Relaxation of Satellite Thinning for 4DEnVar

Typical methods for thinning of satellite data in 3DVar preserve only a single observation per thinning box within a data assimilation window. Among several factors that decide which observations are to be retained, are the proximity to the center of the spatial box and the center of the time window. Within 4DEnVar (or 4DVar), this paradigm can be relaxed by removing the second of these requirements, which allows more data to enter assimilation cycle, better describing temporal variability in a thinning box. This paper presents preliminary tests of such a relaxed satellite thinning algorithm within GSI's 4DEnVar, developed in preparation for assimilation of high temporal frequency of infrared channels on GOES-R. The developed algorithm will be analyzed in terms of its performance measured with standard metrics employed by GSI (such as, parallel scores, fitting of observations and the results of radiance monitoring software). In addition, the effect of increased number of satellite observations on minimization, the choice of an optimal size of the thinning box and the temporal bin will be investigated. The research methodologies of potential interest with an increased number of temporal data also include FOTO (First Order extrapolation To Observation), in order to pinpoint observation within its bin, and superobing, for an improved representativeness, and their applicability will also be reviewed. 37.171.116 on 7-30-2014-->