Product Usage
NCWF-2 produces rapidly-updating probabilistic forecasts of convection that is likely to be hazardous to aviation. The system produces 30, 60, 90 and 120 min probabilistic forecasts every 5 minutes. The frequent update cycle is required to capture rapidly changing nature of thunderstorms that evolve over very short time scales (order 30 min).
The goal of this system is to provide the most up-to-date information on current convective activity and forecasted probable storm locations across the CONUS for a range of aviation users involved in strategic planning including aviation weather forecasters, airline dispatchers, general aviation, and FAA Traffic Management Units and Center Weather Service Units (CWSU). The probabilistic nature of the forecast gives the user additional information over that provided by a deterministic (future “snap shots”) forecast by indicating the uncertainty in the forecasted storm locations. It is noted that forecasts of the storm intensity are implied since this system has been explicitly designed to forecast only those storms likely to impact aviation interests (i.e., storms of VIP 4 or greater).
This section gives an example of how these probabilistic forecasts can be used in both tactical and strategic decision making by the users of aviation weather forecasts.
Probabilistic vs Deterministic Forecasts
A deterministic forecast of convection can be easily produced by identifying existing storms, determining their movement and extrapolating them to their future locations. This is done by a number of available systems including NCWF1 and CIWS. These forecast systems generally assume that the storm motion and organization will remain constant for the forecast period. However, it is well-known that convective storm processes are rapidly evolving so that these assumptions can quickly become invalid, particularly at the longer forecast lead times.
A probabilistic forecast may be generated to account for uncertainties in storm motion and evolution in the storm forecast. For example, if known errors in the motion vectors for a lead time of 2 hr are +/- 5 knts and 30 deg, then the location of the storm center can be off by as much as 30 nm. This is the worst possible forecast assuming the largest possible storm motion error. In reality there is a distribution of possible outcomes due to the range of possible errors and their likelihood of occurrence. The uncertainty of convective storm predictions increases with forecast lead-time and smaller storms are less predictable than larger organized systems (e.g., squall line). This forecast uncertainty can be expressed by applying a spatial filter to the extrapolated field (in NCWF2 we extrapolate a special field called the Hazard Detection) that increases in size with forecast lead-time (see Forecast Section).
Comparison of 2-hr deterministic (WSI reflectivity in dBZ) and probabilistic (mouse over, % likelihood) forecasts of convection.
The figure above can be toggled to see the difference between a deterministic forecast (in which radar reflectivity from WSI is extrapolated) and a probabilistic forecast generated by NCWF2 (which indicates the likely future locations of storms with VIP level 4 and greater). These forecasts were generated with the same set of motion vectors. Toggle the image to see the overlap of the high probability regions with the higher reflectivities. Note the broader area of possible storm locations indicated by probabilistic forecast. The deterministic forecast contains no information as to the certainty of the forecast.
Two-hour deterministic (color scale, dBZ) and probabilistic (contoured in %) forecasts of convection.
This is demonstrated future in the figure just above in which there is a 120 km wide line of probabilities greater than 10% running SW-NE across Mississippi.
Probabilistic Forecast to Deterministic NEXT >