Need filter 'time-derivate' or 'tendency'

[ssenesi]

Some physical schemes may have a timestep shorter than the one used for xios_send(ing)_field.

In that case, time averaged values computed by Xios cannot be as accurate as time averaged values computed by a time-derivation on high-frequency-time-integrated values (assuming here that his time-integration is performed by the high-frequency physical scheme, at its high frequency)

This advocates for providing an Xios time-derivation operation, in addition to the already existing time-integration

[ymipsl]

Hi Stéphane,

What do you mean by time-derivation ? Time interpolation ? Could you give us a concrete example ?

Regards

[ssenesi]

I mean time derivation :

• let X(t) be the field, delivered to XIOS at t, t+dt, t+2*dt ...
• I want XIOS to compute Y(t) = dX(t)/dt , so, in details :

Y(0)=0
Y(t)=(X(t) - X(t-dt))/dt for t > 0