Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

Exchange experience with 4DEnVar

satjemkes

New member
Hi,

We are running wrf 4DEnVar and I have a couple of questions regarding its use. Is there anybody in this forum who is willing to exchange emails on this topic. The questions are related to the fundamentals of 4DEnVar, answers to which i can not find in the publications.
To give an example: OE theory indicates that at any time in the DA-window the analysis is optimal. Though traditional the analysis time is taken at centre point of the DA-window, according to theory we could either pick end or beginning. But that is only the theory. Is there any evidence that an analysis at the end of the DA window versus an analysis at the beginning generates comparable results? The concern i have is that I do not see that the information at either sides are used for the analysis of the end respectively beginning of the DA-window. An analysis time is provided by the user (e.g. beginning of window) but i do not see that the data at end of the window is being used. Or maybe i am not looking at the right place. A related question is the be.dat. In OE there is an expression for the state and for the error of the state (that is why OE is preferred over other methods), I do not see that the be.dat input is updated by the 4DEnVar system. Maybe this is a setting but there is also no indication in the basic papers about this. So do i miss something?

I also have some more technical questions: MPI and Fortran: it appears that IF-ELSE-ENDIF constructs should be avoided in the coding of a da_module (to avoid race conflicts). Can this be confirmed? I have noted that my module produced strange results. after removing all if statements within the do-enddo cycle it appears better, but i would like to get a confirmation that the if-else-endif can cause

Any feedback would be appreciated.
 
Top