Hyperspectral Feature And Mercury Content Inversion Of Reed Leaves Under Mercury Pollution

Mercury is a kind of environmental pollutant with serious biological toxicity.The environment of accumulated water and abundant organic matter in wetland provide the conditions for the accumulation of mercury,and wetland plants can absorb mercury from various ways such as water,soil and air.Therefore,understanding the impact of heavy metal mercury on plant leaves,and seeking a rapid and accurate method to estimate mercury pollution in wetland plants has become an important way to monitor wetland ecosystem.In this study,the wetland plant reed?Phragmites communis?was the research object.Field sampling and control experiments were conducted to analyze the changes in total mercury and chlorophyll content of reed leaves under different levels of mercury pollution,and combined with hyperspectral techniques to study the spectral reflectance characteristics variation of reed leaves under mercury pollution.Mercury contamination information which hidden in spectral information were extracted,and models between the total mercury content of reed leaves and spectral data were established to invert the total mercury content of reed leaves.Our results showed that:?1?With the increase of mercury treatment in soil,the total mercury content of reed leaf gradually increased,the chlorophyll ? content gradually decreased,the chlorophyll ? and total chlorophy content first increased and then decreased.Chlorophyll ? was more sensitive to mercury than chlorophyll ? and total chlorophyll.The correlations between the total mercury content in water,air and the total mercury,chlorophyll content in reed leaves were the same as that in the soil.However,the influence degree of each environmental medium was soil ? water > air.At the same time,with the increase of mercury treatment time,the total mercury content of reed leaf gradually increased,the chlorophyll ?,chlorophyll ? and total chlorophyll content showed a decreasing trend,but the correlation coefficient of total mercury content of reed leaf and chlorophyll ?,chlorophyll ?,total chlorophyll gradually increased,and the related significance gradually increased.?2?The sensitive bands of total mercury content in reed leaves were mainly concentrated in the visible light region 600-750 nm and the mid-infrared light region 1500-2300 nm under different mercury treatment levels.This indicated that the effects of mercury on reed leaves were mostly influenced leaves chlorophyll and moisture content.Among them,low concentration(0-1.0 mg·kg^-1)had a significant effect on leaf water content,and high concentration(1.5-2.0 mg·kg^-1)had a greater effect on leaf chlorophyll content.In addition,the red edge position?WP_r?,the green peak reflectance?Rg?,the red valley reflectance?Rr?,the ratio of red edge area and blue edge area?SDr/SDb?,and the normalization of red edge area and blue edge area??SDr-SDb?/?SDr+SDb??and the normalization of red edge area and yellow edge area??SDr-SDy?/?SDr+SDy??was significantly correlated with the total mercury content in reed leaves,and they could effectively respond to heavy metal mercury pollution.?3?The correlation between the wavelet coefficients after continuous wavelet transform and the total mercury content of reed leaves was higher than that of the spectral reflectance without continuous wavelet transform,and the bands with significant significance were also increased but scattered.Among them,21-25 low-scale wavelet decomposition was more conducive to improving the correlation between spectral data and the total mercury content of reed leaf.The differential spectral index?DSI?,the normalized spectral index?NDSI?and the ratio spectral index?RSI?sensitive areas of the total mercury content in the reed leaves were basically the same.Among them,the best correlation with the total mercury content of reed leaves was the combination of 350 or 355 nm and 1715 nm bands under the control experiment,while in the field experiments was the combination of 350 nm and 520 nm bands.?4?The inversion models of total mercury content of reed leaves established by sensitive bands which selected by spectral transform techniques,the fitting effect and precision of calibration set and validation set of models constructed by the continuous wavelet transform spectral data were generally better than the models constructed by spectral data without continuous wavelet transform.Among them,the NR-CWT-RF model?R²=0.8560,RMSE=0.1062?and the CR-CWT-SMLR model?R²=0.8170,RMSE=0.0173?were the optimal inversion models for total mercury of reed leaves based on the spectral conversion technique in control experiment and field experiment,respectively.The inversion effects of the model constructed using the spectral index were varied.Among them,the NDSI-RF model?R²=0.8611,RMSE=0.0968?under control experiment and the DSI-PLSR model?R²=0.9582,RMSE=0.0195?under field experiment were the optimal inversion models.The inversion of the total mercury content in the reed leaves showed a trend of decreasing with the increase of total mercury content in soil,and the inversion of mercury content in reed leaves by hyperspectral data is more stable and reliable when the total mercury content in soil was less than 1.0 mg·kg^-1.At the same time,it is better to retrieve the total mercury content of reed leaves by using the SMLR model established by the spectral data after the wavelet transform,both in control and field experiments.