changes.txt

Updates in v0.9.1
ANISOTROPY:
1. This is a new module for anisotropy analysis. Currently, only implemented BANX (beamforming anisotropy analysis from noise cross-correlations)

NOISE:
1. get_stationpairs(): added the option to collect station coordinates.
2. extract_corrdata(): added capibility to determine the side.

HELPERS:
1. xcorr_sides(): added "o" for one-sided but unclear negative or positive and "u" for unavailable side (e.g.,empty data)

UTILS:
1. Added cart2pol(), pol2cart(), and cart2compass() to convert coordinates between cartisian and polar.

SETUP:
1. Removed requirement for python<3.12. It seems to work with 3.12 on Purdue cluster, at least.

TYPES:
1. Added capibility to handle more side symbols for corrdata.

=======================================================
Updates in v0.9.0:
SCRIPTS:
1. Added the scripts for the whole workflow of extracting EGFs.
2. Removed some old scripts that are incompatible with the new usage and functions.

SCRIPTS/2_XCORR_MPI.py:
1. Added a switch named correct_orientation for channel orientation correction/conversion.
2. Added pad_thre to define the maximum threshold of size gaps between channels when doing orientation correction.
3. Added a switch named do_rotation for horizontal channel rotation
4. Added a switch named channel_pairs for specifying which channel pairs are saved.

SIMULATION MODEL:
1. Added build_vmodel() to create a layered model with linearly increasing velocity.

TYPES:
1. Added checking for corrdata.data.ndim when initilizing CorrData to make sure substack True for ndim 2 (or >1).

OBSMASTER:
1. Moved correct_orientations() out to utils, but kept the block of calling this function to be compatible with old usage.

UTILS:
1. Moved correct_orientations() here.
2. Fixed a bug where the number of traces in slicing_trace() is 1 trace less than the true number.

NOISE:
1. Added trace_info() to extract trace and orientation info.
2. Added the function to correct horizontal orientations and convert 1/2 channels to N/E channels.
3. Optimized the workflow for cross-correlation with the option to do rotation before xcorr.
3a. Added the function to do channel rotation specified channel pairs
3b. do_correlation(): added an option to specify which channel pairs are saved.
3c. Added assemble_raw() to assemble raw data from h5 files
3d. Updated assemble_fft() to prepare data for FFT with rotation and specified xcorr pairs.
4. Added get_locator() to read inventory and location id of stations
5. Added rotation() to perform horizontal channel rotation

=============================================================
Updates in v0.8.3:
UTILS:
1. Replace obspy.core.util.base._get_function_from_entry_point for 
  taper hann window with scipy.signal.windows.hann for compatibility with new scipy. 

DOWNLOADERS:
1. Added more verbose prints in getdata().
2. Use utils.taper() for tapering the raw data, instead of obspy's taper function,
  which has version errors in scipy hann window.

NOISE:
1. Fixed minor bugs in int and float, which are now excluded from numpy.

NOTEBOOKS:
1. Updated the seisgo_download_xcorr_demo with the latest version of seisgo.

=============================================================
Updates in v0.8.1
TYPES:
1. Types.CorrData.subset: clarified documentary of overwrite option. Default False.
2. Added a new class for plotting shaded relief based on cartopy.

NOISE:
1. In shaping_corrdata(), modified to loop through all pairs and all components, if not specified by the user.

UTILS:
1. qml2list:add option to only extract location and dates.
2. slicing_trace(): fixed a bug when all trace std was computed using the amplitudes instead of the absolute amplitudes.
3. slicing_trace(): changed to compute trace_stdS after demean and detrend. 
added demean and detrend of whole trace at the beginning.
4. import tukey from scipy.signal.windows to be compatible with different scipy version.

DOWNLOADERS:
1. get_events(): added depth range and search with radius in circles.

PLOTTINGS:
1. plot_stations: added options to specific colors and renamed it to gmt_scatters() to be more general. 
    Old function was kept but will throw an error.

OBSMASTER:
1. import tukey from scipy.signal.windows to be compatible with different scipy version.

=============================================================
Updates in v0.8.0
MONITORING:
1. get_dvv():drop negative error data

TYPES:
1. RawData: minor bug fix for attribute names.
2. Added CorrDataEnsemble() placeholder, to store a gather of CorrData from the same virtual source.
3. Fixed a bug when merging corrdata pairs with different channel types (when ignore_channel_type is True).
4. Fixed typos in CorrData.filter(), where highpass() was typed as lhighpass.
5. CorrData.merge(): fixed bugs when errors in joining two data matrices, the time attribute will still be merged, leading 
to inconsistent sizes.
6. DvvData: drop negative error data and added option to specify maximum error in plot().

NOISE:
1. merge_pairs(): fixed minor bugs. Removed try{} block in stack, split, egf, and saving steps. The try block would hide errors
that might be important to address.

DOWNLOADERS:
1. download(): use "PRESSURE" as output for pressure channel.

UTILS:
1. Change np.float to np.float32.
2. save2asdf(): fixed sta_in when it is a list.
=======
2. save2asdf: Fixed a bug when saving multiple traces with a list of station inventory.

DOWNLOADERS:
1. download(): changed pressure channel rmresp output to "PRESSURE", instead of hardcoding it to "VEL".

STACKING:
1. stack: discard NaN traces before passing the data to each method.
2. robust: fixed a bug when outlier traces have huge amplitudes, which would cause the dot product 
  and L2 norms too large to be handeled by the system. The final stack will be scaled back.
=================================================================
Updates in v0.7.7
TYPES:
1. CorrData.merge(): added option to ignore channel types. This is needed when stations update
  their channels types, e.g., EH? to BH?. This new option would allow station pairs with different
  channel types to merge. Default is False.

NOISE:
1. merge_pairs(): added option of ignore_channel_type.

SETUP:
1. Change numpy requirement to be <1.26.0.
=================================================================
Updates in v0.7.6
UTILS:
1. Added smooth functions for 1d, 2d, and 3d grids.
2. Added matrix_in_polygon to extract 2d or 3d values with,in a polygon.
3. Added boundary_points() to get the boundary of a series of points.

=================================================================
Updates in v0.7.5
NOISE:
1. correlate(): discarded lines removing peak at zero lag (demean of the spectrum).

MONITORING:
1. get_dvv: improved plot saving. Fixed a bug where stack_method was not specified for substacking.
2. ts_dvv: trimed data to only compare the original data window, excluding the padded zeros.
3. xc_dvv: newly added method using moving window cross-correlations. needs improvements and more tests.

DISPERSION:
1. get_dispersion_image: use adaptive window size for different periods, increasing with periods.
  This is options. Options to specify minimum number of trace and minimum number of wavelength.
  * added option to specify energy type, "power_sum" or "envelope".
  * added plotting options.
  * fixed bugs when only processing one side.
2. renamed get_dispersion_waveforms() to narrowband_waveforms to avoid using the word "dispersion".

DOWNLOADERS:
1. read_data(): minor bug fixes.
2. get_data() and get_sta_list(): added flexibility of specifying the source, which could be a Client object.
3. download(): added verbose to calling getdata().

STACKING:
1. Added handling of small data size, i.e., 1 trace, which will return the data without stacking.
2. Added option of using DOST for tfpws, based on the implementation by Jared Bryan. It was added as a new function:tfpws_dost.

UTILS:
1. Added two functions to get image gradient and convert xyz to matrix, without interpolation.
2. extract_waveform: read all data if not specifying stations.

PLOTTING:
1. Added facecolor to set the background of plots to white.
2. Added plot_dispersion_image() to plot dispersion image.

SIMULATION:
1. New module for seismic wave simulations.

===================================================================
Updates in v0.7.4
UTILS:
1. Added spectral whitening function: whiten(), for time series data.

MONITORING:
1. Added whitening option in get_dvv().

TYPES:
1. FFTData, fixed a minor bug when getting Nfft for single trace data.
2. CorrData shaping: trim_end option to accommodate the need by FWANT.

PLOTTING:
1. plot_psd() bug fixed

STACKING:
1. Fixed a bug in robust stacking when the trace amplitudes are anomalously small.

DOWNLOADERS:
1. ms2asdf(): added option to specify response file.

DISPERSION:
1. get_dispersion_waveforms: Use period to get the evenly-spaced period vector.
2. get_dispersion_image: new function to extract dispersion image in velocity-period domain/space.

==================================================================
Updates in v0.7.3
NOISE:
1. Added mpi option and fixed error in handling wrong file, in extract_corrdata()
2. Fixed error handling in get_stationpairs().
3. Added split_sides() to split corrfile sides and save separately.
4. Added shaping_corrdata() to wrap shaping function, convolving with wavelet.
5. Added split option in merge_pairs().

TYPES:
1. Added CorrData.shaping() to shape the data with wavelet.
2. Added CorrData.save() to wrap saving functions.
3. In CorrData.to_asdf(), save stack_method.
4. Fixed a bug in CorrData.to_egf() where the zero lag was not handled correctly. The negative side was wrong.

UTILS:
1. Added gaussian() and ricker() as the shaping wavelets.
2. Added box_smooth().

DOWNLOADERS:
1. In read_data(), changed default to False for getstainv.

HELPERS:
1. Added wavelet_labels().

MONITORING:
1. Added vpcluster_evaluate_kmean() to find the optimal number of clusters.

DISPERSION:
1. Improved frequency steps in disp_waveform_bp().

STACKING:
1. Synced with stackmaster functions.

CLUSTERING:
1. Automatically determine the optimal number of clusters.

==================================================================
Updates in v0.7.2
All modules: Cleaned up unused functions.

TYPES:
1. CorrData.subset() changed overwrite to False by default.
2. DvvData, added get_info() to streamline and wrapup key information for other use.
3. DvvData: added save() as a wrapper for saving functions, including to_asdf() and to_pickle.
4. DvvData: added to_pickle() to save data to pickle files. This is mainly aiming to
    workaround with the problem of saving large attributes in h5 file, through asdf.
5. DvvData: added subfreq as a Boolean attribute.
6. DvvData.plot(): added option to plot when subfreq=False. only plots error bars.

MONITORING:
1. extract_dvvdata: added option to read in pickle files.
2. get_dvv(): added option to save to pickle.
3. get_dvv(): added "ts" method.
4. Modified ts_dvv() to provide filter option and streamlined to work as a standalone method for SeisGo.
5. Delted unused functions.

OBSMASTER:
1. Moved some funcitons from utils.py to here.

NOISE:
1. Added reorganize_corrfile() to reorganize corrfile.
2. Added option in do_correlation() to specify the flag of output_structure.
3. Added comments and descriptions to major functions.

HELPERS:
1. Added this new module to provide a place showing a summary of helper function using SeisGo.

==================================================================
Changes in v0.7.0
STACKING:
1. Added clusterstack() and tfpws().

MONITORING:
1. Added option to specify file name when saving to *.h5 file.

==================================================================
Updates in v0.6.6
MONITORING:
1. Fixed a bug in single cpu mode, where the positive side was used for the negative measurements.

STACKING:
1. Cleaned up stacking methods with simpler names with a consistent fashion
2. Added seisstack() as a wrapper/interface to call all methods.

TYPES:
1. CorrData.to_egf(): Added a statement to check whether the data is already EGFs.
2. CorrData.stack(): demean is now an option. ampcut can be turned off now. This is to
    consider situations when all data need to be stacked and/or the overall trend needs to be preserved.
3. Simplified calling stacking through seisstack. this enables easier future development.

NOISE:
1. Simplified calling stacking through seisstack. this enables easier future development.

UTILS:
1. Added get_snr() to get snr of data with distance information.
2. Added rms() to get the rms of data.
3. Fixed a bug in psd() to return half of the PSD and the corresponding frequency vector.

==================================================================
Updates in v0.6.5
SETUP:
1. Added requirements for tslearn and minisom mainly for the clustering module .

CLUSTERING:
1. Added the new module.
2. Added clustering of velocity depth profiles with kmean and som, in two functions.

PLOTTING:
1. Added get_color_cycle() to help assign colors in plotting using matplotlib.
2. Cleaned up unused old NOISEPY functions.
3. Function to plot vmodel clustering results.

NOISE:
1. Optimized merge_chunks() to use less time getting the time range.

UTILS:
1. Added option to use pattern in get_filelist().

==================================================================
Updates in v0.6.4
TYPES:
1. Check lower case only for method in CorrData.stack()
2. Bugs and improvements in DvvData and CorrData to save large attribute to asdf
3. Added psd() method in CorrData to plot psd of the CorrData result.
4. Added plot() method in FFTData to plot the amplitude spectrum of the FFT results.
5. Improved plot() for DvvData with smoothing option.
6. Minor bug fixes and improvement for FFTData in merging and plotting.

DOWNLOADERS:
1. Use inventory to remove response in read_data for miniseed
2. Added ms2asdf() to convert miniseed files to asdf files.

NOISE:
1. Optimize memory usage for merge_chunks()

PLOTTING:
1. Added plot_psd() to plot the psd of an array (works with 1-d and 2-d array only for now.)
2. Updated plot_eventsequence() to have the option of plotting depth as the ydata.

UTILS:
1. Minor bug fixes in read_gmtlines()
2. Added psd() to get the power spectral density of an array

==================================================================
Updates in v0.6.3.1
1. Fixed the size issue when saving "time" to ASDF file. HDF file limits the attribute
  size to 64k or less. We split time to time_mean and np.float32(time) to reduce the size.
  This is a temporary fix. Hopefully HDF could lift the size limit for attributes.

==================================================================
Updates in v0.6.3
NOISE
1. Removed ncomp in do_correlation(). Setup a warning message if old usage is used.
2. Change defaults for acorr_only and xcorr_only both to False in do_correlation().
3. Added option to stack in merging(). This option could replace do_stacking() if no
  rotation. Renamed merging() to merge_pairs(). The old name is kept for compatibility.
4. Updated extract_corrdata() to read "side" attribute if available.
5. Added merge_chunks() to merge correlation files, to reduce the number of files, with
  the option for stacking.


DOWNLOADERS
1. Return inventory in get_event_waveforms()
2. Drop duplicates in get_sta_list() and fixed minor bug when channels might be skipped.
3. Change default region to globe in get_events()

UTILS
1. Added mag_duration(), modified from obspyDMT.utils.event_handler.py
2. Renamed qml_to_event_list() to qml2list(). Added option to convert to pandas dataframe.
3. Fixed a bug in slicing_trace(), where the index was float instead of integer.
4. Added get_filelist() and slice_list().
5. Fixed a bug in sclicing_trace() when returning zeros array with errors. changed to return
    empty arrays.

PLOTTING
1. Added plot_eventsequence() to plot event with time.

TYPES
1. Added "side" attribute in CorrData() to mark whether the corrdata is two-sided or one-side only.
2. Revised CorrData.plot() to check "side" attribute when plotting.
3. Added copy() method in CorrData class to allow the user to copy the object, to avoid directly
  modifying the object values.
4. Added split() method in CorrData class to split the negative and positive sides of the data. This
  is needed when the user wants to analyze the two sides separately.
5. Removed ngood attribute from CorrData, corresponding changes have been implemented for other
  functions in "noise".
6. Added subset() method in CorrData() to subset data by time range.
7. Added filter() method in CorrData() to filter corrdata.data.
8. Added DvvData class to store dvv monitoring data.

MONITORING
1. Added get_dvv() as a wrapper to measure dvv with given CorrData object.
2. Added extract_dvvdata() to extract DvvData object from a ASDF file.

OBSMASTER
1. Removed getdata() and deprecated getobsdata(). Data downloading is now handled all by downloaders.

==================================================================================
==================================================================================
Updates in v0.6.2

PLOTTING
1. Added plot_stations() to plot seismic station map using GMT

2. Fixed minor bugs in plot_waveform()

3. Updated plot_corrfile() to take more options, more consistent with CorrData.plot()


DOWNLOADERS
1. Fixed bug in get_event_waveforms() where only one station was downloaded. now it downloads
all station data.

2. Return Stream() for waveform data.

UTILS
1. Added subsetindex()
2. Added points_in_polygon() and generate_points_in_polygon()
3. Added read_gmtlines() to read in line segments in GMT style.
4. Added read_ncmodel3d to read 3-D model files in netCDF format.
5. Added read_ncmodel2d to read 2-D model files in netCDF format.
6. Added ncmodel_in_polygon to extract seismic model within polygons.

TYPES
1. Pushed cc_len and cc_step in CorrData.
2. Added in CorrData.stack() the option to stack over segmented time windows.

SETUP
1. Added requirement for shapely and netCDF4 packages.