Acta Optica Sinica, Volume. 45, Issue 12, 1201001(2025)

Subsurface CDOM Vertical Distribution Inversion in Ocean Based on Multi-Source Remote Sensing Data

Beibei Xie1,2、*, Wang Gao1, Yingjie Wang1, Kaijie Ma1, and Xiuzhi Kong3
Author Affiliations
  • 1School of Information Science and Engineering, Yanshan University, Qinhuangdao 066004, Hebei , China
  • 2Key Laboratory for Computer Virtual Technology and System Integration of Hebei Province, Qinhuangdao 066004, Hebei , China
  • 3Guoneng Zhishen Control Technology Co., Ltd., Beijing 102211, China
  • show less
    Figures & Tables(18)
    Multi-source data processing flow
    Inversion model structure for vertical distribution of CDOM in the ocean subsurface
    Effect of BGC-Argo data before and after filtering. (a) Vertical distribution of CDOM mass concentration at subsurface before filtering; (b) vertical distribution of CDOM mass concentration at subsurface after filtering
    Initial Rrs443 remote sensing data for the target area on the 1st of each month in 2020
    Rrs443 remote sensing data on the 1st of each month in 2020 after target area reconstruction
    Loss during training
    Independent testing area
    Vertical distribution of subsurface CDOM mass concentration in independent testing area A. (a) Actual measured result; (b) predicted result
    Vertical distribution of subsurface CDOM mass concentration in independent testing area B. (a) Actual measured result; (b) predicted result
    Residual scatter plots of true and predicted values. (a) Residual plot of the test set; (b) residual plot of independent testing area A; (c) residual plot of independent testing area B
    CDOM-SST scatter plots of independent testing area. (a) CDOM-SST scatter plot in area A; (b) CDOM-SST scatter plot in region B
    • Table 1. BGC-Argo buoy information statistics

      View table

      Table 1. BGC-Argo buoy information statistics

      BGC-Argo buoy numberData collection timeSelect CDOM profile depth /mNumber of collected files
      29027502018-09-11‒2019-05-310‒200226
      29027542018-08-30‒2023-11-23339
      29027552018-09-03‒2021-02-10289
      29028572022-07-05‒2023-03-10235
      29033292018-01-28‒2018-04-2082
      29033302018-01-28‒2018-04-2082
      29033932020-08-22‒2022-09-26150
      29033942019-05-26‒2022-04-07218
      29033952019-06-03‒2021-06-24167
      29033962019-08-26‒2020-05-2762
      29036722021-03-05‒2022-10-2951
      29028872023-06-06‒2023-11-3042
      Total--1943
    • Table 2. Remote sensing data information

      View table

      Table 2. Remote sensing data information

      Remote sensing parameterData collection timeTemporal resolutionSpatial resolutionNumber
      B1 (Rrs412)2018-01-01‒2023-12-311 d0.04°2190
      B2 (Rrs443)2190
      B3 (Rrs490)2190
      B4 (Rrs510)2190
      B5 (Rrs560)2190
      B6 (Rrs665)2190
      SST0.25°2190
      Total---15330
    • Table 3. Results of K-fold cross-validation

      View table

      Table 3. Results of K-fold cross-validation

      Data collection timeMAPE /%
      January 1, 202025.77
      February 1, 202018.51
      March 1, 202018.13
      April 1, 202019.49
      May 1, 202021.52
      June 1, 202022.57
      July 1, 202020.35
      August 1, 202016.53
      September 1, 202017.11
      October 1, 202029.04
      November 1, 202017.27
      December 1, 202022.88
    • Table 4. Dataset partition information

      View table

      Table 4. Dataset partition information

      Data typeNumber of samples
      Total1895
      Training set1400
      Test set350
      Independent testing area A72
      Independent testing area B73
    • Table 5. Model input and output parameters

      View table

      Table 5. Model input and output parameters

      NumberInput variableInput parameterNumberOutput variableOutput parameter
      1B1 (Rrs412)

      Target area: 1°×1° grid with

      25×25 data points

      8CDOM

      CDOM profile at center point

      (0‒200 m, 20 levels)

      2B2 (Rrs443)
      3B3 (Rrs490)
      4B4 (Rrs510)
      5B5 (Rrs560)
      6B6 (Rrs665)
      7SST
    • Table 6. Parameters of sensitivity analysis model

      View table

      Table 6. Parameters of sensitivity analysis model

      ModelInput parameter
      Model 1SST, Rrs (B1‒B6)
      Model 2Rrs (B1‒B6)
      Model 3SST
    • Table 7. Model evaluation results

      View table

      Table 7. Model evaluation results

      Test dataRMSE /(μg/L)rR2
      Test set0.140.730.74
      Independent testing area A0.130.810.79
      Independent testing area B0.180.740.69
    Tools

    Get Citation

    Copy Citation Text

    Beibei Xie, Wang Gao, Yingjie Wang, Kaijie Ma, Xiuzhi Kong. Subsurface CDOM Vertical Distribution Inversion in Ocean Based on Multi-Source Remote Sensing Data[J]. Acta Optica Sinica, 2025, 45(12): 1201001

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Atmospheric Optics and Oceanic Optics

    Received: Jul. 28, 2024

    Accepted: Nov. 24, 2024

    Published Online: Mar. 26, 2025

    The Author Email: Beibei Xie (beibeixie@ysu.edu.cn)

    DOI:10.3788/AOS241370

    CSTR:32393.14.AOS241370

    Topics