Estimation of lidar bias over complex terrain using numerical tools

Abstract

For a few years, a new wind measurement instrument has been competing with standard cup anemometers: wind LiDARs. Despite numerous advantages such as ease of deployment and the possibility to scan at multiple heights simultaneously, the performances of this instrument over complex terrain are still a matter of debate. This is mainly due to the flow homogeneity assumption made by the remote sensor which leads to a positive or a negative bias. The objective of this work was to implement a method to evaluate LiDAR bias over complex terrain using OpenFOAM. To accomplish this task, a CFD model capable of dealing with complex terrain and sparse forest was developed in OpenFOAM v1.7. A RANS approach coupled with a modified k-ε turbulence model accounting for extra turbulence generated by the forest was used. To estimate LiDAR bias, the method proposed by Bingöl et al. (2008) was implemented as a post-processing tool. First, a simple verification of the model was carried out by modeling neutrally stratified boundary layer. Apart from the usual overshoot of k in the near-wall cells, results agreed well with analytical solutions. Then, the modifications brought to the solver to account for the effects of forest were validated. In order to do so, flow over and within a dense forest was modeled and results were compared to experimental data of Amiro (1990) and numerical results of Dalpé and Masson (2008). An innovative top boundary condition totally independent of the forest displacement height developed by Lussier-Clément (2012) was also validated. The experiment showed that, in the presence of forest, imposing fixed values for U, k and ε at the top of the domain is not appropriate. The LiDAR bias post-processing algorithm was also validated for flow over an isolated Gaussian hill. The effects of the scanning height as well as the slope of the hill were investigated. For terrain slopes ranging from ∼25% to ∼43%, LiDAR bias ranging from 2% up to 10% was observed. The generalization of the method for large areas revealed to be particulary useful at showing the extant of the bias. Finally, a real case scenario was studied where a LiDAR was sited in the Gaspe peninsula on a complex and densely forested terrain. The assessment of the CFD model for this site firstly revealed the significant impact of both the location and nature of the inlet boundary condition. Then, the LiDAR bias was estimated with the help of OpenFOAM v1.7, MeteoDyn WT 4.0 and WAsP Engineering. Numerical results were compared to experimental data. Despite the presence of terrain complexity up to a distance of eight times the radius of the scanned disc around the remote sensor, very little error was observed, suggesting that the LiDAR is only affected by topographic variations closer to the scanned volume.

Date	24 Jul 2013
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Christian Masson (Supervisor) & Louis Dufresne (Co-supervisor)