This article is a re-edited version of a portion of "Evaluation of the Accuracy of Wind Speed ​​Profiles for LFM-GPV and MSM-GPV," presented at the 47th Wind Energy Utilization Symposium held in November 2025.

Introduction

GPV (Grid Point Value) data* provided by the Japan Meteorological Agency for wind power generation projects1The Local Numerical Weather Forecast Model (LFM) and Mesoscale Numerical Weather Forecast Model (MSM)* are used in a wide range of applications, including initial investigations of wind conditions, long-term corrections, and power generation forecasts.2High-temporal and spatial resolution GPV output from1), 2) are available, and understanding the accuracy of these wind speed estimates is important in practice. 

*1 GPV is data that organizes "future weather and wind forecast results" calculated by a meteorological model into numerical values ​​at regular intervals (grids). 

*2 Both LFM and MSM are numerical forecast models developed by the Japan Meteorological Agency. The main differences are as follows: 
・LFM: Calculates finer areas (local areas) at higher resolution 
・MSM: Stable calculation for a relatively wide weather range 

Data and Methods

This time, we visited the Mutsu Ogawara Offshore Wind Test Site (MTS) in Rokkasho Village, Aomori Prefecture.3)Using one year of observation data from a vertical lidar (ZX300M) installed atAccuracy assessment of initial and forecast wind speed profilesFigure 1 shows the location of the ZX300M, and Table 1 shows an overview of the wind observations. Bias, RMSE, and correlation coefficient were used as evaluation indices for GPV wind speed, and quantitative comparisons were made at multiple altitudes. 

Figure 1: Location of the ZX300M. The exterior of the ZX300M is shown in the lower left of the image. 

Table 1. Overview of wind observation using the ZX300M

Results and Discussion 

Figure 2 shows the vertical profiles of Bias, RMSE, and R for the LFM-GPV wind speed and MSM-GPV wind speed over the MTS year.LFM-GPV tends to significantly underestimate wind speed in the initial valuesOn the other hand, as the forecast time progresses, the error tendency tends to shift in the positive direction.No significant systematic differences were observed between the initial and predicted values ​​in MSM-GPV., and showed relatively stable wind speed reproducibility. 

Figure 2 Vertical profiles of (1) Bias, (2) RMSE, and (3) R for (a) LFM-GPV wind speed and (b) MSM-GPV wind speed at MTS.
In Forecast, 0h means the initial value, and 1-15h means the forecast value. 

Furthermore, a comparison of the annual mean wind speed distribution calculated from the initial values ​​and hourly forecast values ​​of the LFM-GPV (Figure 3) suggested that the magnitude relationship between the initial values ​​and forecast values ​​confirmed by the MTS also exists over a wide area around Japan. This may be related to the fact that wind observation data from the Automated Meteorological Data Acquisition System (AMeDAS) is assimilated into the LFM data assimilation system.

Figure 3: Annual mean wind speed distribution at 10m height obtained from LFM-GPV (a) FH0100 (hourly forecast value) and (b) FH0000 (initial value), and (c) the distribution of the wind speed difference ((a)-(b)). (c) is the relative value (unit: percentage) obtained by dividing the difference between (a) and (b) by (b). The target period is from 00:00 UTC on March 26, 2024 to 23:00 UTC on March 25, 2025. 

Conclusion

From these results,When using the initial values ​​of LFM-GPV as input data for dynamic downscaling or wind analysis, it is necessary to take into account the influence of systematic errors.Going forward, we plan to continue verification using nationwide observation data, and to continuously evaluate the accuracy of the JMA GPV and the results of wind resource simulations that use it as input. 

As wind condition consultants, Rera Tech Inc. will conduct optimal wind condition surveys that combine "observation" and "estimation" for wind power generation. Please feel free to contact us if you have any inquiries regarding wind conditions.

CONTACT

References 

1) Japan Meteorological Business Support Center, Local Numerical Weather Prediction Model GPV (LFM),
URL: https://www.jmbsc.or.jp/jp/online/file/f-online10300.html（アクセス：2025/12/22）. 

2) Japan Meteorological Business Support Center, Mesoscale Numerical Weather Prediction Model GPV (MSM), URL: https://www.jmbsc.or.jp/jp/online/file/f-online10200.html（アクセス：2025/12/22）. 

3) Mutsu Ogawara Offshore Wind Observation Test Site, URL: https://mo-testsite.com(Accessed: 2025/12/22).