Wind vector rotation troubles (rotate_vector)

Start by reproducing the provided Basemap example and compare m.rotate_vector results for the single point and the sliced gridded arrays. Investigate how the routine handles the non-standard, two-dimensional latitude grid and irregular observation coordinates; done means the rotation behavior is consistent or the supported coordinate constraints are clearly established.

I am currently working with a gridded data set that has its wind vectors already in an earth-relative frame. When looking to rotate the wind vectors appropriately for plotting on a Basemap projection I've run into a strange problem where the vectors, quite simply, look rather odd. I ran a few tests with simple test points to make sure the rotate_vector routine appeared to be working correctly, and it seemed fine. But when I run a whole model output grid through the rotate_vector routine, it seems to produce wind fields which don't look quite right (see below).

Here are some stats associated with the point that has the red star on the plot with the title 'Rotated vectors'

Lat and Lon: 39.11, -70.0144

Original U and V:     1.30,  7.14
Rotated U and V:    -1.33 , 7.13

Now note how the rotated wind at this point is aligned along the -70 meridian, which would suggest that in earth relative terms we might expect the wind components to be more along the lines of u=0 and v=7. However, judging from above that's certainly not the case.

Now, if I rotate that same vector, by itself, I get the following rotated wind:

Rotated U and V: -0.03 7.26

And see the second plot for a visual:

It would seem desirable to have the same behavior for the single point that we have for the gridded set of winds.

The only thing special about the gridded data, that I have noticed, is that the latitudes can vary in a non-standard way with increasing x dimension, west to east (e.g. 35.5N, 34.2N, 33.1N, 36.2N, etc.). It wasn't clear if this was okay within rotate_vector. In some simple tests it didn't seem to be a problem.

It would be nice to be able to use this routine for not only a set of gridded data, but observations as well. It's worth noting that observations wouldn't necessarily be ordered in a nice, regular way given the nature that observations tend to be irregularly spaced (e.g. surface stations co-located with airports).

Unfortunately I don't have any suggestions for a solution, but am hopeful others may have an idea.

Below is the snippet of code used to generate the first plot above.

Thanks!
Jacob

import nemsio
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.basemap import Basemap

# Get the input file
f='restart_file'
nio=nemsio.nemsfile(f)

skip=25
u10=u[::skip,::skip,0]
v10=v[::skip,::skip,0]
lats=nio.lats[::skip,::skip]
lons=nio.lons[::skip,::skip]

# Create the figure
fig=plt.figure(figsize=(18, 6)) 

# Domain covers NE CONUS
llcrnrlon=-84.0
llcrnrlat=35.0
urcrnrlon=-60.0
urcrnrlat=49.0
res='l'

m = Basemap(llcrnrlon=llcrnrlon,llcrnrlat=llcrnrlat,urcrnrlon=urcrnrlon,urcrnrlat=urcrnrlat,\
          rsphere=(6378137.00,6356752.3142),\
          resolution=res,projection='lcc',\
          lat_1=25.0,lon_0=-95.0)

#The vector rotation to the Basemap projection just specified
u10_rot, v10_rot, x, y = m.rotate_vector(u10, v10, lons, lats, returnxy=True)

parallels = np.arange(-80.,90,5.)
meridians = np.arange(0.,360.,5.)

# - First sublot is without rotation
ax = fig.add_subplot(121)
ax.set_title('Without rotation')

m.drawmapboundary(fill_color='aqua')
m.fillcontinents(color='#cc9955', lake_color='aqua', zorder = 0)
m.drawcoastlines(color = '0.15')

m.drawparallels(parallels)
m.drawmeridians(meridians)

m.barbs(x, y, u10, v10, pivot='middle', barbcolor='black',zorder=10)

# - Second subplot is with rotation

ax = fig.add_subplot(122)
ax.set_title('Rotated vectors')

m.drawmapboundary(fill_color='aqua')
m.fillcontinents(color='#cc9955', lake_color='aqua', zorder = 0)
m.drawcoastlines(color = '0.15')
m.drawparallels(parallels)
m.drawmeridians(meridians)
m.barbs(x, y, u10_rot, v10_rot, 
    pivot='middle', barbcolor='black',zorder=10)

m.scatter(-70.0144,39.11,s=175,color='red',marker='*',latlon=True)