| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Dep. of Agronomy and Horticulture, Univ. of Nebraska-Lincoln, P.O. Box 830915, Lincoln, NE 68583
* Corresponding author (hyang2{at}unl.edu)
Received for publication May 26, 2005.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONFEATURES OF THE HYBRID-MAIZE...MODEL INPUT REQUIREMENTSAPPLICATIONSDISCUSSION AND CONCLUSIONSREFERENCES |
|---|
Abbreviations: CRM, comparative relative maturity • DAE, days after emergence • GDD, growing degree days • HPRCC, High Plains Regional Climate Center • LAI, leaf area index
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONFEATURES OF THE HYBRID-MAIZE...MODEL INPUT REQUIREMENTSAPPLICATIONSDISCUSSION AND CONCLUSIONSREFERENCES |
|---|
Given this situation, there is a need for developing crop models that are robust in performance, easy to use, flexible in parameter settings, transparent in formulation and well supported with a comprehensive user's guide and program documentation. Moreover, to facilitate the use of crop models among clientele with diverse needs and/or computer literacy, it is desirable for the models to come equipped with auxiliary functions and utilities that facilitate data handling (e.g., weather data acquisition, proofing, and formatting) and intuitive presentation of model ouput.
We have recently developed a new corn simulation model, Hybrid-Maize. Details about this model and its validation are published elsewhere (Yang et al., 2004a, 2004b). Briefly, the Hybrid-Maize model builds on the strengths of existing models by combining the crop-specific attributes of CERES-Maize (Jones and Kiniry, 1986; Kiniry et al., 1997) related to phenology and organ growth with explicit photosynthesis and respiration functions from assimilate-driven generic crop models such as SUCROS, WOFOST, and INTERCOM (van Diepen et al., 1989; Kropff and van Laar, 1993; van Ittersum et al., 2003). Hybrid-Maize also includes additional modifications for several functions based on calibration with experimental data from a field study that produced maize with minimal possible stress—conditions that are required to achieve yield potential (Yang et al., 2004a). Since its release in August 2004, the model is being used by researchers, extension specialists and educators, crop consultants, industry professionals, and farmers. The objective of this paper is to provide an overview of the Hybrid-Maize software, with emphasis on its capabilities and potential applications based on our own experience and feedback from users. We also discuss limitations of the model and opportunities for future improvement.
|
FEATURES OF THE HYBRID-MAIZE SOFTWARE |
|---|
|
TOPABSTRACTINTRODUCTIONFEATURES OF THE HYBRID-MAIZE...MODEL INPUT REQUIREMENTSAPPLICATIONSDISCUSSION AND CONCLUSIONSREFERENCES |
|---|
To support research applications, the Hybrid-Maize model is transparent in formulation, flexible in parameter settings, and comprehensive in documentation to facilitate model validation and further development. It provides tools to evaluate crop growth during the growing season and at maturity in terms of dry matter accumulation in roots, stems, leaves, and developing grain. It also provides information that helps explain crop performance, such as gross assimilation, respiration, actual evapotranspiration, and soil moisture depletion in the root zone. To facilitate its use in crop management decisions, on the other hand, the model also is designed to be easy to operate with a minimum number of input variables compared to other corn simulation models. It has a user-friendly input and output interface and auxiliary functions and utilities that facilitate data file handling such as weather data compilation, unit conversion, editing/saving/retrieving of model settings, default parameters, and output. To avoid unrealistic simulations, Hybrid-Maize confines input settings to reasonable limits or provides warning messages when input values fall outside the range of conditions for which the model has been validated.
User Interface Features
A flowchart of input settings, operation, and presentation of model outputs is presented in Fig. 1
, and these components are implemented through a multi-page graphical user interface (Fig. 2
). These pages are identified by seven tabs: (1) ‘Input’ for input and simulation mode settings, (2) ‘Results’ for summary of numerical outputs, (3) ‘Chart’ for display of any of 17 output variables for across-run comparisons, (4) ‘Growth’ for graphical display of growth dynamics of eight variables, (5) ‘Weather’ for display of growing season climate data, (6) ‘Water’ for growing season soil moisture regime and crop water stress index, and (7) ‘Yield trend’ for displaying trends in in-season yield forecasts. Most input settings have multiple choices, and some choices determine the need for additional input parameters. Therefore, only ‘live’ input boxes are accessible depending on selection of a particular simulation mode while input boxes for parameters that are not needed are grayed out and inaccessible. For example, if the ‘Optimal’ water regime is selected as the simulation mode, then the input boxes for irrigation timing and soil properties are not accessible as they are not needed.
|
|
A unique feature of Hybrid-Maize is that all key internal model parameters are accessible to the user (Fig. 3 ). Each parameter has a text description and some have references to journal papers from which default values were obtained or derived. The user can modify the values for each of these parameters, but there is also the option of restoring all of the original default values with a single mouse click.
|
The software includes a context-sensitive help system that can be accessed by pressing F1 on the keyboard or through a pull-down Help menu. Other utilities include easy access to changing the settings for the initial default values for input parameters, folders for working files and weather files, saving and retrieving of input settings from past simulation runs, and saving and printing of simulation results. Window's default text editor and calculator can also be brought up directly within Hybrid-Maize without exiting the software. A complete publication list of all citations used as background information for developing model formulations and default input parameter settings and a printable users manual with examples of potential applications are provided in the Help menu.
|
MODEL INPUT REQUIREMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONFEATURES OF THE HYBRID-MAIZE...MODEL INPUT REQUIREMENTSAPPLICATIONSDISCUSSION AND CONCLUSIONSREFERENCES |
|---|
Preparation and error checking of weather data is one of the most difficult tasks required before any crop model can be used. In Hybrid-Maize, the weather file is a text file and its format is flexible in terms of data positioning. One file contains all years of weather data for a single site with text description of the site as well as titles and units for variables. For users within the coverage of the High Plains Regional Climate Center (HPRCC, www.hprcc.unl.edu, verified 28 Feb. 2006) located in Lincoln, NE, the model contains an automatic conversion utility to convert the weather data downloaded from the HPRCC to the format and units required by the model. Weather data from other sources can also be prepared easily in a spreadsheet program following the instructions in the program's user guide under the help button. Because errors and missing data are common in large weather datasets, Hybrid-Maize has a utility for identifying outlier values in the daily weather data entries and can replace missing data. Suspicious or missing data are reported and recorded in a text file so that the user can easily locate file errors to correct errant values.
Simulation Modes
Three simulation modes are available: (1) ‘Long-term runs’ using a selected set or all years of available historical site weather data, (2) ‘Single year’ simulations with or without a long-term simulation run for comparison, and (3) ‘Current season prediction’ to the end of the growing season.
The ‘Long-term runs’ mode (1) is used for estimating the long-term site yield potential and its variability under an optimal moisture regime without water stress, or for estimating the attainable water-limited yield potential under rainfed conditions. For both cases, yield potential can be evaluated with respect to different hybrid maturities, planting dates, and plant populations. This mode can be used to estimate the impact on yield from changes to one or more of these management practices. While the default mode simulates across all years, the user can also select specific periods within historical weather database. All simulations results are ranked by predicted grain yield and a detailed summary of results are reported for five rank years (i.e., the years with the highest, 75th percentile, median, 25th percentile, and worst yield). In addition, predicted mean values and coefficients of variation (CV) are computed for key model outputs. Overall probability of frost damage during grain filling is also reported for the time series in the long-term simulation mode. For each rank year, detailed outputs are reported for other parameters such as stover yield and harvest index, days to silking and maturity, total duration of vegetative and grain-filling periods, total cumulative solar radiation, mean maximum, minimum and average daily temperature during the vegetative and grain-filling stages, and total precipitation during the growing season. To investigate site yield potential and yield variability using the long-term run mode, it is recommended that the historical weather file includes at least 10 yr of data.
Single year without long-term runs mode (2) is primarily used to evaluate the influence of management or climate factors on yield in an individual year, or to estimate the size of the exploitable yield gap by comparing simulated yield potential with actual measured yields. Typically, this type of simulation constitutes a post-hoc analysis of a past growing season. Up to six individual runs can be made sequentially and their results can be compared, both numerically and graphically on the output pages. Single year with long-term runs mode is useful for comparing a given year with the long-term site yield potential as simulated for all other years in the weather database for that site. In this mode, the user can investigate why the yield potential in a given year was above or below average and which climatic factors may have contributed to observed differences across years. All simulations are ranked by predicted grain yield and detailed summary results are reported for the chosen year as well as the years that represented the highest, 75th percentile, median, 25th percentile, and worst yield within the population of years in the weather database.
‘Current season prediction’ mode (3) allows in-season (or real-time) assessment of corn growth up to the current date based on the actual weather data up to that point in time, followed by prediction of growth and final yield thereafter based on historical weather data for the remainder of the growing season (Fig. 4 ). Similar to the ‘Long-term runs’ mode, predictions are ranked according to grain yield and results are shown for the scenarios with the highest, 75th percentile, median, 25th percentile, and worst yields.
|
Maturity can be set as either the actual observed date of maturity (e.g., in ‘single-run’ mode) or as total growing degree days (GDD) from emergence to physiological maturity (blacklayer) of the hybrid (e.g., in ‘long-term runs’ and ‘current-season prediction’ modes). Most commercial seed companies publish GDD values for their hybrids. The base temperature for GDD is 10°C (metric system) or 50°F (English system). If information on GDD values to physiological maturity is not available for a hybrid, users can enter a comparative relative maturity (CRM) in days as an alternative. In such cases, the model uses seed brand-specific or generic relationships between GDD and CRM to estimate GDD for reaching maturity.
Accurate estimation of silking date is crucial for simulating grain yield because it represents the transition from vegetative to reproductive growth stages, and grain yield is extremely sensitive to the duration of the grain-filling phase (Yang et al., 2004a). The software offers the option of entering either the observed date of silking (e.g., in ‘single run’ mode) or GDD from emergence to silking. If neither is known, the program will estimate the date of silking using an internal algorithm based on seed brand specific or generic relationships between GDD to physiological maturity and GDD to silking (Yang et al., 2004a).
Setting plant population is straightforward. For settings of both total GDD and plant population, there are default minimum and maximum limits for these values and a warning message will pop up if the settings are outside the range for which Hybrid Maize has been validated. These limits, however, may be modified by the user through the ‘Parameter settings’ menu.
Optimal or water-limited growth simulation can be accomplished by choosing either ‘Optimal’ or ‘Rainfed/Irrigated’ modes (Fig. 2). If ‘Optimal’ is selected, there is no need to enter soil properties (the whole soil panel is grayed out), as soil is assumed to provide optimal water supply. However, if the option ‘Estimate irrigation water requirement’ is selected, soil physical properties must be entered as for rainfed and irrigated crop simulations. Required soil properties include estimated initial gravimetric moisture content of topsoil (i.e., 0–30 cm), maximum rooting depth, bulk density, and textural class of topsoil and subsoil. When ‘Rainfed/Irrigated’ is selected, an irrigation schedule can also be specified (dates and amount); if it is left blank the model will assume rainfed conditions. Different irrigation strategies can be explored by changing irrigation times and amounts and evaluating the impact on yield. The irrigation schedule can be saved/retrieved along with other model settings for use in future simulations.
|
APPLICATIONS |
|---|
|
TOPABSTRACTINTRODUCTIONFEATURES OF THE HYBRID-MAIZE...MODEL INPUT REQUIREMENTSAPPLICATIONSDISCUSSION AND CONCLUSIONSREFERENCES |
|---|
Input settings (in metric units) for this exercise were: Mode = long-term run from 1986 to 2004, start from planting on 1 May with a planting depth of 4 cm, maturity = total GDD 1500 (CRM of about 112 d), population = 79 000 ha–1, water = optimal. The results screen from this simulation provides a summary of the outputs and is shown as "simulation 1" in the upper section of Fig. 5 .
|
Upon analysis of these simulated data, the question arises as to whether yield potential in most years could be increased by shifting grain filling later into September when nighttime temperatures begin to cool, but without significantly increasing the risk of frost. This could be accomplished by planting at a later date, choosing a longer maturity hybrid, or a combination of both. We first explored a later planting date (10 May) while keeping the same hybrid (1500 GDD), which would, on average, initiate grain filling at a later date. Simulation 2 in Fig. 5 shows the results from this change. The model predicted a long-term average yield potential of 14.2 Mg ha–1, ranging from 11.9 to 17.3 Mg ha–1 during the 19-yr period in the weather file. Risk of frost occurrence during grain filling was still zero. Overall, planting the same hybrid 10 d later predicted a slight increase in yield potential, which resulted from a small increase in the grain-filling period (long-term average rDays was 53 for the 10 May planting vs. 51 for the 1 May planting).
The effect of planting a longer maturity hybrid was evaluated in simulation 3 while keeping planting date and other input settings the same as in simulation 1. Because hybrids with a relative maturity rating of about 119 d are available for this environment, the GDD to maturity was set at 1590. The model predicted an average yield potential of about 15.5 Mg ha–1, with a range of 12.7 to 18.2 Mg ha–1 (simulation 3 in Fig. 5). This represents a 12% increase compared to planting a 112-d hybrid with total GDD of 1500. With the longer maturity hybrid the average length of grain filling increased from 51 to 57 d, which is the primary reason for the increased yield compared to simulations with the earlier maturing hybrid.
The results from simulations 1 to 3 raise the issue of whether yield potential could be further increased by use of a longer maturity hybrid in combination with a later planting date. A fourth simulation was run with the same settings as in simulation 3 but with a later planting on 10 May. The results from this combination further increased the predicted average yield potential to 15.9 Mg ha–1 and also narrowed the overall range to 13.7 to 17.8 Mg ha–1, which indicates greater yield stability in addition to higher mean yield (simulation 4 in Fig. 5). The average length of the grain-filling period increased to 59 d and there is only small (5%) risk of frost occurrence before the crop reaches maturity under this scenario.
While additional combinations of planting dates and hybrid maturity can be explored, other factors must be considered in selecting the most appropriate combination. For example, the model suggests that planting the 119-d hybrid on 30 May may further increase yield potential at Lincoln to an average of 17.1 Mg ha–1, but at the increased risk of frost risk (17%). The logistics of harvesting, harvest losses, and cost of grain drying due to much later maturity in mid-October must also be considered.
In summary, simple changes in hybrid selection and planting date resulted in substantial increases in predicted yield potential with greater yield stability and relatively low risk of frost compared to current recommended practices for southeast Nebraska. This modified management regime achieves higher yields through a longer grain-filling period as a result of delaying the grain-filling period into late September when night temperatures are much cooler than late August and early September. It should be noted, however, that mean temperatures at this site within the City of Lincoln are warmer than in surrounding rural environments and therefore these results may not be widely applicable to the surrounding region. However, the exercise demonstrates the use of Hybrid-Maize to evaluate changes in management. It is also noteworthy that recent field studies at the Lincoln research site have confirmed the predicted improvements in yield potential from later planting with a longer maturity hybrid (Dobermann et al., 2005).
Example 2: Estimating Water Requirements for Optimal Yields
Another potential model application is to estimate crop water requirements and irrigation timing. As an example, we will simulate irrigated corn in southwest Nebraska. This is an area with high yield potential (elevation about 1000 m, semiarid climate with high solar radiation), provided that crops are fully irrigated with optimal nutrient supply and pest management. Rainfall is highly variable from year to year and averages about 500 mm annually, with about half coming during the growing season. Standard practices for irrigated corn are to plant around 1 May at 0.76-m row spacing and a final population of 74 000 plants ha–1. Common hybrids in this area require total GDD of 1440. A weather station representing this area is located at Champion, NE (40.40° N, 101.72° W) with daily climate data available for the 1982 to 2003 period. The site simulated is a gently sloping field with a deep, well-drained, fine-loamy soil. Rooting depth is not limited by a hardpan or compacted layer, and general soil quality is good.
A long-term simulation run was performed for this site for the period of 1982 to 2003 using the ‘Optimal’ mode. The average yield potential was 15.9 Mg ha–1 but with a wide range (12.0–20.5 Mg ha–1) and a 27% probability of premature frost. The grain-filling period is relatively long (average of 67 d) due to cooler average temperatures than at Lincoln; nighttime temperatures during grain filling are considerably cooler than at Lincoln (rTmean averaged 21°C). Given these results, there is less potential to extend the growing season because of increased risk of yield loss from frost damage. The key production constraint is water availability for irrigation in this semiarid environment.
To illustrate the severity of water stress in this region, a run was made using the same settings as above but for rainfed conditions. Settings for soil properties were moisture content of topsoil = 25% (w/w) at planting, maximum rooting depth = 1 m, bulk density and texture of topsoil and subsoil were 1.3 g cm–3 and loam, and 1.4 g cm–3 and loam, respectively. Without irrigation, attainable yield averaged 9.1 Mg ha–1 with 264 mm of rainfall during the growing season, but can be as low as 4.4 Mg ha–1 with 105 mm of rainfall with poor distribution, or as high as 17.5 Mg ha–1 with 522 mm of rainfall with good distribution throughout the growing season.
Irrigation is needed at critical growth stages to overcome the water deficit that occurs in this region. To estimate the water requirement for maintaining adequate water supply throughout the crop growth period, we assumed that the maximum amount of deliverable water per (sprinkler) irrigation event was 32 mm (through ‘Settings, Parameter settings’ in the menu). The summary of simulation output is similar to Fig. 5, but an additional column reports the total irrigation water requirement (not shown). On average, 296 mm of additional water (11.7 inches) is required to achieve stress-free growth, but the water requirement varies from about 226 (1999, the best year) to 356 mm (1997, 75% percentile of grain yield). The amount and predicted timing of irrigation for the year of median yield (1991) are shown in Fig. 6 , along with daily rainfall and soil moisture dynamics in three layers (i.e., 0–30; 30–60; and <60 cm). Other years can be selected for display from the dropdown list, and numerical outputs are also available on the ‘Results’ page.
|
Example 3: Real-time Crop Growth Simulation and Yield Forecast
Corn yield potential varies considerably from year to year in the same field as a result of the combined effects of variation in solar radiation and temperature in irrigated systems as well as rainfall in rainfed systems. In-season crop model predictions can be used to guide management and marketing decisions, along with other sources of information, common sense, and experience (Dobermann and Yang, 2004). The ‘Current season prediction’ mode in Hybrid-Maize allows the user to simulate "real-time" crop growth during a growing season up to the date of the simulation run. To interpret real-time simulations, however, we recommend that the weather file contain at least 10 yr of representative historical climate data for the site (or a nearby weather station) in addition to weather data for the current year.
This real-time simulation example involves 2003 data from two fields in eastern Nebraska: one from Lincoln (irrigated) and the other from Mead (rainfed). Yield forecasts were made every 5 d, beginning shortly after planting (Fig. 7 ). At each forecasting date, actual weather data were used in Hybrid-Maize to simulate growth until that date. From that point forward to maturity, the model used historical weather records to simulate possible growth scenarios for the remainder of the season.
|
At Mead, the crop was grown under rainfed conditions in a production field at 59 000 plants ha–1. Predicted median yield was close to the long-term median water-limited yield potential early in the season (Fig. 7b, upper dashed line), but by silking predicted yields began to fall because of less than normal rainfall. Predicted water-limited yield potential continued to decline throughout the remainder of the growing season because of continued drought. By the end of August the range of predicted outcomes indicated a range in final yield of from 6 to 9 Mg ha–1, which is 46 to 69% of the long-term median rainfed yield potential at this site. The final measured grain yield of 8.0 Mg ha–1 confirmed this prediction.
|
DISCUSSION AND CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONFEATURES OF THE HYBRID-MAIZE...MODEL INPUT REQUIREMENTSAPPLICATIONSDISCUSSION AND CONCLUSIONSREFERENCES |
|---|
So far the Hybrid-Maize model has mainly been validated with plant densities of 60 000 to 100 000 plants ha–1 and commercial hybrids that predominantly have a single ear with relatively little prolificacy. The model performance needs further evaluation before it can be used with confidence outside this range in plant population. The formulation proposed recently by Ritchie and Alagarswamy (2003) for predicting yield components provides a promising avenue to potentially improve model performance across a wider range of populations. Although Hybrid-Maize simulates corn growth and yields under optimal water regime as well as water-limited conditions, most of the model validation so far has been conducted under optimal water conditions in the western and central U.S. Corn Belt. Therefore, simulation results under water-limited conditions must be interpreted and used with caution, particularly in areas that are prone to severe drought and heat stress at key growth stages.
Sensitivity analysis in Yang et al. (2004a) indicates that model results are sensitive to dates of critical growth stages, which can either be provided as an input to the model or predicted by the model. The dates of silking and physiological maturity as defined by Ritchie et al. (1992) are particularly sensitive. Those dates should only be entered if accurate measurements have been taken. Otherwise, the model appears to do a reasonable job of estimating the date of these events. Simulated yield can be seriously affected by entering wrong silking or maturity dates, or an incorrect value for GDD to maturity for a commercial hybrid. When in doubt, use the GDD option to let the model predict silking and maturity dates based on the total GDD to maturity for the hybrid used, and be sure to check that the correct GDD value has been entered. In some unusual cases, however, this approach may lead to an unrealistically long predicted growing season when a full-season hybrid is planted late and cool weather predominates during last half of grain filling. There is a need to verify predicted maturity dates with observations, and to use common sense about sufficient time for grain dry down to allow timely harvest and to avoid additional grain drying costs.
As the model is based on mechanistic descriptions of corn growth and development, it is expected to perform well in a wide range of environments. Nevertheless, caution should be exercised when using Hybrid-Maize outside the U.S. Corn Belt as this may require changes in some of the default model parameters. In exploring site yield potential with the model, it is important to note that yield potential can only be achieved under growth conditions that are ‘ideal’ with regard to both crop and soil management. Reasonable soil quality is also required. Although it is theoretically possible to overcome shallow soil depth or a hardpan that restricts root growth by employing more precise management of nutrients and irrigation, it is generally not practical or profitable to do so at a production scale. Likewise, some soil constraints, such as salinity or soil acidity, reduce crop growth directly and therefore make it impossible to achieve yield levels that approach the genetic yield potential of a given hybrid even with optimal management of water, nutrients, and pests. Therefore, in interpreting investigations of site yield potential, model users must be aware of limitations to crop growth that are not considered in the model, such as soil compaction, shallow soil depth, sandy soil texture, or soil acidity or salinity. In general, investigation of site yield potential using Hybrid-Maize is most appropriate for fields in which soil quality is relatively good and there are no obvious constraints to crop growth.
As with all simulation models, Hybrid-Maize still represents a simplification of the ‘real-world’ system and, as such, model simulations may differ from actual outcomes. There are uncertainties or weaknesses in Hybrid-Maize that will have to be addressed in future research and subsequent model releases. They include:
Current and future work to improve the Hybrid-Maize model focuses on (i) addressing some of the uncertainties listed above, especially with regard to crop water stress; (ii) adding a module to help estimate N, P, and K fertilizer requirements; and (iii) adding a module to predict grain moisture content during the dry-down period to harvestable maturity. In addition, we continue to test and validate the model across a wider range of production environments and management levels. More information about the model's development, applications and information update is available at a dedicated website at www.hybridmaize.unl.edu. The model software is available for online purchase at http://estore.adec.edu (verified 3 Mar. 2006).
|
ACKNOWLEDGMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONFEATURES OF THE HYBRID-MAIZE...MODEL INPUT REQUIREMENTSAPPLICATIONSDISCUSSION AND CONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. L. Ping, R. B. Ferguson, and A. Dobermann Site-Specific Nitrogen and Plant Density Management in Irrigated Maize Agron. J., June 23, 2008; 100(4): 1193 - 1204. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 | |||
