不同插值方法对成分数据空间预测结果的影响——以土壤连续分类模糊隶属度值为例

地球科学中成分数据(compositional data)非常普通,其在进行空间插值时必须满足4个条件:每一位置各组分之和为常数,每一组分为非负,插值结果无偏最优.本文以土壤连续分类模糊隶属度值为例,数据经对数正态变换、非对称对数比转换、对称对数比转换后进行普通克里格插值结果和成分克里格插值(compositional kriging)结果进行比较.结果表明,对原始数据和经对数正态变换后数据进行插值,每一位置预测结果隶属度之和不能满足常数1.经非对称对数比转换后,插值结果虽然满足各个位置组分之和为1,但是预测结果精度较低,且预测结果空间分布连续性不明显.数据经对称对数比转换后插值结果和成分克里格插值结果,都能满足成分数据空间插值的4个条件,但二者各有优势.相比较而言,对称对数比转换方法得到的预测结果更能体现土壤空间连续渐变特征,而成分克里格插值结果能保证隶属度本身是最优无偏估计.     

基于各向异性的区域土壤有机碳三维模拟与空间特征分析

为探索更加科学的土壤属性三维空间模拟方法,以各项同性三维普通克里格法为对比方法,采用均方根误差（root mean squared errors,RMSE）和标准化克里格方差（mean squared deviation ratio,MSDR）以及空间模拟方差图等,评价比较了各项同性和顾及各项异性的三维模拟方法的模拟效果。结果显示：三种方法模拟的土壤有机碳三维空间分布格局基本一致。随着土壤深度的不断增加,土壤有机碳含量较高的斑块逐渐减少,垂直方向上总体呈现出土体上部高下部低的格局。顾及各向异性能在一定程度上克服普通克里格法常出现的牛眼和趋中效应等缺陷问题。顾及各向异性基于Markov 的同位置协同格里格法模拟效果最佳。该法的 RMSE 值最小（1.6215）,相比于各项同性三维普通克里格法 RMSE提高将近50%,特异值覆盖比率最大（76.15%）,模拟精度最高,能够更好地突出波动性,体现特异值;该方法的 MSDR最接近1（1.4409）,且模拟的土壤有机碳质量分数总体方差均值最小（2.08）。研究成果将为区域土壤属性三维空间有效模拟提供方法参考。     

表征土壤有机碳区域分布的优化空间插值模型研究——以福建省龙海市为例

如何利用有限的样本数据来获得更为准确的土壤属性空间分布信息是土壤学研究的热点问题之一。利用福建省龙海市采集的1 133个耕地土壤样品,设计了结合地貌类型、土壤类型和土地利用类型等信息的5种克里格插值模型,研究县级尺度上土壤有机碳空间预测优化插值模型及其与样点密度的关系。结果表明:设计的5种插值模型预测精度均高于普通克里格法,但不同样点密度对插值结果影响较大。按0.5 km×0.5 km及以上的格网密度进行样点布设,采用土地利用现状类型结合土壤类型信息的普通克里格法(KDLTR)插值结果误差较小;按2 km×2 km的格网密度布设调查样点时,采用土壤类型信息的普通克里格法(KTR)插值结果误差较小;当格网大于4 km×4 km时,由于样点数少,各种模型的结果相差不大,可直接采用普通克里格法(KYJZ)进行插值。     

基于BP神经网络插值的土壤全氮空间变异

大尺度土壤养分空间变异研究可以为土壤改良分区治理提供基础数据。寻求合适的取样数和插值方法是进行土壤养分空间变异研究的关键。以安徽省舒城县为例,共取得0～20cm土壤表层样品523个,土壤全氮的空间变异由BP神经网络插值方法在不同取样数条件下获得,通过与克里格插值法进行比较得出:样本数在100个时,神经网络插值的预测吻合度(G)比克里格插值高7.75%,均方根误差(RMSE)低0.1,总体精度优于克里格;样本数大于200时,神经网络插值和克里格插值精度基本相同,随着采样数量增加,两种方法的插值精度也在提高,并逐步趋于平稳。在大尺度土壤养分空间变异研究中,在小样本情况下,神经网络插值具有优势。     

基于高光谱的土壤不同颗粒含量预测分析

以典型黄河下游冲积平原区的土壤为研究对象,分析土壤高光谱特征,探讨土壤质地不同粒级颗粒含量的统一估测途径,为土壤质地快速监测评价提供技术支持。选择原始光谱,及其倒数、对数、标准正交变换、多元散射变化、一阶微分、二阶微分共7种光谱变换形式,首先主成分降维,然后分别建立土壤黏粒、粉粒和砂粒含量的支持向量机预测模型,采用决定系数、均值绝对误差、均方根误差3种精度指标来衡量模型的预测能力。结果表明:原始光谱的对数为最佳光谱变换形式,具有最佳的土壤不同颗粒含量估测能力,决定系数R2 ≥ 0.6853,均值绝对误差MAE ≤ 0.1193,均方根误差RMSE ≤ 0.1683;黏粒含量的变化范围相对集中,预测能力整体表现的相对较强,R2 = 0.8127,MAE = 0.0820,RMSE = 0.1248。通过筛选最佳光谱变换处理,主成分降维,支持向量机预测,建立了土壤中黏粒、粉粒和砂粒含量的统一估测途径,实现了简单快捷的高光谱估测。     

典型干旱区土壤盐分空间异质性对比研究

研究目的在于采用不同方法研究典型干旱区土壤盐分空间分布,并对比预测精度。研究方法:通过对研究区域进行移动式磁感调查,获取大地表观电导率信息,经校正解译成土壤盐分信息,拟合最优半方差函数,分别利用克里格插值、条件模拟、径向基函数法以及栅格运算方法对土壤盐分空间分布进行预测。研究结果表明:比较四种方法预测精度,条件模拟和克里格插值具有良好的预测精度,径向基函数法和栅格运算方法误差较大。研究结论:通过对研究区域移动式磁感调查,结合条件模拟或克里格插值方法,可以快速而又比较准确的预测土壤盐分空间分布。     

地学环境变量支持的土壤全钾含量自适应曲面建模——以青海湖流域典型地区为例

土壤属性空间分布受地学环境要素影响,空间分异特征十分明显,单一的全局插值模型在应用中常受到一定条件的限制。对复杂地貌类型区土壤属性插值所面临的空间不连续、全局插值模型精度有限以及适应性差的缺点,提出了一种融合地学环境信息的土壤属性自适应曲面建模方法（Adaptive surface modeling for soil properties,ASM-SP）。利用2013年采集的110个样点数据,以土壤全钾含量为例,利用ASM-SP、普通克里格法（Ordinary Kriging,OK）、回归克里格法（Regression Kriging,RK）、地理加权回归克里格法（Geographically Weighted Regression Kriging,GWRK）和协同克里格（Ordinary Co-Kriging,OCK）5种插值方法,分别模拟了青海湖流域典型地区土壤全钾含量的空间分布。采用平均误差（Mean Error,ME）、平均相对误差（Mean Relative Error,MRE）、均方根误差（Root Mean Square Error,RMSE）、准确度（Accuracy,AC）、相关系数、回归系数和决定系数7类指标系统评价不同插值方法的预测效果。结果表明:（1）利用常规插值（OK）得到的插值曲面较为平滑,具有弱“牛眼”效应,在刻画土壤全钾含量的空间变异性方面存在明显不足,精度有待提高。（2）在融合地学环境信息的插值方法中,RK,OCK,GWRK和ASM-SP模拟精度较OK有不同程度提高,其中ASM-SP在刻画土壤全钾含量的空间变异和局部细节信息方面表现突出,精度较其他插值方法有较大程度提高,其准确度较OK,RK,GWRK和OCK分别提高9.27%,6.29%,2.66%和7.74%。ASM-SP尤其适合复杂地貌类型区,因其考虑了地学环境变量与土壤属性的非线性关系,并融合了多个模型的适应性优势,其在刻画土壤属性空间分异的复杂性方面也更加符合实际情况,为土壤属性的空间模拟提供了新思路。     

Vis-NIR光谱信息辅助的土壤质地协同克里格预测制图

土壤质地传统获取方法费时费力,土壤反射光谱信息是土壤综合信息的反应,以光谱信息为辅助变量的土壤质地协同克里格预测制图值得研究。利用采样点的土壤光谱数据主成分分析综合得分作为协同克里格协变量,并结合GS+软件提供的最佳地统计模型预测土壤质地的空间变异。结果表明以光谱信息为辅助变量的协同克里格方法能提高制图精度,减少样本数量。在主变量减少至60个的情况下,砂粒,粘粒含量的预测值和实测值的均方根误差分别为7.85%,3.03%。     

县域耕地土壤锌含量的协同克里格插值及采样数量优化

耕地土壤锌含量的空间分布和适宜的采样数量研究,为农产品产地评价和认证、土壤环境质量调查提供帮助。以土壤有机质、pH、全氮、碱解氮、有效铁、有效铜、有效锰、高程数据为辅助变量,采用协同克里格插值方法,对吉林省舒兰市耕地土壤锌含量进行了插值精度分析,并对不同样点数量下协同克里格法在县域尺度耕地土壤锌含量空间分布研究中的适用性进行了评价。结果表明,相同采样数量下,土壤锌协同克里格法的均方根误差比普通克里格法低,拟合效果更好;辅助因子对土壤锌协同克里格插值的影响效果大小为:全氮>有效锰>高程>有效铁>碱解氮>pH>有机质>有效铜;当高程和有效铜作为辅助因子时,土壤采样数量可以减少20%,土壤锌协同克里格插值仍能满足精度要求。协同克里格作为一种更为精确和经济的空间插值方法,可为县域尺度耕地土壤重金属含量的空间分布研究提供更多的信息和帮助     

冀西山前冲积扇区土壤机械组成模型制图对比研究

探索适合地形平缓的山前冲积扇地区土壤机械组成的空间预测方法。以河北省灵寿、行唐、曲阳县400 m高程以下区域为研究区,结合地形因子、土壤类型、归一化植被指数、地表温度等环境变量,选择基于对称对数比(SLR)转换的普通克里格法(SLR-OK)、回归克里格法(SLR-RK)、随机森林法(SLR-RF)3种方法,对训练集114个样点表层土壤机械组成的空间分布进行预测,并通过验证集50个样点比较了3种方法的预测精度。(1)从空间预测图来看,砂粒呈现出西北低、东南高的空间分布趋势;粉粒和黏粒与砂粒相反。与SLR-OK法相比,SLR-RK法和SLR-RF法能够更好地反映局部变异并减小平滑效应。(2)对于砂粒和粉粒,SLR-RF法对验证集含量预测的平均绝对误差(MAE)和均方根误差(RMSE)均低于其他两种方法,且决定系数最高,表明SLR-RF的预测精度最高;对于黏粒,SLR-OK法对验证集含量预测的MAE和RMSE均低于其他两种方法,且决定系数最高,表明SLR-OK法的预测精度最高。(3)线性回归预测模型的辅助变量包括高程、土壤类型和风力作用指数;随机森林法模型的辅助变量包括高程、土壤类型、归一化植被指数、地表温度、风力作用指数和流量累积,对于砂粒和粉粒,土壤类型和高程是重要的辅助变量,归一化植被指数、地表温度、风力作用指数和流量累积重要性相对较低。研究区砂粒和粉粒空间预测的最优方法为SLR-RF法,在山前冲积扇地区地形较平缓,砂粒和粉粒对地形变量敏感;黏粒空间预测的最优方法为SLR-OK法。     

An assessment of Sentinel-1 synthetic aperture radar,geophysical and topographical covariates for estimating topsoil particle-size fractions

Data derived from synthetic aperture radar (SAR) are widely employed to predict soil properties, particularly soil moisture and soil carbon content. However, few studies address the use of microwave sensors for soil texture retrieval and those that do are typically constrained to bare soil conditions. Here, we test two statistical modelling approaches—linear (with and without interaction terms) and tree-based models, namely compositional linear regression model (LRM) and random forest (RF)—and both nongeophysical (e.g., surface soil moisture, topographic, etc) and geophysical-based (electromagnetic, magnetic and radiometric) covariates to estimate soil texture (sand %, silt % and clay %), using microwave remote sensing data (ESA Sentinel-1). The statistical models evaluated explicitly consider the compositional nature of soil texture and were evaluated with leave-one-out cross-validation (LOOCV). Our findings indicate that both modelling approaches yielded better estimates when fitted without the geophysical covariates. Based on the Nash–Sutcliffe efficiency coefficient (NSE), LRM slightly outperformed RF, with NSE values for sand, silt and clay of 0.94, 0.62 and 0.46, respectively; for RF, the NSE values were 0.93, 0.59 and 0.44. When interaction terms were included, RF was found to outperform LRM. The inclusion of interactions in the LRM resulted in a decrease in NSE value and an increase in the size of the residuals. Findings also indicate that the use of radar-derived variables (e.g., VV, VH, RVI) alone was not able to predict soil particle size without the aid of other covariates. Our findings highlight the importance of explicitly considering the compositional nature of soil texture information in statistical analysis and regression modelling. As part of the continued assessment of microwave remote sensing data (e.g., ESA Sentinel-1) for predicting topsoil particle size, we intend to test surface scattering information derived from the dual-polarimetric decomposition technique and integrate that predictor into the models in order to deal with the effects of vegetation cover on topsoil backscattering.     

Fertilization effects on organic matter of arable soils in diverse environmental conditions of the Czech Republic

Site specific variables and anthropogenic factors influence composition of soil organic matter (SOM). We evaluated quantity and quality of SOM under different fertilization regimes and site conditions. The study combines data based on repeated measurements obtained from six long-term field experiments, which have been established between 1955 and 1983 at ten locations, resulting in thirteen site and experiment combinations. The experimental sites cover a wide spectrum of pedological and climatic conditions of the Czech Republic. Four basic fertilization regimes were selected: unfertilized plots, mineral-only fertilized plots, plots with application of farmyard manure, and both organic and mineral fertilized plots. The study employs compositional data analysis, principal component analysis, and mixed effect linear models for statistical inference. Under combined organic and mineral fertilization, total soil organic C (SOC) increased by 1?3 g kg^?1. Evidence of possible priming effect was obtained for mineral-only fertilization. Local site conditions were the dominant factor shaping SOM properties. The positive relationship between proportion of clay in soil and decomposition index (DI) was confirmed. In the absence of fertilization, DI was eleven times higher in clay-rich than in clay-poor soil. This effect was moderated by fertilization, decreasing to a seven-fold difference under the full fertilization regime.     

Regionalization of organic pollutants in Bavarian soils: The performance of Indicator Kriging

The evaluation of the impact of additional soil pollutants has to be contrasted against the naturally occurring pollutant concentration, i.e., the background concentration. Because background concentrations have to represent areal entities, point information has to be extrapolated into the area using interpolation methods. Thus, the accuracy of the interpolation method is crucial for the correct designation of background values to the areas. For the area of Bavaria (SE Germany), the actual background values of organic and inorganic soil pollutants were derived from >337,000 data from 5000 horizons based upon 1134 soil profiles. Background values were determined for the different soil depth compartments (O layer, topsoil, subsoil, and parent material) and land uses (agriculture, forestry). For interpolation between the nodes, Indicator Kriging was applied. The kriged total area was subdivided into 6 subareas of different background concentrations using percentile thresholds. To derive representative background concentrations, the reliable segregation of the total area into subareas and, thus, a robust interpolation method is a prerequisite. In this study, the robustness of the applied Indicator Kriging should be tested. Influences of data transformations and different kriging methods upon the demarcation of subareas should be investigated for the organic sum‐parameter EPA‐PAH. Neither a data transformation nor the comparison with Ordinary Kriging yielded significant deviations in the assigned subareas. Furthermore, cross‐validation as well as addition of synthetic noise was used to check the susceptibility of the method to artifacts and changes in the data set. After random splitting of the original data set into 4 subsets and re‐arrangement to 6 half‐sets, subsequent Indicator Kriging produced 6 results with mainly identical subarea configurations. Cross‐validation, i.e., comparison of points from the kriging surface (validation data set) with the calibration data set, yielded considerable residuals between estimates and measurements. Based on these normally distributed residuals, random numbers with identical statistical moments were generated and used as measurement errors for another kriging run. This synthetic noise was added to the corresponding result based on the calibration half‐set. The resulting subareas changed only slightly for the most polluted region, but considerably for the other regions. The chosen interpolation method provides sufficient stability to demarcate the relevant areas with elevated pollution in Bavaria. For other areas, its stability is less clear. Here, additional soil samples are required.     

克里格插值在区域土壤有机碳空间预测中的应用

基于河北省第二次全国土壤普查数据,对比了常用土壤有机碳相关因子土地利用和土壤类型与普通克里格插值结合前后对土壤有机碳密度空间预测精度的差异,探讨了普通克里格插值在区域土壤有机碳空间预测中的应用。研究结果表明,土地利用能够独立解释土壤有机碳密度总方差的19.0%,与普通克里格插值结合以后能够将对土壤有机碳密度总方差的解释程度显著提高到30.2%。低级土壤分类土属能够独立解释土壤有机碳密度总方差的45.0%,但与普通克里格插值结合以后对土壤有机碳密度总方差的解释程度为44.8%,两者相差不大。因此区域空间上能否进一步应用普通克里格插值优化土壤有机碳的空间预测与所选用的土壤有机碳相关因子有关。     

基于遥感与电磁感应仪数据的土壤盐分空间变异性

针对目前黄淮海平原存在的土壤盐渍化问题,以河南省封丘县为研究区,利用遥感影像与磁感应电导率仪 （EM38）在田间定位、定点测量,对研究区土壤盐分的空间变异性进行研究。利用Kriging、指数回归和回归-Kriging 3种不同的方法对土壤盐分空间变异性进行分析。结果表明：土壤电导率在一定的区域范围内具有空间结构特征,符合指数模型分布,具有极强的空间自相关性。通过3种不同的方法对土壤盐分的空间变异性进行分析,结果表明Kriging法、指数回归法及回归-Kriging法在描述土壤盐分空间分布的总体趋势上相似,但指数回归法与回归-Kriging法对于盐分的分布表达更具体,更细致。综合运用多种数据源的方法,即遥感影像数据与EM38测量数据相结合,经典统计学与地统计学方法相结合的方法,能够提高土壤盐分空间预测的精度,结果表明研究区土壤盐分由北向南、自西向东有逐渐升高的趋势,具有明显的趋势效应。     

基于GIS和地理加权回归的砂田土壤阳离子交换量空间预测

土壤阳离子交换量(CEC)反映土壤保水保肥能力,研究CEC空间分布可为土壤改良和田间施肥提供理论依据。本文以宁夏香山地区砂田淡灰钙土为研究对象,在土壤CEC和理化性质相关分析基础上以普通克里格(OK)为对照,探索回归克里格(RK)和地理加权回归克里格(GWRK)在CEC空间插值上的应用,并对三者的插值精度及制图效果进行评价。描述统计表明研究区土壤CEC含量均值为10.145cmol/kg,CEC与有机质含量呈显著正相关,与砂粒含量呈显著负相关;地统计分析表明CEC实测值、OLS残差和GWR残差块金系数分别为8.50%、6.36%和7.02%,比值均小于25%,具有强烈空间自相关;对验证点进行插值精度分析,RK和GWRK的相对模型改进值(RI)分别为40.49%、41.50%,插值精度GWRKRKOK;从成图效果看,GWRK中辅助变量参与了局部回归,成图效果更加精细,揭示了更多空间变化细节。本研究结论可为土壤CEC空间预测研究提供可靠的方法借鉴。     

黄土沟壑区小流域土壤pH值的空间分布及条件模拟

土壤pH值是影响土壤养分有效性和化学物质在土壤中行为的主要因素,研究土壤pH值的空间分布特征对于土壤养分管理和土壤污染预测具有重要意义。该文用地统计学方法研究了环境因素复杂的黄土高原小流域土壤pH值空间分布特征。结果表明,黄土沟壑区小流域土壤pH值具有球形—指数套合模型的空间结构特征,其空间异质性主要来源于流域内土地利用和土壤侵蚀等随机因素。与有机质协同的Kriging法能较好地对土壤pH值进行估值,其估值范围小于实测数据,估值误差来源于复杂的环境因素。序贯高斯条件模拟的土壤pH值范围与实测数据接近,模拟的平均值低于实测数据,模拟误差来源于模拟过程中独特的Kriging算法及高斯特性。     

Usability of soil survey soil texture data for soil health indicator scoring

Soil textural information is an important component underlying other soil health indicators. Soil texture analysis is a common procedure, but it can be labor intensive and expensive. Soil texture data typically are available from the Soil Survey Geographic (SSURGO) database, which may be an option for determining soil health texture groups (SHTG). The SSURGO database provides soil texture information in the soil map unit (SMU) name, taxonomic class category (family), and detailed values (≤ 2 mm soil fraction) of percent sand, silt and clay by soil horizon. The objective of this study was to examine the possibility of using SSURGO data for SHTG at the 147-ha Cornell University Willsboro Research Farm in New York state as an alternative for soil texture data determined manually on collected soil core samples. Comparative results revealed that representative values for soil texture from the SSURGO database generally matched measured mean values for all SMUs.