Introductory exercises with bioRad

Adriaan M. Dokter

These course materials were developed for the 2nd Radar Aeroecology Training School, Aug 14-17 2018, University of Oklahoma, Norman, OK, USA.

1 Basic visualisation of radar scans in R

1.1 Preparation

Radar data needed for this practical can be downloaded from Dropbox. https://www.dropbox.com/s/qkm3ll8l5oe2y54/RadAero18.zip?dl=0

Execute each of the code examples provided below in RStudio, and try to complete the exercises.

All the functions of the bioRad package have fairly elaborated manual pages, that you can refer to at any moment:

In the help window, you can click on each of the links to pull up help pages of specific functions. Click on the Index link at the bottom of each help page to get an overview of all functions and available manual pages.

Start by making a new directory on your local machine that you will use for this practical:

Your R session is now properly set up.

To work with US NEXRAD data, bioRad currently requires a working installation of Docker (Linux/Mac) or Docker for Windows (not Docker Toolbox, this is an older implementation of Docker for Windows operating systems that is not supported). If you managed to successfully install Docker, test whether it works:

If you get a “Hello from Docker!” welcome message, everything is working.

If you did not manage to install Docker, you will not be able to load NEXRAD data into R at this time, but you will be able to continue the exercises with radar data that can be loaded without Docker.

1.2 The structure of polar volumes

Exercise 1: Load the polar volume file, example_pvol.h5 from the zipfile downloaded earlier. Put it in your working directory and load it into R. What is the minimum and maximum scan elevation contained in the volume? And which scan parameters are available? (Check the help file of the read_pvolfile() function for the nomenclature of various available quantities).

1.3 Plotting radar scans

Exercise 2: This case contains convective precipitation, characterized by localized but intense thunderstorms, as well as biological scattering. Make also a ppi plot of the radial velocity and correlation coefficient. Verify how the texture of the radial velocity, and the values of correlation coefficient of the precipitative areas differs from the areas with biological scattering.

Exercise 3: Based on the radial velocity image, are the biological scatterers birds or insects? Why?

1.4 Overlaying radar scans on maps

Exercise 4: Plot the same ppi but now using a cartographic map instead of satellite maps as base layer. See help of bioRad’s map() function for available types of basemaps, specifically the argument maptype.

2 Analysis and visualisation of vertical bird profiles

In this section you will learn to compute, interpret and analyze vertical profiles (vp). A vp consists of the (bird) density, speed and directions at different altitudes at the location of a single radar. It is typically calculated for all data within a cylinder with a 35 km radius around the radar, i.e. only containing data at relatively short distances, where the radar beam is still sufficiently narrow to resolve altitude information.

See section 3 for information on how to process polar volume data into vertical profiles. To save some time, we will start below with a list of pre-processed vertical profiles for the Brownsville radar in Texas (KBRO).

2.1 Loading processed vertical profiles

2.2 Inspecting single vertical profiles

Now that you have loaded a list of vertical profiles, we can start exploring them. We will start with plotting and inspecting single vertical profiles, i.e. a single profile from the list of vp objects you have just loaded.

These two plots look very different, but they are twice the same data plotted on a different scale.

eta = (radar-wavelength dependent constant) * 10^(dbz/10)

So eta and dbz are closely related, the main difference is that reflectivity factors are logarithmic, and reflectivities linear.

The reflectivity factor dbz is the quantity used by most meteorologist. It has the useful property that at different radar wavelengths (e.g. S-band versus C-band) the same amount of precipitation shows up at equal reflectivity factors. The same holds for insects, as well as any other target that is much smaller than the radar wavelength (S-band = 10 cm, C-band = 5 cm), the so-called Rayleigh-scattering limit.

In the case of birds we are outside the Rayleigh limit, because birds are of similar size as the radar wavelength. In this case the reflectivity eta is more similar between S-band and C-band. eta is also more directly related to the density of birds, since eta can be expressed as (bird density) x (radar cross section per bird). For these two reasons, for weather radar ornithologists reflectivity eta is the more conventional unit (instead of reflectivity factor in dBZ).

Exercise 5: If you change your assumption on the bird’s radar cross section in the previous example, and assume it is 10 times as large, what will be the effect on the bird density profile?

The assumed radar cross section can be changed as follows:

Exercise 6: Verify your answers on the previous two questions, by re-plotting the vertical profiles for the bird density quantity.

2.3 Plotting vertical profile time series

We will now examine multiple vertical profiles at once that are ordered into a time series, e.g. the vertical profiles obtained from a single radar over a full day.

Let’s make a plot for a subselection of the time series:

Exercise 7: Interpret the wind barbs in the profile time series figure: what is the approximate speed and direction at 1400 meter at 22 UTC. In the speed barbs, each half flag represents 2.5 m/s, each full flag 5 m/s, each pennant (triangle) 25 m/s.

Exercise 8: Extract the vertical profile at 6 UTC from the time series and plot the vertical profile of ground speeds (quantity ff, see manual for the vp class for full list of available quantities). Check whether your answer to the previous question was approximately correct.

2.4 Vertical and time integration of profiles

Often you will want to sum together all the migrants in the vertical dimension, for example if you want a single index of how many birds are migrating at a certain instant. There are at least two ways in which you can do that:

We will be using bioRad’s integrate_profile() function to calculate these quantities:

The following questions only require pen and paper. Assume a night migration event in which the volume density of birds from 0-1000 meter above ground is 200 birds per cubic kilometer, and from 1000-1500 meter 100 birds per cubic kilometer. In the lower layer birds fly at 50 km/hour, and in the upper layer at 100 km/hour. Above 1500 meter there are no birds. Migration continues for exactly three hours after sunset, and then halts abruptly.

Exercise 9: What is in this case the bird’s vertically integrated density (VID)? Give your answer in units birds/km\(^2\).

Exercise 10: What is in this case the migration traffic rate across a transect perpendicular to the direction of movement? Give your answer in units birds/km/hour.

Exercise 11: How many birds have passed a 1km transect perpendicular to the direction of movement in this night? Give your answer in terms of migration traffic (mt) in units birds/km.

Both MTR, VID and MT depend on the assumed radar cross section (RCS) per bird. If you are unwilling to make any assumptions on RCS, you can alternatively use two closely related quantities that do not depend on RCS:

VIR gives you the total cross-sectional area of air-borne targets per square kilometer of ground surface, whereas RTR gives you the total cross-sectional area of targets flying across a one kilometer line perpendicular to the migratory flow per hour.

Exercise 12: Re-plot the vertically integrated quantities discussed so far using a different radar cross section, and verify which quantities are affected by a change in radar cross section.

2.5 Inspecting precipitation signals in profiles

Precipitation is known to have a major influence on the timing and intensity of migration, therefore it is a useful skill to be able to inspect profiles for presence of precipitation.

Also, although automated bird quantification algorithms become more and more reliable, distinguishing precipitation from birds remains challenging for algorithms in specific cases. It is therefore important to have the skills to inspect suspicious profiles. That may help you to identify potential errors of the automated methods, and prevent your from overinterpreting the data.

An easy way of doing that is plotting the vertical profile of total reflectivity (quantity DBZH), which includes everything: birds, insects and precipitation. Precipitation often has higher reflectivities than birds, and also extends to much higher altitudes.

Exercise 13: Compare your plots for bird density (quantity dens) with a profile plot for total reflectivity (quantity DBZH, showing birds and precipitation combined). Compare the two plots to visually identify periods and altitude layers with precipitation.

3 Processing polar volume data into profiles

3.1 Obtaining radar data

The names of the radars in the networks can be found here:

Useful sites for inspecting pre-made movies of the US composite are http://birdcast.info/live-migration-maps/ and http://www.pauljhurtado.com/US_Composite_Radar/.

3.2 Processing a single polar volume with the vol2bird algorithm

The following steps take you through the process applying the vol2bird algorithm yourself. You need a working installation of Docker (Linux/Mac) or Docker for Windows (not Docker Toolbox, this is an older implementation of Docker for Windows operating systems that is not supported).

We will generate vertical profiles with the automated algorithm vol2bird (https://github.com/adokter/vol2bird), which is included in the bioRad package.

If you get a “Hello from Docker!” welcome message, everything is working and you can start processing.

3.3 Processing multiple polar volumes

This section contains an example for processing a directory of polar volumes into profiles:

Having generated the profiles, we can read them into R:

You can now continue with visualizing and post-processing as we did earlier: