Significant wave height retrieval model of CYGNSS based on multivariate machine learning
-
摘要:
旋风全球导航卫星系统(CYGNSS) 提供了高质量的全球导航卫星系统反射信号 (GNSS-R) 技术数据，能可靠地用于有效波高(SWH)的反演。由于CYGNSS的高动态性，导致接收信号很容易受到环境因素的影响，海况的复杂性使简单模型难以准确反演SWH。为解决上述问题，提出一种基于机器学习的多变量SWH反演模型，根据海浪的形成机理及对CYGNSS参数与SWH之间的相关性结果分析选取出相关参数，并设计5参数、9参数和17参数3种训练方案。随后利用随机森林(RF)和卷积神经网络(CNN)对反演模型进行训练和验证，并将SWH反演结果与欧洲中期天气预报中心(ECMWF)的参考值进行比较。最佳的反演模型是17参数CNN反演模型，均方根误差（RMSE）为
0.1840 m，相关系数R 2=0.9485 。与17参数CNN反演模型相比，9参数CNN反演模型减少了24%的训练时间，并且精度损失很小。但9参数反演模型相较17参数反演模型在泛化评估方面表现不佳。因此，为提高模型的泛化能力，将风速作为参数添加到17参数反演模型中，得到了17+1参数泛化模型。其中，最佳的泛化模型是17+1参数RF泛化模型，RMSE为0.497 1 m，R 2=0.584 6。有效地证明了所提模型在SWH反演中具有良好的潜力。-
关键词:
- 有效波高 /
- 旋风全球导航卫星系统 /
- 星载GNSS-R /
- 机器学习 /
- 卷积神经网络
Abstract:The cyclone global navigation satellite system (CYGNSS) provides high-quality global navigation satellite system reflectometry (GNSS-R) data that can be reliably used for the retrieval of significant wave height (SWH). Due to the high dynamic nature of CYGNSS, the received signal is easily affected by environmental factors, and the complexity of sea conditions makes it difficult for simple models to accurately retrieve SWH. To address the above issues, this article proposed an SWH retrieval model based on multivariate machine learning. According to the mechanism of wave formation and the analysis of the correlation between CYGNSS parameters and SWH, relevant parameters were selected, and three training schemes were designed, involving five parameters, nine parameters, and 17 parameters, respectively. Random forest (RF) and convolutional neural network (CNN) were used to train and verify the retrieval model, and the SWH retrieval results were compared with the reference values of the European Centre for Medium-Range Weather Forecasts (ECMWF). The best retrieval model among them was the 17-parameter CNN retrieval model, with root mean square error（RMSE）was 0.184 0 m and $ {R}^{2} $= 0.948 5. Compared with the 17-parameter CNN retrieval model, the 9-parameter CNN retrieval model reduced training time by 24% and has minimal accuracy loss. However, the 9-parameter retrieval model performed poorly in terms of generalization evaluation compared to the 17-parameter retrieval model. To improve the generalization ability of the model, wind speed was added as a parameter to the 17-parameter retrieval model, resulting in a 17 + 1-parameter generalization model. The best generalization model among them was the 17 + 1 parameter RF generalization model, with RMSE was 0.497 1 m and $ {R}^{2} $= 0.584 6. This effectively proves that the model proposed in this article has good potential in SWH retrieval.
-
图 3 风速与NBRCS和涌浪与LES的关系
Figure 3. Relationship of wind speed with NBRCS and swell with LES
图 8 CYGNSS参数相关系数热力图
Figure 8. Heatmap of correlation coefficients of CYGNSS parameters
图 9 不同反演模型反演结果的相关散点图
Figure 9. Scatter plots of retrieval results related to different retrieval models
图 10 不同反演模型泛化结果的相关散点图
Figure 10. Scatter plots of generalization results related to different retrieval models
图 11 不同泛化模型泛化结果的相关散点图
Figure 11. Scatter plots of generalization results related to different generalization models
表 1 特征参数
Table 1. Characteristic parameters
参数列表 5参数 9参数 17参数 DSNR ○ ○ ○ LES ○ ○ ○ NBRCS ○ ○ ○ Vty ○ ○ ○ Vsy ○ ○ ○ SPANG ○ ○ ZSNR ○ ○ ANT ○ ○ FC ○ ○ DPDR ○ DPDC ○ DSDR ○ DSDC ○ Vtx ○ Vtz ○ Vsx ○ Vsz ○ 注：○表示使用了该参数。 
下载: 导出CSV
表 2 SWH反演模型的反演性能分析
Table 2. Retrieval performance analysis of SWH retrieval models
反演模型 RMSE/m $ {R}^{2} $ 5参数RF反演模型 0.6471 0.3641 5参数CNN反演模型 0.6517 0.3550 9参数RF反演模型 0.3059 0.8578 9参数CNN反演模型 0.1978 0.9405 17参数RF反演模型 0.2107 0.9325 17参数CNN反演模型 0.1840 0.9485
下载: 导出CSV
表 3 SWH反演模型的泛化性能分析
Table 3. Generalization performance analysis of SWH retrieval models
反演模型 RMSE/m $ {R}^{2} $ 9参数RF反演模型 0.5784 0.4375 9参数CNN反演模型 0.6871 0.2064 17参数RF反演模型 0.5175 0.5498 17参数CNN反演模型 0.5990 0.3969
下载: 导出CSV
表 4 SWH泛化模型的泛化性能分析
Table 4. Generalization performance analysis of SWH generalization models
泛化模型 RMSE/m $ {R}^{2} $ 17+1参数RF泛化模型 0.4971 0.5846 17+1参数CNN泛化模型 0.5617 0.4696 文献[27] 0.6430 0.3700
下载: 导出CSV
表 5 SWH不同区间的泛化精度评估
Table 5. Evaluation of generalization accuracy for different intervals of SWH
区间/m RMSE/m 训练集数量 测试集数量 0～1 0.4333 96626 465909 1～2 0.3920 1032123 4972700 2～3 0.4090 1042745 5338182 3～4 0.7727 262176 1146376 4～5 1.2386 47533 267787 >5 1.6599 18797 69085
下载: 导出CSV
-
[1] FOTI G, GOMMENGINGER C, JALES P, et al. Spaceborne GNSS reflectometry for ocean winds: first results from the UK TechDemoSa-1 mission[J]. Geophysical Research Letters, 2020, 42: 5435-5441. [2] CLARIZIA M P, RUF C S. Wind speed retrieval algorithm for the cyclone global navigation satellite system (CYGNSS) mission[J]. IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(8): 4419-4432. doi: 10.1109/TGRS.2016.2541343 [3] ZHANG Y, CHEN X, MENG W T, et al. Wind direction retrieval using support vector machine from CYGNSS sea surface data[J]. Remote Sensing, 2021, 13(21): 4451. doi: 10.3390/rs13214451 [4] CARDELLACH E, LI W Q, RIUS A, et al. First precise spaceborne sea surface altimetry with GNSS reflected signals[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2019, 13: 102-112. [5] LI W Q, CARDELLACH E, FABRA F, et al. Assessment of spaceborne GNSS-R ocean altimetry performance using CYGNSS mission raw data[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(1): 238-250. doi: 10.1109/TGRS.2019.2936108 [6] ZHANG Y, LI B B, TIAN L M, et al. Phase altimetry using reflected signals from BeiDou GEO satellites[J]. IEEE Geoscience and Remote Sensing Letters, 2016, 13(10): 1410-1414. doi: 10.1109/LGRS.2016.2578361 [7] ZHANG Y, TIAN L M, MENG W T, et al. Feasibility of code-level altimetry using coastal BeiDou reflection (BeiDou-R) setups[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2015, 8(8): 4130-4140. doi: 10.1109/JSTARS.2015.2446684 [8] ZHANG Y, HUANG S, HAN Y L, et al. Machine learning methods for spaceborne GNSS-R sea surface height measurement from TDS-1[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021, 15: 1079-1088. [9] LI Y W, CHANG X, YU K G, et al. Estimation of snow depth using pseudorange and carrier phase observations of GNSS single-frequency signal[J]. GPS Solutions, 2019, 23(4): 118. doi: 10.1007/s10291-019-0912-5 [10] YU K G. Theory and practice of GNSS reflectometry[M]. Berlin: Springer, 2021: 1-376. [11] ZHANG Y, MA D H, MENG W T, et al. Antarctic sea ice detection using a shipborne GPS reflectometry setup[J]. Measurement and Control, 2021, 54(5-6): 618-626. doi: 10.1177/00202940211013063 [12] YAN Q Y, HUANG W M. Spaceborne GNSS-R sea ice detection using delay-Doppler maps: first results from the U. K. TechDemoSat-1 mission[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2016, 9(10): 4795-4801. doi: 10.1109/JSTARS.2016.2582690 [13] ZHANG Y, MENG W T, GU Q M, et al. Detection of Bohai bay sea ice using GPS-reflected signals[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014, 8(1): 39-46. [14] ZHANG Y, HANG S J, HAN Y L, et al. Sea ice thickness detection using coastal BeiDou reflection setup in Bohai Bay[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 18(3): 381-385. doi: 10.1109/LGRS.2020.2980106 [15] CHANG X, JIN T Y, YU K G, et al. Soil moisture estimation by GNSS multipath signal[J]. Remote Sensing, 2019, 11(21): 2559. doi: 10.3390/rs11212559 [16] YAN Q Y, HUANG W M, JIN S G, et al. Pan-tropical soil moisture mapping based on a three-layer model from CYGNSS GNSS-R data[J]. Remote Sensing of Environment, 2020, 247: 111944. doi: 10.1016/j.rse.2020.111944 [17] GERLEIN-SAFDI C, RUF C S. A CYGNSS-based algorithm for the detection of inland waterbodies[J]. Geophysical Research Letters, 2019, 46(21): 12065-12072. doi: 10.1029/2019GL085134 [18] AL-KHALDI M M, JOHNSON J T, GLEASON S, et al. Inland water body mapping using CYGNSS coherence detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(9): 7385-7394. doi: 10.1109/TGRS.2020.3047075 [19] MARTÍN-NEIRA M. A pasive reflectometry and interferometry system (PARIS) application to ocean altimetry[J]. ESA Journal, 1993, 17(4): 331-355. [20] SOULAT F, CAPARRINI M, GERMAIN O, et al. Sea state monitoring using coastal GNSS-R[J]. Journal of Geophysical Research, 2004, 31(21): 7385-7394. [21] ALPERS W, HASSELMANN K. Spectral signal to clutter and thermal noise properties of ocean wave imaging synthetic aperture radars[J]. International Journal of Remote Sensing, 1982, 3(4): 423-446. doi: 10.1080/01431168208948413 [22] ALONSO-ARROYO A, CAMPS A, PARK H, et al. Retrieval of significant wave height and mean sea surface level using the GNSS-R interference pattern technique: results from a three-month field campaign[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 53(6): 3198-3209. doi: 10.1109/TGRS.2014.2371540 [23] QIN L Y, LI Y. Significant wave height estimation using multi-satellite observations from GNSS-R[J]. Remote Sensing, 2021, 13(23): 4806. doi: 10.3390/rs13234806 [24] CLARIZIA M P, GOMMENGINGER C P, GLEASON S T, et al. Analysis of GNSS-R delay-Doppler maps from the UK-DMC satellite over the ocean[J]. Geophysical Research Letters, 2009, 36(2): L02608. [25] PENG Q, JIN S G. Significant wave height estimation from space-borne cyclone-GNSS reflectometry[J]. Remote Sensing, 2019, 11(5): 584. doi: 10.3390/rs11050584 [26] YANG S, JIN S G, JIA Y, et al. Significant wave height estimation from joint CYGNSS DDMA and LES observations[J]. Sensors, 2021, 21(18): 6123. doi: 10.3390/s21186123 [27] BU J W, YU K G. Significant wave height retrieval method based on spaceborne GNSS reflectometry[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19: 1503705. [28] WANG F, YANG D K, YANG L. Retrieval and assessment of significant wave height from CYGNSS mission using neural network[J]. Remote Sensing, 2022, 14(15): 3666. doi: 10.3390/rs14153666 [29] LV H Q. Fundamentals of physical oceanography[M]. San Jose : Ocean Press, 2012: 178. [30] PIERSON W J, MOSKOWITZ L. A proposed spectral form for fully developed wind seas based on the similarity theory of S. A. Kitaigorodskii[J]. Journal of Geophysical Research, 1964, 69(24): 5181-5190. doi: 10.1029/JZ069i024p05181 [31] CLARIZIA M P, RUF C S, JALES P, et al. Spaceborne GNSS-R minimum variance wind speed estimator[J]. IEEE Transactions on Geoscience and Remote Sensing, 2014, 52(11): 6829-6843. doi: 10.1109/TGRS.2014.2303831 [32] LIN W M, PORTABELLA M, FOTI G, et al. Toward the generation of a wind geophysical model function for spaceborne GNSS-R[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(2): 655-666. doi: 10.1109/TGRS.2018.2859191 [33] BROWN G. Backscattering from a Gaussian-distributed perfectly conducting rough surface[J]. IEEE Transactions on Antennas and Propagation, 1978, 26(3): 472-482. doi: 10.1109/TAP.1978.1141854 [34] ZHANG Y, YIN J W, YANG S H, et al. High wind speed inversion model of CYGNSS sea surface data based on machine learning[J]. Remote Sensing, 2021, 13(16): 3324. doi: 10.3390/rs13163324 [35] LI X H. Application of random forest model in classification and regression analysis[J]. Chinese Journal of Applied Entomology, 2013, 4: 1190-1197. [36] GUO W F, DU H, GUO C, et al. Information fusion for GNSS-R wind speed retrieval using statistically modified convolutional neural network[J]. Remote Sensing of Environment, 2022, 272: 112934. doi: 10.1016/j.rse.2022.112934 [37] CLARIZIA M P, RUF C S. Statistical derivation of wind speeds from CYGNSS data[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(6): 3955-3964. doi: 10.1109/TGRS.2019.2959715 [38] ZHANG Y, LU Q, JIN Q , et al. Global sea surface height measurement from CYGNSS based on machine learning[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 2022, 16: 841-852. -
下载:
点击查看大图
计量
- 文章访问数: 864
- HTML全文浏览量: 185
- PDF下载量: 31
- 被引次数: 0
