A technique for diagnosing the likely occurrence and severity of convection

A dataset consisting of over 100,000 convective and non-convective events for the period 1957-1994, for the continental United States east of the Rocky Mountains, is assembled. The data include measures of time and location (station, date, time of day, distance from radiosonde location), convective energy and triggers (CAPE, CIN, moisture convergence, topographic lift), and synoptic environment (Miller type, based on a spatial correlation analysis).

These data are randomly split into development and testing datasets, and a robust classification procedure is used to successfully diagnose the occurrence and severity of convection, given these inputs. The applicability of this method to the operational evaluation of convective occurrence and severity will be discussed, as will plans to apply this methodology to climate model predictions to evaluate the possible impact of climate change on convective weather.