Opened 3 weeks ago

Last modified 3 weeks ago

#244 new Support

Problem in grid area size

Reported by: slmeyer Owned by:
Priority: major Milestone:
Component: FP physics/numerics Version: FLEXPART-WRF
Keywords: grid area Cc:

Description

Dear FLEXPART-WRF Community,

I have some problems using WRF fields in very high resolution. My set-up looks as follows:
Using WRF I created up to 5 Domains, leading to a horizontal resolution up to 40m.
My FLEXPART(-WRF) outgrid I desgined in a way to lay directly on the WRF Grid (see attached config file). The origin is shifted half a grid box to the upper right corner. The reason for this is, that I noticed, that the difference of latitude and longitude in the centre of each gridbox in FLEXPART and WRF is smaller when shifting the FLEXPART grid than without.

When looking at the GRIDAREA output inside the header file I noticed a very large range of grid areas as you can see in the attached image 'gridarea_40m.png' (please note the logarithmic scale in this plot)

As you can see, the grid area varies by one magnitude in a systematic pattern. Also there are some very small grid areas (three to four magnitudes smaller) spread around the domain.
My question is, why is this happening?

Due to a coord transfo error (compare ticket #168, https://www.flexpart.eu/ticket/168) I slightly increased the threshold in the map_proj_wrf.f90 file, because the MAPFAC_* values in the corresponding WRF fields seem to be ok (between 1.000075 and 1.000143, so below 2%).

But the strange behaviour of grid areas also occurs in my coarser domains with spatial resolution of 200m and 1000m where I did not touch the threshold. Images of this are also attached to this ticket (see gridarea_200m.png and gridarea_1000m.png).

You can see that in the case of 200m spatial resolution I seem to have an increase of variation when going to the right, while in the 1000m case a have two vertical stripes in the middle of the domain.

I hope someone can help me to understand where the problems come from and to find a solution (best without re-runing all the WRF runs).

Best regards
Sarah

Attachments (4)

40m_run (14.4 KB) - added by slmeyer 3 weeks ago.
Cofig file of my 40m run
gridarea_40m.png (271.5 KB) - added by slmeyer 3 weeks ago.
Grid area with grid designe to have 40m grid distance
gridarea_200m.png (289.5 KB) - added by slmeyer 3 weeks ago.
Grid area with grid designed to have 200m grid distance
gridarea_1000m.png (238.5 KB) - added by slmeyer 3 weeks ago.
Grid area with grid designed to have 1000m grid distance

Download all attachments as: .zip

Change History (5)

Changed 3 weeks ago by slmeyer

Cofig file of my 40m run

Changed 3 weeks ago by slmeyer

Grid area with grid designe to have 40m grid distance

Changed 3 weeks ago by slmeyer

Grid area with grid designed to have 200m grid distance

Changed 3 weeks ago by slmeyer

Grid area with grid designed to have 1000m grid distance

comment:1 Changed 3 weeks ago by pesei

Obviously, the problem has to do with the map projection. I am wondering whether 40 m grid distance works in WRF well at all. One attempt could be to have the compiler double the precision globally. If this helps, one could try to increase precision where actually needed.

As for feedback from the community, it would be best to post on the FLEXPART mailing list.

Note: See TracTickets for help on using tickets.
hosted by ZAMG