基于BP神经网络模型的哨兵SAR反演风速偏差校正

倪晗玥; 董昌明; 刘振波; 杨劲松; 李晓辉; 任林

doi:10.3969/j.issn.1001-909X.2024.03.006

PDF(3937 KB)

海洋学研究 ›› 2024, Vol. 42 ›› Issue (3) : 75-87. DOI: 10.3969/j.issn.1001-909X.2024.03.006

研究论文

基于BP神经网络模型的哨兵SAR反演风速偏差校正

倪晗玥 ¹^,² ,
董昌明 ³ ,
刘振波 ³ ,
杨劲松 ¹^,²^,^* ,
李晓辉 ² ,
任林 ²

作者信息 +

Calibration of Sentinel-1 SAR retrieved wind speed based on BP neural network model

NI Hanyue ¹^,² ,
DONG Changming ³ ,
LIU Zhenbo ³ ,
YANG Jingsong ¹^,²^,^* ,
LI Xiaohui ² ,
REN Lin ²

Author information +

文章历史 +

摘要

该文基于美国国家浮标资料中心(National Data Buoy Center,NDBC) 浮标观测数据对哨兵一号搭载的合成孔径雷达(synthetic aperture radar,SAR) 反演风速数据进行精度分析,并利用BP神经网络 (back propagation neural network) 对SAR反演风速的偏差进行校正;同时针对环境要素、BP神经网络训练输入的样本量以及神经网络结构参数设计了敏感性试验;最后将SAR标量风场数据转换为用u、v矢量表示的风场数据,并对u向风和v向风分别进行了精度分析和校正。实验结果表明:SAR反演风速相较于浮标观测数据出现了低估现象;经过BP神经网络校正后,SAR反演风速数据的精度得到了改善,风速的平均偏差绝对值从0.78 m/s下降到0.04 m/s,均方根误差从1.98 m/s下降到了1.77 m/s;敏感性试验表明输入质量较差的环境要素数据时BP神经网络的校正效果有所下降,而增加训练集样本量能改善校正效果;将标量风场数据转换为u、v矢量风场数据后的校正结果也显示BP神经网络具有较好的校正效果。

Abstract

An accuracy analysis of wind speed data retrieved from Sentinel-1 synthetic aperture radar (SAR) was conducted based on buoy observations from the National Data Buoy Center (NDBC). A back propagation (BP) neural network was utilized to correct the deviation in the SAR-derived wind speeds. Sensitivity experiments were designed for environmental factors, the number of training samples for BP neural network input, and neural network structure parameters. Finally, the SAR wind field data were converted into u and v vector wind data, and the accuracy analysis and correction were performed separately for u and v wind components. The experiment finds that the SAR-derived wind speed is underestimated compared to the buoy data. After calibration using BP neural network, the accuracy of SAR-derived wind speed data is improved, and the absolute value of bias of wind speed decreases from 0.78 m/s to 0.04 m/s, the RMSE of wind speed decreases from 1.98 m/s to 1.77 m/s. The sensitivity experiments suggest that low quality environmental factors input data will decrease the calibration effect of BP neural network, and increasing the sample size of the training set can improve that. The calibration results of converted u and v vector wind field data also show that the BP neural network has good correction effect.

导出引用

倪晗玥, 董昌明, 刘振波, 等. 基于BP神经网络模型的哨兵SAR反演风速偏差校正[J]. 海洋学研究. 2024, 42(3): 75-87 https://doi.org/10.3969/j.issn.1001-909X.2024.03.006

NI Hanyue, DONG Changming, LIU Zhenbo, et al. Calibration of Sentinel-1 SAR retrieved wind speed based on BP neural network model[J]. Journal of Marine Sciences. 2024, 42(3): 75-87 https://doi.org/10.3969/j.issn.1001-909X.2024.03.006

中图分类号： P732;TP183

参考文献

列表( 原文顺序 | 文献年度倒序 | 文中引用次数倒序 ) 可视化分析

[1]	蒋兴伟, 林明森, 张有广. 中国海洋卫星及应用进展[J]. 遥感学报, 2016, 20(5):1185-1198. JIANG X W, LIN M S, ZHANG Y G. Progress and prospect of Chinese ocean satellites[J]. Journal of Remote Sensing, 2016, 20(5): 1185-1198. 本文引用 [1]

[2]	方贺, 杨劲松, 樊高峰, 等. 组合表面Bragg散射模型共极化SAR海表面风速反演[J]. 遥感学报, 2022, 26(6):1274-1287. FANG H, YANG J S, FAN G F, et al. Ocean surface wind speed retrieve from co-polarized SAR using composite surface Bragg scattering model[J]. National Remote Sensing Bulletin, 2022, 26(6): 1274-1287. 本文引用 [1]

[3]	蔡树群, 张文静, 王盛安. 海洋环境观测技术研究进展[J]. 热带海洋学报, 2007, 26(3):76-81. CAI S Q, ZHANG W J, WANG S A. An advance in marine environment observation technology[J]. Journal of Tropical Oceanography, 2007, 26(3): 76-81. 本文引用 [1]

[4]	徐文, 鄢社锋, 季飞, 等. 海洋信息获取、传输、处理及融合前沿研究评述[J]. 中国科学:信息科学, 2016, 46(8):1053-1085. XU W, YAN S F, JI F, et al. Marine information gathering, transmission, processing, and fusion: Current status and future trends[J]. Scientia Sinica: Informationis, 2016, 46(8): 1053-1085. 本文引用 [1]

[5]	范开国, 徐青, 徐东洋, 等. 星载SAR海面风场遥感研究进展[J]. 地球物理学进展, 2022, 37(5):1807-1817. FAN K G, XU Q, XU D Y, et al. Review of remote sensing of sea surface wind field by space-borne SAR[J]. Progress in Geophysics, 2022, 37(5): 1807-1817. 本文引用 [1]

[6]	邓云凯, 赵凤军, 王宇. 星载SAR技术的发展趋势及应用浅析[J]. 雷达学报, 2012, 1(1):1-10. DENG Y K, ZHAO F J, WANG Y. Brief analysis on the development and application of spaceborne SAR[J]. Journal of Radars, 2012, 1(1): 1-10. 本文引用 [1]

[7]	张康宇. 基于C波段SAR的海面风场反演方法与近海风能资源评估[D]. 杭州: 浙江大学, 2019. ZHANG K Y. Sea surface wind field retrieval and offshore wind energy resource assessment based on C-band SAR data[D]. Hangzhou: Zhejiang University, 2019. 本文引用 [1]

[8]	张璇. C波段全极化SAR图像海面风场反演研究[D]. 南京: 南京信息工程大学, 2018. ZHANG X. Research of ocean wind retrival from C-band quad-polarization SAR images[D]. Nanjing: Nanjing University of Information Science & Technology, 2018. 本文引用 [1]

[9]	欧阳伦曦, 李新情, 惠凤鸣, 等. 哨兵卫星Sentinel-1A数据特性及应用潜力分析[J]. 极地研究, 2017, 29(2):286-295. OU-YANG L X, LI X Q, HUI F M, et al. Sentinel-1A data products’ characteristics and the potential applications[J]. Chinese Journal of Polar Research, 2017, 29(2): 286-295. 本文引用 [1]

[10]	朱良, 郭巍, 禹卫东. 合成孔径雷达卫星发展历程及趋势分析[J]. 现代雷达, 2009, 31(4):5-10. ZHU L, GUO W, YU W D. Analysis of SAR satellite development history and tendency[J]. Modern Radar, 2009, 31(4): 5-10. 本文引用 [1]

[11]	孙佳. 国外合成孔径雷达卫星发展趋势分析[J]. 装备指挥技术学院学报, 2007, 18(1):67-70. SUN J. Analysis of the SAR satellite development tendency in the world[J]. Journal of the Academy of Equipment Command & Technology, 2007, 18(1): 67-70. 本文引用 [1]

[12]	LIN H, XU Q, ZHENG Q A. An overview on SAR measurements of sea surface wind[J]. Progress in Natural Science, 2008, 18(8): 913-919. 本文引用 [1]

[13]	FREILICH M H, DUNBAR R S. The accuracy of the NSCAT 1 vector winds: Comparisons with National Data Buoy Center buoys[J]. Journal of Geophysical Research: Oceans, 1999, 104(C5): 11231-11246. 本文引用 [1]

[14]	WAN Y, GUO S, LI L G, et al. Data quality evaluation of Sentinel-1 and GF-3 SAR for wind field inversion[J]. Remote Sensing, 2021, 13(18): 3723. 本文引用 [1]

[15]	PENSIERI S, BOZZANO R, SCHIANO M E. Comparison between QuikSCAT and buoy wind data in the Ligurian Sea[J]. Journal of Marine Systems, 2010, 81(4): 286-296. 本文引用 [2]

[16]	代彭坤. 基于Sentinel-1的SAR台风风场反演研究[D]. 厦门: 厦门大学, 2018. DAI P K. Research of retrieving typhoon wind field based on Sentinel-1 SAR images[D]. Xiamen: Xiamen University, 2018. 本文引用 [1]

[17]	许素芹, 李婷婷, 陈捷, 等. 海洋温度锋区后向散射系数仿真与特性研究[J]. 计算机仿真, 2019, 36(10):214-218,297. XU S Q, LI T T, CHEN J, et al. Research and simulation of backscattering coefficient characteristic across ocean thermal fronts[J]. Computer Simulation, 2019, 36(10): 214-218, 297. 本文引用 [1]

[18]	丁赟. 风浪状态对海面风应力的影响[D]. 青岛: 中国海洋大学, 2005. DING Y. The influence of wind sea state on wind stress[D]. Qingdao: Ocean University of China, 2005. 本文引用 [1]

[19]	殷实. 西北太平洋海面粗糙度分布与变化的初步研究[D]. 青岛: 中国海洋大学, 2009. YIN S. Temporal and spatial variation of the sea surface roughness over the northwest Pacific Ocean[D]. Qingdao: Ocean University of China, 2009. 本文引用 [1]

[20]	余水, 杨劲松, 贺双颜, 等. 降雨对SAR风场反演的影响及校正[J]. 海洋学报, 2017, 39(9):40-50. YU S, YANG J S, HE S Y, et al. The correction of rain effect on SAR wind field retrieval[J]. Haiyang Xuebao, 2017, 39(9): 40-50. 本文引用 [1]

[21]	BENTAMY A, GRODSKY S A, CARTON J A, et al. Matching ASCAT and QuikSCAT winds[J]. Journal of Geophysical Research: Oceans, 2012, 117: C02011. 本文引用 [1]

[22]	MONALDO F, JACKSON C, PICHEL W, et al. Early validation of operational SAR wind retrievals from Sentinel-1A[C]//2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), July 26-31, 2015, Milan, Italy. IEEE, 2015: 1223-1226. 本文引用 [1]

[23]	徐天真, 徐静琦, 楼顺里. 海面风垂直分布的计算方法[J]. 海洋湖沼通报, 1988, 10(4):1-6. XU T Z, XU J Q, LOU S L. A method for calculating the wind variation with height in the marine atmospheric surface layer[J]. Transactions of Oceanology and Limnology, 1988, 10(4): 1-6. 本文引用 [1]

[24]	陈永利, 赵永平, 张必成, 等. 海上不同高度风速换算关系的研究[J]. 海洋科学, 1989, 13(3):27-31. CHEN Y L, ZHAO Y P, ZHANG B C, et al. The study of the relations of wind velocity at different heights over the sea[J]. Marine Sciences, 1989, 13(3): 27-31. 本文引用 [1]

[25]	ATLAS R, BUSALACCHI A J, GHIL M, et al. Global surface wind and flux fields from model assimilation of Seasat data[J]. Journal of Geophysical Research: Oceans, 1987, 92(C6): 6477-6487. 本文引用 [1]

[26]	LIU W T, KATSAROS K B, BUSINGER J A. Bulk parameterization of air-sea exchanges of heat and water vapor including the molecular constraints at the interface[J]. Journal of the Atmospheric Sciences, 1979, 36(9): 1722-1735. 本文引用 [1]

[27]	JOHNSON H K. Simple expressions for correcting wind speed data for elevation[J]. Coastal Engineering, 1999, 36(3): 263-269. 本文引用 [1]

[28]	LIU W T, TANG W. Equivalent neutral winds[R]. Pasadena, CA: Jet Propulsion Laboratory, 1996. 本文引用 [1]

[29]	FAIRALL C W, BRADLEY E F, HARE J E, et al. Bulk parameterization of air-sea fluxes: Updates and verification for the COARE algorithm[J]. Journal of Climate, 2003, 16(4): 571-591. 本文引用 [1]

[30]

陈克海, 解学通, 张金兰, 等. HY-2B卫星散射计海面风场产品质量分析[J]. 热带海洋学报, 2020, 39(6):30-40.

https://doi.org/10.11978/2019138

摘要

星载微波散射计是获取全球海面风场信息的主要手段, HY-2B卫星散射计的成功发射为全球海面风场数据获取的持续性提供了重要保障。本文利用欧洲中期天气预报中心(European Center for Medium-Range Weather Forecasts, ECMWF)再分析风场数据、热带大气海洋观测计划(Tropical Atmosphere Ocean Array, TAO)和美国国家数据浮标中心(National Data Buoy Center, NDBC)浮标获取的海面风矢量实测数据, 对HY-2B散射计海面风场数据产品的质量进行统计分析。分析表明, HY-2B风场与ECMWF再分析风场对比, 在4~24m·s-1风速区间内, 风速和风向均方根误差(root mean square error, RMSE)分别为1.58m·s-1和15.34°; 与位于开阔海域的TAO浮标数据对比, 风速、风向RMSE分别为1.03m·s-1和14.98°, 可见HY-2B风场能较好地满足业务化应用的精度要求(风速优于2m·s-1, 风向优于20°)。与主要位于近海海域的NDBC浮标对比, HY-2B风场的风速、风向RMSE分别为1.60m·s-1和19.14°, 说明HY-2B散射计同时具备了对近海海域风场的良好观测能力。本文还发现HY-2B风场质量会随风速、地面交轨位置等变化, 为用户更好地使用HY-2B风场产品提供参考。

CHEN

K H

, XIE

X T

, ZHANG

J L

, et al. Accuracy analysis of the retrieved wind from HY-2B scatterometer[J]. Journal of Tropical Oceanography, 2020, 39(6): 30-40.

https://doi.org/10.11978/2019138

本文引用 [1] 摘要

Spaceborne microwave scatterometer is the most efficient instrument to measure the global ocean surface wind. The successfully launched HY-2B scatterometer is an important guarantee to continue the measurement of global ocean surface wind. In this paper, we use the European Center for Medium-Range Weather Forecasts (ECMWF) reanalysis wind data, Tropical Atmosphere Ocean Array (TAO) and National Data Buoy Center (NDBC) buoy wind data to estimate the accuracy of the HY-2B retrieved wind. Our statistical results show that at speed range from 4 to 24 m·s^-1, the retrieved wind speed root mean square error (RMSE) with respect to the ECMWF wind speed is 1.58 m·s^-1, and the direction RMSE is 15.34°, while the speed and direction RMSEs with respect to the TAO buoys located in the open ocean are 1.06 m·s^-1 and 15.98°, respectively. Thus, the retrieved wind well satisfies the operational precision requirement of the HY-2B scatterometer, that is, the speed and direction RMSEs should be better than 2 m·s^-1 and 20.00°, respectively, for the open ocean. Moreover, the NDBC buoys are mainly located in immediate offshore areas, HY-2B wind speed and direction RMSEs are 1.60 m·s^-1and 19.14°, respectively, which shows that HY-2B scatterometer has good ability to measure wind of offshore sea areas. We also reveal that the retrieved wind error varies with wind speed and cross track position of wind vector cell, which will provide the users with reference for the effective application of HY-2B wind product.

[31]	DINKU T, RUIZ F, CONNOR S J, et al. Validation and intercomparison of satellite rainfall estimates over Colombia[J]. Journal of Applied Meteorology and Climatology, 2010, 49(5): 1004-1014. 本文引用 [1]

[32]	JAMANDRE C A, NARISMA G T. Spatio-temporal validation of satellite-based rainfall estimates in the Philippines[J]. Atmospheric Research, 2013, 122: 599-608. 本文引用 [1]

[33]	SALIO P, HOBOUCHIAN M P, GARCÍA SKABAR Y, et al. Evaluation of high-resolution satellite precipitation estimates over southern South America using a dense rain gauge network[J]. Atmospheric Research, 2015, 163: 146-161. 本文引用 [1]

[34]	XING J Y, SHI J C, LEI Y H, et al. Evaluation of HY-2A scatterometer wind vectors using data from buoys, ERA-interim and ASCAT during 2012-2014[J]. Remote Sensing, 2016, 8(5): 390. 本文引用 [1]

[35]

朱金台, 董晓龙, 云日升. 利用浮标和NWP风场对HY-2散射计联合定标验证[J]. 电子学报, 2015, 43(11):2237-2242.

https://doi.org/10.3969/j.issn.0372-2112.2015.11.015

摘要

本文对海洋二号卫星微波散射计(Haiyang-2 Scatterometer,HY-2 SCAT)进行了海洋定标算法研究,并使用数值天气预报模型风场(Numerical Weather Prediction,NWP)和浮标数据对定标后反演风场进行联合验证.通过匹配2012年12月份的HY-2 SCAT反演风场、NWP风场及浮标的观测数据,共得到无降雨条件下的3112个25km分辨率的匹配数据.对匹配数据进行分析时,采用基于变量的误差分析方法能够得到比传统线性回归方法更精确的验证结果.选取在风场U、V分量进行联合验证能得到较在风速、风向上更为有利的验证结果.验证结果表明,经过海洋定标法之后的HY-2 SCAT测量后向散射系数的误差残余小于0.15dB,其反演风场与浮标及NWP数据相吻合,U、V分量相对浮标及NWP数据偏差均小于0.23m/s,验证了该定标算法的有效性及定标后反演风场的高精度.

ZHU

J T

, DONG

X L

, YUN

R S

. Calibration and validation of HY-2 scatterometer using NWP and buoy data[J]. Acta Electronica Sinica, 2015, 43(11): 2237-2242.

本文引用 [1]

[36]

孙乾洋, 周利, 丁仕风, 等. 基于人工神经网络的极地船舶冰阻力预报方法[J]. 上海交通大学学报, 2024, 58(2):156-165.

https://doi.org/10.16183/j.cnki.jsjtu.2022.316

摘要

极地船舶冰区航行时,冰阻力的准确预报在保障船舶航行安全方面起着重要作用.近年来,机器学习在船舶方面的应用越来越广泛,其中,人工神经网络(ANN)是机器学习领域中一种常用的方法.本文的重点是设计一个用于预报极地船舶冰阻力的ANN模型.参考传统的经验和半经验公式,选择合适的输入特征参数,通过大量的船舶模型试验数据来训练神经网络,搭建径向基(RBF)神经网络模型,并选用遗传算法(GA)进行模型优化.研究表明,基于7个特征参数输入的遗传算法优化径向基(RBF-GA)神经网络模型具有良好的泛化效果,与模型试验和实船试验数据对比,平均误差在8%左右,具有较高的精度,可作为冰阻力预报工具.

SUN

Q Y

, ZHOU

, DING

S F

, et al. An artificial neural network-based method for prediction of ice resistance of polar ships[J]. Journal of Shanghai Jiao Tong University, 2024, 58(2): 156-165.

本文引用 [1]

[37]

王煦莹, 沈红波, 徐兴周. 基于人工神经网络的金融信息系统风险评价[J]. 信息安全研究, 2022, 8(11):1055-1060.

WANG

X Y

, SHEN

H B

, XU

X Z

. Financial information system risk assessment based on artificial neural network[J]. Journal of Information Security Research, 2022, 8(11): 1055-1060.

本文引用 [1] 摘要

With the rapid development of financial informatization, the risk management of financial information system is becoming more and more important. The core of risk management is risk assessment, which requires scientific assessment of information system risks of financial institutions. This paper analyzes the results of BP artificial neural network by using the artificial neural network algorithm under the condition of big data, and uses the information systems developed by 60 financial institutions at the end of 2021 as samples for experimental verification. The experimental results show that the artificial neural network has high correlation and low relative error, and the numerical fitting effect is good. The risk assessment model of financial information system based on artificial neural network is feasible, which provides a powerful demonstration for the application of big data and artificial neural network in financial information system.

[38]	郭晓萌, 方秀琴, 杨露露, 等. 基于人工神经网络的西辽河流域根区土壤湿度估算[J]. 自然资源遥感, 2023, 35(2):193-201. GUO X M, FANG X Q, YANG L L, et al. Artificial neural network-based estimation of root zone soil moisture in the western Liaohe Rriver Basin[J]. Remote Sensing for Natural Resources, 2023, 35(2): 193-201. 本文引用 [1]

[39]	刘志丰, 丁锋. 基于人工神经网络的沿海风速多步预测研究[J]. 气象科技, 2022, 50(6):851-858. LIU Z F, DING F. Research on application of artificial neural network for sea surface wind speed forecasting[J]. Meteorological Science and Technology, 2022, 50(6): 851-858. 本文引用 [1]

[40]	蒋晨凯, 章四龙. 基于贝叶斯优化与人工神经网络的秦淮河流域水位预报[J]. 水电能源科学, 2022, 40(9):48-51,60. JIANG C K, ZHANG S L. Water level forecasting of Qinhuaihe River Basin based on Bayesian optimization and artificial neural network[J]. Water Resources and Power, 2022, 40(9): 48-51, 60. 本文引用 [1]

[41]

曹若馨, 张可欣, 曾维华, 等. 基于BP神经网络的水环境承载力预警研究:以北运河为例[J]. 环境科学学报, 2021, 41(5):2005-2017.

CAO

R X

, ZHANG

K X

, ZENG

W H

, et al. Research on the early-warning method of water environment carrying capacity based on BP neural network: A case study of Beiyunhe River Basin[J]. Acta Scientiae Circumstantiae, 2021, 41(5): 2005-2017.

本文引用 [1]

[42]	张秀菊, 王柳林, 李秀平, 等. 基于BP神经网络的潇河流域水质预测[J]. 水资源与水工程学报, 2021, 32(5):19-26. ZHANG X J, WANG L L, LI X P, et al. Water quality prediction of the Xiaohe River Basin based on BP neural network model[J]. Journal of Water Resources and Water Engineering, 2021, 32(5): 19-26. 本文引用 [1]

[43]	安慧, 范历娟, 吴海林, 等. 基于BP神经网络的淮河流域水生态足迹分析与预测[J]. 长江流域资源与环境, 2021, 30(5):1076-1087. AN H, FAN L J, WU H L, et al. Analysis and prediction of water ecological footprint of Huaihe River Basin based on BP neural network[J]. Resources and Environment in the Yangtze Basin, 2021, 30(5): 1076-1087. 本文引用 [1]