作者中心
专家审稿
责编办公
编辑办公
Machine Learning-Based Modeling of Wind and Wave Responses in Sentinel-1 SAR Doppler Shifts for Ocean Current Retrieval
-
摘要: 海流在全球气候调节中具有重要作用。合成孔径雷达(SAR)凭借其对海表多普勒频移的观测能力，为高分辨率海流探测提供了有效的数据支撑。然而，SAR多普勒频移含有多种贡献项，要从中准确反演海流，需要进行非地球物理校正，并准确估算风-浪致多普勒频移的贡献。本研究基于Sentinel-1 SAR多普勒频移观测数据，在完成精确的非地球物理校正后，构建了经粒子群算法优化的BPNN与XGBoost模型描述SAR风-浪致多普勒频移与海面风-浪参数间的非线性映射关系，并通过系统对比评估确定了性能更优的模型，实现了海流流速的高精度反演。结果表明，与BPNN模型相比，XGBoost模型在性能上实现了显著提升。XGBoost模型估计的多普勒频移均方根误差(RMSE)约为4.043 Hz，较BPNN模型降低了2.898 Hz；与HYCOM海流相比，XGBoost模型反演海流的RMSE约为0.202 m/s，较BPNN模型降低了0.122 m/s；与HF雷达观测流速对比，XGBoost模型反演海流的RMSE为0.21 m/s，较BPNN降低了16%。本研究为星载SAR海流反演提供了一种更为精确的技术方法。Abstract: Ocean currents play a critical role in global climate regulation. Synthetic aperture radar (SAR) provides high-resolution observational support for ocean current detection by measuring Doppler shifts; however, SAR Doppler shifts contain multiple contributing components. To accurately retrieve ocean currents from these data, nongeophysical contributions must be precisely corrected, and wind- and wave-induced Doppler shifts must be accurately estimated. This paper proposes a machine learning-based method for modeling such shifts and retrieving ocean currents from Sentinel-1 SAR data. First, nongeophysical contributions in the SAR Doppler shift are precisely corrected to remove the effects unrelated to ocean motion. Second, backpropagation neural network (BPNN) and eXtreme gradient boosting (XGBoost) models, optimized using the particle swarm optimization algorithm, are developed to describe the nonlinear relationship between the wind-wave Doppler shift and sea-surface wind-wave parameters derived from SAR data. Finally, the corrected Doppler shift is utilized to retrieve ocean surface current velocities. This paper comparatively evaluates the estimation accuracies of the wind-wave Doppler shifts obtained using the BPNN and XGBoost models, as well as the respective influence of each model's performance on the effectiveness of ocean current retrieval. Results indicate that the XGBoost model achieves superior estimation accuracy compared with the BPNN model. The root mean square error (RMSE) of the Doppler shift estimated by the XGBoost model is approximately 4.043 Hz, which is 2.898 Hz lower than that of the BPNN model. Compared with those of the HYCOM current data, the RMSE of the currents retrieved by the XGBoost model is about 0.202 m/s; this value is reduced by 0.122 m/s compared with that of the BPNN model. Validations against the current velocities detected by HF radar show that the RMSE of currents retrieved by the XGBoost model is 0.21 m/s, representing a 16% reduction compared with that of the BPNN model. These findings indicate that the proposed technical approach for ocean current retrieval using spaceborne SAR is highly accurate.
-
图 1 获取的Sentinel-1A IW模式OCN数据产品的空间分布图
Figure 1. Spatial distribution of Sentinel-1A IW OCN data used in this study
图 7 2019年1月10日12:26获取的Sentinel-1A干涉宽幅(IW)陆地场景的多普勒频移图
Figure 7. Doppler shift map derived from the Sentinel-1A Interferometric Wide (IW) swath acquired over land on January 10, 2019 at 12:26 UTC.
图 9 去除扇形误差${f_{{sca} }}$的过程图
Figure 9. Geometric correction workflow for fan distortion${f_{{sca} }}$
图 10 $ {f}_{\rm{elec}} $+$ {f}_{{\mathrm{others}}} $与入射角的线性关系
Figure 10. Linear fitting of $ {f}_{\rm{elec}} $ + $ {f}_{{\mathrm{others}}} $ with incident angle
图 11 2019年1月22日23:19获取的海面SAR图像中子带1的多普勒频移
Figure 11. Doppler frequency shift of swath 1 in SAR image obtained from sea surface at 23:19 on January 22nd, 2019
图 12 BPNN和XGBoost模型预测风-浪致多普勒频移与SAR观测值的对比散点图
Figure 12. Comparison of wind-wave-induced Doppler shift predictions (BPNN vs. XGBoost) with SAR observations
图 13 不同风速区间BPNN预测风浪多普勒频移与SAR观测值的对比散点图
Figure 13. Comparison of BPNN-predicted wind-wave Doppler shift vs. SAR observations across different wind speed ranges
图 14 不同风速区间XGBoost预测风浪多普勒与SAR观测值的对比散点图
Figure 14. Comparison of XGBoost-predicted wind-wave Doppler shift vs. SAR observations across different wind speed ranges
图 15 BPNN和XGBoost模型预测风浪多普勒频移与SAR观测值之间偏差的误差棒图
Figure 15. Error bar plots of deviations between BPNN/XGBoost predicted wind-wave Doppler shift and SAR observations
图 16 测试集上基于BPNN和XGBoost模型反演海流与HYCOM海流的对比散点图
Figure 16. Scatter plot of retrieved ocean current velocitiesbased on BPNN and XGBoost versus HYCOM currect from the test dataset
图 17 BPNN和XGBoost模型测试集上反演海流与HYCOM海流之间偏差的误差棒图
Figure 17. Error bar plots of deviations between predicted ocean current velocity (from BPNN and XGBoost models) and HYCOM data in the validation region
图 19 在个例上反演海流流速与HYCOM海流流速的对比散点图
Figure 19. Scatter plot of the comparison between the inverted ocean current velocity and the HYCOM ocean current velocity in the individual case
图 20 湾流区域BPNN和XGBoost模型反演海流的空间分布特征比较
Figure 20. Comparison of current characteristics retrieved by BPNN and XGBoost models in the Gulf Stream region
图 21 黑潮区域BPNN和XGBoost模型反演海流的空间分布特征比较
Figure 21. Comparison of current characteristics retrieved by BPNN and XGBoost models in the Kuroshio Current region
图 22 反演海流流速与HF海流流速的对比散点图
Figure 22. Scatter plot comparing retrieved ocean current velocities with HF radar-derived current velocities
表 1 模型超参数设置
Table 1. Model hyperparameter settings
模型 超参数 BPNN Layer1_units:16, layer2_units:8, learning_rate:0.01,
batch:512, epochs:100XGBoost Booster:gbtree, n_estimators:152,
learning_rate:0.04, max_depth:7
min_child_weight:10
下载: 导出CSV
-
[1] MOISEEV A, JOHNSEN H, HANSEN W M, et al. Evaluation of radial ocean surface currents derived from Sentinel‐1 IW doppler shift using coastal radar and Lagrangian surface drifter observations[J]. Journal of Geophysical Research: Oceans, 2020, 125(4): e2019JC015743. doi: 10.1029/2019JC015743. [2] 何宜军, 刘保昌, 张彪, 等. 海面流场卫星遥感方法综述[J]. 广西科学, 2015, 22(3): 294–300. doi: 10.13656/j.cnki.gxkx.2015.03.005.HE Yijun, LIU Baochang, ZHANG Biao, et al. Overview on satellite remote-sensing methods for sea-surface-current measurement[J]. Guangxi Sciences, 2015, 22(3): 294–300. doi: 10.13656/j.cnki.gxkx.2015.03.005. [3] POTIN P, ROSICH B, MIRANDA N, et al. Sentinel-1 mission status[C]. 2017 IEEE International Geoscience and Remote Sensing Symposium, Fort Worth, USA, 2017: 5525–5528. doi: 10.1109/IGARSS.2017.8128255. [4] ROMEISER R, UFERMANN S, and ALPERS W. Remote sensing of oceanic current features by synthetic aperture radar—achievements and perspectives[J]. Annales Des Télécommunications, 2001, 56(11): 661–671. doi: 10.1007/BF02995560. [5] CHAPRON B, COLLARD F, and ARDHUIN F. Direct measurements of ocean surface velocity from space: Interpretation and validation[J]. Journal of Geophysical Research: Oceans, 2005, 110(C7): C07008. doi: 10.1029/2004JC002809. [6] JOHANNESSEN J A, CHAPRON B, COLLARD F, et al. Direct ocean surface velocity measurements from space: Improved quantitative interpretation of Envisat ASAR observations[J]. Geophysical Research Letters, 2008, 35(22): L22608. doi: 10.1029/2008GL035709. [7] MOUCHE A A, COLLARD F, CHAPRON B, et al. On the use of Doppler shift for sea surface wind retrieval from SAR[J]. IEEE Transactions on Geoscience and Remote Sensing, 2012, 50(7): 2901–2909. doi: 10.1109/TGRS.2011.2174998. [8] HANSEN M, COLLARD F, DAGESTAD K, et al. Retrieval of sea surface range velocities from Envisat ASAR Doppler centroid measurements[J]. IEEE Transactions on Geoscience and Remote Sensing, 2011, 49(10): 3582–3592. doi: 10.1109/TGRS.2011.2153864. [9] ELYOUNCHA A, ERIKSSON L E B, JOHNSEN H, et al. Using Sentinel-1 ocean data for mapping sea surface currents along the southern Norwegian Coast[C]. IGARSS 2019 - 2019 IEEE International Geoscience and Remote Sensing Symposium, Yokohama, Japan, 2019: 8058–8061. doi: 10.1109/IGARSS.2019.8898468. [10] MOISEEV A, JOHANNESSEN J A, and JOHNSEN H. Towards retrieving reliable ocean surface currents in the coastal zone from the Sentinel-1 Doppler shift observations[J]. Journal of Geophysical Research: Oceans, 2022, 127(5): e2021JC018201. doi: 10.1029/2021JC018201. [11] YANG Zhonghao, LIU Lei, WANG Jing, et al. Extrapolation of electromagnetic pointing error corrections for Sentinel-1 Doppler currents from land areas to the open ocean[J]. Remote Sensing of Environment, 2023, 297: 113788. doi: 10.1016/j.rse.2023.113788. [12] JOHNSEN H, NILSEN V, ENGEN G, et al. Ocean doppler anomaly and ocean surface current from Sentinel 1 tops mode[C]. 2016 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Beijing, China, 2016: 3993–3996. doi: 10.1109/IGARSS.2016.7730038. [13] MOISEEV A, JOHNSEN H, JOHANNESSEN J A, et al. On removal of sea state contribution to Sentinel-1 Doppler shift for retrieving reliable ocean surface current[J]. Journal of Geophysical Research: Oceans, 2020, 125(9): e2020JC016288. doi: 10.1029/2020JC016288. [14] YANG Zhonghao, WANG Jing, LIU Lei, et al. Estimating effects of wind and waves on the Doppler centroid frequency shift for the SAR retrieval of ocean currents[J]. Remote Sensing of Environment, 2024, 311: 114312. doi: 10.1016/j.rse.2024.114312. [15] SHAO Weizeng, ZHOU Yuhang, HU Yuyi, et al. Range current retrieval fromsentinel-1 SAR ocean product based on deep learning[J]. Remote Sensing Letters, 2024, 15(2): 145–156. doi: 10.1080/2150704X.2024.2305176. [16] FAN Shengren, KUDRYAVTSEV V, YUROVSKY Y, et al. Reconstructing ocean surface current vector field from SAR doppler shift measurements[J]. Remote Sensing of Environment, 2025, 328: 114855. doi: 10.1016/j.rse.2025.114855. [17] CHEN Tianqi and GUESTRIN C. XGBoost: A scalable tree boosting system[C]. The 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, USA, 2016: 785–794. doi: 10.1145/2939672.2939785. [18] BORISOV V, LEEMANN T, SEßLER K, et al. Deep neural networks and tabular data: A survey[J]. IEEE Transactions on Neural Networks and Learning Systems, 2024, 35(6): 7499–7519. doi: 10.1109/TNNLS.2022.3229161. [19] GRINSZTAJN L, OYALLON E, and VAROQUAUX G. Why do tree-based models still outperform deep learning on typical tabular data?[C]. The 36th International Conference on Neural Information Processing Systems, New Orleans, USA, 2022: 37. doi: 10.5555/3600270.3600307. [20] MURPHY D and JANZEN C. Advances in in-situ ocean measurements[M]. VENKATESAN R, TANDON A, D'ASARO E, et al. Observing the Oceans in Real Time. Cham: Springer, 2018: 141–162. doi: 10.1007/978-3-319-66493-4_8. [21] Copernicus. Sentiwiki:S-1mission[EB/OL].2023. https://sentiwiki.copernicus.eu/web/s1-mission. [22] DE ZAN F and MONTI GUARNIERI A. TOPSAR: Terrain observation by progressive scans[J]. IEEE Transactions on Geoscience and Remote Sensing, 2006, 44(9): 2352–2360. doi: 10.1109/TGRS.2006.873853. [23] Copernicus Marine Service. Product User Manual For Global Ocean Waves Analysis and Forecast [EB/OL]. CMEMS-GLO-PUM-001-032,2023. https://documentation.marine.copernicus.eu/PUM/CMEMS-GLO-PUM-001-032.pdf. [24] LAW-CHUNE S, AOUF L, DALPHINET A, et al. WAVERYS: A CMEMS global wave reanalysis during the altimetry period[J]. Ocean Dynamics, 2021, 71(3): 357–378. doi: 10.1007/s10236-020-01433-w. [25] ZHANG Yixuan, YUE Songshan, XU Kai, et al. Performance analysis of global HYCOM flow field using Argo profiles[J]. International Journal of Digital Earth, 2023, 16(1): 3536–3559. doi: 10.1080/17538947.2023.2252407. [26] CHASSIGNET E P, HURLBURT H E, SMEDSTAD O M, et al. The HYCOM (HYbrid Coordinate Ocean Model) data assimilative system[J]. Journal of Marine Systems, 2007, 65(1/4): 60–83. doi: 10.1016/j.jmarsys.2005.09.016. [27] SAVAGE J A, TOKMAKIAN R T, and BATTEEN M L. Assessment of the HYCOM velocity fields during Agulhas Return Current Cruise 2012[J]. Journal of Operational Oceanography, 2015, 8(1): 11–24. doi: 10.1080/1755876X.2015.1014637. [28] KABIR A, LEMONGO-TCHAMBA I, and FERNANDEZ A. An assessment of available ocean current hydrokinetic energy near the North Carolina shore[J]. Renewable Energy, 2015, 80: 301–307. doi: 10.1016/j.renene.2015.02.011. [29] LUECKE C A, ARBIC B K, RICHMAN J G, et al. Statistical comparisons of temperature variance and kinetic energy in global ocean models and observations: Results from mesoscale to internal wave frequencies[J]. Journal of Geophysical Research: Oceans, 2020, 125(5): e2019JC015306. doi: 10.1029/2019JC015306. [30] WANG Haodi, CHEN Shiyao, WANG Ning, et al. Evaluation of multi-model current data in the East/Japan Sea[C]. 2020 IEEE 3rd International Conference on Information Communication and Signal Processing (ICICSP), Shanghai, China, 2020: 486–491. doi: 10.1109/ICICSP50920.2020.9232090. [31] HARLAN J, TERRILL E, HAZARD L, et al. The integrated ocean observing system HF radar network[C]. OCEANS 2015-MTS/IEEE Washington, Washington, USA, 2015: 1–4. doi: 10.23919/OCEANS.2015.7404587. [32] LORENTE P, SOTO-NAVARRO J, ALVAREZ FANJUL E, et al. Accuracy assessment of high frequency radar current measurements in the Strait of Gibraltar[J]. Journal of Operational Oceanography, 2014, 7(2): 59–73. doi: 10.1080/1755876X.2014.11020300. [33] KENNEDY J and EBERHART R. Particle swarm optimization[C]. ICNN'95 - International Conference on Neural Networks, Perth, Australia, 1995: 1942–1948. doi: 10.1109/ICNN.1995.488968. [34] MARCILIO W E and ELER D M. From explanations to feature selection: Assessing SHAP values as feature selection mechanism[C]. 2020 33rd SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI), Porto de Galinhas, Brazil, 2020: 340–347. doi: 10.1109/SIBGRAPI51738.2020.00053. [35] ANDRES M. Spatial and temporal variability of the gulf stream near cape Hatteras[J]. Journal of Geophysical Research: Oceans, 2021, 126(9): e2021JC017579. doi: 10.1029/2021JC017579. -
计量
- 文章访问数:
- HTML全文浏览量:
- PDF下载量:
- 被引次数: 0
