The Earth Lab Analytics Hub helps scientists use Earth data for scientific discovery. The hub works closely with scientists to develop data and software packages that are available to the public. The team includes experts in remote sensing, programming, statistics, natural language processing, computing, and machine learning.

Machine Learning in Earth Analytics

Machine learning is in essence teaching a computer how to learn through experience - similar to how humans learn. The computer tests how two or more things are related. It then uses these relationships to create a model that makes decisions and predictions. It works much like how you would learn in school: the computer performs a task, reports the answer, gets feedback on the answer, and makes adjustments to do better in the future.

 

Schematic on how to label 3D Point Clouds. Source: Huang and You.

In more technical terms, machine learning describes a family of algorithms that can establish relationships between input datasets and (possibly labeled) outputs. Machine learning algorithms are useful for modelling complex relationships where there is lots of data available to train the computer. For example, machine learning can be used with remote sensing data to calculate vegetation cover, biomass or to perform land cover classification.

Earth Lab uses machine learning to combine data from remote sensing, social media, sensors, and other data sources in novel ways. Through this, Earth Lab uncovers environmental and social trends,produces innovative data products, and ultimately helps address pressing environmental problems.

Using Remote Sensing Data for Science

Remote sensing data can come from satellites, drones, and airplanes. Scientists use this data to understand how our environment is changing. Analyzing remote sensing data can be time consuming and require specialized computing resources. The Analytics Hub develops openly available software tools that make it easier for anyone to work with remote sensing data. The hub investigates how remote sensing data can be used to do things like detect, predict and measure the impact of wildfires, measure how individual trees respond to drought, or distinguish between grass and shrubs in city landscapes.

Earth Data Analytics in the Cloud

As satellite technology improves and more satellites are launched into space  and computing technology improves, more earth data is becoming available every day. As data sets get larger and more complex, the typical computer no longer has enough power and memory to effectively process earth data. Cloud computing allows scientists to get more computing power without investing in a high performance computer. The Analytics Hub is helping train scientists and developing new software tools to access powerful computing in the cloud.

The Analytics Hub helps researchers  focus on science rather than operating systems and computing technologies.

The hub provides Earth Lab researchers with access to data storage and computing in the cloud so they can work with large, complex data sets and compute-intensive analytics. You can also check out some of the hub's earth analytics Docker containers here.