You are on page 1of 26

GEE toolbox: efficient computing using

+
August 19, 2020

Leandro Parente - leandro.parente@opengeohub.org


Post-doctoral researcher - OpenGeoHub Foundation
https://opengeohub.org/people/leandro-parente
Presentation Outline
● Google Earth Engine introduction

● R integration

● Code demo

● Concluding remarks

+
Google Earth Engine introduction
https://earthengine.google.com

A (free) cloud-computing platform capable to analyse (on-the-fly) Earth


science data in a planetary-scale
Google Earth Engine introduction
https://earthengine.google.com

A (free) cloud-computing platform capable to analyse (on-the-fly) Earth


science data in a planetary-scale

Regular workflow

...

Several EO data providers

Download
the data

Customized algorithms
and analysis
Google Earth Engine introduction
https://earthengine.google.com

A (free) cloud-computing platform capable to analyse (on-the-fly) Earth


science data in a planetary-scale

Regular workflow Earth Engine workflow

... ...

Several EO data
in one place
Several EO data providers

Download
the data
Upload
the code

Customized algorithms Customized algorithms


and analysis and analysis
Public Available
Datasets

All Landsat (Since 1985) and


Sentinel 1/2/3 Archive

SRTM and ALOS data

OpenLandMap soil data

Several climate datasets

and much more in Earth


Engine Data Catalog
Google Earth Engine introduction

Source: Noel et al., 2017


https://doi.org/10.1016/j.rse.2017.06.031
Google Earth Engine introduction

Annual Median for NDVI - Landsat-8 SR - 2019

https://code.earthengine.google.com/885924eea422efb01335ac4d99b146d9
Source: Noel et al., 2017
https://doi.org/10.1016/j.rse.2017.06.031
Google Earth Engine introduction

Annual Median for NDVI - Landsat-8 SR - 2019

https://code.earthengine.google.com/885924eea422efb01335ac4d99b146d9
Source: Noel et al., 2017
https://doi.org/10.1016/j.rse.2017.06.031
Google Earth Engine introduction

Annual Median for NDVI - Landsat-8 SR - 2019

Temporal reduction (median): 0.4434

https://code.earthengine.google.com/885924eea422efb01335ac4d99b146d9
Source: Noel et al., 2017
https://doi.org/10.1016/j.rse.2017.06.031
Google Earth Engine introduction

Flexible Workflow

...
Data
Landsat archive
exportation
(since 1985) Download
the output

Upload
the code Customized algorithms
and analysis
Landsat temporal
reduction

Source: Noel et al., 2017


https://doi.org/10.1016/j.rse.2017.06.031
Purely Earth Engine
Google Earth Engine introduction
https://code.earthengine.google.com/de9ecc00be820384b8b9001c1a0b219c?noload=1

Global NDVI (AVHRR)


Google Earth Engine introduction The smooth and gapfill used
Greenbrown package
R integration

...and about R ?

Source: Noel et al., 2017


Gif Source https://doi.org/10.1016/j.rse.2017.06.031
R integration

https://r-spatial.github.io/rgee/

Unofficial Earth Engine API,


but a great piece of software

Source: Noel et al., 2017


https://doi.org/10.1016/j.rse.2017.06.031
R integration
How does rgee work?

reticulate

Adapted: https://github.com/r-spatial/rgee

Source: Noel et al., 2017


https://doi.org/10.1016/j.rse.2017.06.031
R integration
Survival guide for rgee

● Sign up for Google Earth Engine

● Install the rgee package (virtualenv or docker image)

● Setup and authenticate (Google Earth Engine, Drive, Cloud Storage)

● Read the best practices

● Start by rgee examples (the same of GEE user guide, but in R)

● Remember that the official GEE documentation is

only available for javascript in the code editor

Image source
Code demo
Land-Cover classification for one Landsat scene (226/67) in the Amazon Forest

Upload the training


samples to GEE

Run the temporal


reduction

Overlay the samples

Tune the RF
hyperparameters

Run the classification


for the whole area

Export the
classification result
Code demo
Land-Cover classification for one Landsat scene (226/67) in the Amazon Forest

Source code and other files: https://tinyurl.com/y6np8gke


Code demo
To try yourself...

● Expand the feature space using more percentiles or others reducers:

● Why not use Sentinel-2 ?

To understand the valid.expr access the thread “Sentinel-2 cloud free mosaic from the MSK_CLDPRB and SLC bands”
Concluding remarks
● The use of Google Earth Engine by scientific
community is increasing

According to SCOPUS 697 document contains the


term “Google Earth Engine” (August 17, 2020)
Concluding remarks
● The use of Google Earth Engine by scientific
community is increasing

Some relevant
special issues
According to SCOPUS 697 document contains the
term “Google Earth Engine” (August 17, 2020)
Concluding remarks
● GEE is a outstanding platform that allows the users to test large-scale geospatial analysis quickly and
on-the-fly

● The best way to use GEE in your workflow is exporting only the final, or intermediate, result of a
specific analysis (e.g. annual median image). Export the raw images is possible, but it's not
recommended

● Multiple Google accounts can operate together to scale a particular processing (e.g. land-cover
classification for a whole country/continent), generating results in a specific Google Cloud Storage
bucket

● It’s possible to use GEE to serve images in Web-GIS applications.

● Join the Geo for Good Summit to find out the latest news about GEE
Leandro Parente
leandro.parente@opengeohub.org
https://opengeohub.org/people/leandro-parente
Extra
What are the alternatives to Google Earth Engine ?

pipsCloud

https://doi.org/10.3390/rs12081253
Extra
What are the alternatives to Google Earth Engine ?

pipsCloud

https://doi.org/10.3390/rs12081253

You might also like