Currently, the hourly forecasts of precipitation are produced by forecasters using several NWP model outputs, and the precipitation start/stop times are generated by radar extrapolation algorithms. Therefore, inconsistencies between the two short-term forecasts are unavoidable. The unified nowcasting is intended to solve the problem of inconsistencies while aiming at a higher temporal and spatial accuracy.
The advantage of radar observations is in their frequency of update and high spatial resolution (1 km) which makes the radar extrapolation a favorable option for precipitation nowcasting within the first few hours. However, the skill of radar nowcasts drops rapidly with the forecast lead-time. On the other hand, NWP forecasts provide a lower but more stable skill over a longer period. The HRRR sub-hourly forecast with its hourly update cycles and 3 km spatial resolution provides an opportunity to generate a seamless blend with radar extrapolations.
The performance of The Weather Network’s unified nowcasting is investigated and demonstrated in several cases, including a convective rain, a stratiform rain, a snow storm, an ice pellet and a freezing rain events during 2015 and 2016 in selected Canadian locations. Local METAR observations and radar observations are used for verification of timing, intensity, and type of the precipitation. The results show that the unified nowcasting maintains a high accuracy within the first six hours and outperforms both hourly forecasts and radar extrapolation within hour 2-3.