“Good evidence” is not the kind that promises a single, neat number, but one that makes its own limits visible, leaves room for different ways of knowing, and still helps societies make fair decisions in the face of uncertainties that are not going away.
Evidence-informed public policy has spread around the world with a strong promise: if knowledge is better produced, organised, and shared, public decisions will become more rational, transparent, and effective. But when climate governance in the global South enters the picture, that promise quickly runs into a basic problem: uncertainty. Climate evidence used for adaptation, risk management, and infrastructure planning is not a settled fact. It is built from projections, observations, and models, all shaped by natural climate variability, scenario choices, differences between models, and limits in spatial and temporal resolution.
International frameworks promoted by organisations such as the WHO, the Global Commission on Evidence, and Evidence-Informed Policy try to sooth the relationship between research and decision-making through ideas such as evidence ecosystems, enabling conditions, and pathways for change. But these frameworks often assume that science has already produced sufficiently solid evidence and that the main task is simply to translate it for policymakers. In Brazilian climate policy though, the challenge comes earlier: what counts as evidence when knowledge is probabilistic, operates across multiple scales, and is shaped by structural inequalities in global scientific production?
That is why applying these frameworks as they are takes more than technical adjustment. It requires epistemological and institutional tropicalisation. In practice, this means rethinking how evidence is produced, interpreted, and used in contexts where uncertainty is not a temporary obstacle but a permanent condition. Tropicalising means creating decision-making processes that can work with a range of possible futures instead of relying on definitive answers. It also means taking local knowledge and grounded observation seriously alongside climate models, while recognising that the ability to produce data and projections depends on infrastructures of power: supercomputers, satellites, protocols, and specialist expertise. Without these changes, climate policy risks importing an ideal of evidence that does not reduce asymmetries, but simply formalises them.
Probabilities and risks in climate policy
Climate change is a classic wicked problem: its boundaries are hard to define, its causes are spread out, its effects unfold across many places and times at once, as natural and social systems are tightly entangled. In this setting, scientific uncertainty is not a side issue; it is built into the problem. Policymaking cannot wait for complete certainty. It has to move through probabilities, scenarios, and risks. This is one reason climate policy looks different from areas such as health, where the scale of evidence and the scale of decision-making often fit together more neatly.
Climate modelling makes this challenge clear. Global models are designed for planetary processes, but public policy needs information that is local or sector-specific. Downscaling tries to bridge this gap, but that step brings new interpretive and computational challenges. Uncertainty also comes from several directions at once: future emissions depend on human choices; Earth systems such as oceans, ice, and the biosphere respond in complex ways; and the climate has its own internal variability, including phenomena such as El Niño and La Niña. On top of that are uncertainties about which equations to use, which processes to include, and how to capture nonlinear feedbacks. So even when models work with the same scenario they can still produce different outcomes, reflecting uncertainty inherent in the act of modelling itself.
Climate science does not remove these uncertainties; it tries to map and govern them. The Intergovernmental Panel on Climate Change (IPCC), for example, has developed a calibrated language that combines confidence and probability. In that language, evidence does not mean certainty; it means a standardised way of saying how strong the available knowledge is and how much agreement exists across studies. Yet this grammar is not neutral. It is shaped by discussion and negotiation among experts and government delegations, and it often gives more weight to quantitative data and computer models than to local knowledge, lived experience, or debates about who carries the greatest burden of climate change.
Climate uncertainty is geopolitical and infrastructural
The Brazilian case shows that uncertainty cannot be separated from scientific infrastructure. Since the 1990s, Brazil has been developing its own climate-modelling capacity at the National Institute for Space Research (INPE). This infrastructure is strategic for a country the size of a continent, vulnerable to droughts, floods, energy shocks, and food insecurity. Without it, Brazil would depend entirely on diagnoses produced in the Global North to shape its policies and negotiate within the international climate regime. The replacement of the Tupã supercomputer with Jaci in 2025, expanding INPE’s modelling resolution, illustrates how reducing some uncertainties requires major investment in supercomputing, specialised teams, and sustained institutional capacity.
Seen from this angle, climate uncertainty is not only statistical but also geopolitical and infrastructural. Few countries in the global South have supercomputers, satellites, domestic models, and full access to international intercomparison protocols. That inequality shapes who gets to produce evidence and whose standards define the ‘state of the art.’ Control over models, protocols, and infrastructure therefore shapes what is accepted as legitimate evidence in international climate policy. Brazil’s climate-modelling capacity reflects a broader investment in epistemic and political sovereignty. It is a reminder that every piece of evidence carries a faint map of the world within it, and that tropicalising evidence frameworks means making those geographies of power visible.
So tropicalising evidence frameworks is not just about translating international guidelines into local settings. It is also about disputing who defines relevant uncertainty, what enters the science-policy interface as evidence, and which forms of knowledge remain excluded. This points to the need for a multi-evidence approach, in which local observations, theory, models, and narratives can coexist and be institutionally negotiated.
To say that evidence should serve climate justice is to reverse a silent hierarchy. It is not society that should adapt to dominant forms of evidence production. Evidence itself should be judged by its capacity to reduce vulnerabilities, address inequalities, and protect those most exposed to the climate crisis. In Brazil, evidence-informed climate policies need to be technically robust, but also socially legitimate, contextually just, and grounded in territory. More than importing ready-made models, the real task is to build a politics of evidence capable of embracing uncertainty without erasing the plurality of knowledges and without reproducing the unequal global geography of expertise.
