南疆干旱区天然胡杨春尺蠖虫害高光谱遥感监测机理及评价模型研究

The natural euphrates poplar forests is the principal part to protect the desert ecosystem around Tarim river basin, but a large area of euphrates poplar forests die of apocheima cinerarius every year, which worsens the ecological damage.Hyperspectral remote sensing is a kind of effective method to monitor insect pests,however, the research on hyperspectral monitoring mechanism and evaluation model of hazard index through ground experiment is the prerequisite of the specific application.So this project takes apocheima cinerarius which is most serious pest in euphrates poplar forests in the southern arid area of Xinjiang as the research object，takes study on hyperspectral response of physiological changes and biochemical changes of the canopy which pests caused as the breakthrough point through factitious mutagenic experiment of pests and hyperspectral reflectance data, physiological data, biochemical data of the canopy and data on drought. Correlation analysis, principal component analysis, wavelet analysis, regression analysis and discriminant analysis methods will be used to study the changes of the hyperspectral data which pests caused, discuss and find hyperspectral monitoring mechanism of insect pests,reveal the influence law of drought on hyperspectral characteristics of pest monitoring,build hyperspectral feature library of pest monitoring,and establish a model estimating physiological and biochemical parameters and a model evaluating hazard index of pests which are suitable for different drought index in the south of xinjiang.And then,the research results can provide theoretical basis and basic data for the application of hyperspectral remote sensing technology in montoring insect pest in the southern arid area of Xinjiang,provide some guidance for accurate spraying of "Aircraft control program about apocheima cinerarius in natural euphrates poplar forests in xinjiang" as well.

胡杨林是维系塔里木河流域荒漠生态系统的主体，但每年都有大片的胡杨林因春尺蠖虫害而枯死，生态环境整体恶化的趋势严重。高光谱遥感技术是监测虫害的一种有效手段，而通过地面试验进行虫害高光谱监测机理和危害指标评价模型的研究是具体应用的前提条件。因此，项目以南疆干旱区天然胡杨林危害最严重的春尺蠖虫害为研究对象，根据人工诱虫试验和实测冠层高光谱反射率数据、理化参数数据、干旱指标数据，采用相关分析、主成分分析、小波分析、回归分析和判别分析等方法，以研究虫害导致的冠层各理化参数变化的高光谱响应为切入点，探寻胡杨虫害的高光谱遥感监测机理，揭示干旱对虫害高光谱特征的影响规律，构建虫害监测的高光谱特征库，并从干旱指标分级角度建立各理化指标估测模型和虫害危害指标评价模型。进而为高光谱遥感技术在南疆干旱区虫害监测方面的应用提供理论依据和地面数据，对"新疆天然胡杨林春尺蠖飞机防治项目"的精准喷药也具有一定的指导意义。

胡杨林经常面临春尺蠖危害，高光谱监测机理和危害指标评价模型的研究是具体遥感监测应用的前提条件。根据春尺蠖的危害特性，以叶面积指数和冠层含水量分别作为春尺蠖危害的一级评价指标和二级评价指标。① 建立胡杨LAI反演模型：使用差值植被指数(DVI)、归一化植被指数(NDVI)、环境植被指数(EVI)和土壤调整植被指数(SAVI)分别建立LAI估测模型。其中采用NDVI指数建立的LAI反演模型或者使用450nm、550nm、680nm和800nm波段建立多元回归模型，获得较满意的结果。.LAI-NDVI模型：y=7.21324*x-2.02458（R2=0.88871，RMSE=0.42853）。.多元回归模型：y=3.01919-5.44648*x1-4.21237*x2-14.57995*x3+10.5085*x4（R2=0.86272，RMSE=0.21216）。.② 叶片水分含量反演模型：多元散射矫正（MSC）、连续投影算法（SPA）和偏最小二乘回归算法的结合获得较好的叶片水分含量检测结果，首先MSC去除了背景噪声对模型精度的影响，其次SPA算法筛选了1169nm、1371nm、1422nm、1581nm和1661nm共五个光谱波段，最后通过偏最小二乘回归算法建立了胡杨叶片水分含量反演模型，预测精度 为 97.734%，相关系数 r=0.78163，RMSE=0.016776。③ 胡杨叶绿素含量反演模型：红边归一化植被指数与胡杨叶片叶绿素含量具有较好的相关性，可用于胡杨叶片叶绿素含量的检测，敏感波段为750nm和705nm。NDVI705与叶绿素a、叶绿素b和叶绿素a+b的相关系数分别为0.68,0.57和0.63。④ 改进胡杨春尺蠖虫害发生期的预测精度：提出了基于粗糙集和PSO-BP神经网络的虫害预测方法，提高网络训练速度和识别精度，识别准确性达到94.8%。⑤ 春尺蠖危害等级评价：按照春尺蠖啃食叶片的比例将虫害分为5级，无啃食为0级，啃食1/4为1级，啃食2/4为2级，啃食3/4为3级，啃食4/4为4级。对应叶面积指数的变化率可直接作为危害等级的一级评价指标，叶面积指数的直接响应因子NDVI 的变化率可作为春尺蠖危害的评价指标。冠层水分含量可作为二级评价指标，作为一级评价指标的补充。进而为高光谱遥感技术在南疆干旱区虫害监测方面的应用提供理论依据和实用模型。