Improving lightning representation in global model through convective cloud based Price and Rind 1992 scheme.

Accurate representation of lightning activity in coarse-resolution medium-range weather forecasting models remains a major challenge due to the fine scale nature of electrification processes. In the National Centre for Medium-Range Weather Forecasting Global Unified Model (NCUM-G), lightning flash rates are computed following the widely adopted Price and Rind (1992) (PR92) formulation. However, the current parameterization relies on storm detection criterion solely a cloud-thickness threshold, resulting in a large number of false alarms over India. To address the limitations of the current parameterization scheme, this study evaluates two key components of the lightning parameterization framework: (i) storm detection criteria and (ii) the formulation for estimating flash counts estimation, with a focus on enhancing performance over the Indian subcontinent. The analysis highlights pronounced seasonal variability in both the storm occurrence and convective intensity, highlighting the need for refined criteria for storm identification. Three approaches were explored to improve storm detection: (a) modifying the definition of cloud depth, (b) redefining cloud-depth limits, and (c) incorporating additional dynamical parameters along with cloud depth to account for seasonal variations in convection characteristics. Among all combinations tested, the approach that uses column-maximum vertical velocity and cold cloud depth for storm identification, together with convective cloud-top height in the PR92 flash-count formulation, shows the most promising results. Fractional Skill Score (FSS) analysis confirms that this modified scheme enhances the skill in representing the observed lightning flash counts in NCUM-G at medium-range time scales. Consistent with this, domain-scale categorical verification metrics, including critical success index (CSI) and equitable threat score (ETS), indicate that the modified configuration achieves a more balanced performance by reducing forecast bias and false alarms while maintaining improved detection skill, demonstrating a more realistic representation of lightning-producing convection. Despite these improvements, certain regional discrepancies persist, largely attributable to systematic model biases in simulating convective processes.