Many simulation models use genetic characteristics in the form of rate coefficients or other system constants in crop growth. These coefficients or constants can be evaluated during the validation and sensitivity processes. Duncan and colleagues (1978) have shown, with a simulation model of peanut, dramatic yield increases solely from changes in flowering time and some aspect of photosynthate partitioning. Breeding work can be undertaken to benefit from these useful characteristics so as to produce high-yielding, insect-disease-resistant cultivars capable of competing with weeds.
With traditional scientific techniques, it is almost impossible to obtain data on the many physiological processes that are important from a crop physiology point of view. For example, Whisler and colleagues (1986) stated that turgor pressure in the cells is the driving force for leaf expansion, but unfortunately, it cannot be measured directly. These can only be inferred from measurements using a dynamic simulation model.
If the models are comprehensive in nature, they can be tested for various growth processes and can help in eliminating unrealistic hypotheses, saving time and energy. Furthermore, exercises using a model may give rise to many more experiments to test various hypotheses. This is the way by which models can be used to probe the physiology of plants, which is not experimentally accessible. Modeling and experimentation can be mutually supportive in developing our understanding of crop physiology.
In the sequence or crop-rotation analysis, one or more crop rotations can be analyzed. In this mode, different cropping sequences are simulated across multiple years. It is critical that in a crop-rotation analysis water, nitrogen, and carbon are simulated as a continuum. The main goal of a crop-sequence application is to determine the long-term change of soil variables as a function of different crop-rotation strategies (Bowen, Thornton, and Hoogenboom, 1998). Several models have been specifically developed to study the long-term dynamics of nitrogen and organic matter in soil. APSIM (Agricultural Production Systems Simulation Model) is used to evaluate crop sequences in the northern grain belt of Australia (McCown et al., 1996). Others have been specially developed to study the long-term sustain-ability of cropping systems (Thornton et al., 1995).
In strategic applications of crop simulation models and decision support systems, the models are mainly run to compare alternative crop management scenarios. This allows for the evaluation of various options that are available with respect to one or more management decisions (Tsuji, Hoog-
enboom, and Thornton, 1998). To account for the interaction of these management scenarios with weather conditions and the risk associated with unpredictable weather, simulations are conducted for at least 20 to 30 different weather seasons or weather years (James and Cutforth, 1996). In most cases, daily historical weather data are used as input, and the assumption is made that these historical weather data will represent the variability of weather conditions in the future. In addition, the biological outputs and management inputs can be combined with economic factors to determine the risk associated with the various management practices being evaluated (Thornton and Wilkens, 1998).
Hammer, Holzworth, and Stone (1996) calculated the benefits of seasonal forecasting for tactical management of nitrogen fertilizer and cultivar maturity of wheat at Goondiwindi, Australia. Using the SOI phase system, they found an increase in profit of 20 percent, or about $10.00/ha. They also showed the risk of making a loss could be reduced by as much as 35 percent. Marshall, Parton, and Hammer (1996) also used APSIM simulations for wheat to investigate how risk and planting conditions changed with an SOI-based forecast. Similar benefits of seasonal forecasts were observed on sorghum, sunflower, corn, and peanuts (Hammer, Carberry, and Stone, 2000; Meinke, 2000).
In tactical applications, crop models are actually run prior to or during the growing season to integrate the growth of a crop with the current observed weather conditions and to decide, on a daily basis, which management decisions should be made. In this regard, the uncertainty of weather conditions in modeling applications has to be managed. For any crop model run, only the weather data up to the previous day will be available. If the weather forecasts are provided in some type of quantitative format, they can also be included with the simulation. There are various methods for handling the uncertainty of future weather conditions. The first one is to use historical weather data and to run the system for multiple years. Instead of historical weather data, generated data can also be used. If multiple years of historical or generated weather data are used as input, a mean and associated error variable can be determined for predicted yield as well as for other predicted variables. Over time, the error will become smaller, as the uncertain weather forecast data are being replaced with observed weather variables. If two or more management alternatives are being compared, one can evaluate the risk associated with each management decision, using both mean and error values of each predicted variable.
Computer models and expert systems are extensively used in irrigation. Packages are available that deal with irrigation scheduling, irrigation system evaluation, crop planning and selection of crop varieties, and irrigation system operation.
In the area of pest and disease management, especially integrated pest management (IPM), the application of models has been shown to be very profitable (Pusey, 1997). As the application of pesticides is rather expensive, farmers are interested in minimizing their use, from both economic and environmental viewpoints.
Another area of application is in climate change impact assessment. As climate change deals with future issues, the use of general circulation models (GCMs) and crop simulation models provides a scientific approach to study the impact of climate change on agricultural production and world food security. Similarly, the issue of climate variability especially related to the variation in sea-surface temperature (SST) of the Pacific Ocean or El Niño/Southern Oscillation (ENSO) has opened an area in which crop simulation models also can play an important role. They can potentially be used to help determine the impact on agricultural production due to ENSO and recommend alternative management scenarios for farmers that might be affected, thereby mitigating the expected negative impacts of ENSO and capitalizing on the opportunities in better seasons.
The application of crop simulation models for forecasting and yield prediction is very similar to the tactical applications. However, in the tactical decision application, a farmer or consultant is mainly concerned with the management decisions made during the growing season. In the forecasting application of the crop models, the main interest is in the final yield and other variables predicted at the end of the season. Most of the national agricultural statistics services provide regular updates during the growing season of total area planted for each crop, as well as the expected yield levels. Based on the expected yield, the price of grain can vary significantly. It is important for companies to have a clear understanding of the market price so that they can minimize the cost of their inputs. Traditionally, many of the yield forecasts were based on a combination of scouting reports and statistical techniques. However, it seems that crop simulation models can play a critical role in crop-yield forecasting applications if accurate weather information is available, both with respect to observed conditions and to weather forecasts. The STIN (Stress Index) model (Stephens, Walker, and Lyons, 1994) has been officially used for forecasting Australian wheat production. Accurate applications of crop simulation models require, in many cases, some type of evaluation of the model with locally collected data. Especially for yield forecasting, it is critical that yields are predicted accurately, as pol icy decisions related to the purchase of food could be based on the outcome of these predictions. One option is to use remotely sensed data that are being used to estimate yield, based on a greenness index. A more advanced application would be to link physical remote sensing with crop simulation models. With this approach, the simulated biomass can be adjusted during the growing season, based on remotely sensed or satellite data, and yield predictions can be improved based on these adjusted biomass values (Maas, 1993).
One of the limitations of current crop simulation models is that they can simulate crop yield only for a particular site for which weather and soil data as well as crop management information are available. One recent advancement is the linkage of crop models with a geographic information system (GIS). A GIS is a spatial database in which the value of each attribute and its associated x- and y-coordinates are stored. To describe a specific situation, all the information available on a territory, such as water availability, soil types, forests, grasslands, climatic data, and land use are used. Each informative layer provides to the operator the possibility to consider its influence on the final result. However, more than the overlap of various themes, the relationships of the various themes is reproduced with simple formulas or with complex models. The final information is extracted using graphical representation or precise descriptive indexes (Hartkamp, White, and Hoog-enboom, 1999; Maracchi, Pérarnaud, and Kleschenko, 2000). This approach has opened a new field of crop modeling applications at a spatial scale, from the field level for site-specific management to the regional level for productivity analysis.
In seasonal analysis applications, a simulation model is used to evaluate a management decision for a single season. This can include crop and cultivar selection; plant density and spacing; planting date; timing and amount of irrigation applications; timing, amount, and type of fertilizer applications (Hodges, 1998); and other options a particular model might have. Model applications can also include investment decisions, such as those related to the purchase of irrigation systems.
Was this article helpful?