Skip to main content

Codes to perform Dynamic Time Warping Based Hierarchical Agglomerative Clustering of GPS data

Project description

Dynamic Time Warping based Hierarchical Agglomerative Clustering

Codes to perform Dynamic Time Warping Based Hierarchical Agglomerative Clustering of GPS data


Installation and usage information can be obtained from the documentation: dtwhaclustering.pdf

Complete documentation at: dtwhaclustering-docs


This package include codes for processing the GPS displacement data including least-square modelling for trend, co-seismic jumps, seasonal and tidal signals. Finally, it can be used to cluster the GPS displacements based on the similarity of the waveforms. The similarity among the waveforms will be obtained using the DTW distance.


Least-squares modeling

Load Pickle Data into Pandas DataFrame

from dtwhaclustering.leastSquareModeling import lsqmodeling
final_dU, final_dN, final_dE = lsqmodeling(dUU, dNN, dEE,stnlocfile="helper_files/stn_loc.txt",  plot_results=True, remove_trend=False, remove_seasonality=True, remove_jumps=False)

LSQ Model

Plot station map

from dtwhaclustering import plot_stations
plot_stations.plot_station_map(station_data = 'helper_files/selected_stations_info.txt', outfig=f'{outloc}/station_map.pdf')

Plot linear trend

df = pd.read_csv(slopeFile, names=['stn','lon','lat','slope'], delimiter='\s+')
plot_linear_trend_on_map(df, outfig=f"Maps/slope-plot_U.pdf")

Note: slopeFile is obtained from lsqmodeling.

Dynamic Time Warping Analysis

from dtwhaclustering.dtw_analysis import dtw_signal_pairs, dtw_clustering
import numpy as np
from scipy import signal
import matplotlib.pyplot as plt

# sampling parameters
fs = 100   # sampling rate, in Hz
T = 1      # duration, in seconds
N = T * fs  # duration, in samples

# time variable
t = np.linspace(0, T, N)

SNR = 0.2 #noise

XX0 = np.sin(2 * np.pi * t * 7+np.pi/2) #+ np.random.randn(1, N) * SNR
XX1 = signal.sawtooth(2 * np.pi * t * 5+np.pi/2) #+ np.random.randn(1, N) * SNR
# XX1 = np.abs(np.cos(2 * np.pi * t * 3)) - 0.5
s1, s2 = XX0, XX1

dtwsig = dtw_signal_pairs(s1, s2, labels=['S1', 'S2'])

dtwsig.plot_matrix(windowfrac=0.6, psi=None) #Only allow for shifts up to 60% of the minimum signal length away from the two diagonals.


  1. Kumar, U., Chao, B.F., Chang, E.T.-Y.Y., 2020. What Causes the Common‐Mode Error in Array GPS Displacement Fields: Case Study for Taiwan in Relation to Atmospheric Mass Loading. Earth Sp. Sci. 0–2.


© 2021 Utpal Kumar

Licensed under the Apache License, Version 2.0

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

dtwhaclustering-1.0.15.tar.gz (26.8 kB view hashes)

Uploaded Source

Built Distribution

dtwhaclustering-1.0.15-py3-none-any.whl (28.2 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page