Should Machine Learning Experts Respond To Climate Change Call To Action?

Republished By Plato

Followers: 0

machine learning for climate change response

Our planet’s proper functioning and survival rely on a delicate balance of a vast heterogeneity of animal, plant, and microorganism species that contribute to the ecosystem established on Earth.

Of all the organisms, there is one that has had a great impact on the planet, so great that it was capable of upsetting its balance, causing entire ecosystems to disappear and threatening its very existence: humans.

If this in-depth educational content is useful for you, subscribe to our AI research mailing list to be alerted when we release new material.

Activities such as intensive fishing have destroyed the oceans, livestock farming and our gigantic demands for meat have dramatically increased carbon dioxide emissions into the atmosphere, and have prompted ever more reckless farming using pesticides and stressful techniques that have destroyed soils halfway around the world and accelerated the phenomenon of desertification.

This has brought us to where we are today, in a society that is only now beginning to recover from one of the greatest disasters in our history, the COVID-19 pandemic, and which must inevitably prepare to face an even greater and more important challenge.

This pandemic has undoubtedly taught us many things. It has taught us that prevention is better than cure and that if we do not prepare in advance for a problem, it will inevitably overwhelm us. It has taught us that our equilibrium is fragile and that although society seems to be moving forward inexorably and unstoppably, it takes very little to bring everything to a halt and cause huge social and economic damage. And finally, it has taught us that although governments are always a little slower to react and plan for the long term, the scientific community is far from clueless and their work before the problem occurs is crucial.

The speed with which vaccines, for example, have developed is also the result of previous studies carried out by scientists in the pre-pandemic years, and their expertise was indispensable in tackling this dark chapter in our history.

But it is not only medical experts who are making a difference in our world; on the contrary, Machine Learning has proven to be a discipline as powerful as it is capable of influencing every sector of society.

When we hear the call to action to do something about climate change, many of us think that it is aimed exclusively at governments, but they will never be able to solve a problem of this magnitude without the support and commitment of technical experts. That’s where this article comes from, to encourage experts in Machine Learning, this fascinating and revolutionary discipline, to take an interest in this problem and understand that even a single person can make a difference when you have such a powerful tool in your hands.

This whole article is inspired by the work of David Rolnick et al. in “Tackling Climate Change with Machine Learning” which created a very extensive analysis of possible ML solutions to fight climate change. I really suggest you read it for a more deepen understanding of the topic.

However, some of the areas focused on in the research are inaccessible to most people and require resources and knowledge that are beyond the reach of everyone. Therefore, I will list below some possible areas and problems that can be handled by Machine Learning and for which it is possible to work even without being a big company or the most important expert in the specific field. But, as we all know, to obtain a good model we need good data, and fortunately, in recent years, many interesting datasets have been published which we can use and combine to obtain significant results.

For every topic of the article, I collected a list of useful datasets which may be a good starting point for your project. In some cases, they may be only a partial start point for you but the journey of a thousand miles begins with one step and the simple model created today on a limited dataset and the problems encountered during the development may be useful for the research of tomorrow.

Energy Management

Although not the only cause of global warming, emissions from energy production are certainly among the first things that come to mind when talking about this problem. For too long we have been tied to the use of fossil fuels and used energy unwisely, but this is where Machine Learning can help.

Machine Learning has already been successfully applied in some virtuous environments to optimize energy consumption. To date, however, these techniques are not particularly widespread, but they could nevertheless reduce energy demand and consequently CO2 emissions into the air.

Opportunities to reduce GHG emissions from electricity systems using machine learning (Source: [1])

In this area, machine learning can also be useful for making increasingly accurate energy consumption forecasts and making electricity production and distribution increasingly strategic.

Where can you start from?

Cities and Transportation

When climate change began to be discussed, much attention was immediately paid to vehicles emissions, and over time many initiatives have been taken to improve their mechanic, making them greener, to reduce their use, and today we are trying to push more and more towards electric cars.

We also tried to encourage people to use public transport to reduce the number of vehicles on the road. However, this requires careful and complex planning to understand which means of transport to use and how to organize them, and to anticipate the needs of the population.

Strategies to mitigate GHG emissions from transportation using machine learning (Source: [1])

The same optimization work can be done for the construction of buildings so that they can be built in a way that reduces their environmental impact and the amount of travel required, but Machine Learning can also be useful in improving their energy efficiency. Approaches based on Reinforcement Learning, for example, have been used for this purpose and have led to promising results.

Selected strategies to mitigate GHG emissions from buildings and cities using machine learning (Source: [1])

Where can you start from?

Industry and Waste

Sustainable industrial production is key to building an environmentally friendly society and one of the areas where Machine Learning can make a difference is in optimizing production and supply chains. The more these are optimized, the more the environmental impact of these activities will be reduced.

Not only that but the consumption of resources produced by industries and society could also be made more efficient. It is estimated that in developing countries, 40% of the food produced is wasted in harvesting and processing or retail. While in industrialized countries, 40% of the waste occurs at the end of the supply chain, in shops, restaurants, and consumers’ homes.

Great strides have also been made in predictive maintenance, which is used to predict when a plant will need to be serviced and to intervene quickly. There is also a lot of interest in this field from the industries themselves, who also see an economic advantage.

Where can you start from?

Meteorological Forecasting

In all likelihood, and we are already experiencing this, no matter how quickly we manage to reverse this trend, we will suffer at least some of the consequences of what we have done in the past years, and when this will happen, it will be essential to be ready to react.

We are already seeing an intensification of destructive natural phenomena such as hurricanes, fires, and floods, and being able to correctly identify the weather conditions that can lead to such events can literally save thousands of lives.

Fortunately, the presence of many satellites in orbit around the planet has provided us with large amounts of data that can be used to build predictive models that could already have a major impact on society today and certainly will in the near future.

In the most extreme cases, floods and droughts may make some places inhospitable, forcing populations to move, generating waves of migration never seen before. Here again, the creative use of past data could produce useful models for tomorrow’s world.

Where can you start from?

Breeding and Agriculture

Livestock farming is now a major cause of pollution and land exploitation. The intensive livestock farms that exist today to satisfy the gigantic demand for meat in developed countries are very costly, requiring the production of huge amounts of food and the consumption of water. In fact, it is estimated that 40% of the grain produced in the world is used to feed farm animals.

To meet this demand, more and more land has been used for cultivation and cultivated in an increasingly intensive and often inconsiderate manner. This approach has proven to be unsustainable. Cultivation by the plow and the use of pesticides impoverishes the soil and damages it more and more, increasing desertification.

Predicting the impact of crops on the land is, therefore, another area where machine learning can be useful, as well as enabling the use of new farming tools and improving crop organization and herd management.

Another phenomenon that plagues agriculture is microbes and diseases that end up ruining entire crops, causing economic and social damage and wasting a large number of resources. Early classification of a diseased plant can therefore be extremely helpful in fighting this phenomenon.

Strategies to mitigate GHG emissions from land use using Machine Learning (Source: [1])

Also, forest fire management, estimating the capacity of vegetation to absorb carbon dioxide, improve herd management, and precision farming, are some possible applications of Machine Learning. The possibilities in this field are endless and Machine Learning can truly be a game-changer and the key to more sustainable farming and cultivation.

Where can you start from?

Ocean and Fishing

One of the ecosystems most affected by human activity is certainly the marine ecosystem. The oceans have warmed, and although this is only a few degrees change, it has been enough to kill off a large part of the coral reefs and disrupt the entire marine ecosystem. As if that were not enough, the huge amount of waste that ends up in the sea, including a large number of fishing nets, and its pollution has already decimated an incredible number of marine species.

This is perhaps one of the most complex areas in which to apply machine learning and the scarcity of good data certainly does not help. However, one could think of models to help fish more healthily without over-stressing ecosystems, or models that can identify the first signs of marine disruption as a result of human activity.

Where can you start from?

Machine Learning Carbon Emissions

Yes, like many other human activities, Machine Learning, and in particular Deep Learning, has an environmental impact. It has been estimated by researchers at the University of Massachusetts that training a large deep-learning model produces 626,000 pounds of planet-warming carbon dioxide, comparable to the lifetime emissions of five cars.

Fortunately, there seem to be many solutions and numerous researchers are working to make this discipline more sustainable. For example, new photonics-based hardware is being developed to create photonic neural networks, which are faster to train on more energy-efficient hardware.

Conclusions

The climate emergency is a reality and while we wait for governments around the world to make their move, anyone can make a difference. This article has tried to provide some ideas and starting points for Machine Learning applications that will be useful in the near future, but the possible innovations are virtually limitless. Combining, manipulating, and analyzing this and other data from online sources and exploiting it for advanced modeling could end up with something crucial to the emergency that lies ahead.

[1] “David Rolnick et al”. “Tackling Climate Change with Machine Learning”

[2] “Srinivasan Iyengar et al”. “WattScale: A Data-driven Approach for Energy Efficiency Analytics of Buildings at Scale”

[3] “Soloplan”, “How machine learning and artificial intelligence revolutionise your transport planning”

[4] “Amirreza Farahani et al”. “Online Multimodal Transportation Planning using Deep Reinforcement Learning”

[5] “Yongyi Ran et al”. “A Survey of Predictive Maintenance: Systems, Purposes and Approaches”

[6] “Andreas Kamilaris et al”. “Deep Learning in agriculture: A survey”

[7] “Nazneen N Sultana et al”. “Reinforcement Learning for Multi-Product Multi-Node Inventory Management in Supply Chains”

[8] “Suwei Yang et al”. “Predicting Forest Fire Using Remote Sensing Data And Machine Learning”

[9] “Piyush Jain et al”. “A review of machine learning applications in wildfire science and management”

[10] “Alexandre Lacoste et al”. “Quantifying the Carbon Emissions of Machine Learning”

[11] “Lorenzo De Marinis et al”. “Photonic Neural Networks: A Survey”

This article was originally published on Towards Data Science and re-published to TOPBOTS with permission from the author.