As the development of agricultural models has progressed, controversy has not been lacking. Traditionalists felt that any attempt to simulate highly complex biological systems using mathematical algorithms in computers was bound to fail. Others argued that agricultural models should have a largely heuristic role in research rather than use as predictive tools. Further controversy regarding agricultural models stems from problems of complexity, testability, and parametrization. A commonly accepted truism is that agricultural models should be no more complex than the level of theory and measurements used to build them, and that components of models should also be balanced in terms of levels of complexity. Although this concept is extremely hard to quantify, it is important to note that this definition of model complexity is related to the kinds of questions the model is designed to answer rather than how much empiricism the model contains or the overall number of parameters.
What is incontrovertible is that the use of agricultural models to solve significant issues related to the economic and environmental sustainability of agroecosystems is on the rise rather than on the decline. Furthermore, the collective experiences of agricultural model developers and users show that, even though they are far from perfect, agricultural system models can be very useful in guiding field research, aiding technology transfer, and generating credible assessments of various impacts (e.g., sustainability, management, environmental impacts, climate change) on farming systems. However, a number of needs remain to be addressed that could improve agricultural system models and their application. Important issues include:
1. better quantification of how abiotic factors (e.g., water, temperature, light, nutrients) affect plant growth for both species and subspecies, and most importantly, genotype x environment interactions;
2. relationships among plant growth and other biotic factors (e.g., weeds, insects, diseases), including quantification of the competitive component to allow for better estimation of the effect of dynamic variations in plant population on growth and partitioning and the competitive aspects of crop-weed and crop-pest interactions;
3. development of comprehensive and common shared experimental databases based on existing standard experimental protocols, with measured values related to modeling variables so that conceptual model parameters can be experimentally verified;
4. improved collaboration between agricultural model developers and field scientists for appropriate experimental data collection, and for evaluation and application of models;
5. better methods of determining model parameters for different spatial and temporal scales, and for scaling and aggregating simulation results from plots to fields and larger scales;
6. continued development and improvement of environmental modeling frameworks to encourage replacement of monolithic agricultural models with modular component-based modeling tools where (a) each model component can be independently tested, improved, and easily substituted; (b) model components can vary with the scale of application; (c) hierarchical parameter estimation from varying degrees of input information is a component of the model; and (d) assembled agricultural systems models are kept compact and easy to use by customizing them to specific problems and regions; and
7. better coordination of international efforts is needed in the future to improve agricultural system modeling and to encourage model developers and field scientists to work on identified knowledge gaps and research priorities.
Finally, while many important challenges and opportunities exist in model development, the greatest challenge facing the practitioners of agricultural system modeling in the future may center on demonstrating relevance to real-world decision-making rather than on building more accurate or comprehensive models.
Was this article helpful?