v4.1:
  * add request_window_tolerance parameter (in seconds, default 0.5) to allow for missing data at the edges of request window
  * add --overwrite-conf to allow to set all available config options from the command line
  * allow to set parameters for removal of response ("full" option)
  * several small improvements in imaging module
v4.0:
  * complete refactoring of Qopen script to use subcommands for the different tasks
  * add source subcommand to calculate sorce spectra and derive source parameters including moment magnitude with the help of previous results
  * add command line arguments to dump opt and fit pkl files for later plotting
  * add command line arguments to turn all ploting on or off
  * add option to plot uniform color in sds plot, do not sort events by event id in all_sds plot
v3.0:
  * prefix option is more stringent now and adds prefix also to log file
  * add option to filter event catalog, add option to resolve full SEED id, add corresponding command line options
  * add option to use coda normalization, no inversion for site effects and event spectra in this case
  * add option to not perform optimization, but only invert for b ("optimize": null)
  * add 2 Greens functions for 3d diffusion approximation, one for exact diffusion and one usable for determining coda Q within Qopen
  * add debug option (--pdb)
  * use viridis_r cmap instead of hot_r cmap by default
  * type of cmap can now be set with the config
  * indicate number of observations with color in source displacement spectrum plots
  * allow more fine grained control of plot file names with str.format method
  * do not plot or calculate b and g0 for --fix-params
    this fixes a bug when using --fix-params with robust mean
  * fix bug when using newly introduced figsize and dpi options
  * fix NaN values in json output (are converted to None aka null)
  * fix warning due to conversion of masked values to nan in gmean
  * add label with excluded stations in sites.pdf
v2.6:
  * add dpi and figsize options for all plots (alternatively use a matplotlibrc file)
  * fix spelling error in invert_events_simultaneously (has to be fixed in old configuration files
    if these are used with new qopen version)
  * fix error when plotting site effects for simultaneously inverted events and present None values
  * fix wrong offset in optimization plots for simultaneously inverted events
v2.5:
  * add official support for python 3.7 and 3.8
v2.4:
  * Qopen can now be used on Windows, tests are passing,
    but multiprocessing is probably not yet working (see issue #3)
v2.3:
  * add Green's function for 2d and 1d radiative transfer
  * changed "G_module" config option to "G_plugin"
  * changed default tolerance for "optimize" config option
  * fix plotting error if events are inverted simultaneously
  * fix error when frequencies are given by fbands parameter
  * fix error when fitting source model to sds if gamma is not fixed
  * fix annotation of sds plot
v2.2:
  * use multiprocessing instead of joblib for parallel processing
  * remove joblib as hard dependency, it is only needed for waveform caching
  * fix: averaging did not work when v0 was determined by picks
v2.1:
  * fix: use matplotlib agg backend by default to circumvent problems with parallel plotting on MacOS
v2.0:
  * Qopen is now Python3-only, if you need support for Python2 use the 1.x releases
  * remove parallel option, set njobs to 1 for sequential processing
v1.6:
  * rename omM to sds (source displacement spectrum)
  * site: add option to ignore some stations when aligning site responses
  * allow to pin site response of a set of stations (previously only a single station or all stations)
  * minor bugfixes
v1.5:
  * core: report fit error for source displacement spectra
  * core: collect_results returns dict instead of list
  * imaging: labels for plots of source displacement spectra
  * site: align sites now correctly aligns around the given response,
          before it was shifted by a small value
  * site: recalculate mean site amplifications after alignment
  * site: greatly enhance test suite
  * fixed some minor bugs
  * other small enhancements
v1.4:
  * move some functions from core to new source and site modules
  * API documentation now available on readthedocs
  * bugfix: fix plotting of source model and seismic moment in sds plots
  * core: if picks are used, S picks are compulsory and P picks are optional now
  * core: fix some more ObsPy depreciation warnings
v1.3:
  * fix ObsPy depreciation warnings
  * remove support for ObsPy<1.0
v1.2:
  * bugfix: observed energy now correctly calculated with a factor of 1/2 instead of 1/4
  * core: rename "correct sensitivities" to "align sites"
  * core: "align sites" now only computes on largest contiguous area
  * core: move free surface correction term from qopen.rt module to the calculation of observed energy
  * core: free surface correction is configurable
  * core: new command line option "--calc-source-params" for re-calculating source parameters
  * core: use source model of Abercrombie 2015 to fit source displacement spectra and
          allow to invert for M0, fc, n, and gamma
  * core: add possibility to filter Inventory with Inventory.select
  * core: other new, not very important, but useful command line options
v1.1:
  * bugfix: energy envelope and not velocity seismogram is now divided by filter width
  * bugfix: bulk wave was not correctly handled if beginning of coda window was before S-onset
  * bugfix: seismic moment magnitude now calculated more precisely
  * core: allow to cut the coda at local minimum (useful to cut out aftershocks etc.)
  * imaging: allow to configure grid layout (nx, ny)
v1.0:
  * initial release