【作者】 王航；

【导师】 曹卫星；

【作者基本信息】 南京农业大学 ， 作物栽培学与耕作学， 2011， 硕士

【摘要】 作物生长模型和遥感信息是作物生长监测预测的2个有力工具,将水稻生长模型和遥感有机结合,有助于模型的过程性、机理性与遥感的空间性、实时性优势互补,既提高水稻生长模型的区域应用能力,又增强遥感监测预测的机理性。本研究提出了一种基于地空遥感信息与生长模型耦合的水稻产量预测方法,该方法基于同化策略,以不同生育时期的水稻叶面积指数(LAI)和叶片氮积累量(LNA)为信息融合点,将地面光谱数据(ASD)及HJ-1 A/B数据与水稻生长模型(RiceGrow)相耦合,反演得到区域尺度生长模型运行时难以准确获取的部分管理措施参数(播种期/移栽期、播种量和施氮量),在此基础上实现了对水稻产量的有效预测。结果表明,将LNA和LAI共同作为遥感和模型的耦合点反演效果较好,粒子群同化算法(PSO)相比于复合型混合演化算法(SCE_A)更适合耦合同化过程,对模型输入参数如播种期/移栽期、播种量和施氮量反演的RMSE值分别为1.25 d、4.52 kg·hm-2和3.69 kg·hm-2.在此基础上实现了对水稻生长状况和产量的有效预测。将过程更新策略(updating strategy)和同化策略(assimilation strategy)相结合,研究建立了基于更新与同化结合策略的遥感模型耦合技术。在遥感信息与水稻生长模型(RiceGrow)的耦合过程中,先利用集合平方根滤波(EnSRF)算法,综合考虑模型模拟和遥感反演生理指标的误差,得到叶面积指数(LAI)和叶片氮积累量(LNA)的更新值,再以这一更新的生理指标序列作为遥感与模型的耦合点,利用初始化策略获取关键初始参数(播种期/移栽期、播种量和施氮量)。结果表明,结合更新和同化策略后模拟的LAI、LNA相比与模型模拟值和遥感监测值,更加接近实测值。且将获取到的参数输入模型运行得到的产量结果也优于单纯同化策略,田块和区域尺度产量预测结果与实测值和统计值之间有很好的一致性。因此,基于更新和同化策略相结合的遥感信息与水稻生长模型耦合技术提高了模型的预测精度,可为水稻生长监测和生产力预测提供重要技术依据。根据面向对象程序设计原理,以Microsoft.NET Framework 3.5为开发环境,C#为编程语言实现系统整体架构和界面定制,调用IDL组件实现遥感信息处理功能,调用GIS组件ESRI ArcGIS Engine 9.3实现GIS空间分析和专题制图功能,另外将利用C#语言编写的水稻生长模型组件和上述耦合技术的代码进行集成,在此基础构建了基于遥感信息与生长模型耦合的水稻生长预测系统(RGMPS),实现了影像预处理与光谱信息提取、遥感监测、模拟预测、专题制图等功能。更多还原

【Abstract】 Rice (Oryza sativa L.) is the most important food crop in China, whose total yield takes the first location in the world. Under the complex climate and economic condition, estimating the rice’s growth status and production information in time and accurately is important for China’s food security and agricultural sustainable development. Crop model and remote sensing are both useful tools in predicting crop growh and productivities. Integration of remote sensing (RS) and crop growth model is an important approach to improving the ability of monitoring and predicting crop growth. Rice growth prediction techniques were developed by integrating the ground-based and space-borne remote sensed data and RiceGrow model based on the assimilation strategy and updating strategy in this paper. Leaf area index (LAI) and leaf nitrogen accumulation (LNA) of rice were estimated using ASD field spectrometer and HJ-1 A/B CCD, based on statistical remote sensing estimation models. This information was integrated with the RiceGrow model for got three management parameters included sowing date, sowing rate, and nitrogen rate. Then the rice yield can be predicted by inputting these parameters. This integrated technique was tested based on independent datasets. The results showed that both LNA and LNA, was better than either as an integrated parameter for crop model parameter initialization. And the Particle Swarm Optimization (PSO) algorithm was more suitable for using in the initialization process than the Shuffled Complex Evolution-University of Arizona (SCE-UA) optimization algorithm.Coupling remotely sensed information and crop growth model can improve the prediction accuracy of crop model in regional scale. A new technique was developed for estimating rice growth parameters and grain yield in both field and regional scales, based on coupling remotely sensed information and rice growth model (RiceGrow) by combing the updating and assimilation strategies. This technique assimilated two growth parameters into RiceGrow model including remotely sensed leaf area index (LAI) and leaf nitrogen accumulation (LNA). The results showed that the predicted values of LAI, LNA and grain yield for RiceGrow model after using updating and assimilation strategy were more close to measured values than ones only using RiceGrow model or assimilation strategy. In addition, the new developed technique also well described the temporal and spatial distributions of rice growth status and grain yields in the study area, with less than 20% of the RE values for both growth parameter and regional total grain yield. Which indicated that the new crop growth and yield prediction technique had a good prediction precision and actual application for rice crop in both field and reginal scales.Rice growth monitoring and predicting system (RGMPS) based on integrating remote sensing and RicetGrow model was developed using object oriented programming technology. Which was developed by taking Microsoft.NET Framework 3.5 as the development environment and C# as the programming language to definite the system structure and interface, and integrating crop growth model components, ESRI ArcGIS Enginec 9.3 and the remote sensing processing modules developed by IDL. Varied functions were realized by this system, such as image processing and spectral information extraction, growth monitoring based on RS, growth and yield simulating and predicting, thematic mapping, and so on.更多还原