Agronomy Journal Grow Your Career With ASA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (17)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Cavero, J.
Right arrow Articles by Faci, J. M.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Cavero, J.
Right arrow Articles by Faci, J. M.
Agricola
Right arrow Articles by Cavero, J.
Right arrow Articles by Faci, J. M.
Related Collections
Right arrow Crop Growth and Development
Right arrow Water Stress
Right arrow Maize
Right arrow Crop Models
Right arrow Irrigation
Agronomy Journal 92:679-690 (2000)
© 2000 American Society of Agronomy

MODELING

Simulation of Maize Yield under Water Stress with the EPICphase and CROPWAT Models

Jose Caveroa, Inma Farreb, Philippe Debaekec and Jose M. Facib

a Dep. Genetica y Produccion Vegetal, Estacion Experimental de Aula Dei (CSIC), Apdo. 202, 50080 Zaragoza, Spain
b Unidad de Suelos y Riegos, Servicio de Investigacion Agroalimentaria (DGA), Apdo. 727, 50080 Zaragoza, Spain
c INRA, Station d'Agronomie, BP 27, 31326 Castanet-Tolosan cedex, France

jcavero{at}eead.csic.es


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Materials and methods
 NOTES
 Results and discussion
 Conclusions
 REFERENCES
 
Models that simulate the effects of water stress on crop yield can be valuable tools in irrigation. We evaluated the crop growth simulation model EPICphase and the model CROPWAT on their ability to simulate maize (Zea mays L.) grain yield reduction caused by water stress under semiarid conditions. The simulation of evapotranspiration (ET), harvest index (HI), leaf area index (LAI), and final biomass was also evaluated. Data from three field experiments were used to test the models. In one sprinkler-irrigated experiment, different water amounts (0–592 mm) were applied, producing a continuous water deficit. The other two experiments were flood-irrigated and water stress was imposed at given development stages of maize. EPICphase simulated the ET with a root mean square error (RMSE) of 40 mm. The regression of the EPICphase simulated vs. measured values of HI and yield had intercepts that were not significantly different from 0 and slopes not different from 1. EPICphase overestimated the biomass in the more water-stressed treatments (intercept of simulated vs. measured values = 5.25 t ha-1) due to overestimation of LAI. Modifications of EPICphase relative to the effect of water stress on LAI growth and on the light extinction coefficient improved the simulations of LAI, biomass, HI, and yield. CROPWAT calculated maize grain yield with a RMSE of 14% but overestimated ET in the flood-irrigated treatments . Better simulation of ET by EPICphase makes this model more consistent for calculating yield reduction due to water stress.

Abbreviations: BD, bulk density • ET, evapotranspiration • HI, harvest index • k, light extinction coefficient • LAI, leaf area index • LER, leaf elongation rate • PAR, photosynthetically active radiation • PET, potential evapotranspiration • RMSE, root mean square error


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Materials and methods
 NOTES
 Results and discussion
 Conclusions
 REFERENCES
 
IN THE SEMIARID AREAS of the world, high yields of field crops can be attained if irrigation water is applied properly. However, because of the high demand for irrigation water by crops in these areas, yields can be very low if water is not supplied adequately both in quantity and in time (Musick and Dusek, 1980; Singh and Singh, 1995), and water can be misused. This misuse could lead to excessive percolation, which has environmental consequences and diminishes water reserves. In addition, water scarcity in these areas is increasing. Thus, optimized use of the available water for irrigation is very important.

Models that adequately simulate the effects of water stress on yield can be valuable tools in irrigation management. These models can be used to optimize the allocation of irrigation water between different crops and/or the distribution of water during the crop season (Bryant et al., 1992; Cabelguenne et al., 1995; Cabelguenne et al., 1997; Howell et al., 1989; Stewart et al., 1975; Wenda and Hanks, 1981). Complete testing of a model is needed before it can be used for irrigation planning in a particular area. This will ensure that the model correctly simulates the main physiological processes that affect crop yield under water stress.

Among the models that can be used for this task, a distinction can be made between crop growth simulation models, which simulate main processes of crop growth (leaf area growth, biomass production and partition), such as CERES-maize (Jones and Kiniry, 1986), CropSyst (Stockle et al., 1994), EPIC (Williams et al., 1984), GOSSYM (Reddy et al., 1997), and SUCROS (Penning de Vries and Van Laar, 1982), and those models that do not explicitly simulate crop growth but that have been developed for irrigation planning. CROPWAT (Smith, 1992) is the best known among the latter. In this model, crop potential evapotranspiration (PET) is calculated using the crop coefficient concept (Doorenbos and Pruitt, 1977). The effect of water stress on crop yield is considered by using the crop response factors (Stewart et al., 1975) that have been derived from different experiments (Doorenbos and Kassam, 1979). However, the crop response factor can be different between locations because of different evaporative demand (Howell, 1990).

Crop growth simulation models can be divided between those more mechanistic (such as SUCROS) and those more empirical (such as CERES-maize, CropSyst, and EPIC). However, in contrast to CROPWAT, these crop growth simulation models usually differentiate between the effects of water stress on photosynthesis (or biomass production, in the more empirical models), leaf area growth, and harvest index (Cabelguenne et al., 1999; Reddy et al., 1997; Villalobos et al., 1996). Those are the main processes affected by water stress (Hsiao, 1990), so if they are adequately simulated, these crop growth simulation models could be more universally applicable than models such as CROPWAT. In addition, crop growth simulation models usually provide simulated data from other processes (such as nitrate leaching) that can be important in the efficient management of water resources.

Maize is one of the most important crops in irrigated semiarid areas of the world. It has high irrigation requirements and is very sensitive to water stress (Rhoads and Bennett, 1990). Recently, decreasing prices of maize grain in Europe have decreased the net return from this crop. Thus, adequate irrigation management of maize is important not only for saving water, but also for improving crop profitability.

Our aim was to evaluate and compare a crop growth simulation model and the model CROPWAT in their ability to simulate maize yield reduction caused by water stress under different types of water stress (continuous or at given development stages). The simulation of evapotranspiration, harvest index, leaf area index, and final biomass was also evaluated. Among the existing different crop growth simulation models, we chose EPICphase (Cabelguenne et al., 1999). This model is mainly empirical and is generic, allowing simulation of different crops. EPICphase is a modification of the extensively tested EPIC model (Beckie et al., 1995; Bryant et al., 1992; Cabelguenne et al., 1990; Cavero et al., 1999a; Kiniry et al., 1992a), especially improved for water and N stress modeling. EPICphase does not use crop coefficients to calculate crop PET. The ET component of this model is Ritchie's model (Ritchie, 1972).


   Materials and methods
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
NOTES
 Results and discussion
 Conclusions
 REFERENCES
 
Field Experiments
Three experiments were conducted at Zaragoza, Spain (41°43' N, 0°48' W, 225 m altitude) on the experimental farm of the Agronomic Research Service (SIA). The soil is loamy and is classified as Typic Xerofluvent. The climate is Mediterranean semiarid with mean annual maximum and minimum daily air temperatures of 20.9 and 8.5°C, respectively, precipitation of 322 mm, and reference ET around 1200 mm (Faci and Martínez-Cob, 1991).

Continuous Water Deficit Experiment
This experiment was conducted in 1986 in a field with soil depth >1.5 m. Other soil characteristics are given in Table 1 . Barley (Hordeum vulgare L.) cv. Georgia was planted on 28 February to dry the soil profile and was incorporated in the soil with the preplant tillage operations. Maize cv. Adour 640 (FAO 700) was planted on 30 April in rows 0.75 m apart at a planting density of 60000 plants ha-1. Fertilization consisted of 80 kg ha-1 P2O5 and 32 kg ha-1 K2O applied preplanting and 200 kg ha-1 N applied sidedress. Weeds were controlled with atrazine [6-chloro-N-ethyl-N'-(1-methylethyl)-1,3,5-triazine-2,4-diamine] applied preemergence at 2 kg a.i. ha-1. Maize was irrigated with a sprinkler line source 50 m long located in the middle of the field and parallel to the maize rows. Different irrigation treatments were established as a consequence of different distances of maize rows from the sprinkler line source (Hanks et al., 1976) (Fig. 1) . Perpendicularly to the sprinkler line source, six plots at 3-m intervals were defined on both sides of the sprinkler line source. This was done at two different places along the sprinkler irrigation line, so there were four replications for each of the six water amount treatments (T1 to T6 treatments) (Fig. 1). Cylindrical metal cans (0.155 m in diameter and 0.165 m high) painted orange were located at soil level in each plot to measure the amount of irrigation water applied (Fig. 1). Collected water was measured immediately after each irrigation event. The cans were raised as the crop grew, so they were always above the maize canopy. Thirty-nine irrigation applications were made between 9 June and 30 September, each 20 to 30 min in duration. The most irrigated treatment (closest to the sprinkler source line, T6) received a water amount (irrigation plus rain) corresponding to the maize ET calculated with a Class A evaporation pan using a pan coefficient of 0.8 (Faci, 1986) and the crop coefficients for maize (Doorenbos and Pruitt, 1977). Volume per application was 8 to 36 mm in this treatment. Seasonal irrigation water amounts applied in the different plots were 0 (T1), 88 (T2), 289 (T3), 437 (T4), 567 (T5), and 592 mm (T6). Soil water measurements were made with a neutron probe (Model 3320, Troxler Electronic Laboratories, Research Triangle Park, North Carolina1) in 16 of the 24 plots (Fig. 1) during the crop season to a depth of 1.80 m at 0.3-m intervals beginning 0.15 m from the soil surface. The neutron probe was calibrated by taking 12 undisturbed soil samples at different depths where the volumetric water content was determined. The regression of the neutron probe measurements against the measured moisture content of soil samples had an r2 of 0.96 and an RMSE of 0.016 m3 m-3.


View this table:
[in this window]
[in a new window]
  		Table 1 Soil characteristics in the different experiments

 


View larger version (67K):
[in this window]
[in a new window]
  		Fig. 1 Experimental layout of the continuous water deficit experiment (1986). Numbers 1 to 6 indicate the position of the different irrigation treatments

 
Evapotranspiration (mm) in each plot was determined from the soil water balance equation , where P is precipitation (mm), I is irrigation (mm), R is runoff/run-on (mm), SD is soil water depletion (mm), and D is drainage (mm) below the rooting depth considered (1.5 m). Runoff/run-on was assumed to be insignificant because the field was level-smoothed to zero slope and bordered with earth berms and irrigation volume per application was low. Soil water depletion was calculated as the difference between the beginning and ending total soil water contents for the season. Drainage below the rooting depth was not determined, so ET could not be calculated directly from the measurements. However, ET was estimated by using the drainage simulated from the EPICphase and CROPWAT models. Saad (1999) found good agreement between measured and EPIC-simulated drainage in a lysimeter study with maize in our experimental site. Nguyen et al. (1996) also reported accurate simulation of drainage in lysimeters with EPIC. This estimated ET will be referred to here as adjusted ET. Precipitation totaled 204 mm during the crop season.

Crop phenology was monitored during the growing season. Length and maximum width of leaves from three plants per plot were measured at different times during the season and leaf area was calculated by multiplying the leaf length by the leaf width and by 0.75, according to Norman and Campbell (1989). LAI was calculated from leaf area and plant density. The aboveground biomass of plants from 5-m transects of the two rows on either side of the catch cans (7.5-m2 area) were harvested (Fig. 1). Plants were divided into leaves, stems, cobs, and grain and dried at 70°C. Grain yield (dry matter), dry biomass and HI were determined.

Experiments of Water Stress at Different Maize Development Phases
Experiments were conducted in 1995 and 1996 to study the effect of water stress at different stages of maize development using flood irrigation. Soil depth was variable within the field due to a gravel layer between 0.8 and 1.7 m in depth. Mean soil characteristics are provided in Table 1. Maize cv. Prisma (FAO 700) was planted on 17 May 1995 and 16 May 1996 in rows 0.75 m apart. The planting density was 80000 plants ha-1. Fertilization consisted of 150 kg ha-1 P2O5 and 150 kg ha-1 K2O applied preplant and 300 kg ha-1 N, one-third applied preplant and the rest as sidedress. Weeds and pests were adequately controlled. Crop phenology was monitored during the growing season. The maize growing season was divided into three phases: (i) from emergence to tassel emergence, (ii) from tassel emergence to milk stage of grain, and (iii) from milk stage to physiological maturity. In each of the phases, irrigation was supplied either in the amount necessary to meet the PET of the crop (I treatment), or at about one-third of this amount (0 treatment) by skipping some of the irrigation events or applying a lower depth (Table 2) . Under water scarcity in areas that are surface-irrigated, water volume per application cannot be easily modified and water stress usually develops because the number of irrigation events is reduced. An additional treatment (iii) consisting of half of the water used in the fully irrigated treatment (III), was included. Irrigation management in the fully irrigated treatment was based on the common practice in the area, which consists of flood irrigation at 10- to 14-d intervals, with application depths ranging from 55 to 79 mm. Irrigation amounts and timing were adjusted according to the average reference ET measured in a weighing lysimeter (Faci et al., 1994) and to the crop coefficients for maize (Doorenbos and Pruitt, 1977).


View this table:
[in this window]
[in a new window]
  		Table 2 Irrigation water amounts applied (mm) in the experiments of water stress at different developmental phases of maize (1995 and 1996)

 
The experimental design was a randomized block with three replicates. The experimental plot unit was 50 m2 and was delimited by ridges to allow the establishment of different irrigation treatments. Irrigation was applied from gated pipes with a total discharge into each plot of 3.4 L s-1. Water applied at each irrigation event was measured with a volumetric flow meter (MC0100, McCrometer, Hemet, CA). Soil water content was determined gravimetrically in each 0.2- (1995) or 0.3-m (1996) layer down to a 1.2-m depth at planting and at harvest. The same neutron probe used in the 1986 experiment was used in 1995 to measure the soil water content, approximately every week, during the crop season to a depth of 1.20 m at 0.2-m intervals. The neutron probe was calibrated by taking 24 undisturbed soil samples at different depths where the volumetric water content was determined . ET in each plot was determined with the soil water balance equation as in the 1986 experiment. The gravimetric measurements of soil water content were used to determine the soil water depletion. Precipitation during the crop season was 38 mm in 1995 and 103 mm in 1996. In 1995, the aboveground biomass of maize plants was sampled from 0.5 m2 of each plot at different times during the crop season. Biomass partitioning was determined. Both years, leaf area at flowering was determined as in the continuous water deficit experiment. Leaf area index was determined for other stages of growth in 1995 by multiplying the dry weight of leaves by the specific leaf area using a value for maize of 15 m2 kg-1 (Cavero et al., 1999b). The percentage of photosynthetically active radiation (PAR) intercepted by the crop was measured every 1 or 2 wk with a light probe (Sunfleck ceptometer, Decagon Devices, Pullman, WA). Fraction of PAR intercepted was calculated from measurements taken above and below the maize canopy at noon. The light extinction coefficient (k) was calculated from the intercepted PAR and the LAI data (Flenet et al., 1996). At harvest, the aboveground biomass from an area of 9 m2 (1995) or 15 m2 (1996) was collected, divided into different parts as in the 1986 experiment and dried at 70°C. Grain yield (dry matter), dry biomass, and HI were determined.

Models
EPICphase
EPICphase was developed from EPIC to improve the simulation of the effects of water and N stresses in crops. Complete details can be found in Cabelguenne et al. (1999). The main differences between EPIC and EPICphase are that EPICphase incorporates division of the developmental period of the crops into physiological phases, crop specific-water extraction capacities, accelerated decline of leaf area under water stress, drought adaptation of sunflower and soybean, inclusion of the effects of water and N stress on the harvest index, and the possibility for a crop to have an ET higher than the reference ET.

In EPICphase, daily actual ET is equal to crop PET if the available water for the crop is higher than crop PET, but it is equal to the available water for the crop if the available water is lower than crop PET. The available water for the crop depends on soil water content, rooting depth, rooting shape, and soil properties. EPICphase considers that water stress has daily effects on the LAI growth, the biomass production, and the HI. The model also considers that the effect of water stress on HI differs with the physiological phases of the crop. In the case of maize, it considers that HI is affected by water stress during Phase 2 (maximum LAI to end of anthesis) and Phase 3 (end of anthesis to pasty grain) with reductions being double for the same water stress intensity in Phase 2 as compared with Phase 3. A conical soil water extraction pattern with depth is used for maize according to its rooting distribution and density with depth (Cabelguenne and Debaeke, 1998).

Crop parameter values for maize used in this study were basically the same as those that Cabelguenne et al. (1999) proposed for southwestern France. However, for some parameters that are cultivar-dependent, such as maximum leaf area index (which also depends on planting density [Kiniry et al., 1992b]), development of leaf area index with time, fraction of season when LAI declines, duration of different developmental phases, and harvest index, values were obtained from the experimental data (Table 3) . These values were derived from the fully irrigated treatment in the 1986 (cv. Adour) and 1995 (cv. Prisma) experiments, because they could be calculated from the measurements made (LAI, phenology, and HI). Data from the other five (1986 experiment), eight (1995 experiment), and nine (1996 experiment) treatments were not used to derive those values.


View this table:
[in this window]
[in a new window]
  		Table 3 EPICphase crop parameter values used for maize in this study

 
Different values from those used by Cabelguenne et al. (1999) were used for some crop parameters for both cultivars. Base temperature was set to 8°C and maximum temperatures to 25°C as in EPIC (Williams et al., 1984). The light extinction coefficient was set to 0.50 according to our data and Cavero et al. (1999b). The water extraction beyond the wilting point was set to 0.8 of the wilting point according to the soil water measurements. The maximum rooting depth considered was 1.50 m according to the soil water measurements from this study. Similar maximum depths are reported in other studies (Cabelguenne and Debaeke, 1998; Dardanelli et al., 1997). The biomass energy conversion was considered to be constant for the entire period before the decline of leaf area, but a lower value (30 instead of the default value of 39 kg MJ-1) was used for Adour due to the lower biomass production. This round value, which was derived to fit the biomass production in the fully irrigated treatment of the 1986 experiment, is within the range of values found for maize (Birch et al., 1999; Kiniry et al., 1989) and close to those found for other cultivars at our site (Cavero et al., 1999b).

The original Penman equation was used to calculate the PET because of its accuracy in our conditions (Faci et al., 1994). Dugas and Ainsworth (1985) have pointed out the consequences of using different methods to calculate the PET in crop model simulations. Daily weather data from a nearby climatic station was used as model input. The maximum ET of maize was allowed to be 20% higher than the PET according to our climatic conditions (ASCE, 1996).

Cropwat
CROPWAT 5.7 was used (Smith, 1992). Default values suggested for maize were used, except values for phase duration (derived from the experiments' phenological data), maximum crop coefficient (which was set to 1.20 as in EPICphase), and maximum rooting depth (set to 1.5 m) (Table 4) . The crop coefficient for the initial phase was calculated for each experiment, considering the time interval between wetting events, the evaporative power of the atmosphere and the magnitude of the wetting events (Allen et al., 1998) (Table 4). Average monthly data of rain and reference ET, calculated with the original Penman equation, were used as inputs.


View this table:
[in this window]
[in a new window]
  		Table 4 CROPWAT crop parameter values used for maize in this study

 
CROPWAT calculates crop PET using crop coefficient values. The actual ET is equal to the crop PET if soil water depletion is below the allowable depletion limit. Above that limit, ET is reduced linearly as soil water depletion increases. Yield is reduced in each crop growth stage according to the crop response factor (Doorenbos and Kassam, 1979), which indicates the percent of yield reduction per percent of ET reduction compared to PET. Main differences between EPICphase and CROPWAT are summarized in Table 5 .


View this table:
[in this window]
[in a new window]
  		Table 5 Main differences between EPICphase and CROPWAT

 
Model Runs
Because of the nonuniform soil depth in the field used in the experiments of 1995 and 1996, the models were run for each individual plot, considering soil depth as the only soil characteristic that was different between plots. Soil depth in every plot was determined during the installation of access tubes for neutron probe soil water measurements and interpolation of those data with SURFER (1995). This was not done for the 1986 experiment, where soil depth was uniform.

Data Analysis
The mean measured values were compared with the mean simulated values of ET, yield, final biomass, LAI, and HI for each irrigation treatment. Bias and RMSE (calculated as described by Retta et al. [1996]) and linear regression of simulated vs. measured values were also used to evaluate model performance:

where S and M are the simulated and measured values for the ith observation and N is the number of observations.


   Results and discussion
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 NOTES
 Results and discussion
Conclusions
 REFERENCES
 
EPICphase Simulations
Model simulations indicated that there was drainage below the rooting depth in some treatments. Thus, the adjusted ET, which was calculated using the EPICphase simulated drainage values, was compared with model simulations of ET (Fig. 2) . Low values of bias (-1.51 mm) and RMSE (39.8 mm) indicated a good agreement between adjusted and simulated values of ET. Saad (1999) found good agreement between measured and EPIC-simulated ET in maize. In our study, there was a tendency to overestimate the ET when the adjusted ET was low and to underestimate the ET when it was high (intercept > 0, slope < 1) (Table 6) . Some disagreement in the 1995 and 1996 experiments could be due to the fact that soil water depletion was calculated to a depth of 1.20 m, because in these experiments soil water was measured only until that depth, while the model considered a rooting depth of 1.5 m. However, this difference was not important, because in our conditions soil water depletion below 1.20 m in maize fields is low (Cosculluela and Faci, 1992) and because EPICphase simulates low water uptake below this depth due to the rooting characteristics of maize (Cabelguenne et al., 1999). Different studies have found that root activity of maize under irrigation is usually concentrated in the top soil (Dardanelli et al., 1997; Otegui et al., 1995) with little water depletion below 1.0 m in depth (Gordon et al., 1995).



View larger version (29K):
[in this window]
[in a new window]
  		Fig. 2 Comparison of seasonal simulated evapotranspiration (ET) from EPICphase and CROPWAT with adjusted ET (calculated from the soil water balance and the simulated deep percolation by each of the two models). Each point represents the mean of each treatment of the different experiments. The diagonal line represents the 1:1 relationship

 

View this table:
[in this window]
[in a new window]
  		Table 6 Simulated values of evapotranspiration (ET) as a function of adjusted ET values for all the experiments with EPICphase, EPICphase modified, and CROPWAT.{dagger}

 
The model adequately simulated the and yield of maize (Fig. 3 , Table 7) . For these two variables, the linear regression of simulated vs. measured values had intercepts not significantly different from zero and slopes not significantly different from 1 (P > 0.05) (Table 7). However, there was a tendency to overestimate the aboveground biomass in the more water-stressed treatments (Fig. 3 and 4 , Table 7). This was possibly due to the fact that the model overestimated the LAI under water stress during the crop season (Fig. 4). The maximum LAI was clearly overestimated in most of the treatments (Fig. 3, Table 7). Steduto et al. (1995) reported similar problems with simulations of wheat LAI under water stress with EPIC.



View larger version (31K):
[in this window]
[in a new window]
  		Fig. 3 Comparison of maize grain yield, aboveground biomass, harvest index, and maximum leaf area index (LAI) measured and simulated with EPICphase. Each point represents the mean of each treatment of the different experiments. The diagonal line represents the 1:1 relationship

 

View this table:
[in this window]
[in a new window]
  		Table 7 Simulated values of maize yield, biomass, harvest index, and maximum leaf area index (LAI) as a function of measured values for all the experiments with EPICphase and EPICphase modified.{dagger}

 


View larger version (29K):
[in this window]
[in a new window]
  		Fig. 4 Comparison of measured and simulated leaf area index (LAI) and aboveground biomass in selected treatments of the 1986 and 1995 experiments. Symbols are the mean values; error bars are the standard deviation. Lines represent the mean simulated values with EPICphase (solid lines) or EPICphase modified (dotted lines)

 
Hsiao (1990) has indicated that although attention in the past has been directed to the effects of water stress on source intensity (rate of photosynthesis per unit of source area), source size (leaf area) is considerably more sensitive to water stress, and therefore would be more critical during the period of canopy development. Boyer (1970) pointed out that in maize and other crops leaf expansion is more sensitive than photosynthesis to water stress. Several other studies have shown the greater sensitivity of maize leaf expansion to water stress (Acevedo et al., 1971; Beadle et al., 1973; Tanguilic et al., 1987). In these studies, transpiration rate was also measured. Although previous studies were conducted under different conditions, a relationship between leaf photosynthesis or leaf elongation rate (LER), expressed as a fraction of the potential leaf photosynthesis or LER (without water stress), and the transpiration rate, expressed as a fraction of the potential transpiration rate, could be derived (Fig. 5) . In EPICphase, water stress affects LAI growth and biomass production (photosynthesis) according to the ratio of actual to potential ET (water stress reduction factor). The relationship used in EPICphase between the water stress reduction factor and the reduction of biomass production is similar to the results of the cited studies for the reduction of photosynthesis (Fig. 5). However, the relationship used in EPICphase between the water stress reduction factor and the reduction of LAI growth is very different from the results of the cited studies for the reduction of LER (Fig. 5). This relationship makes model simulations of LAI growth less sensitive to water stress than it was found to be in the mentioned studies (Fig. 5).



View larger version (22K):
[in this window]
[in a new window]
  		Fig. 5 Relationships of the leaf photosynthesis or leaf elongation against the relative transpiration rate obtained from different studies with maize and the relationships used in EPICphase and EPICphase modified to reduce biomass production or LAI growth as function of the water stress reduction factor (actual ET/potential ET)

 
Reduction of leaf area growth decreases the solar irradiance intercepted by the crop. The consequences of this reduction in terms of biomass production and yield will depend on the time period of crop growth when it happens and the final LAI attained (Bradford and Hsiao, 1982; Singh and Singh, 1995). In determinate species such as maize, the reduction of LAI can be irreversible (Bennett et al., 1989).

When plants are under water stress, leaf rolling is a mechanism to reduce transpiration of the crop (Bradford and Hsiao, 1982; Turner, 1986). Leaf rolling reduces the solar irradiance intercepted by the crop because it reduces the light extinction coefficient (k), as was found in our experiments (Fig. 6) . However, EPICphase does not take into account this reduction of k as leaf rolling develops.



View larger version (20K):
[in this window]
[in a new window]
  		Fig. 6 (A) Mean values of the light extinction coefficient (k) calculated for the different treatments in the flood-irrigated experiments at maize flowering. (B) Relationship between the water stress reduction factor (actual evapotranspiration [ET]/potential ET) and the correction factor for k used in the modified EPICphase model

 
The rate of water uptake required to sustain normal plant growth at any given time depends not only upon soil water status but also upon the atmospheric conditions and properties of the plants (Ahuja and Nielsen, 1990). Under semiarid conditions, water stress during the vegetative growth phase is not uncommon. The reduction in leaf area growth leads to reduced PAR interception, and consequently to reduced biomass production and yield (Muchow, 1989). In other studies, EPICphase has been tested for maize in a milder climate, where water stress usually develops later (Cabelguenne et al., 1999).

Testing of crop models often does not include LAI measurements, but in the case of water stress this seems to be important. We made an additional run of EPICphase considering a modified effect of water stress on LAI growth that was intermediate between the measured results in the different experiments mentioned (Fig. 5). In this run, we also considered a reduction of the light extinction coefficient as water stress increases with a function derived from our experimental data (Fig. 6).

Simulation of the effect of water stress on LAI growth during the crop season was improved with the modified EPICphase model (Fig. 4). Simulation of the maximum LAI was especially improved as shown by the lower bias and RMSE values and the higher r2 and lower intercept values of the regression of simulated vs. measured values (Fig. 7 , Table 7). Simulation of ET was slightly improved (Table 6). Modifications of EPICphase improved the simulation of biomass (lower bias and RMSE, intercepts closer to 0) (Table 7) because both modifications altered the solar irradiance intercepted by the crop. The small improvement in HI simulation (lower bias and RMSE, slope closer to 1) may have been due to the slightly better simulation of ET (Table 6). Yield in EPICphase is calculated as the product of biomass and HI. Thus, simulation of yield was improved (lower bias and RMSE, slopes closer to 1) (Table 7), mainly because of the improved simulation of biomass. In any case, the improvement in biomass and yield simulation was not very significant (Fig. 3 and 7). This was possibly because reductions of leaf area when LAI is >4 usually have small consequences for biomass production (El-Sharkawy and Cock, 1987). Considering only the treatments where the measured biomass reduction was higher than 25% with respect to the fully irrigated treatment, the RMSE for biomass was reduced by 24% from the original EPICphase model, compared with 11% reduction when all the data were included. Thus, the modified model improved simulation of biomass production, especially when water stress was greater.



View larger version (32K):
[in this window]
[in a new window]
  		Fig. 7 Comparison of maize grain yield, aboveground biomass, harvest index, and maximum leaf area index (LAI) measured and simulated with EPICphase modified. Each point represents the mean of each treatment of the different experiments. The line represents the 1:1 relationship

 
To derive the relationship between water stress and reduction of leaf elongation, we used long-term (at least 24 h) responses of maize because the time step of EPICphase is 1 d. However, Acevedo et al. (1971) found very short-term responses of maize leaf elongation to water stress and reduction of leaf elongation rate when there was no apparent reduction of transpiration. They also found that after severe water stress, leaf elongation rate did not recover to a similar level as nonstressed plants. NeSmith and Ritchie (1992) have indicated that water stress in maize causes not only short-term effects, such as reduced daily LAI increase, but long-term effects, such as reduced leaf number and leaf area. The reduction in expansion rate of leaves while the blade is growing inside the whorl of the plant is ignored by the model, but it affects final leaf size and therefore LAI. Those effects could explain why the simulations of LAI by the modified EPICphase model seem to be delayed as compared with the measured values in the more water-stressed treatments (Fig. 4). Such effects constitute a limit to LAI modeling as it is done in EPICphase, even modified. Ben-Haj-Salah and Tardieu (1996) have recently proposed to model maize LER taking into account the reduction in LER due to evaporative demand.

Some disagreement between measured and simulated values is expected because EPICphase does not consider some processes, such as the direct effect of water stress on root growth and the changes in biomass partitioning between shoots and roots as a consequence of water stress (Debaeke et al., 1996). Besides, if water stress is severe enough, the photosynthetic capacity of leaves can be affected irreversibly (Boyer and McPherson, 1975), leading to lower biomass production than expected.

CROPWAT Model
Model simulations indicated that there was drainage below the rooting depth, as in the case of EPICphase runs. As with EPICphase, the adjusted ET, calculated using the CROPWAT simulated drainage values, was compared with the model-simulated ET (Fig. 2). The agreement between adjusted and simulated values of ET was less than with EPICphase, as shown by the higher bias and RMSE values and the lower values for the slope and higher values for the intercept of the regression of simulated against adjusted ET (Table 6). CROPWAT overestimated ET mainly in the flood-irrigated experiments . In general, reduction of maize yield in the different irrigation treatments was adequately addressed by CROPWAT, with mean simulated values mostly within one standard deviation of the measured values (Fig. 8) .



View larger version (48K):
[in this window]
[in a new window]
  		Fig. 8 Comparison of measured and simulated yield reduction with EPICphase modified and CROPWAT in the different experiments. Values are means of each treatment; error bars represent standard deviations

 
Comparison of the Two Models
Comparison of the models' performance was only possible in terms of drainage, ET, and grain yield reduction. In the case of EPICphase, simulated yield reduction was calculated as a fraction of the mean yield in the fully irrigated treatment.

In general, simulated drainage was slightly higher with the EPICphase model (Fig. 9) . However, EPICphase simulated drainage losses at around 70 mm for the two most irrigated treatments in the 1986 experiment, while the CROPWAT model simulated insignificant losses for these treatments. As indicated, Saad (1999) found good agreement between measured and EPIC-simulated drainage in a lysimeter study with maize at our site, which suggests that drainage losses were underestimated by CROPWAT in these treatments.



View larger version (21K):
[in this window]
[in a new window]
  		Fig. 9 Comparison of the drainage below the rooting depth simulated with EPICphase and CROPWAT. Each point represents the mean of each treatment of the different experiments. The diagonal line represents the 1:1 relationship. Numbers inside the symbols indicate the number of points that are coincident

 
The poorer simulation of ET with CROPWAT was possibly related to the fact that EPICphase considers water extraction from deep soil layers to be limited, the limitation being dependent on crop species due to differences in root distribution and density (Turner, 1986). However, CROPWAT does not consider this limitation, so it simulated higher soil water depletion, and consequently higher ET. This higher soil water depletion simulated by CROPWAT could also be responsible for the lower simulated drainage. Another reason for the poorer simulation of ET by CROPWAT could be that EPICphase calculates ET using the Ritchie model. When flood irrigation is used and a small number of irrigations are applied until the crop completely covers the soil, the Ritchie model could better calculate the ET, especially during the initial period (10% cover). The use of a mean value of the crop coefficient for the initial period by CROPWAT, even if different values were used for the experiments, led to simulated ET values of 51 and 92 mm for 1995 and 1996, respectively, while EPICphase simulated an ET of 44 and 66 mm. The interpolation of climatic data made by CROPWAT to calculate the daily water balance could also be responsible for the discrepancies.

Both models followed a similar trend in yield reduction between the different treatments in all the experiments, which was similar to the observed trend (Fig. 8). No difference between model performances was found (RMSE was 16.3 % for EPICphase modified and 14.0 % for CROPWAT). Both models underestimated maize yield when water stress was imposed during the vegetative phase but irrigation was applied in any of the subsequent phases. This could be related to the conditioning effect indicated by some authors (Stewart et al., 1975; Jama and Ottman, 1993), which reduces the consequences of late water stress if the crop has suffered water stress during the vegetative period. However, Garrity et al. (1983) indicated that the magnitude of a drought stress conditioning response will depend on the genotype used, the phenological timing of the treatment and the irrigation scheduling, which makes modeling it difficult.

The poorer simulation of ET by CROPWAT in the flood-irrigated experiments indicates that results from this model should be considered with some caution. In addition, CROPWAT uses developmental time in days, whereas EPICphase uses degree-days, which makes calculations of development stages more accurate and has positive consequences for yield calculations.


   Conclusions
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 NOTES
 Results and discussion
 Conclusions
REFERENCES
 
Our results indicated that EPICphase can be used to calculate maize yield reduction under water stress in semiarid conditions. The model was tested under two different water stress experiments, where stress occurred either continuously or at given growth phases, which better illustrated its performance. However, the model must be modified to adequately address the effect of water stress on leaf area growth. The modified model that takes into account this question, as well as the reduction of the light extinction coefficient as a consequence of leaf rolling, improved the simulations of maize LAI, biomass, HI, and yield under water stress. Improvement was greatest for those treatments where LAI and biomass reduction were highest.

The model CROPWAT adequately calculated yield reduction caused by water stress, which makes this model a valuable tool for irrigation planning in maize. However, the model overestimated ET in two of the three years, so maize yield reductions calculated by CROPWAT should be considered with caution. Comparison of the two models indicated that there is no advantage in using one or the other in terms of yield reduction calculation due to water stress. However, better simulation of ET by EPICphase makes this model more consistent, even if more input data are needed.SURFER 1995


   ACKNOWLEDGMENTS
 
This work was supported by the CICYT (HID96-1380-C02-02). The useful comments received from the associate editor (Dr. S.R. Evett) and the reviewers are acknowledged.


   NOTES
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 NOTES
Results and discussion
 Conclusions
 REFERENCES
 
1 Mention of a product does not imply approval of this product to the exclusion of other products. Back

Received for publication August 15, 1999.
   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 NOTES
 Results and discussion
 Conclusions
 REFERENCES
 




This article has been cited by other articles:


Home page
 Agron. J.Home page
Y. Zhang, Q. Yu, C. Liu, J. Jiang, and X. Zhang
Estimation of Winter Wheat Evapotranspiration under Water Stress with Two Semiempirical Approaches
Agron. J., January 1, 2004; 96(1): 159 - 168.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
A. A. Andales, L. R. Ahuja, and G. A. Peterson
Evaluation of GPFARM for Dryland Cropping Systems in Eastern Colorado
Agron. J., November 1, 2003; 95(6): 1510 - 1524.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
J. K. F. Roygard, M. M. Alley, and R. Khosla
No-Till Corn Yields and Water Balance in the Mid-Atlantic Coastal Plain
Agron. J., May 1, 2002; 94(3): 612 - 623.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
J. Cavero, E. Playan, N. Zapata, and J. M. Faci
Simulation of Maize Grain Yield Variability within a Surface-Irrigated Field
Agron. J., July 1, 2001; 93(4): 773 - 782.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Ci