Atmospheric Convection Research And Operational Forecasting Aspects PDF Download

The advent of modern geostationary satellite infrared (IR) imagery has ushered in a new era in atmospheric observations. In recent years, tremendous progress has been made towards improving the forecasts of tropical cyclones and mid-latitude weather by the ensemble data assimilation (DA) of these IR observations. However, there is comparatively little work on assimilating these observations to improve the analysis and forecast of tropcial mesoscale convective systems (MCSs). Because these systems are the primary source of rainfall over the Tropics and influence global weather and climate, it is essential to accurately predict these systems. As such, the ultimate goal of my dissertation research is to improve the analysis and forecast of tropical MCSs through assimilating geostationary infrared radiance (GeoIR) observations.\\ My dissertation research has two aspects: 1) demonstrating the impacts of GeoIR ensemble DA on the analysis and forecast of tropical MCSs, and 2) devising a new computationally efficient ensemble DA algorithm to better exploit GeoIR observations. To demonstrate aspect 1, I start with assimilating real GeoIR water vapor channel brightness temperature (WV-BT) observations from the Himawari-8 geostationary satellite in a tropical squall line case over the Sumatra-Malaysia-Borneo region. The ensemble DA is executed through the state-of-the-art Pennsylvania State University Ensemble Kalman Filter (PSU-EnKF) system. Relative to a control experiment where only in-situ and satellite-derived atmospheric motion vectors (AMVs) are assimilated, the inclusion of WV-BT observations has been found to improve the analysis and forecast of the tropical squall line. Specifically, the analyses' squall line outflow positions and the analysis and forecast of the squall line's clouds were improved. These improvements scale with the frequency of GeoIR DA.\\ Since no MCS-resolving tropical reanalysis product currently exist over the Maritime Continent and GeoIR DA has demonstrated positive impacts in my tropical squall case, I have created a high-resolution tropical MCS reanalysis (TMeCSR). This is done by combining an ensemble of MCS-resolving model runs (9-km horizontal grid spacing), in-situ observations, AMV observations and WV-BT observations through the PSU-EnKF. To further enhance the TMeCSR, I have also introduced large-scale information from the European Center for Medium-Range Weather Forecasts (ECMWF) Reanalysis version 5 (ERA5). The TMeCSR is available during June, July and August of 2017, and spans a region covering most of the tropical Indian Ocean, most of continental Asia, the Maritime Continent, and the West Pacific. \\ Comparisons of TMeCSR and ERA5 against independent satellite retrievals indicate that TMeCSR's cloud and multiscale rain fields are better than those of ERA5. Furthermore, TMeCSR better captures the diurnal variability of rainfall and the statistical characteristics of MCSs. Forecasts that are initialized from TMeCSR also have more accurate rain and clouds than those that are initialized from ERA5. The TMeCSR and ERA5 forecasts have similar performances with respect to sounding and surface observations. These results indicate that TMeCSR is a promising MCS-resolving dataset for tropical MCS studies.\\ The second aspect of my dissertation research is to create a new computationally efficient ensemble DA algorithm: the bi-Gaussian ensemble Kalman filter (BGEnKF). The BGEnKF is motivated by the differences in the atmospheric dynamics, and thus the statistics, of clear atmospheric columns and cloudy atmospheric columns. Unlike the EnKF, which does not distinguish the differences between these two types of columns, the BGEnKF explicitly treats clear atmospheric columns separately from cloudy atmospheric columns. Furthermore, unlike earlier formulations in the literature\footnote{Note that the BGEnKF is a two-kernel form of Gaussian mixture model EnKFs (GMM-EnKFs). As such, the ``earlier formulations'' reference here are really earlier formulations of GMM-EnKFs.}, my formulation is computationally efficient and does not require laborous derivations and programming concerning stochastic subspaces.\\ To examine the advantages of my BGEnKF over the EnKF, I have implemented the BGEnKF into the PSU-EnKF system and have performed observing systems simulation experiments (OSSEs) using a case of tropical convection over the equatorial Indian Ocean. This case occurred during the onset of the October 2011 Madden-Julian Oscillation event. Only synthetic infrared window channel brightness temperatures (Window-BT) from the Meteorological Satellite 7 are assimilated. My results indicate that the BGEnKF outperforms the EnKF in this semi-idealized setting. These performance advantages are found in the horizontal wind vector components, temperature, specific humidity and WV-BT fields. \\ My OSSE tests with the BGEnKF are among the first to test a Gaussian mixture model EnKF (GMM-EnKF) with a realistic weather model. The encouraging results motivate further testing and development of the BGEnKF algorithm. If the BGEnKF consistently outperforms the EnKF at assimilating GeoIR observations, the BGEnKF will replace the EnKF in operational weather forecasting systems in the future.