Interactive Radar Visualization

Overview

Within this cookbook, we will detail how to create interactive plots of radar data!

Reading data with Xradar
Creating your first interactive figure with Xradar + hvPlot
Combining your plots into a single dashboard
Filtering and Checking Data Quality
Create a Dashboard to Analyze ZDR Bias

Imports

import xradar as xd
import fsspec
import pyart
from open_radar_data import DATASETS
import hvplot.xarray
import holoviews as hv
import panel as pn

hv.extension("bokeh")

## You are using the Python ARM Radar Toolkit (Py-ART), an open source
## library for working with weather radar data. Py-ART is partly
## supported by the U.S. Department of Energy as part of the Atmospheric
## Radiation Measurement (ARM) Climate Research Facility, an Office of
## Science user facility.
##
## If you use this software to prepare a publication, please cite:
##
##     JJ Helmus and SM Collis, JORS 2016, doi: 10.5334/jors.119

Prerequisites

It is recommended that you are familiar with working with weather radar data, the core data structures, and the basics of reading in different radar datasets.

Concepts	Importance	Notes
Intro to Cartopy	Necessary
Xradar User Guide: Plot a PPI	Necessary
Understanding of NetCDF	Helpful	Familiarity with metadata structure

Time to learn: 30 minutes

Creating Your First Interactive Figure with Xradar and hvPlot

hvPlot is helpful tool when working with interactive visualizions! It is a tool built on top of several other packages, that we can use with Xarray.

By default, this visualization plots azimuth along the y-axis and range along the y-axis. While this is desired in certain cases, we cannot gather much spatial information from this.

ref = radar["sweep_0"].DBZH.hvplot.quadmesh(cmap='pyart_ChaseSpectral',
                                            title='Horizontal Reflectivity (dBZ)',
                                            clim=(-20,60))
ref

Refining Our Plot - Recreating a Plan Position Indicator (PPI)

We instead would like to create a Plan Position Indicator (PPI) plot. Since we already georeferenced the dataset, we set x/y to be x and y, or the distance away from the radar, as well as tuning some additional parameters. We set rasterize=True to lazily load in the data, which renders the plot more quickly and increases resolution as we zoom in.

ref = radar["sweep_0"].DBZH.hvplot.quadmesh(x='x',
                                           y='y',
                                           cmap='pyart_ChaseSpectral',
                                           clabel='Horizontal Reflectivity (dBZ)',
                                           title=f'Horizontal Reflectivity \n {radar.attrs["instrument_name"]} Radar',
                                           clim=(-20, 60),
                                           height=500,
                                             rasterize=True,
                                           width=600,)

ref

zdr = radar["sweep_0"].ZDR.hvplot.quadmesh(x='x',
                                             y='y',
                                             cmap='pyart_ChaseSpectral',
                                             clabel='Differential Reflectivity (dB)',
                                             title=f'Horizontal Reflectivity \n {radar.attrs["instrument_name"]} Radar',
                                             clim=(-1, 6),
                                             height=500,
                                             rasterize=True,
                                             width=600,)
zdr

Combining your plots into a single dashboard

You can combine plots using the + syntax to add plots side-by-side, or * to add them to the same plot. For example, let’s combine our reflectivity and velocity plot.

(ref + zdr).cols(1)

Create a Dashboard to Analyze ZDR Bias

A common data quality check is differential reflectivity bias. This value should be around 0 for low values of horizontal reflectivity. We use a few steps here to create this visualization

Unstack the dataset so we are left with a single dimension - the single range gate (single points)
Create histograms (.hist) and a 2-dimensional histogram (.hexbin) to visualize the data
Stack these into single view using gridspec

ds = ds.stack({"gate": {"azimuth", "range"}}).reset_index("gate")

hist_dbz = ds.hvplot.hist("DBZH",
                          width=500,
                          height=200,
                          title="Horizontal Reflectivity Distribution",)
hist_zdr = ds.hvplot.hist("ZDR",
                          height=400,
                          title="Differential Reflectivity Distribution",
                         ).opts(invert_axes=True)
hexbin = ds.hvplot.hexbin(x='DBZH',
                          y='ZDR',
                          title='Reflectivity vs. Differential Reflectivity Distribution',
                          width=500,
                          height=400) *  hv.HLine(0,
                                                  label='Differential Reflectivity = 0 Line').opts(color='red',
                                                           line_width=1)

gspec = pn.GridSpec(width=800, height=400)

gspec[0,   0:2  ] = hist_dbz
gspec[1:3,   0:2  ] = hexbin
gspec[1:3,   2  ] = hist_zdr

gspec

Summary

Within this notebook, we covered how to use interactive visualizations with your weather radar data, including applications to checking data quality.

What’s Next?

Next, we will continue to explore methods of cleaning and visualizing data!

Interactive Radar Visualization

Overview

Imports

Prerequisites

Reading Data with Xradar

Read the data in xradar

Creating Your First Interactive Figure with Xradar and hvPlot

Refining Our Plot - Recreating a Plan Position Indicator (PPI)

Combining your plots into a single dashboard

Filtering and Checking Data Quality

Determine Mask Thresholds

Double Check our Filtered Data

Create a Dashboard to Analyze ZDR Bias

Summary

What’s Next?

Resources and References