Two types of high density GOES satellite derived winds, cloud-drift infrared and cloud-topwater vapor winds, are now used operationally at NCEP. Various data thinning algorithms have been tested to reduce the effect of correlated errors in assimilating these winds in the NCEP globe data assimilation system. Numerous experiments have been conducted with T62 global data assimilation and forecast system from August 1 to 31, 2000. These experiments show that thinning overall improves the forecast skill for all thinning algorithms, but the improvements are extremely small. Based on the standard test scores, the best thinning algorithm is to screen the data based on the quality indicator (QI) developed by EUMETSAT with pre-filtered by recursive filter flag (RFF) developed by Cooperative Institute for Meteorological Satellite Studies (CIMSS). This algorithm will be implemented into operation later in 2001