辽东湾海水透明度的遥感估算模型

Remotely Sensed Estimation Model of Ocean Water Transparency in Liaodong Gulf

海水透明度和水质存在很好的相关性,是水质的重要量度之一。利用卫星遥感估算获取透明度可弥补常规采样方法的数据离散化、耗时长且费用高的缺陷,可得到大尺度海洋透明度分布。以辽东湾为研究区,利用现场实测光谱模拟了MODIS波段从而削除大气干扰,在分析MODIS特征波段基础上,通过比较单波段算法和波段比值法等算法,建立了透明度遥感估算模型,并进行了模型验证。结果表明:在辽东湾海域,MODIS 667 nm波段和透明度的相关性最好,而且单波段建立的估算模型要优于其他特征波段的组合。未考虑传统的蓝绿波段的667 nm单波段二次多项式估算模型相关系数r为0.973(p<0.01),检验平均相对误差13%,表明了模型稳定而有良好的预测能力。模型具有明显的区域特色,结果是否具有普适性,还需进一步验证。

As one of the important indicators which describe water visibility and optical property,ocean water transparency has good relationship with water quality.Measurement of ocean water transparency from satellite remotely sensed image can compensate for the deficiencies of discrete data,long time and expensive cost of field measurement.Moreover,it can obtain the large-scale distribution of ocean water transparency.So it can be applied to long term observation of ocean water transparency at large scale.Liaodong Gulf was selected as the study area.Ocean water reflectance of MODIS bands were simulated to reduce the disturbance of atmosphere based on in situ filed measurement spectrum.According to the spectral simulation,characteristic bands of MODIS were indentified for water transparency estimation model.The correlationship was respectively ananlyzed in Liaodong Gulf.Then the estimation model of water transparency using MODIS of the area was developed and the model test was also performed.The result shows that in Liaodong Gulf,the band of 667 nm of MODIS has the best correlation with ocean water transparency.Moreover,the estimation model based on single band performs better than band combine approach such as band ratio.Without the incorporation of the classical bands of blue and green bands for oceancolor remote sensing,the estimation model of two-order polynomial of single band 667 nm has the correlation coefficient r of 0.973（p0.01） and the relative mean error of model test is 13%.It shows the stability and good forecast performance of the model.As the model was developed based on the dataset of area of the experiment,it has distinct local character.It needs further study to test whether it can be applied to other areas.