[A-AS03] Extreme Events and Mesoscale Weather: Observations and Modeling

[AAS03-20] Error Quantification in Multi-Physics and Initial Conditions

*Gaurav Tiwari¹, Atsushi Okazaki^1,2 (1. Center for Environmental Remote Sensing, Chiba University, 2.Advanced Academic Research (IAAR), Chiba University, Japan)

Keywords:Numerical Weather Prediction, Heavy Rainfall Events, Cloud-Microphysics Schemes, Large Member Ensemble

Quantification of uncertainties in initial conditions and parameterization physics is a decisive aspect of numerical weather prediction (NWP) modelling. In the NWP models, various physical processes, controlled by different mathematical formulations and numerical methods, are integrated. The precise representation of diverse processes and their interactions within the unified framework raises inherent complexity, causing model uncertainty. The initial conditions are also a hefty source of uncertainty/error in the forecast. These errors are primarily caused by several factors including inaccurate observational data, observation operator, and data assimilation algorithm used to initialize the model. This research quantifies uncertainties in model physics and initial conditions (MPIC) by employing Scalable Computing for Advanced Library and Environment Local Ensemble Transform Kalman Filter (SCALE-LETKF) generating a large member ensemble (800 members). The frequent manifestation of large-scale heavy rainfall events (HREs) has been evidenced in recent years. So, the heavy rainfall case of 27-28 July 2020 in the Tohoku region of Japan is considered in this regard. The SCALE model was initialized with 0.25^o x 0.25^o National Centers for Environmental Prediction (NCEP’s) FNL (Final) operational global analysis supported by the assimilation of NCEP’s Global Upper Air and Surface Weather Observations. A suit of sixteen combinations was designed using cumulus parameterization, cloud microphysics, and ocean dynamics within the scope of the SCALE model. The simulation setup consists of outer and inner domains with horizontal resolutions of 25 and 5 km, respectively. The findings highlight the importance of precise initial conditions and vigorous physical parameterizations in enhancing the reliability of complex simulations.