李珊, 卢霞, xxxx. 黄河三角洲滨海湿地碱蓬叶绿素荧光参数Fv/Fm高光谱遥感反演[J]. 海洋科学进展, x(x): xx-xx. doi: 10.12362/j.issn.1671-6647.20230505001.
引用本文: 李珊, 卢霞, xxxx. 黄河三角洲滨海湿地碱蓬叶绿素荧光参数Fv/Fm高光谱遥感反演[J]. 海洋科学进展, x(x): xx-xx. doi: 10.12362/j.issn.1671-6647.20230505001.
LI S, LU X, xxxx. Hyperspectral remote sensing inversion of the chlorophyll fluorescence parameters Fv/Fm of suaeda salsa in the yellow river delta wetland[J]. Advances in Marine Science, x(x): xx-xx. DOI: 10.12362/j.issn.1671-6647.20230505001
Citation: LI S, LU X, xxxx. Hyperspectral remote sensing inversion of the chlorophyll fluorescence parameters Fv/Fm of suaeda salsa in the yellow river delta wetland[J]. Advances in Marine Science, x(x): xx-xx. DOI: 10.12362/j.issn.1671-6647.20230505001

黄河三角洲滨海湿地碱蓬叶绿素荧光参数Fv/Fm高光谱遥感反演

Hyperspectral Remote Sensing Inversion of the Chlorophyll Fluorescence Parameters Fv/Fm of Suaeda Salsa in the Yellow River Delta Wetland

  • 摘要: 为了实现对植物健康状况的快速、无损监测,本研究以黄河三角洲滨海湿地土著植物碱蓬为研究对象,采集了碱蓬滩湿地无人机高光谱影像和123个碱蓬样方的叶绿素荧光参数Fv/Fm,分析了光谱比值、差值和归一化植被指数与叶绿素荧光参数Fv/Fm的相关性,采用统计分析算法和机器学习算法分别优选植被指数和Fv/Fm的遥感反演模型并进行评价,利用最优模型获取基于无人机影像的Fv/Fm空间分布,以期为动态监测滨海湿地土著植物碱蓬的生长提供理论基础。研究结果表明:黄河三角洲滨海湿地碱蓬Fv/Fm与光谱植被指数D450/D756D612D849和(D517D851)/(D517+D851)的相关性较高。其中,利用Fv/Fm和植被指数(D517D851)/(D517+D851)构建的统计分析模型精度最高(R2=0.606, RMSE=0.006),模型预测值和实测值之间的R2为0.739,RMSE为0.724;利用Fv/Fm和植被指数D612D849构建的机器学习模型精度最高,其中基于BP神经网络算法构建的反演模型R2为0.821,RMSE为0.261;基于PSO-DELM算法构建的反演模型R2为0.982,RMSE为0.140。该研究可以为无人机高光谱影像的碱蓬叶绿素荧光参数反演研究提供重要参考,为大面积碱蓬叶绿素荧光参数估测提供技术支持。

     

    Abstract: Chlorophyll fluorescence parameters, especially primary photochemical maximum mass yield (Fv/Fm), are the indicative probes reflecting plant photosynthesis. The combination of remote sensing and chlorophyll fluorescence techniques can realize rapid and non-destructive monitoring of plants. It is of great significance for the vegetation growth monitoring and the ecological restoration in the coastal wetlands. The Suaeda salsa in the Yellow River Delta wetland was taken as the research object, and the chlorophyll fluorescence parameter Fv/Fm in 123 sample plots and an unmanned aerial vehicle (UAV) hyperspectral image of 400 m \subseteq 600 m were collected. Based on this, the correlation analysis between some vegetation indices such as ratio, difference, and normalized difference vegetation index and the chlorophyll fluorescence parameter Fv/Fm was carried out. The inversion models of Fv/Fm by using statistical regression and machine learning algorithms were constructed and evaluated by R2 and RMSE. The best estimation model of Fv/Fm was then applied to the UAV image and the spatial distribution of Fv/Fm was achieved. The results showed that the Fv/Fm was highly correlated with vegetation indices of D450/D756, D612D849 and (D517D851)/(D517+D851). Among statistical regression models, the model accuracy based on Fv/Fm and vegetation index (D517D851)/(D517+D851) was the highest (R2=0.606, RMSE=0.006), and the R2 and RMSE between predicted and measured values were 0.739 and 0.724, respectively. Among machine learning models, the model accuracy based on Fv/Fm and vegetation index (D612D849) was the highest. The R2 and RMSE of inversion model based on BP neural network algorithm were 0.821 and 0.261, respectively, while the R2 and RMSE of inversion model based on the PSO-DELM algorithm were 0.982 and 0.140, respectively. This study provides not only an important reference for estimating the chlorophyll-fluorescence parameters by using the UAV hyperspectral images, but also the technical support for estimating the chlorophyll-fluorescence parameters at the macro scale.

     

/

返回文章
返回