Estimation And Analysis Of Total Suspended Matter By Remote Sensing In The Lower Reaches Of Minjiang

Compared with the traditional water quality monitoring method,remote sensing water quality monitoring has the advantages of fast speed,wide range and low cost.Remote sensing technology can better monitor the temporal and spatial variation of water quality parameters in the study area.The total suspended matter(TSM),one of the important parameters of water quality,is commonly used to evaluate the quality of water.Taking the lower reaches of Minjiang as the study area,the measured water quality data and spectral data,the GF-1 WFV image and the Landsat 8 image as the data sources from 2015 to 2017,the PSO-RBF neural network model is applied to estimate and analyze the total suspended matter by remote sensing.The main research contents and conclusions are as follows:(1)Construction and analysis of the inversion model of total suspended matter in the lower reaches of Minjiang.Using the measured spectrum simulation of GF-1 WFV and Landsat 8 bands,together with GF-1 WFV1 and Landsat 8 bands reflectance data to analysize the correlation between remote sensing factors and TSM.Hyperspectral inversion for TSM and the sensitive band combination with R468,8656 and R656/R468,GF-1 sensitive band combination with b3,b3/b2 and b3/b1,Landsat 8 sensitive band combination with b4,b4/b3 and b4/b2.The use of sensitive bands and combinations were established hyperspectral,GF-1 WFV and Landsat 8 statistical model,BP neural network model and PSO-RBF neural network model.From the modeling accuracy and error analysis,the PSO-RBF neural network model is the optimal inversion model of the same kind of sensor.(2)The results and analysis of suspended substance concentration in the lower reaches of Minjiang.Based on the well-trained PSO-RBF neural network model and BP neural network model,the TSM in the lower reaches of Minjiang is retrieved by GF-1WFV and Landsat 8 remote sensing images.The result of TSM inversion in PSO-RBF model better than the BP neural network model,and remote sensing image retrieval results accuracy was significantly higher than that of Kriging interpolation results.The distribution of TSM in the lower reaches of Minjiang,the overall performance of West East high low,especially from Mawei to the entrance to the sea of Minjiang,low values are concentrated in the upstream of the research area.From 2015 to 2017,the total suspended matter in the lower reaches of Minjiang showed a relatively stable state.Summer suspended matter concentration inversion model is applied to the remote sensing image in autumn the quantitative inversion,results show that the fall of the RMSE is 3 times of the summer synchronous image inversion results of RMSE,the average relative error is almost 6 times.Inversion model of TSM using summer datas for quantitative inversion of remote sensing images in autumn,the results show that RMSE in autumn is 3 times as high as that of RMSE in summer,and the difference of the MRE is nearly 6 times.(3)Spatial variation analysis of TSM in the lower reaches of Minjiang.Spatial variability analysis of TSM in the lower reaches of Minjiang by spatial self-correlation method and spatial heterogeneity method.The results showed that Moran’s I coefficient is greater than 0.5,Z value is greater than 2.58,P value is less than 0.01;the minimum range of variation is 13550m,others are greater than 20000 meters,the D fractal dimension is higher,between 1.31 and 1.85,the C0/0+C is all less than 25%,the largest is only 20%,the smallest is 0.1%,and the C/C0+C is all greater than 0.84.This shows that there is a significant spatial positive correlation in the concentration of suspended matter in Minjiang,and the variation of regional heterogeneity is caused by regional self-correlation.