I try to convert Lambert conformal coordinates to lat/lon (WGS84) and I have used wgrib2, but the result is biased.
Command:
wgrib2 "mypath" -match "10m...." -new_grid_winds grid -new_grid_interpolation neighbor -new_grid latlon 108:129:0.25 16:65:0.25 "outputpath"
results with:
while it should be like that (from windy.com)
grib file:
I think there might be some flaws in the initial grib file. I converted the grib file to netCDF using wgrib2 and after that made some plots using Python and the result is not good.
Thing is, when I make the plot of temperature and overlay that with wind vectors, it looks ok. Problem is, when I also add the coastline, I see that the location of Taiwan island and also the main continent does not match with coastline drawn from the database.
Therefore I assume there is something bad in the initial gribfile - either the coordinates (start and endpoint or the step) are not very good and the coordinates written to the netCDF are not correct.
My code is here, if interested:
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.basemap import Basemap
from netCDF4 import Dataset
import json
# -------------------------------
# read the json file:
with open('2018091312.json','r') as f:
data = json.load(f)
# -------------------------------
lo1,lo2,la1,la2 = 108,142.75,16,23.75
dx,dy=0.25,0.25
nx,ny=140,32
udata=np.array(data[0]['data'],dtype='float32');udata=np.reshape(udata,(ny,nx));
vdata=np.array(data[1]['data'],dtype='float32');vdata=np.reshape(vdata,(ny,nx));
londata=np.arange(lo1,lo2+dx,dx);
latdata=np.arange(la1,la2+dy,dy);
londata,latdata=np.meshgrid(londata,latdata)
# -------------------------------
# -------------------------------
ncin=Dataset('test.nc');
lons=ncin.variables['longitude'][:];
lats=ncin.variables['latitude'][:];
u10=np.squeeze(ncin.variables['UGRD_10maboveground'][:])
v10=np.squeeze(ncin.variables['VGRD_10maboveground'][:])
t2=np.squeeze(ncin.variables['TMP_surface'][:])
ncin.close();
# -------------------------------
xlim=(np.min(lons),np.max(lons));
ylim=(np.min(lats),np.max(lats));
# -------------------------------
plt.figure(figsize=(8, 8))
m = Basemap(projection='cyl', resolution='i',
llcrnrlat=ylim[0], urcrnrlat=ylim[1],
llcrnrlon=xlim[0], urcrnrlon=xlim[1], )
xx,yy=m(lons,lats);
m.pcolormesh(lons,lats,t2,vmin=273.,vmax=300.);
skipx=skipy=16
m.quiver(xx[::skipy,::skipx],yy[::skipy,::skipx],u10[::skipy,::skipx],v10[::skipy,::skipx],scale=20.0,units='inches');
# ------------------------------------------
plt.savefig('test_withoutland.png',bbox_inches='tight')
m.drawcoastlines()
m.drawlsmask(land_color = "#ddaa66")
plt.savefig('test_withland.png',bbox_inches='tight')
plt.show()
# ------------------------------------------
skipx,skipy=2,2
plt.figure(figsize=(8, 8))
m = Basemap(projection='cyl', resolution='i',
llcrnrlat=ylim[0], urcrnrlat=ylim[1],
llcrnrlon=xlim[0], urcrnrlon=xlim[1], )
xx,yy=m(londata,latdata);
m.pcolormesh(lons,lats,t2,vmin=273.,vmax=300.);
m.quiver(xx[::skipy,::skipx],yy[::skipy,::skipx],udata[::skipy,::skipx],vdata[::skipy,::skipx],scale=20.0,units='inches');
# ------------------------------------------
m.drawcoastlines()
m.drawlsmask(land_color = "#ddaa66")
plt.savefig('test_json.png',bbox_inches='tight')
plt.show()
And the result looks like this (the test with JSON file):
The conversion from grib to newCDF, I did like this:
wgrib2 M-A0064-000.grb2 -netcdf test.nc
There are some weird definitions in the WRF LCC that you need to keep in mind when doing your reprojections. This website (unaffiliated) details most of it using python.
