Plotting#

pyorc offers a number of convenient plotting functions that allow you to summarize the output data into figures. The plotting functionality is built such, that it can combine a view of a frame, with a view on your surface velocity estimates, and transect velocities into one figure that displays all the steps of your processing in one single view. The views can be made either in a camera perspective, providing an augmented reality view, a local perspective, meaning the perspective after orthorectification, or a geographical perspectives, plotting on a geographical map if the coordinate reference system is known. There may be use cases for all three perspectives. A full example of how to establish a plot from our example dataset is shown below, with more information below that.

Command-line

To generate a plot through a recipe, you may include a plot section. In this section you can configure as many plots as you like in subsections. In the example, we only make one, called plot_quiver. Under the plot_quiver examples, we specify 3 subsections frames, velocimetry and transect to plot the respective components. Besides these subsections, you can specify a few other things:

mode: select in what projection mode the plot should be made. This can be camera, local or geographical as explained above. Try it out to see what this gives you.
reducer: because plotting is only 2-dimensional, somehow you have to reduce the velocimetry results over the time dimension so that you only have the spatial dimensions left. This is done with a so-called reducer operation. The available ones are mean and median.
write_pars: here you pass several parameters that may be used whilst writing to file. Please note that the available parameters are described in the matplotlib documentation, more specifically this page. Only the parameter fname cannot be supplied as this is the filename for the output. This filename will be constructed by pyorc and has the naming convention <figure-name>.jpg.

...
...

plot:
  plot_quiver:
    frames:
    velocimetry:
      alpha: 0.4
      cmap: rainbow
      vmax: 0.6
    transect:
      transect_1:
        cmap: rainbow
        add_colorbar: True
        # width: 2  # uncommenting woul make the arrows 2x thicker than default
        # scale: 0.5  # uncommenting would make the arrows twice as long as default
        vmin: 0
        vmax: 0.6
    mode: camera
    reducer: mean
    write_pars:
      dpi: 100
      bbox_inches: tight

API

To generate a plot with all three components, you therefore need access to at minimum one frame from a pyorc.Frames object, a pyorc.Velocimetry object that is somehow reduced over time (e.g. a mean or median) and a pyorc.Transect object that has an effective velocity, perpendicular on the stream direction. Let’s do a full example, starting with an already processed video. Below we re-open the video, also open the velocimetry results, retrieve a transect from it, and plot everything together in one figure. The specifics are explained after that.

import xarray as xr
import pandas as pd
import pyorc
from matplotlib.colors import Normalize
ds = xr.open_dataset("examples/ngwerere/ngwerere_masked.nc")

# also open the original video file, only one frame needed
video_file = "examples/ngwerere/ngwerere_20191103.mp4"

video = pyorc.Video(video_file, start_frame=0, end_frame=1)

# borrow the camera config from the velocimetry results
video.camera_config = ds.velocimetry.camera_config

# get the frame as rgb
da_rgb = video.get_frames(method="rgb")

# read a cross section
cross_section = pd.read_csv("examples/ngwerere/ngwerere_cross_section.csv")
x = cross_section["x"]
y = cross_section["y"]
z = cross_section["z"]

# get data over the cross section
ds_points = ds.velocimetry.get_transect(x, y, z, crs=32735, rolling=4)

# get vertically integrated velocities
ds_points_q = ds_points.transect.get_q(fill_method="log_interp")

# plot the rgb frame first. We use the "camera" mode to plot the camera perspective.
norm = Normalize(vmin=0., vmax=0.6, clip=False)

# plot the first frame and return the mappable
p = da_rgb[0].frames.plot(mode="camera")

# extract mean velocity and plot in camera projection
ds.mean(dim="time", keep_attrs=True).velocimetry.plot(
    ax=p.axes,  # use the axes already created with the first mappable
    mode="camera",  # show camera perspective
    cmap="rainbow",  # choose a colormap
    alpha=0.3,  # transparency (smaller is more transparent)
    norm=norm,  # color scale
)

# plot velocimetry point results in camera projection
ds_points_q.isel(quantile=2).transect.plot(
    ax=p.axes,  # refer to the axes already created
    mode="camera",  # show camera perspective
    cmap="rainbow",  # choose a colormap
    norm=norm,  # color scale
    add_colorbar=True  # as final touch, add a colorbar
)

The resulting figure is shown below. Let’s dive a little more into the details of the three different components.

Frames#

Command-line

To plot a frame you only have to specify frames: without any further argument. Automatically, the first frame in the video will then be added to the figure.

API

The steps to get a frame plotted on your figure canvas, are the following:

get the frames from the video with get_frames.
If you want to plot in local or geographical mode, then call frames.project on your Frames object and move on with the result of this.
select a (one!) frame
call frames.plot on the result with your desired projection mode defined in the parameter mode. By default mode="local". As you may note

Note that in the example, we do the frame selection (with da_rgb[0]) and plotting in one line of code as follows:

p = da_rgb[0].frames.plot(mode="camera")

Velocimetry#

Getting your 2-dimensional velocimetry results in the figure can be done in several ways with for each method also other parameters that may be supplied. Below we describe briefly what the options are and we refer to the respective web pages where the input parameters may be found (x and y being supplied by pyorc so you should not supply these). For all methods, you can supply the parameters vmin and vmax, which indicate the minimum and maximum value you wish to show on a color scale. Furthermore, you may add an option add_colorbar: True to establish a colorbar in the lower left corner. This colorbar will be bounded by the values supplied with vmin and vmax. If you do not supply vmin and vmax, then a standard colorbar with limits of 0 and 3 with pretty breaks is used. Note that any parameter added is not mandatory! If you leave the parameters out, then default values will be used instead.

quiver (default): “quivers” are arrows pointing into the vector direction, with the length and (possibly) color representing the scalar value of the velocity. See this page. Especially the scale parameter is a little counter- intuitive. A larger value results in smaller arrows!!
pcolormesh: this is a gridded colored mesh that defines the scalar values only. See this page.
scatter: similar to pcolormesh, but then the velocities are represented by dots instead of grid cells. See this page.
streamplot: this only works on local mode and shows lines how particles would move over the surface. See this page.

Specific default values for quiver plots include the width and scale parameters. These are set such that they always look pretty on all plot modes. A width of 1 results in the default width. Smaller (larger) than 1 in thinner (thicker) quivers. A scale of 0.5 (2.0) means the quiver length will be twice as large (small) as default.

Command-line

Within the recipe, under the selected plot (in the example plot_quiver) and the velocimetry subsection you can define the plot method, and below that define the parameters. Please read the referred pages to see what options there are.

API

The plot methods are available under a subclass pyorc.Velocimetry.plot. Before applying them, you must first reduce any time variable results over time, for instance with:

# assume that processed and masked results are in piv
piv_mean = piv.mean(dim="time", keep_attrs=True)

Note

The keep_attrs=True flag is quite important here as we may need the attributes to reproject data from the default x, y projection to the original camera perspective for instance.

After reducing, we can use a set of methods to make plots, in a very similar manner as used for plotting frames. As shown in the example, you can smartly combine plots from a frame, with plots of your velocimetry results and by doing so create beautiful augmented reality views, or geospatial views. The different plotting method all use an underlying matplotlib.pyplot function and are also named accordingly. Hence they can receive keyword arguments specific to these underlying functions. In addition, the additional important keywords described above can be supplied as follows:

add_colorbar=True: this flag will add a default colorbar to the axes
mode: this flag can be set to local (default), geographical (for plots in a geospatial view) and camera for augmented reality in the camera’s original perspective.

Each plotting method always returns the mappable so that you can make your own colorbars, and refer to its parent axes and in turn the axes parent figure.

The different plotting methods are summarised below.

Plot function	Description
`pyorc.Velocimetry.plot.pcolormesh`	computes scalar velocities and plots these as a gridded mesh
`pyorc.Velocimtry.plot.scatter`	computes scalar velocities and plots as colored dots
`pyorc.Velocimetry.plot.streamplot`	draws a streamplot through the x and y-directional velocities. Only works with `mode="local"`
`pyorc.Velocimetry.plot.quiver`	draws a quiver plot using the x and y-directional velocities. If `pyorc.Velocimetry.plot` is called directly, this method is used

Transect#

Plotting a transect is very similar to plotting of 2-dimensional velocimetry. Only the methods quiver and scatter are available, as streamplots and meshes do not apply to 1-dimensional datasets.

Another nice example of a augmented reality view over the Geul river in The Netherlands is shown below.