SPAC系统中农作物水循环知识融合模型研究

Agricultural water cycle-water balance is very important to the scale of river basin agricultural water calculation and decision making. Since the 1990s, the water cycle analysis of crops has been studied by SPAC model. However, the system parameters are difficult to be determined by some uncertain factor, such as soil, crop, fertilizer, water and so on. In the recent years, the knowledge model method has been introduced for the analysis. But it is one of the difficult point how fuse data and knowledge between models.. In view of the above, with the Heihe River basin as an example, the project will focus on the following researches: .1, To study the parameters contraction, parameters calibration and scale conversion method of SPAC quantitative analytical model in each stage, and to establish SPAC generalization model in irrigation district scale;.2, To study the relationship between discrete data and crop growth mechanism in the crops water cycle system during the crops growth, and to solve the problem of representation knowledge for uncertainty factors of water cycle under the complex environment;.3, To study the methods and mechanism of information transmission, structure transformation and collaborative work between experimental data and all levels of knowledge, and to establish a knowledge reasoning model of crops water cycle by knowledge driving..The implementation of this project will solve many key issues in the process of models fusion, such as knowledge abstraction, expression and mapping, parameters contraction and conversion and so on. According to these foundations, the agro-ecology water cycle model will be constructed in the Heihe River basin. Furthermore, they will provide new technology for analysis of agricultural water.

农业水循环-水平衡对流域尺度的农业用水计算与决策极为重要。90年代以来，有关学者利用SPAC模型开展了农作物水循环研究，但受土壤、作物、肥水等不确定因素影响，系统参数难以确定。近年来开展的在农业水循环模型中融入知识模型的方法研究，模型之间数据与知识如何融合，是其中的难点。.依此，将以黑河流域为例，重点开展以下工作： 1、SPAC定量解析模型各阶段的参数简约、定标及尺度转化方法，建立灌区尺度下SPAC泛化模型；2、农作物水循环系统中离散数据与作物生长机理之间的关联关系，解决复杂环境下不确定因素的知识表示问题；3、实验数据与农作物各生长阶段知识间的信息传递机制、协同工作机制，建立农作物水循环知识融合模型。.项目实施后将解决模型融合过程中知识抽取、表达、映射，参数简约、转化等关键问题，并据此构造黑河流域农业生态水循环模型，为农业用水分析提供新的技术手段。

农业水循环-水平衡对流域尺度的农业用水计算与决策极为重要。90年代以来，有关学者利用SPAC模型开展了农作物水循环研究，但受土壤、作物、肥水等不确定因素影响，系统参数难以确定。近年来开展的在农业水循环模型中融入知识模型的方法研究。模型之间数据与知识如何融合，是其中的难点。依此，将以黑河流域为例，重点开展以下工作：1、SPAC定量解析模型各阶段的参数简约、定标及尺度转化方法，建立灌区尺度下SPAC泛化模型；2、农作物水循环系统中离散数据与作物生长机理之间的关联关系，解决复杂环境下不确定因素的知识表示问题；3、实验数据与农作物各生长阶段知识间的信息传递机制、协同工作机制，建立以农作物水循环知识融合模型。.项目实施过程中，结合甘肃当地的实际情况，提出了基于气象数据和改进的BP神经网络的农作物需水为核心的月度ETo估算模型以及基于辐射数据和随机森林算法与BP-NN算法的日度ETo估算模型，并进一步融合农业知识，提出了基于旬数、气象数据和随机森林日度ETo估算模型，分别利用黑河流域酒泉气象站和玉门镇气象站的1964年1月到2014年7月的数据进行了实验和模型验证，表明我们提出的模型在均方根误差、平均偏差、平均绝对百分误差以及决定系数等评价标准上优于之前的多种经验模型。.通过本项目的研究，解决了模型融合过程中知识抽取、表达、映射，参数简约、转化等关键问题，并据此构造黑河流域农业生态水循环模型，尤其是提出的更为精准的蒸散发模型，为农业用水分析提供了新的技术手段和方法参考。