Google Earth Engine: An Early ReviewFitness Gear & Equipment
Google Earth Engine is the latest offering from Google. The product has the potential to be the most useful and affordable technology in the fight against global warming. The tool also brings the possibility of greater transparency, review and monitoring of environmental issues to the reach of common people. However, because of the complex technologies involved and lack of proper documents and training materials, it is still out of reach for common people to appreciate the pros and cons of its potential implementation. This objective of this document is to clear the haze and to provide an early review of the platform to the common man, the most important stakeholder in fight against global warming.
For a high-level overview you can skip the functional overview section which includes some basic understanding of related technology and calculations.
What is Google Earth Engine?
Google Earth Engine is a product from Google that allows real time monitoring of natural resources across the globe. As per Google definition, it is “an online environment monitoring platform”. It has a huge repository of satellite images of last 25 years integrated with Google Map (3D projection) and powered by Google Cloud infrastructure. Google cloud provides the infrastructure and computing power needed to process the images. In calculating different metrics and alerts, it embeds the algorithms of last 20 years in the area of environmental science and deforestation. Like other products of Google, it will be available to the world as a freeware. Applications will be built on using the exposed API. These applications can be used by government, corporate, NGO and regulatory bodies to implement REDD (Reduced Emission from Deforestation and Forest Degradation in Developing Countries) vision of International Climate Conference. Although REDD vision is yet not agreed upon by all stakeholders, it is anticipated that this will become an inevitable need of the hour in fight against global warming. Google Earth Engine will implement the mission as derived from the REDD vision of MRV (Monitor – Report – Verify) once implemented completely across the globe especially in Tropical countries.
Timeline – chronicle of events
The product was launched on 2nd December 2010 at International Climate Change Conference in Cancun, Mexico. The product is still in beta version. The API is not yet available to the general public. One can request to access the API for independent research to Google. At the moment, the product is lacking proper documentation. The current version is a beta one with limited read-only views exposed to the outside world. There is no timeline or product road map shared publicly so as to apprehend its complete availability. Personally I feel this will be dependent on a common agreement on some methodology/techniques by the participants of last environmental summit participants.
Product USP –Technology behind
The technology behind is the USP or rather a probable monopoly. Technology used to filter raw satellite images to remove clouds and hazes. Transforming raw images to quantifiable data requires huge computing. In a normal client server mode or vanilla network, it might take very high time making it impossible to make any real time decision making. Google cloud offers this computing in almost real time. For example it might even take multiple weeks to compute the deforestation of Amazon for a top line desktop computer. This task is performed by Google cloud in less than a second! So the question of USP does not arise as there is no second provider in the business that can either provide the data or the computing power required for the data. And this is not all! Google Earth Engine also ensures to have the best algorithms available till date elaborated in stakeholder section.
Major Stakeholders of the project
There are different kind of partners and stakeholders in this venture with Google. They can be classified into different categories like:
Landsat is a joint initiative by U.S. Geological survey (USGS) and NASA. Landsat has a series of Earth observing satellite missions since 1972. Landsat has archives of space-based moderate-resolution land remote sensing data. This is being used currently by scientists and researchers globally in planning and monitoring agriculture, regional planning, forestry, geology, mining, global change research etc. It is also invaluable asset for disaster relief and emergency response. Landsat has collaborated with Google and has shared the archive to be used in Google Earth Engine.
Greg Asner of Carnegie Institution for Science – Carnegie has developed algorithms and a software program called CLASlite forest monitoring for last 12 years.
CLASlite is a part of the product engine.
Carlos Souza, Jr. of Imazon – Imazon is a non-profit research institution classified as OSCLIP – Civil Society Public Interest Organization based in Brazil. Imazon is working for last 17 years for sustainable development in Amazon. Like ClASlite, Imazon has very powerful algorithms built in the Imazon software developed by them.
Matt Hansen of Geographic Information Center at South Dakota State University (SDSU) – The Geographic Information Science Center of Excellence is joint initiative by SDSU and U.S. Geological survey’s National Center for Earth Resource Observation and Science (EROS).
Google Earth Outreach team lead by Rebecca Moore is an inspired team of excellent engineers who makes a difference to the existence of indigenous people and earth itself. This team is responsible for Earth Engine design and Surui implementation. Outreach gives non-profits and public benefit organizations the resources including training to use Google Earth/Map.
The development is funded by Google itself and other US governmental bodies and NGOs. The most significant funding partner is Gordon and Betty Moore Foundation who has committed a funding of 12 Million USD for the development. The Moore foundation has also funded US Geological Survey for the same cause. Note that US Geological Survey is already a partner of the project. Moore Foundation is chaired by Gordon and Betty Moore. Gordon Moore is a co-founder of Intel Corporation and chairman of Emeritus.
Local bodies and communities
CONAFOR – CONAFOR is the National Forestry Commission of Mexico. Matt Hansen and CONAFOR with Google Earth Engine team have created an extensive forest cover and water reserve map for Mexico. Jose Carlos Fernandez is the key sponsor from CONAFOR.
Surui is discussed separately under Surui project.
You can access the platform by logging into the product. It consists of four tabs – Home, Data Catalog, Workspace and Map Gallery. The UI is very simple in fact too simple and not as per with best of the Google UIs we are used to enjoy in other products. In fact, if you browse through workspace and try to experiment with different data sets, the AJAX labels seems to append giving wrong or at times irritating user experience. Following is a brief description of each tab and their content. I have explained a few technical terms briefly for the benefit of non technical readers.
The Home tab has no functionality. Nor it is a dashboard for the user customized as per him/her. It only talks about the brief of the product and lists some featured maps with hyperlink. If you click the link, it opens the map under Map Gallery tab.
This is the most important tab of the product. It lists the data set available for viewing and analysis at present. These data majorly come from two major NASA scientific research satellites Terra and Aqua. Both these were launched on 1999. These data are commonly termed as Terra data and Aqua data. All the data under this section are orthorectified meaning they are geometrically corrected such that the scale is uniform. The data is grouped under three major categories:
Loosely speaking, a surface reflectance is the reflected light data captured by the satellite through its sensors. Terra for example, has seven band sensors built in within. These sensors and the associated data are maintained by MODIS team under NASA. MODIS stands for Moderate-Resolution Imaging Spectroradiometer. A Surface Reflectance product is a raw product which provides major input utilized in the generation of several land products: Vegetation Indices, BRDF (Bidirectional Reflectance Distribution Function), Land Cover, Snow Cover, Thermal Anomalies, and LAI/FPAR (LAI – Leaf Area Index, FPAR – Fraction of Photosythetically Active Radiation absorbed by vegetation) etc. Terra and Aqua has in built sensors to capture almost all surface reflectance data. These data are most valuable inputs to environmental science and climate system. MODIS team produces different products after pre-processing the raw satellite images. These products or data is loosely called MODIS products. The MODIS surface reflectance products provide an estimate of the surface spectral reflectance as it would be measured in ground level in the absence of atmospheric scattering or absorptions.
Currently there are three MODIS products under Data Catalogue. The data set MCD43A4 is derived from both terra and aqua data and hence named differently under MCD series. The data is available for the period Jan 1, 2010 to September 30, 2010. The rest of the two data sets are to be used in conjunction with each other to get meaningful information. These are MOD09GA and MOD09GQ. These data is available for the period Mar 2, 2010 to July 1, 2010.
Mosaic data sets are grouped under 2 different categories viz. 8-day composites and 32-day composites. These data are essentially of three different composites like BAI composites, EVI composites and NBRT composites. Data set under this are made of L1t orthocentric scenes. All the composite data are available for the period Jan 1, 1984 to November 25, 2010. Following is a brief description of these three composites. The difference is considering the number of days in each scene. For 8 day and 32 days, each scene is composed of all scenes for 8 days and 32 days respectively.
Both these sets present the BAI (Burn Area Index). Loosely speaking these data sets are more to do with forest degradation. BAI is a metrics developed by Martin and Chuvieco in 1998. BAI is used in many forms for forest fire detection. BAI gradient for example, can be useful for studying spatial evolution of forest fire. The purpose of BAI is to highlight the burned area and can be obtained as:
In general, BAI is not a normalized index and thus inappropriate for multitemporal study. Moreover it is not clear to me at least whether the BAI is calculated after removing cloud and hazes which can cause significant errors in BAI. L1T datasets are terrain corrected by incorporating ground control points (GCP). However, as per Landsat, outside U.S., the accuracy of terrain corrected points depend on availability of the ground control points or GCPs. The GCP data for 1 Terrain correction comes from GLS2005 data set. For certain scenes, those do not have ground control or elevation data necessary for precision or terrain correction, respectively. In these cases, the best level of correction will be applied (Level 1G-systematic).
EVI stands for Enhanced Vegetation Index. Loosely speaking, it measures the growth of the forest in accordance with different weather patterns. The most significant work using the EVI is done by Alfredo Huete and his team of University of Arizona in 2006. They have shown significant difference in growth of Amazon in dry season which impacts the understanding of our carbon cycle and global warming. It is yet to be seen if these observations are factually supported. The formula for EVI is as following:
EVI has become extremely popular among scientific communities after the launch of Terra and Aqua, the two MODIS sensors. EVI is currently distributed freely by USGS.
NBRT stands for Normalized Burn Ratio Thermal. It is a Landsat data derived Burn Severity Index (BSI). This is essentially calculated from pre-fire and post-fire Normalized Burn Ratio (NBR). Upon comparing the actual data of Post Fire Effect (PFE) and different BSIs, NBRT is chosen by many scientists as a cheaper alternative with high degree of accuracy. This is highly used to decipher the burned and unburned regions.
Once you select a data catalog item, you can open it in the workspace. It shows the data mapped onto Google Map. You can visually see the data and application of different computations as available for the data. The computations currently available are: i. Normalized Difference Vegetation Index ii. Enhanced Vegetation Index iii. Normalized Burn Ratio Thermal iv. Burn Area Index v. Normalized Difference Water Area Index vi. Normalized Difference Snow Index
You can define and view the data for your own viewport. This is useful if you want to study the behavior of a zone or country. The sample maps available in map gallery section are created by defining custom viewport. The current workspace provides only a read-only view of one particular day or composite (8-day, 32-day). It does not provide any other function to save the image or data and a means of comparison the data with any other day/composite. That functionality must be in testing phase as this is a beta release. Currently Google Earth / Google Map API do not support Earth Engine integration.
This section features five beautiful maps showcasing the actual power of the platform. One of the maps demonstrates the most comprehensive scale map of Mexico’s forest and water resources ever made. That project alone would have taken three years to process using a single computer, Google officials say, but took just one day using Google Earth Engine. Another sample under this section is NDFI over Amazon. NDFI is a new Index (Normalized Difference Fraction Index) designed to detect deforestation and forest degradation. NDFI is used by Imazon for monitoring forest change. If you look at the forest canopy (canopy is the green cover) damage visually through map, you will actually be scared and disturbed. The Congo Forest Cover Loss map for just 10 years (2000 – 2010) is another example why REDD is important to agree upon. Just have a look at the map and count every color other than green is a loss. You will know what I am talking about.
A note about data and algorithm
It is to be noted that the different indices and patterns as calculated from analyzing different satellite images, need to be validated and corrected with ground data. Ground level data is obtained from physical inspection and data collection. Country specific research and development wing has to collaborate with Earth Engine team for these data. While creating the water resource map for Mexico, CONAFOR has provided this data for Mexico.
Google Earth Engine review is incomplete without discussing the Surui project. “What you need is a living breathing daily update of the planet. Google Earth was not designed for that. That’s why we are building Google Earth Engine” – Rebecca Moore, Google Earth Engine Manager. While the map gallery in the product demonstrates some best of the maps of resources, the Surui project highlights the daily online living update. Surui is a distinct community in the Rodonia region of Brazil who has a distinct separate culture from contemporary western culture. Like many tribal communities, the survival of the community is under threat from deforestation and non sustainable development. Google through its outreach program of Google Earth is associated with the community since 2008. The community people are trained in using Google Earth through Android. They have put their cultural map information in Google Earth and uploaded the videos in YouTube. As part of Surui carbon project under Google Earth Engine, they have now started using the technology to record and quantify the carbon measurements by filling up the form in the handhold Android device. Also, they keep a close eye on the regular surface reflectance data and easily identify any suspicious activity leading to deforestation or forest degradation. This way, the community is serving the great cause to the mankind and also making money for their livelihood which would otherwise have to come from economic growth leading to deforestation and forest degradation. The voice of Chief Almir Surui talks about a future – “a future with conscience”. No matter what you think about Earth Engine, this is the vision of tomorrow if we all have to be a part of that bright sunny day.
Platform vision and mission
This is a grey area. I did not find either of vision or mission statement of the product in any of the released artifact so far. Hope Google will soon update us on this.
Officially speaking, Google is offering this technology to the world as not-for-profit service. At its product launch event, Google has announced to donate 10 Million CPU hours a year over next 2 years to build the platform. The product has many potential commercial usage prospects like mapping water resources, ecosystem services, mining and deforestations. Most of these are invaluable applications to governments and large corporate bodies. It is still not very clear whether the technology can be used free of cost for commercial reason. However, as I have explained the USP or rather monopoly of technology, the following points can easily be derived.
• Most of the Tropical underdeveloped countries do not have the data, technology and infrastructure for this kind of project. Hence once they start using it in deforestation MRV (Monitor – Review – Verify) they will be highly dependent on Google. Remember the REDD vision talks about significant grants towards achieving this. Thus Google will have a big say in environmental policy or decision. This is one way increasing the political influence over majority of the Tropical countries. On the other hand, U.S. and rest of the developed countries will look up to Google to get things done in their favor.
• So far Google has donated resources required for this. The declared agenda is to support the development of the platform for next couple of years. We never know, once fully developed and implemented, who will bear the cost of running it. If there is no competitive platform like this, one never knows what that cost will be translated in dollar.
• This is also a beautiful way to get all the data and computations under Google Cloud platform. Consider a national ID card and monitoring project in countries like China and India where populations are well above 1000 million. You can never have anything else for a real time data processing other than Google Cloud.
• The project demonstrates and develops API needed for high end multimedia data processing algorithms. As the world has moved on from simple character based input to click of the mouse, it is anticipated that future will see inputs from sophisticated multimedia devices like webcam and so on. Investing in this technology thus gives Google an edge over competitors in terms of reduced time required to develop fresh API.
• As the product is dedicated to the mankind, it guarantees Google to get significant tax benefit. Also, this also ensures that in a recession time like 2009, Google will get sufficient grant from government to keep the engine running.
• The association of Landsat and Google opens up the possibility of Google having a greater say in U.S. security, monitoring, space research and other programs under USGS (U.S. Geological Survey). Google might become the preferred or rather obvious vendor for all of these agencies.
Having mentioned all these points, one must acknowledge the importance of the product in the history of mankind. Today, probably Global Warming is the biggest challenge to our very existence. Even terrorism can be fought and localized but Global Warming can cause a much greater damage. One must acknowledge the contribution of Google Labs in fighting this monster beyond any other point.
- Google Earth Engine: http://www.google.org/earthengine/ - Official Google blog introducing Google Earth Engine: http://googleblog.blogspot.com/2010/12/introducing-google-earth-engine.html - Official Google Blog Seeing the forest through the cloud: http://googleblog.blogspot.com/2009/12/seeing-forest-through-cloud.html - UN REDD programme: http://www.un-redd.org/ - COP16/CMP6: http://www.cc2010.mx/en/ - Landsat: http://landsat.usgs.gov/ - USGS: http://www.usgs.gov/ - Integration and Application Network: http://ian.umces.edu/ - Geo FCT portal: http://portal.geo-fct.org/ - Gordon and Betty Moore Foundation: http://www.moore.org/ - Wikipedia: http://en.wikipedia.org/wiki/Main_Page - IEEE Xplore: http://ieeexplore.ieee.org/Xplore/guesthome.jsp - The IEEE Computer Society: http://www.computer.org/portal/web/csdl
Google Earth Engine Overview: http://www.youtube.com/watch?v=MnCf9Gjz720 Google Earth Engine Partners: http://www.youtube.com/watch?v=omKFvpJok3I Google Earth Engine and the Surui: http://www.youtube.com/watch?v=riYSJBD8gEM Google Earth Engine Set to be an Important Tool for Conservation : http://www.youtube.com/watch?v=v0nuC8ZV-BQ