资源一号02D高光谱影像内陆水体叶绿素a浓度反演

2019-09-12成功发射的资源一号02D卫星(ZY-102D)搭载了新一代可见短波红外高光谱相机AHSI(Advanced Hyperspectral Imager),其丰富的细分波段和较高的空间分辨率在内陆湖库水质监测方面具有较大潜力,但数据可用性有待分析和验证。本研究以中国华东和华北平原的典型富营养湖库(太湖、于桥水库)和中营养湖库(小浪底水库)为研究区,开展基于ZY-102D高光谱影像的叶绿素a浓度反演研究。以3个研究区在46个采样点地面测量的叶绿素a浓度和同步获取影像的遥感反射率作为数据源,基于5种典型的叶绿素a半经验模型进行了模型参数优化和叶绿素a反演精度验证。结果表明,基于中心波长为705 nm和671 nm的波段比值模型叶绿素a反演精度最高,模型R²为0.78,平均无偏相对误差(AURE)和均方根误差(RMSE)分别为13.5%和4.5 mg/m³。研究表明,ZY-102D高光谱数据在内陆水体叶绿素a浓度高精度反演方面具有重要潜力,但未来需要通过多星组网提升监测能力,以及发展针对于ZY-102D水体图像的降噪和大气校正方法。

Inland water chlorophyll-a retrieval based on ZY-1 02D satellite hyperspectral observations

China’s ZY-102D satellite was successfully launched on September 12,2019.It carries the new-generation Advanced Hyperspectral Imager(AHSI),which has 166 bands in the visible to short-wave infrared bands.AHSI can acquire images at 30 m spatial resolution with a 60 km swath.ZY-102D satellite shows great potential for inland water quality monitoring application,owing to its abundant narrow bands and relatively high spatial resolution.However,this satellite has been launched for a short period,and the applicability of this data needs to be further analyzed and tested.Taihu Lake(eutrophic),Yuqiao reservoir(eutrophic),and Xiaolangdi Reservoir(mesotrophic)in China were used as study areas for the Chlorophyll-a(Chla)retrieval based on the ZY-102D hyperspectral images.Within one day of the ZY-102D satellite overpass,in situ spectra,and Chla concentrations were collected at sampling sites in these study areas.We selected five typical Chla semi-empirical models based on spectral indices,namely,Band Ratio(BR),Normalized Difference Chlorophyll Index(NDCI),Three-Band Index(TBI),Enhanced Three-Band Index(ETBI),and the Baseline Height(BH).We used in situ measured Chla concentration at 46 sampling sites in the three study areas and simultaneously acquired ZY-102D images to optimize the parameters in these models.We evaluated the accuracies of image-derived R_rs at sampling sites,and then conducted accuracy analysis for estimated Chla concentrations using optimized empirical models.ZY-102D image-derived R_rs were consistent with in situ measured R_rs in the 671 and 705 nm,whereas the 731 and 748 nm band R_rs had greater uncertainties because they were more likely to be affected by the image noise.In addition,the accuracy analysis for the estimated Chla concentrations shows that the model based on the 705 to 671 nm band ratio achieves the highest accuracy,with an R² of 0.78.In addition,the mean unbiased relative error(AURE)and Root Mean Square Error(RMSE)are 13.5% and 4.5 mg/m³,respectively.On the contrary,models based on the ETBI and BH yield Chla concentration estimates with low accuracies.In conclusion,ZY-102D hyperspectral data show good potential in terms of accurate retrieval of Chla concentration for inland waters.We plan to conduct more in situ experiment when the ZY-102D satellite overpasses to improve the Chla concentration retrieval model applied on the ZY-102D data.In the future,the monitoring capacity should be improved through establishing a hyperspectral satellite constellation,and noise reduction and atmospheric correction methods should be developed for ZY-102D’s inland water application.