What is Google Earth Engine?
This post will discuss a geospatial technology from Google called Google Earth Engine (GEE). Maybe some people think that Google Earth Engine is similar to Google Earth (GE). But both are much different. Google Earth is an application for visualization of geospatial data with a programming language called Keyhole Markup Language (KML). While Google Earth Engine is a cloud-based platform for geospatial data analysis, mainly raster data. Those definition clearly give the differences between Google Earth and Google Earth Engine, which is: Google Earth is for geospatial visualization while Google Earth Engine is a platform for data processing.
Then, what are the advantages of GEE? So it is worthy to be called a "Geospatial technology innovation". There are many reasons for that. They are:
- Have access to satellite imagery and other data with a very large amount (petabytes), and updated regularly. (Please see dataset). So we don't need to explore and search geospatial data from different data sources. More than that, most of the data are free.
- Parallel data processing on Google cloud platform.
- Numerous data processing algorithms are available and continue to be refined by Google Engineer and by experts around the world who are using GEE for their research.
- Application Programming Interface (API) available for Java Script and Python, allowing users to perform more complex data processing in accordance with their needs.
- Comprehensive documentation available for beginner and advanced user.
From the reasons above, imagine! You can access thousands of existing data on the cloud, no big data storage is required and process such data without purchasing any software. What you only need is an internet connection.
Google Earth Engine Application Example
Maybe it will be more clear by an example: You are assigned to monitor changes in landcover of an area of between 1999 and 2008 using Normalized Difference Vegetation Index (NDVI) indicator. In general, the approach may be as follows:
- Finding and downloading Landsat satellite imagery for the two years. If one scene per year is 1 GB for multi-spectral imagery, then at least 2 GB of data must be downloaded.
- Process the images into a geospatial/remote sensing software and calculate each NDVI for respective year.
- Compare the NDVI value of each image to determine changes in the region's land.
Yes. That's at least three steps necessary to resolve the case. But it will be much different using GEE platform, because it will be done through an editor with the following lines of code:
|Picture 1. NDVI 1999|
|Picture 2. NDVI 2008|