Harald Richter's research interest spans a variety of aspects of deep moist convection, primarily in the extra-tropics. His viewpoint on convection has become quite operational in nature, having worked as a thunderstorm and radar trainer at the Bureau of Meteorology in Australia between 2004 and 2009. Focal areas of his work in the past have been convective initiation, the diagnosis of Doppler radar signatures and, more recently, the day 1 convective forecast problem.
Recently Harald has been working on an ensemble-based calibrated probabilistic thunderstorm prediction methodology similar to the approach presented by David Bright and co-authors at the Conference on Severe Local Storms in 2005. Additionally he is investigating selected Australian severe thunderstorm case studies with an aim of improving their operational predictability and the associated warning decisions.