Changing paradigm in soil and plant sciences: design of balances between components

组成土壤和植物的数据接近于某个整体，并具有作为自组织系统的连接部分的组件之间的层次关系。传统的土壤和植物模型是基于19世纪初和20世纪上半叶在其他基本假设下发展起来的概念。这种方法严重威胁了当前植物营养、土壤保持和土壤肥力理论的连贯性和进步。它否认了成分数据相互作用的事实。数字偏差在统计分析中产生噪声的原因是数据冗余(根据差异计算，封闭系统的一个组件是冗余的)、尺度相关性(尺度的变化产生次成分不一致)和非正态分布(成分空间被限制在0和整体之间，而实际空间不受约束)。克服偏差最有希望的技术是2003年提出的等距对数比(ILR)变换，它去除了引起相互作用的噪声(1dof)。我们使用数字平衡的概念，涉及到人类看待自然现象的一种直观方式。比率是两个组件之间的平衡，而ILR是组件子集的几何平均值之间的平衡。ILR具有欧几里得几何。根据合理的理论，形成系统子集的组分，如阴离子和阳离子物种、微聚集体和大聚集体或生化C或P组分，被安排成正交天平。如果目标是得出一个多元距离，那么平衡的设计方式并不重要。否则，它们必须对研究人员和最终用户有意义。利用ILR变量和受试者工作特征(ROC)曲线来诊断系统的“健康”，我们评估了诊断组织和土壤测试在作物营养(‘植物健康’)、土壤有机残留物分解模型(C封存)、土壤磷转化模型和土壤团聚体指数(‘土壤健康’)方面的性能。在植物营养方面，平衡是根据植物生理、土壤生物地球化学和作物管理中的养分相互作用而设计的。在土壤肥力方面，天平有助于确定肥料和石灰的需求。微团聚体，即固碳，与大团聚体相平衡，以定义土壤健康。磷的形态被平衡为地球化学库和生物库，每个库又细分为磷有效性递减的库。这些平衡模型将使用在加拿大、巴西和新西兰收集的数据集以及从文献中收集的数据进行测试。土壤和植物的测试性能将使用ROC进行测量，并与临床生物学中的测试进行比较，以衡量人类健康。我们设想将平衡概念应用于土壤分析的光谱技术、水培中的养分平衡以及土壤-植物系统中养分转移的模拟。全世界都需要维持作物产量，同时将环境和经济风险降至最低。在解释荟萃分析计量经济学(作物反应)和土壤-植物模型的结果时，成分数据的无偏模型的公式将比当前的模型减少噪声。该研究计划将在这些问题上保持和扩大国际合作(例如，拉瓦尔与西班牙大学、巴西大学和悉尼大学的合作)。迫切需要形成HQP，以使平衡概念适应现实世界。HQP将校准和验证植物营养和土壤质量的大型数据集，而农学、生物地球化学(包括生物指示植物)、营养和生态学方面的论文和理论将基于对自然界平衡的合理知识进行重新审查。这项计划引领了农业科学向平衡的范式转变。

Compositional soil and plant data are closed to some whole and have a hierarchy of relationships between components as connected parts of a self-organizing system. Traditional soil and plant models are based on concepts developed at the beginning of the 19th and the first half of the 20th century under the ceteris paribus assumption. This approach is a serious threat to the coherence and progress in current theories on plant nutrition, soil conservation and soil fertility. It denies the fact that compositional data interact. Numerical biases produce noise in statistical analysis due to data redundancy (one component of a closed system is redundant as computed by difference), scale dependency (change of scale produces sub-compositional incoherence) and non-normal distribution (compositional space is constrained between 0 and the whole while the real space is unconstrained). The most promising technique to overcome biases is the isometric log ratio (ilr) transformation elaborated in 2003 that removes noise (one dof) causing interaction. We use a numerical concept of balance, relates to an intuitive way for human beings to look at natural phenomena. A ratio is a balance between two components while an ilr is a balance between geometric means of subsets of components. The ilrs have Euclidean geometry. Groups of components that form system subsets such as anionic and cationic species, micro- and macro-aggregates or biochemical C or P fractions are arranged into orthogonal balances based on sound theory. If the objective is to derive a multivariate distance, the way balances are designed is not important. Otherwise, they must make sense to the researcher and the end-user. Using ilr variables and the receiver operating characteristic (ROC) curve to diagnose system ‘health’, we evaluate the performance of diagnostic tissue and soil tests for crop nutrition (‘plant health’), decomposition models of organic residues in soil (C sequestration), soil P transformation models and soil aggregate indices (‘soil health’). In plant nutrition, balances are designed according to nutrient interactions in plant physiology, soil biogeochemistry and crop management. In soil fertility, balances assist determining fertilizer and lime requirements. Micro-aggregate, that sequester C, are balanced with macro-aggregates to define soil health. The P forms are balanced into geochemical and biological pools, each subdivided into pools of decreasing P availability. Those balance models will be tested using data sets collected in Canada, Brazil and New-Zealand and from literature. Test performance for soils and plants will be measured using ROC and compared to tests in clinical biology to measure human health. We envisage applications of the balance concept to spectroscopic techniques for soil analysis, nutrient balance in hydroponics and the modeling of nutrient transfer in soil-plant systems. There is a need worldwide to sustain crop productivity while minimizing the environmental and economic risks. The formulation of unbiased models for compositional data will reduce noise compared to current models when interpreting results of meta-analysis econometric (crop response) and soil-plant models. This research program will maintain and expand international collaboration on these issues (e.g. Laval with UPC-Spain, UNESP-Brazil and University of Sydney). There is urgent need to form HQP for adapting balance concepts to the real world. The HQP will calibrate and validate large datasets in plant nutrition and soil quality while papers and theories in agronomy, biogeochemistry (including bio-indicator plants), nutrition and ecology will be revisited based on sound knowledge of balances in nature. This program leads a paradigm shift to balances in agricultural sciences.