Designing sampling schemes for effect monitoring of nutrient leaching from agricultural soils

A general methodology for designing sampling schemes for monitoring is illustrated with a case study aimed at estimating the temporal change of the spatial mean P concentration in the topsoil of an agricultural field after implementation of the remediation measure. A before‐after control‐impact (BACI) sample‐pattern is proposed, with stratified random sampling as a spatial sampling design. The strata are formed as compact blocks of equal area, so that the sample locations cover the field very well. Composite sampling, where the aliquots of a composite come from different strata, is proposed in order to save laboratory costs. The numbers of composites and aliquots per composite are optimized for testing the hypothesis that the mean P concentration didn’t change or has increased. Initially, this is done for a known variogram, temporal correlation, variance of laboratory measurement error, initial mean P concentration, and time needed for fieldwork. The optimal sample size to achieve a power of 0.90 at a 10% decrease of the mean P concentration is six composites of six aliquots each. Next, the effect of uncertainty about these model parameters on the optimal sample size and on the power of the test for a fixed sample size is analyzed. This analysis showed that, to obtain a probability of 95% that the power ≥ 0.90, the sample size must be increased to 7 composites of 10 aliquots each.     

Testing for additivity in chemical mixtures using a fixed-ratio ray design and statistical equivalence testing methods

Fixed-ratio ray designs have been used for detecting and characterizing interactions of large numbers of chemicals in combination. Single-chemical dose-response data are used to predict an “additivity curve” along an environmentally relevant ray. A “mixture curve” is estimated from the mixture dose-response data along the ray. A test of additivity is equivalent to a test of coincidence of these two curves, which is based on the traditional hypothesis testing framework that assumes additivity in the null hypothesis and rejects with evidence of interaction. However, failure to reject may be due to lack of statistical power, making the claim of additivity problematic. As a solution we have developed rigorous methodology to test for additivity using statistical equivalence testing logic in which additivity is claimed based on pre-specified biologically important additivity margins, if the data support such a claim. Using the principle of confidence interval inclusion, a confidence region about the difference of meaningful functions of model parameters from the mixture model and that predicted under additivity is computed. When the confidence region is completely contained within the additivity margins then additivity is claimed with a Type I error rate chosen a priori to be some acceptably small value. The method is illustrated using an environmentally relevant fixed-ratio mixture of nine haloacetic acids where cytotoxic response is measured.     

基于超参数优化算法下的随机森林模型预测奶牛呼吸频率

奶牛呼吸频率是评估环境造成的奶牛热应激程度的重要指标之一。该研究基于随机森林（random forest,RF）算法提出了适用于生产条件下的奶牛个体呼吸频率准确预测模型,为了平衡模型精度与计算效率问题,利用遗传算法（genetic algorithm,GA）、差分进化（differential evolution,DE）算法、粒子群优化（particle swarm optimization,PSO）算法、贝叶斯优化（Bayesian optimization,BO）算法对模型超参数进行优化,并与网格搜索（grid search,GS）下的人工神经网络（artificial neural network,ANN）和极限梯度提升机（extreme gradient boosting,XGBoost）模型进行了对比分析。研究结果表明,使用融合环境参数的修正温湿指数（adjusted temperature-humidity index,ATHI）、时间区域、奶牛产奶量、泌乳天数、身体姿势以及胎次作为输入特征时,基准RF模型的预测性能最佳。在此基础上,4种智能优化算法下的RF模型性能优于GS-ANN和GS-XGBoost,其中BO-RF的综合性能最优,其决定系数、平均绝对误差、平均绝对百分比误差以及均方根误差分别为0.614、7.723、14.4%、9.737,超参数优化耗时约为DE-RF的1/220。特征重要性分析表明,输入因子对奶牛呼吸频率的影响程度不同,ATHI是影响力最高的因子,相对重要性（relative importance,RI）为0.71,其次是时间区域（RI=0.09）和奶牛产奶量（RI=0.07）。研究为奶牛生产、健康评价及牛舍环境精准调控提供了有效方法和基础。     

A Score Test for Testing a Marginalized Zero-Inflated Poisson Regression Model Against a Marginalized Zero-Inflated Negative Binomial Regression Model

Marginalized zero-inflated count regression models (Long et al. in Stat Med 33(29):5151–5165, 2014) provide direct inference on overall exposure effects. Unlike standard zero-inflated models, marginalized models specify a regression model component for the marginal mean in addition to a component for the probability of an excess zero. This study proposes a score test for testing a marginalized zero-inflated Poisson model against a marginalized zero-inflated negative binomial model for model selection based on an assessment of over-dispersion. The sampling distribution and empirical power of the proposed score test are investigated via a Monte Carlo simulation study, and the procedure is illustrated with data from a horticultural experiment. Supplementary materials accompanying this paper appear on-line.     

Nonparametric Bootstrap Confidence Intervals for Variance Components Applied to Interlaboratory Comparisons

Exact confidence intervals for variance components in linear mixed models rely heavily on normal distribution assumptions. If the random effects in the model are not normally distributed, then the true coverage probabilities of these conventional intervals may be erratic. In this paper we examine the performance of nonparametric bootstrap confidence intervals based on restricted maximum likelihood (REML) estimators. Asymptotic theory suggests that these intervals will achieve the nominal coverage value as the sample size increases. Incorporating a small-sample adjustment term in the bootstrap confidence interval construction process improves the performance of these intervals for small to intermediate sample sizes. Simulation studies suggest that the bootstrap standard method (with a transformation) and the bootstrap bias-corrected and accelerated (BC _a ) method produce confidence intervals that have good coverage probabilities under a variety of distribution assumptions. For an interlaboratory comparison of mercury concentration in oyster tissue, a balanced one-way random effects model is used to quantify the proportion of the variation in mercury concentration that can be attributed to the laboratories. In this application the exact confidence interval using normal distribution theory produces misleading results and inferences based on nonparametric bootstrap procedures are more appropriate.     

Simultaneous nitrification and diffusion in soil. II. The effects at levels of ammonium chloride which inhibit nitrification

The physicochemical and microbiological changes occurring in a fine sandy loam soil following the application of ammonium chloride were followed experimentally and with a simulation model. Two levels of ammonium addition were used corresponding to application rates of 37 and 143 kg ha^?1. The measured concentration profiles of ammonium nitrate and pH, which developed in soil columns as a result of the diffusion and simultaneous nitrification of the added NH₄⁺, were measured at different incubation times. The measured profiles suggested that nitrification was inhibited at the site of application of the ammonium salt. This inhibition was attributed to an effect of increased osmotic pressure or chloride ion in the soil. A simulation model was developed to account for the inhibition by examining and testing two hypotheses about the response of nitrifiers to a fluctuating osmotic pressure. These were the irreversible inhibition model, which assumed that exposure to high osmotic pressures irreversibly inactivated a portion of the nitrifier population, and the reversible inhibition model, which assumed that the nitrifiers would recover after exposure to high osmotic pressures. The model included terms for the adsorption equilibria of NH⁴⁺ and soil acidity with the soil solid phase, and the influence of other ions on the rate of diffusion of each diffusing ion. The inputs to the model were based on parameters obtained independently of the diffusion experiments. Good agreement was found between experimental and predicted concentration profiles for both models although the reversible inhibition model gave the better simulation of the data.     

基于机器学习的沟谷地貌识别模型对比——以黄土高原典型流域为例

探索沟谷地貌空间分布与环境控制特征之间的联系并构建沟谷地貌准确提取模型,对大尺度范围沟谷提取具有重要意义。基于人工提取黄土高原典型流域沟谷地貌样本,结合不同时期的Landsat8 OLI影像数据和DEM数据,建立随机森林模型确定黄土高原沟谷地貌提取最佳影像时期和最佳组合特征,基于最优模型参数,对比其与支持向量机和人工神经网络沟谷提取模型效果,验证模型泛化能力。结果表明:(1)黄土高原沟谷提取的最佳影像时期为12月,最佳组合特征集为Red、Blue、H、SWIR1、PNT、Coastal、GLCM4和NIR;(2)3种方法提取测试区域的沟谷空间分布一致,从定性和定量角度进行比较,随机森林模型提取效果最好,验证样区平均总体精度为80.48％,相较于支持向量机模型和人工神经网络模型分别提高4.00和8.63个百分比;(3)测试区域中,沟谷地貌面积约占总面积的56.91％,且呈现西北至东南方向逐渐集中的特点。研究表明随机森林模型在黄土高原地区高精度沟谷地貌识别研究中综合表现最佳,可大范围推广。     

Bootstrap Confidence intervals for the Shannon biodiversity index: A simulation study

Biodiversity determination at any organization level is a heuristic tool in environmental management. The Shannon biodiversity index stands out among the various available indexes that combine both components of the concept: species number and their relative abundance. Whatever the index used, an aspect that has received little attention in the study of biodiversity is the uncertainty quantification, precluding hypothesis testing. Applying standard bootstrap methods to estimate the index allows us to build confidence bounds, but it yields underestimated indexes due to unsampled species. This article explores the empirical mean coverage of confidence intervals for the Shannon biodiversity index when the total number of species in the sample itself is subject to random variation. I analyze their dependence on evenness, total abundance, and number of species within the community. Very low empirical coverages of bootstrap confidence intervals were obtained using the standard and bias-corrected techniques. I propose a new technique using the difference of the original sample and each bootstrap replication to build an empirical, adjustment for the bias, and get an adjusted point estimator and corresponding confidence interval. An example based on field data, is considered.     

水稻秧盘自动码垛机的设计与试验

针对水稻工厂化育秧播种完成的秧盘人工取盘和层叠堆放劳动强度大、效率低、后续运输不便等问题,该研究设计了一种水稻秧盘自动码垛机,可完成秧盘精确夹取、转移和码垛;阐述了码垛机的基本结构和工作原理,对关键部件进行设计,以可编程逻辑控制器(Programmable Logic Controller,PLC)作为控制核心,搭建了...     

集成随机配置网络在养殖水质监测中的应用

为解决集约化水产养殖过程水体氨氮浓度无法实时检测的问题,提出基于Bagging集成随机配置网络(stochastic configuration network,SCN)的建模方法,利用养殖过程采集的相关水质参数对养殖水体氨氮浓度进行软测量。该方法首先采用Bootstrap方式生成多个不同的训练子集,然后并行训练多个SCN模型,最后将各个SCN模型的输出结果取均值作为Bagging-SCN模型的输出。为验证方法的有效性,分别通过UCI标准数据库中的机翼自噪声数据集和集约化海水养殖过程数据集进行了仿真试验,将该研究提出的Bagging-SCN模型与单一SCN模型、以及目前应用最广泛的随机权向量函数连接网络(random vector functional-link net,RVFL)模型、Bagging-RVFL模型的测量效果进行了比较。试验结果表明:该文所提模型对机翼自噪声数据集中缩放声压级测量的均方根误差、平均绝对百分比误差和最大绝对误差分别为4.225 dB、2.599%和17.500 dB;在对集约化海水养殖过程中水体氨氮浓度测量的均方根误差、平均绝对百分比误差和最大绝对误差分别为0.0628 mg/L、27.851 mg/L和0.189 mg/L均优于其他测量模型;进一步说明该模型具有较高的测量精度和稳定性,更适合应用于集约化水产养殖水质监测过程。     

基于改进全卷积网络的棉田冠层图像分割方法

针对传统的全卷积网络分割精度低、效果差等问题,该文提出一种结合条件随机场的改进全卷积网络棉田冠层图像分割方法。首先通过提取和学习图像特征对全卷积网络进行训练以优化其分割性能,得到初步分割结果和训练后的全卷积网络模型;接着将初步分割结果以像素和像素对应的分类向量形式输入到条件随机场中,同时结合像素间相对关系构建能量函数再进行训练,对初步分割结果进行优化得到训练后的条件随机场模型;进而通过验证过程对全卷积网络和条件随机场模型参数进一步调优,得到最优的全卷积网络和条件随机场;最后结合最优的全卷积网络和条件随机场实现棉田冠层图像分割并进行试验。试验结果表明:该方法的平均像素准确率为83.24%,平均交并比为71.02%,平均速度达到0.33 s/幅,与传统的全卷积网络分割性能相比分别提升了16.22和12.1个百分点,改进效果明显;与Zoom-out和CRFas RNN(conditional random fields as recurrent neural networks)分割方法进行对比,平均像素准确率分别提升了4.56和1.69个百分点,平均交并比分别提升了7.23和0.83个百分点;与逻辑回归方法和SVM(support vector machine)方法进行对比,平均像素准确率分别提升了3.29和4.01个百分点,平均交并比分别提升了2.69和3.55个百分点。该文方法在背景复杂、光照条件复杂等环境下可以准确分割出冠层目标区域,鲁棒性较好,可为棉花生长状态自动化监测提供参考。     

On spatial prediction of soil properties in the presence of a spatial trend: the empirical best linear unbiased predictor (E-BLUP) with REML

Geostatistical estimates of a soil property by kriging are equivalent to the best linear unbiased predictions (BLUPs). Universal kriging is BLUP with a fixed‐effect model that is some linear function of spatial co‐ordinates, or more generally a linear function of some other secondary predictor variable when it is called kriging with external drift. A problem in universal kriging is to find a spatial variance model for the random variation, since empirical variograms estimated from the data by method‐of‐moments will be affected by both the random variation and that variation represented by the fixed effects. The geostatistical model of spatial variation is a special case of the linear mixed model where our data are modelled as the additive combination of fixed effects (e.g. the unknown mean, coefficients of a trend model), random effects (the spatially dependent random variation in the geostatistical context) and independent random error (nugget variation in geostatistics). Statisticians use residual maximum likelihood (REML) to estimate variance parameters, i.e. to obtain the variogram in a geostatistical context. REML estimates are consistent (they converge in probability to the parameters that are estimated) with less bias than both maximum likelihood estimates and method‐of‐moment estimates obtained from residuals of a fitted trend. If the estimate of the random effects variance model is inserted into the BLUP we have the empirical BLUP or E‐BLUP. Despite representing the state of the art for prediction from a linear mixed model in statistics, the REML–E‐BLUP has not been widely used in soil science, and in most studies reported in the soils literature the variogram is estimated with methods that are seriously biased if the fixed‐effect structure is more complex than just an unknown constant mean (ordinary kriging). In this paper we describe the REML–E‐BLUP and illustrate the method with some data on soil water content that exhibit a pronounced spatial trend.     

Spatio-temporal modeling of fine particulate matter

Studies indicate that even short-term exposure to high concentrations of fine atmospheric particulate matter (PM_2.5) can lead to long-term health effects. In this article, we propose a random effects model for PM_2.5 concentrations. In particular, we anticipate urban/rural differences with regard to both mean levels and variability. Hence we introduce two random effects components, one for rural or background levels and the other as a supplement for urban areas. These are specified in the form of spatio-temporal processes. Weighting these processes through a population density surface results in nonstationarity in space. We analyze daily PM_2.5 concentrations in three midwestern U.S. states for the year 2001. A fully Bayesian model is implemented, using MCMC techniques, which enables full inference with regard to process unknowns as well as predictions in time and space.     

Detecting interaction(s) and assessing the impact of component subsets in a chemical mixture using fixed-ratio mixture ray designs

An important environmental and regulatory issue is the protection of human health from potential adverse effects of cumulative exposure to multiple chemicals. Earlier literature suggested restricting inference to specific fixed-ratio rays of interest. Based on appropriate definitions of additivity, single chemical data are used to predict the relationship among the chemicals under the zero-interaction case. Parametric comparisons between the additivity model and the model fit along the fixed-ratio ray(s) are used to detect departure from additivity. Collection of data along reduced fixed-ratio rays, where subsets of chemicals of interest are removed from the mixture and the remaining compounds are at the same relative ratios as considered in the full ray, allow researchers to make inference about the effect of the removed chemicals. Methods for fitting simultaneous confidence bands about the difference between the best fitting model and the model predicted under additivity are developed to identify regions along the rays where significant interactions occur. This general approach is termed the “single chemicals required” (SCR) method of analysis. A second approach, termed “single chemicals not required” (SCNR) method of analysis, is based on underlying assumptions about the parameterization of the response surface. Under general assumptions, polynomial terms for models fit along fixed-ratio rays are associated with interaction terms. Consideration is given to the case where only data along the mixture rays are available. Tests of hypotheses, which consider interactions due to subsets of chemicals, are also developed.     

基于机器学习算法的蒸发量模型评估

[目的]探讨合理的气候因子个数,建立蒸发量模型,提出基于特征选择算法筛选最优特征集。[方法]以陕西榆林、泾河和汉中3个气象站16年(2005-01至2021-03)的逐小时观测资料为研究对象,利用特征选择函数和遍历循环方法对模型参数、特征变量个数进行优化。基于最佳参数结合随机森林模型和多元线性回归模型两种机器学习算法建立榆林、汉中和泾河地区蒸发量模型,采用平均绝对误差、均方根误差和平方相关系数三项指标评估模型的预测精度。[结果]特征变量和随机森林模型中的决策树个数分别是8,61时,模型预测效果最佳。采用优化的随机森林模型、多元线性回归模型评估3个地区的平均绝对误差均为0,均方根误差除泾河地区相等外,榆林、汉中地区的均方根误差均小于优化的多元线性回归模型。优化的随机森林模型预测榆林、泾河和汉中地区蒸发量拟合效果分别为0.85,0.90,0.86,优化的多元线性回归模型的拟合效果分别为0.77,0.83,0.79。[结论]整体而言,优化的两种模型都具有良好的预测效果且随机森林模型的预测效果优于多元线性回归模型。     

Testing statistical hypotheses using standard error bars and confidence intervals

Abstract

When agricultural researchers construct figures or graphs displaying sample means from an experiment, a popular technique to display the relative variation is to include standard error bars. This technique can be very informative but misleading. Researchers will sometimes try to draw inference to the equality of their means by using these standard error bars. This paper explores the use of standard error bars in comparing population parameters and exhibits how conclusions drawn by this method will often be faulty. The use of confidence intervals to test hypotheses is also presented. Some simple mathematical derivations are presented, along with a small computer simulation study. If a researcher utilizes standard error bars in an attempt to test a hypothesis, he or she will be performing a test with an approximate type I error rate of α=0.16. In situations in which it is difficult to perform a test, but confidence intervals are available, an alternative for performing a=0.05 test is to evaluate 85% confidence intervals and reject the hypothesis that the parameters are equal if the intervals fail to overlap.     

改进SSD的灵武长枣图像轻量化目标检测方法

针对加载预训练模型的传统SSD（Single Shot MultiBox Detector）模型不能更改网络结构,设备内存资源有限时便无法使用,该研究提出一种不使用预训练模型也能达到较高检测精度的灵武长枣图像轻量化目标检测方法。首先,建立灵武长枣目标检测数据集。其次,以提出的改进DenseNet网络为主干网络,并将Inception模块替换SSD模型中的前3个额外层,同时结合多级融合结构,得到改进SSD模型。然后,通过对比试验证明改进DenseNet网络和改进SSD模型的有效性。在灵武长枣数据集上的试验结果表明,不加载预训练模型的情况下,改进SSD模型的平均准确率（mAP,mean Average Precision）为96.60%,检测速度为28.05帧/s,参数量为1.99×106,比SSD模型和SSD模型（预训练）的mAP分别高出2.02个百分点和0.05个百分点,网络结构参数量比SSD模型少11.14×106,满足轻量化网络的要求。即使在不加载预训练模型的情况下,改进SSD模型也能够很好地完成灵武长枣图像的目标检测任务,研究结果也可为其他无法加载预训练模型的目标检测任务提供新方法和新思路。     

基于径向基神经网络与粒子群算法的双叶片泵多目标优化

针对双叶片泵存在水力性能比相同比转速的多叶片离心泵低的缺陷,该文以一台型号为80QW50-15-4的双叶片污水泵作为研究对象,将其设计流量点的扬程和效率定为优化目标,运用ANSYS CFX(computational fluid dynamics x)进行数值模拟获得性能数据,采用径向基(radial basis function,RBF)神经网络建立结构参数与扬程、效率性能间的预测模型,并将其用作粒子群算法的适应值评价模型,在样本空间内进行最优值求解,获得扬程和效率的Pareto解。选取扬程最优个体和效率最优个体进行数值模拟,研究其在输运不同介质时的性能与内流场差异,并与初始模型的数值模拟数据相比较。经试验验证,清水介质中设计流量点扬程最优个体的扬程较初始个体增加0.96 m,增幅达到5.5%;效率最优个体的效率较初始个体提升了10.11个百分点。该优化方法改善了叶轮水力特性,使双叶片泵性能得到提高。     

基于时序模拟的配电系统可靠性评估不确定性研究

基于时序模拟的配电系统可靠性评估中,采用状态持续时间抽样,各元件的状态时间段是随机产生的,属于不确定性问题。求解不确定性问题的可靠性指标的收敛判据不同于常规数值迭代算法,需要运用概率知识描述。针对配电系统的随机时序模拟问题,研究了独立事件的选取,提出了以随机初始种子,相同循环次数作为广义独立事件;分析了在给定置信概率下总体均值的置信区间,其置信区间长度与样本均值的绝对误差有关,从而得出置信区间长度与给定计算精度的不等关系式,提出了以绝对误差作为计算精度的实用收敛判据。通过对RTBS测试系统的分析验证了该方法的合理性与有效性。     

土壤空间变异研究中的半方差问题

简要回顾了土壤空间变异的研究。根据地质统计学理论和多年从事土壤空间变异研究的经验,对土壤空间变异研究的关键问题——半方差函数的基本假设、取样、模型选取及模型的检验进行了讨论,并对确定半方差函数模型应注意的问题提出建议。在保证取样样本容量的前提下,检查测定数据是否服从内蕴假设;注意提高每一个估算值的置信水平;尽量选择安全型模型作为半方差函数模型;对确定的半方差模型进行统计检验。由此可以求得较为客观合理的半方差模型。