Xingou Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Xingou Xu has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Ad Stoffelen
Satellite Observations, Royal Netherlands Meteorology Institute, De Bilt, 3730 AE, the Netherlands
12 shared publications - Xiaolong Dong
National Space Science Center, Chinese Academy of Sciences, Beijing, China
8 shared publications Di Zhu
Key Laboratory of Microwave Remote Sensing, National Space Science Center, Chinese Academy of Sciences, Beijing, China
6 shared publications- Wenming Lin
School of Marine Sciences, Nanjing University of Information Science and Technology, Nanjing, China
5 shared publications Shuyan Lang
National Satellite Ocean Application Service, Beijing, China
4 shared publicationsLei Zhang
DFH Satellite co., Ltd, Beijing, 100080, China
2 shared publications
Latest Publications
- Open Access
- 6 November 2025
- Published by Copernicus GmbH in Atmospheric Measurement Techniques
- Vol. 18 (21) , 6329-6344
An improved one-dimensional ocean wave description based on SWIM observations
The one-dimensional ocean wave spectra (1D spectra) describing the total energy of the ocean waves are vital for providing ocean surface roughness information in remote sensing simulations. Most existing wave spectrum models deviate from real ocean surface descriptions due to limitations of observation methods and approximation in theories applied to generating them. In this research, the widely applied Goda and Elfouhaily spectra in their 1D form are compared with the remote sensing products from the Surface Waves Investigation and Monitoring instrument (SWIM) on-board the China France Oceanography Satellite (CFOSAT). Differences between models and the measurements are addressed, and the causes are analysed before concluded in terms of sea states. Then, a Combined spectrum (C spectrum) considering varied sea states is proposed as a closer model to the observations of the real sea, where parameterization of the spectral peak enhancement factor (γ) is achieved by the inverse wave age and wave steepness for multiple sea states. Then the specific values of the all-state sea are obtained from SWIM observations. The validation of the C spectrum is achieved by comparisons with SWIM measurements not utilized during the model establishment, and with buoy measurements. The difference index (DI) and the R-squared (R2), are calculated for evaluation of the results, indicating that the C spectrum demonstrates closer fitting to the SWIM and buoy measurements than both Goda and Elfouhaily spectra. The DI and R2 for the C spectrum are compared to the Goda spectrum, which is closer to SWIM measurements than E spectrum. Results suggest the C spectrum is suitable for providing information required for remote sensing applications.
- 6 November 2025
- Published by Copernicus GmbH in Atmospheric Measurement Techniques
- Vol. 18 (21) , 6329-6344
Open AccessAn improved one-dimensional ocean wave description based on SWIM observations
The one-dimensional ocean wave spectra (1D spectra) describing the total energy of the ocean waves are vital for providing ocean surface roughness information in remote sensing simulations. Most existing wave spectrum models deviate from real ocean surface descriptions due to limitations of observation methods and approximation in theories applied to generating them. In this research, the widely applied Goda and Elfouhaily spectra in their 1D form are compared with the remote sensing products from the Surface Waves Investigation and Monitoring instrument (SWIM) on-board the China France Oceanography Satellite (CFOSAT). Differences between models and the measurements are addressed, and the causes are analysed before concluded in terms of sea states. Then, a Combined spectrum (C spectrum) considering varied sea states is proposed as a closer model to the observations of the real sea, where parameterization of the spectral peak enhancement factor (γ) is achieved by the inverse wave age and wave steepness for multiple sea states. Then the specific values of the all-state sea are obtained from SWIM observations. The validation of the C spectrum is achieved by comparisons with SWIM measurements not utilized during the model establishment, and with buoy measurements. The difference index (DI) and the R-squared (R2), are calculated for evaluation of the results, indicating that the C spectrum demonstrates closer fitting to the SWIM and buoy measurements than both Goda and Elfouhaily spectra. The DI and R2 for the C spectrum are compared to the Goda spectrum, which is closer to SWIM measurements than E spectrum. Results suggest the C spectrum is suitable for providing information required for remote sensing applications.
- Open Access
- 18 July 2025
- Published by Copernicus GmbH
- Vol. 2025, 1-17
Comment on egusphere-2024-3840
Ocean surface sea ice screening is important for a wide variety of research and applications. While the normalized radar cross-sections (NRCS) have been long applied in sea ice classifications and quantifications, the beam-filling problem still exists. Besides, for the melting ice cover, the NRCS are not providing accurate links to sea ice. However, if considering the NRCS within the traditional wind retrieval unit, despite the existence of fractal sea ice and the different mixing quantities of ice and sea water, there will be signatures in both the derivations from wind retrieval model and the extend of heterogeneity within the unit. The former has been well researched and together with the ice model from scatteroemters, applied now in ice labelling for scatterometer operational wind products in the indicator MLE. In this research, we uncover the properties of the later due to sea ice, in terms of the indicator Joss. The sea ice and iceberg concentration from products derived in AMSR-2 and Sentinel-1 are applied as references. The scatterometer data are from collocations of C- and Ku-band scatterometer products from ASCAT-A, ASCAT-B, OSCAT-2 and HSCAT. The ice screening ability in combination of MLE and Joss are concluded. Finally, the iceberg identification potential is discussed along with the application of the joint observations of C- and Ku-band scatteroemter observations.
- 18 July 2025
- Published by Copernicus GmbH
- Vol. 2025, 1-17
Open AccessComment on egusphere-2024-3840
Ocean surface sea ice screening is important for a wide variety of research and applications. While the normalized radar cross-sections (NRCS) have been long applied in sea ice classifications and quantifications, the beam-filling problem still exists. Besides, for the melting ice cover, the NRCS are not providing accurate links to sea ice. However, if considering the NRCS within the traditional wind retrieval unit, despite the existence of fractal sea ice and the different mixing quantities of ice and sea water, there will be signatures in both the derivations from wind retrieval model and the extend of heterogeneity within the unit. The former has been well researched and together with the ice model from scatteroemters, applied now in ice labelling for scatterometer operational wind products in the indicator MLE. In this research, we uncover the properties of the later due to sea ice, in terms of the indicator Joss. The sea ice and iceberg concentration from products derived in AMSR-2 and Sentinel-1 are applied as references. The scatterometer data are from collocations of C- and Ku-band scatterometer products from ASCAT-A, ASCAT-B, OSCAT-2 and HSCAT. The ice screening ability in combination of MLE and Joss are concluded. Finally, the iceberg identification potential is discussed along with the application of the joint observations of C- and Ku-band scatteroemter observations.
- Open Access
- 26 May 2025
- Published by Copernicus GmbH
- Vol. 2025, 1-17
Comment on egusphere-2024-3840
Ocean surface sea ice screening is important for a wide variety of research and applications. While the normalized radar cross-sections (NRCS) have been long applied in sea ice classifications and quantifications, the beam-filling problem still exists. Besides, for the melting ice cover, the NRCS are not providing accurate links to sea ice. However, if considering the NRCS within the traditional wind retrieval unit, despite the existence of fractal sea ice and the different mixing quantities of ice and sea water, there will be signatures in both the derivations from wind retrieval model and the extend of heterogeneity within the unit. The former has been well researched and together with the ice model from scatteroemters, applied now in ice labelling for scatterometer operational wind products in the indicator MLE. In this research, we uncover the properties of the later due to sea ice, in terms of the indicator Joss. The sea ice and iceberg concentration from products derived in AMSR-2 and Sentinel-1 are applied as references. The scatterometer data are from collocations of C- and Ku-band scatterometer products from ASCAT-A, ASCAT-B, OSCAT-2 and HSCAT. The ice screening ability in combination of MLE and Joss are concluded. Finally, the iceberg identification potential is discussed along with the application of the joint observations of C- and Ku-band scatteroemter observations.
- 26 May 2025
- Published by Copernicus GmbH
- Vol. 2025, 1-17
Open AccessComment on egusphere-2024-3840
Ocean surface sea ice screening is important for a wide variety of research and applications. While the normalized radar cross-sections (NRCS) have been long applied in sea ice classifications and quantifications, the beam-filling problem still exists. Besides, for the melting ice cover, the NRCS are not providing accurate links to sea ice. However, if considering the NRCS within the traditional wind retrieval unit, despite the existence of fractal sea ice and the different mixing quantities of ice and sea water, there will be signatures in both the derivations from wind retrieval model and the extend of heterogeneity within the unit. The former has been well researched and together with the ice model from scatteroemters, applied now in ice labelling for scatterometer operational wind products in the indicator MLE. In this research, we uncover the properties of the later due to sea ice, in terms of the indicator Joss. The sea ice and iceberg concentration from products derived in AMSR-2 and Sentinel-1 are applied as references. The scatterometer data are from collocations of C- and Ku-band scatterometer products from ASCAT-A, ASCAT-B, OSCAT-2 and HSCAT. The ice screening ability in combination of MLE and Joss are concluded. Finally, the iceberg identification potential is discussed along with the application of the joint observations of C- and Ku-band scatteroemter observations.
- Open Access
- 14 January 2025
- Published by Copernicus GmbH in EGUsphere
- Vol. 2025, 1-17
Sea Ice Screening Ability in Ku Band and C Band Wind Scatterometry
Ocean surface sea ice screening is important for a wide variety of research and applications. While the normalized radar cross-sections (NRCS) have been long applied in sea ice classifications and quantifications, the beam-filling problem still exists. Besides, for the melting ice cover, the NRCS are not providing accurate links to sea ice. However, if considering the NRCS within the traditional wind retrieval unit, despite the existence of fractal sea ice and the different mixing quantities of ice and sea water, there will be signatures in both the derivations from wind retrieval model and the extend of heterogeneity within the unit. The former has been well researched and together with the ice model from scatteroemters, applied now in ice labelling for scatterometer operational wind products in the indicator MLE. In this research, we uncover the properties of the later due to sea ice, in terms of the indicator Joss. The sea ice and iceberg concentration from products derived in AMSR-2 and Sentinel-1 are applied as references. The scatterometer data are from collocations of C- and Ku-band scatterometer products from ASCAT-A, ASCAT-B, OSCAT-2 and HSCAT. The ice screening ability in combination of MLE and Joss are concluded. Finally, the iceberg identification potential is discussed along with the application of the joint observations of C- and Ku-band scatteroemter observations.
- 14 January 2025
- Published by Copernicus GmbH in EGUsphere
- Vol. 2025, 1-17
Open AccessSea Ice Screening Ability in Ku Band and C Band Wind Scatterometry
Ocean surface sea ice screening is important for a wide variety of research and applications. While the normalized radar cross-sections (NRCS) have been long applied in sea ice classifications and quantifications, the beam-filling problem still exists. Besides, for the melting ice cover, the NRCS are not providing accurate links to sea ice. However, if considering the NRCS within the traditional wind retrieval unit, despite the existence of fractal sea ice and the different mixing quantities of ice and sea water, there will be signatures in both the derivations from wind retrieval model and the extend of heterogeneity within the unit. The former has been well researched and together with the ice model from scatteroemters, applied now in ice labelling for scatterometer operational wind products in the indicator MLE. In this research, we uncover the properties of the later due to sea ice, in terms of the indicator Joss. The sea ice and iceberg concentration from products derived in AMSR-2 and Sentinel-1 are applied as references. The scatterometer data are from collocations of C- and Ku-band scatterometer products from ASCAT-A, ASCAT-B, OSCAT-2 and HSCAT. The ice screening ability in combination of MLE and Joss are concluded. Finally, the iceberg identification potential is discussed along with the application of the joint observations of C- and Ku-band scatteroemter observations.
- 20 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Magazine
- Vol. 13 (1) , 153-174
Wind and Sea State Signatures in Wind Scatterometry: An analysis
A better understanding of how wind vectors are obtained in scatterometer retrievals may be useful to further develop wind scatterometry. Scatterometer ocean surface wind estimation involves air-sea and wind-wave interactions in rather diverse conditions. Curiously, scatterometer Geophysical Model Functions only depending on the wind vector prove very successful operationally, and sea state (ocean wave) parameters are not considered. It has been suggested in theories that, on one hand, wind-wave features are modulated by fetch or wave age and affected by longer waves or swells in terms of non-linear interactions and, on the other, direct wind-induced wave movements in the short ocean-wave spectrum, modulated by gusty winds. Though generally supported by limited sampling, some theories are associated with in-situ observations and laboratory experiments, scatterometer wind-wave-induced ocean roughness back-scattered signals can be associated with winds quite accurately, without considering the wave parameters. In this research, sea-state and wind variability parameters are specifically investigated for their effect on the scatterometer measurements and retrieved winds. Results confirm that the scatterometer backscatter values are generally not affected by non-local sea state and this is further discussed. Finally, an extreme wind case is provided for a better demonstration. The conclusions provide good references for the high-wind model modification and the next generation of scatterometers designed for simultaneous measurements of ocean surface wind, wave motion and ocean currents.
- 20 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Magazine
- Vol. 13 (1) , 153-174
Wind and Sea State Signatures in Wind Scatterometry: An analysis
A better understanding of how wind vectors are obtained in scatterometer retrievals may be useful to further develop wind scatterometry. Scatterometer ocean surface wind estimation involves air-sea and wind-wave interactions in rather diverse conditions. Curiously, scatterometer Geophysical Model Functions only depending on the wind vector prove very successful operationally, and sea state (ocean wave) parameters are not considered. It has been suggested in theories that, on one hand, wind-wave features are modulated by fetch or wave age and affected by longer waves or swells in terms of non-linear interactions and, on the other, direct wind-induced wave movements in the short ocean-wave spectrum, modulated by gusty winds. Though generally supported by limited sampling, some theories are associated with in-situ observations and laboratory experiments, scatterometer wind-wave-induced ocean roughness back-scattered signals can be associated with winds quite accurately, without considering the wave parameters. In this research, sea-state and wind variability parameters are specifically investigated for their effect on the scatterometer measurements and retrieved winds. Results confirm that the scatterometer backscatter values are generally not affected by non-local sea state and this is further discussed. Finally, an extreme wind case is provided for a better demonstration. The conclusions provide good references for the high-wind model modification and the next generation of scatterometers designed for simultaneous measurements of ocean surface wind, wave motion and ocean currents.
- Open Access
- 14 May 2024
- Published by Copernicus GmbH in EGUsphere
- Vol. 2024, 1-23
An Improvement of the One-dimensional Ocean Wave Description based on SWIM Observations
The one-dimensional ocean wave spectra (1D spectra) describing the total energy of the ocean waves are vital for providing ocean surface roughness information in remote sensing simulations. Most existing wave spectrum models deviate from real ocean surface descriptions due to limitations of observation methods and approximation in theories applied to generating them. In this research, the widely applied Goda and Elfouhaily spectra in their 1-D form are compared with the remote sensing products from the Surface Waves Investigation and Monitoring instrument (SWIM) on-board the China France Oceanography Satellite (CFOSAT). Differences between models and the measurements are addressed, then the causes are analyzed and concluded in terms of sea states. Then, a Combined spectrum (C spectrum) considering varied sea states is proposed as a closer model to the observations of the real sea, where parameterization of the spectral peak enhancement factor (γ) is achieved by the inverse wave age and wave steepness for multiple sea states. Then the specific values of the all-state sea are obtained from SWIM observations. The validation of the C spectrum is achieved by comparisons with SWIM measurements not utilized during the model establishment, and with buoy measurements. The difference index (DI) and the R-squared (R2), are calculated for evaluation of the results, indicating that the C spectrum demonstrates closer fitting to the SWIM and buoy measurements than both Goda and Elfouhaily spectra. The DI and R2 for the C spectrum are compared to the Goda spectrum, which is closer to SWIM measurements than E spectrum, and values are 0.780 and 0.909 respectively. Results suggest the C spectrum is suitable for providing information required for remote sensing applications. Further research would be focusing on implementing description in different azimuthal directions.
- 14 May 2024
- Published by Copernicus GmbH in EGUsphere
- Vol. 2024, 1-23
Open AccessAn Improvement of the One-dimensional Ocean Wave Description based on SWIM Observations
The one-dimensional ocean wave spectra (1D spectra) describing the total energy of the ocean waves are vital for providing ocean surface roughness information in remote sensing simulations. Most existing wave spectrum models deviate from real ocean surface descriptions due to limitations of observation methods and approximation in theories applied to generating them. In this research, the widely applied Goda and Elfouhaily spectra in their 1-D form are compared with the remote sensing products from the Surface Waves Investigation and Monitoring instrument (SWIM) on-board the China France Oceanography Satellite (CFOSAT). Differences between models and the measurements are addressed, then the causes are analyzed and concluded in terms of sea states. Then, a Combined spectrum (C spectrum) considering varied sea states is proposed as a closer model to the observations of the real sea, where parameterization of the spectral peak enhancement factor (γ) is achieved by the inverse wave age and wave steepness for multiple sea states. Then the specific values of the all-state sea are obtained from SWIM observations. The validation of the C spectrum is achieved by comparisons with SWIM measurements not utilized during the model establishment, and with buoy measurements. The difference index (DI) and the R-squared (R2), are calculated for evaluation of the results, indicating that the C spectrum demonstrates closer fitting to the SWIM and buoy measurements than both Goda and Elfouhaily spectra. The DI and R2 for the C spectrum are compared to the Goda spectrum, which is closer to SWIM measurements than E spectrum, and values are 0.780 and 0.909 respectively. Results suggest the C spectrum is suitable for providing information required for remote sensing applications. Further research would be focusing on implementing description in different azimuthal directions.
- Open Access
- 30 November 2021
- Published by Copernicus GmbH in Atmospheric Measurement Techniques
- Vol. 14 (12) , 7435-7451
Support vector machine tropical wind speed retrieval in the presence of rain for Ku-band wind scatterometry
Wind retrieval parameters, i.e. quality indicators and the two-dimensional variational ambiguity removal (2DVAR) analysis speeds, are explored with the aim to improve wind speed retrieval during rain for tropical regions. We apply the well-researched support vector machine (SVM) method in machine learning (ML) to solve this complex problem in a data-oriented regression. To guarantee the effectiveness of SVM, the inputs are extensively analysed to evaluate their appropriateness for this problem, before the results are produced. The comparisons between distributions and differences between data of rain-contaminated winds, corrected winds and good quality C-band winds illustrate that the rain-distorted wind distributions become more nominal with SVM, hence much reducing the rain-induced biases and error variance. Further confirmation is obtained from a case with synchronous Himawari-8 observation indicating rain (clouds) in the scene. Furthermore, the estimation of simultaneous rain rate is attempted with some success to retrieve both wind and rain. Although additional observations or higher resolution may be required to better assess the accuracy of the wind and rain retrievals, the ML results demonstrate benefits of such methodology in geophysical retrieval and nowcasting applications.
- 30 November 2021
- Published by Copernicus GmbH in Atmospheric Measurement Techniques
- Vol. 14 (12) , 7435-7451
Open AccessSupport vector machine tropical wind speed retrieval in the presence of rain for Ku-band wind scatterometry
Wind retrieval parameters, i.e. quality indicators and the two-dimensional variational ambiguity removal (2DVAR) analysis speeds, are explored with the aim to improve wind speed retrieval during rain for tropical regions. We apply the well-researched support vector machine (SVM) method in machine learning (ML) to solve this complex problem in a data-oriented regression. To guarantee the effectiveness of SVM, the inputs are extensively analysed to evaluate their appropriateness for this problem, before the results are produced. The comparisons between distributions and differences between data of rain-contaminated winds, corrected winds and good quality C-band winds illustrate that the rain-distorted wind distributions become more nominal with SVM, hence much reducing the rain-induced biases and error variance. Further confirmation is obtained from a case with synchronous Himawari-8 observation indicating rain (clouds) in the scene. Furthermore, the estimation of simultaneous rain rate is attempted with some success to retrieve both wind and rain. Although additional observations or higher resolution may be required to better assess the accuracy of the wind and rain retrievals, the ML results demonstrate benefits of such methodology in geophysical retrieval and nowcasting applications.
- Open Access
- 27 August 2021
- Published by Copernicus GmbH
- p. 1-23
Reply on RC1
Wind retrieval parameters, i.e., quality indicators and the 2DVAR analysis speeds, are explored with the aim to improve wind speed retrieval during rain for tropical regions. We apply the well-researched support vector machine (SVM) method in machine learning (ML) to solve this complex problem in a data-orientated regression. To guarantee the effectiveness of SVM, the inputs are extensively analysed to evaluate their appropriateness for this problem, before the results are produced. Subsequently, triple collocation shows that the similarity of the resolved Ku-band (OSCAT-2) wind speed in rain is better than the 2DVAR speed, with respect to the collocated C-band (ASCAT) speed, which is much less affected by rain. The comparisons between distributions and differences between data of rain-contaminated winds, corrected winds and good quality C-band winds, illustrate that the rain-distorted wind distributions become more nominal with SVM, hence eliminating rain-induced biases and error variance. Further confirmation is obtained from a case with synchronous Himawari-8 observation indicating rain (clouds) in the scene. Furthermore, the determination of simultaneous rain rate is attempted to retrieve both wind and rain. Although, additional observations or higher resolution may be required to better assess the accuracy of the wind and rain retrievals, the Machine Learning (ML) results demonstrate benefits of such methodology in geophysical retrieval and nowcasting applications.
- 27 August 2021
- Published by Copernicus GmbH
- p. 1-23
Open AccessReply on RC1
Wind retrieval parameters, i.e., quality indicators and the 2DVAR analysis speeds, are explored with the aim to improve wind speed retrieval during rain for tropical regions. We apply the well-researched support vector machine (SVM) method in machine learning (ML) to solve this complex problem in a data-orientated regression. To guarantee the effectiveness of SVM, the inputs are extensively analysed to evaluate their appropriateness for this problem, before the results are produced. Subsequently, triple collocation shows that the similarity of the resolved Ku-band (OSCAT-2) wind speed in rain is better than the 2DVAR speed, with respect to the collocated C-band (ASCAT) speed, which is much less affected by rain. The comparisons between distributions and differences between data of rain-contaminated winds, corrected winds and good quality C-band winds, illustrate that the rain-distorted wind distributions become more nominal with SVM, hence eliminating rain-induced biases and error variance. Further confirmation is obtained from a case with synchronous Himawari-8 observation indicating rain (clouds) in the scene. Furthermore, the determination of simultaneous rain rate is attempted to retrieve both wind and rain. Although, additional observations or higher resolution may be required to better assess the accuracy of the wind and rain retrievals, the Machine Learning (ML) results demonstrate benefits of such methodology in geophysical retrieval and nowcasting applications.
- Open Access
- 17 August 2021
- Published by Copernicus GmbH
- p. 1-23
Comment on amt-2021-200
Wind retrieval parameters, i.e., quality indicators and the 2DVAR analysis speeds, are explored with the aim to improve wind speed retrieval during rain for tropical regions. We apply the well-researched support vector machine (SVM) method in machine learning (ML) to solve this complex problem in a data-orientated regression. To guarantee the effectiveness of SVM, the inputs are extensively analysed to evaluate their appropriateness for this problem, before the results are produced. Subsequently, triple collocation shows that the similarity of the resolved Ku-band (OSCAT-2) wind speed in rain is better than the 2DVAR speed, with respect to the collocated C-band (ASCAT) speed, which is much less affected by rain. The comparisons between distributions and differences between data of rain-contaminated winds, corrected winds and good quality C-band winds, illustrate that the rain-distorted wind distributions become more nominal with SVM, hence eliminating rain-induced biases and error variance. Further confirmation is obtained from a case with synchronous Himawari-8 observation indicating rain (clouds) in the scene. Furthermore, the determination of simultaneous rain rate is attempted to retrieve both wind and rain. Although, additional observations or higher resolution may be required to better assess the accuracy of the wind and rain retrievals, the Machine Learning (ML) results demonstrate benefits of such methodology in geophysical retrieval and nowcasting applications.
- 17 August 2021
- Published by Copernicus GmbH
- p. 1-23
Open AccessComment on amt-2021-200
Wind retrieval parameters, i.e., quality indicators and the 2DVAR analysis speeds, are explored with the aim to improve wind speed retrieval during rain for tropical regions. We apply the well-researched support vector machine (SVM) method in machine learning (ML) to solve this complex problem in a data-orientated regression. To guarantee the effectiveness of SVM, the inputs are extensively analysed to evaluate their appropriateness for this problem, before the results are produced. Subsequently, triple collocation shows that the similarity of the resolved Ku-band (OSCAT-2) wind speed in rain is better than the 2DVAR speed, with respect to the collocated C-band (ASCAT) speed, which is much less affected by rain. The comparisons between distributions and differences between data of rain-contaminated winds, corrected winds and good quality C-band winds, illustrate that the rain-distorted wind distributions become more nominal with SVM, hence eliminating rain-induced biases and error variance. Further confirmation is obtained from a case with synchronous Himawari-8 observation indicating rain (clouds) in the scene. Furthermore, the determination of simultaneous rain rate is attempted to retrieve both wind and rain. Although, additional observations or higher resolution may be required to better assess the accuracy of the wind and rain retrievals, the Machine Learning (ML) results demonstrate benefits of such methodology in geophysical retrieval and nowcasting applications.
- Open Access
- 26 July 2021
- Published by Copernicus GmbH in EGUsphere
- Vol. 2021, 1-23
Support vector machine tropical wind speed retrieval in the presence of rain for Ku-band wind scatterometry
Wind retrieval parameters, i.e., quality indicators and the 2DVAR analysis speeds, are explored with the aim to improve wind speed retrieval during rain for tropical regions. We apply the well-researched support vector machine (SVM) method in machine learning (ML) to solve this complex problem in a data-orientated regression. To guarantee the effectiveness of SVM, the inputs are extensively analysed to evaluate their appropriateness for this problem, before the results are produced. Subsequently, triple collocation shows that the similarity of the resolved Ku-band (OSCAT-2) wind speed in rain is better than the 2DVAR speed, with respect to the collocated C-band (ASCAT) speed, which is much less affected by rain. The comparisons between distributions and differences between data of rain-contaminated winds, corrected winds and good quality C-band winds, illustrate that the rain-distorted wind distributions become more nominal with SVM, hence eliminating rain-induced biases and error variance. Further confirmation is obtained from a case with synchronous Himawari-8 observation indicating rain (clouds) in the scene. Furthermore, the determination of simultaneous rain rate is attempted to retrieve both wind and rain. Although, additional observations or higher resolution may be required to better assess the accuracy of the wind and rain retrievals, the Machine Learning (ML) results demonstrate benefits of such methodology in geophysical retrieval and nowcasting applications.
- 26 July 2021
- Published by Copernicus GmbH in EGUsphere
- Vol. 2021, 1-23
Open AccessSupport vector machine tropical wind speed retrieval in the presence of rain for Ku-band wind scatterometry
Wind retrieval parameters, i.e., quality indicators and the 2DVAR analysis speeds, are explored with the aim to improve wind speed retrieval during rain for tropical regions. We apply the well-researched support vector machine (SVM) method in machine learning (ML) to solve this complex problem in a data-orientated regression. To guarantee the effectiveness of SVM, the inputs are extensively analysed to evaluate their appropriateness for this problem, before the results are produced. Subsequently, triple collocation shows that the similarity of the resolved Ku-band (OSCAT-2) wind speed in rain is better than the 2DVAR speed, with respect to the collocated C-band (ASCAT) speed, which is much less affected by rain. The comparisons between distributions and differences between data of rain-contaminated winds, corrected winds and good quality C-band winds, illustrate that the rain-distorted wind distributions become more nominal with SVM, hence eliminating rain-induced biases and error variance. Further confirmation is obtained from a case with synchronous Himawari-8 observation indicating rain (clouds) in the scene. Furthermore, the determination of simultaneous rain rate is attempted to retrieve both wind and rain. Although, additional observations or higher resolution may be required to better assess the accuracy of the wind and rain retrievals, the Machine Learning (ML) results demonstrate benefits of such methodology in geophysical retrieval and nowcasting applications.