基于Landsat8影像的Himawari-8叶绿素a空间降尺度研究

新型地球静止气象卫星Himawari-8由于空间分辨率较低,其叶绿素a产品难以满足空间异质性高的近岸海域水质监测要求。为了克服这个限制,基于非线性的随机森林算法,利用陆地资源卫星Landsat8的波段反射率数据和Himawari-8的叶绿素a产品,通过构建降尺度模型,以提高Himawari-8的叶绿素a数据的空间分辨率。结果表明,2个秋季模型和2个冬季模型的模型决定系数(R²)分别达到0.6、0.72、0.71和0.85;均方根误差(RMSE)为别为1.47,1.05,1.89,0.76mg/m³。通过实测站点数据对比分析表明,降尺度模型生成的叶绿素a与葵花叶绿素a数据具有较高的一致性,R²达到了0.81,能较好的反映近岸海域叶绿素a浓度的空间变化特征。

Spatial downscaling of chlorophyll A in Himawari-8 based on Landsat 8 images

The chlorophyll-a products of the new geostationary meteorological satellite Himawari-8 are difficult to meet the requirements of water quality monitoring in near-shore waters with high spatial heterogeneity due to their low spatial resolution. To overcome this limitation, a non-linear random forest algorithm was used to improve the spatial resolution of the chlorophyll a data from Himawari-8 by constructing a downscaling model using the band reflectance data from Landsat 8and the chlorophyll-a products from Himawari-8. The results showed that the coefficients of determination (R²) of the two autumn models and two winter models reached 0.6, 0.72, 0.71 and 0.85, respectively, and the root mean square errors (RMSE) were 1.47, 1.05, 1.89, 0.76mg/m³, respectively. The comparative analysis of the measured site data showed that the chlorophyll-a data generated by the downscaled model had a high consistency with the chlorophyll-a data of Himawari-8, and the R² reached 0.81 The spatial variation of chlorophyll-a concentration in the near-shore sea area is well reflected by the spatial variation of chlorophyll a data.