The Texas component of the Wind Forecast Improvement Project (WFIP) is a joint collaboration between the Department of Energy, the National Oceanic Atmospheric Administration, and a group of private sector partners that is being managed by AWS Truepower, LLC. The goal of this project is to improve short term wind power forecasts through the use of an expanded set of targeted atmospheric observations and advanced modeling techniques. An integral part of this project is an ensemble of high resolution NWP forecasts that will assimilate observations from SODARs, wind profilers and meteorological towers, which have been deployed in targeted locations for this project. This high resolution ensemble includes a wide variety of numerical weather prediction models and data assimilation systems. Accurate estimates of the background state and boundary conditions play a key part in forecast performance and can help determine how observations are treated during the assimilation process. For this project the first guess and boundary conditions for the high resolution ensemble members are obtained from two sources: the Earth System Research Lab (ESRL) Rapid Refresh (RR) Model and an AWS Truepower Ensemble Kalman Filter (EnKF) prediction system. The EnKF prediction system is an attractive method to initialize high resolution ensemble members because it offers estimates of forecast uncertainty along with a range of forecast solutions including the ensemble mean. The WFIP EnKF prediction system is composed of 24 Weather Research and Forecasting (WRF) numerical weather prediction ensemble members assimilating observations intermittently through the Data Assimilation Research Testbed (DART) EnKF scheme. The EnKF ensemble produces a 30 hour forecast every 6 hours using the Global Forecast System (GFS) as boundary conditions and a previous EnKF forecast as the initial conditions. Data is assimilated into the EnKF from a number of different sources including, satellite derived winds, rawinsondes, and surface observations. This presentation will focus on the performance of the real-time EnKF prediction system and its impact on the Advanced Regional Prediction System (ARPS) Data assimilation System (ADAS) and the WRF Variational (VAR) data assimilation system when the EnKF is used for the initial and boundary conditions. Also, the use of various background model states (ranging from a specific EnKF ensemble member representative of the ensemble mean to the ensemble mean itself) and their impact on an 80-m wind speed forecast will be shown. A hybrid technique, which uses ensemble spread to estimate background error covariance in the WRFVAR data assimilation system and its impact on the high resolution WRF forecasts will also be discussed.