**7. Conclusions**

In the present study, field observation wind data from the time of the wind turbine blade damage accident on Shiratakiyama Wind Farm, which was studied by Uchida [1], were analyzed in detail. In parallel, high-resolution LES turbulence simulations were performed in order to examine the model's ability to numerically reproduce terrain-induced turbulence.

First, a comparison was made between terrain elevation data with 10.0 m and 50.0 m spatial resolutions from the Geospatial Information Authority of Japan (GSI). When a uniform grid spacing of 50.0 m was used for the horizontal grid for the airflow calculation, the accuracy of the simulation results was approximately the same between the simulations which used the two datasets. Furthermore, in order to capture unsteady fluid properties of terrain-induced turbulence, a sufficiently small time increment (Δ*t* = 2 × 10−<sup>3</sup> *h*/*U*) is required. Here, *h* is the difference between the minimum and maximum terrain elevations within the computational domain; *U* is the wind velocity at the inflow boundary at the height of the maximum terrain elevation within the computational domain.

Secondly, vertical profiles of the standard deviations of the three wind velocity components at Wind Turbine No. 17 were examined based on the non-dimensional simulation outputs. This examination revealed that the value of the standard deviation for the *y*-component of wind velocity exceeded that for the *x*-component of wind velocity at the wind turbine hub height (=85.0 m) and below.

Thirdly, based on the rescaled-simulation outputs, we try to examine the model's ability to numerically reproduce terrain-induced turbulence (turbulence intensity) under strong wind conditions (8.0–9.0 m/s at wind turbine hub height). Since the wind velocity and time acquired from the numerical simulation are dimensionless, they are converted to full scale. As a consequence, both the standard deviation of the horizontal wind speed (m/s) and turbulence intensity evaluated from the field observation and simulated wind data are successfully in close agreement.

Finally, to investigate the cause of the wind turbine blade damage accident on Shiratakiyama Wind Farm, a power spectral analysis was performed on the fluctuating components of the observed time series data of wind speed (1 s average values) for a 10 min period (total of 600 data) by using a fast Fourier transform (FFT). It was suggested that the terrain-induced turbulence that caused the wind turbine blade damage accident on Shiratakiyama Wind Farm was attributable to rapid wind speed and direction fluctuations that were caused by vortex shedding from Tenjogadake (elevation: 691.1m) located upstream of the wind farm.

**Funding:** This work was supported by JSPS KAKENHI Grant Number 17H02053.

**Acknowledgments:** In carrying out the present study, the author received valuable input from Takashi Maruyama of the Disaster Prevention Research Institute, Kyoto University, Japan. In addition, the field observation wind data from Shiratakiyama Wind Farm which were analyzed in the present study were provided by Kinden Corporation. The author wishes to express his gratitude to those who have contributed to the present study.

**Conflicts of Interest:** The authors declare no conflict of interest.
