Open jesusff opened 1 year ago
Projections are definitely different in the urban fraction on nextcloud and the orog and land fraction files on ESGF.
On ESGF (orog_EAS-22_NCC-NorESM1-M_historical_r1i1p1_ICTP-RegCM4-4_v0_fx.nc
), it is Lambert conformal:
int crs ;
crs:longitude_of_central_meridian = 115. ;
crs:semi_major_axis = 6371229. ;
crs:inverse_flattening = 0. ;
crs:grid_mapping_name = "lambert_conformal_conic" ;
crs:standard_parallel = 20., 50. ;
crs:latitude_of_projection_origin = 35. ;
crs:false_easting = -12500. ;
crs:false_northing = -12500. ;
Nextcloud files (data/RegCM/urbanfraction/orig/EAS-22.nc
) are in rotated Mercator:
:projection = "ROTMER" ;
:grid_size_in_meters = 25000. ;
:latitude_of_projection_origin = 35. ;
:longitude_of_projection_origin = 116. ;
:grid_north_pole_latitude = 35. ;
:grid_north_pole_longitude = 116. ;
These must come from different simulations. Likely, the evaluation runs (not on ESGF) are not on the same computational grid as scenario runs.
Features are now well located, but clearly on different grids:
A fixed UF file for EAS-22 has been provided by @graziano-giuliani, thanks! The same problem arises for NAM-22 simulations, which were carried out by ISU but the UF provided from ICTP does not match the geographical projection. ISU simulations are on
char oblique_mercator ;
oblique_mercator:grid_mapping_name = "oblique_mercator" ;
oblique_mercator:azimuth_of_central_line = 90. ;
oblique_mercator:latitude_of_projection_origin = 46.5 ;
oblique_mercator:longitude_of_projection_origin = 263. ;
oblique_mercator:scale_factor_at_projection_origin = 1. ;
oblique_mercator:false_easting = 0. ;
oblique_mercator:false_northing = 0. ;
while our current UF file has:
// global attributes:
:experiment = "NACOR" ;
:projection = "ROTMER" ;
:grid_size_in_meters = 25000. ;
:latitude_of_projection_origin = 47.28 ;
:longitude_of_projection_origin = 263. ;
:grid_north_pole_latitude = 47.28 ;
:grid_north_pole_longitude = 263. ;
:grid_factor = 0. ;
This is currently being "fixed" by overwritting the geographical information from other variable (orog) in https://github.com/FPS-URB-RCC/CORDEX-CORE-WG/blob/13f1c3f5592359cec2a69ac68ea9dbb8ee7a0932/uhi/urban_areas.py#L155
However, a half-cell shift still exists and needs to be fixed in the original data.
There is a problem with the alignment of urban and land fraction. The hypothesis used to remove equal number of gridcells from the borders of the domains (see #7) seems wrong in this case. See last plot in this notebook. It is likely that there is a problem with the projection, since the same files look aligned over Tokyo.