An Objective Quality Control of Surface Contamination Observations for ABI Water Vapor Radiance Assimilation

Abstract

A quality control (QC) process which handles surface impacts is an important step toward successful assimilation of the Advanced Baseline Imager (ABI) water vapor (WV) band radiances. If the QC is too relaxed, many surface contaminated radiances get assimilated. If the QC is too stringent, useful radiances are rejected. Either way can result in reduced or even compromised observation impacts. A new machine learning‐based QC scheme for the three ABI WV bands is developed and optimized to help understand the importance and effectiveness of the scheme. Unlike previous schemes which are dependent on the background, this scheme extracts and blends the surface information from 7 ABI bands (bands 8–10, 13–16) to determine if a WV radiance is affected by the surface. Simulation studies show that the new QC scheme is effective in retaining radiances that are either unaffected by the surface or have very small surface contamination. It is highly effective in rejecting radiances with large surface contamination. Numerical experiments from a single case study of Hurricane Harvey (2017) were carried out to optimize the QC and to understand the potential impacts on forecasts. The use of the new QC scheme shows that radiances from each WV band have substantially added value. Combining them has a positive impact on hurricane track forecasts compared with existing QC schemes. Hence, it is critical that an optimized QC scheme is used for infrared WV radiance assimilation. It provides a balance between positive impacts from useful radiances and negative impacts from surface contaminated radiances.

An Enhanced Storm Warning and Nowcasting Model in Pre-Convection Environments

“…Note that these sample sizes are for the convective candidates that were to be monitored and n via the RF model, rather than convective storm systems in synoptic definition. Inherited from the experience of SWIPEv1, the random forest (RF) algorithm, which is very useful for classification problems [47,48], was adopted as the machine learning framework for the enhanced SWIPEv2. For a detailed description of the RF algorithm, readers are referred to the classic paper on this method [42] and the brief introduction by coauthors of this article in the Appendix of the paper on SWIPEv1 [29].…”