4th GEOSS Science and Technology Stakeholder Workshop
March 24-26, 2015, Norfolk, VA, USA

Scope | Deadlines | Venue | Logistics | Committees | Participation | Output | Abstracts | Registration | Accommodation | Sessions | Schedule | Program

Workshop infos

Co-Located Events

Relevant Links and Documents


Utilizing the emerging data super nova for Earth observations (Text for the Video on Mobile Data and Big Data)

Ben Burford, Mobile Science, USA

Hello, my name is Ben Burford. I would like to share some information and thoughts on the Internet of Things, big data analysis, and what they could mean to GEOSS.

This slide is kind of busy but it shows the variety of data that the Internet of Things will produce - air pollution, forest fire detection, health of structures like bridges, ships and ship cargo movement, perimeter access control, radiation levels, electromagnetic levels, traffic congestion, smart roads, smart lighting, noise maps, water leakage, waste management, smart parking, water quality, and others.

A large part of IoT data will be Earth observation data. And of the “non-Earth Observation” data, much of it will be useful for GEOSS purposes when combined with EO data. Which brings up an important point - integration of multiple data types is more important than quantity of data and the Internet of things will produce a tremendous variety of data.

Lets look at some examples to understand the scale of the Internet of Things.

The sensors on jet engines, for all flights globally, produce about 12 exabytes (thats 12,000 petabytes or 12 million terabytes) of data per day. In contrast, NASA’s EOSDIS captures about 1 terabyte of data per day. That’s a ratio of 12 million to 1.

You’ve probably heard this before - “A full 90% of all the data in the world has been generated over the last two years.” This is so called “Big Data”, such as cell tower call logs and search query logs. Much of this data is actually useful for GEOSS purposes, for example processing Google search queries to predict epidemics, and processing cell tower logs to track population movement in an epidemic.

Cell phones will be a major source of data. In 2014, 800 million smartphone subscriptions were added worldwide taking the total number to 2.7 billion. The number of sensors on smart phones is up to about 18, and more, such as temperature and humidity sensors, will be added. In addition, lets look at some examples of sensors that can be connected to smart phones.

This is a chemical sniffer developed by NASA. It has 32 nanosensors where each responds to different chemicals and it can monitor concentrations in real-time at levels of parts per million to parts per billion. It can be manufactured for pennies per chip, and plugs into a smart phone to operate

Researchers have developed a cradle for smart phones that contains optics and uses photonics crystal technology which can detect and measure biological materials, such as proteins, cells, pathogens or DNA. The cradle holds about $200 of optical components but it performs as accurately as a $50,000 spectrophotometer. The device is not only portable it is also affordable for fieldwork in developing countries, for example, a test for iron and vitamin a deficiency in expectant mothers.

Other cell-phone devices include Ultrasound imagers, EEG recorders, lensless holographic imagers, wide-field fluorescent imagers, spectrometers and instrument packages.

Unmanned Aerial Vehicle use is growing rapidly. UAVs can carry RGB cameras, infrared cameras, multispectral cameras, lidar and a variety of sensors.

All in one software solutions process the data and provide a range of data products such as orthophotography, lidar point clouds, digital surface models, 3D modeling, contour maps and volumetric measurement. Several studies have pointed out that 80% of UAV use will be for agriculture.

The world of satellites is being massively transformed. Planet Labs has received $160 million in funding and has launched 71 cubesat satellites (and it would be 97 except for the Antares rocket explosion last October). They’re mass produced to minimize cost and they provide imagery in the 3 to 5 meter resolution range.

Skybox imaging built and launched 2 refrigerator sized satellites that produce 1 meter resolution imagery before they were bought by Google for $500 million.

Spire has received $27 million in funding and is preparing 100 CubeSats, the size of wine bottles, to be launched by the end of 2017. Collectively, they're designed to provide nearly real-time weather data and as many as 10,000 weather readings daily, about five times more than publicly funded satellites deliver.

Rocket Lab has received funding and is developing a new rocket that is 18m long and can launch a 110kg payload. It is designed to launch cubesats at a cost of $4.9 million per launch and will be operational in late 2016. More examples could be given.

But how can this vast quantity and variety of Internet of Things data be utilized? “Big Data analysis” is the close companion to the Internet of Things. Fortunately Big Data analysis is receiving a great deal of attention and investment from the business world, and this is resulting in the rapid advancement of software to utilize Big Data.

Hadoop is the software system most commonly used for Big Data analysis. It allows thousands (even tens of thousands) of servers to be applied simultaneously on a single problem, at low cost using cloud computing.

However, there is a new Big Data analysis system called “Spark” that is faster and more flexible than Hadoop. Like Hadoop it can operate on thousands of servers simultaneously, but Spark is 10 to 100 times faster than Hadoop.

More importantly, Spark is a general purpose analysis system providing a machine learning library, and soon providing a raster data processing library of routines that can be used on satellite data.

One example of Earth observation data processing on Spark is already under development, a system to downscale climate model output data and use analysis tools that enable decision makers to examine community vulnerability at the micro-level. This will enable them to develop proactive strategies to address impending climate change impacts specific to their regions.

Machine learning can support predictive analytics to produce actionable information. In the business world predictive analytics are used to predict what a consumer will buy, where they will go and what advertising will be most successful.

It can also be used on Earth observation data to provide actionable information for decision makers to deal with problems in health, water, agriculture, ecosystems, in other words, in all of the GEOSS SBAs.

I have worked with scientists from developing countries on a variety of projects. The scientists were highly motivated, educated and skilled, but they lacked computers and data processing tools. These tools now exist, can be used at low cost, and they are just as accessible in developing countries as they are in developed countries, using a $200 laptop computer with a modest internet connection.

I plan to organize an Architecture Implementation Pilot project with Earth observation data on Spark. The topic hasn’t been finalized. Please contact me if you’re interested, or even if just curious.

Thank you for your time and attention.