基于表观电导率与实测光谱的干旱区湿地土壤盐分监测

以新疆艾比湖滨盐渍化土壤为对象,利用磁感应电导仪和光谱仪测得的盐渍土表观电导率和可见光/近红外光谱数据,选取与EM38解译的土壤盐分相关性最好的光谱变换形式和特征波长,分别建立多元逐步回归、偏最小二乘回归和支持向量回归的土壤盐分监测模型。结果表明：（1）表观电导率两种模式相结合建立的盐分含量解译模型的拟合优度达到0.91,即在该区域内电磁感应技术可用于土壤盐分含量的间接监测。（2）一阶微分处理优于二阶微分,经一阶微分变换后的光谱可以较好地预测土壤盐分含量。（3）3种建模方法中,支持向量回归的建模精度最高,偏最小二乘回归和多元逐步回归次之。干旱区湖滨湿地土壤盐分含量的估测模型宜选取基于平滑后的原始一阶微分光谱数据建立的支持向量回归模型。     

基于NDVI与偏最小二乘回归的荒漠化地区植被覆盖度高光谱遥感估测

以星载高光谱影像Hyperion为数据源,系统比较了NDVI与偏最小二乘回归（PLS）估测荒漠化地区植被覆盖度的能力,模型的建立（n=46）与独立检验所用样本（n=10）均为地面实测数据。研究结果表明,基于星载高光谱数据的NDVI与PLS模型可以有效地估测荒漠化地区植被覆盖度。相比于宽波段NDVI（RMSEP=10.5618）及基于803.3/671.02 nm计算的标准高光谱NDVI（RMSEP=8.3863）,选择特定高光谱波段（823.65/701.55 nm）构建的NDVI预测植被覆盖度的误差明显较低（RMSEP=6.5189）。基于高光谱所有波段原始反射率、一阶导数及包络线去除光谱的PLS回归模型表现,要明显优于仅利用两个波段信息的NDVI,其中基于原始反射率的PLS回归模型表现最佳,RMSEP为4.4998,约为因变量平均值的23%。     

克里雅河流域土壤盐分光谱定量分析

本文对克里雅河流域进行野外调查、采集土壤样品及其光谱反射特性的测量,通过比较不同光谱预处理的方法建立偏最小二乘回归(PLSR)模型,并利用决定系数(R2)、均方根误差(RMSEP)、残留预测偏差(RPD)对模型的稳定性和预测能力进行检验。结果表明:反射率一阶微分是预测土壤样本盐分含量的最佳光谱指标。PLSR模型在建立土壤光谱与盐分含量关系时较为适用,R2、RMSE和RPD分别为0.77、0.25和1.88。利用反射光谱估算土壤中盐分含量,通过各种光谱预处理方法可以提高估算精度,可以为该区土壤盐渍化评价和生态环境调查提供依据。     

基于多光谱遥感的典型绿洲棉田春季土壤盐分反演及验证

为探索快速提取典型绿洲棉田土壤盐分的有效方法,获取区域尺度的土壤盐渍化特征及空间分布,进而为土壤盐渍化防治提供参考。以新疆兵团农二师31团为研究区域,2019、2021年春季Landsat 8 OLI多光谱影像和野外实测土壤含盐量为数据源,将波段组、光谱指数组和全变量组作为模型输入变量组,采用多元逐步回归（Multiple stepwise regression, MSR）、偏最小二乘回归（Partial least squares regression, PLSR）、极限学习机（Extreme learning machine, ELM）、支持向量机（Support vector machine, SVM）和BP神经网络（Back propagation neural network, BPNN）构建基于3个输入变量组的土壤盐分遥感反演模型,探究输入变量和建模方法对模型精度的影响效果,通过对比确定春季土壤盐分最优反演模型,定量反演地表土壤含盐量。结果表明：（1） 研究区主要为非盐化土和轻度盐化土,总样本变异系数为0.67,呈中等变异性;光谱反射率与土壤盐渍化程度的关系表现为土壤盐渍化越重,光谱反射率越高。（2） 海岸波段（b1）、蓝波段（b2）、绿波段（b3）、红波段（b4）和盐分指数（SI1、SI2、SI3、SI4、S3、S4、S5）均通过显著性检验P<0.01,相关系数均达到0.4以上。（3） 所有模型中,基于全变量组建立的BPNN反演模型精度最高,建模集R²为0.705;验证集R²为0.556。（4） 由反演结果可知,2019、2021年春季耕作区土壤主要为非盐化土,分别占耕作区总面积的55.55%和64.62%,其次为轻度盐化土,分别占44.31%和35.17%;2021年土壤盐渍化程度较2019年有所减轻。     

基于热红外光谱的干旱区土壤盐分监测研究

土壤盐渍化是新疆最常见的土地退化过程,已经严重威胁到了当地的农业生产、生态稳定和经济发展。通过对渭库绿洲土壤含盐量和土壤热红外光谱分析,探讨了土壤含盐量与热红外发射率之间的定量关系。结果表明：（1）盐渍化土壤的发射率随着含盐量的变化而发生变化,当土壤盐分增加时,发射率也随之增大。（2）土壤含盐量与热红外发射率光谱数据相关性在8.5～9.5 μm波段范围内表现尤为显著,相关系数超过0.8,最高为0.90,对应波段范围9.259～9.271 μm。（3）运用回归模型一阶导数变换形式下建模效果和预测精度都是最优的,R²达到了0.899,RMSE最小为1.734。热红外光谱技术可以反演土壤含盐量,为利用热红外遥感识别土壤盐分信息提供技术支撑。     

土壤氯化钠含量高光谱估算模型研究

基于光谱反射率快速、无损的检测优势,以于田地区不同氯化钠含量土壤光谱反射率作为信息源,探讨利用反射光谱估算土壤氯化钠含量的可行性。于2012年7月对于田地区进行野外调查,在测定土壤光谱反射率及氯化钠含量的基础上,对光谱反射率数据进行预处理并建立了逐步多元线性回归(SMLR)模型和偏最小二乘回归(PLSR)模型,并利用决定系数R2和均方根误差RMSE对模型稳定性和预测能力进行检验,进而比较不同预处理方法和不同模型估算结果的适用性。结果表明:反射率二阶微分光谱是预测土壤样本氯化钠含量的最佳光谱指标;偏最小二乘回归(PLSR)模型是建立土壤光谱与氯化钠含量关系的最优模型,R2和RMSE分别为0.812和0.105。利用反射光谱估算土壤氯化钠含量,通过各种光谱预处理方法提高估算精度,可实现在区域尺度上的土壤盐渍化监测和评价。     

MLR和PLSR的沙壤土盐分含量光谱检测对比研究

为了快速有效检测南疆地区典型土壤（沙壤土）的盐分含量变化,利用光谱仪和电导仪测得南疆阿拉尔市红枣种植区盐渍土近红外高光谱和电导率数据,基于7种不同光谱预处理方法和2种特征波长选择算法,分别建立多元线性回归（MLR）和偏最小二乘回归（PLSR）的土壤盐分监测模型。结果表明：7种预处理方法中,归一化,多元散射,变量标准化和一阶导数能够有效提高土壤盐分的预测模型精度。基于多元逐步回归（SMR）波长选择方法的多元线性回归（SMLR）模型的R_val²>0.948 9,RPD>6.294 9,RMSEP<0.435 6;基于连续投影算法（SPA）的多元线性回归（SPA-MLR）模型的R_val²>0.956 8,RPD>6.922 1,RMSEP<0.361 6,预测结果要优于偏最小二乘回归（PLSR）模型,其中基于归一化处理后的SMLR和SPA-MLR的预测精度最为理想,分别为R_val²=0.979 2,RPD=9.907 8,RMSEP=0.287 6和R_val²=0.980 5,RPD=10.50,RMSEP=0.278 3,而且筛选的特征波长较少。说明归一化是更有效的光谱预处理方法,多元线性回归（MLR）更适合建立南疆典型沙壤土盐分含量的预测模型。     

阿拉善地区土壤盐渍化的遥感反演及分布特征

土壤盐渍化是威胁干旱区土地的重要环境问题。利用遥感技术对土壤盐渍化进行动态监测,分析土壤盐度水平与空间分布,有利于掌握土壤盐渍化现状,为土地资源可持续利用提供理论依据。现有研究多在田间尺度,随着土壤环境问题涉及的范围越来越大,区域斑块信息的提取已无法满足宏观地模拟和展示整体土壤环境的空间分布。以阿拉善地区为例,结合遥感光谱指数与实测土壤盐分数据,运用偏最小二乘回归（PLSR）方法,构建区域尺度范围的土壤盐渍化反演模型,实现大面积地区土壤盐度的精准模拟和定量监测。结果表明,构建的模型验证精度达到0.8788,达到极显著水平,预测结果与实际情况相符,可以较准确地模拟研究区土壤盐渍化状况。受地形、气候、景观类型、农业活动以及土地管理等因素的综合影响,阿拉善地区约20%的区域土壤呈现出不同程度的盐渍化,其中黑河下游河岸带、雅布赖山西侧及贺兰山西侧冲积扇土壤盐渍化程度最为严重。本研究可为大面积区域土壤盐分状况的快速监测及遥感定量反演提供可行的方法,同时为该区域不同程度盐渍化土壤的治理和土地利用管理提供依据。     

翅碱蓬高光谱植被指数对土壤化学性质的响应

植被覆盖区土壤化学性质遥感监测一直是一个难点,往往只能通过生物地球化学的方法,利用上覆植被信息间接地反映。该研究通过野外采样分析的17个翅碱蓬(Suaeda salsa)光谱数据和其下土壤样品的理化分析配对数据,探讨土壤化学性质与翅碱蓬高光谱植被指数间的关系。结果表明:上覆翅碱蓬高光谱植被指数与土壤有机质、全氮、速效钾之间均无显著相关,高光谱植被指数(NDNI)可用于初步反映土壤全磷的含量变化,NDVI705可用于初步反映土壤pH值的变化,而高光谱植被指数(MSI)可以很好地反映土壤盐分含量的变化。在此基础上,建立了土壤全盐量与19个高光谱植被指数的偏最小二乘回归模型,这为翅碱蓬覆盖区域土壤盐渍化遥感监测提供了一种方法。     

基于地表光谱建模的区域土壤盐渍化遥感监测研究

土壤盐渍化是土地荒漠化和土地退化的主要类型之一,也是世界性资源问题和生态问题,盐渍化土壤主要分布在干旱半干旱地区,而土壤盐渍化的动态监测和预报是目前研究的重要手段。新疆渭干河-库车河三角洲绿洲是我国土壤盐渍化比较严重也比较典型区域之一,以实测获取的绿洲区域范围内不同恶化程度的盐渍化土壤的高光谱反射率及其土壤含盐量为基础数据源,从中优选出对不同盐渍化程度土壤最为敏感的光谱波段,结合Landsat-TM多光谱遥感影像构建于地表光谱建模的最佳土壤盐渍化监测模型,并将此模型实现大尺度范围内的高精度土壤盐分遥感定量反演。结果表明:(1)反射率一阶微分光谱变换方式最好,与土壤含盐量相关性最佳,最大相关系数可达0.987。(2)利用抛物线模型与最佳土壤盐分指数SI3,构建的地表光谱模型效果最为理想(R2=0.885 2)。(3)与单纯以传统多光谱遥感技术构建的土壤盐分监测模型(R2=0.592 5)相比,经尺度转换后的模型,其精度与效果(R2=0.708 4)则更优。为今后可实现高精度干旱区区域大尺度卫星遥感土壤盐渍化监测提供一种科学、有效的途径。     

Development of a chironomid-based air temperature inference model for the central Canadian Arctic

Subfossil midge remains were identified in surface sediment recovered from 88 lakes in the central Canadian Arctic. These lakes spanned five vegetation zones, with the southern-most lakes located in boreal forest and the northern-most lakes located in mid-Arctic tundra. The lakes in the calibration are characterized by ranges in depth, summer surface-water temperature (SSWT), average July air temperature (AJAT) and pH of 15.5 m, 10.60°C, 8.40°C and 3.69, respectively. Redundancy analysis (RDA) indicated that maximum depth, pH, AJAT, total nitrogen-unfiltered (TN-UF), Cl and Al capture a large and statistically significant fraction of the overall variance in the midge data. Inference models relating midge abundances and AJAT were developed using different approaches including: weighted averaging (WA), weighted averaging-partial least squares (WA-PLS) and partial least squares (PLS). A chironomid-based inference model, based on a two-component WA-PLS approach, provided robust performance statistics with a high coefficient of determination (r ² = 0.77) and low root mean square error of prediction (RMSEP = 1.03°C) and low maximum bias. The use of a high-resolution gridded climate data set facilitated the development of the midge-based inference model for AJAT in a region with a paucity of meteorological stations and where previously only the development of a SSWT inference model was possible. David Porinchu and Nicolas Rolland contributed equally to the work.     

The modern distribution of chironomid sub-fossils (Insecta: Diptera) in the Sierra Nevada, California: Potential for paleoclimatic reconstructions

Surface lake sediment was recovered from 57 lakes along an elevation gradient in the central, eastern Sierra Nevada of California. The surface sediment was analysed for subfossil chironomid remains in order to assess the modern distribution of chironomids in the region. The lakes sampled for the calibration dataset were between 2.0 and 40.0 m in depth, spanned an altitudinal gradient of 1360 m and a surface water temperature gradient of approximately 14 °C. Redundancy analysis (RDA) identified that five of the measured environmental variables – surface water temperature, elevation, depth, strontium, particulate organic carbon – accounted for a statistically significant amount of the variance in chironomid community composition. Quantitative transfer functions, based on weighted-averaging (WA), partial least squares (PLS) and weighted-averaging partial least squares (WA-PLS), were developed to estimate surface water temperature from the chironomid assemblages. The best model was a WA model with classical deshrinking, which had a relatively high coefficient of determination (r² = 0.73), low root mean square error of prediction (RMSEP = 1.2 °C) and a low maximum bias (0.90 °C). The results from this study suggest that robust quantitative estimates of past surface water temperature can be derived from the application of these models to fossil chironomid assemblages preserved in late-Quaternary lake sediment in this region.     

Diatom-based water chemistry reconstructions from northern Sweden: a comparison of reconstruction techniques

The diatom composition in surface sediments from 119 northern Swedish lakes was studied to examine the relationship with lake-water pH, alkalinity, and colour. Diatom-based predictive models, using weighted-averaging (WA) regression and calibration, partial least squares (PLS) regression and calibration, and weighted-averaging partial least squares (WA-PLS) regression and calibration, were developed for inferences of water chemistry conditions. The non-linear response between the diatom assemblages and pH and alkalinity was best modelled by weighted-averaging methods. The lowest prediction error for pH was obtained using weighted averaging, with or without tolerance downweighting. For alkalinity there was still some information in the residual structure after extracting the first weighted-averaging component, which resulted in a slight improvement of predictions when using a two component WA-PLS model. The best colour predictions were obtained using a two component PLS model. Principal component analysis (PCA) of the prediction errors, with some characteristics of the training set included as passive variables, was performed to compare the results for the different alkalinity predictive models. The results indicate that calibration techniques utilizing more than one component (PLS and WA-PLS) can improve the predictions for lakes with diatom taxa that have broad tolerances. Furthermore, we show that WA-PLS performs best compared with the other techniques for those lakes that have a high relative abundance of the most dominant taxa and a corresponding low sample heterogeneity.     

基于TM影像的荒漠绿洲交错带土壤有机质含量反演模型

选取Landsat TM影像的光谱反射率（R）、反射率之倒数（1/R）、反射率倒数之对数（lg（1/R））、反射率一阶导数（FDR）、波段深度（D）等5种光谱指标,分别建立了奇台县土壤有机质（SOM）含量的反演模型,并利用F检验来验证模型的显著性。结果表明：用各光谱指标建立的土壤各层和不同深度SOM含量的反演模型精度值由低到高的顺序均为lg（1/R）<R<1/R<FDR<D,以D反演SOM含量的模型效果最好,且对10～20 cm的SOM含量的反演精度最高,适用于对研究区SOM含量的反演,FDR的反演效果次之,1/R和R的模型精度一般,而lg（1/R）的模型精度最差;各层拟合模型的反演精准度由低到高的顺序为50～60 cm <40～50 cm <30～40 cm <20～30 cm <0～10 cm <10～20 cm,不同深度反演模型的优劣为0～60 cm <0～50 cm <0～40 cm <0～30 cm <0～10 cm <0～20 cm。     

Fourier-transform Infrared Spectroscopy (FTIRS), a New Method to Infer Past Changes in Tree-line Position and TOC using Lake Sediment

This study tests the hypothesis that Fourier-transform infrared spectroscopy (FTIRS) of lake sediments can be used to infer past changes in tree-line position and total organic carbon (TOC) content of lake water. A training set of 100 lakes from northern Sweden spanning a broad altitudinal and TOC gradient from 0.7 to 14.9 mg/l was used to assess whether vegetation zones and TOC can be modelled from FTIR spectra of surface sediments (0–1 cm) using principal component analysis (PCA) and partial least squares (PLS) regression. Preliminary results show that FTIRS of lake sediments can be used to reconstruct past changes in tree line and the TOC content of lake water, which is hardly surprising since FTIRS registers the properties of organic and minerogenic material derived from the water mass and the drainage area. The FTIRS model for TOC gives a root mean squared error (RMSECV) of calibration of 1.4 mg/l (10% of the gradient) assessed by internal cross-validation (CV) yielding an R_cv² of 0.64. This should be compared with a near-infrared spectroscopy (NIRS) and diatom transfer function for TOC from the same set of lakes, which have a R_cv² of 0.61 and 0.31, and RMSECV of 1.6 and 2.3 mg/l, respectively. The FTIRS-TOC model was applied to a Holocene sediment core from a tree-line lake and the results show similar trends as inferences from NIRS and pollen from the same core. Overall, the results indicate that changes in FTIR spectra from lake sediments reflect differences in catchment vegetation and TOC, and that FTIRS-models based on surface-sediment samples can be applied to sediment cores for retrospective analysis.     

A TEST OF SIGNIFICANCE FOR PARTIAL LEAST SQUARES REGRESSION

Partial least squares (PLS) regression is a commonly used statistical technique for performingmultivariate calibration, especially in situations where there are more variables than samples. Choosingthe number of factors to include in a model is a decision that all users of PLS must make, but iscomplicated by the large number of empirical tests available. In most instances predictive ability is themost desired property of a PLS model and so interest has centred on making this choice based on aninternal validation process. A popular approach is the calculation of a cross-validated r~2 to gauge howmuch variance in the dependent variable can be explained from leave-one-out predictions. Using MonteCarlo simulations for different sizes of data set, the influence of chance effects on the cross-validationprocess is investigated. The results are presented as tables of critical values which are compared againstthe values of cross-validated r~2 obtained from the user's own data set. This gives a formal test forpredictive ability of a PLS model with a given number of dimensions.