Time Series By Regions¶
Let’s see how to extract time series by region (or regions) with eemont!
Before anything, let’s import our modules and authenticate in Google Earth Engine:
import ee, eemont, geemap
import pandas as pd
import numpy as np
ee.Authenticate()
ee.Initialize()
Now, we are ready to go!
Overview¶
The eemont package extends the ee.ImageCollection classes with the methods getTimeSeriesByRegion() and getTimeSeriesByRegions():
ee.ImageCollection¶
|
Gets the time series by region for the given image collection and geometry (feature or feature collection are also supported) according to the specified reducer (or reducers). |
|
Gets the time series by regions for the given image collection and feature collection according to the specified reducer (or reducers). |
Usage¶
The getTimeSeriesByRegion() and getTimeSeriesByRegions() methods extract time series by region (or regions) according to the specified ee.Geometry, ee.Feature or ee.FeatureCollection.
Let’s create an ee.FeatureCollection as example (two crops in France with two identifiers: A and B):
f1 = ee.Feature(ee.Geometry.Point([3.984770,48.767221]).buffer(50),{'ID':'A'})
f2 = ee.Feature(ee.Geometry.Point([4.101367,48.748076]).buffer(50),{'ID':'B'})
fc = ee.FeatureCollection([f1,f2])
Let’s take the Sentinel-2 SR image collection as example (compute NDVI and EVI to extract their values):
S2 = (ee.ImageCollection('COPERNICUS/S2_SR')
.filterBounds(fc)
.filterDate('2020-01-01','2021-01-01')
.maskClouds()
.scale()
.index(['EVI','NDVI']))
Time Series By Region¶
Let’s assume that we want the mean of all pixels inside our collection as a single geometry. In that case, we’ll use the getTimeSeriesByRegion():
ts = S2.getTimeSeriesByRegion(reducer = ee.Reducer.mean(),
geometry = fc,
bands = ['EVI','NDVI'],
scale = 10)
Here, we are extracting the EVI and NDVI time series from the S2 collection by the geometry of our feature collection. If required, we can use more than one reducer:
ts = S2.getTimeSeriesByRegion(reducer = [ee.Reducer.mean(),ee.Reducer.median()],
geometry = fc,
bands = ['EVI','NDVI'],
scale = 10)
Now we are not extracting just the mean values, but also the median values. A column named reducer is created indicating the corresponding reducer.
We can also add arguments that are valid for the reduceRegion() method:
ts = S2.getTimeSeriesByRegion(reducer = [ee.Reducer.mean(),ee.Reducer.median()],
geometry = fc,
bands = ['EVI','NDVI'],
scale = 10,
bestEffort = True,
maxPixels = 1e13,
tileScale = 2)
By default, the output date column is named reducer, but it can be modified:
ts = S2.getTimeSeriesByRegion(reducer = [ee.Reducer.mean(),ee.Reducer.median()],
geometry = fc,
bands = ['EVI','NDVI'],
scale = 10,
dateColumn = 'my_date_column')
The date value is by defult retrieved in the Standard ISO format, but it can be changed by ms (milliseconds) or any other format:
ts = S2.getTimeSeriesByRegion(reducer = [ee.Reducer.mean(),ee.Reducer.median()],
geometry = fc,
bands = ['EVI','NDVI'],
scale = 10,
dateFormat = 'YYYYMMdd')
When the region is masked and the reducer doesn’t retrieve any value, a NA value of -9999 is used, but if required, it can be modified:
ts = S2.getTimeSeriesByRegion(reducer = [ee.Reducer.mean(),ee.Reducer.median()],
geometry = fc,
bands = ['EVI','NDVI'],
scale = 10,
naValue = -9999999)
Time Series By Regions¶
Now, if we want a time series by each feature in our feature collection, we require the getTimeSeriesByRegions() method:
ts = S2.getTimeSeriesByRegions(reducer = [ee.Reducer.mean(),ee.Reducer.median()],
collection = fc,
bands = ['EVI','NDVI'],
scale = 10)
The same parameters of the getTimeSeriesByRegion() method can be used here, except for bestEffort and maxPixels:
ts = S2.getTimeSeriesByRegions(reducer = [ee.Reducer.mean(),ee.Reducer.median()],
collection = fc,
bands = ['EVI','NDVI'],
scale = 10,
naValue = -99999999,
dateColumn = 'my_date_colum',
dateFormat = 'ms')
Conversion to Pandas¶
The time series is always retrieved as an ee.FeatureCollection. To convert the collection to a pandas data frame, we’ll use the geemap package:
tsPandas = geemap.ee_to_pandas(ts)
Then we can convert the NA value to a real NA and the date column to a datetime class:
tsPandas[tsPandas == -9999] = np.nan
tsPandas['date'] = pd.to_datetime(tsPandas['date'],infer_datetime_format = True)