Quality control (QC) is essential before assimilating Global Navigation Satellite System (GNSS) radio occultation (RO) data. Conventional QC methods, applied globally or by latitudinal bands, can inadvertently remove valid data that contain significant meteorological signals. Using Constellation Observing System for Meteorology, Ionosphere, and Climate-2 (COSMIC-2) bending angle data—a primary RO observable—from the summer of 2023, we demonstrate that this latitude-based QC (LatQC) flags an excessive number of outliers near an altitude of 8 km, a region typically considered highly reliable. These outliers are not random but are strongly clustered near the Solomon Islands, coinciding with frequent summer altostratus clouds with tops exceeding 6 km, which systematically perturb bending angles. Another area with high outlier concentration is the moist lower troposphere within the Intertropical Convergence Zone (ITCZ). Although GNSS RO signals are less affected by clouds due to their long wavelength, the aggressive LatQC in cloudy areas often misclassifies real weather-induced signals as noise. To address this, we developed a local QC method (LocQC), establishing statistical benchmarks within 2.5° × 2.5° longitude–latitude grids instead of broad latitudinal bands. The LocQC significantly improves the spatial distribution of identified outliers: it mitigates over-removal in the 6–8 km layer over the Solomon Islands and aligns outlier detection better with liquid water path and humidity patterns in the 2–4 km layer of the ITCZ. The quality of RO bending angle data is influenced not merely by latitude but significantly by local weather systems with distinct longitudinal signatures, such as clouds and moisture. Consequently, the proposed LocQC, accounting for two-dimensional atmospheric structures, is more appropriate than LatQC. It ensures robust filtering of erroneous data while better preserving valuable local weather signals crucial for numerical weather prediction and climate studies.

This study first modifies a conventional method for locating the center of a southwest vortex (SWV) and applies it to center identification in the fifth generation ECMWF reanalysis (ERA5) and the Global Forecast System (GFS) analysis from the NCEP of the U.S. for an SWV event on 26–27 June 2020. Significant discrepancies are found between the SWV trajectories in ERA5 and GFS. Secondly, according to observations from the Advanced Himawari Imager (AHI), a row of isolated clouds over the eastern Qinghai–Xizang Plateau (QXP) exhibits characteristics of mountain gravity waves and generates SWVs as they propagate eastward. The SWV-related cloud/rain structure in cloud-top pressure and cloud type varies consistently with the observed lower brightness temperatures (TBs). Thirdly, using ERA5 and GFS as inputs to a radiative transfer model, we generate all-sky simulations of TBs at AHI channel 13. On the synoptic scale, the ERA5 simulations more closely match the observed TBs, whereas for the small-scale cloud/rainband structure of the SWV, the GFS-simulated TBs bear a closer resemblance to the observations. The ERA5 liquid water clouds extend to altitudes as high as 140 hPa, with temperatures as low as −32°C. The AHI liquid water cloud tops are located in the middle and upper troposphere, while the GFS liquid water cloud tops are confined to the lower troposphere.

The world witnessed an accelerated development of various types of meteorological observing technology, an evolution of numerical weather prediction (NWP) models from single atmospheric component to coupled multi-components of the earth system, as well as the multi graphics processing unit technology in computer sciences, a new era for rapidly advancing data assimilation science and technology development has arrived. The multi-source data assimilation is important not only for NWP but also for further understanding of global and regional weather changes. This article firstly selectively reviews past methods of multi-source data assimilation. New opportunities are then discussed for future development of data assimilation system framework, for innovative uses of high-resolution observations, and for applications of artificial intelligence machine learning in meteorological data assimilation.

Geostationary Operational Environmental Satellite-16 (GOES-16) Advanced Baseline Imager (ABI) observations of brightness temperature (TB) are used to examine the temporal evolutions of convection-affected structures of Hurricane Irma (2017) during its rapid intensification (RI) period from 0600 to 1800 UTC 4 September 2017. The ABI observations reveal that both an elliptical eye and a spiral rainband that originated from Irma’s eyewall obviously exhibit wavenumber-2 TB asymmetries. The elliptical eye underwent a counterclockwise rotation at a mean speed of a wavenumber-2 vortex Rossby edge wave from 0815 to 1005 UTC 4 September. In the following about 2 hours (1025–1255 UTC 4 September), an inner spiral rainband originated from the eyewall and propagated at a phase speed that approximates the vortex Rossby wave (VRW) phase speed calculated from the aircraft reconnaissance data. During the RI period of Irma, ABI TB observations show an on–off occurrence of low TB intrusions into the eye, accompanying a phase lock of eyewall TB asymmetries of wavenumbers 1 and 2 and an outward propagation of VRW-like inner spiral rainbands from the eyewall. The phase lock leads to an energy growth of Irma’s eyewall asymmetries. Although the eye remained clear from 1415 to 1725 UTC 4 September, an inner spiral rainband that originated from a large convective area also had a VRW-like outward propagation, which is probably due to a vertical tilt of Irma. This study suggests a potential link between convection sensitive GOES imager observations and hurricane dynamics.