A method for automatic verification of thunderstorm nowcasts

India Meteorological Department (IMD) operationally issues district-level nowcasts for the occurrence of thunderstorms and associated weather for all districts of India round the clock at three-hour intervals since 2018. These impact-based nowcasts are updated in real-time on the IMD website (https://mausam.imd.gov.in/responsive/districtWiseNowcast.php) and are also communicated to various state and national level stakeholders during thunderstorm events. While a Standard Operating Procedure (SOP) utilizing climatology, observations from various sources, and direct and derived model products is in place for issuing nowcasts, objective verification of these district-level nowcasts has presented a challenge till date. The current study details a method for automatic verification of these district-level operational nowcasts of thunderstorms w.r.t. observations of lightning as a measure of the occurrence of thunderstorms. The data from the ground-based lightning sensor network of the Indian Institute of Tropical Meteorology has been used as the observation dataset. The method is customizable to automatically derive categorical verification scores on a daily, monthly, or seasonal basis or for any arbitrary time period of choice. The verification of the three hourly nowcasts on the monthly scale for the period of April–November of 2022 and 2023 indicates that the all-India Probability of Detection (POD) verification score was in the range of 0.57–0.62, the False Alarm Ratio (FAR) score was about 0.70 to 0.77, the Critical Success Index (CSI) score was in the range of 0.20–0.24 while the Equitable Threat Score (ETS) score was in the range of 0.15–0.20. Verification results for individual districts at a seasonal scale during 2022–2023 indicate that scores were poor during the winter season (<0.2) (except over southeast peninsular India) and improved thereafter during the rest of the year. This may be due to the fact that thunderstorms during the winter season are infrequent, comprise weak cells with short lifetimes over most parts of India (except the southeast peninsula). In non-winter months, the POD score was generally high over the plain regions of the north peninsula, central, and North India, with the best score during the monsoon months (>0.8). However, the FAR score was better (lower values) over peninsular India (0.2 to 0.4). This translated to improved CSI and ETS over peninsular India. FAR score was generally better during the pre-monsoon season compared to other seasons. These district-level nowcast verification results were not confined as per state boundaries but were common across states and parts of multiple states and other years, for each of which a different meteorological centre issues nowcasts. The spatially systematic errors over East India are due to the nature and movement of weather systems over this region. The spatially systematic errors over the western and eastern Himalayan region may be a consequence of the quality of the lightning detection network data used for verification over this region.