of severe convection, lightning, hail, turbulence and strong wind gusts. Traditionally,
the forecasting of lightning risk has been performed using large-scale indices, but
these often tend to over-predict the area at risk of experiencing thunder, especially in
marginal situations. An alternative to this approach is to parametrize the lightning
strike incidence directly and since 2015, this has been done operationally over the
UK and Europe using the Met Office Unified Model coupled to a lightning
parametrization (McCaul et al, 2009, Weather and Forecasting).
In this study, the performance of two high-resolution (4.4km and 1.5km) operational
Met Office NWP models over Europe at producing thunderstorms is examined over
several months. The model data is compared to the Met Office ATDnet lightning
detection system and a verification system is employed (Wilkinson 2017, Weather
and Forecasting) which examines thunderstorm forecasts in terms of three
properties: distance to lightning events, coverage and lightning strike intensity.
Results show that the model has a reasonable skill in forecasting thunderstorm onset
locally, especially over the summer months, but over the UK the intensity of the
forecast thunderstorms is overestimated, even when observational error is taken into