What is Nowcasting
Nowcasting is defined as forecasting the weather over the first 1 hour of time from the present. Typically, this involves extrapolation of current observation and trends into the immediate future. Extrapolation-based forecasting, i.e. nowcasting, leads to much better weather forecasts in comparison to numerical weather prediction models, or other known prediction methods, in this initial time frame.
Nearcasting, or short-range forecasting, covers the 1 to 6 hour time frame. Typically, numerical weather prediction models begin outperforming nearcasting or blended techniques after about 4 to 6 hours.
Nowcasting is particularly useful when forecasting complicated processes such as rainfall, clouds, and rapidly developing or changing storms. Nowcasting has been a viable method since the first radar and satellite data sets became available. For radar, this was roughly the late 1950s. For satellite, this was the early 1970s. Today nowcasting is as heavily supported in the research world as numerical prediction model development. The ultimate goal would be that both nowcasting and numerical weather prediction techniques will converge in terms of accuracy, but this is likely several decades away.
The main challenges associates with nowcasting are the timeliness and resolution of data. Specifically, forecasting clouds and rainfall on scales less than 5 kilometers is key, yet this demands large, voluminous data sets transferred and processed very quickly (on the order of 2 – 5 minutes). Fortunately, the newest radar data are available at 5 minute time and less than 1 kilometer space resolution, whereas the forthcoming geostationary satellites will provide data at up to 30 second time and 1 kilometer spatial resolution. Computer processing speeds and storage capability resources are rapidly advancing, which makes processing terabytes of data quickly, as required for nowcasting, possible. Advanced algorithms are continually under development which can exploit these new data streams.