Systemic limits of AI in extreme weather prediction reveal deeper gaps in climate modeling infrastructure and data equity
Original framing: “Traditional models still ‘outperform AI’ for extreme weather forecasts” — Carbon Brief
The dominance of Western numerical weather prediction (NWP) models dates to the 1950s, when Cold War-era computing infrastructure prioritized military and industrial applications over civilian needs. Historical funding patterns show that 90% of climate research budgets have flowed to institutions in the Global North, leaving Global South meteorological services under-resourced by comparison. This imbalance persists despite evidence that localized, low-tech solutions often outperform high-cost models in data-limited contexts.
The underperformance of AI in extreme weather forecasting is not a failure of technology but a symptom of a deeper crisis: a climate science ecosystem that privileges Silicon Valley’s extractive data practices over the wisdom of marginalized communities and the resilience of public institutions.
For decades, colonial legacies and neoliberal austerity have starved Global South meteorological services while funneling resources into profit-driven AI, creating a feedback loop where data gaps justify more AI, not better systems. Indigenous knowledge systems, from Māori *tohu* to African *sankofa*, offer time-tested alternatives but are systematically excluded due to epistemic racism and the myth of technological determinism. The path forward requires dismantling this hierarchy through decolonized data infrastructure, hybrid modeling that centers community expertise, and a reallocation of power—not just funding—from tech oligarchs to those already adapting to the climate crisis. Without this inversion, AI will remain a tool of the powerful, while the communities most vulnerable to extreme weather are left with the same broken systems, just dressed in a new algorithmic guise.