A new approach to laser wind sensor measurement validation is described and demonstrated. The new approach relies on the paired-t statistical method to generate a time series of differences between two sets of measurements. This series of differences is studied to help identify and explain time intervals of operationally significant differences, which is not possible with the traditional approach of relying on the squared coefficient of variation as the primary metric. The new approach includes estimating a confidence interval for the mean difference and establishing a level of meaningful difference for the mean difference, and partitioning the data set based on wind speed.

To demonstrate the utility of the new approach, measurements made by a laser wind sensor mounted on a floating buoy are compared first with those made by a second laser wind sensor mounted on a nearby small island for which the co-efficient of variation is high (> 99%). It was found that time intervals when high differences in wind speed occurred corresponded to high differences in wind direction supporting a hypothesis that the two laser wind sensor units are not always observing the same wind resource. Furthermore, the average difference for the 100m range gate is positive, statistically significant (α=0.01) and slightly larger than the precision of the gages, 0.1m/s. One possible cause of this difference is that the surface roughness over land is slowing the wind at 100m slightly.

A second comparison was made with previously existing cup anemometers mounted on a metrological mast located on-shore. The cup anemometers are about 8m lower than the center of the lowest range gate on the laser wind sensor. The data was partitioned into three sets: not windy (average wind speed at the cup anemometers ≤ 6.7m/s) windy but no enhanced turbulence (average wind speed at the cup anemometers > 6.7m/s), and windy with enhanced turbulence. Periods of enhanced turbulence are associated with the passage of a cold frontal boundary.

The paired-t analysis for the not windy data set showed a difference in the average wind speeds of -0.096m/s, less in absolute value than the precision of the gages. The negative sign indicates slower wind speed over land as well as at a lower height, which is expected. Similar results were obtained for the windy with no enhanced turbulence data set. In addition, the average difference was not statistically significant (α=0.01).

The windy with enhanced turbulence data set showed significant differences between the buoy mounted laser wind sensor and the on-shore mast mounted cup anemometers. The sign of the average difference depended on the direction of the winds in the periods of enhanced turbulence. Mean turbulent kinetic energy was measured to be greater when air flow into Muskegon Lake was predominantly from over land versus when air flow was predominantly from Lake Michigan. The higher mean turbulent kinetic energy for flow originating over land would likely be due to greater surface roughness experienced by the overland flow.

Overall, the value of the new approach in obtaining validation evidence has been demonstrated. In this case, validation evidence is obtained in periods of no enhanced turbulence. Differences in wind speed during periods of enhanced turbulence are isolated in time, studied and are correlated in time with differences in wind direction.


The following article has been submitted to/accepted by the Journal of Renewable and Sustainable Energy. After it is published, it will be found at http://jrse.aip.org/.