Xarray - Examples

Contents

Xarray - Examples#

import xarray as xr
import numpy as np

mvbs.Sv.plot(x="ping_time", y="depth", #define axes
             col="channel", #define face or columns
             col_wrap=2,  #how many columns per row
             cmap='RdYlBu_r', #colourmap
             yincrease=False, # increasing or decreasing y axis
             vmin=-85, #colorbar min
             vmax=-45 #colorbar max
            )

<xarray.plot.facetgrid.FacetGrid at 0x1f4fce06ea0>

_images/c256a526264cfe9fd078db3abc6b33e3b70edbc5760fa73871697595aa978913.png

Interactive plotting of ´xarray´ datasets is easiest in ´hvplot´

import hvplot.xarray 

mvbs.Sv.hvplot(
    groupby="channel",
    cmap="RdYlBu_r",
    x='ping_time',
    y='depth',
    clim=(-85,-45)).opts(invert_yaxis=True)

Simple Processing#

First we add a linear ´Sv_lin´ to our dataset, to allow for simple operations in linear space:

Median#

We can apply a rolling median with a specific window size along time/distance and depth/range:

rm = 2 #range / depth window size
tm = 10 #distance window size 

#we use Sv_lin and want to compute a rolling function, here we select median
ds_median = mvbs.Sv_lin.rolling(depth=rm, #we wanth the depth dim to be summarised by rm
                              distance = tm, #we want distance to be summarised y tm
                              center=True).median(skipna=True) #we want the median to be computer

# We do a back transformation into log space
ds_median = 10 * np.log10(ds_median) 

# and we do a nice interactive plot
ds_median.hvplot(groupby="channel", #grouping element
               cmap="RdYlBu_r", #colormap
               x='distance', #x axis
               y='depth',#y axis
               clim=(-75,-32)).opts(invert_yaxis=True) #colormap limits and inverted y axis

Gaussian Blur#

Another commonly used alorithm is gaussian blur.
We import the ´gaussian_filter´ from the ´scipy.ndimage´ package and apply it:

from scipy.ndimage import gaussian_filter

#Gaussian Blur
sigma=3 #sigma parameter
blur = xr.DataArray(
    gaussian_filter(mvbs.Sv_lin, sigma=sigma), #apply gaussian blur to Sv_lin
    coords=[mvbs.channel.values, mvbs.distance.values, mvbs.depth.values],
    dims=["channel","distance","depth"])

#backtransformation into log space
blur = 10 * np.log10(blur)

# and we do a nice interactive plot
blur.hvplot(groupby="channel", #grouping element
               cmap="RdYlBu_r", #colormap
               x='distance', #x axis
               y='depth',#y axis
               clim=(-75,-32)).opts(invert_yaxis=True) #colormap limits and inverted y axis