【作者】 陈立平；

【导师】 陈阜；

【作者基本信息】 中国农业大学 ， 作物栽培与耕作学， 2003， 博士

【摘要】 本研究针对精准农业技术体系中变量施肥这一关键环节，围绕变量施肥处方生成技术及应用系统开发，探讨了土壤养分空间分布图生成技术，提出了产量数据的误差处理方法，模拟研究了变量施肥的尺度效应问题，设计并开发了具有田间导航采样、产量数据处理和决策支持功能的变量施肥处方生成系统。通过初步的实践验证和应用效果分析，本研究成果有较好的实用价值。主要研究结果包括： 1．围绕土壤养分空间分布图生成，研究了土壤采样策略制定方法，探讨了采用Cokriging方法降低采样密度和采用电导率综合评价土壤生产潜力的可能性。结果表明，在进行土壤养分采样和插值，必须同时考虑养分的空间变异性和数值变异性，而N／S值和变异系数是比较好的指标；产量与0-30.5.cm、0-91.5cm、30.5-91.5cm三层土壤电导率之间都存在极显著的负相关关系，二次多项式较好地拟合了产量与电导率的关系，电导率图可以参与进行土壤生产潜力区的划分。 2．研究分析产量数据的获取过程，采用层次分析的原则，将可能造成产量数据误差的因素归纳为直接因素、间接因素和产生原因。分别针对不同产量数据的误差来源，设计并实现了5种过滤器：产量域值过滤器，速度过滤器、填充和清空延时过滤器、邻域均值过滤器、邻域系数过滤器；在分析比较不同过滤器过滤效果的基础上，提出了1种综合过滤器，过滤后使产量数据的正态分布性得到显著改善，对已测产小区的估计误差明显下降。 3．针对精准变量施肥的尺度问题，采用计算机模拟技术，研究不同土壤硝态氮均值、方差和相关距下，施肥单元面积对玉米的平均施肥量、产量和肥料增产效率的影响。结果表明：a．随着变量施肥单元面积的减小，平均单位面积的施肥量增加或保持不变，产量一般增加，肥料增产效率也增加；b．随着土壤养分均值增加，平均施肥量降低，平均产量增加，肥料增产效率降低，且施肥单元面积越大，不同土壤养分均值间的施肥量、产量和肥料增产效率的差异越大；c．在养分均值低于施肥临界土壤养分浓度时，随着土壤养分变异系数的增加，平均施肥量增加，平均产量降低，肥料增产效率降低；d．随着土壤养分相关距的增加，平均施肥量增加，平均产量增加，肥料增产效率增加。这些结果对我们进行变量施肥的实践有一定的指导意义。 4．在分析变量施肥决策流程的基础上，采用OLE/ActiveX技术，设计并实现了集成应用GPS、GIS和ES技术的变量施肥处方生成系统。该系统在推理机中嵌入模型语法分析构件，并采用对象连接和嵌入的自动化方式实现与GIS功能构件之间的互操作功能；将知识／模型／数据采用一体化方法存储在数据库中，通过ODBC规范实现基于多数据源的知识调度与推理；应用基于矢量网格的空间叠加分析方法解决了栅格和常规矢量数据结构在处方生成中的局限。系统可以提供田间导航采样、产量数据处理和决策支持功能，最终生成可供变量施肥机具使用的处方图，可以用于指导变量施肥实践。 5．进行了基于土壤肥力测定和基于地物光谱数据的小麦变量施肥试验研究，结果表明，在施肥总量相同的情况下，与常规均一施肥对照区相比，变量施肥使产量略降，穗粒数、千粒重、叶绿素和蛋白质变异系数增大，但降低了土壤硝态氮浓度，减小了污染地下水的可能性，生态效果明显。更多还原

【Abstract】 Variable-rate fertilization is one of the key technologies in precision farming. In this dissertation, different topics related to variable-rate fertilization prescription were studied. Soil sampling strategies and yield data processing technologies were analyzed and studied, the influence of management zone size on variable-rate fertilization was studied theoretically, and DSS for variable-rate fertilization precision was also designed and implemented. According to the experiment, these results can provide support for variable-rate fertilization implementation. The main results are as follows:1. Soil nutrients in the same locations are statistically related to each other and spatially related. Both spatial variability and traditional statistics need to be considered in sampling and interpolation of soil nutrients. N/S value and CV are good indicators of spatial variability and traditional statistics. Soil EC is a compositive indicator of soil physical-chemical properties and can show the productivity of farmland to certain extent. Veris 3100 EC Meter is a type of quick and cheap equipment for acquirement of soil system information.2. The acquisition process of yield data was studied, and the principle of AHP (Analytic Hierarchy Process) was applied to systematically classify factors leading to yield data errors into direct and indirect factors. Five types of filters were designed and realized to deal with different error sources in yield data, including yield limits filter, harvester speed filter, filling and cleaning delay filter, neighbor mean filter, and neighbor coefficient filter. Based on analysis of different filters, a synthesized filter was provided to filter outliers in yield data for acquirement of reliable yield map.3. The influence of management zone size on variable-rate fertilization under different soil fertilities was studied. Field distribution dataset series of soil nitrate concentration in 500X500m2 field were generated based on given values of mean, standard deviation and related distance. Each dataset is composed of nitrate concentration values in 104 cells, with the unit cell size of 5x5 m2. The fertilization rate was calculated using the fertilization models established by D.W. Franzen and L.J. Cihacek in North Dakota based on a common potential yield and the mean value hi each management cell size. The yield was calculated based on combination of fertilization amount and soil nitrate storage in each location with the same yield potential. Based on the fertilization rate and yield, the NUE (nitrogen use efficiency, yield increment with unit weight of fertilizer) was calculated. The result shows increasing or constant fertilization rate, increasing yield and NUE when the size of management zone for variable rate fertilization decreased. When the soil fertility increased, the fertilization rate decreased and the yield and NUE increased. As the management zone for variable rate fertilization becomes larger, the difference of the fertilization rate, yield and NUE between different soil fertility levels becomes more significant With the increase of standard deviation, the increment of the fertilization rate from large management zone size to small management size scale 1 to 100 become quicker, the difference of yield between different soil fertility levels at the same management zone size become larger, and the difference of the NUE between different soil fertility levels at the same management zone size become smaller.4. Based on analysis of variable rate fertilization decision-making process, OLE/ActiveX technologies were applied to integrated GPS, GIS, ES and MS into decision support system, which has the functions of navigation for sampling, yield map processing, decision support function, and variable-rate fertilization prescription. Analytic component of model expression was embedded into the inference engine. The OLE and automation made it possible for inter-operation between the system and GIS functional components. Knowledge, models,更多还原