Modelling of Spatio-Temporal Variability of Extreme Rainfall Events and Improving Model Forecast using Data Assimilation

Predicting extreme rainfall is a critically important endeavour in the field of meteorology, as damages caused by it have profound impacts on the human life and economy. Every year, many people across the globe endure the devastating consequences of extreme rainfall events (EREs) like flood, agricultural loss etc. This research aims to enhance the prediction of EREs by refining the representation of physical processes in the weather research and forecasting (WRF) model through parameterization schemes and improving the initial atmospheric conditions via data assimilation technique. A comprehensive study is conducted over India, specifically Karnataka, to examine the spatio-temporal variability in rainfall as well as heavy rainfall. It reveals that certain regions in India exhibit a negative trend in rainfall during different seasons, while others experience a significant positive trend. The southern state of India, Karnataka demonstrates a highly diverse rainfall pattern. The northern and southern parts of the state fall into one group, while the coastal area is categorised under another group. This study highlights the drastic increase in rainfall, as well as heavy rainfall events (>100mm/day), in the south-interior region of Karnataka, accompanied by a decrease in coastal Karnataka. The significant change in rainfall patterns and the rise of heavy rainfall events in unexpected regions of Karnataka highlighted the need for accurate modelling and prediction of EREs in the state. Identification of EREs poses a significant challenge in the field of climate science due to the complex and diverse spatio-temporal variation in rainfall pattern. One of the main challenges is selecting the appropriate threshold of EREs for each region. To address this, a flood modelling study is conducted using the Strom Water Management Model (SWMM) to analyse four rainfall events with different intensities. This study aims to identify regions and rainfall intensities that are prone to floods. Specifically, the study focused on the KC Valley, a low-lying area in the highly urbanized metropolitan city of Bengaluru, Karnataka. It is one of the four major watershed valleys in Bengaluru and frequently experiences flash floods during heavy rainfall events. The study proposes a threshold of 60mm/day for identifying EREs over Bengaluru, ensuring that the identified events are both representative of extreme rainfall and indicative of flood-prone conditions. Accurate modelling and prediction of EREs hinges on factors like selecting an appropriate parameterization scheme that accurately represents the specific physical processes and establishing optimal conditions to initialise the model. Cloud microphysics processes are particularly crucial in determining rainfall characteristics. Improving the representation of microphysics can enhance the accuracy and reliability of ERE prediction. To achieve this, a study is conducted using the WRF model, focusing on 30 EREs (15 localised and 15 nonlocalised) occurred in Bengaluru city. The study employs four domains with different horizontal resolutions, with the fourth domain (1.33km) specifically focusing on Bengaluru city. The results indicate that the model effectively captured atmospheric parameters associated with both types of EREs using various microphysics schemes, particularly for nonlocal EREs. However, the accuracy of the model forecasts varies depending on the chosen microphysics schemes. The New-Thompson and Morrison double moment scheme exhibit superior performance in predicting both nonlocalised and localised EREs compared to other schemes. Furthermore, this research aims to enhance the accuracy of ERE prediction at a regional scale by utilizing various Global Navigation Satellite Systems (GNSS) observations to improve initial conditions. Specifically, the study focuses on assimilating GNSS ground-based and space-based observation data to improve forecast skill. The study conducts experiments for 16 EREs occurring between October 2019 and December 2020 in Karnataka state. Using the 3D-Var technique, zenith tropospheric delay data from GNSS ground-based observations and atmospheric refractivity data from GNSS space-based observations are assimilated into the model. Results indicate significant improvements in rainfall forecasting throughout Karnataka, particularly in the interior region, through the assimilation of both types of GNSS observations. Furthermore, experiments with newly launched COSMIC2 observations demonstrate substantial enhancements in weather parameters such as humidity, temperature, and rainfall prediction compared to other GNSS observations. These findings underscore the important performance enhancement of the model in predicting EREs through the assimilation of COSMIC2 observations.