基于地面高光谱遥感的降香黄檀黑痣病病情指数反演

利用美国Spectra Vista Corporation(以下均用简称SVC)HR-1024i非成像高光谱仪采集不同病情程度的降香黄檀冠层光谱数据,并结合地面同步调查获得的降香黄檀黑痣病病情指数数据,对光谱数据进行重叠校正(scan matching/overlap correction)和白光板反射率校正(white plate reflectance correction)。采用主成分分析法(PCA法)对与降香黄檀黑痣病病情指数相关性较高的敏感波段进行降维。利用53个训练集,将敏感波段和PCA法处理后的敏感波段分别作为输入变量,训练降香黄檀黑痣病的BP神经网络。两种输入变量建立的神经网络计算出的预测值与实际值之间的决定系数(R2)均达到99%。利用27个验证集做进一步精度检验,结果表明,通过这两种输入变量训练的BP神经网络,得到的预测值与实际值之间的决定系数(R2)分别为0.951 9和0.706 0,均方根误差(RMSE)分别为5.998 0和12.919 3。直接以敏感波段作为变量输入和PCA法处理后的敏感波段作为变量输入训练BP神经网络是一种有效的方法,其中,直接以敏感波段作为变量输入精度更高。     

基于PLSR-BP复合模型的绿洲土壤pH高光谱反演

以土壤pH、野外实测光谱以及多元散射校正(MSC)预处理后的光谱数据为基础,利用数学方法（主成分回归PCA、偏最小二成回归PLSR、BP神经网络模型）分别建立了土壤pH的预测模型。结果表明：土壤实测光谱和经过MSC方法预处理的光谱数据均与pH存在良好的相关性,并呈极显著水平,后者的相关性更高。PCA和PLSR两种土壤pH估测模型均具有良好的预测能力。BP神经网络模型则因输入变量多,预测精度较低。但利用PCA和PLSR模型所获得主成分,作为BP神经网络的输入变量所建立的复合模型,可明显提高模型稳定性和预测能力。     

基于流形学习的土壤高光谱数据特征提取研究

尾矿重金属污染是当今矿区环境污染面临的主要问题之一,精确反演土壤重金属含量对矿区土壤污染监测和治理具有非常重要的意义。以陕西金堆城矿区尾矿为研究区,利用ASD光谱仪测量土壤光谱,通过实验室化学分析获取土壤样本铜元素含量;将Isomap流形学习方法应用于土壤高光谱数据降维,利用随机森林方法对矿区尾矿土壤的Cu含量进行反演建模,并与原始高光谱数据反演结果和PCA降维后的反演结果进行对比。结果表明:土壤铜含量反演模型在经过Isomap降维后的光谱数据集上预测铜元素含量的相关系数R2为0.7272,均方根误差RMSE为1140.20,在预测的准确性方面均优于原始高光谱数据。研究结果为探索土壤高光谱数据特征提取提供了理论依据,同时对尾矿重金属污染监测具有重要的现实指导意义。     

干旱区土壤盐渍化信息遥感建模

以新疆塔里木盆地北缘的渭干河-库车河三角洲绿洲为研究区,利用GF~(-1)与Landsat8 OLI影像数据作为基本数据源,从影像上提取15个盐分指数和5个光谱植被指数,通过灰度关联分析法,对0~10 cm表层土壤含盐量与影像光谱指数进行分析和筛选,确定出与土壤含盐量相关性较高的综合光谱指数,采用多元线性回归,偏最小二乘法回归,支持向量机回归三种方法分别对GF~(-1)与Landsat8 OLI影像构建基于实测数据和影像数据的综合指数土壤含盐量估算模型,并选出最优模型。结果表明:(1)在20个光谱指数中,相关性较好的光谱指数是SR、CSRI、SI、BI、S6、ARVI、SAVI、NDSI,关联系数均达到0.7以上,并基于这8个光谱指数构建综合光谱指数。(2)3种估算模型:基于GF~(-1)多元线性回归模型决定系数R~2为0.6856,高于决定系数R2为0.5142的Landsat8 OLI;偏最小二乘回归模型1~8个主成分,GF~(-1)决定系数2个3个1个,其中2个主成分最高可达0.6104,Landsat决定系数4个3个2个,其中4个主成分最高可达0.549;支持向量机模型3种函数,GF~(-1)决定系数RBFPolynomialLinear,其中RBF函数最高可达0.7969,Landsat决定系数PolynomialRBFLinear,其中Polynomial函数最高可达0.7154。对比3种模型可知,支持向量机回归模型的R2最高,因此该模型相对于多元线性回归和偏最小二乘回归更适于土壤盐渍化估算。     

干旱区盐渍化土壤高光谱遥感信息分析与提取

以干旱区典型区域新疆渭干河-库车河三角洲绿洲为研究区,以环境小卫星高光谱影像及野外实测土壤含盐量为主要数据源,进行光谱反射率及其变换形式与土壤含盐量的相关性分析,筛选盐渍化土壤响应最敏感波段,利用多元线性回归分析方法,建立基于HSI影像的研究区土壤含盐量定量反演模型。结果表明:研究区土壤含盐量与HSI波段的敏感性随着波长的增加而增强,位于近红外波段范围(797.826-923.913nm)的相关系数R普遍较高,基本在0.7左右。土壤光谱反射率对数的倒数一阶微分变换在628.261nm和923.913nm的波段组合为最佳敏感波段,所构建的土壤含盐量反演模型为最优模型,模型方程为Y=-11.731-114.996X628.261-186.637X923.913,模型及检验的决定系数R2都在0.85以上,均方根误差RMSE约为2.7。该模型的建立为地区土壤含盐量信息的提取及监测提供了参考。     

基于垂直植被指数的东北黑土区玉米LAI反演模型研究

本文旨在探讨以不同波段组合垂直植被指数所建立的高光谱模型对玉米叶面积指数(Leaf Area Index,LAI)的反演精度。在不同水肥耦合作用条件下,实测玉米冠层的高光谱反射率与叶面积指数数据以及裸土的高光谱反射率数据,在高光谱红光波段(631~760 nm)与近红外波段(760~1050 nm)逐波段构建土壤线,并在此基础上构建垂直植被指数(Perpendicolar Vegetation Index,PVI),找出与LAI具有最佳相关性波段组合PVI,建立玉米LAI估算模型。结果显示,采样波段间隔越窄,反演精度越高,在采样波段间隔1.4 nm的PVI(R677,R918)反演2004年的玉米LAI模型中,最佳回归方程是指数函数,精度达91.1%,标准差为0.1997,RMSE=0.0399,通过了0.01极显著验证。采用高光谱数据构建的PVI植被指数对玉米LAI的估算可以取得较高的精度。     

基于Landsat 8 OLI影像的三江源区表层土壤全氮空间格局反演

利用Landsat 8 OLI影像反演三江源区玉树、称多及玛多县的表层土壤全氮含量空间分布格局,选取光谱反射率(R)、光谱反射率的倒数(1/R)、光谱反射率倒数的对数〔lg(1/R)〕3个光谱指标,与表层土壤(0~30 cm)全氮实测数据进行相关性分析,筛选相关性最高的光谱指标,以达到显著性相关水平波段的主成分分量建立回归模型。结果表明:OLI影像的B1~B4和B7的R、1/R、lg(1/R)均与实测全氮数据达到显著性相关水平,以lg(1/R)变换最为明显;利用这5个波段lg(1/R)的第一、第二主成分建立负二次多项式回归模型,其中建模样本的R2为0.621,RMSE为2.075,验证样本的R2为0.730,RMSE为1.493,RPD为1.849,反演模型精度较高,稳定性较好。利用OLI影像可较好的估算表层土壤全氮含量的空间分布格局。     

基于高光谱的区域土壤质地预测模型建立与评价——以河套灌区解放闸灌域为例

为了运用光谱反射率快速确定土壤质地,对河套灌区6种不同类型土壤质地在室内进行光谱反射率测试,分别运用一元线性回归、逐步多元回归及BP神经网络三种方法建立光谱反射率与土壤砂粒含量及粉粒含量的拟合模型,并利用估测数据对样品进行土壤质地的模拟。结果显示:三种预测模型精度及其预测能力均较为满意,其中BP神经网络的拟合效果最好,砂粒,粉粒估测模型的决定系数R2均为0.86,外部检验决定系数R2分别为0.88,0.90。利用BP神经网络预测得出的粒径含量对样本质地重新判定,发现达到91.74%的样本符合类别分类要求。研究结果为利用高光谱图像大范围确定土壤质地奠定了基础,对于未来区域模型模拟和土壤水力参数推求具有重要指导意义和应用价值。     

BP神经网络的沙漠化土地信息提取研究

以塔克拉玛干沙漠南缘策勒绿洲为例,探讨了基于主成分融合的沙漠化信息的提取方法.由于Landsat-7 ETM 的全色波段与多光谱波段有相同成像条件,影像获取时间一致,两种不同分辨率的数据可以不经配准而实现高精度融合.首先,对Landsat-7ETM 的全色图像与多光谱图像进行主成分融合处理,再利用BP神经网络模型,以相同的训练样本分别对融合前后的影像进行分类,在此基础上进行沙漠化信息的提取.结果表明:主成分变换融合图像的光谱信息保持性、信息量以及空间分解力都较高,且分类精度比Landsat-7ETM 多光谱图像有较大提高,是监测沙漠化土地变化的有效手段.     

不同地下水埋深条件下胡杨叶片光谱特征分析

用HR-768型光谱仪,实地测定塔里木河下游不同地下水埋深条件下胡杨叶片高光谱数据及其含水率。结果表明:胡杨叶片含水率随地下水埋深下降而降低,但其实测光谱曲线对地下水埋深梯度变化并无明显的响应规律。用ENVI软件去除实测光谱数据包络线后,得出1455nm附近胡杨叶片归一化光谱曲线对地下水埋深梯度变化具有明显的响应规律。采用相关系数法分析所测光谱与含水率的关系,结果表明:1466～1646nm波段是胡杨叶片光谱对水分响应的敏感波段。以1466～1646 nm波段胡杨叶片光谱水分吸收深度和实测含水率构建胡杨含水率估测模型:Y=7.309X1474-5.77X1482-0.629(R2=0.954),模型预测值与实测值的相对误差在1.02%～2.73%之间,平均相对误差为1.59%,模型精度较高。这一研究将为使用高光谱数据进行胡杨生境监测与反演提供依据。     

旺长期烤烟对不同程度干旱胁迫的光谱响应

基于不同时长和程度的干旱胁迫试验,采用ASD光谱仪,研究了不同干旱胁迫处理下不同基因型烤烟叶片水分含量与光谱特征的变化规律,分析不同烤烟叶片水分指标FMC(叶片相对含水量)、LEWT(叶片等效水厚度)与光谱特征参数间的相关关系,构建烤烟叶片水分指标的高光谱特征参数模型。结果表明:不同程度干旱胁迫下烤烟叶片含水量和光谱反射率均随干旱程度的加重而降低,在不同干旱时长下,FMC的变化并不明显,而LEWT对其较为敏感,说明不同干旱胁迫处理间EWT的差异要比FMC显著,不同基因型烤烟品种表现一致,表明LEWT比FMC更适合反映旺长期烤烟叶片水分状况。利用光谱参数建立FMC和LEWT的一元及多元线性模型和BP神经网络模型中,均以BP神经模型网络模型效果最好,其模型R2(决定系数)分别为0.8650、0.9464,RMSE(均方根误差)分别达到0.0049、0.0047,表明模型的精度和稳定性均较好。     

组合灰色预测模型在入库流量预测中的应用

本文将灰色GM(1,1)模型、BP人工神经网络和马尔柯夫链相结合,利用历年入库流量及千河径流量建立组合模型对入库流量进行预测。GM(1,1)模型主要预测趋势,其前半部分与实测值拟合较好,BP神经网络模型后半部有波动部分与实测值拟合较好,二者结合使相对误差最小建立组合模型,同时运用马尔柯夫链预测入库流量的变化范围。预测2001和2002年的入库流量对模型进行检验:GM(1,1)模型预测的相对误差分别为0.359和-0.017;BP神经网络预测的相对误差分别为0.032和-0.251,组合模型相对误差分别为0.164和0.117,组合预测值在预测区间之内,该组合模型预测结果合理有效,能更精确预测冯家山水库入库流量。     

适用于我国的稻纵卷叶螟长期预测模型

为给我国稻纵卷叶螟Cnaphalocrocis medinalis防治提供前期预警,使用R语言软件对我国15个省市区稻纵卷叶螟发生等级与全球海温场资料进行遥相关分析,绘制相关系数的时空间分布图,筛选出显著相关海温区作为预测因子,根据各省市区虫情数据组建回归模型+判别模型、BP神经网络模型和支持向量机(SVM)模型,比较3种模型的历史回检率和预测完全准确率。结果显示,3种模型对稻纵卷叶螟发生等级均有一定的预测能力,其中判别模型+回归模型效果最好,预检完全准确率可达到75.0%,BP神经网络模型次之,预检完全准确率为68.2%,SVM模型预测效果最差,预检完全准确率为54.5%。进一步分析建模所使用的50个预测因子的空间位置,在南印度洋和北大西洋确定3个预测指标,预检准确率为94.4%。通过海温场数据建立的我国15个省市区稻纵卷叶螟发生等级预测模型,适用于长期预测预报。判别模型+回归模型更适合在样本量少、预测因子相关性强的地区建模,而根据预测因子空间分布选择的预测指标进行定性预测准确率更高。     

The application of artificial neural networks in plant protection*

Artificial neural networks are powerful predictive tools that have the ability to detect and approximate non‐linear relationships from the data. In an explorative analysis, artificial neural networks were used to predict the geographic distribution of groups of polyphagous plant pests. Using climate variables as predictors, artificial neural network models were compared with binary logistic models for predicting insect distribution. Using bootstrapping, artificial neural networks were shown to predict insect presence and absence significantly better than the binary logistic regression models. Results from the study suggest that artificial neural networks have the potential for application in many areas of plant protection and biosecurity.     

谷子叶绿素含量高光谱特征分析及其反演模型构建

基于高光谱数据综合分析不同施肥条件下谷子各生长期冠层叶绿素含量的高光谱特征,在分析各光谱特征参数与叶绿素相关性的基础上,基于偏最小二乘法和人工神经网络构建叶绿素含量的遥感反演模型.结果表明:NDVI(归一化植被指数)、GNDVI(绿色归一化植被指数)、PSNDa(特殊色素归一化指数a)、PSSRc(特征色素简单比值指数...     

西北地区小麦黄矮病流行分析及人工神经网络模型的构建

对西北地区半干旱气候区小麦黄矮病1992—2009年发生、流行情况进行长期监测、分析,选择制约小麦黄矮病发生、流行的23个因素,利用三层人工神经网络可以逼近任意连续函数,对非线性预测系统进行模拟处理的特点,分析所选预测分子,提出一套完整的建立BP人工神经网络模型的方法,并建立陕西省BP神经网络长期预测模型。对1992—2006年数据进行网络训练,利用2007—2009年数据进行测试。结果表明,以发病率为指标,输出结果误差在0.001~0.034之间;以发病级别作为预测结果,模型计算得出的数值与实际病级完全吻合,准确率为100%。说明利用神经网络建立小麦黄矮病预测模型是可行的。     

Weather-based prediction of anthracnose severity using artificial neural network models

Data were collected and analysed from seven field sites in Australia, Brazil and Colombia on weather conditions and the severity of anthracnose disease of the tropical pasture legume Stylosanthes scabra caused by Colletotrichum gloeosporioides . Disease severity and weather data were analysed using artificial neural network (ANN) models developed using data from some or all field sites in Australia and/or South America to predict severity at other sites. Three series of models were developed using different weather summaries. Of these, ANN models with weather for the day of disease assessment and the previous 24 h period had the highest prediction success, and models trained on data from all sites within one continent correctly predicted disease severity in the other continent on more than 75% of days; the overall prediction error was 21·9% for the Australian and 22·1% for the South American model. Of the six cross-continent ANN models trained on pooled data for five sites from two continents to predict severity for the remaining sixth site, the model developed without data from Planaltina in Brazil was the most accurate, with >85% prediction success, and the model without Carimagua in Colombia was the least accurate, with only 54% success. In common with multiple regression models, moisture-related variables such as rain, leaf surface wetness and variables that influence moisture availability such as radiation and wind on the day of disease severity assessment or the day before assessment were the most important weather variables in all ANN models. A set of weights from the ANN models was used to calculate the overall risk of anthracnose for the various sites. Sites with high and low anthracnose risk are present in both continents, and weather conditions at centres of diversity in Brazil and Colombia do not appear to be more conducive than conditions in Australia to serious anthracnose development.     

Influences of drip and flood irrigation on soil carbon dioxide emission and soil carbon sequestration of maize cropland in the North China Plain

The need is pressing to investigate soil CO_2(carbon dioxide) emissions and soil organic carbon dynamics under water-saving irrigation practices in agricultural systems for exploring the potentials of soil carbon sequestration. A field experiment was conducted to compare the influences of drip irrigation(DI) and flood irrigation(FI) on soil organic carbon dynamics and the spatial and temporal variations in CO_2 emissions during the summer maize growing season in the North China Plain using the static closed chamber method. The mean CO_2 efflux over the growing season was larger under DI than that under FI. The cumulative CO_2 emissions at the field scale were 1959.10 and 1759.12 g/m~2 under DI and FI, respectively. The cumulative CO_2 emission on plant rows(OR) was larger than that between plant rows(BR) under FI, and the cumulative CO2 emission on the irrigation pipes(OP) was larger than that between irrigation pipes(BP) under DI. The cumulative CO_2 emissions of OP, BP and bare area(BA) under DI were larger than those of OR, BR and BA under FI, respectively. Additionally, DI promoted root respiration more effectively than FI did. The average proportion of root respiration contributing to the soil CO_2 emissions of OP under DI was larger than that of OR under FI. A general conclusion drawn from this study is that soil CO_2 emission was significantly influenced by the soil water content, soil temperature and air temperature under both DI and FI. Larger concentrations of dissolved organic carbon(DOC), microbial biomass carbon(MBC) and total organic carbon(TOC) were observed under FI than those under DI. The observed high concentrations(DOC, MBC, and TOC) under FI might be resulted from the irrigation-associated soil saturation that in turn inhibited microbial activity and lowered decomposition rate of soil organic matter. However, DI increased the soil organic matter quality(the ratio of MBC to TOC) at the depth of 10–20 cm compared with FI. Our results suggest that the transformation from conventional FI to integrated DI can increase the CO2 emissions and DI needs to be combined with other management practices to reduce the CO_2 emissions from summer maize fields in the North China Plain.     

Field models for the prediction of leaf infection and latent period of Fusicladium oleagineum on olive based on rain,temperature and relative humidity

Although much is known about the effect of climatic conditions on the development of peacock leaf spot of olive, field‐operational models predicting disease outbreaks are lacking. With the aim of developing such models, a 10‐year survey was conducted to relate leaf infection to climate parameters that can be easily monitored in the field. As outbreaks of disease are known to be linked to rain, models were evaluated for their ability to predict whether infection would occur following a rain event, depending on air temperature and duration of relative humidity above 85%. A total of 134 rain events followed by confirmed leaf infection and 191 rain events not followed by detectable infection were examined. The field data were adequately fitted (both specificity and sensitivity >0·97) with either a multilayer neural network or with two of six tested regression models describing high boundary values of high humidity duration, above which no infection occurred over the temperature range, and low boundary values below which no infection occurred. The data also allowed the selection of a model successfully relating the duration of latent period (time between infection and the first detection of leaf spots) as a function of air temperature after the beginning of rain (R² > 0·98). The predictive abilities of these models were confirmed during 2 years of testing in commercial olive orchards in southern France. They should thus provide useful forecasting tools for the rational application of treatments and foster a reduction in fungicide use against this major disease of olive.