About Artificial Intelligence and climate adaptation in fragile places

Lately I have been researching Artificial Intelligence (AI), trying to get a good understanding of its uses for climate adaptation. Specifically, I am interested to know how AI could help develop a climate action plan for a concrete location, in a fragile and conflict-affected country.

I understand AI could be particularly useful in what I consider the first part of developing a climate project proposal: the overall mapping exercise, namely on the climate and environmental data analysis, mapping climate risks, stakeholders, national policies and priorities… But I am not so certain about the second part, this is, proposing concrete context-tailored solutions based on the mapping results.

I decided to ask ChatGPT. It told me AI can propose tailored solutions, taking into account it would need to integrate local data, expert knowledge, and community input. Which is not an easy task for a human or for a machine. In these places data is not abundant, and very often it is also not updated; people in communities move continuously (due to flooding, conflict); and expert knowledge is many times limited. ChatGPT insisted that even in data-scarce environments, AI can generate insights by combining satellite imagery, remote sensing data, and historical climate patterns – still, in my opinion, that gives us an approximate idea of the situation, but not a detailed local context analysis with best adaptive options.

Some of the tasks AI says it could do:

1.     Scenario Modeling – It can simulate different future climate scenarios and suggest the best adaptation options for each, helping policymakers and communities make informed choices; again, the relevance of this would depend on the quality of the data available (always thinking about remote concrete locations).

2.     Community-Driven Solutions – This is so interesting: it says AI-powered chatbots or voice-based systems could gather feedback from local populations and refine solutions based on real-world experiences. If only people in those communities could have such a tool, it could be used for much more that climate adaptation purposes!

3.     Flood Prediction & Early Warning – analyzing satellite data, rainfall patterns, and river levels to provide more accurate flood forecasts. This type of AI application is already out there, helping local authorities and aid organizations prepare in advance.

4.     Optimized Infrastructure Planning – suggesting the best locations for flood-resistant shelters, raised roads, or drainage systems by analyzing topography and hydrology data (basically engineering)

5.     Climate-Resilient Agriculture – AI can recommend flood or drought-resistant crops and improved farming techniques tailored to context, helping communities sustain food production despite climate shocks; this could be a great help, matching available varieties with contexts.

Still, what about other non-climatic factors that are key to develop a project, like local conflict, or environmental pollution? Could AI incorporate these factors when proposing solutions? It would need multiple data sources, including conflict risk maps, pollution levels (if there is any reliable data available), and socio-economic conditions. Conflict-aware or conflict-sensitive solutions need daily-feed and updated information about movements, discussions, evolution of disputes, in brief, clear patters of local violence and displacement, to ensure that whatever adaptation efforts are proposed do not unintentionally worsen tensions (e.g., by favoring one group over another in resource allocation).

Regarding environmental pollution, this is a high politically sensitive matter. I understand that AI can process satellite imagery and sensor data to detect oil spills, water contamination, or waste buildup. But in case that such information would be available, I think there could arise interests to hide it, or to manipulate it. Again, not so easy.

So while I consider what so far I know about AI a great mean to improve climate work, I am convinced that, for now, only a human-machine mix could get climate action to succeed for the most vulnerable in fragile contexts.