Introduction: Artificial Intelligence and Climate Change
Artificial Intelligence and Climate Change: Hardly any day goes by without both climate change and artificial intelligence (AI) making headlines. With evidence of climate change mounting up worldwide and AI becoming ever more powerful in different sectors, it is worth considering the relationship between those two fields.
Leading scientists consider climate change to be the biggest challenge facing our planet right now. Their fears are backed up by numbers. Columbia University reports more than 770 weather and disaster events for the year of 2016, three times as many as the earth experienced in 1980. The scientists also cite the growing numbers of extinction of various species as an example of climate change.
Arguably, the most commonly known sign of climate change is global warming. By 2100, average temperatures are likely to be 3°C higher than they are now. This is only one of the reasons why climate activists continue to urge governments to act faster and more decisively to stem the tide of climate change.
The World Economic Forum (WEF) identified six areas in which our planet is facing critical challenges:
- Global warming
- Ocean chemistry
- Water security
- Air quality
- Disaster and weather resilience
Artificial intelligence technology (AI) is offering potential solutions to some of these challenges. In short, AI refers to computer systems that have learned to sense their environment. They can think, learn more, and act by combining what they have sensed with predetermined objectives. According to the authors of the WEF report, AI is quickly becoming “the “electricity” of the Fourth Industrial Revolution.”
Aside from the potential of AI, there are some concerns when it comes to the relationship between AI and climate change. Information technology, which is at the heart of AI, is one of the biggest contributors to climate change with a sizable carbon footprint. For example, training one algorithm for natural language processing may emit an amount of carbon dioxide that is equivalent to five times the lifetime emissions of a car.
With that said, AI has huge potential to help address climate change challenges. As a positive force, artificial intelligence technology can improve a population’s resilience to hazards created by climate change. AI may also help measure and reduce emissions and help develop more innovative business models to tackle global warming and other problems.
AI to Create Climate Zones
Humans have referred to climate zones for nearly 140 years to predict weather patterns and better understand changes happening to the planet. The Koppen climate classification was introduced in 1884 and refined in the 1960s. This process produced the Trewartha climate classification. Both are based on prior rules, such as average temperature and precipitation.
Using AI to define and perhaps rearrange existing climate zones would involve another approach. Machine learning, one aspect of AI, allows scientists to let the data speak for itself rather than following pre-set rules.
By using this approach, computer systems quickly evaluate a large amount of data to combine locations with similar traits. The goal is to group locations that are as similar as possible into one climate zone. Individual climate zones, on the other hand, would show characteristics that clearly differentiate them from one another.
Because AI-derived climate zones are based on real-world data rather than predefined factors, they have the potential to allow more accurate predictions. One such example is the climate model developed by the National Center for Atmospheric Research (NCAR). The model aims to project both precipitation and temperatures up to 2050, following a variety of scenarios.
AI technology allows researchers to create and compare different scenarios within minutes. The NCAR research projects changes based on the assumption that greenhouse gases are aggressively reduced worldwide. The model then compares those predictions to what may happen if the world continues to operate as usual.
Compared to traditional climate zones, AI-based predictions are rooted in actual data. By putting different outcomes side-by-side, researchers can make a more compelling argument for the need to change our daily habits.
The scientists at the Mercator Research Institute on Global Commons and Climate Change in Germany have harnessed AI’s data processing power for attribution studies. Analyzing vast quantities of data has allowed the researchers to prove that the impact of climate change can already be felt almost universally around the world.
Their research highlighted changes in the climate of a specific location and then identified the driver behind the changes. The results have been published as a map, and the authors believe it can provide “orientation for the global fight against global warming.”
At this stage, however, the map is far from complete. Data is missing for some of the countries that have been affected most directly by climate change. Several low-income countries simply do not have the amount of depth of research compared to their more developed counterparts.
AI for Planting Native Trees
The planting of native trees is one promising application for artificial intelligence technology. At a time when deforestation is considered one of the biggest contributors to greenhouse gas emissions, this could be a game changer.
In New Zealand, government-owned scientific research institute Scion teamed up with remote sensing specialists Indufor to improve native biodiversity. Together, they developed a model that allows scientists to monitor the seedlings of radiata pine trees. Remote sensing identifies these seedlings and other pockets of native forest. Scientists use the results to better understand indigenous biodiversity on a site. This most recent deep learning model requires less data, and the imagery is cheaper than previous options. AI is making technology more accessible and affordable for projects with limited budgets.
Indufor’s technology helps foresters protect indigenous species growing on their land and integrate them with planned developments. Supporting and preserving these forests provides vital habitats for native birds and other species. By visualizing where these pockets exist, scientists can assess whether indigenous forest areas are sufficient to support these species. If the areas are too small, foresters can look into steps to join them together.
Combined with drones, AI is transforming reforestation. The World Wildlife Fund (WWF) and other sources estimate that the world is losing the equivalent of up to one football field of forest every second. It is far more than what humans could realistically replant by hand. Plus, planting by hand is expensive and time-consuming.
British company Dendra has developed drone technology that can plant up to 120 seedpods per minute. The company has pledged to plant 500 billion trees by 2060. Drones make it easier to re-plant in places that are hard to reach, and they can distribute more seedlings faster than workers could. Data analysis is used to identify planting locations.
Technology like this offers an opportunity for reforestation to match the rate of deforestation or at least close the gap between the two.
AI to Detect Rainfall
Weather forecasts have a notorious (and often undeserved) reputation for being inaccurate. Conventional forecasting technology is limited, especially when it comes to short-term predictions.
With that in mind, Google-owned Deep Mind worked with the University of Exeter in the UK and the British Met Office to create a better alternative. Their “Precipitation Nowcasting” model is capable of predicting average or heavy rainfall over the coming 90 minutes. Being able to forecast with this degree of accuracy can help government agencies prepare for weather emergencies.
The model may also help with outdoor event planning, air traffic and ground traffic control as well as water management. Once again, AI outperforms traditional technology because of its ability to process large quantities of data fast.
Deep Mind uses high-resolution radar data to measure the humidity in the air and the speed at which water vapor is moving. A machine learning algorithm then uses this data to predict rainfall.
Despite having been the subject of countless jokes, traditional weather forecasts are rather accurate in the medium to long term. However, predicting precipitation in the very near future has proven challenging. This is where AI can close a gap.
AI to Detect Ground Water
The World Economic Forum’s report on harnessing the potential of AI for the earth identified water security as a major challenge. The paper stated that as early as 2030, the world may fall 40% short of the amount of freshwater required to support the global economy.
This is where AI technology can be invaluable both in predicting groundwater availability and in assessing water quality.
South African scientists have successfully used a machine learning approach to help mitigate drought and ensure a sustainable supply of water. AI and data analysis allowed researchers to understand what happened to an area’s water supply when the flow of rivers declined. Insights based on real data helped them see the stages leading up to drought.
These results are already proving useful for the evaluation of groundwater sources of drinking water and their protection. They will enable water resource managers to create groundwater monitoring programs that are targeted to their area. Plus, the amount of data analyzed by AI will even let them anticipate the water quality in areas that have not yet been sampled.
In India, AI-based models are being used in the Ganges river delta to assess whether groundwater has been exposed to arsenic pollution. In this case, researchers from the Indian Institute of Technology in Kharagpur combined traditional, statistical research methods with artificial intelligence techniques.
Another part of their research investigated and evaluated the health of people who had already been exposed to arsenic. By analyzing vast quantities of data, the researchers identified the worst affected regions. They also learned that groundwater-fed irrigation was the leading factor in causing arsenic hazards.
Their findings have been transposed onto a map to visualize where interventions are needed most urgently. The same results are also the baseline for India’s commitment to providing safe drinking water to every family by 2024.
Similar projects are underway in Malaysia, where a group of researchers used intelligent calculating tools to simulate the groundwater level in the Langat Basin area over seven years. The Malaysian scientists measured rainfall, humidity, minimum and maximum temperature as well as evaporation. These parameters were input into different models, whose accuracy over time the scientists compared.
Whilst the Langat Basin research compared the accuracy of two different AI techniques. Another recent study compared four different techniques to identify the one most suited to groundwater quality forecasting. Despite different accuracy delivered by different approaches, one thing is clear: AI is already likely to make better decisions under critical conditions than humans can.
Artificial intelligence technology has outstanding potential to help fight the consequences of climate change. Whilst having a large carbon footprint itself, AI applications and techniques also have the power to improve climate change outcomes.
Groundwater detection and monitoring, reforestation, rainfall prediction, and mapping of climate zones are four prime examples of AI’s potential. Technology may not replace humans when it comes to climate change research. But its capacity to analyze vast quantities of data and facilitate scenario planning faster than humans could make this technology invaluable.