Identifying Precipitation types using RADOLAN data (RW Product)

Hello everyone,

I currently working on RADOLAN binary data. I am trying to classify my data according to different Precipitation types available in the documentation Precipitation Types
Please guide me here if it is possible with my data available in the variable “clip_shape” be used to classify different precipitation types. My data variable is “RW”. You can see the file and data format below in the code. Thanks in advance.

import numpy as np
import pandas as pd
import scipy
import wradlib as wrl
import xarray as xr
import rioxarray 
import geopandas as gpd
import cartopy
import as ccrs
import shapely
from shapely.geometry import mapping
import matplotlib.pyplot as plt
import os
import sys
import tarfile
import fiona
import datetime
dict(xarray=xr.__version__, rioxarray=rioxarray.__version__, geopandas=gpd.__version__, cartopy=cartopy.__version__, 
     shapely=shapely.__version__, wradlib=wrl.__version__)
{'xarray': '2023.2.0',
 'rioxarray': '0.13.4',
 'geopandas': '0.12.2',
 'cartopy': '0.21.1',
 'shapely': '2.0.1',
 'wradlib': '1.19.1'}
# binary data files downloaded from

#Step 1:--------- load radolan raster data
fname = "Extracted Data/raa01-rw_10000-*-dwd---bin.gz"
rad = xr.open_mfdataset(fname, engine="radolan")
rad.RW.encoding["_FillValue"] = 65536             # fix encoding _FillValue

#Step 2:------------setup projection
proj_radolan = ccrs.Stereographic(
    true_scale_latitude=60.0, central_latitude=90.0, central_longitude=10.0
)"x", y_dim="y", inplace=True), inplace=True)
#Step 3:-----------Load German federal states shapefile
germany = gpd.read_file(filepath + "VG2500_LAN.shp") = "EPSG:31467"
#Step 4:-----------Extract Brandenburg
brandenburg = germany.loc[[11], "geometry"]
#Step 5:-----------Clip using rioxarray clip_box
bounds = brandenburg.to_crs(proj_radolan).bounds.iloc[0]

clip_box =*bounds)

fig = plt.figure(figsize=(12, 6))
ax = fig.add_subplot(projection=ccrs.PlateCarree())
clip_box.RW[0].plot(ax=ax, transform=proj_radolan)
ax.gridlines(draw_labels=True, x_inline=False, y_inline=False)
#Step 6:-----------clip using rioxarray with shape
clip_shape =,, drop=True)
Dimensions:      (time: 8756, y: 254, x: 256)
  * time         (time) datetime64[ns] 2006-01-01T00:45:00 ... 2006-12-31T23:...
  * y            (y) float64 -4.171e+06 -4.17e+06 ... -3.919e+06 -3.918e+06
  * x            (x) float64 8.904e+04 9.004e+04 9.104e+04 ... 3.43e+05 3.44e+05
    spatial_ref  int32 0
Data variables:
    RW           (time, y, x) float32 dask.array<chunksize=(1, 254, 256), meta=np.ndarray>
    radarid:         10000
    formatversion:   2
    radolanversion:  01.01.00
    radarlocations:  ['bln', 'drs', 'eis', 'emd', 'ess', 'fbg', 'fld', 'fra',...

@SandeepAllampalli Your data RW are hourly rain sums. It’s not possible to derive the precipitation type from that.

For hydrometeorclassification you would need at least some of the radar moments (eg. horizontal reflectivity ZH, differential reflectivity ZDR, specific differential phase KDP, crosscorrelation coefficient RHOHV).

Update: DWD is providing RE product free of charge which contains the percentage of solid precip and a hail-flag. wradlib is able to digest RE (RADOLAN Product Showcase — wradlib) but not with all features, so care has to be taken and possible adaptions to wradlib are needed. Not sure this will help with your research but this is all I can recommend right now.

If others have some more ideas, you are welcome to add to this discussion.

Thank you for the reply. I am curious to know about the kind of analysis/ inferences could be extracted from the RW product.