• Remote Sensing Technology and Application
  • Vol. 39, Issue 1, 98 (2024)
Jiao WANG*, Wei LI, Weiquan ZHAO, Zulun ZHAO..., Liang HUANG and Jiafang YANG|Show fewer author(s)
Author Affiliations
  • Institute of Mountain Resources,Guiyang 550001,China
  • show less
    DOI: 10.11873/j.issn.1004-0323.2024.1.0098 Cite this Article
    Jiao WANG, Wei LI, Weiquan ZHAO, Zulun ZHAO, Liang HUANG, Jiafang YANG. Retrieving CODMn Concentration in Karst Plateau Deep Lake Reservoir Using Sentinel-2 Data[J]. Remote Sensing Technology and Application, 2024, 39(1): 98 Copy Citation Text show less
    Spatial distribution of study area and in-situ sites in “Liang Hu”
    Fig. 1. Spatial distribution of study area and in-situ sites in “Liang Hu”
    Remote sensing reflectance of MSI bands corresponding to CODMn in-situ locations
    Fig. 2. Remote sensing reflectance of MSI bands corresponding to CODMn in-situ locations
    Flowchart for CODMn concentration estimation of Karst Plateau Deep Lake Reservoir using Machine Learning methods
    Fig. 3. Flowchart for CODMn concentration estimation of Karst Plateau Deep Lake Reservoir using Machine Learning methods
    Comparison of "Two Lakes" based on Sentinel-2 data's prediction results and actual measured values (training set)
    Fig. 4. Comparison of "Two Lakes" based on Sentinel-2 data's prediction results and actual measured values (training set)
    Comparison of "Two Lakes" based on Sentinel-2 data's prediction results and actual measured values (test set)
    Fig. 5. Comparison of "Two Lakes" based on Sentinel-2 data's prediction results and actual measured values (test set)
    "Two lakes" CODMn spatiotemporal distribution
    Fig. 6. "Two lakes" CODMn spatiotemporal distribution
    波段分辨率Sentinel-2A/nmSentinel-2B/nm意义
    /m中心波长带宽中心波长带宽
    B160443.927442.345监测近岸水体/大气气溶胶
    B210496.698492.198蓝、绿、红可见光波段
    B310560.04555946
    B410664.53866539
    B520703.919703.820红边监测植被健康信息
    B620740.218739.118
    B720782.528779.728
    B810835.1145833133近红外波段(宽)
    B8a20864.83386432近红外波段(窄)
    B960945.026943.227水蒸气波段
    B10601 373.5751 376. 976卷云波段(短波红外)
    B11201 613.71431 610.4141短波红外波段
    B12202 202.42422 185.7238
    Table 1. The major specifications of Sentinel-2 MSI
    采样日期成像日期Sentinel-2 (A/B)(.SAFE)
    2018/4/122018/4/9S2A_MSIL1C_20180409T032541_N0206_R018_T48RXQ_20180409T070457
    2018/10/282018/10/31S2A_MSIL1C_20191031T032851_N0208_R018_T48RXQ_20191031T062949
    2019/4/102019/4/9S2B_MSIL1C_20190409T032539_N0207_R018_T48RXQ_20190409T075904
    2019/8/122019/8/17S2B_MSIL1C_20190817T032539_N0208_R018_T48RXQ_20190817T081113
    2019/10/312019/10/31S2A_MSIL1C_20191031T032851_N0208_R018_T48RXQ_20191031T062949
    2020/4/252020/4/28S2A_MSIL1C_20200428T032541_N0209_R018_T48RXQ_20200428T062927
    2020/8/202020/8/26S2A_MSIL1C_20200826T032541_N0209_R018_T48RXQ_20200826T062908
    2020/11/52020/11/14S2A_MSIL1C_20201114T033011_N0209_R018_T48RXQ_20201114T062024
    Table 2. Sentinel-2 (A and B) image references and sample dates
    方法超参数Grid-Search值CODMn最终模型参数
    RFRn estimators10、20、50、100、20050
    min_samples_leaf1、5、10、151
    min_samples_split2、5、10、15、2510
    max_depth2、3、4、5、6、7、NoneNone
    SVRC1、2、5、10、25、50、120、256256
    gamma24* np.logspace(-6, 6, num=12)0.007 5
    GPRn-restarts-optimizer4、8、10、12、16、20、32、6410
    Table 3. Hyper parameters setting and results of cross-validation with Grid-Search
    波段B1B2B3B4B5B6B7B8A
    相关系数0.580.580.700.540.570.490.490.50
    Table 4. Pearson correlation coefficients between MSI bands and in-situ CODMn
    波段模型训练集(49个)验证集(21个)
    R2RMSE(mg/L)

    MAPE

    /%

    R2RMSE(mg/L)

    MAPE

    /%

    bandSVR0.610.3912.670.570.379.36
    RFR0.930.175.640.840.225.84
    GPR0.570.4113.810.500.3910.24
    logbandSVR0.670.3611.400.690.318.57
    RFR0.930.175.830.800.256.57
    GPR0.590.4013.290.560.3710.21
    Table 5. Training set and validation set inversion results
    Jiao WANG, Wei LI, Weiquan ZHAO, Zulun ZHAO, Liang HUANG, Jiafang YANG. Retrieving CODMn Concentration in Karst Plateau Deep Lake Reservoir Using Sentinel-2 Data[J]. Remote Sensing Technology and Application, 2024, 39(1): 98
    Download Citation