Please use this identifier to cite or link to this item:
Title: Utilizing large-eddy simulation in high-resolution wind resource modelling : a coastal low-level jet case study
Authors: Tay, Ken
Keywords: DRNTU::Engineering::Mathematics and analysis::Simulations
DRNTU::Science::Physics::Meteorology and climatology
Issue Date: 2017
Source: Tay, K. (2017). Utilizing large-eddy simulation in high-resolution wind resource modelling : a coastal low-level jet case study. Doctoral thesis, Nanyang Technological University, Singapore.
Abstract: The challenges of nesting large-eddy simulation (LES) within the Weather Research Forecast (WRF) model are examined and addressed in this work. Coastal low-level jet (LLJ) events in field campaigns CBLAST-LOW 2001 and NEAQS-ITCT 2004 are used as case studies. Nested LES models offer an advantage in LLJ modeling by encompassing a wide range of length scales. Broad mesoscale conditions that drives LLJ events are taken into account in the coarse outer domains while local turbulence interactions are resolved within the nested high-resolution LES model. The accuracy of nested LES models are largely dependent on the mesoscale parent domains used. The sensitivity of the mesoscale domains to initial/boundary conditions (IBC), planetary boundary layer (PBL) scheme and sea-surface temperature (SST) forcing are characterized using a suite of diagnostic metrics. In the CBLAST-LOW 2001 case study, IBC was the greatest source of variability due to the differences in the representation of a warm frontal passage For a hypothetical wind turbine located in the center of the domain, over the duration of the LLJ event, the variability in the simulations results in 5% difference in predicted wind power. In the NEAQS-ITCT 2004 case study, a continuous 48-hour period that featured two different LLJ events, a WEAK LLJ (∼10 m s −1 ) and a STRONG LLJ (∼18 m s −1 ), was selected for analysis. All the simulations were able to successfully model both LLJ events to varying degree. The simulations underestimated the WEAK LLJ peak speed, which also arrived before the observed LLJ; whereas the STRONG LLJ peak speed was overestimated and lagged the observed LLJ. Regardless of IBC or PBL used, the simulations showed an average cold bias (∼ 2 K) in the surface temperature when compared to ASOS data. The normalized bias (NBIAS) and normalized root-mean-square error (NRMSE) of wind velocity prediction were used to score model performance. IBC was the greatest source of variability in wind velocity prediction during the WEAK LLJ event. Different SST forcings, which include warming the SST within the Gulf of Maine, had little to no impact on the NRMSE scores during this event. However, during the STRONG LLJ event, regardless of model configurations, the simulations scored similarly in NBIAS within the subjet layer (< 100 m). The NBIAS peaked at approximately the height of the LLJ core (∼ 100 m). Above this layer, the variability induced by the choice of IBC, PBL and SST were much more pronounced. In bothWEAK and STRONG LLJ events, the maximum NRMSE occurs within the subjet layer, highlighting the need to improve parameterization near the surface layer. Based on the NBIAS and NRMSE scores, GFS MYNN was the best performing configuration for the mesoscale simulation and hence selected as the parent domain for nested WRF-LES simulation. Two sets of nested WRF-LES simulations were performed, one each for the two LLJ events identified. A new method to spin-up turbulence between mesoscale and LES domain was also tested. WRF-LES reduced the NBIAS and NRMSE for both WEAK and STRONG LLJ within the subjet and jet core layer as compared to the parent domain. Spectral analysis showed that the Cell-Perturbation Method (CPM) reduced the fetch required for the turbulence spectra to reach equilibrium. CPM was more effective in STRONG than WEAK. CPM was able to inject the missing turbulence scales from the transition between the mesoscale and LES domain when the horizontal resolution was sufficient to resolve most of the LES eddies. The improvement in turbulence representation due to CPM also resulted in further improvements in the NBIAS and NRMSE scores of the modeled LLJ. CPM was shown to be an effective tool for real nested WRF-LES studies.
DOI: 10.32657/10356/72652
Fulltext Permission: open
Fulltext Availability: With Fulltext
Appears in Collections:IGS Theses

Files in This Item:
File Description SizeFormat 
Ken-Thesis_FINAL_SUBMITTED.pdf25.77 MBAdobe PDFThumbnail

Page view(s)

Updated on Apr 19, 2021

Download(s) 50

Updated on Apr 19, 2021

Google ScholarTM




Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.