基于TM影像的沱江中游土壤重金属预测研究

发布时间：2018-01-15 04:18

本文关键词：基于TM影像的沱江中游土壤重金属预测研究　出处：《四川农业大学》2015年硕士论文　论文类型：学位论文

【摘要】：为找到一种简便快捷的土壤重金属污染定量化监测方法,本文利用TM遥感影像的各波段光谱反射率、9种植被指数、4种地面辅助因子和80%样点土壤重金属含量实测值建立线性回归预测模型,将未参与建模但与土壤重金属含量显著相关的因子作为修正因子进行蒙特卡罗模拟。剩余的20%样点作为检验点对各类模型进行趋势分析和误差统计检验。最后采用普通克里格方法反演实测值、预测值的空间分布状况,对比其差异。主要结论如下：运用TM影像光谱信息和其他因子可以通过建立模型预测土壤中重金属砷、汞、铬、镉、铜、铅、有效锰、有效铁、有效锌的含量,其中有效铁的预测效果最优。使用的建模因子是从TM遥感影像、数字高程模型中提取的,容易获取,能有效节约成本和时间。遥感影像各波段光谱反射率单独建模、结合地面因子共同建模、不同地貌分别建模这三种建模均可预测研究区土壤重金属含量(P0.01),R2值、平均误差、总均方根误差及平均相对误差值显示三种建模方式的精度是递增的。因此在地貌复杂的地区利用遥感研究土壤重金属有必要分地貌建立预测模型。蒙特卡罗模拟对预测模型有修正作用,砷、汞、铬、镉、铜、铅、有效锰、有效铁、有效锌蒙特卡罗模拟修正后的预测模型其均方根误差比修正前的线性回归预测模型分别降低了22.7%、18.0%、37.7%、36.2%、13.5%、30.5%、5.4%、25.2%、11.9%,表明蒙特卡罗模拟可以提高模型精度。但修正因子与目标值相关程度的差异会使不同修正因子的效果不同。趋势分析、误差统计检验以及空间分布图显示：砷的波段2和大气抗阻植被指数平均法修正预测模型效果最佳,汞的波段3修正预测模型效果最佳,铬的归一化植被指数修正预测模型效果最佳,镉的坡度修正预测模型效果最优,铜的有效叶面积植被指数修正预测模型效果最优,铅的波段5和土壤显色指数几何平均法修正预测模型效果最佳,有效锰的波段3修正预测模型效果最佳,有效铁的波段1修正预测模型效果最佳,有效锌的波段7和有效叶面积植被指数几何平均法修正预测模型效果最佳。
[Abstract]:In order to find a simple and fast quantitative monitoring method for soil heavy metal pollution, the spectral reflectance of TM remote sensing image was used to measure the soil heavy metal pollution by using the spectral reflectance of the TM image. A linear regression prediction model was established for 4 kinds of ground auxiliary factors and 80% sample soil heavy metal contents. Monte Carlo simulation was carried out by using factors which were not involved in modeling but were significantly related to soil heavy metal content. The remaining 20% samples were used as test points for trend analysis and error statistical test of various models. Then the common Kriging method is used to retrieve the measured data. The main conclusions are as follows: using TM image spectral information and other factors, we can predict the heavy metals arsenic, mercury, chromium, cadmium, copper and lead in soil by establishing models. The content of available manganese, iron and zinc, among which the prediction effect of available iron is the best. The modeling factors used are extracted from TM remote sensing image, digital elevation model, easy to obtain. It can save cost and time effectively. The spectral reflectance of each band of remote sensing image can be modeled separately and combined with ground factors. The three models can predict the soil heavy metal content (P0.01C ~ (2)) and average error in the study area. The total root mean square error and the average relative error show that the accuracy of the three modeling methods is increasing. Therefore, it is necessary to establish a geomorphological prediction model by remote sensing in areas with complex geomorphology. It can modify the prediction model. For arsenic, mercury, chromium, cadmium, copper, lead, available manganese, available iron and available zinc, the root-mean-square error of the modified model is 22.7% lower than that of the linear regression model. 18.0 37.7%, 36.2%, 30.5%, 30.5%, and 30.5%, about 25.20.5% and 11.9%. The results show that Monte Carlo simulation can improve the accuracy of the model, but the difference between the correction factor and the target value will make the effect of different correction factors different. The error statistical test and spatial distribution map showed that the modified prediction model of arsenic in band 2 and the average method of atmospheric resistance vegetation index was the best, and the modified model of mercury in band 3 was the best. The effect of normalized vegetation index correction model of chromium is the best, that of cadmium slope correction model is the best, and that of copper effective leaf area vegetation index correction model is the best. Band 5 of lead and geometric average of soil color index were the best, band 3 of effective manganese and band 1 of effective iron were the best. Band 7 of available zinc and geometric average of vegetation index of effective leaf area are the best.
【学位授予单位】：四川农业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：X87;X833

【相似文献】