The Role of Observation Uncertainty in the Verification of Global Forecasts of Convective Hazards

In preparation for an upcoming global convective forecast verification task, an investigation into possible observation sets was undertaken. Radar is a reliable and trustworthy observation, but has limited coverage around the globe. Lightning, also, is a sure indicator of convection but with detection efficiencies that can be quite low over certain regions. Satellite observations investigated include radar onboard the Global Precipitation Measurement (GPM) mission core observatory, the Integrated Multi-satellitE Retrievals for GPM (IMERG) rainfall estimates, cloud top height (CTH) fields and anvil/overshooting top fields derived from geostationary satellites. The GPM core satellite has global radar coverage but with a very limited view at any instant. The anvil/overshooting top fields focus on features that are not present with every thunderstorm, while the IMERG and CTH fields are focused on features that are not necessarily convective.

In the absence of global data set observing thunderstorm features, it is necessary to make use of multiple observations sets, utilizing their strengths and accommodating their weaknesses. For instance, where lightning and overshooting top observations exist, there is a high confidence of the presence of thunderstorms, but the absence of either does not necessarily imply the absence of a thunderstorm. Conversely, thunderstorms will not be found where there are no clouds, so the CTH field can act as a bound in the opposite direction. In addition to the uncertainties resulting from the nature of the instruments, the uncertainties arising from the spatial and temporal mismatch between forecasts and observations will also be explored.