Time Series By Regions (with Conversion to Pandas and Visualization with Seaborn)

Tutorial created by **David Montero Loaiza**: GitHub | Twitter

GitHub Repo: https://github.com/davemlz/eemont
PyPI link: https://pypi.org/project/eemont/
Conda-forge: https://anaconda.org/conda-forge/eemont
Documentation: https://eemont.readthedocs.io/
More tutorials: https://github.com/davemlz/eemont/tree/master/docs/tutorials

Let’s start!

If required, please uncomment:

[1]:

#!pip install eemont
#!pip install geemap

Import the required packages.

[2]:

import ee, eemont, geemap
import pandas as pd
import numpy as np
import seaborn as sns
from matplotlib import pyplot as plt

Authenticate and Initialize Earth Engine and geemap.

[3]:

Map = geemap.Map()

Let’s use some center-pivot crops in Egypt:

[4]:

pivots = ee.FeatureCollection([
    ee.Feature(ee.Geometry.Point([27.724856,26.485040]).buffer(400),{'pivot':0}),
    ee.Feature(ee.Geometry.Point([27.719427,26.478505]).buffer(400),{'pivot':1}),
    ee.Feature(ee.Geometry.Point([27.714185,26.471802]).buffer(400),{'pivot':2})
])

Let’s pre-process and process our image collection:

[5]:

L8 = (ee.ImageCollection('LANDSAT/LC08/C01/T1_SR')
      .filterBounds(pivots)
      .maskClouds()
      .scaleAndOffset()
      .spectralIndices(['EVI','GNDVI']))

Time Series By Regions

Let’s get the L8 time series for our buffer. Checklist:

Image Collection: The Landsat 8 collection.
Bands to use for the time series: GNDVI and EVI.
Feature Collection: Our center-pivot crops.
Statistics to compute: Mean and Median.
Scale: 30 m.

[6]:

ts = L8.getTimeSeriesByRegions(collection = pivots,
                               bands = ['EVI','GNDVI'],
                               reducer = [ee.Reducer.mean(),ee.Reducer.median()],
                               scale = 30)

Conversion to Pandas

The time series is retrieved as a feature collection. To convert it to a pandas dataframe we’ll use geemap (This may take a little bit):

[7]:

tsPandas = geemap.ee_to_pandas(ts)

Let’s check our pandas data frame:

[8]:

tsPandas

[8]:

	date	pivot	EVI	reducer	GNDVI
0	2013-04-01T08:40:36	0	0.585979	mean	0.648333
1	2013-04-01T08:40:36	1	0.131134	mean	0.312830
2	2013-04-01T08:40:36	2	0.642244	mean	0.697033
3	2013-04-11T08:40:46	0	0.290635	mean	0.482938
4	2013-04-11T08:40:46	1	0.099826	mean	0.297480
...	...	...	...	...	...
1141	2021-09-08T08:37:40	1	0.327657	median	0.507243
1142	2021-09-08T08:37:40	2	0.289696	median	0.451613
1143	2021-10-10T08:37:49	0	0.154731	median	0.366600
1144	2021-10-10T08:37:49	1	0.446784	median	0.630140
1145	2021-10-10T08:37:49	2	0.199530	median	0.403407

1146 rows × 5 columns

What can we see here?

The values for each band (GNDVI and EVI) are in separated columns.
There are some -9999 values in the GNDVI and EVI columns. These values represent the NA values (e.g. Clouds or shadows). The -9999 can be changed by modifying the naValue parameter in the getTimeSeriesByRegion method (e.g. naValue = -10000).
Multiple reducers can be used. In the output dataframe they are specified by a single column named reducer: mean, median.
The date is a string that needs to be converted to a date data type.
The attributes of the original feature collection are attached to the data frame: pivot.

Given this, let’s curate our data frame!

First, let’s get rid of the -9999 value:

[9]:

tsPandas[tsPandas == -9999] = np.nan

And now, let’s convert the date to a date data type:

[10]:

tsPandas['date'] = pd.to_datetime(tsPandas['date'],infer_datetime_format = True)

We can also gather the GNDVI and EVI columns into a single column to make the data frame more ‘tidy-er’ (This is optional):

[11]:

tsPandas = pd.melt(tsPandas,
                   id_vars = ['reducer','date','pivot'],
                   value_vars = ['GNDVI','EVI'],
                   var_name = 'Index',
                   value_name = 'Value')

Let’s check our curated data frame:

[12]:

tsPandas

[12]:

	reducer	date	pivot	Index	Value
0	mean	2013-04-01 08:40:36	0	GNDVI	0.648333
1	mean	2013-04-01 08:40:36	1	GNDVI	0.312830
2	mean	2013-04-01 08:40:36	2	GNDVI	0.697033
3	mean	2013-04-11 08:40:46	0	GNDVI	0.482938
4	mean	2013-04-11 08:40:46	1	GNDVI	0.297480
...	...	...	...	...	...
2287	median	2021-09-08 08:37:40	1	EVI	0.327657
2288	median	2021-09-08 08:37:40	2	EVI	0.289696
2289	median	2021-10-10 08:37:49	0	EVI	0.154731
2290	median	2021-10-10 08:37:49	1	EVI	0.446784
2291	median	2021-10-10 08:37:49	2	EVI	0.199530

2292 rows × 5 columns

Visualization

Now, let’s visualize our time series using seaborn:

[13]:

g = sns.FacetGrid(tsPandas,row = 'pivot',height = 3,aspect = 4)
g.map_dataframe(sns.lineplot,x = 'date',y = 'Value',hue = 'Index',style = 'reducer')
g.add_legend()

[13]:

<seaborn.axisgrid.FacetGrid at 0x7f8cb5dd34f0>

../_images/tutorials_014-Time-Series-By-Regions-Pandas_33_1.png

Time Series By Region (with Conversion to Pandas and Visualization with Seaborn)

Histogram Matching