A sodar campaign in complex terrain for data quality evaluation and methodological investigations

Remote sensing appears more and more as a valuable alternative to high masts for wind resource assessment and power curve measurement. Many campaigns have shown that sodars and lidars provide data in very good agreement with anemometers measurements. However, few long campaigns have been performed in complex terrain. Moreover, there is still a need to define a methodology for sodar (or lidar) use in an operational context. A 8-months campaign has been performed on a very complex and isolated terrain in southern France between October 2007 and June 2008. The instrumental set-up has been installed at the top of a hill in a clearing surrounded by forest and strong slopes. It includes one Scintec SFAS sodar, and a 82m mast equipped with 5 cup anemometers, 2 vanes, 4 sonic anemometers, 4 temperature and 1 humidity sensors. Due to the vicinity of a high mast, and the noise generated by the trees, the working conditions were very difficult for the sodar. The rate of availability of validated sodar data remains higher than 90% up to 75m. The statistical comparison between the sodar and the anemometers measurements of wind speed indicates a good correlation, but the sodar underestimates the wind speed, contrary to what was observed on flat terrain in a previous campaign for the same sodar. This underestimation is probably due to the fact that the assumption of local homogeneity on which relies the remote sensing measurement is not valid in complex terrain. This is confirmed by an analysis by direction sector, which shows a clear link between the amplitude of the sodar underestimation and the terrain steepness upwind to the instrumental site. However, a simple correction based on the regression between sodar and mast measurement at 45m allows to reduce the sodar error at 80m from 12% to about 2%. In order to take advantage of the possibility to move easily a remote sensing sensor, a method using 2 months of sodar data combined with 1 year of a small mast data has been tested to retrieve the wind speed at 80m. This method consists in calculating vertical profiles normalized at 40m level and averaged by direction sectors and stability classes. The comparison of the annual distribution of wind speed calculated at 80m with anemometer measurements shows that this method is able to provide the Weibull parameters A and k with an error of about 2%.