Landscaping the use of AI for Climate Change: Addressing Challenges and Risks


Incubated by the Boston Consulting Group, CO2 AI uses AI technologies to help large corporates track, measure and reduce their carbon footprint. [1] Israel-based Albo Climate has combined use of geospatial technology and AI tools to monitor carbon and track sequestration in nature-based projects. [2 ] And the United Nations Environment Programme’s World Environment Situation Room offers an AI-enabled real-time analysis of sensor data collected on atmospheric CO2 concentration, glacier mass, and seal levels. [3] Such technologies can support climate research and enhance our ability to predict changes and create simulations, and inform design strategies to make architecture, buildings and designs more sustainable and climate-friendly.[4]

How can we use this technology in India? The Ministry of Earth Sciences’ ACROSS program has plans to deploy AI to build weather models and visualization systems, and to create more accurate weather and natural disaster forecasting systems. But there is more to be done. Government agencies hold over 125 years’ worth of meteorological, geospatial, biological and astronomical data.[5] In the absence of enabling policies and a prudent appreciation of existing challenges, we are massively underutilizing the potential of AI and ML.[6]

Policy recognition of AI and climate change in India

Key decision makers in India have recognized the possibilities of using AI in climate change across a number of statements. In 2022, Union Minister Dr. Jitendra Singh said that AI will be a key pillar in a Science Mission on Climate Research.[7] Union Minister of Finance, Ms Nirmala Sitharaman, has also acknowledged the role of start-ups in developing AI-enabled solutions addressing climate change.[8]

In 2021, the Science and Engineering Research Board (SERB) called for applications from institutions to set up centres of excellence (CoEs) for the application of AI ad ML models in improving weather and climate geohazard predictions.[9] The next step now is for this political recognition to translate into policy implementation and strategy.

Challenges to deploying AI for climate change in India

Any strategic thinking and policy making on AI in the climate change context must take cognizance of the following challenges:

a. Availability and use of data

Quality data is central to AI models’ capability. A robust data infrastructure, i.e., well populated datasets, servers and data centres, smooth data sharing mechanisms, reliable data governance, and definitive data standards build a foundation for a strong AI-enabled predictive model.[10] However, the data relevant to AI’s climate change applications is often not available or is improperly collected and maintained.[11] For AI systems to use data, datasets often have to be cleaned, labelled, and sorted – which involves time and costs. This may discourage smaller companies from enhancing their work in AI-driven climate change research and innovation.

Geographical diversity in climate data directly impacts the effectiveness of the AI models that can be built from it. For instance, it may be easier to collect climate-relevant data in developed countries and urban areas compared to under-developed regions. But it is often under-developed regions that face an exacerbated risk of climate change. A concentration of data collection practices in developed / urban areas leads to inadequate representation for other regions. This ultimately reduces the usefulness and viability of AI as a large-scale solution for climate research (particularly for climate danger zones). Inadequate representation can also increase the risk of harm from using AI for under-represented populations and regions.[12]

Where data is available, it should be accessible for use. The NITI Aayog, for instance, noted that the government ministries hold large swathes of data (like climate data, remote sensing data, soil health data, etc.), but in silos.[13] It has recommended that data be made accessible through inter-ministerial and inter-departmental data-sharing mechanisms, along with adequate privacy safeguards.[14]

b. The concentration of technological development in the Global North

The persistent technology divide between the Global North and the Global South[15] gives rise to two scenarios for the deployment of AI-driven climate technologies. First, solutions developed and deployed in the Global North do not reach the Global South, and second, solutions designed in the context of the Global North are deployed in the Global South without adequate adjustment.[16]

In the first scenario, concentrated technological development in the Global North may exclude at-risk regions from beneficial research and development activities like modelling, climate prediction, etc. At-risk areas of the Global South may also see delayed adoption of AI solutions. In the second scenario, if AI solutions are not contextualized locally, outcomes may be counterproductive. This is particularly critical because holistic climate change mitigation solutions require cooperation and inputs from indigenous communities.[17] Prediction models developed outside local social, economic and political contexts are unlikely to solve complex challenges of climate change. For example, a prediction model developed and trained in Canada may produce unusable outcomes if deployed as-is in Maldives.

c. Emissions from data centres

Typically, AI and machine learning models rely on heavy computing power, which utilizes vast amounts of energy. This creates large carbon footprints. For instance, data farms in the EU and Canada reportedly use as much energy as small cities.[18] According to a study by the University of Massachusetts, training a large-scale AI model for human language tools can emit about 300,000 kg of carbon dioxide equivalent – this is about 5 times the emissions of an average car’s manufacture and usage).[19] It is unclear if improving hardware architecture and energy efficiencies can keep up with the increasing requirement of computing power. Regulatory frameworks need to introduce mechanisms to measure the climate impact of such technologies.

Policy recommendations

Policy approaches going forward should be cognizant of emerging challenges to deploying AI for climate change applications. The following considerations may be relevant to alleviate the challenges discussed above:

a. Standards and protocols for data collection and sharing

First, the work that is already underway — in December 2022, the government rolled out the National Geospatial Policy 2022.24 It aims to create the National Geospatial Data Registry and ease the private sector’s access to geospatial data.[20] In 2022, the government also published a new draft of the National Data Governance Framework Policy – which aims to promote sharing of government-held data and ease the private sectors and startups’ access to government-held data.[21] In addition, the upcoming Digital India Act may govern non-personal data (NPD) and anonymised data, which will further dictate the use of climate data in AI applications.

Once these bases are covered, there is more that can be built on it. In addition to enabling policies for data sharing, there is also a need to create standards for data quality and protocols for data sharing.[22] This includes standards for data storage and data distribution. There is also a need to enable equitable data collection from different regions and populations. This can be done through government-backed programs for data collection. Government-backed incentive programs can promote private sector participation in collection, sorting and distributing climate-relevant datasets. For example, in the US, the National Centers for Environmental Information maintains a portal for data accessibility – which covers data related to climate data records, coastal indicators, geomagnetism, and others.[23] Similarly, in the UK, the University of East Anglia maintains a Climate Research Unit- which offers access to datasets under the Open Database License.[24]

The government may also consider establishing data taskforces in critical sectors relevant to climate change – like energy, transportation, manufacturing, and others.[25] This can be modelled on the UK’s Energy Data Task Force that has been tasked with integrating data to further UK’s climate change efforts.[26] These taskforces can be in charge of setting standards for data necessary for climate action, propose to promote data collection in the industry, and create data sharing mechanisms on a fair, reasonable and non-discriminatory (FRAND) basis. Voluntary, compensation-linked data sharing mechanisms can encourage private sector participation and data transfer across sectors.

b. Mechanisms for national, and international knowledge transfer across sectors

Technology, knowledge transfer, and diffusion are key to ensuring that their benefits reach the most vulnerable populations.[27]There is acknowledgement for this at the multilateral level. The Paris agreement codified international cooperation for climate technology development, transfer and capacity building.[28]  In 2022, India and the EU signed an Intent of Cooperation on High-Performance Computing, Weather Extremes and Climate Modeling and Quantum Technologies.[29] Under the agreement, India and the EU intend to set up the EU-India Trade and Technology Council (TTC) to focus on developing advanced technology solutions.[30] In December 2022, the Indo-US Science and Technology Forum discussed avenues for India and the US to work together on emerging technologies for sustainable development, including artificial intelligence, machine learning, etc.[31] Strategic cooperation with countries like the US – which ranked first in climate technology investment in 2021 – can help India accelerate its ability to meet its climate action plan.[32]Such knowledge transfer between governments and the private sector can be further facilitated through combined funding for research institutions and knowledge centres, and joint workshops for policymakers on key issues like cross border transfers of climate-relevant data, data sharing mechanisms, and data standards .[33] Further, a centralized coordination centre hosted by the government could act as an incubator for start-ups, platform for exchanging research & development where appropriate and bringing stakeholders together.[34]

c. Incentives for AI-driven climate innovations

Going forward, it will be important to explicitly recognize climate change applications as a public interest use case in AI regulation and policies. This can create a foundation for the government to create incentive programmes, and promote AI-driven climate innovation. The government can also explore fiscal incentives for AI-driven climate innovation.[35] This can include Production Linked Incentives, grants to universities and companies for Research Development, and funding consortiums for global exchange and cooperative research.

With a growing global focus on operationalizing AI for public welfare, policy frameworks will need to catch up. Governments should approach AI policies with a tech-first, inclusive, and multistakeholder approach to push AI adoption and innovation for climate change.

d. Reducing the adverse impacts of AI on climate change

AI driven technologies themselves could have an adverse impact on the climate due to emissions generated by data centres. Therefore, policies must be devised to mandate the tracking and appropriate disclosure of statistics on Green House Gas (GHG) emissions and energy consumption by AI solutions aiming to counter climate change.[36]

This post has been authored by Ikigai team

Image credits: Adobe

For more on the topic please reach out to us at












[12] AI and the Global South: Designing for Other Worlds – Chinmayi Arun

[13], page 111.

[14], page 111.

[15] AI and the Global South: Designing for Other Worlds – Chinmayi Arun

[16] Ibid












[28], page 27. See also,

[29] See,





[34] Can be modelled on the private sector equivalent here:,accelerate%20action%20on%20climate%20change.



the status quo

Dividing by zero...