基于MODIS模拟的辽东湾叶绿素a的遥感反演模型

叶绿素a浓度对于水质和富营养化评价、赤潮探测及海洋生产力研究有重要意义.利用卫星遥感反演获取叶绿素a浓度弥补了传统采样方法的数据离散化、耗时长且费用高的缺陷,可以实时和全面地观测大尺度海洋叶绿索a分布.通过实测光谱模拟卫星波段,可以在一定程度上削除大气干扰,有助于准确地确定特征波段和建立遥感反演模型.本文以辽东湾为研究区,利用现场实测光谱模拟了MODIS波段,并在光谱特征分析基础上进行叶绿素a的特征波段选择,通过比较单波段和波段比值法等算法,建立了2个具有区域特色的叶绿素a遥感反演模型,并进行了模型验证.结果表明:在辽东湾海域,运用红光附近特征波段667和678 nm来反映叶绿素特征比传统蓝绿光波段(440和550nm)似乎更能削弱其它水体组分的干扰,突出辽东湾水体的特点.利用667、678和551nm的三波段模型要比单纯考虑667和678nm波段比值的模型要好一些,该模型r为0.856,标准差为0.190.模型验证预测值和实测值的相关系数为0.93(p<0.01),表明了模型良好的预测能力.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.

Remotely sensed retrieval model of chlorophyll a in Liaodong Gulf based on MODIS band simulation

Chlorophyll a is important for water quality and eutrophication assessment, red tide detection as well as ocean primary production. Retrieving chlorophyll α by remote sensing can compensate for the deficiency of the traditional time-consume sampling method which acquire discrete data at high cost, and it provides information of ocean chlorophyll a nearly in real-time at large-scale. This paper simulate satellite band using field spectrum to remove atmospheric impact. It is helpful for the identification of characteristic bands and development of remote sensing retrieval model. Here Liaodong Gulf was selected as study area, Based on the field spectral data MODIS bands were simulated for identification of characteristic bands of chlorophyll a, We developed 2 area-specific remotely sensed retrieval models for chlorophyll α by comparison of single band and band ratio algorithms based on spectral feature analysism, result of the model shows that in Liaodong Gulf the 667 and 678 nm bands work better than traditional blue and green bands of 440 and 550 nm in revealing chlorophyll a features. They seemed to weaken the impact of other water constituent better and reveal the water property in Liaodong Gulf. The retrieval model based on combination of 667, 678 and 551 nm bands is better than the two-band model of 667 and 678 nm bands. The correlation coefficient r of the model is 0.856 and standard error is 0.19. In model test the r between the derived value and the in situ value is 0.93(p<0.01), which show good performance of forecast. The model was area-specific and it needs further tested to identify whether it can be applied to other areas.