Opened 7 years ago

Last modified 5 years ago

#168 reopened Support

the coordinate transfo error exceeds 0.02*dx

Reported by: victoria.sinclair@… Owned by: adingwell
Priority: minor Milestone:
Component: FP other Version: FLEXPART-WRF
Keywords: coordinate transform, mercator grid Cc:

Description

I am trying to run FLEXPART-WRF (forward) using WRF input data on a mercator grid. I keep getting the error:

test_xyindex_to_ll_wrf calling map_set -- map_proj_id, map_set_proj_code = 3 1

test_xyindex_to_ll_wrf -- lgrid, rmserr (km) = 0 1.03E+03


  • *
  • WARNING - the coordinate transfo error exceeds 0.02*dx *
  • *


problem with the projection. wrf+flexpart stops.

I have attached my input file I am using. The WRF output is at 9km grid spacing and has 249 points in the east-west direction and 255 points in the north-south direction. Any advice in how to solve this problem would be appreciated.

Attachments (1)

flexwrf.input.forward1 (6.4 KB) - added by victoria.sinclair@… 7 years ago.
input file

Download all attachments as: .zip

Change History (6)

Changed 7 years ago by victoria.sinclair@…

input file

comment:1 Changed 7 years ago by adingwell

  • Owner set to adingwell
  • Status changed from new to assigned

Hi Victoria,

the error you get is due to large scaling errors from your map projection. FLEXPART-WRF makes a test for this to prevent the user from setting up too distorted grids. You can view the scaling errors in your WRF-output (the MAPFAC_* fields). FLEXPART-WRF is hard-coded to abort if the error is above 2%.

You can solve this problem by:

  • Changing the map projection to something more suitable for the region you're studying
  • Setting up a smaller model domain in flexpart-wrf (the highest errors are usually near the domain borders unless you're using an unsuitable map projection)
  • Changing the threshold in map_proj_wrf.f90 (not recommended)

comment:2 Changed 7 years ago by adingwell

  • Resolution set to fixed
  • Status changed from assigned to closed

I'm marking this as solved due to inactivity.

comment:3 follow-up: Changed 5 years ago by ahilboll

We're running in the same problem here with WRF output on 40m resolution. The MAPFAC_ values are all below 1.0002, using Lambert projection. So in principle, the domain should not affected by scaling.

So I don't understand how come that the RMS is above 2%. Do WRF and FLEXPART-WRF use different formulas to calculate lat/lon from the map projection parameters?

And, for now, more importantly: Can we safely use the WRF output in this situation?

comment:4 in reply to: ↑ 3 ; follow-up: Changed 5 years ago by pesei

  • Resolution fixed deleted
  • Status changed from closed to reopened

Replying to ahilboll:

So I don't understand how come that the RMS is above 2%. Do WRF and FLEXPART-WRF use different formulas to calculate lat/lon from the map projection parameters?


They should not but maybe there have been changes in newer versions of WRF? Might be worthwhile to compare.

comment:5 in reply to: ↑ 4 Changed 5 years ago by ahilboll

Replying to pesei:

Replying to ahilboll:

So I don't understand how come that the RMS is above 2%. Do WRF and FLEXPART-WRF use different formulas to calculate lat/lon from the map projection parameters?


They should not but maybe there have been changes in newer versions of WRF? Might be worthwhile to compare.

I just compared map_proj_wrf_subaa.f90 from FLEXPART-WRF v3.3.2 to module_llxy.F from WRF v4.1 and the respective ijll_lc subroutines seem to be equivalent.

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