Abstract:Summary key findings: Near‐surface wind observations from offshore platforms are estimated in a suboptimal way based on higher‐elevation measurements and are operationally distributed to the meteorological community via the Global Telecommunication System (GTS). Applying these inaccurate observations in numerical weather prediction (NWP) models reduces the accuracy of the NWP analysis and makes the interpretation of the verification results difficult as it may alter the sign of biases and the ranking of forecast systems. However, applying these observations in a more optimal way can lead to improved forecasts and increase our insight into offshore wind, for example related to biases, NWP errors in different stability regimes and in the development of forecast accuracy over time. Histograms of differences in estimated 10‐m wind speed from observations at six Norwegian platforms for the inaccurate GTS approach minus a more sophisticated stability‐dependent estimate which takes into account the actual vertical profiles. The x‐axis shows differences, the y‐axis shows the number of cases. The comparison is done for 2017–2023 and different stability classes are indicated by colors. At Norwegian offshore platforms, wind speed is observed at heights between approx. 35 and approx. 150 m a.s.l. These observations are used to estimate 10‐m wind speed, by a power law with a constant wind shear coefficient. The resulting 10‐m wind speed is then distributed to the Global Telecommunication System (GTS) and used for assimilation and verification purposes by the meteorological community. However, multiple methods with varying results exist for the reduction of wind speed from the original measurement height to 10 m. The resulting estimate of 10‐m wind speed is therefore sensitive to the choice of method. The use of such observations may therefore be suboptimal in forecast analysis and make the forecast verification difficult to interpret. This study investigates how the use of these observations impacts forecasts from the operational regional forecast system used at MET Norway and the Global Integrated Forecast System operated by ECMWF for the period 2017–2023. Both forecast systems assimilated the observations over major parts of the period and the presented results indicate that the use of these observations increased the initial errors in the forecasts. This is confirmed in a data denial experiment over a shorter period of time. The verification results show that both forecast systems underestimate the wind speed at the original measurement heights, while the results for 10‐m verification vary with the interpolation method between original sensor height and 10 m. The forecasts show in general more wind than the GTS‐distributed estimates and are in better agreement with more advanced interpolation methods. The verification results show a strong dependency on the vertical stability of the atmosphere. The major differences between the forecast systems are that the regional system shows less bias, while the standard deviations of the errors are slightly lower in the global system.

Evaluation of forecasts by a global data-driven weather model with and without probabilistic post-processing at Norwegian stations

Evaluation of forecasts by a global data-driven weather model with and without probabilistic post-processing at Norwegian stations

Multivariate post‐processing of probabilistic sub‐seasonal weather regime forecasts

The Compatibility between the Pangu Weather Forecasting Model and Meteorological Operational Data

Regional data-driven weather modeling with a global stretched-grid

Comparison of statistical post-processing methods for probabilistic NWP forecasts of solar radiation

Multivariate post-processing of probabilistic sub-seasonal weather regime forecasts

How inaccurate offshore wind observations impact the quality of operational numerical weather prediction

The rise of data-driven weather forecasting

A Machine Learning Outlook: Post-processing of Global Medium-range Forecasts

Statistical post-processing of multiple meteorological elements using the multimodel integration embedded method

Uncertainty quantification for data-driven weather models

Using numerical weather prediction to assess climate models

A transformer neural network for predicting near‐surface temperature

A novel approach to improve numerical weather prediction skills by using anomaly integration and historical data

Improving Boreal Summer Precipitation Predictions From the Global NMME Through Res34‐Unet

A framework for probabilistic weather forecast post-processing across models and lead times using machine learning

Deep learning for post-processing global probabilistic forecasts on sub-seasonal time scales

Statistical Modelling of 2m Temperature and 10m Wind Speed Forecast Errors

Leveraging data-driven weather models for improving numerical weather prediction skill through large-scale spectral nudging