Skip to main content
ARS Home » Research » Publications at this Location » Publication #233345
Title: A Genetic/Heuristic Approach to Simulating Plant Height in Winter Wheat

Author
item WEISS, A - UNIVERSITY OF NEBRASKA
item BAENZIGER, P - UNIVERSITY OF NEBRASKA
item McMaster, Gregory
item Wilhelm, Wallace
item AL-AJLOUNI, Z - UNIVERSITY OF NEBRASKA

Submitted to: Symposium Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 10/20/2008
Publication Date: 11/2/2008
Citation: Weiss, A., Baenziger, P.S., Mcmaster, G.S., Wilhelm, W.W., Al-Ajlouni, Z.I. 2008. A Genetic/Heuristic Approach to Simulating Plant Height in Winter Wheat. Symposium Proceedings.

Interpretive Summary: An explosion of genetic information is rapidly emerging, yet understanding this information and using it to improve our crop simulation models is proving to be a difficult challenge. The problem is further exacerbated in that gene expression is likely influenced by environmental conditions, particularly water deficits in semi-arid production regions. Several approaches have tried to go from the gene to the whole plant in crop simulation modeling with varying success. In this paper, our objective was to use winter wheat (Triticum aestivum L.) as a case study to discuss how to better address these challenges by simulating plant height across a range of environments in Nebraska using genetic information. We chose plant height as part of our case study because although it is a complex trait, it is far simpler than many traits, and the major genes influencing plant height are discrete and well characterized genetically and phenotypically. We used two data sets from Nebraska in this study. We used an existing crop simulation model, CropSyst, and modified it by incorporating a different phenology algorithm and adding a plant height algorithm that we developed. The plant height algorithm was based on the ratio of the mean unstressed final height of a height class (tall semidwarf, semidwarf, short semidwarf) to the mean unstressed final heights for two tall cultivars, an environmental index based on the fraction of transpirable water, and the development stage. Although we could predict final plant height for the different height classes very well, simulating plant height during the growing season was not as accurate. However, this problem was related to the phenology algorithm, which simulated flowering too early because of the unusual weather conditions of the season (e.g., above average precipitation and below normal temperatures and solar radiation) and not our plant height algorithm. While our main goal of predicting final plant height was met and we were able to capture the genotype by environment interaction, the newly developed plant height algorithm was partially constrained by the phenology algorithm for accurately predicting within-season plant height due to some unique input requirements of the crop simulation model used in this effort. It is likely that we would run into problems such as this with any model, just that the constraints would change with the model as well as problems associated with genotype by environment ineractions.

Technical Abstract: A challenge for crop simulation modeling is to incorporate existing and rapidly emerging genomic information into models to develop new and improved algorithms. The objective of this effort was to simulate plant height in winter wheat (Triticum aestivum L.) across a range of environments in Nebraska using genetic information. Plant height is influenced by major genes that are discrete and well characterized genetically and phenotypically. Although plant height is a complex trait, it is far simpler than many traits. Two data sets were used in this study, one from a PhD dissertation where final plant heights were measured at five locations across Nebraska over two years (data set 1) and another study where plant height was measured 13 times before anthesis in one year at one location for one genotype (data set 2). The crop simulation model CropSyst was modified by incorporating a different phenology algorithm and adding a plant height algorithm. The plant height algorithm was based on the ratio of the mean optimum final height of a height class (tall semidwarf, semidwarf, short semidwarf) to the mean optimum final heights for two tall cultivars, an environmental index based on the fraction of transpirable water, and the development stage. Final plant heights were simulated reasonably well for the different height classes with a root mean square error of 3.4 cm (data set 1). For data set 2, although the final plant height was well predicted and there was a very good correlation between observed and simulated intermediate heights (r2 = 0.99), the absolute height values did not agree well. This problem was related to the phenology algorithm, which simulated anthesis too early; specifically the use of the optimum development rate in this algorithm. This algorithm also did not respond well to the unusual weather conditions of this season, where above average precipitation occurred accompanied by below normal temperatures and solar radiation. While the initial goal was generally successful, the plant height algorithm was partially constrained by the phenology algorithm and some unique input requirements of the crop simulation model used in this effort. This conclusion would be true for any model, just that the constraints would change with the model. Development of robust crop simulation models incorporating genetic information can be used to address the inverse problem, how does one go from a model to a gene, as well as problems associated with genotype by environment interactions.