Dynamic prediction on leaf length of maize based on metabolic model GM(1,1)

Abstract: The research background of this paper focused mainly on two aspect elements. The first one was fast development on study of plant virtualization under the support policy of agricultural information in China and abroad. The other one was further necessary combination of subjects among management science, computer science, and agronomy. The Grey modeling method from management science, tools from computer science, and experimental design methods from agronomy were all used in this paper. Through this research, correct metabolic GM (1,1) from management science, which is helpful to dynamic simulation growth of maize leaf in research of virtual plant, was expected. The research method and content in general were as follows: To achieve the dynamic prediction of virtual maize, this paper, taking the maize three-ear-leaves as the research object, analyzed the dynamic changes of length of maize leaf under different nitrogen levels (150, 300 and 450kg/hm2) with the method of metabolic GM (1,1). GM (1,1), which has the advantages of requiring fewer message parameters, simplicity, and easy-building, is a grey-dynamic-prediction model. Through the effect of sequence operator to few-data, developing regulations of objects was researched, and the dynamic developing regulation was also simulated. Field experiments which had the same experimental design were made in years 2010 and 2011. Then, the sample data from 2010 and 2011 were used to construct and test models. Taking the data of inferior leaf in three-ear-leaves from 2010 as an example, the proper modeling steps were followed. First, the sample data should be processed with the method that the length of leaf should be divided by the effective accumulated temperature which was needed during the growth of the leaf, which produced a series of data sequences whose units were cm/℃. Second, the final data sequence should be formed by adding the newest data and rejecting the oldest data in that sequence. Finally, the model parameters, ratios of mean square deviation, and mean relative errors of models were identified through a series of calculations including one accumulating generation, mean generation with consecutive neighbors, and least square estimation method. Meanwhile, in order to verify the universality of this model, the sample data of 2011 was also used; to be precise, all the methods and measures used in the process of building models with sample data from 2010 were all used with the sample data of 2011. The experimental results showed that the ratios of mean square deviation and the mean relative errors from the models that were created by the data sequences were all less than 0.0811 and 0.0471 respectively, and that the accuracy of these models was better than level 2, which revealed a high accuracy. The results obviously verified the feasibility and effectiveness of the metabolic GM (1,1) in simulating the length of maize leaf and also provided a theoretical reference for the dynamic change simulation of maize during its growth.