Evaluating Empirical Lightning Parameterizations in Global Atmospheric Models

Lightning fundamentally influences atmospheric chemistry as it is the largest natural source of NOx in the upper troposphere. To include lightning, global models rely on experimental parameterizations. We use global atmospheric model diagnostics (2.0 x 2.5 degrees horizontal resolution, 38 degrees S-38 degrees N, for the years 2012-2013) to evaluate three lightning parameterizations that are based on: (1) cloud top height, (2) vertical ice-mass flux, and (3) meteorological variables including aerosols. We apply conversions between convective/cloud area and model grid box area in these calculations to compare each parameterization's spatial, temporal, and spectral characteristics with an observed lightning climatology from the Tropical Rainfall Measuring Mission Lightning Imaging Sensor. Domain-wide median values of parameterized lightning differ from observations by a factor of 0.6-3.1. The three parameterizations depict spatial (Pearson correlation (r = 0.80-0.84) and temporal patterns (r = 0.79-0.95) that match observations well. The parameterized median land-ocean lightning contrast ranges from a factor of 2.82 (using environmental factors with aerosols) to 596.0 (using cloud top height), compared to 22.9 in observations. One-to-one comparisons suggest that lightning parameterizations can explain approximately 55%-64% of the observed monthly lightning variance, depending on the parameterization. The highest correlation and minimum error bias statistics (e.g., Logarithmic Mean Bias, LMB) are found for the lightning parameterization that employs environmental factors with aerosols (r = 0.80, LMB = +0.002), whereas comparable correlation and generally higher error bias are found using vertical ice fluxes (r = 0.80, LMB = +0.27) and cloud top height (r = 0.74, LMB = -0.08). We show how the biases in lightning estimates are sensitive to uncertainties in the convective area and suggest ways to minimize bias overall.

Automated summary

Lightning plays a fundamental role in atmospheric chemistry as the largest natural source of NOx in the upper troposphere. Global models rely on experimental parameterizations to include lightning, with different parameterizations showing varying degrees of correlation and bias compared to observations. Utilizing environmental factors with aerosols for lightning parameterization yields the highest correlation and the smallest error bias, while biases in lightning estimates are shown to be sensitive to uncertainties in convective area.