Agronomy Journal Journal of Natural Resources and Life Sciences Education
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 Google Scholar
Google Scholar
Right arrow Articles by Staggenborg, S. A.
Right arrow Articles by Vanderlip, R. L.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Staggenborg, S. A.
Right arrow Articles by Vanderlip, R. L.
Agricola
Right arrow Articles by Staggenborg, S. A.
Right arrow Articles by Vanderlip, R. L.
Related Collections
Right arrow Soybean
Right arrow Dryland Cropping Systems
Right arrow Sorghum
Right arrow Wheat
Right arrow Agricultural Systems
Right arrow Crop Models
Right arrow Crop Ecology
Published in Agron. J. 97:378-384 (2005).
© American Society of Agronomy
677 S. Segoe Rd., Madison, WI 53711 USA

Symposium Papers

Crop Simulation Models Can be Used as Dryland Cropping Systems Research Tools

S. A. Staggenborg* and R. L. Vanderlip

Dep. of Agron., Kansas State Univ., Manhattan, KS 66506

* Corresponding author (sstaggen{at}ksu.edu)

Received for publication February 5, 2004.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Dryland cropping systems research in the semiarid Great Plains region requires a substantial investment in land, labor, and other resources. The objective of this analysis was to illustrate that crop simulation models can assist scientists in making more efficient use of these resources by providing insight on potential plant responses to alterations in cropping systems before conducting field research. Models included in DSSAT 3.5 were used to simulate two cropping systems studies that evaluated the inclusion of grain sorghum [Sorghum bicolor (L.) Moench] into a traditional wheat (Triticum aestivum L.)–fallow system in western Kansas and soybean [Glycine max (L.) Merr.] into continuous grain sorghum in north-central Kansas. CERES-Wheat overestimated wheat yields by 16% although no consistent reason was identified for these errors. The model also simulated complete plant stand losses from winter injury in 5 yr when no stand losses were observed. CERES-Sorghum underestimated grain sorghum yields by approximately 27% across both studies. Overestimating the impact of water stress on plant growth appeared to be common at the western site, and a lack of response to N when grown in rotation with soybean appeared to be the primary sources of error at the northern site. Using uniform genetic coefficients to span a 19-yr study also contributed to errors in simulating sorghum yields. CROPGRO simulated soybean within 20% and closely mimicked annual responses of soybean yields to weather patterns. If researchers used these results to evaluate the objectives of both studies before conducting fieldwork, despite the errors, the overall trends would have been similar to those measured in the field. These results would have also enabled researchers to focus their research efforts, thus more efficiently using their resources.

Abbreviations: LAI, leaf area index • RMSE, root mean square error • WF, wheat–fallow • WSF, wheat–sorghum–fallow


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
CROPPING SYSTEMS RESEARCH generally requires significant commitments in large field areas to accommodate crop rotation treatments, field equipment and instrumentation, and diverse technical expertise to manage multiple crops and collect relevant soil and crops data necessary to provide insight into extremely dynamic systems. One of the greatest challenges in dryland cropping systems research in the Great Plains is the necessity of conducting trials over a long time period. Because of extreme variations in growing season temperatures and precipitation, short-term (1 or 2 yr) cropping system evaluations may not accurately represent long-term production potential of a given treatment. Crop simulation models can facilitate examining long-term treatment responses where the only limitation to the number of years examined in a modeling framework is availability of weather data.

Cropping systems research often entails investigating the adoption of new crops and other practices that might be viewed as risky by producers. For example, Norwood et al. (1990) conducted trials for 14 yr to compare a wheat–sorghum–fallow rotation to the traditional wheat–fallow (WF) system. Research in Colorado that compared WF to wheat–corn (Zea mays L.)–fallow or wheat–sorghum–fallow rotations was conducted from 1987 through 1992 (Kolberg et al., 1996) and continued in some fashion through 2000 (Shaver et al., 2002). If long-term simulations are coupled with field research data, the simulated data may provide insight on the production potential of the new crops after only two or three seasons of field data.

As resources to conduct long-term experiments become more limited, researchers increasingly need tools to evaluate cropping systems and their potential economic risk. Crop simulation models have existed in some simple forms since the conclusion of World War II (Sinclair and Seligman, 1996). Since then, they have become more complex and potentially more useful (Boote et al., 1996). In semiarid environments, simple empirical models that estimate grain yields as a function of water use, as described by Krieg (1988) and Stewart et al. (1983), may be useful in characterizing year-to-year yield variability since growing season water supply is often limiting. Sinclair and Seligman (1996) suggested that simple models have been shown to outperform more complex models when the research objective was very narrow or related to a single component such as water use (Asare et al., 1992). However, the weakness to this approach is that these types of models are incapable of accounting for the direct impact of temperature and water stresses on plant growth and are often limited by the constraints of the data from which they were developed.

Crop simulation models such as SORKAM (Rosenthal et al., 1989), GOSSYM (Baker et al., 1983), and CERES (Ritchie et al., 1985) estimate crop growth and development on a daily basis throughout the crop's life cycle. These models attempt to integrate numerous factors that affect crop growth and development such as plant available soil water, temperature, wind, genetics, management choices, and pest infestations. The strength of these models as research tools is their ability to capture the soil–environment–plant interaction across a wide range of environmental conditions.

Over the past two decades, crop models have been used to assess risk in marginal environments (Muchow et al., 1994) and to predict potential corn yields (Hodges et al., 1987), cotton water use (Staggenborg et al., 1996; Asare et al., 1992), soybean (Egli and Bruening, 1992) and wheat (Aggarwal and Kalra, 1994) yield responses to planting date, soybean (Boote et al., 1996) and peanut (Singh et al., 1994) response to plant population and row spacing, and replant decisions for grain sorghum (Heiniger et al., 1997). More recently, crop simulation models have been used in site-specific crop management (O'Neal et al., 2002). However, the limitation of most of these examples from a cropping systems approach is that they only encompass a single crop. Crop rotation is more common than continuous cropping throughout the USA because of the benefits that multicrop systems provide. Integrated systems such as the Decision Support System for Agrotechnology Transfer (DSSAT) (Jones, 1998) and Agricultural Production System Simulator (APSIM) (APSRU, 2004) make modeling cropping systems easier as inputs can be efficiently utilized by several models and outputs from a simulation can be seamlessly passed to the next model in the cropping sequence.

Several issues that limit the widespread adoption of crop models as cropping systems tools include model sensitivity to variables of interest such as N, atmospheric C, or pests; level of calibration or adaptability to environmental conditions within a given region; and availability and quality of input data. The objectives of this exercise were to illustrate that crop simulation models can be used to evaluate dryland cropping decisions. More specifically, these simulations were conducted in the context that if the researchers had run these simulations before initiating the field research, would they have made the same decisions or recommendations to producers regarding the viability of the systems being tested?


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
The objectives were to illustrate the utility of crop simulation models as research tools before conducting field research. In that context, the researcher would be faced with using basic knowledge about a crop rotation, planting dates, soil types, and hybrids (genetic coefficients) and would likely not alter them drastically during simulations so as to fit the results to some previously measured result. Therefore, many input decisions were made based on what was considered average or commonly accepted production practices. For instance, soil physical properties were not adjusted if it was believed that simulated water stress was too severe; hybrids and varieties were selected based on availability at the time each field research study was started; planting dates and plant densities were selected based on published recommendations, not actual planting dates used in the study; and no attempts were made to adjust genetic coefficients to address potential shortcomings that arose during the simulation of these three crops.

Wheat–Fallow vs. Wheat–Sorghum–Fallow
The first system simulated included rotation treatments of WF and wheat–sorghum–fallow (WSF). Research comparing these two systems under both conventional and reduced tillage was conducted at Tribune, KS, from 1974 through 1997 (Norwood et al., 1990). Yields were averaged each year across tillage treatments for use in the model evaluation as it was expected that simulated soil evaporation and runoff would be similar to the conventional treatments and water losses and movement with the soil would be similar to the reduced-tillage treatments since the simulations would not have soil moisture losses associated with the tillage treatments.

Wheat (Godwin et al., 1989) and sorghum (Alargarswamy and Ritchie, 1991) models contained with the generic CERES framework of DSSAT 3.5 (International Consortium for Agricultural Systems Application, Honolulu, HI) were used to simulate these treatments. Soil type at this location is a Ulysses silt loam (fine-silty, mixed, superactive, mesic Aridic Haplustolls). Soil physical properties were characterized at this site by Stone et al. (1987) and served as the source for the simulation input data (Table 1). Weather data used for these simulations were measured at Tribune from 1961 to 2000 using automated weather stations to record maximum and minimum temperatures and incident solar radiation while rainfall was measured manually. This span of years was chosen largely as a result of data availability at this location.


View this table:
[in this window]
[in a new window]
  		Table 1. Soil physical properties of a Ulysses silt loam near Tribune, KS (Stone et al., 1987), and Crete silt loam near Scandia, KS (Soil Survey Staff, 2004; Stone et al., 1978). Both were used as input for CERES-Wheat, CERES-Sorghum, and CROPGRO to simulate a cropping systems study at each location.

 
Wheat and grain sorghum planting dates were set at 1 October and 24 May, respectively. Simulated plant densities of 2100000 plants ha–1 were used for wheat and 20000 plants ha–1 for grain sorghum. Fertilizer application rates were set at similar levels as those used in the field experiments. Genetic coefficients supplied with the DSSAT system that describe the wheat variety ‘Newton’ and grain sorghum hybrid ‘Pioneer 8333’ were used as they represented cultivars most similar to those being used in the region at the time the experiments were initiated. A series of simulations with staggered starting years were used so that each crop in each crop rotation would be simulated each year.

Soybean–Sorghum Rotation
The second rotation study involved a summer crop rotation. Measured data were collected as part of a study conducted at the North Central Experiment Field near Belleville, KS. The study was conduced from 1982 through 2000 and included N treatments of 0, 27, 54, and 80 kg N ha–1 superimposed on continuous grain sorghum and a soybean–grain sorghum rotation (Gordon et al., 2001). Measured data were used to evaluate output from the sorghum model contained in the generic CERES and soybean model in CROPGRO 5.42 (Jones et al., 1989). Soil type at this site is a Crete silt loam (fine, smectitic, mesic Pachic Argiustolls). Soil physical properties were characterized at this site by Stone et al. (1978) and supplemented with data from the NRCS Soils Characterization Database (Soil Survey Staff, 2004) to serve as the input data source (Table 1). Weather data used for these simulations were measured at Scandia, KS, approximately 12 km from the research site, from 1961 to 2000, using automated weather stations to record maximum and minimum temperatures and incident solar radiation. Precipitation data were measured at the research site. Planting date of 24 May was used for both crops as it represents the optimum recommended planting date for both crops in the region. Simulated plant densities of 150000 plants ha–1 for grain sorghum and 450000 plants ha–1 for soybean were used. Genetic coefficients supplied with the DSSAT system that describe the grain sorghum hybrid Pioneer 8333 and Generic Group III soybean were used as they represented cultivars most similar to those being used in the region at the time the experiments were initiated. A series of simulations with staggered starting years were used so that each crop in each crop rotation would be simulated each year.

Statistics
When measured values were available, several statistics were used to describe the relationships between measured and simulated values. Willmott (1982) suggested that bias (Eq. [1]) and root mean square error (RMSE) (Eq. [2]), although less sensitive to extremes, are among the most appropriate overall measures of model performance.


[1]

[2]


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Wheat–Fallow vs. Wheat–Sorghum–Fallow
Comparisons of simulated and measured yields during the 15 yr from which data were collected illustrate that CERES-Wheat performed better than CERES-Sorghum (Table 2). CERES-Wheat overestimated wheat yields by 10% (RMSE = 1477 kg ha–1) with a bias of –227 kg ha–1 when simulated in a WSF system and overestimated wheat yields by 22% (RMSE = 1439 kg ha–1) with a bias of –564 kg ha–1 when wheat was simulated in a WF system. On average, CERES-Wheat overestimated wheat yields by 16% with a bias of –395 kg ha–1.


View this table:
[in this window]
[in a new window]
  		Table 2. Simulated and measured wheat (W) and grain sorghum (S) yields in both a wheat–fallow (W-F) and wheat–sorghum–fallow (WSF) trial conducted at Tribune, KS, from 1974 to 1997.

 
Our results contain greater errors in simulating wheat than errors reported by others using CERES-Wheat. Bannayan et al. (2003) reported a relative error for CERES-Wheat of approximately 3% with a RMSE of 839 kg ha–1. Jamieson et al. (1998) reported similar results from Australia with a relative error of approximately 2% with a RMSE of 902 kg ha–1. Both of these reported higher mean yields than experienced in our simulations, suggesting that CERES-Wheat does not perform as well under dryland conditions such as those experienced in the semiarid Great Plains.

The largest simulation errors were associated with overestimates of yield. In the WSF rotation, errors in simulated wheat yields were primarily related to 6 yr wherein the difference between simulated and measured yields exceeded 1000 kg ha–1 (Fig. 1A). In four of these years, simulated yields were higher than measured yields, with the greatest difference of 3580 kg ha–1 occurring in 1981. In 2 of the 6 yr, simulated yields were lower than measured yields where the greatest difference of 2525 kg ha–1 occurred in 1985. Similarly in the WF rotation, errors in simulated wheat yields were the result of 6 yr in which the differences between simulated and measured yields exceeded 1000 kg ha–1 (Fig. 1B). However, in five of these years, simulated yields were higher than measured yields, with the greatest difference of 4644 kg ha–1 also occurred in 1980.



View larger version (26K):
[in this window]
[in a new window]
  		Fig. 1. Simulated ({circ}) and measured (•) crop yields from 1961 to 2000, and wheat yields at Tribune, KS, from 1974 to 1987 for (A) wheat in a wheat–sorghum–fallow (WSF) rotation, (B) wheat in a wheat–fallow (WF) rotation, and (C) grain sorghum in WSF rotation.

 
It is unclear why CERES-Wheat overestimated wheat yields in both the WSF and WF systems. Jamieson et al. (1998) reported that CERES-Wheat overestimated leaf area index (LAI). If LAIs were overestimated in our simulations, it is possible that when soil water was adequate, these higher LAI values translated into higher growth rates and subsequently higher biomass and yield estimates compared with measured values. Simulated wheat yields were closely correlated with simulated water use, further implicating the impact of overestimating LAI on simulated yields. Since LAIs were not directly measured by Norwood et al. (1990), it is difficult to determine the actual reasons for the overestimations.

An additional source of potential errors in simulated wheat yields occurred as a result of CERES-Wheat predicting winter temperature damage more frequently than observed in wheat performance tests during this same period (Walter, 1989, 1990; Roozeboom, 1991, 1996, 1997). Simulated wheat yields of zero occurred five times in both the WSF and WF rotation as a result of complete simulated plant death from extremely low temperatures. The variety used for these simulations, Newton (Heyne and Niblett, 1978), is rated as having adequate winter hardiness for the region. Since winter hardiness is not a variable considered in the genetics coefficients used to distinguish different wheat cultivars, these results indicate that the methods implemented in CERES-Wheat to simulate cold weather damage need to be evaluated before adoption as a viable cropping systems research tool in the Great Plains.

CERES-Sorghum underestimated sorghum yields by 25% (RMSE = 1589 kg ha–1) with a bias of 722 kg ha–1 in the WSF system (Table 2). CERES-Sorghum underestimated sorghum yields in 10 of the 15 yr simulated (Fig. 1C), with errors exceeding 1000 kg ha–1 in 6 of the 10 yr. Simulated grain sorghum yields were consistently lower than measured yields from 1980 through 1986. Simulated wheat yields were overestimated several times during this period, suggesting that overestimated wheat yields may have led to overestimated water use, thus reducing the water available to the subsequent grain sorghum simulation. However, in 5 of the 7 yr, adequate precipitation during the fallow period eliminated soil water deficits before the start of sorghum simulations. Periods of drought or high temperatures were experienced at the research site from 1980 to 1986. Examination of the simulated growth and water use data indicates that when CERES-Sorghum encountered water stress, it subsequently slowed or ceased growth processes in response to the simulated stress. Apparently, CERES-Sorghum estimates water stress effects more severely than under actual condition, as indicted by the underestimated yields. It also appeared that the model was unable to recover from stresses if the high temperature or water stress were relieved, which occurred in the simulations for 1981 and 1984 growing seasons. Conditions in 1981 were characterized as hot and dry early but improved as the summer progressed (Walter and Fjell, 1981). The simulated growth data indicated that LAI peaked at 1.23 approximately 40 d after planting and then declined rapidly to plateau at 0.66, despite several rains during this period. Conditions in 1984 at the research site were characterized as hot and dry in June although good yields were reported as a result of excellent stored soil moisture at planting (Walter, 1985). Simulated soil water values followed the same trend of a gradual decline as the season progressed with only two significant precipitation events. Apparently, simulated water stress levels were more severe than actually occurred as the measured yield was 4344 kg ha–1 and simulated yields were 1517 kg ha–1 (Fig. 1C). These results are similar to those of Matowo (1993), who reported that CERES-Sorghum underestimated sorghum yields under simulated drought stress conditions.

It is not likely that climatic data input errors are the source of errors for either wheat or grain sorghum as values were measured at or near the location of the research trials and data are filtered to remove outliers in the data set. It is doubtful that soil physical properties are a significant source of errors as soil physical properties at the research site have been documented (Stone et al., 1987). It is difficult to determine if improvements in simulated yields could be realized by more accurately characterizing the genetic coefficients for the hybrid used in this study since most errors were attributed to overestimating water stress and water stress functions are not controlled by the genetic coefficients employed by CERES-Sorghum.

Soybean–Sorghum Rotation
As with the WSF simulations, CERES-Sorghum underestimated sorghum yields by 29% (RMSE = 2033 kg ha–1) with a bias of 383 kg ha–1 (Fig. 2). When simulating continuous sorghum, CERES-Sorghum underestimated yield by 22% (RMSE = 1525 kg ha–1) with a bias of 859 kg ha–1, and when simulating sorghum grown in rotation, the error was 35% (RMSE = 2436 kg ha–1) with a bias of 1735 kg ha–1.



View larger version (20K):
[in this window]
[in a new window]
  		Fig. 2. Simulated ({square}) and measured ({blacksquare}) grain sorghum yields in continuous grain sorghum and simulated ({circ}) and measured (•) grain sorghum yields in a soybean–sorghum rotation at Belleville, KS, at four N rates from 1982 to 2000.

 
Simulated grain sorghum yields and associated errors increased with time in both continuous cropping and soybean rotation systems (Fig. 3). Average simulated yields from 1982 to 1991 were 2641 and 2817 kg ha–1 for the continuous and rotated sorghum, respectively. Measured yields during this same time period were 2844 and 3725 kg ha–1 with errors of 7 and 24% for the same systems, respectively. From 1992 to 2000, simulated yields for continuous sorghum were 3253 kg ha–1, and simulated yields for sorghum grown in rotation with soybean were 3612 kg ha–1. When compared with measured values of 4757 and 6344, respectively, for the two systems, the simulated errors increase to 32% for the continuous system and 43% for the rotation. It is likely that selecting one hybrid for simulating the entire period did not account for the genetic improvement that occurred over this period in the field plots. Simulated yields did increase from one time period to the next, lending some support for variations in environmental factors (e.g., higher growing season precipitation and cooler temperatures) during one 10-yr period compared with another, but the increase in measured values between these two periods was twice that of the simulated values (20% compared with 41%). Hybrids selected for use in this study were re-evaluated approximately every 5 yr based on local performance tests, and leading hybrids were used over time to reflect local conditions. The use of genetic coefficients that characterized Pioneer 8333 appeared to be more suitable for simulating hybrids and conditions during the first 10 yr of the study compared with the last 9 yr. The use of one hybrid in the previous section appeared to have less impact than drought stress in the more semiarid environment of west-central Kansas compared with north-central Kansas.



View larger version (31K):
[in this window]
[in a new window]
  		Fig. 3. Simulated ({circ}) and measured (•) grain sorghum yields and associated errors ({blacktriangleup}) at Belleville, KS, from 1982 to 2000 a (A) continuous grain sorghum and (B) grain sorghum–soybean rotation.

 
Another source of errors for CERES-Sorghum occurred as a result of it not being responsive to N inputs in rotation with soybean. In both continuous sorghum and rotated sorghum simulations, errors increased as N rates increased (data not shown). Errors increased when sorghum was rotated with soybean (r = 0.95) compared with continuous cropping (r = 0.70). Measured yields responded in a quadratic manner with yields optimized at approximately 27 kg N ha–1, whereas no N response was observed with the simulated yields. In the continuous sorghum system, simulated yield response to N was similar to the measured response except that yields were 22% lower. Zewdie (1999) reported that CERES-Sorghum did not respond to N under water stress but was responsive when irrigation reduced water stress. Our results suggest that the model may not be responsive in dryland systems where water stress is not as frequent.

The soybean model simulated soybean with a high degree of accuracy (Fig. 4). Although the relative error between measured and simulated was 20%, the RMSE was extremely small (648 kg ha–1). Another encouraging aspect was that the simulated response to annual climatic conditions followed a similar trend as the measured data.



View larger version (20K):
[in this window]
[in a new window]
  		Fig. 4. Simulated ({circ}) and measured (•) soybean yields at Belleville, KS, from 1982 to 2000.

 

   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
The primary goal of crop simulation models as a research tool for dryland cropping systems, such as those of interest in the semiarid Great Plains, should be to adequately characterize expected system responses to crop management decisions such as potential crops to be considered and feasibility of intensifying rotations.

When CERES-Wheat and CERES-Sorghum were used to simulate WSF and WF rotations, CERES-Wheat overestimated wheat yields by approximately 16%. The model may be overestimating leaf area development, resulting in higher estimations of biomass production when adequate water supplies were available. Modifications may be needed in the model subroutines as CERES-Wheat estimated complete stand losses resulting from low winter temperature damage in five separate years when no stand losses were recorded in local wheat variety performance tests.

CERES-Sorghum consistently underestimated sorghum by approximately 27% across the two studies simulated. In a dryer environment, annual simulation errors were greater in years with simulated and observed water stress. It was also not responsive to N treatments when simulations of grain sorghum grown in rotation with soybean were examined. Some of the overall errors in simulating sorghum yields were attributed to the use of one set of genetic coefficients over the life of a 19-yr experiment that likely do not represent the more advanced hybrids used in the field study during the past decade.

Soybean yield simulations by CROPGRO were within 20% of measured values but did respond in a similar pattern on an annual basis as the measured values. Despite the underestimation, had these simulations been performed before initiating the experiment in 1982, it is likely that researchers would have included soybean in rotation experiments in the region.

Despite the simulation errors, if the cropping system simulations had been conducted before the initiation of both field research studies, the simulated wheat, grain sorghum, and soybean data would have led researchers to similar conclusions as the field studies, with the exception of the appropriate N response of grain sorghum following soybean. These results illustrate that with the correct input data (soil physical properties and weather and genetic data), crop simulation models can serve as effective cropping systems research tools. These simulation results might have culminated in a more efficient use of resources by identifying potential problems such as the need for winter-hardy wheat varieties and a better understanding of the N dynamics in a rotation containing soybean.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 





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 Google Scholar
Google Scholar
Right arrow Articles by Staggenborg, S. A.
Right arrow Articles by Vanderlip, R. L.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Staggenborg, S. A.
Right arrow Articles by Vanderlip, R. L.
Agricola
Right arrow Articles by Staggenborg, S. A.
Right arrow Articles by Vanderlip, R. L.
Related Collections
Right arrow Soybean
Right arrow Dryland Cropping Systems
Right arrow Sorghum
Right arrow Wheat
Right arrow Agricultural Systems
Right arrow Crop Models
Right arrow Crop Ecology

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Crop Science Vadose Zone Journal
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome