum, and millet,
respectively. These four, together with Jeff White,
1, p.1 .1"i: CIMMYT, would constitute the GCTE
T, I'p .,I Cereals V '... I I 1, Group.
Also published as GCTE .. Document 28. http:/ /mwnta.nmw.ac.uk/GCTEFocus3/networks/tropical.htm
Please see the GCTE web site I.': .... I. .. ..It .. ) for more information about GCTE Focus 3: Agroecology
and Production Systems, and about the Crop Networks in particular.
Session 1: Collation of models for maize,
sorghum, and pearl millet
An informal review of models available for the target
crops identified numerous examples, as summarized
in Table 1.
Session 2: Developing metadata for models
There is a need to document models and differences in
models, and it was agreed to base these efforts on the
format developed for the GCTE Wheat Network. An
additional field in the d ?i ?,..l, .i.t,, -., Model
Uff..,i.,n... is also needed to describe differences
between versions of the same model. Thus, CERES-
Maize would have a single metadata description while
variants of the model would be presented on a form
that stated the contact individual, and highlighted or
listed I.l .. between CERES-Maize and the
Session 3: Data availability and describing
Key individuals were identified as possible contacts
for data (Table 2).
Table 1. Crop simulation models identified dealing with maize, sorghum
or pearl millet.
Maize Sorghum Pearl Millet
F:- M.:.i, i,,i, release) CERES-Sorghum (IFDC) CERES-Millet (IFDC)
(Du Toit, Batchelor, o.:.j.ell. CERES-Grignon (Gabrielle) APSIM (Carberry)
.i ...... SORKAM (Vanderlip) EPIC/ALMANAC
CERES-Grignon (Roche) APSIM (G. Hammer)
CAMS-Maize v 2.0 (Wang Futang) WINSORG ,dM Mi:..)
i' ,i : ,,i.. EPIC/ALMANAC
:: ... GPFARM
Tom Sinclair's model RESCAP (S. Huda)
GPFARM :'. i, i. i
Stewart model i., :,,:li
EPIC .Wal s.,
Note: Names in parentheses indicate developers or key contacts.
Indicating data quality
A section for data describing .u .1 ;ii needs to be
added to the metadata files. Several entries will be
requested from the data contributors, along with
several entries from model testers.
* SD (standard deviation) for yield and other traits
* Mean for yield and other traits
* Comments on data ,i; i, from data collector
* Consistency of data: what are the three strongest and
weakest features of the data?
* Lack of model fit and causes for lack of fit
* Consistency among components (seasonal trends,
water content trends, etc.)
Session 4: Evaluating models using datasets
and error plots
Model Performance Using Actual Data
This analysis will be highly dependent upon what
data are contributed to the group. Four types of
validation are envisioned:
1. Development performance The purpose of this
analysis would be to test the model's ability to
capture the impact of climate on plant development.
Several datasets that represent diverse environments
Table 2. Possible sources of data for evaluating crop models.
Maize Sorghum Pearl Millet
(Batchelor, leader) (Vanderlip, leader) (Singh, temporary leader)
CERES list ICRISAT ICRISAT
(Batchelor) (S. Huda, Vanderlip) (S. Huda, Algarswamy)
CIMMYT (Risk Kansas (Vanderlip)1 Kansas
French data APSIM W. Africa
(Remain) (Hammer) (Kropff)
FACE Maize C02 European Network APSIM
experiment in 1 and 2
Illinois (Long) (Gabrielle, M. Martin)
New Zealand data INTSORMIL C02
(Jamieson) (Vanderlip) (Bidinger and Vanderlip)
CIMMYT Maize Program FACE (Kimball)
APSIM data? (Keeting) DSSAT data
(v. 3.5) (Singh)
1 Mainly US data
and that have good measurements of phenological
events would be selected. The model would be run
to determine the accuracy with which it can predict
these events. The graph generated as a result would
illustrate % error vs. measured phenological data
(i.e., anthesis date, etc.). The .. ti phenological
data selected would depend upon the data
measured in the dataset.
2. Growth performance The purpose of this analysis
would be to test the model's ability to capture the
impact of climate on plant growth. Several datasets
would be selected for analysis. These datasets would
represent diverse environments and would have
good measurements of growth characteristics,
i,,.:I7;l.,i final yield and biomass at harvest. After
running the model, a plot of % error vs. data (i.e.,
yield, biomass, etc.) would be generated.
3. Temporal performance -The purpose of this analysis
would be to test the model's ability to capture the
impact of climate on temporal plant growth. The
model would be run for several datasets that have
time series measurements of growth, and a plot of
predicted and measured data (i.e., leaf, stem, seed,
pod weights) vs. time would be .. 11. .1 .I for each
4. Water stress performance The 1 ... > of this
analysis would be to test the model's ability to
respond to water stress. The model would be run for
several datasets .' itii ;I is.t.ed and non-irrigated
treatments. Percent error plots would be generated
for any measured data.
Statistics such as RMSE, D, and r2 could be used as
indicators of accuracy for all four analyses.
Sensitivity Analysis to Climate Factors
The purpose of this analysis would be to demonstrate
the general response of key model outputs to changes
in climatic inputs, including daily temperature, CO2,
daylength, and drought.
1. Temperature A dataset would be selected and the
model would be run with the daily temperatures
found in the dataset. A sensitivity analysis would
then be conducted by ro...i. f .. the daily
temperature by + 5'C. Results to be plotted include
duration to flowering, duration from flowering to
physiological maturity, total biomass at harvest, and
yield as a function of daily temperature change.
2. Carbon dioxide A dataset would be selected and the
model would be run with the CO2 levels found in the
dataset. A sensitivity analysis would then be
conducted by iir. i:;l- ;, i:_ CO2 levels over a certain
range. Results to be plotted include duration to
fl... .. in;: duration from i -.... im, to physiological
maturity, total biomass at harvest, and yield as a
function of CO2.
3. i.. .'...;:: A dataset would be selected and the
model would be run with the weather data for a
given cultivar found in the dataset. A sensitivity
analysis would then be conducted by modifying
parameters that affect .i. 1. L ... li response while
changing the latitude in the weather file. Results to be
plotted include duration to fl. ., .,:_ duration from
)'ii .... .. to I'l, ;..)I ...i ,1 i n it., total biom ass at
harvest, and yield as a function of the .i.. i. w11,i and
4. Drought -A dryland dataset that shows (1 I i .1l
effects would be selected and the model would be
run with the weather data for a given cultivar found
in the dataset. A sensitivity analysis would then be
conducted by :i1 fia .,;,1 rainfall over a range of +
as well as i. Il. i;: a fully irrigated treatment.
Results to be plotted include duration to flowering,
duration from flowering to pl,-. ;... -.;. i-I maturity,
total biomass at harvest, and yield as a function of
cumulative rainfall or water available in the season.
Standard Output File
A standard output file could be created by each model
to facilitate developing the graphs discussed above.
This file would have the f.. 11, l,. i,. information:
* Anthesis date
* Physiological maturity date
* Biomass at maturity
* Level of the factor used for the sensitivity analysis
* Some identification of the dataset
Session 5: Model runs for different
experiments and sensitivity analysis
Datasets from Hawaii and South Africa were supplied
by Upendra Singh of IFDC and Andre du Toit of ARC-
GCI, respectively. These data represent tropical and
semi-arid environments and include low- to high-
water stress environments. It became evident from the
exercise that there is a need to revisit the definition on
physiological maturity and clarify any possible
confusion in this regard.
The simulations should be seen as a 1., t11. p1 -' and a
practical indication of possible problems. The
experiments took longer to set up than expected, and it
became evident while running the experiments that it
was quite easy to use the wrong soil and genotype input
files. While it is clear that more analysis will be needed
before it will be possible to indicate model differences,
the exercise proved valuable in launching the process
and identifying where more work is needed.
Friday 27 April
Session 6: Next steps/actions
Meta-database for crop h n i experiments
The meta-database should include a further question
(Q15),"What do you consider to be the strongest (and
weakest) points of this experiment/dataset?"
Respondents should follow the standard GCTE format
provided (GCTE Crop Networks: Metadata format for
crop growth experiments) and return information to
crop coordinators by 15 May. If a full report is not
feasible, at least a preliminary report could be sent.
Crop coordinators are to follow up with contributors
and/or the contacts for the datasets.
Potential contributors for metadata were identified as:
Maize (Leader Bill Batchelor)
CERES-Maize standard data (15)
Contributors/contacts for other datasets:
Jeff White, Peter Jamieson, Steve Long, Maria
Travasso, Romain Roche, B. I ..; J. Kiniry,
T. Tolenaar, Jorge Bolaios
Sorghum (Leader Richard Vanderlip)
CERES-Sorghum standard data (2)
Contributors/contacts for other datasets:
Richard Vanderlip, S. Huda, B. 1 l.lii.,,l G.
Hammer, C -..l. .. M. Martin, Scott Chapman
Millet (Leader Upendra Singh)
Contributors/contacts for datasets:
Richard Vanderlip, S. Huda, G. .1.i:.- ...,amy,
M. I ..ff F. FT;I-,l ,- APSIM.
Meta Database for Crop Growth Models
Respondents should follow the standard GCTE format
provided (GCTE Crop Networks: Metadata format for
crop growth models). For different versions of models,
or for models derived/developed from an existing
model, the appropriate contact person and the major
S.Ii f..,,n. -. from the original model should be
indicated. In addition, the model itself (in the form of
an .exe file with user instructions) should be provided
to crop coordinators for independent model runs, etc.
General Actions for Network and Crop Coordinators
* Invite other modelers to participate in the network
(i i, I.... I Coordinator)
* Send friendly reminders, provide pressure, etc. to
get datasets from contributors/contacts
* Send datasets that meet MDS requirements to
* Set established procedures for data and model
* Distribute results of model performance, sensitivity
analyses, and .1.ildi. checks among members
* Discuss above pg .-... --. in approximately 12 months
Topics for Next Meeting
* Formalizing the structure of network with all
p. :i ;.:; i members
* Constraints to I I .
* Examination of procedures for ',.. .. model
* Gaps in data coverage
* Funding of data collection when there is no member
in a i. .
* Funding for less popular crops (ie, millet)
* Funding of members from d .....1-.. i'-. countries
A few points were noted regarding potential Network
* Membership would be comprised of volunteer/
invited experimenters and modelers who value the
scientific interactions and relationships that develop
in the Network. These activities would offer
S.,., tc,,,o 4.- for publications.
* P, 1. voluntary, the priority given to Network
activities may be low, and deadlines may not be met
* The long-term membership will comprise of a highly
motivated: q. -' i that views the i.. 1, 1It as
outweighing the additional time commitment.
Session 7: Outcomes from GCTE Tropical
Cereals Working Group meeting
The indications are that the GCTE Tropical Cereals
Network will initially consist of about 15 members.
Sp-t.. :, : actions include:
1. Andre du Toit will set up a web page to host the
different datasets. The web page will be password
2. Jeff White will put a map together to indicate the
1' ii distribution of the different data points.
3. The next workshop was to be held at CIMMYT
within 12 months time, with ICRISAT as an option
for a meeting venue in 2003. However, due to
changes in GCTE Focus Group 3 mi .-i,_ ,. t the
2002 workshop was not held.
4. The next meeting will complete the present work
and start looking at tillage-climate interactions.
Participants: See combined list for CIMMYT Modeling
Workshop and GCTE Workshop.
Note: Full details on how to apply to join the gcte tropical
crops network, as well as downloadable forms, can be found
on the GCTE Focus 3 web site: http://mwnta.nmw.ac.uk/
Participants and Contact Information
A 570, Plant and Soil Sciences Building
Department of Crop and Soil Sciences
East il, ,::m Ill 48824-1325
Fax: +1 517-355-0270
William D. Batchelor
Agricultural and %.., .I-",.. CIE ..:.:".'
219b Davidson Hall
Iowa 0:'i, i." *,f
Ames, IA 50011
Tel.: +1 515-294-9906
Fax: +1 515-294-2552
v;;, ..lr ;.. .., be,) .... -.. ,, ,,Ju
Andre Sim6n du Toit
Agricultural Research Council
Summer Grain Centre
Private Bag X1251
Tel.: +27 018-299-6249
103 Frazier Rogers Hall, Room 259
Agricultural and Biological Engineering
P.O. Box 110570
Gainesville FL, 32611-0570
Richmond, K29 Bldg
Locked Bag 1797
Penrith South DC NSW 1797
Tel.: +61 2-4570-1390
Fax: +61 2-4570-1750
L. Anthony Hunt
Plant Agriculture, Crop Science Division
University of Guelph
Guelph, ON NIG 2W1
Tel.: +1 519-824-4120 ext. 3595/2414
Fax: +1 519-763-8933
"* li.;I ih .. .., .r. l l... ,
NERC Centre for Ecology and Hydrology
wv;:i.,,i...,J OX10 8BB
Tel.: +44 1491-692-410
Web site: mwnta.nmw.ac.uk/GCTEFocus3
New Zealand Institute for Crop & Food
Private Bag 4704
Tel.: +64 3-325-6400/
+64 3-325-6401 Ext. 3606
Fax: +64 3-325-2074
4 1. 1, .. ... p :',
Agricultural and .- :. -. rir .' .1
219b Davidson Hall
Iowa State University
Ames, IA 50011
Tel.: +1 515-294-7350
Stephen P. Long
-:; ;-,.n.. n: ..fI "..; Science & Plant
University of Illinois
190 Edward R. Madigan Laboratories
1201 '-. .-'.., .,' Drive
Urbana, IL 61801-3838
Fax: +1 217-244-7563
(Dept. +1 217-244-1336)
Great Plains Systems Research Unit
2150 Centre Avenue, Building D
Fort Collins, CO 80526
Tel.: +1 970-492-7340
Em ail: (.! 'i !.'- l i i .: .... i .l ,
Web site: http://gpsr.ars.usda.gov
University of Hawaii
1955 >.: I W. .I i:..,,i: Room 206
Dept. of Tropical Plant and Soil Science
Honolulu, HI 96822
Missoula, MT 59812
Tel.: +1 406-243-6228
,i ,;.I1 ,,:, -10.11i ;.n ;i .
,;e ...i ,,, .. -. ..,: : ., ,.:.g '
j .i...,, n ..i '
F78850 TI,: r.1 ,.,,1 ,
Tel.: +33 1-3081-5506
Fax: +33 1-3081-5563
% ,n,;.I ,,,. I;,:, I. "l.,; ,,:, .,....,; ,,",,; fr
Int. Fertilizer Development Center
P.O. Box 2040
Muscle Shoals, AL 35662
Tel.: +1 256-381-6600
Fax: +1 256-381-7408
Maria I. Travasso
Institute de Clima y Agua
CNIA- CASTELAR 1712
Tel.: +54 11-4621-0125
Fax: +54 11-4621-5663
: n I ., Wll h n I: .,, I'f-
2004 Throckmanton Plant Science Center
Manhattan, KS 66506-5501
Tel.: +1 785-532-7248
Fax: +1 785-532-6094
Chinese Academy .l .:. :, ..l..i .. .i ......
Zhongguancunnandalia No.46, 100081
Tel.: +86 10-6840-7046 (Office)
Tel.: +86 10-6840-8395 (Home)
Fax: +86 10-6217-5931
Jeffrey W. White
Natural Resources Group
J,. P,.:, l .',I ,i
06600 Mexico D.F.
Tel.: +52 5804-2004
Fax: +52 5804-7558 and 5804-7559
Environmental and Plant Dynamics Research
U.S. '..,,1I, ,' ; I,-,. --',.;.. ... USDA-ARS
4331 E Broadway Rd.
Phoenix, AZ 85040-8834
Tel.: +1 602-437-1702 ext. 268
Fax: +1 602-437-5291
Paul W. Wilkens
Int. Fertilizer Development Center
P.O. Box 2040
Muscle Shoals, AL 35662
Tel.: +1 256-381-6600
Fax: +1 256-381-7408