Assessing WRF PBL Schemes for Wind Energy Applications Branko Kosović, Yubao Liu, Youwei Liu, Will Cheng NCAR Workshop May 12, 21 NATIONAL CENTER FOR ATMOSPHERIC RESEARCH
In the Past PBL Parameterizations Have Not Been Evaluated with Respect to Wind Forecasting at 8m Accurate representation of internal PBL processes and PBL interaction with surface and upper troposphere is important for accurate wind forecasting in PBL. We are trying to determine optimal PBL parameterization configuration for wind forecasting: PBL scheme, vertical resolution, and input parameters. We focus on PBL parameterizations available in WRF.
We Need to Identify and Address Limitations of PBL Schemes that Impact Wind Forecasting What is the level of uncertainty in external forcing? When are wind forecast errors largest? daytime or nightime are there seasonal differences under what synoptic conditions, etc. How can we improve performance of PBL schemes? through improved representation of physical processes by reducing uncertainty in model parameters better accounting for uncertainties in model parameters
Surface Weather Map March 5, 21 Lake Benton Wildorado 28, University Corporation for Atmospheric Research. All rights reserved.
High and Low Temperatures March 5, 21
During Several Days at the Beginning of March We Observed Significant Under Prediction of Wind Speed
Qualitative Comparison with Data from A Profiler Shows that Upper Level Winds are Accurately Predicted 4 Wildorado Profiler m/s 24 3 2 Height [m] 2 16 12 1 8 4 3 4 5 6 7 March 21
Qualitative Comparison with Data from A Profiler Shows that Upper Level Winds are Accurately Predicted 4 Wildorado Wind Speed Forecast m/s 24 3 2 Height [m] 2 16 12 1 8 4 3 4 5 6 7 March 21
Operational Forecast for XXXX Wind Farm Underestimated Wind Speed Between March 3 and 7
Qualitative Comparison with Data from A Profiler Shows that Upper Level Winds are Accurately Predicted 4 XXXX Profiler m/s 24 3 2 Height [m] 2 16 12 1 8 4 3 4 5 6 7 March 21
Qualitative Comparison with Data from A Profiler Shows that Upper Level Winds are Accurately Predicted 4 Wind Speed Forecast m/s 24 3 2 Height [m] 16 2 12 1 8 4 3 4 5 6 7 March 21
We used several PBL parameterizations available in WRF in our SCM study Yonsei University - YSU Melor-Yamada-Janic - MYJ Melor-Yamada-Nakanishi-Niino MYNN (2.5) Melor-Yamada-Nakanishi-Niino MYNN3 Quasi Normal Scale Elimination QNSE Initial conditions and forcing were derived from the operational NCEP GFS (1 deg) analysis with 6h data frequency to force SCM simulations We used 56 and 84 grid points in vertical direction
Temperature from SCM with YSU Parameterization and 56 Grid Points at XXXX Wind Farm 8 6 Height [m] 4 2 3 4 5 6 7 March 21
Wind Speed from SCM with YSU Parameterization and 56 Grid Points at XXXX Wind Farm 8 6 Height [m] 4 2 3 4 5 6 7 March 21
Wind Speed Difference Between SCM Simulations with YSU and MYJ Schemes are Negligible m/s 8 1. 6.5 Height [m] 4. 2 -.5 3 4 5 6 7 March 21-1.
Wind Speed Difference Between SCM Simulations with YSU and MYNN Schemes are Small 8 m/s 1. 6.5 Height [m] 4. 2 -.5 3 4 5 6 7 March 21-1.
Wind Speed Difference Between SCM Simulations with YSU and MYNN3 Schemes are Negligible m/s 8 1. 6.5 Height [m] 4. 2 -.5 3 4 5 6 7 March 21-1.
Wind Speed Difference Between SCM Simulations with YSU and QNSE are Small m/s 8 1. 6.5 Height [m] 4. 2 -.5 3 4 5 6 7 March 21-1.
GFS Analysis Does Not Include Snow Cower Over the Domain of Interest at the Beginning of March During the time of interest near Lake Benton snow cover was estimated at ~2in. However, without snow cover the grassland surface roughness, z =.5, while for snow cover it is significantly lower, z =.1. We modified input data to account for snow cover and rerun SCM simulations.
Wind Speed Difference Between SCM Simulations with YSU Scheme with and without Snow Cover 8 2. 6 1. Height [m] 4. 2-1. 3 4 5 6 7 March 21-2.
Wind Speed Difference Between SCM Simulations with High-Resolution YSU with and YSU m/s 8 3. 6 1.5 Height [m] 4. 2-1.5 3 4 5 6 7 March 21-3.
Comparison of Measured Wind Speed at XXXX on with SCM Prediction with YSU Scheme 14 Solid line measurements Dotted line SCM 12 1 Wind Speed [m/s] 8 6 4 2 3 4 5 6 7 March 21
Temperature from SCM with YSU Parameterization and 56 Grid Points at XXXX Wind Farm 8 6 Height [m] 4 2 3 4 5 6 7 March 21
Wind Speed from SCM with YSU Parameterization and 56 Grid Points at XXXX Wind Farm 8 6 Height [m] 4 2 3 4 5 6 7 March 21
Wind Speed Difference Between SCM Simulations with YSU and MYJ at XXXX Wind Farm m/s 8 1. 6.5 Height [m] 4. 2 -.5 3 4 5 6 7 March 21-1.
Comparison of Measured Wind Speed at XXXX with SCM Prediction with YSU Scheme 14 Solid line measurements Dotted line SCM 12 1 Wind Speed [m/s] 8 6 4 2 3 4 5 6 7 March 21
Summary and Next Steps SCM captures southerly LLJ but the magnitude of the wind at turbine hub is significantly underestimated Little difference between SCM simulations with different PBL schemes SCMs over predict 1m winds while under predicting hub-height winds It is important to assimilate reliable, quality controlled local data not assimilated in large scale forecasts (analysis) By correctly accounting for surface roughness and by increasing vertical resolution we reduced the error in wind speed prediction by 1.5 m/s More and better data are needed to further study and better understand PBL processes that affect LLJ We will further analyze surface layer parameterizations and PBL parameterizations 29, and University their Corporation interaction for Atmospheric Research. All rights reserved.
For Wind Energy Applications We Need to Improve Wind Forecasting in Planetary Boundary Layer (PBL) Hypothesis: Accurate representation of internal PBL processes and PBL interaction with surface and upper troposphere is important for accurate wind forecasting in PBL Objective: Improved wind forecasting in PBL Steps: Analyze simple PBL parameterizations w.r.t. wind forecasting below 3m Analyze PBL parameterizations in WRF Analyze how PBL parameterization(s) affect wind forecasting in WRF
During Several Days at the Beginning of March We Observed Significant Under Prediction of Wind Speed
From A. Beljaars ECMWF training course on BLs In the Past PBL Parameterizations Have Not Been Evaluated with Respect to Wind Forecasting at 8m A number of reasons exists to have a realistic representation of the boundary layer in a large scale model: The large-scale budgets of momentum heat and moisture are considerably affected by the surface fluxes on time scales of a few days. Model variables in the boundary layer are important model products. The boundary layer interacts with other processes e.g. clouds and convection.
To Improve PBL Parameterizations We Need to Answer Some of Following Questions What is the level of uncertainty in external forcing? What is the level of uncertainty in representation of internal processes (or parameters)? Are there processes that are not represented at all or that are not represented accurately? Which processes and parameters affect wind forecasting the most? What can we do to improve representation of these processes or reduce uncertainty in parameters toward improving wind forecasting in PBL?
Surface Weather Map March 4, 21 Lake Benton Wilderado
Wind Speed Difference Between SCM Simulations with YSU and QNSE at XXXX Wind Farm 8 6 Height [m] 4 2 March 21