Image for post
Image for post
Photo by @korpa Unsplash

Adaptation to Climate Change with Machine Learning

Climate Prediction, Societal Impacts and Solar Geoengineering

Yesterday I had a look at the topic Mitigation of Climate Change with Machine Learning, this was an excerpt from the paper posted on arXiv called Tackling Climate Change with Machine Learning published the 10th of June. I decided I want to explore this paper further and thus look into the second way to tackle climate change: adaptation.

The paper starts by outlining a few known facts within the area of climate change with rising emissions that show no trend of decreasing. Then it rolls into the idea of mitigation (reducing emissions) and adaptation (preparing for unavoidable consequences). These diverse problems can be seen as an opportunity to have an impact.

Adaptation to Climate Change

The section within the paper dedicated to adaptation is shorter than mitigation. In my mind it makes sense to focus more on mitigation than adaptation, due to the need to truly work to reverse the trend that is ongoing (the carbon emissions are not pointing downwards). Yet on the other hand we find ourselves in a situation where climate change has gone so far to the point that it will happen to a large degree: the question is simply how much will it change? Towards this point there are three different approaches that the paper proposes: (1) climate prediction; (2) societal impact; (3) solar geoengineering.

According to the paper it seems the first global warming prediction was made in 1896, when Arrhenius estimated that burning fossil fuels could eventually release enough CO2 to warm the Earth by 5 ◦C. That is surely a while ago, however it was not clear at the time how accurate this prediction was. Apparently the predominant predictive tools are climate models, known as general circulation models (GCMs) or Earth system models (ESMs). There has become a growing opportunity for ML techniques to contribute to creating climate models, and why is this relevant now?

  1. New and cheaper satellites are creating petabytes of climate observation data.
  2. Massive climate modeling projects are generating petabytes of simulated climate data.
  3. Climate forecasts are computationally expensive (a simulations could take three weeks to run on NCAR supercomputers), but ML applications are driving the design of next-generation supercomputers that could ease current computational bottlenecks.

It is suggested work can be done through uniting data, ML and climate science as well as forecasting extreme events. There is a mention of new models needing to leverage existing knowledge. In ten years, the written statement hopes that in the future there will be: “…more satellite data, more interpretable ML techniques, hopefully more trust from the scientific community, and possibly a new climate model written in Julia.”

As such Julia mentioned here is a programming language. You can read far more on their website (or their JuliaCon on 22nd-26th of Julia). However since it is not described in the paper I thought it may be useful to give a general description:

Julia is a high-level general-purpose dynamic programming language designed for high-performance numerical analysis and computational science. It is also useful for low-level systems programming, as a specification language, and for web programming: both for server web use and for web client programming.”

Well, that was a side note. In the paper it is said that: “…climate scientists have recently begun to explore ML techniques, and are starting to team up with computer scientists to build new and exciting applications.” The examples claim to show that many components of large climate models can be replaced with ML models at lower computational costs.

It is not only the atmosphere and the weather that is changing. Due to these rapid changes there will be prolonged ecological and socioeconomic stresses. There is a mention too of possible severe, societal disruptions. Crop yield and food shortages is certainly not impossible. This section starts by raising two questions:

  • How do we reduce vulnerability to climate impacts?
  • How do we support rapid recovery from climate-induced disruptions?

In the last article I wrote about mitigation and there were plenty og suggestions in regards to energy, transportation and construction. Yet if a sever disruption does happen according to the text there is a need to be:

  • Sounding alarms: Identifying and prioritizing the areas of highest risk, by using evidence of risk from historical data.
  • Providing annotation: Extracting actionable information or labels from unstructured raw data.
  • Promoting exchange: Making it easier to share resources and information to pool and reduce risk

The suggest ecosystem monitoring and biodiversity monitoring for the ecology. Design (adapted to climate) and maintenance (what can function well under stress) for the infrastructure. In terms of social system there is a need to think of: food security, resilient livelihoods and migration. In a crisis there can be epidemics and there will be a need for disaster relief.

Outside of the context of this article, but related, I did write about how Epigram uses cognitive machine learning to help protect and do research on wildlife. Additionally I did an interview with Morten Goodwin on how AI can be used during emergency situations. As such certain actors in the field of AI are already moving to address this issue of societal impact.

Perhaps the most controversial point for me in this report is solar geoengineering, perhaps because it seems so foreign. The idea is to shift the balance between how much heat the Earth absorbs and how much it releases. It has been described as a last-ditch option, but this paper however seems more optimistic:

“Airships floating through the sky, spraying aerosols; robotic boats crisscrossing the ocean, firing vertical jets of spray; arrays of mirrors carefully positioned in space, micro-adjusted by remote control: these images seem like science fiction, but they are actually real proposals for solar radiation management, commonly called solar geoengineering”

Elsewhere it is argued geoengineering could threaten crop yields by reducing crops’ access to sunlight, and that it does not address ocean acidification, a significant environmental threat associated with climate change. I think it is healthy to be sceptical of those claiming minimal side effects. Thankfully in the paper I am looking at there is a mention of a sceptical outlook:

  • Potential side effects and governance challenges.
  • It cannot simply reverse the effects of climate change
  • Comes with the risk of termination shock (fast, catastrophic warming if humanity undertakes solar geoengineering but stops suddenly)

While presenting these doubts and that the hardest problems are ‘non-technical’, this section still takes a technical approach on how machine learning techniques can be used. First it looks at primary candidate methods in geoengineering:

  1. Marine cloud brightening, making low-lying clouds more reflective.
  2. Cirrus thinning, making high-flying clouds trap less heat.
  3. Stratospheric aerosol injection (which the paper has a focus on)
  4. “white-roof” methods (although this seems debunked or questionable)
  5. Launching sunshades into space

What is considered a leading candidate both because of its technological and economic feasibility is injecting sulfate aerosols into the stratosphere. Another reason strongly emphasised is ‘we have data’. Specifically volcanic eruption data. Once injected into the air, sulfates circulate globally and remain aloft for 1 to 2 years.

Thus there is a proposition of: (1) designing aerosols; (2) aerosol modelling; (3) engineering a planetary control system; and (4) impact modelling.

Summarised there was proposed three points for adaptation (1) climate prediction, (2) societal impact, and (3) solar geoengineering. The two first seem like far more viable options than the last. Using machine learning or the field of artificial intelligence to tackle climate change is already important and could be a focus for parts of the AI research community and startups within the field of AI.

Thank you for reading. This is day 57 of #500daysofAI. I write one new article every day on the topic of artificial intelligence.

Written by

AI Policy and Ethics at Student at University of Copenhagen MSc in Social Data Science. All views are my own.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store