Pre-Processing, Quality Assurance, and Use of Global Atmospheric Motion Vector Observations in CRA

Assimilation of atmospheric motion vectors (AMVs) is important in the initialization of the atmospheric state in numerical weather prediction models, especially over oceans and at high latitudes where conventional data are sparse. This paper presents a detailed description of the pre-processing, quality assurance, and use of global AMVs in China’s first generation of the 40-yr (1979–2018) CRA global atmospheric reanalysis product. A new AMV archive is integrated from near real-time operational Global Telecommunication System data and reprocessed AMV datasets released or produced mainly during 2014–2016 according to a priority principle. To avoid the misuse of data with systematic quality problems, the observations of all 18 types of AMVs from 54 satellites are pre-evaluated over the whole time series. The pre-evaluation system developed by the CRA team is based on the NCEP Gridpoint Statistical Interpolation (GSI) three-dimensional variational assimilation system and the ERA-Interim reanalysis product. The AMVs in the new AMV archive are denser than the AMVs prepared for the Climate Forecast System Reanalysis product, the bias and root-mean-square values are smaller, and the time series are steadier. The new AMV archive is assimilated in the CRA product based on the NCEP GSI assimilation procedure and quality control configuration with reference to the pre-evaluation results. This is the first time that the reprocessed AMVs from Fengyun-2 satellites from June 2005 to July 2017 are assimilated in a reanalysis product. The assimilation features inspire confidence in the accuracy and stability of these data. The mean root-mean-square values of the observation minus analysis infrared, water vapor, and visible AMV were 1.5–3.4, 2.7–3.6, and 1.3–2.1 m s⁻¹, respectively. This experience of integrating, pre-evaluating, and assimilating AMV observations is valuable for the next generation of reanalysis products.