[CF-metadata] ... wind speed

From: Jonathan Gregory <j.m.gregory>
Date: Mon, 21 Feb 2011 11:25:14 +0000

Dear Martin

> 15,WINDSPEED,m/sec,"Vector wind speed; m/sec.",NUMBER,"16,4",
> and
> 23,WINDSPEED_SCALAR,m/sec,"Scalar wind speed; m/sec.",NUMBER,"16,4",
>
> The first appears to be a vector average = sqrt( mean(u)**2 + mean(v)**2 )
> the second is = mean( sqrt(u**2 + v**2) )
>
> Are both tied to "wind_speed" as standard_name and should they differ in the cell_method attribute?

Yes, I think these are both wind_speed. The difference between them is some-
thing like cell_methods, but it doesn't entirely fit into cell_methods, which
is a way to record how the value given relates to subgrid variation. The second
of these is really a time-mean in the sense that cell_methods implies, I would
say; wind_speed is defined at higher temporal resolution and your values are
means over time-intervals. The first isn't really a time-mean. It's as if the
data from which wind_speed had been calculated had a low temporal resolution in
the first place, since the higher temporal resolution has already been lost by
calculating the time-mean of u and v. Since this is an intensive quantity, if
it's not a time-mean, it really has to be a "point" as if it were instantaneous
in cell_methods. That doesn't sound quite right, does it? It makes sense in the
respect that, for model output, you would probably record the u and v, and
hence the wind_speed, as instantaneous (point) values in time, even if the
model timestep was very long, because for a model there is truly no information
at higher temporal resolution than the timestep.

Best wishes

Jonathan
Received on Mon Feb 21 2011 - 04:25:14 GMT