基于实测光谱的杭州湾悬浮物浓度遥感反演模式

悬浮物是我国近海水质遥感监测的主要参数之一.利用ASD地物光谱仪测量杭州湾水体的反射光谱，同步采集表层水样获取悬浮物浓度，模拟水色卫星MODIS和MERIS的波段设置提取遥感反射率，基于人工神经网络分别建立2种悬浮物浓度的遥感反演模式.实测结果表明，杭州湾水体中悬浮物浓度较高，泥沙含量远远大于叶绿素浓度，平均值分别为705 mg/L和1.164 mg/m³；2个监测站位的悬浮物浓度表现出时空上的差异，尤其是随潮汐变化的短周期变异十分显著；实测水体反射率光谱曲线在650～700 nm之间与800 nm附近分别出现峰值，光谱曲线一阶微分结果显示第一反射峰的位置随着悬浮物浓度增大向长波方向移动.基于神经网络建立的模拟算法充分利用了卫星不同通道的光谱信息，可以同时模拟水体中色素与非色素悬浮物的浓度，模型取得了较好的拟合效果，R²均大于0.95，可以应用于实际卫星遥感反演，尤其是MERIS数据，因其具有相对较高的空间分辨率，在近海水环境遥感监测中的应用前景更为广阔.

Remote Sensing Inversion Mode of Suspended Particles Concentration in Hangzhou Bay Based on in situ Measurement Spectrum

Suspended particles are one of major parameters of coastal water color remote sensing in China. The reflectances spectral of water were measured using an ASD field spectroscope, synchronously the suspended particles concentrations of surface water were acquired in Hangzhou Bay. Two remote sensing inversion models for suspended particles concentrations (SPC) were developed based on the simulated reflectance of MODIS & MERIS channels using artificial neural network (ANN) algorithm. Measurement results show that the total SPC of Hangzhou bay is comparatively high where the suspended sediments concentrations (SSC) are far more than chlorophyll concentrations, averagely 705 mg/L and 1.164 mg/m³, reseparately. The SPC in two measurement stations appears spatio-temporal variation, especially the short period change due to tidal cycle. There are two reflectance peaks in the measured spectral curves, one is between 650 nm to 750 nm, and the other is near 800 nm. The first order derivative curves of reflectance spectral indicate that the first reflectance peaks shift to long wavelength according to the increased SSC. The ANN models make full use of the spectral information in different channels which can simulate the pigment and non-pigment particles concentrations at same time. The fitting effects are preferable with R² more than 0.95 for each model. The ANN mode can be used for satellite remote sensing inversion, especially MERIS data, because of its comparatively high spatial resolution.