Large-scale trending of radar data

Tsonis and Austin (1981) first investigated the use of trends in echo size and intensity to improve forecasts out to 30 minutes. Trending was further tested by Wilson et al. (1998) using the Thunderstorm Identification, Tracking, Analysis and Nowcasting system (Dixon et al., 1993) for forecasts ranging between 6 and 36 minutes. In both cases, little if any improvement was found. Wilson et al. (1998) concluded that "essential physical processes that dictate the change in rainfall with time are not necessarily observable in the past history of a particular echo development." However, in an analysis of 0-2 hour forecasts, MIT/Lincoln Labs suggests that trending provides benefit. The difference in these studies is one of scale. In the earlier studies, individual cells were tracked and trended. In the MIT/Lincoln Labs work, the area change of the region around an individual grid point was evaluated. Instead of evaluating whether an individual cell was growing or dissipating, all the cells within a region were examined.

The National Convective Weather Forecast - 2 (NCWF-2) product provides a short-range (1 and 2 hour) probabilistic nowcast of convection based on radar, lightning and RUC data. The NCWF-2 software trends large-scale growth and dissipation. Trends are calculated in Lagrangian space (based on motion vectors at forecast time) using a weighted linear fit of a mean value within a given radius circle over a given period. This paper reviews the trending methodology and shows results of applying large-scale trending to the 1 and 2 hour forecasts.