The EPIC Crop Growth Model

J. R. Williams, C. A. Jones, J. R. Kiniry, D. A. Spanel

ABSTRACT include weather simulation, hydrology, erosion-

he EPIC plant growth model was developed to sedimentation, nutrient cycling, crop growth, tillage, soil
T estimate soil productivity as affected by erosion
throughout the U.S. Since soil productivity is expressed
temperature, economics, and plant environment control.
Several of these components have been described
in terms of crop yield, the model must be capable of elsewhere (Williams et al., 1984; Jones et al., 1984a,
simulating crop yields realistically for soils with a wide 1984b; Jones, 1985; Sharpley et al., 1984).
range of erosion damage. Also, simulation of many crops Only one of the model components, crop growth, is
is required because of the wide variety grown in the U.S. described here. Since soil productivity is expressed in
EPIC simulates all crops with one crop growth model terms of crop yield, crop growth is one of the most
using unique parameter values for each crop. The important processes simulated by EPIC. To evaluate the
processes simualted include leaf interception of solar effect of erosion on crop yield, the model must be
radiation; conversion to biomass; division of biomass sensitive to crop characteristics, weather, soil fertility,
into roots, above ground mass, and economic yield; root and other soil properties. The processes simulated
growth; water use; and nutrient uptake. The model has include crop interception of solar radiation; conversion
been tested throughout the U.S. and in several foreign of intercepted light to biomass; division of biomass into
roots, above-ground biomass, and economic yield; root
countries.
growth; water use; and nutrient uptake. Potential plant
growth is simulated daily and constrained by the
INTRODUCTION
minimum of five stress factors (water, nitrogen,
Recently, a mathematical model called EPIC (Erosion-
phosphorus, temperature, and aeration). Root growth is
Productivity Impact Calculator) was developed to
constrained by the minimum of three stress factors (soil
determine the relationship between soil erosion and soil
strength, temperature, and aluminum toxicity). Through
productivity in the U.S. (Williams et al., 1984). EPIC is
its effects on soil properties and plant and root growth
composed of physically based components for simulating
stress factors, erosion affects crop production indirectly.
erosion, plant growth, and related processes. It also
includes economic components for assessing the cost of EPIC simulates all crops with one crop growth model
erosion, determining optimal management strategies, using unique parameter values for each crop. EPIC is
etc. The physical processes involved are simulated capable of simulating crop growth for both annual and
simultaneously and realistically using readily-available perennial plants. Annual crops grow from planting to
inputs. Since erosion can be a relatively slow process, harvest date or until the accumulated heat units equal
EPIC is capable of simulating hundreds of years if the potential heat units for the crop. Perennial crops
necessary. (alfalfa, grasses, and pine trees) maintain their root
The EPIC model was used to analyze the relationships systems throughout the year, although the plant may
among erosion, productivity, and fertilizer needs as part become dormant after frost. They start growing when the
of the Soil and Water Resources Conservation Act (RCA) average daily air temperature exceeds the base
analysis for 1985. EPIC provided erosion-productivity temperature of the plant.
relationships for about 900 benchmark soils and 500,000 The EPIC crop table contains parameter values for
crop/tillage/conservation strategies throughout the U.S. soybeans, alfalfa, corn, grain sorghum, wheat, barley,
Thus, the model had to be generally applicable, oats, sunflower, cotton, pasture, peanuts, potatoes,
computationally efficient, and capable of computing the Durham wheat, winter peas, faba beans, rapeseed,
effects of management decision. Model components sugarcane, sorghum hay, range grass, rice, cassava,
lentils, and pine trees.

Article was submitted for publication in March, 1988; reviewed and

MODEL DESCRIPTION
approved for publication by the Soil and Water Div. of ASAE in The framework of an early version of the EPIC crop
November, 1988. model was described previously (Williams et al., 1984).
Contribution from the USDA-Agricultural Research Service, in
cooperation with Texas Agricultural Experiment Station, Texas A&M
The model has been modified, expanded, and tested
University System. extensively since that time. Thus, the objectives here are
The authors are: J.R. WILLIAMS, Hydraulic Engineer, USDA- to describe in detail the 1988 version of the EPIC crop
Agricultural Research Service; C.A. JONES, Professor, TAES; J.R. model and to present test results for several crops and
KINIRY, Research Agronomist; and D.A. SPANEL, Biological locations.
Technician, USDA-Agricultural Research Service, Temple, TX.
Acknowledgment: The measured data represent work by researchers
from several disciplines in several countries (Table 1). Their willingness Phenological development
to share this information is deeply appreciated. The phenological development of plants is accelerated

Vol. 32(2):March-April, 1989 497

by warm temperatures. For example, high temperatures t•ha~^ BE is the crop parameter for converting energy to
shorten times to emergency (Angus et al., 1981), to biomass in kg-ha-MJ-^-m^, HRLT is the daylength in h,
anthesis (Tompsett, 1976), and the grain filling period and AHRLT is the change in daylength in h-d^^ The
(Balasko and Smith, 1971). Several indices have been daylength function of equation [4] increases potential
proposed to quantify the effects of temperature on growth during the spring and decreases it in the fall. It
development rate (Gilmore and Rogers, 1958; Coelho represents an attempt to mimic the observed, though
and Dale, 1980). poorly understood, effects of rate of change of daylength
In EPIC, phenological development of the crop is on plant growth (Baker et al., 1980).
based on daily heat unit accumulation. It is computed Daylength is a function of the time of year and latitude
using the equation as expressed in the equation

'^mx,K "^'^m n , K \ . 27r

HUt ^b,J HUK>0 -sin( LAT) sin(SD). - 0.044
<
HRLT: = 7.64 cos-1 360 '
[1] cos( LAT)cos(SD)-
360 '
where HU, T^.^, and T^^^ are the values of heat units, .[5]
maximum temperature and minimum temperature in °C
for any day K, and T^ is the crop-specific base where LAT is the latitude of the watershed in degrees
temperature in °C (no growth occurs at or below T^) of and SD, the sun's declination angle, is defined by the
crop j . A heat unit index (HUI) ranging from 0 at equation
planting to 1 at physiological maturity is computed as
follows.
SDi = 0.4102 sin/"— (1-80.25)") [6]

HUI-
(i)
PHU:
HU.
.[2]
In most crops, leaf area index (LAI) is initially zero or
very small. It increases exponentially during early
vegetative growth when the rates of leaf primordia
development, leaf appearance, and blade expansion are
where HUI is the heat unit index for day i and PHU is the linear functions of heat unit accumulation (Tollenaar et
potential heat units required for maturity of crop j . The al., 1979; Watts, 1972). In vegetative crops such as
value of PHU may be provided by the user or calculated sugarcane and some forages, LAI reaches a plateau at
by the model from normal planting and harvest dates. which senescence and growth of leaf area approximately
Date of harvest, leaf area growth and senescence, equal. In many crops, LAI decreases after the maximum
optimum plant nutrient concentrations, and partition of LAI is reached and approaches zero at physiological
biomass among roots, shoots, and economic yield are maturity. In addition, leaf expansion, final LAI, and leaf
affected by HUI. duration are reduced by stresses (Acevedo et al., 1971;
Eik and Hanway, 1965).
Potential growth LAI is simulated as a function of heat units, crop
Interception of solar radiation is estimated with a stress, and crop development stages. From emergence to
Beer's law equation (Monsi and Saeki, 1953) the start of leaf decline, LAI is estimated with the
equations
PARi = 0.5 (RA)i [1. - exp(-0.65 LAIjl^ .[3]
LAIi = LAIi_i + ALAI [7]
where PAR is intercepted photosynthetic active radiation
in MJ-m~^ RA is solar radiation in MJ-m~2, LAI is the
leaf area index, and subscript i is the day of the year. The ALAI = (AHUF)(LAI^^) ( L - exp(5.(LAIi_i
constant, 0.5, is used to convert solar radiation to
photosynthetically-active radiation (Monteith, 1973).
Experimental studies indicate that the extinction •LAI^x)) ^VREGi .[8]
coefficient varies with foliage characteristics, sun angle,
row spacing, row direction, and latitude (Thornley, where LAI is the leaf area index, HUF is the heat unit
1976). The value used in EPIC (0.65) is representative of factor, and REG is the value of the minimum crop stress
crops with narrow row spacings (Uchijima et al., 1968). factor discussed in more detail below. Subscript mx is
A somewhat smaller value (0.4-0.6) might be appropriate the maximum value possible for the crop and A is the
for tropical areas in which average sun angle is higher daily change. The exponential function of equation [8]
and for wide row spacings (Begg et al., 1964; Bonhomme prevents LAI from exceeding LAI^.^ when HUF is
et al., 1982; Muchow et al., 1982). Using Monteith's adjusted for vernalization of certain crops. The heat unit
approach (Monteith, 1977), potential increase in factor is computed using the equation.
biomass for a day can be estimated with the equation HUI,
HUF: .[9]
ABp i = 0.001 (BE)j(PAR)i(l + AHRLTi)^ [4] ' HUIiH-exp(ah.^l-(ahj^2)(HUIi))

where ABp is the daily potential increase in biomass in Two pairs of values for HUF and HUI are specified as

498 TRANSACTIONS of the ASAE

crop parameters and determine the sigmoid relationship The economic yield of most grain, pulse, and tuber
described by equation [9] for crop j . The parameters ahj, crops is a reproductive organ. Crops have a variety of
and ahj2 are calculated by simultaneous solution of mechanisms which insure that their production is neither
equation [9] using the two pairs of values for H U F and too great to be supported by the vegetative components
HUI. nor too small to insure survival of the species. As a result,
From the start of leaf decline to the end of the growing harvest index (economic yield/above-ground biomass) is
season, LAI is estimated with the equation often a relatively stable value across a range of
environmental conditions. In EPIC, crop yield is
/ 1. - HUI: \ ad; estimated using the harvest index concept.
LAL = LAI ( ) ^ [1 0]
° V L - HUI. / [15]
YLDj = ( H I J ) ( B A G )
where ad is a parameter that governs LAI decline rate for
where YLD is the amount of economic yield that could be
crop j and HUI^ is the value of H U I when LAI starts
removed from the field in t-ha~S HI is the harvest index,
declining.
and BAG is the above-ground biomass in t-ha~^ for crop j .
Crop height is estimated with the relationship
For non-stressed conditions, harvest index increases non-
linearly from zero at planting to HI at maturity using the
CHTi = HMXj VHUFi [11] equation
where CHT is the crop height in m and H M X is the
maximum height for crop j .
The fraction of total biomass partitioned to root HIAi = HIj ^ S AHUFHj^j [16]
system normally decreases from 0.3 to 0.5 in the seedling
to 0.05 to 0.20 at maturity (Jones, 1985). The model
where HIA is the harvest index on day i and H U F H is the
estimates the fraction of crop growth that is partitioned
to the root system to range linearly from 0.4 at heat unit factor that affects harvest index.
emergency to 0.2 at maturity. Thus, the potential daily The harvest index heat unit factor is computed with
change in root system weight is computed with the the equation
equation
HUI,
HUFH, [17]
ARWT- = ABp i(0.4 - 0.2 HUI^) [12] HUIj + e x p ( 6 . 5 0 - 1 0 . 0 H U I i )
where ARWT is the change in root weight in t-ha~^ on
day i. The distribution of root growth through the root The constants in equation 17 are set to allow HUFHj to
zone is simulated as a function of plant water use in each increase from 0.1 at HUIi==0.5 to 0.92 at H U I i = 0 . 9 .
layer of soil using the equation This is consistent with economic yield development of
grain crops which produce most economic yield in the
U;i,C second half of the growing season. The effects of stresses
ARWi g = (ARWTi) [13] on harvest index are discussed in the section on growth
M
Z u i,C constraints.
2=1

where RW is the root weight in soil layer i in t-ha~S M is Water Use

the total number of soil layers, and u is the daily water The potential water use, Ep, is estimated as a fraction
use in layer i in m m . of the potential evaporation using the leaf-area-index
As discussed in the section describing water use, this relationship developed by Ritchie (1972).
method of estimating root growth produces realistic
exponential decreases in root weight with depth when soil LAL
water and other properties do not constrain growth. Eoi(-r-)' Epi<Eo [18]
-Pi
When a soil layer is dry or root stress factors (strength,
aluminum saturation, or aeration) limit root function,
both water use and root growth in the layer are reduced. where E^ is the potential evaporation and LAI is the leaf-
Rooting depth normally increases rapidly from the area-index on day i.
seeding depth to a crop-specific maximum. In many The potential water use from the soil surface to any
crops, the maximum is usually attained well before root depth is estimated with the function
physiological maturity (Borg and Grimes, 1986). Rooting
depth is simulated as a function of heat units and
potential root zone depth. ^Pi
UPi (l--exp(-A(^))) [19]
1. - exp(-A)
RDi = 2.5(RDMXj)(HUIi), RD- < RZj [14]
where Up is the total water used in m m to depth Z in m
where RD is the root depth in m, R D M X is the on day i, RZ is the root zone depth in m, and '^ is a water
maximum root depth for crop j in ideal soil in m, RZ is use distribution parameter. The amount used in a
the soil profile depth in m, and the constant 2.5 allows p a r t i c u l a r layer can b e calculated by t a k i n g t h e
root depth to reach its maximum when HUI reaches 0.4. difference between Upj values at the layer boundaries.

Vol. 32(2):March-April, 1989 499

 -Pi FCn - WPo
^PE l._exp(-A(—)) Un = U P C
1. - exp(-A) sw,> 4.
+ WPg [24]

where SW is the soil water content in layer i on day i in

( -
exp(-A(-
l^'O) [20] mm and FC and WP are the soil water contents at field
capacity and wilting point for layer i .
where Up^ is the potential water use from layer I in mm. Nutrient uptake
Equation 20 represents a soil that provides poor Nitrogen
conditions for root development when A is set a high
Crop use of N is estimated using a supply and demand
value like 10. The high A value gives high water use near
approach. The daily crop N demand is the difference
the surface and very low use in the lower half of the root between the crop N content and the ideal N content for
zone. Since there is no provision for water deficiency day i. The demand is estimated with the equation
compensation in any layer, considerable water stress may
result using equation [20]. To overcome this problem, i=l
equation [20] was modified to allow plants to compensate UNDi = (cNB)i(B)i 2 UN, [25]
K=l *
for water deficiency in a layer by using more water from
other layers. Total compensation can be accomplished by where UND is the N demand of the crop in kg-ha~^ UN
taking the difference between Upj at the bottom of a layer is the actual N uptake in kg-ha~S C^B is the optimal N
and the sum of water use above a layer. concentration of the crop in kg•t~^ and B is the
accumulated biomass in t-ha"^ for day i. The optimal
•-Pi crop N concentration declines with increasing growth
^P£ (l.-exp(-A(^))) stage (Jones, 1983a) and is computed as a function of
1. -exp(-A)
growth stage using the equation
C-1
[21] ^NBi = ^ ^ 1 + ^ ^ 2 e x p ( - b n 3 HUI^) [26]
K=l
where HUI (heat unit index) is the fraction of the
growing season and bnj, bn2, and bnj are parameters
where u^ is the actual water use in mm for all layers
calculated from crop-specific concentrations of N in the
above layer i. Thus, any deficit can be overcome if a
plant at the seedling stage, halfway through the season,
layer is encountered that has adequate water storage.
and at maturity.
Neither equation [20] (no compensation) nor equation
Mineral nutrients move to plant roots primarily by
[21] (total compensation) is satisfactory to simulate a
mass flow and diffusion. Mass flow is the movement of
wide range of soil conditions. A combination of the two
nutrients to roots in the soil water absorbed and
equations, however, provides a very general water use
transpired by the plant. Mass flow rarely provides exactly
function.
the amount of nutrient absorbed by the plant; therefore,
the concentration of nutrient near the root may increase
•-Pi
-^pe 1. - exp(-A ( or decrease in response to the balance between mass flow
1. - exp(A) and absorption (Barber, 1984). Diffusion causes the
nutrient concentration gradient between the root to the
e-i
(1. - UC) ( . . - , .p(-A( •UC i: XXyr bulk soil to decrease, and when soil solution
K=l ^ concentrations are low it can provide a significant
fraction of N absorbed by the plant. Barber (1984)
[22] suggests that mass flow normally accounts for about
80% of the N uptake of corn roots, and a combination of
where UC ranges (0.0 to 1.0) and is the water deficit mass flow and diffusion can reduce soil NO3—N to very
compensation factor. In soils with a good rooting low levels.
environment, UC = 1. gives total compensation. The In EPIC, mass flow of NO3—N to the roots is used to
other extreme, poor conditions, allow no compensation distribute potential N uptake among soil layers
(UC = 0.). The procedure for estimating UC is described
in the Growth Constraints section. WN03o
The potential water use in each layer (neglecting the U N o ,i == ^c,i(- [27]
effects of root constraints) calculated with equation [22] SWn
is reduced when the soil water storage is less than 25% of
plant-available soil water (Jones and Kiniry, 1986) using where UN is the amount of N supplied by the soil in
the equation. kg-ha-S WN03 is the amount of NO3-N in kg-ha-^,
SW is the soil water content in mm, u is water use in mm,

f( 4.(SWBi-WPg) and subscript i refers to the soil layers. The total N

Ug = Upg e x p available for uptake by mass flow is estimated by
(FC,-WP,) •
) ) summing mass flow of all layers.
FCg-WPg M
SWgi< + WPo [23] UNS:= 2 UN,fi,i [28]
K=l

500 TRANSACTIONS of the ASAE

 Since mass flow rarely provides the exact amount of N WN03
FXN= 1 . 5 - 0 . 0 0 5 ( 100.<WNO3<300.
required by the crop, UN values obtained from equation RlT
[27] are adjusted. [38]
UND-
UNaei=UNgi ( UNag i<WN03e^i. . . [29] FXN = 1.0, WN03 < 100. kg-ha-l-m-l ,[39]
UNS:

where UNa is the adjusted N supply in kg-hg-^ for layer Where W N 0 3 is the weight of NO3—N in the root zone
L Equation 29 assures that actual N uptake cannot in kg-ha-^ and RD is the root depth in m. This approach
exceed the plant demand when mass flow exceeds reduces N fixation when the NO3—N content of the root
demand. It also provides for increased N supply from the zone is greater than 100 kg-ha"^ and prohibits N fixation
layer (by diffusion) when mass flow does not meet crop at N contents greater than 300 kg-ha"^
demand but NO3—N is available in the soil.
Phosphorus
Nitrogen Fixation: Daily N fixation is estimated as a Crop use of P is estimated with the supply and demand
fraction of daily plant N demand by legumes. approach described in the N model. The daily plant
demand is computed with equation [25] written in the
WFXi = FXRi • UND^, WFX < 6.0 [30] form

where WFX is the amount of N fixation in kg-ha~^ and

FXR is the fraction of uptake for day i. The fraction, UPDi = (cpB)i(B)i- 2 UPK [40]
K— 1
FXR, is estimated as a function of soil NO3 and water
content and plant growth stage.
where UPD is the P demand for the plant in kg-ha~S U P
is the actual P uptake in kg-ha~S and Cpg is the optimal
FXR = min (1.0, FXW, FXN) • FXG [31] P concentration for the plant. As in the case of N, the
optimal plant P concentration is computed with equation
where FXG is the growth stage factor, FXW is the soil
[26] in written in the form
water content factor, and FXN is the soil NO3 content
factor. The growth stage factor inhibits N fixation in
CpBi = ^Pl + t)p2 exp(-bp3 HUIj) [41]
young plants prior to development of functional nodules
and in old plants with senescent nodules (Patterson and
where bpi, bp2, and bp3 are parameters calculated from
LaRue, 1983).
crop-specific optimum P concentrations at the seedling
stage, halfway through the season and at maturity. Soil
FXGi = ^-0' ^U^i ^ ^-15, HUI > 0.75 . . . . [32] supply of P is estimated using the equation

FXGi = 6-67 HUIi - 1.0, 0.15 < HUIi < 0.3 . . . [33] M RWo
UPS: 1.5 UPDi X (LF^)e ( [42]
FXG- = 1.0, 0.3 < HUIi < 0.55 [34]
where UPS is the amount of P supplied by the soil in
FXGi = 3.75 - 5.0 HUIi, 0.55 < HUIi < 0.75 . . [35] kg'ha~\ LFu is the labile P factor for uptake, RW is the
root weight in layer i in kg•ha~^ and RWT is the total
root weight on day i in kg-ha~^ The constant 1.5 allows
where HUI is the heat unit index for day i. The soil water
2 / 3 of the roots to meet the P demand of the plant if
content factor reduces N fixation when the water content
labile P is not limiting. This approach is consistent with
at the top 0.3 m is less than 85% of field capacity
studies suggesting that roots of P-deficient plants (or
(Albrecht et al., 1984; Bouniols et al., 1985) using the
plants whose root systems have been pruned) can absorb
equation
P faster than the roots of normal plants (Andrews and
Newman, 1970; DeJager, 1979; Jungk and Barber,
SW3i - WP3 1974).
FXU^ =
0.85(FC3-WP3) The labile P factor for uptake ranges from 0.1 to 1.0
according to the equation
SW3 < 0.85 (FC3 - WP3) + WP3 [36]
0.9 c LPfi
L F , , = 0.1 + . . . . [43]
where SW3, WP3, and FC3 are the water contents in the CLPC + 117. exp(-0.283 CLPC)
top 0.3 m of soil on day i, at wilting point, and at field
capacity. where CLP is the labile P concentration in soil layer i in
The amount of NO3 in the root zone can affect N g-t~^. Equation 43 allows optimum uptake rates when
fixation (Harper, 1976; Bouniols et al., 1985) determines CLP is above 20 g-t~^ This is consistent with critical labile
the soil NO3 factor, FXN. P concentrations for a range of crops and soils (Sharpley
et al., 1989). Sharpley et al. (1984, 1985) described
methods of estimating CLP from soil test P and other soil
FXN = 0., WN03 > 300. kg-hsT^-m-^ [37] characteristics.

Vol. 32(2):March-April, 1989 501

 Growth Constraints optimal values. The stress factors vary non-linearly from
Potential crop growth and yield are usually not 1.0 at optimal N and P contents to 0. when N or P is half
achieved because of constraints imposed by the plant the optimal level (Jones, 1983a). In the case of N, the
environment. The model estimates stresses caused by scaling equation is
water, nutrients, temperature, aeration, and radiation.
These stresses range from 0.0 to 1.0 and affect plants in
several ways. In EPIC, the stresses are considered in 2 UN K
K=l
estimating constraints on biomass accumulation, root SNs,i = 2 [47]
growth, and yield. The biomass constraint is the (cNB)i(S)i
minimum of the water, nutrient, temperature, and
aeration stresses. The root growth constraint is the where SNs is a scaling factor for the N stress factor, C^B is
minimum of soil strength, temperature, and aluminum the optimal N concentration of the crop on day i, B is the
toxicity. Though topsoil aluminum toxicity can have a accumulated biomass in kg-ha~S and UN is the crop N
direct effect on shoot growth, EPIC simulates only its uptake on day K in kg-ha-^. The N stress factor is
indirect effects through its inhibition of root growth and computed with the equation
water use. A description of the stress factors involved in
determining each constraint follows. SN S,i
SN: = 1 - - [48]
SNg i + exp(3.39 - 10.93 SNg ^j
Biomass
The potential biomass predicted with equation [4] is where SN is the N stress factor for day i. The P stress
adjusted daily if any of the five plant stress factors is less factor is computed with equations [47] and [48] written
than 1.0 using the equation in P terms.
Aeration Stress: When soil water content approaches
AB= (ABp) (REG) [44] saturation, plants may suffer from aeration stress. The
water content of the top 1 m of soil is considered in
where REG is the crop growth regulating factor (the estimating the degree of stress.
minimum stress factor).
Water Stress: The water stress factor is computed by SWI
considering supply and demand in the equation SAT = CAR [49]
POI
M

.^1""-'
WSi = [45] SAT
AS:= 1.-- -, SAT > 0.0
-Pi SAT + exp(-1.291 - 56.1 SAT)

where WS is the water stress factor, u is the water use in [50]

layer i, and Ep is the potential plant water use on day i.
This is consistent with the concept that drought stress where SAT is the saturation factor, SWl is the water
limits biomass production in proportion to transpiration content of the top 1 m of soil in mm, POl is the porosity
reduction (Hanks, 1983). of the top 1 m of soil in mm, CAP is the critical aeration
Temperature Stress: The plant temperature stress is factor for crop j (ssO.85 for many crops), and AS is the
estimated with the equation aeration stress factor. This approach allows the model to
restrict crop growth both when water tables are high (but
a layer of aerated soil occurs near the surface) and when
TSrsine(^t^-i:^)), 0 < TS, < 1 [46] slow internal drainage causes poor aeration near the soil
surface. Several studies suggest that when water-filled
pore space (WFP) exceeds 60% (Linn and Doran, 1984;
where TS is the plant temperature stress factor Tg is the Grable and Seimer, 1968; Trouse, 1964), root growth is
average daily soil surface temperature in °C, T^ is the inhibited by poor aeration. EPIC produces similar
base temperature for crop j , and T^ is the optimal results when CAF = 0.85, WFP exceeds 60% in the
temperature for crop j . Equation [46] produces surface 0.5 m, and the soil is saturated from 0.5 to 1.0 m.
symmetrical plant growth stress about the optimal Growth of flood-tolerant crops like rice can be simulated
temperature and it is driven by average daily soil surface by setting CAF = 1.0.
temperature. This approach allows growth of small Finally, the value of REG is determined as the
plants to respond realistically to low soil surface minimum of WS, TS, SN, SP, and AS.
temperatures found in temperate regions in the spring.
The presence of soil residues can retard simulated soil Root Growth
warming and reduce crop growth. As the crop canopy As described in equation [13], root growth is
develops, it shades the soil surface, and simulated proportional to water use. Water use from a soil layer is
average soil surface temperature approaches average air estimated as a function of soil depth, water content, and
temperature. A more detailed description of the soil a compensation factor using equation [22] and [23]. Soil
temperature model is given in Williams et al. (1984). strength, temperature, and aluminum toxicity stress
Nutrient Stress: The N and P stress factors are based factors are calculated from soil properties. The
on the ratio of simulated plant N and P contents to the minimum of these three stresses, the root growth stress

502 TRANSACTIONS of the ASAE

factor, constains root growth by governing the water use Aluminum ( A l ) toxicity can limit root growth in some
compensation factor. acid soil layers, and A l saturation is a widely used index
Cold soil temperature may limit root growth, of its effects (Abruna et al., 1982; Brenes and Pearson,
especially when subsoil layers warm slowly in the spring 1973; Pavan et al., 1982). Because crops and cultivars
(Taylor, 1983). Temperature stress for each soil layer is differ in their sensitivities to A l toxicity (Foy et al., 1972,
computed by substituting soil temperature at the center 1974; Mugwira et al., 1980), EPIC considers both the A l
of the layer for soil surface temperature in equation [46]. toxicity of the soil and crop sensitivity to it. Root growth
Numerous studies have shown that root growth is stress caused by aluminum toxicity is estimated with the
affected by soil strength. Three important strength equation
determinants are bulk density, texture, and water
content (Eavis, 1972; Monteith and Banath, 1965; [57]
ATS,= ALSg > ALOj
Taylor et al., 1966). All three of these variables are 100-ALO:
considered in estimating the EPIC soil strength stress
factor using the following equation.
ATSn = 1 . 0 , ALSg < ALOj [58]
0.9 BDn
SSg = 0.1 + [51]
BDg + exp(bti +bt2(BDg)) where ATS is the aluminum toxicity root growth stress
factor for soil layer i , ALS is the aluminum saturation in
where SS is the soil strength factor in layer i , BD is the % , and ALO is the maximum ALS value crop j can
soil bulk density adjusted for water content in t-m~^ and tolerate without stress in % . Crop specific values of ALO
bt, and bt2 are parameters dependent upon soil texture. are determined from the equation
T h e values of bt^ a n d bt2 are o b t a i n e d from a
simultaneous solution of equation [51] by substituting ALOj = 10 + 20(ALTj- 1) [59]
boundary conditions for stress. The lower boundary
where essentially no stress occurs is given by the equation where ALTj is the aluminum tolerance index number for
(Jones, 1983b) crop j . Values of ALT range from 1 to 5 (1 is sensitive; 5
is tolerant) for various crops. Finally, the root growth
BDL = 1.15 + 0.00445 SAN [52] stress factor, R G F , is the minimum of SS, ATS, and TS.

where BDL is the bulk density near the lower boundary Water use
(SS = 1.) for a particular percent sand, SAN. The upper Plant water use is governed by the root growth stress
boundary is given by equation (Jones, 1983b) factor using the water deficit compensation factor of
equation [22]. Recall that the water deficit compensation
BDU= 1.5+ 0.05 SAN [53] factor, UC, allows total compensation if the value is 1.0
and no compensation at 0.0. The value of UC for any
where BDU is the bulk density near the upper boundary layer is estimated as the product of the root growth stress
(SSJ5:0.2) for a particular percent sand, SAN. The factors for the layer and all layers above.
equations for estimating bt^ and bt2 are

UCo 2 RGF [60]

ln(0.112 BDL) - ln(8. BDU) K
bto [54] K=l
BDL-BDU
Thus, a low R G F K greatly reduces water compensation
bt^ = ln(0.0112 BDL) - (bt2)(BDL) [55] for layer K and all layers below K. The final estimates of
water use for each layer are obtained by multiplying the
equations [23] and [24] u^ values by R G F .
Equations [54] and [55] assure that equation [51] gives
SS values of 1.0 and 0.2 for BD equal BDL and BDU u*e = (ug)(RGF)g [61]
(Jones, 1983b).
The water-content adjusted bulk density used in Crop Yield
equation [51] is estimated with Grossman's equation Crop yield may be reduced through reductions in
(Grossman et al., 1985)
harvest index caused by water stress. Most grain crops
are particularly sensitive to water stress from shortly
BDg i = BD3 + (BDD - BD3)
before until shortly after anthesis, when major yield
components are determined (Doorenbos and Kassman,
FCg-SW,i 1979). Optimum conditions for growth may reduce
.[56] harvest index slightly if dry matter accumulation is large
FCg-WPg(4.083-3.33 BDD^/^ > and economic yield is limited by sink size. The harvest
index is affected by water stress using the equation
where BD is the water content adjusted bulk density on
day i, BD3 is the bulk density at 33 kPa water content,
HIA-HIAi_i-HIj(l---
BDD is the oven dry bulk density, FC is the field (WSYFj)(FHUi)(0.9-WSi)/
capacity, W P is the wilting point, and SW is the soil
water content for layer i on day i. [62]

Vol. 32(2):March-April, 1989 503

where HIj is the normal harvest index for crop j , HI A is short daylength. The daylength reduction factor is
the adjusted harvest index, WSYFj is a crop parameter estimated with the equation
expressing the sensitivity of harvest index to drought for
crop j , FHU is a function of crop growth stage, and WS HRLT,
is the water stress factor for day i. Notice that harvest FHR^ = 0.35 (1.0- -) [64]
index may increase slightly on days with WS values HRLT^, + 1
greater than 0.9. The crop stage factor, FHU, is
estimated with the equation where FHR is the daylength reduction factor, HRLTi is
the daylength on day i, and HRLTmn is the minimum
daylength for the location. The frost reduction factor is
FHU: = sin ' . 3 < H U I i < 0 . 9 . .[63] estimated with the equation
12 0.3 I
- T mn,i
whe FRSTi =
FHUi = 0., HUIi < 0.3 -Tmn,i-^^P(^^j,l+^^j,2-Tnin,i)

HUIi>0.9 T < 1 °C .[65]

Thus, water stress only affects harvest index between 0.3 where FRST^ is the frost damage factor, T^.^ is the
and 0.9 of maturity with the greatest effect occurring at minimum temperature on day i in °C, and afj j and afj 2
0.6. are parameters expressing the crop's frost sensitivity.
The reduction in standing live biomass is estimated with
Winter Dormancy the equation
The daylength growth constraint is used to simulate a
winter dormant period for fall planted crops. These ^ ^ A G , i = 0.5-BAG,i(1.0-HUIi).max(FHRi,FRSTi)
constraints are only imposed for areas that have less than
12 months growing season. A 12-month growing season [66]
for warm-season crops is defined in the model as one that
has no months with mean minimum temperature less
than 5°C. If there is a dormant winter period, it is where ABAC is the reduction in above ground in t-ha~i on
defined as the time when daylength is within 1 h of the day i, HUI is the heat unit index, and B^G is the above
location's minimum daylength. ground biomass in t-ha"^ on day i. Note that frost
If a crop experiences a winter dormant period, the damage is greater when plants are small (HUI«0)and
heat unit summation (equation [2]) is set to zero. This approaches zero near maturity.
provides for rapid new growth when temperatures
increase in the spring. During the dormant period, the MODEL TESTING
plants are not allowed to grow. The standing live biomass The model was tested with data from several locations
is actually decreased during this period due to frost and with considerable variation in soil and weather

TABLE 1. Data For Rice, Sunflower, Barley and Soybean Used for Yield Demonstration of the EPIC Model

Crop Location Years Researchers

Corn Yuma, AZ 1974 Stewart et al. (1977)

Davis, CA 1974-1975 Stewart et al. (1977)
Fort Collins, CO 1974-1975 Stewart et al. (1977)
Logan, UT 1974-1975 Stewart etal. (1977)
Mandan, ND 1968-1970 Alessi and Power (1977)
Florence, SC 1980-1982 Karlen and Camp (unpubl.)
Temple, TX 1979 Pietsch and Gerik (1980)
Bushland, TX 1975-1977 Musick and Dusek (1980)
Story, Ringgold, and Monona, lA 1960-1970 Laflen (unpubl.)

Wheat Phoenix, AZ 1977-1978 Dusek (unpubl.)

Garden City, KS 1980-1981 Wagger (1983)
Hutchinson, KS 1979-1980 Wagger (1983)
Manhattan, KS 1981 Wagger and Kissel (impubl.)
Pendleton, OR 1980-1981 Klepper et al. (1983)
Bushland, TX 1977 Musick and Dusek (unpubl.)
Temple, TX 1977 Monk, Arkin, Maas, and Ritchie (unpubl.)
Lind, WA 1976 Johnson (1978)
Pullman, WA 1972 Thill (1976)

Rice Far East, Southeast Asia, 1983-1984 Oldemanet al., (1986)

Southern Asia, South America

Toulouse, France 1983-1986 INRA, Toulouse, France (unpubl.)

Soybeans Ringgold, lA 1963-1969 Laflen (unpubl.)

504 TRANSACTIONS of the ASAE

characteristics (Table 1). The tests involved were: (1) the relationship between simulated and measured yields.
response of corn {Zea mays L.) grain yield to soil depth This consisted of testing whether the slope of the
for three locations; (2) measured and simulated grain regression line was significantly different from zero. If it
yields for six crop species in a wide range of was not, the model failed to show superiority over simply
environments; (3) the ability of the model to simulate using the mean measured yield for prediction.
com yield response to irrigation for four arid locations in If the regression line had a significant positive slope, it
the U.S.; and (4) the ability of the model to simulate was compared with the ideal line through the origin with
N-response for six corn planting dates in Hawaii. a slope of 1.0. The simulations were checked for bias by
constructing a confidence band for the regression line
Materials and Methods and checking to see if the ideal line was outside this
General band.
EPIC requires a number of crop-specific inputs. Once Finally, the r^ provided a final means of describing
these input paramaters were set for a crop species, they how well the simulated and measured yields agreed.
were not adjusted for individual data sets or locations,
however, potential heat units from planting to maturity Simulating Corn Grain Yield Response to Irrigation
may vary at different locations for the same crop. These simulations were designed to test the model's
In all but two cases, the water balance was simulated ability to simulate yield under various soil moisture
using the Penman method (Penman, 1956) of estimating conditions. The data came from a four-location, two-
potential evaporation. In testing the rice {Oryza sativa year study with a wide range of irrigation levels and
L.) yields and in testing the N-response of corn in measured yields (Stewart et al., 1977). The high
Hawaii, lack of wind data necessitated use of the evaporative demand at these sites was desirable for
Priestley-Taylor equation (Priestley and Taylor, 1972; testing the water balance. Simulated yields were
Ritchie, 1972). graphically compared to measured yields.

Simulating Corn Grain Yield Response N-Response of Corn in Hawaii

to Decreased Soil Depth The ability of EPIC to simulate N-response for corn
The purpose of this test was to demonstrate EPIC's grain yields was tested for four planting dates at one
ability to simulate the effect of soil depth on corn yield. location and one planting at each of two others, all in
Three locations with a range in annual rainfall were Hawaii. The soil at all locations was Hydric Dystrandept.
chosen for the test (Columbia, MO — 970 mm; Temple, For each planting, there were six nitrogen treatments
TX — 860 mm; and Bushland, TX — 460 mm). The ranging from zero to a maximum of between 185 and 225
same silt loam soil was used at all three locations. Soil kg/ha. Based on description of the data (Singh, 1985),
characteristics were assumed uniform throughout the the initial organic N in the soil was between 0 and 30
940 mm soil profile depth (bulk density =1.4 t-m~3, tons/ha throughout the profile. Within this range, initial
wilting point = 0.16 m-m-^, field capacity = 0.315 organic N values were adjusted to give reasonable
m-m-i, sand=20%, silt=70%, andpH=7.0). For each simulation of humus mineralization. Close agreement
location, grain yield was predicted for 13 consecutive between measured and simulated yields for the 0 applied
years of generated weather. The first three years were N treatment indicates proper humus mineralization
used to remove the effect of initial estimates of soil water simulation. The simulated yields for all the treatments
and NO3. Thus, only results from the last 10 years were were then graphically compared to the measured yields.
used in the analysis. Three simulations were performed
using the same weather and the same soil, except 7, 14, TABLE 2. Pertinent Crop Parameters Used in the EPIC Runs
or 22 cm of soil was removed from the bottom of the 94 Crop
cm soil profile. Removal was from the bottom of the Parameter Com Wheat Rice Sunflower Soybeans Barley
profile to standardize the top layers across depth BE 40.0 20.0 60.0 25.0 25.0
35.0
treatments. The surface layer was 15 cm in all cases. HI 0.50 0.42 0.50 0.25 0.31 0.42
Each soil profile was given a runoff curve number To 25.0 15.0 24.0 25.0 25.0 15.0
appropriate for its total depth. Tb 8.0 0.0 8.0 6.0 10.0 0.0
LAImx 5.0 8.0 6.5 5.0 9.0 7.0
HUIo 0.80 0.80 0.78 0.55 0.60 0.75
Simulating Grain Yield of Six Crop Species
ahj 15.05 15.01 20.01 15.01 15.01 15.01
Several statistics were used to evaluate the model's ah2 50.95 50.95 70.95 50.95 50.95 50.95
effectiveness in estimating yields for six crops at various afi 5.01 5.01 1.50 5.15 5.01 5.01
locations. First the means and standard deviations of the af2 15.05 15.10 2.95 15.95 15.05 15.10
measured and simulated yields were compared. Close adj 1.00 1.00 0.50 1.00 1.00 1.00
agreement in mean and standard deviation indicates ALT 3.0 2.0 3.0 3.0 3.0 1.0
CAF 0.85 0.85 1.0 0.85 0.85 0.85
similarity in yield probability distributions. Yield
HMX 2.5 1.2 0.8 2.5 1.5 1.2
probability distributions are useful in decision making.
RDMX 2.0 2.0 0.9 2.0 2.0 2.0
For each crop, a paired comparison t-test was used to WSYF 0.050 0.010 0.010 0.010 0.010 0.010
determine if the difference between measured and bni 0.044 0.060 0.05 0.05 0.0524 0.06
predicted values was significantly different from zero bn2 0.0164 0.0231 0.0200 0.0230 0.0320 0.0231
(SAS, 1982). bn3 0.0128 0.0134 0.0100 0.0146 0.0286 0.0130
Model performance was evaluated further by bpi 0.0062 0.0084 0.0060 0.0063 0.0074 0.0084
0.0023 0.0032 0.0030 0.0029 0.0037 0.0032
regressing simulated on measured yields. The first aspect bp2
bp3 0.0018 0.0019 0.0018 0.0023 0.0035 0.0019
of this approach was to check for a significant

Vol. 32(2):March-April, 1989 505

TABLE 3. Values for PHU for Various Locations and Years Unless point was the frost damage curve, af2; the critical
Otherwise Noted, the Values were Input aeration factor, CAP; and maximum rooting depth,
Crop Location PHU
RDMX (Appendix A).
Rice was also unusual in that all the values for PHU
Corn Florence, SC 2000 were input. A preliminary test of the data indicated that
Bushland, TX 2000 the values for PHU from planting to reported dates of
Columbia, MO 1820 maturity fell into two divergent groups. The means for
Man dan, ND 1020*
Bloomington, XL 1970*
the two groups were approximately 2100 C and 1600 C.
Logan, UT 1625 Thus, these values were used for the simulations. It was
Davis, CA 1760 interesting to note that with only two exceptions, all the
Fort Collins, CO 1205 locations in the 2100 C PHU group were at latitudes
Yuma, AZ 2835* greater than 20°. Those in the 1600 C PHU group, with
Manona, lA 1730-1830*
Story,lA 1670-1830* one exception, were at latitudes less than 20°. The
Ringgold, lA 1850-2000* variety of IR36 is sensitive to photoperiod, with a 7.5 d
delay in flower per hour increase in photoperiod above 14
Wheat Pullman, WA 1790-2690* hours (Vergara and Chang, 1985). Thus, photoperiod
Pendleton, OR 1710-1962* probably was at least partially responsible for the
Garden City, KS 1614-1835*
Hutchinson, KS 1463*
differences in development rate at the different latitudes.
Manhattan, KS 1942* Values for PHU of sunflower (Helianthus annus L.),
Bushland, TX 1650-1965* soybeans (Glycine max (L.) Merr.), and barley
Temple, TX 1502* (Hordeum vulgare L.) were calculated based on the
Phoenix, AZ 1485-2346*
reported harvest dates. These values can be used for
Rice Cuttack, India 2100 further applications of EPIC when the harvest date is not
Coimbatore, India 2100 reported.
Kapurthala, India 2100
Nanjing, China 2100 Simulating Corn Grain Yield Response
Muara, Indonesia 2100
Parwanipur, Nepal 2100 to Decreased Soil Depth
Sakha, Egypt 2100 The response of yield to reduced soil depth was
Suweon, South Korea 2100 dependent on amount of rainfall (Fig. 1). Mean
Milyang, South Korea 2100 simulated yields tended to decline as the soil profile was
Ahero, Kenya 1600 reduced at Columbia and Temple, but such a tendency
Palmira, Columbia 1600
Pintung, China 1600 was not evident at Bushland. Thus, the soil water
Sanpatong, Thailand 1600 holding capacity of the profile was a limitation only at
Los Banos, Philippines 1600 the first two locations. Even a profile of 72 cm was deep
Masapang, Philippines 1600 enough to accommodate the limited water at Bushland.
Sunflower France 1442-1849* Perhaps the most noteworthy aspect of this analysis
was the magnitude of the variability associated with each
Soybeans France 1123-1282* mean yield as evidenced by the large LSD values. At all
Ringgold, lA 1576* three locations, yields were not significantly different
among soil depths. Previously, the effect of weather on
Barley France 1524-2000*
grain yield reduced yield prediction accuracy using a
productivity index (Kiniry et al., 1983). Thus, a process-
oriented, weather-dependent model appears to be a more
Results feasible means for quantifying yield response to erosion.
General
As discussed above, a number of input values were 6000 r • COLUMBIA. MO 95 cm
required to describe each crop (Tables 2 and 3). o TEMPLE,TX 88cm
Definitions of parameters are given in Appendix A. Only A BUSHLAND.TX 49cm
the potential heat units (PHU) from planting to maturity
varied among cultivars within a species. Thus, while the
number of these inputs was great, only one was adjusted 4000
for the proper cultivar for a location.
Most PHU values for corn were input, but some were
calculated by the model using average temperatures
between planting and harvest (Table 3). In the latter
case, harvesting was assumed to occur at maturity. The 2000
PHU for corn ranged from 1020 C for Mandan, ND, to
2835 C for Yuma, AZ. The PHU for wheat {Triticum O
o
aestivum L.) was calculated in all cases and ranged from
1463 to 2690 C.
Wetland rice, partially because of its unusual soil "94 87 80 72
environment, was often the divergent species for a TOTAL SOIL DEPTH ( c m )
parameter. Noteworthy cases were for the two points on Fig. 1—Simulated response of com yield to reduced soil depth for three
the leaf area development curve, ahi and ah2; the second locations.

506 TRANSACTIONS of the ASAE

 TABLE 4. Mean and Standard Deviation of Measured and fell well within the 95% confidence band for the
Simulated Yield for the Six Crop Species regression line.
Yield
Results with wheat were similar to those with corn.
t-values There was a wide range of measured yields, with some
Crop N Measured Simulated for less than 1 t-ha~^ and one of 9 t-ha~^ Again, there was a
differences
X SD X SD in meanst
significant relationship between measured and predicted
yield. The value of r^ was 0.80 and the fitted line fell
s/ha) -- within the 95% confidence band for the regression line.
Corn 118 6.5 3.1 6.3 3.1 -1.0 Measured wheat yields in both the high and low yielding
Wheat 20 3.8 2.1 3.7 1.8 -0.5 environments were simulated reasonably well. However,
Rice 33 5.6 1.7 5.3 1.3 -1.1 similar to corn the slope of the regression line was less
Sunflower 27 2.9 0.9 2.7 0.9 -1.4 than 1.0. This indicated that the model was not as
Barley 19 3.5 1.2 3.3 0.7 -0.8
Soybeans 10 2.1 0.4 2.2 0.3 0.7
responsive to the environment as it should have been.
Data for rice {Oryza sativa (L.)) was a subset from a
tNone were significant at the 95% confidence level. multilocation wetland study done by IRRI (Oldeman et
al., 1986). These data were collected at locations in the
Simulating Grain Yield of Six Crop Species Philippines, Asia, Africa, and South America. Testing
E P I C ' S mean simulated yields were always within 7% was done using 33 of the original 63 data sets. Those not
of the mean measured yields (Table 4). With the included had unusual problems with pests or disease,
exception of the soybean simulations, mean simulated unusual management practices, or some other anomaly.
yield of each crop was greater than the mean measured Only data for the standard variety IR36 were included.
value. The wheat crop had the smallest percent However, two values for PHU were used as discussed
difference between mean simulated and measured yields, above. As in the previous two comparisons, there was a
while sunflower had the greatest. None of the crops had a significant relationship between measured and simulated
significant difference between measured and simulated yields. The value of r^ was not as high, 0.41, but the 1:1
mean yields. line was within the 95% confidence band except in the
The SDs of the simulated yields were similar to the extreme upper range of the data. The simulated yields
SDs of the measured, but the percent differences tended to be less than the measured yields in the higher
between simulated and measured SDs were greater (up yielding environments. However, EPIC simulated the
to 34% for barley) in general than for the means. Two second highest measured yield nearly exactly. Again, the
exceptions to this were corn and sunflower, with 0 and low value for the slope of the regression line, 0.48,
4% differences in SD, respectively. implied that the model was not as responsive to the
EPIC simulated corn yields reasonably in the high, environment as it should have been.
intermediate, and low yielding conditions (Fig. 2). The Simulated yields of sunflower were also significantly
slope of the regression line was less than 1.0. Measured related to the measured yields. The r^, 0.42, was similar
yields ranged from zero in some dryland studies in the to the value for rice. The regression line was close to the
southwestern U.S. to almost 14 t-ha~^ at two irrigated 1:1 line and within the 95% confidence band throughout
sites. The slope was significantly different from zero at the range of the data. Simulated yields were similar to
the 95% confidence level. The value of r^ was high (0.65) measured yields at both the lowest and the highest
and the fitted line was close to the 1:1 line. The 1:1 line values.

CORN
n °II8
rZ = 0.65
Y = 0 . 8 I X + I.|
/
/ •

r2«0.20 rZ « 0 . 2 0
Y «0.48Xf2,47

o MONOCULTURE

MEASURED YIELD (Mg/ha)

Fig. 2—Simulated and measured yield of six crop species.

Vol. 32(2):March-April, 1989 507

 to account for the variability in yield of these two crops
MEASURED
SIMULATED
MEASURED
SIMULATED
but gave a reasonable average yield.

Simulating Corn Grain Yield Response to Irrigation

Simulated corn yield response to irrigation was similar
to measured response from Davis, CA (Fig. 3). At the
lowest values, the model tended to slightly underpredict
grain yield. In contrast the high yielding treatments were
slightly overpredicted. However, in general, EPIC
simulated the effects of water stress on grain yield at this
\' location reasonably well.
Simulations of corn yield at Logan, UT were close to
FORT C O L L I N S , CO. YUMA.AZ.
measured yields in the lowest irrigation treatments.
MEASURED
1974
•
1975
o MEASURED
1974
•
Similar to the results at Davis, the model showed a yield
SIMULATED A A SIMULATED A response throughout the range of irrigation levels.
However, the simulated yields were less than the
measured values in the high yielding treatments.
Simulations of corn yield at Fort Collins, CO, were
closer to measured values in 1975 than in 1974. In all but
the lowest yielding treatment in 1974, measured yield
was considerably greater than simulated. It appeared
that the simulated evaporative demand was too great or
the input soil water holding capacity was too low in 1974.
Finally, EPIC overestimated yield response to
AMOUNT OF IRRIGATION APPLIED (cm) irrigation in Yuma, AZ, in 1974. Measured yields were
Fig. 3—Simulated response of corn yield to irrigation for four increased only 1.6 t-ha~^ while the model simulated a 5.6
locations. t-ha~^ increase. However both simulated and measured
yields showed some yield response throughout the range
of irrigation treatments.
The small range of measured yields of soybeans and
barley restricted the amount of variability the model N-Response of Corn in Hawaii
could account for and thus restricted the r^. Soybean Generally, simulated yields were similar to measured
yields ranged from 1.5 to 3.0 t-hd~S and barley yields throughout the range of N application (Fig. 4). In
ranged from 2.4 to 4.2 t-ha~S excluding the two highest addition, for up to about 100 kg/ha applied-N, both
yielding data sets. Slopes for the simulated: measured simulated and measured yields increased with increased
regression line were not significantly different from zero applied N. However, the model usually failed to
for either crop. The values for r^ were low 0.20 — for adequately simulate the plateau in response at the upper
both cases. However, there was no obvious bias in range of applied N values.
simulated yields and the 1:1 line fell within the The measured yield showed a plateau in the response
confidence band for both crops. EPIC, therefore, failed above approximately 110 to 130 kg-ha"^ applied-N in

LOCATION I PLANTING I 12 " LOCATION 1 PLANTING 2 LOCATION I PLANTING 3

• MEASURED
A SIMULATED
10
A
•
8 • 1
• • A
6 A

1 1 1 1

40 80 120 160 200 240 40 80 120 160 200 240 40 80 120 160

LOCATION I PLANTING 4 LOCATION 3

40 80 120 160 200 240 0 40 80 120 160 200 240 40 80 120 160 200 240

NITROGEN APPLIED (kgho"')

Fig. 4—Simulated response of corn yield to applied-nitrogen for six plantings in Hawaii.

508 TRANSACTIONS of the ASAE

planting 1, 2, and 4 of location 1 and at location 2. Relationship of nitrogenase activity to plant water stress in field-grown
However, in all these cases, the simulated yield soybeans. Field Crops Res. 8:61-71.
4. Alessi, J. and J.F. Power. 1975. Effects of plant population, row
continued to increase up to the maximum amount of spacing and relative maturity on dryland corn in the northern plains. I.
applied-N. Thus, the model overestimated the impact of Corn forage and grain yield. Agron. J. 66:316-319.
applied-N on yield above about 120 kgN-ha"^ 5. Andrews, R.E. and E.I. Newman. 1970. Root density and
application. competition for nutrients. Oecol. Plant. 5:319-334.
6. Angus, J.F. 1981. Phasic development in field crops. I. Thermal
Two cases where the measured and simulated yield response in the seedling phase. Field Crops Res. 3:365-378.
responded similarly at the high N application rates were 7. Baker, C.K., J.N. Gallagher and J.L. Monteith. 1980.
location 1, planting 3 and location 3. In the first Daylength change and leaf appearance in winter wheat. Plant, Cell and
situation, measured and simulated yields increased with Environ. 3:285-287.
an increase in applied N up to 135 kg-ha~^ N, with a 8. Balsko, J. A. and D. Smith. 1971. Influence of temperature and
nitrogen fertilization on the growth and composition of switchgrass
slight decrease at the greatest amount of applied-N. (Panicum virgatum L.) and timothy {Phleum pratense L.) at anthesis.
Likewise for location 3, both measured and simulated Agron. J. 63:853-857.
showed very similar increases in yield for each additional 9. Barber, S.A. 1984. Soil nutrient bioavailability. New York:
increment of applied-N, throughout the range of John Wiley & Sons, Inc.
10. Begg, J.E. 1964. Diurnal energy and water exchanges in bulrush
treatments. millet in an area of high solar radiation. Agric. Meteorol. 1:294-312.
11. Bonhomme, R., F. Ruget, M. Derieux and P. Vincourt. 1982.
SUMMARY AND CONCLUSIONS Relations entre production de matiere seche aerienne et energie
interceptee chez differents genotypes de mais. C.R. Acad. Sci. Paris
The EPIC plant growth model was developed to 294:393-398.
estimate soil productivity as affected by erosion 12. Borg, H. and D.W. Grimes. 1986. Depth development of roots
throughout the U.S. Since soil productivity is expressed with time: An empirical description. Transactions of the ASAE
29(1):194-197.
in terms of crop yield, the model must be capable of 13. Bouniois, A. et al., 1985. Influence des conditions
simulating crop yields realistically for soils with a wide d'alimentation hydrique ou (et) azotee a differents stades du
range of erosion damage. Also, simulations of many developpement sur la production de grains et la nutrition azotee du
crops are required because of the wide variety grown in soja. Eurosoya 3:55-61.
the U.S. 14. Brenes, E. and R.W. Pearson. 1973. Root responses of three
Gramineae species to soil acidity in an Oxisol and an Ultisol. Soil Sci.
To meet the requirements, a general crop model that is 116:295-302.
sensitive to the plant environment (climate, nutrient 15. Coelho, D.T. and R.F. Dale. 1980. An energy-crop growth
supply, and soil characteristics) was developed. The variable and temperature function for predicting corn growth and
EPIC crop parameter table presently contains development: Planting to silking. Agron. J. 72:503-510.
16. DeJager, A. 1979. Localized stimulation of root growth and
parameters for about 25 crops. Parameters for other phosphate uptake in Zea mays L. resulting from restricted phosphate
crops may be obtained from experts or from the supply, pp. 391-403. In J.L. Harley and R. Scott Russell, eds. The Soil-
literature. Root Interface. New York: Academic Press.
The plant growth model has been tested throughout 17. Doorenbos, J. and A.H. Kassam. 1979. Yield response to water.
Irrigation and Drainage Paper 33. Food and Agric. Org. of the United
the U.S. and in several foreign countries. The simulation Nations, Rome.
of crop yield by EPIC predicted the response of long- 18. Eavis, B.W. 1972. Soil physical conditions affecting seedling
term yield to various management strategies. Mean root growth. I. Mechanical impedance, aeration, and moisture
simulated yields for crops and the associated standard availability as influenced by bulk density and moisture levels in a sandy
deviations were remarkably close to values for measured loam soil. Plant Soil 36:613-622.
19. Eik, K. and J.J. Hanway. 1965. Some factors affecting
yields. None of the simulated means were significantly development and longevity of leaves of corn. Agron. J. 57:7-12.
different from the measured means at the 95% 20. Foy, C D . , A.L. Fleming and G.C. Gerloff. 1972. Differential
confidence level. Likewise, there was a significant aluminum, tolerance in two snapbean varieties. Agron. J. 64:815-818.
(a=0.05) relationship between measured and simulated 21. Foy, C D . et al., 1974. Aluminum tolerance of wheat cultivars
related to region of origin. Agron. J. 66:751-758.
yields for all crops except soybeans and barley. 22. Gilmore, E . C and J.S. Rogers. 1958. Heat units as a method of
EPIC also accurately simulated corn yield response to measuring maturity in corn. Agron. J. 50:611-615.
irrigation; thus, it could be used for irrigation 23. Grable, A.R. and E.G. Seimer. 1968. Effects of bulk density,
scheduling. However, the regression lines of predicted aggregate size and soil water suction on oxygen diffusion, redox
yield as a function of measured yield for the various crops potential and elongation of corn roots. Soil Sci. Soc. Amer. Proc.
32:180-186.
consistently had slopes less than 1.0. This indicated that 24. Grossman, R.B., W.D. Nettleton and B.R. Brasher. 1985.
the model was not as responsive to the environment as it Application of pedology to plant response prediction for tropical
should have been. These errors could be especially vertisols. In Proc. Fifth Int. Soil Classification Workshop, Sudan,
important when making real-time irrigation or fertilizer November 2-11, 1982.
25. Hanks, R.J. 1983. Yield and water-use relationships: An
application decisions. Similar results at Bushland, TX overview, pp. 393-411. In H.M. Taylor, W.R. Jordan and T.R.
(Steiner et al., 1987) indicated that EPIC is best suited Sinclair, eds. Limitations to Efficient Water use in Crop Production.
for making long-term management decisions. Amer. Soc. Agron., Crop Sci. Soc. Amer., Soil Sci. Soc. Amer.,
Madison, WI.
26. Harper, J. E. 1976. Contribution of dinitrogen and soil or
fertilizer nitrogen to soybean {Glycine max L. Merr.) production, p.
References 101-107. In L. D. Hill, ed. World soybean research conference.
1. Abruna, F., J. Rodriquez and S. Silva. 1982. Crop response to Danville, ILL: The Interstate Printers and Publishers, Inc.
soil acidity factors in Ultisols and Oxisols in Puerto Rico. VI. Grain 27. Johnson, R.C. 1978. Apparent photosynthesis and
sorghum. /. Agric. Univ. of Puerto Rico 61:28-38. evapotranspiration of winter wheat in a field environment. M.S. thesis,
2. Acevedo, E., T.C. Hsiao and D.W. Henderson. 1971. Washington State Univ., Pullman.
Immediate and subsequent growth responses of maize leaves to 28. Jones, C A . 1983a. A survey of the variability in tissue nitrogen
changes in water status. Plant Physiol. 48:631-636. and phosphorus concentrations in maize and grain sorghum. Field
3. Albrecht, S.L., J.M. Bennett and K.J. Boote. 1984. Crops Res. 6:133-147.

Vol. 32(2):March-April, 1989 509

 29. Jones, C.A. 1983b. Effect of soil texture on critical bulk control of water and salinity levels in the soil. U t a h W a t e r Res. L a b .
densities for root growth. Soil Sci. Soc. Amer. J. 47:1208-1211. P R W G 151-1. 191 p .
30. Jones, C.A. 1985. C-4 grasses and cereals. New York: John Wiley 59. Taylor, H . M . 1983. A p r o g r a m to increase plant available water
& Sons, I n c . through rooting modification, p p . 463-472. In Root Ecology and Its
3 1 . Jones, C.A. a n d J . R . Kiniry, e d s . 1986. CERES-Maize. College Practical Application. I n t . Symp., G u m p e n s t e i n . September 27-29,
Station: Texas A & M Univ. Press. 1982. Budndesanstalt fur alpenlandische Landwirtschalf, A-8952
32. Jones, C.A., A. N . Sharpley a n d J. R. Williams 1984a. A Irding.
simplified soil a n d plant phosphorus model: I. Documentation. Soil 60. Taylor, H . M . , G . M . Robertson a n d J.J. Parker, Jr. 1966. Soil
Sci. Soc. Amer. J. 48:800-805. strength — Root penetration relations for medium to coarse-textured
33. Jones, C.A., A . N . Sharpley a n d J.R. Williams. 1984b. A soil materials. Soil Sci. 102:18-22.
simplified soil a n d plant phosphorus model: I I I . Testing. Soil Sci. Soc. 6 1 . Thill, D . C . 1976. Physiological a n d environmental effects on
Amer. J. 48:800-805. yield potential in early- a n d late-planted winter wheat. M . S . thesis,
34. Jungk. A. a n d S.A. Barber. 1974. Phosphate u p t a k e of corn Washington State Univ., P u l l m a n .
roots as related to the proportion of the roots exposed to phosphate. 62. Tollenaar, M . , T . B . D a y n a r d a n d R . B . H u n t e r . 1979. Effect of
Agron. J. 66:554-557. t e m p e r a t u r e on rate of leaf a p p e a r a n c e a n d flowering date of maize.
35. Kiniry, L.N., C.L. Scrivner a n d M . E . Keener. 1983. A soil Crop Sci. 19:363-366.
productivity index based upon predicted water depletion and root 63. Thornley, J. H . M . 1976. Mathematical models in plant
growth. Univ. Missouri Agric. E x p . Sta. Res. Bull. 1051, 2 6 p . physiology. London: Academic Press.
36. Klepper, B . , T . W . T u c k e r a n d B . D . D u n b a r . 1983. A 64. Tompsett, P . B . 1976. Factors affecting the flowering of
numerical index to assess early inflorescence development in wheat. Andropogon gayanus K u n t h . Responses t o photoperiod, t e m p e r a t u r e ,
Crop Sci. 23:206-208. a n d growth regulators. Ann. Bot. 40:695-705.
37. Linn, D . M . a n d J . W . D o r a n . 1984. Effect of water-filled pore 6 5 . Trouse, A . C . , Jr. 1964. Effects of compression of some
space on carbon dioxide a n d nitrous oxide production in tilled a n d subtropical soils on t h e soil properties a n d upon root development.
nontilled soils. Soil Sci. Soc. Amer. J. 48:1267-1272. P h . D . dis., Univ. Hawaii, Honolulu.
38. Monsi, M . a n d T . Saeki. 1953. U b e r den Lichfaktor in den 66. Vergara, B . S . a n d T . T . C h a n g . 1985. The flowering response of
Pflanzengesellschaften u n d sein Bedeutung fur die Stoffproduktion. the rice plant to photoperiod. Intl. Rice Res. Inst., Manila. 61 p .
Japan J. Bot. 14:22-52. 67. Wagger, M . G . 1983. Nitrogen cycling in t h e plant-soil system.
39. Monteith, J.L. 1973. Principles of Environmental Physics. P h . D . thesis, K a n s a s State Univ., M a n h a t t a n .
London: Edward Arnold. 68. W a t t s , W . R . 1972. Leaf extension in Zea mays. II. Leaf
40. Monteith, J.L. 1977. Climate a n d the efficiency of crop extension in response to independent variation of the t e m p e r a t u r e of
production in Britain. Phil. Trans. Res. Soc. London B. 281:277-329. the apical meristem, of the air a r o u n d the leaves, a n d of the rootzone.
4 1 . Monteith, N . H . a n d C.L. B a n a t h . 1965. T h e effect of soil / . Exp. Bot. 23:713-721.
strength on sugarcane growth. Trop. Agric. 42:293-296. 69. Williams, J . R . , C.A. Jones a n d P . T . Dyke. 1984. A modeling
42. Muchow, R . C . et a l . , 1982. Growth a n d productivity of approach to determining t h e relationship between erosion a n d soil
irrigated Sorghum bicolor (L. Moench) in Northern Australia. I. Plant productivity. Transactions of the ASAE 27(1)1:129-144.
density a n d a r r a n g e m e n t effects of light interception a n d distribution,
and grain yield, in the hybrid Texas 610SR in low a n d m e d i u m APPENDIX A
latitudes. Aust. J. Agric. Res. 33:773-784. Notation
43. Mugwira, L . M . , S.J. Patel a n d A.L. Fleming. 1980. A l u m i n u m A plant water use rate-soil depth parameter
effects on growth a n d Al, Ca, Mg, K, a n d P levels in triticale, wheat, afi,af2 crop parameters for frost sensitivity
and rye. Plant Soil 57:467-470. ahi,ah2 crop parameters that determine the shape of the
44. Musick, J . T . a n d D . A . Dusek. 1980. Irrigated corn yield leaf-area-index development curve
response to water. Transactions of the ASAE 23(l):92-98, 103. ad crop parameter that governs leaf area index decline
45. O l d e m a n , L . R . , D . V . Seshu a n d F . B . Cady. 1986. Response of rate
rice to weather variables. Intl. Rice Res. Inst. Report. Manila, ALO maximum ALS value a crop can tolerate without
Philippines. stress (%)
46. Patterson, T . G . a n d T . A . L a R u e . 1983. Nitrogen fixation by ALS aluminum saturation of the soil(%)
soybeans: Cultivar a n d seasonal effects of comparison of estimates. ALT aluminum tolerance index number
Crop Sci. 23:488-492. AS aeration root growth stress factor (0-1)
47. Pavan, M . A . , F . T . Bingham a n d P . F . Pratt. 1982. Toxicity of ATS aluminum toxicity root growth stress factor (0-1)
a l u m i n u m to coffee in Ultisols a n d Oxisols a m e n d e d with C a C 0 3 , bnj,bn2,bn3 crop parameters for plant N concentration equation
M g C 0 3 , a n d C a S 0 4 - 2 H 2 0 . Soil Sci. Soc. Amer. J. 46:1201-1207. bpi,bp2,bp3 crop parameters for plant P concentration equation
48. P e n m a n , H . L . 1956. Evaporation: An introductory survey. bti,bt2 soil strength parameters related to soil texture
Neth. J. Agric. Sci. 4:9-29. B accumulated plant biomass (above ground and
49. Pietsch, D . a n d T.J. Gerik. 1980. Corn hybrid performance. roots) (t-ha~0
Texas Agric. E x p . Sta. Bull. PR3724, 13p. above ground biomass of growth crop (t-ha"^)
50. Priestley, C . H . B . a n d R.J. Taylor. 1972. O n t h e assessment of Bp potential crop biomass (kg-ha~^)
surface heat flux a n d evaporation using large-scale p a r a m e t e r s . Mon. BD soil bulk density (t-m"-^)
Weather Rev. 100:81-92. BDD oven dry soil bulk density (t-m~^)
5 1 . Ritchie, J . T . 1972. A model for predicting evaporation from a BD3 33 kPa water content bulk density (t-m"-^)
row crop with incomplete cover. Water Resources Res. 8:1205-1213. BDL bulk density near lower boundary of stress (causes
52. SAS Institute. 1982. SAS User's Guide: Statistics. SAS Inst., no root growth stress) (t-m~-^)
Cary, NC, 584. BDU bulk density near upper stress boundary (root
5 3 . Sharpley, A . N . et al., 1984. A simplified soil a n d plant growth stress ~0.2 (t-m"-^)
phosphorus model: I I . Prediction of labile, organic, a n d sorbed BE crop parameter — converts energy to biomass
phosphorus. Soil Sci. Soc. Amer. J. 48:800-805. (kg.ha-MJ-i-m-2)
54. Sharpley, A . N . et al., 1985. A detailed phosphorus CjLP labile P concentration in the soil (g-t~^)
characterization of seventy-eight soils. ARS-31. 32 p . optimal N concentration for a plant (g'g~0
55. Sharpley, A . N . , C.A. Jones a n d J . R . Williams. 1989. T h e CPB optimal P concentration for a plant (g-g~0
nutrient submodel of E P I C . In A.N. Sharpley a n d J . R . Williams, eds. critical aeration factor for a crop
CAF
The Erosion-Productivity Impact Calculator (EPIC), (in press). Model crop height (m)
CHT
Documentation. clay content of the soil (%)
CLA
56. Singh, U . 1985. A crop growth model for predicting corn Eo potential evaporation ( m m ' d ~ 0
performance in the tropics. P h . D . diss., Univ. Hawaii. Ep potential plant evaporation rate (mm-d~0
57. Steiner, J.L., J . R . Williams a n d O . R . Jones. 1987. Evaluation of FCi field capacity (33 kPa for many soils) water content
the E P I C simulation model using a dryland wheat-sorghum-fallow crop of soil layer 1 (mm)
rotation. Agron. J. 79:732-738. FHU crop stage (heat unit) factor governing water stress-
58. Stewart, J.L et al., 1977. Optimizing crop production through harvest index

510 TRANSACTIONS of the ASAE

 winter dormancy day length factor for reducing SAN sand content of a soil layer (%)
FHR
standing live biomass (0-1) SAT saturation index for top 1 m of soil
FRST frost damage factor (0-1) SD sun's declination angle (radians)
HI potential harvest index •— ratio of crop yield to SN N stress factor for crop growth (0-1)
above ground biomass SNs N stress scaling factor
HIA actual harvest index for a crop ss soil strength root growth stress factor (0-1)
HMX maximum crop height (m) SWi soil water content in layer 1 (mm)
HRLT daylength (h) t time (h)
HU daily heat units — average daily temperature minus T temperature (°C)
base temperature of crop (°C) Tb base temperature for a crop (plants start growing)
HUF heat unit factor for driving leaf-area-index (°C)
development (0-1) T soil surface temperature (°C)
HUFH heat unit factor for estimating harvest index T daily minimum air temperature (°C)
^ mn
development (0-1) daily maximum air temperature (°C)
HUI heat unit index — ratio of accumulated to potential T optimal temperature for a crop (°C)
*• mx
heat units (0-1) T temperature stress factor affecting harvest index
HUI, heat unit index value when leaf area index starts TH (0-1)
declining TS temperature stress factor for crop growth (0-1)
LAI leaf area index — area of plant leaves relative to plant water use rate in soil layer 1 (mm-d"^)
the soil surface Si potential plant water use rate in soil layer
LAI^, maximum LAI potential for a crop (mm-d"^)
LAI, leaf area index value at start of decline UN, N supply rate in layer 1 (kg-ha~*-d~0
LAT latitude of watershed (degrees) Up daily potential plant water use from the root zone
LFu labile P factor for crop uptake (0-1) (mm)
PAR photosynthetically active radiation (MJ-m~^) uc soil water deficit compensation factor (0-1)
PHU potential heat units for a crop (°C) UN N uptake rate of plant (kg-ha~^-d~^)
RA solar radiation (ly) UND plant N demand rate (kg-ha~'-d~^)
RAF solar radiation factor affecting harvest index UNS N supply rate of soil (kg-ha~^-d~')
RD root depth (m) UP P uptake rate of plant (kg-ha~^-d~^)
RDMX maximum root depth for a crop (m) UPD plant demand rate (kg'ha~^'d~^)
REG crop growth constraint (minimum stress factor — UPS P supply rate of soil (kg-ha"~*-d~')
water, nutrients, aeration, temperature) WNO3 weight of NO3 in a soil layer (kg-ha~^)
RGF root growth constraint (minimum stress factor — WPj soil water content of layer 1 at wilting point (mm)
soil strength, temperature, aluminum toxicity) WS water stress factor for crop growth (0-1)
RWj root weight in soil layer 1 (t-ha~^) WSYF water stress factor for adjusting harvest index
RWT total root weight (t-ha~^) YLD crop yield (t-ha~0
RZ root zone depth (m) Z soil depth from the surface (m)

Vol. 32(2):March-April, 1989

511