湿地芦苇光谱对富营养化响应的多尺度高光谱遥感研究

近年来大多数滨海河口湿地都受到了人类活动的影响,水体富营养化加剧的问题造成了深远的影响,亟待在不同时空尺度上以快速、稳定、准确的途径监测滨海河口水体富营养化。论文采用能够提取植被氮吸收特征细微变化的高光谱遥感技术,基于3个时相的野外实验测量数据和HJ-1A HSI高光谱遥感数据,在叶片、冠层和景观3种尺度上,研究湿地芦苇的高光谱特征对滨海河口水体氮素浓度变化的时空响应特性,以此避免营养物质脉冲效应导致的错误水体富营养化水平评估结论。研究结果表明：在叶片和冠层尺度,邻近不同氮浓度的水体,湿地芦苇的光谱特征存在差异。而且,所采用的高光谱植被指数能够对邻近水体氮浓度的变化作出响应,各指数与水体氮浓度的相关性在冠层比叶片尺度有所降低,其中TCARI/OSAVI指数与水体氮浓度的相关性较高（叶片尺度R²为0.73,冠层尺度R²为0.7）;在景观尺度,采用基于Bootstrap的SVM模型,构建高光谱指数TCARI/OSAVI的水体总氮浓度估算模型,应用于3个时相HJ-1A HSI高光谱图像。精度检验显示,估算值和实测值之间存在较好的相关性（R²=0.61,RMSE=2.96 mg/L）。综合叶片、冠层和景观3种尺度的研究结果,证明湿地植被的高光谱响应特征可用于滨海河口水体氮浓度的时空分布估算,进而可以指示湿地所在区域水体富营养化水平的变化状况。

Study on Spectral Response of Wetland Reed to Eutrophication by Multiscale Hyperspectral Remote Sensing

In recent years, most coastal estuarine wetlands were influenced by human activities which caused the eutrophication of water body. However, the worse situation of water eutrophication has controversy brought deep influence to social economy. The fast, stable, and correct way monitoring the water eutrophication of coastal estuarine was expected on different temporal and spatial scale. In this paper, hyperspectral remote sensing technology was adopted. It can extract the little change of vegetation’s nitrogen absorption. The data measured in field at three different times and the HJ-1A HSI hyperspectral remote sensing data were used. The hyperspectral feature of wetland vegetation reed located in coastal estuarine region was studied. The spatio-temporal response feature of reed to water eutrophication was researched based on three scales including leaf, canopy, and landscape scale. By this means, the wrong evaluation of water eutrophication induced by pulse effect of nutrients can be corrected. The results showed: There exists difference in the spectral features of leaf and canopy of reed when the nitrogen concentration of water are different; on leaf and canopy scales, all hyperspectral vegetation indices can respond to the change of total nitrogen concentration of water. Compared with leaf scale, the correlation of various hyperspectral vegetation indices with total nitrogen concentration of water on canopy scale was lower. Among them, TCARI/OSAVI index has obvious positive correlation with total nitrogen concentration of water (R² was 0.73 on leaf scale, and R² was 0.7 on canopy scale); on landscape scale, SVM model based on bootstrap was adopted to construct the estimation models of total nitrogen concentration based on hyperspectral index TCARI/OSAVI. The model was used on three time phases of HJ-1A HSI hyperspectral images. The precision validation showed good correlations between the simulated values and the measured values (R²=0.61, RMSE=2.96 mg/L). The spatial distribution maps of total nitrogen concentration of water in Shuangtaizi estuarine region acquired from the images of the three time phases showed obvious spatio-temporal distribution patterns. The research on three kinds of scales of leaf, canopy and landscape proved that the hyperspectral response feature of wetland vegetation can be used to estimate the water nitrogen concentration in coastal estuarine region to indicate the water eutrophication level of wetlands.