NOAA Grand Challenge Forecasts Verification

Onset and Duration Plots: Description and Examples

As defined in the Demuth WAF paper:

 Onset time plots: 
 Values indicate the hour of the day when a threshold is first exceeded at a given point.
 Blank areas indicate points where the threshold is not exceeded in any of the
 ensemble members, at any time.  This plot is an "average" of multiple ensemble members.
 If only one member exceeds a threshold during the entire time of interest, the color is
 assigned to the first hour of exceedance in that member; if multiple members exceed the
 threshold, the color is assigned to an average of the first hours in each member.

 Duration plots:
 Values indicate the total number of hours during which a threshold is exceeded, at a
 given point.  As for the onset time plots, ensemble averaging is applied.  If only one
 member exceeds a threshold during the entire time of interest, the color is assigned to the
 duration of exceedance in that member; if multiple members exceed the threshold, the
 color is assigned to an average of the exceedance duration from each member.

These definitions lead to spreading out of features as seen in these loops:

precip threshold duration over a 12-hour window: all hours, forecast hour 12, forecast hour 24, forecast hour 36, forecast hour 48, forecast hour 60, forecast hour 72
precip threshold onset time during a 12-hour window: all hours, forecast hour 12, forecast hour 24, forecast hour 36, forecast hour 48, forecast hour 60, forecast hour 72
wind speed threshold duration over a 12-hour window: all hours, forecast hour 12, forecast hour 24, forecast hour 36, forecast hour 48, forecast hour 60, forecast hour 72
wind speed threshold onset time during a 12-hour window: all hours, forecast hour 12, forecast hour 24, forecast hour 36, forecast hour 48, forecast hour 60, forecast hour 72

As depicted in the forecast hour 48 and 60 loops for item 4 above, notice how the coherence of the onset of the wind is smeared out by this. Also, since the above definitions allow it, a single member can and does dominate these plots; and, two members that have a wind or rain event enter the domain in different locations can make an event look really large:

Member 1	Member 2	Sample of others	Mean	Full Loop
				full loop
				full loop
				full loop