Access MRMS datasets from siphon

[raysweather]

I am new to metpy but successfully using it to plot local radar data.

Can we access MRMS datasets through siphon in near real time?

Thanks

[dopplershift]

(Moved to newly created Siphon discussions)

Absolutely! Here's an example:

from datetime import datetime, timedelta
from siphon.catalog import TDSCatalog

cat = TDSCatalog('http://thredds.ucar.edu/thredds/catalog/grib/NCEP/MRMS/BaseRef/catalog.xml')
dataset = cat.datasets['Full Collection Dataset']

ncss = NCSS(dataset.access_urls['NetcdfSubset'])
query = ncss.query().variables('MergedBaseReflectivityQC_altitude_above_msl')
query.time(datetime.utcnow() - timedelta(hours=1))

mrms_ds = ncss.get_data(query)

[raysweather]

Thanks so much.

[raysweather]

We made lots of progress but cannot seem to get it to the finish line. I know this is a big ask, but...

Here's our code. It's getting data but the plot is a blank map with state borders, centered on NC.

What are we missing on getting latest reflexivity data to the map.
Thanks

from datetime import datetime, timedelta
from siphon.catalog import TDSCatalog
from siphon.radarserver import RadarServer
from siphon.ncss import NCSS
import cartopy
import cartopy.feature as cfeature
import cartopy.crs as ccrs
import matplotlib.pyplot as plt
from metpy.plots import USCOUNTIES
from metpy.plots import ctables

cat = TDSCatalog('http://thredds.ucar.edu/thredds/catalog/grib/NCEP/MRMS/BaseRef/catalog.xml')

dataset = cat.datasets['Full Collection Dataset']
ncss = NCSS(dataset.access_urls['NetcdfSubset'])

query = ncss.query()
query.lonlat_box(north=37, south=33, east=-75, west=-85)
now = datetime.utcnow()
query.time_range(now - timedelta(hours=2), now)
query.accept('netcdf')
query.variables('MergedBaseReflectivityQC_altitude_above_msl')
mrms_ds = ncss.get_data(query)

#variables defs
field_name = 'MergedBaseReflectivityQC_altitude_above_msl'
radar_lat = 35
radar_long = -80
t_step = 0
radar_data = mrms_ds.variables[field_name][t_step][0][:]
x = mrms_ds['lat']
y = mrms_ds['lon']

#making map
proj = cartopy.crs.LambertConformal(central_longitude=radar_long,
                                    central_latitude=radar_lat)
fig = plt.figure(figsize=(10,10))
ax = fig.add_subplot(1, 1, 1, projection=proj)
norm, cmap = ctables.registry.get_with_steps('NWSReflectivity', 100, -30)
mesh = ax.pcolormesh(y, x, radar_data, transform = proj, norm = norm, cmap=cmap, zorder=0)
ax.add_feature(cfeature.STATES.with_scale('50m'))
#ax.add_feature(USCOUNTIES.with_scale('20m'),edgecolor='gray',zorder=2,linewidth=1.2) #US counties
dist_in_degrees = 5
ax.set_extent([radar_long - dist_in_degrees,
               radar_long + dist_in_degrees,
               radar_lat - dist_in_degrees,
               radar_lat + dist_in_degrees])

[dopplershift]

You're calling pcolormesh with lat/lon data. That means you should be passing in ccrs.PlateCarree() for transform in that call since your x/y are lat/lon. Also, the colortable wasn't quite right. With the following two lines:

norm, cmap = ctables.registry.get_with_steps('NWSReflectivity', 5, 5)
mesh = ax.pcolormesh(y, x, radar_data, transform=ccrs.PlateCarree(), norm = norm, cmap=cmap, zorder=0)

Your original approach was telling it to interpret values like (-80, 35) as coordinates in meters within a Lambert Conformal coordinate system, which is definitely not right.

Also, if you're really only interested in a single radar, why not just use straight Level 2 data, and use pyart if you need gridding?

[raysweather]

Thanks yet again. I had never written one line of metpy program before 3 weeks ago, but I do know python. There's so much to get right but I'm getting there. So many things to get right and more difficult to inspect given the massive data sets. Every mistake gets me closer thanks to your guidance. I had never worked with a grib dataset, so I was fishing in those waters to find the problem.

The goal is an animation but I decided to get a single image before taking it to animation. Divide and conquer. I have our local radar animations; so, I should be off and running now... until I try the next new (to me) thing.
Thanks again

[dopplershift]

No worries, learning happens best from these mistakes. I've both written enough broken images, and debugged enough from others, that it's second nature upon seeing an empty image: double check the transform. 😉