spacepy_logo

Source code for spacepy.plot.spectrogram

#!/usr/bin/env python
# -*- coding: utf-8 -*-

"""
Create and plot generic 'spectrograms' for space science.
This is not a signal processing routine and does not apply
Fourier transforms (or similar) to the data. The functionality
provided here is the binning (and averaging) of multi-dimensional
to provide a 2D output map of some quantity as a function of two
parameters. An example would be particle data from a satellite mission:
electron flux, at a given energy, can be binned as a function of
both time and McIlwain L, then plotted as a 2D color-map,
colloquially known as a spectrogram.

In many other settings 'spectrogram' refers to a transform of data
from the time domain to the frequency domain, and the subsequent plotting
of some quantity (e.g., power spectral density) as a function of time and
frequency. To approximate this functionality for, e.g., time-series magnetic field
data you would first calculate a the power spectral density and then use
:class:`spectrogram` to rebin the data for visualaization.

Authors: Brian Larsen and Steve Morley
Institution: Los Alamos National Laboratory
Contact: balarsen@lanl.gov, smorley@lanl.gov
Los Alamos National Laboratory

Copyright 2011 Los Alamos National Security, LLC.

.. autosummary::
    :template: clean_class.rst
    :toctree: autosummary

    spectrogram
"""

import bisect
import datetime

import numpy as np
import matplotlib
from matplotlib.dates import date2num
from matplotlib.colors import LogNorm

import spacepy.datamodel as dm
import spacepy.toolbox as tb

__contact__ = 'Brian Larsen, balarsen@lanl.gov'

[docs]class spectrogram(dm.SpaceData): """ This class rebins data to produce a 2D data map that can be plotted as a spectrogram It is meant to be used on arbitrary data series. The first series "x" is plotted on the abscissa and second series "y" is plotted on the ordinate and the third series "z" is plotted in color. The series are not passed in independently but instead inside a :class:`~spacepy.datamodel.SpaceData` container. Parameters ========== data : :class:`~spacepy.datamodel.SpaceData` The data for the spectrogram, the variables to be used default to "Epoch" for x, "Energy" for y, and "Flux" for z. Other names are specified using the 'variables' keyword. All keywords override .attrs contents. Other Parameters ================ variables : list keyword containing the names of the variables to use for the spectrogram the list is a list of the SpaceData keys in x, y, z, order bins : list if the name "bins" is not specified in the .attrs of the dmarray variable this specifies the bins for each variable in a [[xbins], [ybins]] format xlim : list if the name "lim" is not specified in the .attrs of the dmarray variable this specifies the limit for the x variable [xlow, xhigh] ylim : list if the name "lim" is not specified in the .attrs of the dmarray variable this specifies the limit for the y variable [ylow, yhigh] zlim : list if the name "lim" is not specified in the .attrs of the dmarray variable this specifies the limit for the z variable [zlow, zhigh] extended_out : bool (optional) if this is True add more information to the output data model (defualt False) Notes ===== Helper routines are planned to facilitate the creation of the SpaceData container if the data are not in the format. .. autosummary:: ~spectrogram.plot .. automethod:: plot """ # TODO # ==== # Allow for the input of a list of SpaceData objects for different sats # Make "subclasses" that allow for data to be passed in directly avoiding the data model def __init__(self, data, **kwargs): """ """ super(spectrogram, self).__init__() ## setup a default dictionary to step through to set values from kwargs self.specSettings = dm.SpaceData() self.specSettings['variables'] = ['Epoch', 'Energy', 'Flux'] self.specSettings['bins'] = None # this is the linspace over the range with sqrt() of the len bins self.specSettings['xlim'] = None self.specSettings['ylim'] = None self.specSettings['zlim'] = None self.specSettings['extended_out'] = False self.specSettings['axisDates'] = False # if the key exists in kwargs replace setting with it, otherwise its an error for key in kwargs: if key not in self.specSettings: raise(KeyError('Invalid keyword specified ' + str(key))) self.specSettings[key] = kwargs[key] # check to see if the variables are in the spacedata for var in self.specSettings['variables']: if not var in data: # TODO could check other capitalization raise(KeyError('"{0}" not found in the input data'.format(var) )) # if the variables are empty error and quit if len(data[self.specSettings['variables'][0]]) == 0: raise(ValueError('No {0} datapassed in'.format(self.specSettings['variables'][0]))) if len(data[self.specSettings['variables'][1]]) == 0: raise(ValueError('No {0} datapassed in'.format(self.specSettings['variables'][1]))) if len(data[self.specSettings['variables'][2]]) == 0: raise(ValueError('No {0} datapassed in'.format(self.specSettings['variables'][2]))) # set limits, keywords override those in the data if self.specSettings['xlim'] == None: try: if 'lim' in data[self.specSettings['variables'][0]].attrs: self.specSettings['xlim'] = (data[self.specSettings['variables'][0]].attrs['lim'][0], data[self.specSettings['variables'][0]].attrs['lim'][1]) else: dum1 = np.min(data[self.specSettings['variables'][0]]).tolist() #TODO: tolist here is a workaround for a bug in dum2 = np.max(data[self.specSettings['variables'][0]]).tolist() # datamodel's min method self.specSettings['xlim'] = (dum1, dum2) except AttributeError: # was a numpy array not dmarray self.specSettings['xlim'] = (np.min(data[self.specSettings['variables'][0]]), np.max(data[self.specSettings['variables'][0]])) if self.specSettings['ylim'] == None: try: if 'lim' in data[self.specSettings['variables'][1]].attrs: self.specSettings['ylim'] = (data[self.specSettings['variables'][1]].attrs['lim'][0], data[self.specSettings['variables'][1]].attrs['lim'][1]) else: self.specSettings['ylim'] = (np.min(data[self.specSettings['variables'][1]]), np.max(data[self.specSettings['variables'][1]])) except AttributeError: # was a numpy array not dmarray self.specSettings['ylim'] = (np.min(data[self.specSettings['variables'][1]]), np.max(data[self.specSettings['variables'][1]])) if self.specSettings['zlim'] == None: try: if 'lim' in data[self.specSettings['variables'][2]].attrs: self.specSettings['zlim'] = (data[self.specSettings['variables'][2]].attrs['lim'][0], data[self.specSettings['variables'][2]].attrs['lim'][1]) else: self.specSettings['zlim'] = (np.min(data[self.specSettings['variables'][2]]), np.max(data[self.specSettings['variables'][2]])) except AttributeError: # was a numpy array not dmarray self.specSettings['zlim'] = (np.min(data[self.specSettings['variables'][2]]), np.max(data[self.specSettings['variables'][2]])) # are the axes dates? forcedate = [False] * 2 if isinstance(data[self.specSettings['variables'][0]][0], datetime.datetime): forcedate[0] = True if isinstance(data[self.specSettings['variables'][1]][0], datetime.datetime): forcedate[1] = True self.specSettings['axisDates'] = forcedate # set default bins if self.specSettings['bins'] == None: # since it is not set by keyword was it set in the datamodel? attr_bins = ['bins' in data[var].attrs for var in self.specSettings['variables']] if dm.dmarray(attr_bins[0:2]).all(): self.specSettings['bins'] = [dm.dmarray(data[self.specSettings['variables'][0]].attrs['bins']), dm.dmarray(data[self.specSettings['variables'][1]].attrs['bins']),] # TODO this is not a hard extension to doing one with bins and one default else: # use the toolbox version of linspace so it works on dates self.specSettings['bins'] = [dm.dmarray(tb.linspace(self.specSettings['xlim'][0], self.specSettings['xlim'][1], np.sqrt(len(data[self.specSettings['variables'][0]])))), dm.dmarray(tb.linspace(self.specSettings['ylim'][0], self.specSettings['ylim'][1], np.sqrt(len(data[self.specSettings['variables'][1]]))))] # copy all the used keys for key in self.specSettings['variables']: self[key] = data[key] try: self[key].attrs = data[key].attrs except: pass # do the spectrogram self._computeSpec() def _computeSpec(self): """ Method operates on the input data to bin up the spectrogram and adds it to the spectrogram class data """ # this is here for in the future when we take a list a SpaceData objects sz = (self.specSettings['bins'][1].shape[0]-1, self.specSettings['bins'][0].shape[0]-1) overall_sum = dm.dmarray(np.zeros(sz, dtype=np.double)) overall_count = dm.dmarray(np.zeros(sz, dtype=np.long)) # the valid range for the histograms _range = [self.specSettings['xlim'], self.specSettings['ylim']] # if x/y is a time need to convert it to numbers (checking first element) var_time = [False]*2 for ivar, var in enumerate(self.specSettings['variables']): if isinstance(self[var][0], datetime.datetime): try: self.specSettings['bins'][ivar] = matplotlib.dates.date2num(self.specSettings['bins'][ivar]) except AttributeError: # it is already changed to date2num pass try: #weird issue with arrays and date2num _range[ivar] = matplotlib.dates.date2num(_range[ivar].tolist()) except AttributeError: try: # ugg if this is not an array this breaks _range[ivar] = [matplotlib.dates.date2num(val.tolist()) for val in _range[ivar]] except AttributeError: _range[ivar] = [matplotlib.dates.date2num(val) for val in _range[ivar]] var_time[ivar] = True # ok not as a dmarray since it is local now plt_data = np.vstack((self[self.specSettings['variables'][0]], self[self.specSettings['variables'][1]])) for ival, val in enumerate(var_time): if val: plt_data[ival] = date2num(plt_data[ival]) if plt_data.dtype.name == 'object': # why was this an abject plt_data = plt_data.astype(float) # go through and get rid of "bad" counts zdat = np.ma.masked_outside(self[self.specSettings['variables'][2]], self.specSettings['zlim'][0], self.specSettings['zlim'][1]) zind = ~zdat.mask # ma has the annoying feature of if all the masks are the same just giving one value try: if len(zind) == len(zdat): pass except TypeError: # no len to a scalar # ok not as a dmarray since it is local now zind = np.asarray([zind]*len(zdat)) # get the number in each bin H, xedges, yedges = np.histogram2d(plt_data[0, zind], plt_data[1, zind], bins = self.specSettings['bins'], range = _range, ) # this is here for in the future when we take a list a SpaceData objects np.add(overall_count, np.require(H.transpose(), dtype=overall_count.dtype), overall_count) # get the sum in each bin H, xedges, yedges = np.histogram2d(plt_data[0, zind], plt_data[1, zind], bins = self.specSettings['bins'], range = _range, weights = zdat.data[zind] ) np.add(overall_sum, H.transpose(), overall_sum) overall_count = np.ma.masked_array(overall_count, overall_count == 0) data = np.ma.divide(overall_sum, overall_count) ## for plotting #ind0 = data.data <= 0 #data[ind0] = np.ma.masked # add to the mask #data = np.ma.log10(data) self['spectrogram'] = dm.SpaceData() self['spectrogram'].attrs = self.specSettings self['spectrogram']['spectrogram'] = dm.dmarray(data, attrs={'name':str(self.specSettings['variables'][0]) + ' ' + str(self.specSettings['variables'][1]), 'xedges':'xedges', 'yedges':'yedges',}) self['spectrogram']['xedges'] = dm.dmarray(xedges, attrs={'name':str(self.specSettings['variables'][0]), 'lim':[self.specSettings['xlim']],}) self['spectrogram']['yedges'] = dm.dmarray(yedges, attrs={'name':str(self.specSettings['variables'][1]), 'lim':[self.specSettings['ylim']],}) if self.specSettings['extended_out']: self['spectrogram']['count'] = overall_count self['spectrogram']['sum'] = overall_sum def add_data(self, data): if not self.specSettings['extended_out']: raise(NotImplementedError('Cannot add data to a spectrogram unless "extended_out" was True on inital creation')) b = spectrogram(data, **self.specSettings) # if they are both masked keep them that way mask = self['spectrogram']['count'].mask & b['spectrogram']['count'].mask # turn off the mask by setting sum and count to zero where it masked for self an be sure b mask doesnt it self self['spectrogram']['count'][self['spectrogram']['count'].mask] = 0 b['spectrogram']['count'][b['spectrogram']['count'].mask] = 0 # put the mask back they are both bad self['spectrogram']['count'].mask = mask b['spectrogram']['count'].mask = mask self['spectrogram']['count'] += b['spectrogram']['count'] self['spectrogram']['sum'] += b['spectrogram']['sum'] self['spectrogram']['spectrogram'][...] = np.ma.divide(self['spectrogram']['sum'], self['spectrogram']['count'])
[docs] def plot(self, fignum=None, axis=None, **kwargs): """ Plot the spectrogram Other Parameters ================ title : str plot title (default '') xlabel : str x axis label (default '') ylabel : str y axis label (default '') colorbar_label : str colorbar label (default '') DateFormatter : matplotlib.dates.DateFormatter The formatting to use on the dates on the x-axis (default matplotlib.dates.DateFormatter("%d %b %Y")) zlog : bool plot the z variable on a log scale (default True) colorbar : bool plot the colorbar (default True) axis : matplotlib axis object axis to plot the spectrogram to zlim : np.array array like 2 element that overrides (interior) the spectrogram zlim (default spectrogram.specSettings['zlim']) figsize : tuple (optional) tuple of size to pass to figure(), None does the default """ # go through the passed in kwargs to plot and look at defaults import matplotlib.pyplot as plt plotSettings_keys = ('title', 'xlabel', 'ylabel', 'DateFormatter', 'zlim', 'colorbar', 'colorbar_label', 'zlog', 'xlim', 'ylim', 'figsize') for key in kwargs: if key not in plotSettings_keys: raise(KeyError('Invalid keyword argument to plot(), "' + key + '"')) self.plotSettings = dm.SpaceData() if 'title' in kwargs: self.plotSettings['title'] = kwargs['title'] else: self.plotSettings['title'] = '' if 'xlabel' in kwargs: self.plotSettings['xlabel'] = kwargs['xlabel'] else: self.plotSettings['xlabel'] = '' if 'ylabel' in kwargs: self.plotSettings['ylabel'] = kwargs['ylabel'] else: self.plotSettings['ylabel'] = '' if 'zlog' in kwargs: self.plotSettings['zlog'] = kwargs['zlog'] else: self.plotSettings['zlog'] = True if 'DateFormatter' in kwargs: self.plotSettings['DateFormatter'] = kwargs['DateFormatter'] else: self.plotSettings['DateFormatter'] = matplotlib.dates.DateFormatter("%d %b %Y") if 'zlim' in kwargs: self.plotSettings['zlim'] = kwargs['zlim'] elif 'zlim' in self.specSettings: self.plotSettings['zlim'] = self.specSettings['zlim'] else: self.plotSettings['zlim'] = self.specSettings['zlim'] if 'colorbar' in kwargs: self.plotSettings['colorbar'] = kwargs['colorbar'] else: self.plotSettings['colorbar'] = True if 'colorbar_label' in kwargs: self.plotSettings['colorbar_label'] = kwargs['colorbar_label'] else: self.plotSettings['colorbar_label'] = '' if 'xlim' in kwargs: self.plotSettings['xlim'] = kwargs['xlim'] elif 'xlim' in self.specSettings: self.plotSettings['xlim'] = self.specSettings['xlim'] else: self.plotSettings['xlim'] = None if 'ylim' in kwargs: self.plotSettings['ylim'] = kwargs['ylim'] elif 'ylim' in self.specSettings: self.plotSettings['ylim'] = self.specSettings['ylim'] else: self.plotSettings['ylim'] = None if fignum is None and axis is None: if 'figsize' in kwargs: if kwargs['figsize'] != None: fig = plt.figure(figsize=kwargs['figsize']) else: fig = plt.figure() else: fig = plt.figure() ax = fig.add_subplot(111) elif axis is None: fig = plt.figure(fignum) ax = fig.add_subplot(111) else: ax = axis bb = np.ma.masked_outside(self['spectrogram']['spectrogram'], *self.plotSettings['zlim']) if self.plotSettings['zlog']: pcm = ax.pcolormesh(self['spectrogram']['xedges'], self['spectrogram']['yedges'], np.asarray(bb), norm=LogNorm()) else: pcm = ax.pcolormesh(self['spectrogram']['xedges'], self['spectrogram']['yedges'], np.asarray(bb)) if self.specSettings['axisDates'][0]: time_ticks = self._set_ticks_to_time(ax, 'x') elif self.specSettings['axisDates'][1]: time_ticks = self._set_ticks_to_time(ax, 'y') ax.set_title(self.plotSettings['title']) ax.set_xlabel(self.plotSettings['xlabel']) ax.set_ylabel(self.plotSettings['ylabel']) if self.plotSettings['ylim'] != None: ax.set_ylim(self.plotSettings['ylim']) if self.plotSettings['xlim'] != None: ax.set_xlim(self.plotSettings['xlim']) if self.plotSettings['colorbar']: if 'cmap' in kwargs: self.plotSettings['cmap'] = kwargs['cmap'] else: self.plotSettings['cmap'] = matplotlib.cm.rainbow cb =plt.colorbar(pcm) cb.set_label(self.plotSettings['colorbar_label']) return ax
def vslice(self, value): """ slice a spectrogram at a given position along the x axis, maintains variable names from spectrogram Parameters ========== value : float or datetime.datetime the value to slice the spectrogram at Return ====== out : datamodel.SpaceData spacedata containing the slice """ # using bisect find the index of the spectrogram to use if isinstance(value, datetime.datetime): value = date2num(value) ind = bisect.bisect_right(self['spectrogram']['xedges'], value) ans = dm.SpaceData() ans[self['spectrogram'].attrs['variables'][1]] = tb.bin_edges_to_center(self['spectrogram']['yedges']) ans['yedges'] = self['spectrogram']['yedges'].copy() ans['xedges'] = self['spectrogram']['xedges'][ind:ind+2].copy() ans[self['spectrogram'].attrs['variables'][2]] = self['spectrogram']['spectrogram'][:,ind:ind+1] return ans def hslice(self, value): """ slice a spectrogram at a given position along the y axis, maintains variable names from spectrogram Parameters ========== value : float or datetime.datetime the value to slice the spectrogram at Return ====== out : datamodel.SpaceData spacedata containing the slice """ # using bisect find the index of the spectrogram to use if isinstance(value, datetime.datetime): value = date2num(value) ind = bisect.bisect_right(self['spectrogram']['yedges'], value) ans = dm.SpaceData() ans[self['spectrogram'].attrs['variables'][0]] = tb.bin_edges_to_center(self['spectrogram']['xedges']) ans['yedges'] = self['spectrogram']['yedges'][ind:ind+2].copy() ans['xedges'] = self['spectrogram']['xedges'].copy() ans[self['spectrogram'].attrs['variables'][2]] = self['spectrogram']['spectrogram'][ind, :] return ans def _set_ticks_to_time(self, axis, xy): """ given the axis change the ticks to times """ timeFmt = self.plotSettings['DateFormatter'] if xy == 'x': ticks = axis.get_xticks() axis.set_xticklabels(matplotlib.dates.num2date(ticks)) axis.xaxis.set_major_formatter(timeFmt) axis.get_figure().autofmt_xdate() elif xy == 'y': ticks = axis.get_yticks() axis.set_yticklabels(matplotlib.dates.num2date(ticks)) axis.yaxis.set_major_formatter(timeFmt) axis.get_figure().autofmt_ydate() def __str__(self): return "<spectrogram object>" __repr__ = __str__