A Sensivity Analysis of the WRF Boundary Layer Parameterizations for the Tropical Conditions

We evaluated the performance of the WRF (Weather Research and Forecasting) - ARW (Advanced Research WRF) model (version 3.2.1) to predict the wind profile (from the surface up to 5 000 m ) at a site which is located at 2 S in a coastal area (Atlantic Ocean). The InterTropical Convergence Zonce and trade winds are the dominant meteorological features and the zonal component (associated with the trade winds) is higher than the meridional wind (associated with the sea breeze). This analysis was made using the default boundary layer parameterizations used by the WRF model: ACM2, MYJ, MYNN2.5 and YSU. Seasonal analyzes of the model behavior were performed against two radiosonde data sets, which represent the dry (strong winds) and rainy (weak winds) seasons made at 00, 06, 12 and 18 UTC using the radiosonde Vaisala RS92-SVG. The 72 hours forecasts were made from initial conditions (at 00 UTC) provided by the Global Forecasting System (GFS/NCEP) obtainded at http://nomads.ncep.noaa.gov/pub/data/nccf/com/gfs/prod/ for 0.5° X 0.5°. The land use/topography data determined by United States Geological Survey (USGS) (available at: http://www.mmm.ucar.edu/wrf/src/wps_files/geog.tar.gz) for 30xx of spatial resolution. Simulations were configured for three grids nested (spatial resolutions of 9 X 9, 3 X 3 and 1 X 1 km) and 42 vertical levels (25 within the lower tropopause – 5000 m) and the outputs were compared with observations every 6 hours using the Willmott index of agreement (d). The spin-up period was assumed to be 6 hours. Both radiosondes and outputs from model were linear interpolated to each 50 m for the analysis. Although the differences amongst the 4 tested parameterizations were small, the parameterizations ACM2 (for the dry season) and MYNN2.5 (for the wet season) provided the slightly better results against the observations. If is necessary, either ACM2 or MYNN2.5 can be used for the entire year without any big problem. In general, the outputs from WRF were able to represent the wind profile in a reasonable way: the model overestimated the bulk windspeed (an average value from the surface up to 5000 m) with maximum values around 2.0 m/s for a specific level, being an average value of 0.85 m/s for the dry and 0.70 m/s for the wet season. There is a good agreement for the wind direction (differences less than 10 º). There is no better performance for dry or wet seasons, however, the Wilmottxs index metric used presented better results for the wet season (the values ranged from 0.6-0.9 instead of 0.5 – 0.9 for the dry season).