What is Machine Learning and why do Geoscientists use it?

---

As technology develops, Machine Learning is becoming ever more integrated into our daily lives. From the voice assistant on your phone, to facial and voice recognition, to recommendations for shopping, to self-driving cars; the uses of Artificial Intelligence seem endless! But how does Machine Learning work, and how do geoscientists use it to study the planet?

What is Machine Learning?

Machine Learning is essentially where computers are ‘taught’ to carry out a task by learning from a pattern.  For example, when you are watching a video on Netflix or YouTube, the app is able suggest other videos that you might find interesting to watch. This is because a Machine Learning ‘algorithm’ (a set of rules) has been trained to understand what you would enjoy watching from the information and data of the previous videos you watched. We can apply this same Machine Learning technique to geoscience data.

Geoscientists collect all kinds of data from different areas of the Earth system. Our task is to then analyse this information and make interpretations that advance our understanding of the planet and its processes. Because we have been collecting a lot of data for a very long time, we can now train computers to analyse and interpret it for us by learning from a pattern of how we do it. Machine Learning makes our analysis much faster and can also help us to predict new outcomes.

For example, if you collect data and photos of different rock types on family holidays, you can use an algorithm to teach the computer to learn how to recognise the name, location and characteristics of the images/data of the different rock types. In future, the computer will then be able to recognise and accurately classify any new similar image from that area. This is geoscience Machine Learning!

Why is Machine Learning important for Geoscience and Society?

Geoscience is an important and well-studied field in our society, and as a result we have an enormous amount of publicly available datasets. We can apply Machine Learning techniques to analyse geoscience datasets to solve urgent problems facing our Earth and environmental systems. 

For example, Machine Learning can be used to analyse and find patterns in spatial and temporal atmospheric datasets. This can help us understand the movement and formation of climate features such as weather fronts and atmospheric rivers, detecting trends and predicting climate events like tropical cyclone and tornadoes. 

Machine Learning can be used to analyse remote sensing data to produce satellite-based estimates of surface water, aquifer vulnerability, the health of vegetation, forest cover, urban growth and land use.  These estimates can help us understand the changes happening in our environment and can be used for making management decisions and adaptation policies.

In terms of natural resources, we can use Machine Learning to improve predictions of exploration targets for oil, gas and minerals.  My current project involves combining my geoscience knowledge with Machine Learning and Artificial Intelligence techniques on a large volume of mineral datasets (satellite, geological, structural, geochemical and geophysical ) collected by geoscientists in Africa.  First, I will process and analyse the datasets. Then I will apply Machine Learning and Artificial Intelligence algorithms to a small subset of training data to find patterns in the data that match key characteristics of past mineral discoveries, allowing the computer to predict the best areas to find mineral resources. The project will help us understand the potential distribution of undiscovered mineral resources, saving time and money on exploration activities and hopefully making new discoveries that will benefit the African continent in terms of investment in the mining sector.

Geoscience knowledge combined with Machine Learning and Artificial Intelligence techniques on mineral data is a powerful tool that can be used to identify new mineral targets.

We can even use Machine Learning to accelerate our fight against the climate crisis. Machine Learning can be used in automatic monitoring of methane concentrations in the air with remote sensing, forecasting extreme climate events, generating low-carbon electricity, improving building and vehicle efficiency, and understanding personal carbon footprint with smart meters that can facilitate behavioural changes.

Get learning!

Machine Learning is gradually changing our world with great benefits to society. With abundant data available, anyone with programming language skill can train a computer to learn how to efficiently perform a wide variety of tasks that can tirelessly solve complex problems in our society.  I believe that anyone can learn and understand how Machine Learning works; you don’t need to have a background in computer science or data science to understand Machine Learning.

You can use this free online tool to learn how to carry out some Machine Learning exercises: https://machinelearningforkids.co.uk/