the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 26 Mar 2026
Improving offshore wind data from reanalyses using ship-based lidar measurements
Abstract. This study addresses the challenge of integrating ship-based lidar measurements with numerical weather prediction models to improve offshore wind characterisation. Accurate wind measurements are vital for the development of offshore wind energy, yet traditionally used fixed devices, such as meteorological masts and platform- or buoy-based lidars, are expensive and scarce. Ship-based lidar systems offer a flexible, cost-effective alternative by collecting wind data over large areas; however, the non-stationarity of ships results in low data density at any specific location. To overcome this challenge, we propose a novel calibration methodology to assimilate ship-mounted lidar observations into the ERA5 reanalysis by statistically adjusting its wind speed outputs. Inspired by observational nudging, which influences model state variables over time to match observational data, our approach applies a weighted correction directly to the model’s wind speed output, preserving the model’s underlying physics while ensuring computational efficiency and flexibility. The calibration parameters, including calibration strength, temporal window, and spatial radius of influence, were optimised to maximise the impact and accuracy of the calibration process. The comparison between ERA5 before and after the calibration demonstrates that the methodology effectively reduces the systematic underestimation of wind speeds, particularly in coastal regions where ERA5 struggles with complex flow dynamics. The methodology has been validated against independent measurements from a fixed Doppler lidar system deployed on an island in the northern Baltic Sea, demonstrating the calibration’s effectiveness in reducing bias and error spread at this location as well. However, it highlights that the calibration effect is strongly dependent on the distance between the ship and the lidar station, with a bias reduction of 0.2 m s-1 when the ship is within 60 km, compared to 0.05 m s-1 when considering data within 90 km, as a consequence of the intermittent influence of the ship-based lidar data.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Wind Energy Science.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(5517 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on wes-2026-50', Anonymous Referee #1, 20 Apr 2026
The comment was uploaded in the form of a supplement: https://wes.copernicus.org/preprints/wes-2026-50/wes-2026-50-RC1-supplement.pdfCitation: https://doi.org/
10.5194/wes-2026-50-RC1 - RC2: 'Comment on wes-2026-50', Anonymous Referee #2, 07 May 2026
-
RC3: 'Comment on wes-2026-50', Anonymous Referee #3, 26 Jun 2026
The journal paper “Improving offshore wind data from reanalyses using ship-based lidar measurements” by Rubio et al., proposes a methodology to correct ERA5 model outputs using offshore observations from a shipborne observational dataset. Although, in general observations are crucial for offshore model validations or improvements, the proposed methodology is too simplistic and has some issues that the reviewer feels the authors are ignoring.
Below are some major comments:
1. Fundamental limits of “Post‑Processing” calibration
The methodology proposed in the paper adjusts ERA5 wind outputs retrospectively, rather than addressing the underlying model physics or integrating corrections through data assimilation at an earlier stage. As a result, this approach is unable to rectify fundamental structural biases inherent in the model, such as those associated with surface layer parameterizations, representation of coastal flows, or stability-related inaccuracies. Furthermore, the corrections applied are confined to specific locations and moments in time, lacking the spatial and temporal propagation characteristics of genuine data assimilation techniques. This limitation means that ERA5 retains internal inconsistencies, with only wind speed being corrected while other crucial variables, including wind direction, vertical shear, turbulence, and stability remain unaltered. Consequently, the calibrated product may lack internal physical consistency, which is particularly problematic for applications that require reliable multi-variable relationships, such as load calculations, wake modeling, and comprehensive resource assessments.
2. Dependence on non‑stationary and intermittent ship data
Ship-based lidar data are not collected at fixed points, instead, measurements are acquired as the vessel moves along its route. This movement introduces spatial and temporal variability in the sampling process, resulting in uneven coverage across the study area. Additionally, the proximity to the coastline and the operational schedule of the ferry lead to sampling biases. For example, there is a concentration of data points near harbors where the ship spends more time, while data are sparse in offshore regions where the vessel passes less frequently. As a result, the calibration effect achieved through these measurements is localized and intermittent, as noted by the authors. This means that corrections are strongest along the ship's track but diminish in areas farther away, both in space and time. This situation creates a “patchwork” calibration pattern, where the signal is strong near the ship’s route but weak elsewhere. Such a limitation poses significant challenges for regional wind resource assessments, as it undermines the ability to achieve consistent corrections over the broader domain.
3 Correlation-based vs. Physics-based correction
The methodology described in the paper relies on a weighted statistical correction to adjust model outputs, rather than employing a physically constrained data assimilation process. This means that instead of integrating observations into the model in a way that respects the underlying physical laws and constraints, the approach uses statistical relationships to determine corrections. A notable issue with this method is that the weighting radii and time windows used for the corrections are chosen empirically, without physical justification. As a result, the selection of these parameters is not guided by an understanding of the relevant atmospheric processes, but rather by statistical or practical considerations. There is no guarantee that the corrections applied through this approach preserve essential physical balances, such as mass continuity or momentum budgets. Because the corrections are not informed by physical constraints, the adjusted model output may violate these underlying principles. Furthermore, improvements achieved at one location may inadvertently degrade the accuracy at nearby grid points. This is due to the decorrelation between the corrections applied at different points, which means that the statistical adjustments do not propagate in a physically consistent manner across the model domain. As a result, the approach is vulnerable to overfitting, noise, and inconsistency. Since corrections are based on local statistical relationships rather than robust physical mechanisms, there is a risk that the method may respond excessively to random fluctuations in the observational data, leading to unreliable and inconsistent results.
4. Sensitivity to measurement errors
Ship-based lidar data are imperfect and introduce several sources of uncertainty into the calibration process. Motion correction algorithms, which are necessary due to the movement of the vessel, inherently add uncertainty to the measurements. Even though Carrier-to-Noise Ratio (CNR) filtering is applied to remove degraded data, it reduces the overall availability of observations and cannot fully eliminate data quality issues. Errors such as instrument tilt and contamination from precipitation can directly affect the calibration results. These measurement errors are not always adequately filtered or corrected, meaning that they can imprint themselves onto the model calibration. Because the methodology applies corrections based explicitly on the differences between observations and model outputs, any error present in the lidar data is transferred directly into the model, potentially degrading its accuracy.
5. Spatial representativeness problems
Lidar observations, which are essentially point or near-point measurements, are being used to adjust the mean values of entire ERA5 grid cells. This poses a fundamental representativeness issue because the spatial footprint of lidar measurements is quite small, whereas ERA5 grid cells cover much larger areas. In regions with pronounced coastal and mesoscale gradients, the lidar data may not accurately reflect the conditions across the broader grid cell. This mismatch is particularly problematic when there are significant spatial variations within the cell, as the localized lidar readings may not capture the heterogeneity present. The methodology’s reliance on the Cressman radius-of-influence approach further complicates matters. This technique assumes isotropic and stationary conditions, which are not valid for coastal flows that often exhibit strong directionality and variability. As a result, the calibration process can misrepresent the spatial structure of the wind field, especially where mesoscale heterogeneity is pronounced. These factors contribute to the risk of miscalibration, particularly in areas with complex spatial variability. When mesoscale heterogeneity is strong, adjustments based on small footprint lidar observations may fail to accurately improve the broader grid cell, potentially introducing errors rather than correcting them.
6. Scalability and generalizability issues
For broader application of the calibration methodology, several practical limitations emerge that affect its scalability and generalizability. Effective calibration requires either dense ship traffic or multiple coordinated vessels to provide adequate spatial and temporal coverage. In many marine environments, this level of observational density is unrealistic, making widespread implementation impractical. Furthermore, the calibration is valid only within approximately 60–90 km of the ship database. Areas beyond this range receive no benefit from the correction process, limiting its geographical reach and impact. Achieving meaningful long-term climatological improvements would require years of repeated ship tracks in the same region. In many cases, such extensive historical data are not available, further restricting the method’s applicability. Thus, while the approach holds promise for localized calibration, it cannot be considered generalizable without a significantly expanded and sustained observational database.
7. Limited improvement at fixed locations
Validation results at Uto demonstrate that the calibration methodology offers only modest reductions in bias. The improvements are primarily observed when the ship providing measurements is in close proximity to the location. When the vessel is farther away, the calibration effect diminishes, underscoring the spatial limitations of the approach. Analysis of time-series data reveals that the corrections made are small compared to the variability inherent in ERA5 wind estimates. This indicates that the method’s ability to significantly alter or improve the data at fixed sites is limited. Ultimately, these findings highlight the fundamental challenge: mobile platforms, such as ships, cannot effectively stand in for stationary reference systems when the objective is detailed, site-specific wind characterization. The transient nature of ship-based measurements restricts their utility for improving local wind assessments at fixed locations.
8. One‑variable calibration is incomplete
The current calibration methodology addresses only wind speed, leaving other essential variables uncorrected. Wind direction remains a significant source of error, particularly in coastal zones. Despite its importance, it is not corrected in the calibration process, which limits the ability to reduce ERA5 inaccuracies in these regions. Additionally, vertical wind shear and atmospheric stability are unchanged. This omission diminishes the value of the calibration for applications such as turbine performance modeling, where these factors are critical. The methodology also does not address wake interactions or extreme wind conditions, both of which are crucial considerations for offshore wind planning. Because only wind speed is calibrated, the method cannot provide a full physical characterization of the wind environment. This limitation restricts its usefulness in applications requiring comprehensive wind data.
Overall, to enhance the scientific value and effectiveness of the calibration methodology, it seems essential for the authors to collaborate more closely with the European Centre for Medium-Range Weather Forecasts (ECMWF). Strengthening this partnership would facilitate the integration and assimilation of shipborne observations into existing datasets, thereby improving the accuracy and representation of wind fields. Another recommendation is to utilize ERA5 reanalysis products as inputs and apply the Weather Research and Forecasting (WRF) model for downscaling. By employing standard data assimilation techniques within WRF, it would be possible to modify the boundary layer structure using shipborne observations and results would be more consistent with physics. This approach could help address current limitations in spatial and temporal coverage and provide a more refined depiction of local wind conditions. Without these improvements in assimilation and modeling practices, the reviewer believes that the current calibration methodology does not contribute significant scientific value toward enhancing the representation of wind fields.
Citation: https://doi.org/10.5194/wes-2026-50-RC3
Status: closed
-
RC1: 'Comment on wes-2026-50', Anonymous Referee #1, 20 Apr 2026
The comment was uploaded in the form of a supplement: https://wes.copernicus.org/preprints/wes-2026-50/wes-2026-50-RC1-supplement.pdf
- RC2: 'Comment on wes-2026-50', Anonymous Referee #2, 07 May 2026
-
RC3: 'Comment on wes-2026-50', Anonymous Referee #3, 26 Jun 2026
The journal paper “Improving offshore wind data from reanalyses using ship-based lidar measurements” by Rubio et al., proposes a methodology to correct ERA5 model outputs using offshore observations from a shipborne observational dataset. Although, in general observations are crucial for offshore model validations or improvements, the proposed methodology is too simplistic and has some issues that the reviewer feels the authors are ignoring.
Below are some major comments:
1. Fundamental limits of “Post‑Processing” calibration
The methodology proposed in the paper adjusts ERA5 wind outputs retrospectively, rather than addressing the underlying model physics or integrating corrections through data assimilation at an earlier stage. As a result, this approach is unable to rectify fundamental structural biases inherent in the model, such as those associated with surface layer parameterizations, representation of coastal flows, or stability-related inaccuracies. Furthermore, the corrections applied are confined to specific locations and moments in time, lacking the spatial and temporal propagation characteristics of genuine data assimilation techniques. This limitation means that ERA5 retains internal inconsistencies, with only wind speed being corrected while other crucial variables, including wind direction, vertical shear, turbulence, and stability remain unaltered. Consequently, the calibrated product may lack internal physical consistency, which is particularly problematic for applications that require reliable multi-variable relationships, such as load calculations, wake modeling, and comprehensive resource assessments.
2. Dependence on non‑stationary and intermittent ship data
Ship-based lidar data are not collected at fixed points, instead, measurements are acquired as the vessel moves along its route. This movement introduces spatial and temporal variability in the sampling process, resulting in uneven coverage across the study area. Additionally, the proximity to the coastline and the operational schedule of the ferry lead to sampling biases. For example, there is a concentration of data points near harbors where the ship spends more time, while data are sparse in offshore regions where the vessel passes less frequently. As a result, the calibration effect achieved through these measurements is localized and intermittent, as noted by the authors. This means that corrections are strongest along the ship's track but diminish in areas farther away, both in space and time. This situation creates a “patchwork” calibration pattern, where the signal is strong near the ship’s route but weak elsewhere. Such a limitation poses significant challenges for regional wind resource assessments, as it undermines the ability to achieve consistent corrections over the broader domain.
3 Correlation-based vs. Physics-based correction
The methodology described in the paper relies on a weighted statistical correction to adjust model outputs, rather than employing a physically constrained data assimilation process. This means that instead of integrating observations into the model in a way that respects the underlying physical laws and constraints, the approach uses statistical relationships to determine corrections. A notable issue with this method is that the weighting radii and time windows used for the corrections are chosen empirically, without physical justification. As a result, the selection of these parameters is not guided by an understanding of the relevant atmospheric processes, but rather by statistical or practical considerations. There is no guarantee that the corrections applied through this approach preserve essential physical balances, such as mass continuity or momentum budgets. Because the corrections are not informed by physical constraints, the adjusted model output may violate these underlying principles. Furthermore, improvements achieved at one location may inadvertently degrade the accuracy at nearby grid points. This is due to the decorrelation between the corrections applied at different points, which means that the statistical adjustments do not propagate in a physically consistent manner across the model domain. As a result, the approach is vulnerable to overfitting, noise, and inconsistency. Since corrections are based on local statistical relationships rather than robust physical mechanisms, there is a risk that the method may respond excessively to random fluctuations in the observational data, leading to unreliable and inconsistent results.
4. Sensitivity to measurement errors
Ship-based lidar data are imperfect and introduce several sources of uncertainty into the calibration process. Motion correction algorithms, which are necessary due to the movement of the vessel, inherently add uncertainty to the measurements. Even though Carrier-to-Noise Ratio (CNR) filtering is applied to remove degraded data, it reduces the overall availability of observations and cannot fully eliminate data quality issues. Errors such as instrument tilt and contamination from precipitation can directly affect the calibration results. These measurement errors are not always adequately filtered or corrected, meaning that they can imprint themselves onto the model calibration. Because the methodology applies corrections based explicitly on the differences between observations and model outputs, any error present in the lidar data is transferred directly into the model, potentially degrading its accuracy.
5. Spatial representativeness problems
Lidar observations, which are essentially point or near-point measurements, are being used to adjust the mean values of entire ERA5 grid cells. This poses a fundamental representativeness issue because the spatial footprint of lidar measurements is quite small, whereas ERA5 grid cells cover much larger areas. In regions with pronounced coastal and mesoscale gradients, the lidar data may not accurately reflect the conditions across the broader grid cell. This mismatch is particularly problematic when there are significant spatial variations within the cell, as the localized lidar readings may not capture the heterogeneity present. The methodology’s reliance on the Cressman radius-of-influence approach further complicates matters. This technique assumes isotropic and stationary conditions, which are not valid for coastal flows that often exhibit strong directionality and variability. As a result, the calibration process can misrepresent the spatial structure of the wind field, especially where mesoscale heterogeneity is pronounced. These factors contribute to the risk of miscalibration, particularly in areas with complex spatial variability. When mesoscale heterogeneity is strong, adjustments based on small footprint lidar observations may fail to accurately improve the broader grid cell, potentially introducing errors rather than correcting them.
6. Scalability and generalizability issues
For broader application of the calibration methodology, several practical limitations emerge that affect its scalability and generalizability. Effective calibration requires either dense ship traffic or multiple coordinated vessels to provide adequate spatial and temporal coverage. In many marine environments, this level of observational density is unrealistic, making widespread implementation impractical. Furthermore, the calibration is valid only within approximately 60–90 km of the ship database. Areas beyond this range receive no benefit from the correction process, limiting its geographical reach and impact. Achieving meaningful long-term climatological improvements would require years of repeated ship tracks in the same region. In many cases, such extensive historical data are not available, further restricting the method’s applicability. Thus, while the approach holds promise for localized calibration, it cannot be considered generalizable without a significantly expanded and sustained observational database.
7. Limited improvement at fixed locations
Validation results at Uto demonstrate that the calibration methodology offers only modest reductions in bias. The improvements are primarily observed when the ship providing measurements is in close proximity to the location. When the vessel is farther away, the calibration effect diminishes, underscoring the spatial limitations of the approach. Analysis of time-series data reveals that the corrections made are small compared to the variability inherent in ERA5 wind estimates. This indicates that the method’s ability to significantly alter or improve the data at fixed sites is limited. Ultimately, these findings highlight the fundamental challenge: mobile platforms, such as ships, cannot effectively stand in for stationary reference systems when the objective is detailed, site-specific wind characterization. The transient nature of ship-based measurements restricts their utility for improving local wind assessments at fixed locations.
8. One‑variable calibration is incomplete
The current calibration methodology addresses only wind speed, leaving other essential variables uncorrected. Wind direction remains a significant source of error, particularly in coastal zones. Despite its importance, it is not corrected in the calibration process, which limits the ability to reduce ERA5 inaccuracies in these regions. Additionally, vertical wind shear and atmospheric stability are unchanged. This omission diminishes the value of the calibration for applications such as turbine performance modeling, where these factors are critical. The methodology also does not address wake interactions or extreme wind conditions, both of which are crucial considerations for offshore wind planning. Because only wind speed is calibrated, the method cannot provide a full physical characterization of the wind environment. This limitation restricts its usefulness in applications requiring comprehensive wind data.
Overall, to enhance the scientific value and effectiveness of the calibration methodology, it seems essential for the authors to collaborate more closely with the European Centre for Medium-Range Weather Forecasts (ECMWF). Strengthening this partnership would facilitate the integration and assimilation of shipborne observations into existing datasets, thereby improving the accuracy and representation of wind fields. Another recommendation is to utilize ERA5 reanalysis products as inputs and apply the Weather Research and Forecasting (WRF) model for downscaling. By employing standard data assimilation techniques within WRF, it would be possible to modify the boundary layer structure using shipborne observations and results would be more consistent with physics. This approach could help address current limitations in spatial and temporal coverage and provide a more refined depiction of local wind conditions. Without these improvements in assimilation and modeling practices, the reviewer believes that the current calibration methodology does not contribute significant scientific value toward enhancing the representation of wind fields.
Citation: https://doi.org/10.5194/wes-2026-50-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 335 | 100 | 19 | 454 | 22 | 26 |
- HTML: 335
- PDF: 100
- XML: 19
- Total: 454
- BibTeX: 22
- EndNote: 26
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1