Sentinel-2 MSI 和 Landsat 8 OLI 数据在玉米秸秆覆盖度遥感估算应用中的比较研究

Comparison between Sentinel-2 MSI and Landsat 8 OLI data in the application of remote sensing estimation for crop residue cover

    摘要
  • 摘要:

    秸秆覆盖度（Crop residue cover, CRC）的遥感估算可以在短时间内获取大范围耕地秸秆覆盖度数据，对于政府部门监测保护性耕作的实施情况有重要的现实意义。本研究基于Sentinel-2 MSI和Landsat 8 OLI数据，分别计算了多种光谱指数，并与野外实测的秸秆覆盖度数据进行相关性分析，挑选出极显著性相关的光谱指数。在此基础上，构建其与秸秆覆盖度之间的相关模型，并通过决定系数（R2）和均方根误差（RMSE）所表征模型的精度比较Sentinel-2 MSI和Landsat 8 OLI数据由于光谱和空间尺度的差异对秸秆覆盖度反演模型的影响。结果表明：6种光谱指数与CRC的相关性系数均大于0.4，相关性较高的是Sentinel-2 MSI 20 m分辨率数据获取的NDTI和STI，相关系数分别为0.878、0.894，相关性最低的为Sentinel-2 MSI 10 m分辨率数据获取的NDSVI，相关系数为0.476；利用一元线性回归法构建模型时，Sentinel-2 MSI 20 m分辨率数据构建的光谱指数STI和NDTI，模型精度最高，R2分别为0.810和 0.800，RMSE分别为6.84%和7.01%，而30 m重采样数据的R2分别为0.770和0.771，RMSE分别为7.52%和7.50%，随着空间分辨率的降低呈现出下降趋势；Sentinel-2 MSI 30 m重采样数据获取的光谱指数构建的所有模型精度均略大于Landsat 8 OLI数据构建的模型。因此，Sentinel-2 MSI 数据获取NDTI和STI这两个光谱指数更加适合本研究区域秸秆覆盖度的估算。

     

    Abstract:

    The remote sensing estimation of crop residue cover (CRC) can quickly obtain the information of straw cover returning to the field in a large range of cultivated land, which has important and practical significance for the government departments to carry out conservation tillage planning and promote straw returning to the field. Based on sentinel-2 MSI and Landsat 8 OLI data, this study calculated a variety of spectral indexes, analyzed the correlation with the field measured straw coverage data, and selected the spectral indexes with most significant correlation. Then constructed the correlation model between index and straw coverage, and compared the accuracy of the model characterized by the coefficient of determination (R2) and root mean square error (RMSE) to compare the influence of CRC inversion model caused by the difference of spectral and spatial scales between sentinel-2 MSI and Landsat 8 OLI data. The results showed that the correlation coefficients between the six spectral indices and CRC were greater than 0.4. The higher correlation was NDTI and STI obtained from sentinel-2 MSI 20 m resolution data, and the correlation coefficients were 0.878 and 0.894, respectively. The lowest correlation was NDSVI obtained from sentinel-2 MSI 10 m resolution data, and the correlation coefficient was 0.476. When using univariate linear regression method to construct the model, the spectral indices STI and NDTI constructed from sentinel-2 MSI 20 m resolution data had the highest model accuracy with R2 of 0.810 and 0.800, RMSE of 6.84% and 7.01%; while R2 of 0.770 and 0.771, RMSE of 7.52% and 7.50% in 30 m resampled data, both showing a downward trend with the reduction of spatial resolution. The precision of all models constructed by the spectral index obtained from sentinel-2 MSI 30 m resampled data was slightly higher than that constructed by Landsat 8 OLI data. Therefore, the spectral indexes NDTI and STI obtained from sentinel-2 MSI data would be more suitable for the estimation of straw coverage in the study area.

     

    • HTML全文
    • 参考文献(36)
    • 相关文章
    • 施引文献
  • 资源附件(0)

/

返回文章
返回