Research On High-Efficient Use Of Water And Fertilizer For Rice And Its Models

This thesis has summarized a large number of documents both inland and abroad of utilization ratio of water-fertilizer, the production function of the crop and the water-fertilizer interaction overally systematically, combined with the field experiment data from Tuanlin Irrigation Experiment Station in 1999, 2000 and 2003 and Dengjiapu Farm in Yingtan, Jiangxi province in 2003, and utilized detailed discussion to high water-fertilizer use efficiency of rice. The main content is as follows:First of all, study according to the purpose of this thesis, has made the high-standard, high-accuracy mutual field testing program that influences of water and fertilizer of rice,which had been carried into execution in the field experiment. Using test materials from two places of tuanlin and yingtan in 2003, it has studied the change law of water-fertilizer biological characteristic of the rice, drawn the main composition factor of influencing the output effect utilizing analysis of variance and relevant analysis, and put forward the high-efficient optimum management mode of utilizing water and fertilizer of the rice.Secondly, because of the understanding of water-fertilizer interaction law of rice, based on Morgan model, through adopting the index of relative evapotranspiration as the effect function instead of the index of the relative moisture, introduced effect function of fertilizer and three kinds expression formulas of water-lack function, dynamic revision Morgan models have been set up and mutual precision has been compared with each other by using the material. The result indicates that the Morgan model of using the function of segment linear has higher precision.Finally, the dynamic neural network (DBP) model based on time series has been set up which was firstly used on the yield prediction of rice under different condtions of water and fertilizer, and which resolves unmatched problems between static model and time system, overcomes the localization of using the data from a certain growth phase to erect the static network. The prediction precision is improved by fully utilizing the sample data characteristic of the production process. Compared with the revision Morgan model, don't need to set up concrete water production function model, adaptability is stronger. The testing result indicates the prediction precision of DBP model is relatively high. And according to the final conclusion, there is not essential difference in yield prediction precision of rice by comparing DBP model, traditional BP model with revision Morgan model.