Studies On Models For Remotely Estimating Total Suspended Matter Concentration In Xin’anjiang Reservoir

Total suspended matter(TSM)plays an important role in determining the underwater light climate and further affecting lake primary production.Therefore,TSM concentration is an important parameter of lake water quality and water environment assessment.This study develops an empirical estimation model and a semi-analytical model based on in situ hyperspectral data and the matching Landsat 8 OLI data after 6S atmospheric correction for the clean Xin’anjiang Reservoir.The results show:(1)Combined with MODIS atmospheric observation data,we proceed the Landsat 8 OLI data of Xin’anjiang Reservoir for atmospheric correction by using 6S atmospheric radiative transfer model,we also compare the reflectance data after 6S atmospheric radiative transfer model and in site measured reflectance data as well as reflectance product after USGS operational algorithm,it showes that 6S atmospheric radiative transfer model with MODIS atmospheric observation data gives a higher precision at 542 nm,600 nm and 668 nm,the determination coefficient between reflectance data after 6S atmospheric radiative transfer model and in site measured reflectance data are 0.71,0.70 and 0.68,respectively.The products after 6S atmospheric correction can be better used for estimating the TSM concentration for the clean Xin’anjiang Reservoir.(2)Among all the single band and multiband algorithms based on measured hyperspectral remote sensing reflectance,the linear combination of two bands(a×Rrs(λ1)±b×Rrs(λ2)+c)at spectral range of 550 nm-700 nm gets better correlation with determination coefficient above 0.8.(3)Among all the single band and multiband algorithms based on landsat 8 OLI products after 6S atmospheric correction,the linear models using Band 2,Band 3 and Band 8 can give a reasonable and satisfied estimation accuracy.The determination coefficient,mean relative error and root mean square error are 0.92,0.16 mg L"1 and 11.49%,respectively for the three-band combination model.(4)The overall TSM concentration of Xin’anjiang Reservoir is low,resulting in unobvious variation in the reflectance in region 700-750 nm(near infra-red,NIR),and it keeps consistently lower than the reflectance in the blue,green and red ranges.Thus,it is difficult to estimate the TSM concentration by calculating the backscattering of particulate matter based in the region of NIR reflectance.It is found by analysis that there was a significant positive correlation between the TSM concentration and ap(λ),with the highest determination coefficient(R2>0.9)at 542 nm.Through the transition of ap(542),we built a semi-analytical model between the TSM concentration and reflectance of Landsat 8 OLI images.(5)We calibrated and validated this semi-analytical model and characterized the temporal-spatial variations of TSM concentrations in Xin’anjiang Reservoir.The TSM concentration exhibited a significant seasonal difference,with significantly higher TSM concentrations in autumn and summer(mean TSM of 2.94±3.03 mg L-1 and 2.57±1.60 mg L-1,respectively)than in winter and spring(mean TSM of 1.86±2.91 mg L-1 and 1.72±2.00 mg L-1,respectively).In addition,the TSM concentration exhibited a significant spatial difference in the Xin’anjiang Reservoir,ranging from 0.002 to 18.50 mg L-1 with a mean of 2.37±2.05 mg L-1.The temporal heterogeneity of the TSM concentration was mainly caused by the seasonal rainfall and seasonal growth of phytoplankton.The spatial heterogeneity is derived from watershed inputs and dredging activity.Our study shows that the new semi-analytical model for the Landsat 8 OLI imagery database can be successfully used for the quantitative monitoring of the TSM concentration in the slightly turbid Xin’anjiang Reservoir.