Improving nowcasting by blending extrapolation and NWP model forecasts

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 6 January 2015: 9:15 AM
129A (Phoenix Convention Center - West and North Buildings)
Yunsung Hwang, CIMMS/Univ. of Oklahoma/NOAA/NSSL, Norman, OK; and V. Lakshmanan, A. Clark, and S. Koch

Improving short-term (0-8 hours) forecasts (nowcasts) is important for the prevention of accidents in the fields of aviation, agriculture, construction, and emergency management. Predicting possible locations of convective cells can lower the economic costs associated with rescheduling or rerouting commercial flights in aviation. As numerical weather prediction models increase in skill and spatial and temporal resolution, it has become feasible to use predicted weather fields to support decision making. However, extrapolation forecasts remain more skillful over the very short term (0-2 hours). It has been suggested that blending extrapolation and model forecasts would provide improved skill over using only extrapolation or only numerical weather forecasts. In this paper, we describe a novel image processing method of morphing based on saliency-based cross-dissolve between extrapolation and model forecasts to create nowcasts.

The proposed method is applied and tested using echo-top heights from Weather Surveillance Radar 1988 Doppler (WSR-88D). This observed data is morphed with forecasts of echo top heights from High Resolution Rapid Refresh (HRRR) model in the convective season in Continental United States (CONUS) from mid-May to mid-June. Two skill scores based on neighborhood method and airplane-route-based method showed that the saliency-based cross-dissolve method performs better than advection, HRRR or a linear cross-fade method in the entire 0-8 hour forecast period. The new method improves upon the individual and cross-fade methods by emphasizing the strong intensities from nowcast and forecast, thus capturing the information about the growth and decay of storms that is present in longer-period forecasts from the HRRR.