def low_pass(series, window_size=193, T=40, dt=360):
"""
This function applies a lanczos filter on a pandas time-series and
returns the low pass data series.
series : Pandas series
window_size : Size of the filter windows (default is 96+1+96).
T : Period of the filter. (The default of 40 hours filter should
get rid of all tides.)
dt : time_delta in seconds. Default is 360 (6 minutes).
"""
from oceans import lanc
T *= 60 * 60. # To seconds.
freq = dt / T
mask = np.isnan(series)
avg = series.mean()
series = series - avg
series.interpolate(inplace=True)
wt = lanc(window_size, freq)
low = np.convolve(wt, series, mode='same')
low = ma.masked_array(low, mask)
return low + avg
def low_pass(series, window_size=193, T=40, dt=360):
"""
This function applies a lanczos filter on a pandas time-series and
returns the low pass data series.
series : Pandas series
window_size : Size of the filter windows (default is 96+1+96).
T : Period of the filter. (The default of 40 hours filter should
get rid of all tides.)
dt : time_delta in seconds. Default is 360 (6 minutes).
"""
from oceans import lanc
T *= 60*60. # To seconds.
freq = dt/T
mask = np.isnan(series)
avg = series.mean()
series = series - avg
series.interpolate(inplace=True)
wt = lanc(window_size, freq)
low = np.convolve(wt, series, mode='same')
low = ma.masked_array(low, mask)
return low+avg
shape(ts1h) # <codecell> roll_l = pd.rolling_mean(df['u'], window=40, center=True, freq='1H') df.plot() # <codecell> import numpy as np from oceans import lanc window_size = 96+1+96 freq = 1./40 wt = lanc(window_size, freq) df['low'] = np.convolve(wt, df['u'], mode='same') df['high'] = df['u'] - df['low'] # <codecell> import pandas as pd roll_l = pd.rolling_mean(df['u'], window=40, center=True, freq='1H') roll_h = df['u'] - roll_l # <codecell> import iris from iris.pandas import as_cube