Detection of Convective Storms From PPI Images of Polarimetric Doppler Weather Radar Using Deep Learning

The convective storm is caused by the convection process in the atmosphere which may lead to heavy rainfall, thunderstorms and other extreme weather events. This paper suggests an innovative approach for detecting convective storms. The convective storm occurs in a short duration and over a small spatial scale. In this work, the Polarimetric Doppler weather radar data is used for detecting these convective storms since it has a large spatial and temporal resolution. The Plan Position Indicator (PPI) plots of horizontal reflectivity are plotted from the radar data. The PPI images are labeled as convective or non-convective based on the hourly rainfall data at similar periods, collected from rain gauges in the same area. This labeled dataset is used to train a Convolutional Neural Network model and obtained an accuracy of 95.91 percent using a test dataset. The result suggests the potential use of deep learning models for detecting convective storms. This method can be used for creating labeled datasets which can be used in training deep learning models for forecasting and classification of convective storms.