Earth Today | ‘Human in the loop’

“HUMAN SUPERVISION remains indispensable” to optimising the use of artificial intelligence for decision-making on a global scale, including on matters related to climate change.

This is according to a 2026 publication of the United Nations Environment Programme, which flags two other necessary conditions for making best use of AI, including machine-learning technologies.

“AI systems synthesise patterns based on training with historical data, but lack contextual understanding needed for decision-making. When addressing critical, site-specific environmental problems, human verification and judgment is needed before stakeholders act on AI outputs,” noted the publication titled Spotlighting Opportunity: How artificial intelligence is accelerating methane action.

This, it said, is demonstrated in the use of AI models for the Methane Alert and Response System (MARS) of UNEP’s International Methane Emissions Observatory.

MARS detects major methane emission events and notifies governments and companies to inform mitigation actions. Methane, like other greenhouse gases including carbon dioxide, drives global warming and other climate change impacts, including sea level rise and extreme weather events.

“AI model outputs are the result of statistical inferences, not conclusive judgment. When identifying methane emission sources across diverse landscapes, each case presents unique characteristics that require human judgment to interpret and act on,” the report explained.

“MARS analysts verify all AI-flagged detections, following a strict protocol that requires at least two analysts’ independent review. This hybrid human-machine system, also known as ‘human in the loop’, is designed to improve accuracy and speed, at the same time preserving scientific rigour,” it added.

Of note is that detections that analysts identify without an AI flag are not regarded as “a lapse in the system”.

Instead, they “feed directly into ongoing model refinement, helping the models improve over time”, the report said.

Another important condition is to have sustainability built into the development of the technologies.

“AI tools can help responsibly and rigorously accelerate global climate progress when sustainability is embedded by design. Through initiatives like MARS, UNEP is demonstrating how AI tools can be scaled up to address a major climate challenge. The result is greater efficiency, speed and scale amidst a rapidly evolving methane data landscape, without undercutting the climate benefits that MARS can catalyse,” the report noted.

Also important is the integration of AI technology in order to take full advantage of space-based data, as part of efforts to turn the corner on climate change.

“As new satellite missions and data collection sources become available in the coming years, the volume of space-based, environmental data continues to grow rapidly. Translating this data into climate action will depend on the development and utilisation of AI-enabled tools and data sets. AI models and AI-enhanced analysis are essential for deriving actionable intelligence from raw satellite observations,” it said.

pwr.gleaner@gmail.com