5. Slopes analysis, SOM…¶
# common
import sys
import os
import os.path as op
# basic
import xarray as xr
import numpy as np
import pandas as pd
from datetime import timedelta as td
from matplotlib import pyplot as plt
from pandas.plotting import register_matplotlib_converters
# interactive
from ipywidgets import interactive
# advanced
import plotly.express as px
import plotly.graph_objects as go
import plotly.figure_factory as ff
import dash # (version 1.12.0) pip install dash
import dash_core_components as dcc
import dash_html_components as html
from dash.dependencies import Input, Output
# warnings
import warnings
warnings.filterwarnings("ignore")
# dev library
sys.path.insert(0, os.getcwd())
# RBF module
from slopes import Slopes_SOM
from slopes import consult
# pip renders for jupyter book
import plotly.io as pio
pio.renderers.default = "notebook"
# load the data
data = pd.read_pickle(op.join(os.getcwd(), '..', 'data', 'reconstructed',
'surfbreaks_reconstructed_final.pkl'))
data['Index'] = data['Index'].where(data['Index']<1, 1) * 10
data = data.dropna(how='any', axis=0)
data.info()
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 1008091 entries, 1979-02-01 02:00:00 to 2020-02-29 20:00:00
Data columns (total 27 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Hs 1008091 non-null float64
1 Tp 1008091 non-null float64
2 Dir 1008091 non-null float64
3 Spr 1008091 non-null float64
4 W 1008091 non-null float64
5 DirW 1008091 non-null float64
6 DDir 1008091 non-null float64
7 DDirW 1008091 non-null float64
8 ocean_tide 1008091 non-null float64
9 Omega 1008091 non-null float64
10 H_break 1008091 non-null float64
11 DDir_R 1008091 non-null float64
12 Slope 1008091 non-null float64
13 Iribarren 1008091 non-null float64
14 Hb_index 1008091 non-null float64
15 Tp_index 1008091 non-null float64
16 Spr_index 1008091 non-null float64
17 Iribarren_index 1008091 non-null float64
18 Dir_index 1008091 non-null float64
19 DirW_index 1008091 non-null float64
20 Index 1008091 non-null float64
21 Hour 1008091 non-null int64
22 Day_Moment 1008091 non-null object
23 Month 1008091 non-null int64
24 Season 1008091 non-null object
25 Year 1008091 non-null int64
26 beach 1008091 non-null object
dtypes: float64(21), int64(3), object(3)
memory usage: 215.4+ MB
data.Index.plot.box()
<AxesSubplot:>
data.Hb_index.hist()
<AxesSubplot:>
data.Index.hist()
<AxesSubplot:>
consult(data=data,
beaches=['segunda', 'laredo', 'forta'],
day='2017-01-13',
columns=['H_break', 'Hb_index', 'Index'])
| H_break | Hb_index | Index | beach | |
|---|---|---|---|---|
| 2017-01-13 03:00:00 | 3.857180 | 0.3 | 3.376 | segunda |
| 2017-01-13 08:00:00 | 3.926694 | 0.3 | 2.532 | segunda |
| 2017-01-13 13:00:00 | 3.804837 | 0.3 | 3.376 | segunda |
| 2017-01-13 18:00:00 | 4.560021 | 0.0 | 1.000 | segunda |
| 2017-01-13 23:00:00 | 3.389903 | 0.5 | 5.088 | segunda |
| 2017-01-13 03:00:00 | 3.747338 | 0.3 | 2.664 | laredo |
| 2017-01-13 08:00:00 | 3.819149 | 0.3 | 2.664 | laredo |
| 2017-01-13 13:00:00 | 3.717768 | 0.5 | 4.384 | laredo |
| 2017-01-13 18:00:00 | 4.273107 | 0.3 | 3.776 | laredo |
| 2017-01-13 23:00:00 | 3.435028 | 0.5 | 4.608 | laredo |
| 2017-01-13 03:00:00 | 1.646689 | 0.7 | 4.128 | forta |
| 2017-01-13 08:00:00 | 1.679265 | 0.7 | 5.216 | forta |
| 2017-01-13 13:00:00 | 1.637391 | 0.7 | 7.568 | forta |
| 2017-01-13 18:00:00 | 2.068835 | 1.0 | 9.438 | forta |
| 2017-01-13 23:00:00 | 1.480459 | 0.7 | 8.250 | forta |
consult(data=data,
beaches=['segunda', 'laredo', 'forta'],
day='2019-12-22',
columns=['H_break', 'Hb_index', 'Index'])
| H_break | Hb_index | Index | beach | |
|---|---|---|---|---|
| 2019-12-22 01:00:00 | 3.646619 | 0.5 | 4.320 | segunda |
| 2019-12-22 06:00:00 | 3.164886 | 0.5 | 4.860 | segunda |
| 2019-12-22 11:00:00 | 2.948261 | 0.8 | 6.264 | segunda |
| 2019-12-22 16:00:00 | 3.183460 | 0.5 | 4.320 | segunda |
| 2019-12-22 21:00:00 | 2.933703 | 0.8 | 5.856 | segunda |
| 2019-12-22 01:00:00 | 3.782744 | 0.3 | 3.664 | laredo |
| 2019-12-22 06:00:00 | 3.431986 | 0.5 | 5.382 | laredo |
| 2019-12-22 11:00:00 | 3.289403 | 0.5 | 5.058 | laredo |
| 2019-12-22 16:00:00 | 3.717900 | 0.5 | 4.496 | laredo |
| 2019-12-22 21:00:00 | 3.537950 | 0.5 | 4.976 | laredo |
| 2019-12-22 01:00:00 | 1.630818 | 0.7 | 4.212 | forta |
| 2019-12-22 06:00:00 | 1.441102 | 0.7 | 4.356 | forta |
| 2019-12-22 11:00:00 | 1.392555 | 0.7 | 4.212 | forta |
| 2019-12-22 16:00:00 | 1.624850 | 0.7 | 4.212 | forta |
| 2019-12-22 21:00:00 | 1.505351 | 0.7 | 4.356 | forta |
consult(data=data,
beaches=['brusco', 'canallave', 'laredo'],
day='2003-06-20',
columns=['H_break', 'Hb_index', 'Index'])
| H_break | Hb_index | Index | beach | |
|---|---|---|---|---|
| 2003-06-20 02:00:00 | 2.131825 | 1.0 | 6.798 | canallave |
| 2003-06-20 07:00:00 | 1.979701 | 1.0 | 5.168 | canallave |
| 2003-06-20 12:00:00 | 2.064915 | 1.0 | 5.168 | canallave |
| 2003-06-20 17:00:00 | 2.103380 | 1.0 | 1.000 | canallave |
| 2003-06-20 22:00:00 | 1.854655 | 1.0 | 7.106 | canallave |
| 2003-06-20 02:00:00 | 1.792632 | 1.0 | 6.886 | brusco |
| 2003-06-20 07:00:00 | 1.678462 | 0.7 | 3.472 | brusco |
| 2003-06-20 12:00:00 | 1.858943 | 1.0 | 1.000 | brusco |
| 2003-06-20 17:00:00 | 1.997746 | 1.0 | 1.000 | brusco |
| 2003-06-20 22:00:00 | 1.703752 | 1.0 | 1.000 | brusco |
| 2003-06-20 02:00:00 | 1.234815 | 0.7 | 1.000 | laredo |
| 2003-06-20 07:00:00 | 1.291503 | 0.7 | 1.000 | laredo |
| 2003-06-20 12:00:00 | 1.678732 | 0.7 | 1.000 | laredo |
| 2003-06-20 17:00:00 | 1.842311 | 1.0 | 1.000 | laredo |
| 2003-06-20 22:00:00 | 1.498014 | 0.7 | 1.000 | laredo |
def plot_scatter(x, y):
fig = px.scatter(data[::100], x=x, y=y, color='beach')
fig.show()
interactive_plot = interactive(plot_scatter,
x=data.columns,
y=data.columns)
interactive_plot
def plot_bar(grouper, variable):
fig = px.bar(data.groupby([grouper, 'beach']).mean().reset_index(),
y=variable,
x=grouper,
color='beach',
barmode='group')
fig.show()
interactive_plot = interactive(plot_bar,
grouper=['Hour', 'Day_Moment', 'Month', 'Season', 'Year'],
variable=data.columns)
interactive_plot
def plot_box(period, y):
data_box = data.copy()
if isinstance(period, np.int64):
fig = px.box(data_box.where(data_box['Month']==period).dropna(how='all', axis=0),
x='beach', y=y, title='Month: '+str(period))
elif period=='all':
fig = px.box(data_box, x='beach', y=y, title='Points: '+period)
elif period in ['Winter', 'Spring', 'Summer', 'Autumn']:
fig = px.box(data_box.where(data_box['Season']==period).dropna(how='all', axis=0),
x='beach', y=y, title='Season: '+period)
else:
fig = px.box(data_box.where(data_box['Day_Moment']==period).dropna(how='all', axis=0),
x='beach', y=y, title='Day moment: '+period)
fig.show()
period = ['all'] + list(data.Season.unique()) + list(data.Month.unique()) + list(data.Day_Moment.unique())
interactive_plot = interactive(plot_box,
period=period,
y=data.columns)
interactive_plot
def plot_hist(period, x):
data_hist = data[::100].copy()
if isinstance(period, np.int64):
fig = px.histogram(data_hist.where(data_hist['Month']==period).dropna(how='all', axis=0),
x=x, color='beach', marginal='box',
hover_data=data.columns, title='Month: '+str(period))
elif period=='all':
fig = px.histogram(data_hist,
x=x, color='beach', marginal='box',
hover_data=data.columns, title='Points: '+period)
elif period in ['Winter', 'Spring', 'Summer', 'Autumn']:
fig = px.histogram(data_hist.where(data_hist['Season']==period).dropna(how='all', axis=0),
x=x, color='beach', marginal='box',
hover_data=data.columns, title='Season: '+period)
else:
fig = px.histogram(data_hist.where(data_hist['Day_Moment']==period).dropna(how='all', axis=0),
x=x, color='beach', marginal='box',
hover_data=data.columns, title='Day moment: '+period)
fig.show()
period = ['all'] + list(data.Season.unique()) + list(data.Month.unique()) + list(data.Day_Moment.unique())
interactive_plot = interactive(plot_hist,
period=period,
x=data.columns)
interactive_plot
def plot_prob(beach):
histcolor = ['blue', 'green', 'yellow', 'orange', 'red', 'purple', 'black']
data_prob = data.where(data['beach']==beach).dropna(how='all', axis=0).copy()
data_prob = data_prob.groupby([data_prob.index.dayofyear,
pd.cut(data_prob['Index'],
[0,1,3,5,7,8,9,10],
right=True)])\
.count().mean(axis=1) / (len(data_prob)/366)
data_prob.name = 'Probability of RSI'
fig = px.histogram(data_prob.reset_index(),
x='level_0', y='Probability of RSI',
color='Index',
color_discrete_map={key: value for (key, value) in zip(
data_prob.reset_index()['Index'].unique(),
histcolor)},
nbins=366, range_y=[0,1],
labels={'level_0': 'Day of year'},
title='Beach: ' + beach, width=900, height=400)
fig.show()
interactive_plot = interactive(plot_prob,
beach=data.beach.unique())
interactive_plot
def plot_prob(beach, ini_year, end_year):
histcolor = ['blue', 'green', 'yellow', 'orange', 'red', 'purple', 'black']
data_prob = data.where(data['beach']==beach).dropna(how='all', axis=0).copy()
data_prob = data.where((data['Year']>=ini_year) & (data['Year']<=end_year))\
.dropna(how='all', axis=0).copy()
data_prob = data_prob.groupby([pd.Grouper(freq='M'),
pd.cut(data_prob['Index'],
[0,1,3,5,7,8,9,10],
right=True)])\
.count().mean(axis=1) / (len(data_prob)/(12*(end_year-ini_year + 1)))
data_prob.name = 'Probability of RSI'
fig = px.histogram(data_prob.reset_index(),
x='level_0', y='Probability of RSI',
color='Index',
color_discrete_map={key: value for (key, value) in zip(
data_prob.reset_index()['Index'].unique(),
histcolor)},
nbins=12*int(end_year-ini_year + 1), range_y=[0,1],
labels={'level_0': 'Historical month'},
title='Beach: ' + beach, width=900, height=400)
fig.show()
interactive_plot = interactive(plot_prob,
beach=data.beach.unique(),
ini_year=data.Year.unique(),
end_year=data.Year.unique())
interactive_plot
def plot_prob_grouper(beach, grouper):
histcolor = ['blue', 'green', 'yellow', 'orange', 'red', 'purple', 'black']
data_prob = data.where(data['beach']==beach).dropna(how='all', axis=0).copy()
data_prob = data_prob.groupby([data_prob[grouper],
pd.cut(data_prob['Index'],
[0,1,3,5,7,8,9,10],
right=True)])\
.count().mean(axis=1) / (len(data_prob)/len(data[grouper].unique()))
data_prob.name = 'Probability of RSI'
fig = px.histogram(data_prob.reset_index(),
x=grouper, y='Probability of RSI',
color='Index',
color_discrete_map={key: value for (key, value) in zip(
data_prob.reset_index()['Index'].unique(),
histcolor)},
range_y=[0,1], nbins=len(data[grouper].unique()),
labels={'level_0': grouper},
title='Beach: ' + beach, width=900, height=400)
fig.show()
grouper = ['Month', 'Season', 'Year']
interactive_plot = interactive(plot_prob_grouper,
beach=data.beach.unique(),
grouper=grouper)
interactive_plot
def plot_rose(beach):
data_prob = data.where(data['beach']==beach).dropna(how='all', axis=0).copy()
data_prob = data_prob.groupby([pd.cut(data_prob['Dir'],
[0,20,40,60,80,100,120,140,160,180,
200,220,240,260,280,300,320,340,360],
right=True),
pd.cut(data_prob['Tp'],
[0,6,8,10,12,14,16,18,24],
right=True)]).mean()
data_prob = data_prob['H_break'].rename('H_break_value', inplace=True).reset_index()
data_prob = data_prob.astype({'Dir': 'str', 'Tp': 'str'}, copy=True)
fig = px.bar_polar(data_prob, r='H_break_value', theta='Dir',
color='Tp', color_discrete_sequence=px.colors.sequential.Plasma_r)
fig.show()
interactive_plot = interactive(plot_rose,
beach=data.beach.unique())
interactive_plot
slopes_som = Slopes_SOM(data)
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=80000, plot_results=True)
slopes_som.plot_results(som, data_mean, data_count, plot_beaches=True, plot_months=True)
The following data will be trained:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 1008091 entries, 1979-02-01 02:00:00 to 2020-02-29 20:00:00
Data columns (total 27 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Hs 1008091 non-null float64
1 Tp 1008091 non-null float64
2 Dir 1008091 non-null float64
3 Spr 1008091 non-null float64
4 W 1008091 non-null float64
5 DirW 1008091 non-null float64
6 DDir 1008091 non-null float64
7 DDirW 1008091 non-null float64
8 ocean_tide 1008091 non-null float64
9 Omega 1008091 non-null float64
10 H_break 1008091 non-null float64
11 DDir_R 1008091 non-null float64
12 Slope 1008091 non-null float64
13 Iribarren 1008091 non-null float64
14 Hb_index 1008091 non-null float64
15 Tp_index 1008091 non-null float64
16 Spr_index 1008091 non-null float64
17 Iribarren_index 1008091 non-null float64
18 Dir_index 1008091 non-null float64
19 DirW_index 1008091 non-null float64
20 Index 1008091 non-null float64
21 Hour 1008091 non-null int64
22 Day_Moment 1008091 non-null object
23 Month 1008091 non-null int64
24 Season 1008091 non-null object
25 Year 1008091 non-null int64
26 beach 1008091 non-null object
dtypes: float64(21), int64(3), object(3)
memory usage: 215.4+ MB
None
slopes_som = Slopes_SOM(data, beach='farolillo')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: farolillo
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='bederna')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: bederna
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='oyambre')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: oyambre
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='locos')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: locos
The sum off all probabilities is: 0.9999999999999999
slopes_som = Slopes_SOM(data, beach='valdearenas')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: valdearenas
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='canallave')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: canallave
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='madero')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: madero
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='segunda')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: segunda
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='primera')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: primera
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='pueblo')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: pueblo
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='curva')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='DirW_index', plot_months=True)
The following data will be trained in: curva
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='brusco')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Dir_index', plot_months=True)
The following data will be trained in: brusco
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='forta')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Spr_index', plot_months=True)
The following data will be trained in: forta
The sum off all probabilities is: 1.0
slopes_som = Slopes_SOM(data, beach='laredo')
som, data_mean, data_count = slopes_som.train(som_shape=(20,20), sigma=0.8, learning_rate=0.5,
num_iteration=50000, plot_results=False)
slopes_som.plot_results(som, data_mean, data_count, second_plot='Hb_index', plot_months=True)
The following data will be trained in: laredo
The sum off all probabilities is: 1.0
