Sensitivity of the WRF model surface wind simulations to initial conditions and planetary boundary layer parameterization schemes (case study: over Persian Gulf)

In this work‎, ‎sensitivity and performance of the Weather Research and Forecasting (WRF) model for surface wind field simulations are evaluated under several initial and‎ boundary conditions‎, ‎along with‎ ‎different planetary boundary layer (PBL) schemes during several dates over the Persian Gulf region‎. Since there are differences between production approaches and development periods of analysis and reanalysis data (namely‎, ‎assimilation system) on the one hand‎, ‎and‎ differences in applied method for PBL parameterization by any scheme on the other hand‎, ‎this paper aims to identify a suitable set up among the whole configurations which‎ are under examination‎. ‎To this end‎, ‎three datasets (two reanalyses and one analysis) including‎, ‎ERA-Interim‎, ‎NCEP-R2‎, ‎and NCEP-FNL and six PBL‎ schemes (two local and four nonlocal) including ACM2‎, ‎BouLac‎, ‎MYJ‎, ‎MYNN‎, ‎QNSE‎, ‎and YSU accompanied by their relevant surface-layer schemes are used‎. To assess the WRF model wind simulations, available observational wind data including 22 synoptic weather stations located in the region and‎ observations of QuikSCAT and ASCAT satellites are employed‎. Findings of this study indicate that when the wind simulations are compared with synoptic weather stations observations‎, ‎irrespective of the type of PBL scheme‎, ‎ERA-Interim and‎ NCEP-FNL datasets exhibit better performance in comparison with the NCEP-R2 and ‎when PBL schemes are also considered‎, ‎results show that combination of YSU scheme and‎ ERA-Interim reanalysis data leads to a better estimate of wind speed and combination of YSU and NCEP-FNL data generates less error for wind direction‎. Moreover‎, ‎comparison of model wind simulations and observations of QuikSCAT and ASCAT satellites shows that there are no substantial differences between various‎ configurations‎. ‎However‎, ‎using YSU and ACM2 scheme‎s, ‎WRF model generates speed and direction of the wind close to the observations of QuikSCAT‎. Although all tests have almost similar results as ASCAT satellite observations‎, ‎YSU scheme estimates are slightly better than other schemes. Overall, the results of this study revealed that the major difference between WRF wind simulations and measured winds arises from the choice of initial conditions data and it does not depend on different PBL schemes. Consequently, changing initial and boundary conditions data has a noticeable impact on the model wind results. Thus, in future studies, emphasis must be more on reanalysis and analysis datasets and the option of WRF PBL parameterization schemes should be the second priority. Due to the fairly good similarity of the model surface wind with QuikSCAT and ASCAT observations, the choice of WRF model simulations as offshore wind database can be a valid available alternative instead of QuikSCAT and ASCAT wind, particularly when meeting their limitation in spatial resolution (swath data) or temporal sampling.