6.2 Climate Change: Causes and Impacts

Data collected over time by weather stations, observatories, radar and satellites allows the study of both climate change and land-use change. Long-term data sets record temperature and greenhouse gas concentrations directly. Where direct records do not reach, indirect (proxy) measurements fill the gap: isotope measurements from ice cores, dendrochronology (tree-ring dating), and pollen taken from peat cores all preserve a usable record of past conditions. Both direct and indirect measurements feed into building climate models.

Global climate models manipulate inputs to climate systems to predict possible outputs, using equations to represent the physical processes and interactions that drive Earth's climate: solar radiation, ocean and atmospheric circulation, cloud formation, ice-albedo feedback, and the carbon cycle are all represented mathematically and run forward in time on powerful supercomputers, often simulating the whole planet at grid resolutions of tens to a few hundred kilometers. Because so many interacting processes are involved, even small differences in how a model represents one process, cloud behavior is a well-known example, can produce meaningfully different projected outcomes between models.

Model validity can be tested through hindcasting: running the model backward from a past starting point, feeding it only the inputs that were actually true at that time, and checking whether its output matches what is already independently known to have happened since. A model that successfully hindcasts the 20th century's observed warming trend, including the timing and rough magnitude of that warming, is treated as more credible for projecting the 21st century forward. There is still real uncertainty in the inputs, including the use of proxy data for periods before direct instrumental records, and this input uncertainty propagates through to the outputs, which is exactly why models are reported as a range of possible future outcomes rather than one single number.

Climate models use different scenarios, built from different assumptions about future variables such as population growth, economic development and emission rates, to predict possible impacts of climate change: sea-level rise, local temperature change and precipitation patterns can all be modeled under a range of plausible futures rather than one falsely precise number. The IPCC's current framework uses five Shared Socioeconomic Pathways (SSPs, covered in full on 6.3's HL Extension tab), ranging from a very low-emissions pathway broadly consistent with limiting warming to around 1.5°C, to a very high-emissions pathway assuming continued heavy fossil fuel use. Projected global warming by 2100, relative to the 1850-1900 pre-industrial baseline, differs enormously between these pathways, from roughly 1-1.8°C under the lowest-emissions scenario to roughly 3.3-5.7°C under the highest, illustrating just how much of the future still depends on choices made now rather than on physics already locked in.

Presenting a range, rather than a single prediction, is a more honest reflection of real scientific uncertainty about how the world will actually behave: it separates the physical uncertainty in the climate system itself from the social and political uncertainty in which pathway humanity actually follows, and it lets policymakers see the difference in outcomes between taking strong action and taking little or none.

Climate models show the Earth may approach a critical threshold, shifting to a new equilibrium; local systems have their own thresholds, or tipping points, too. Global critical thresholds may be rapid, unanticipated and potentially catastrophic, often driven by positive feedback loops that push a system past the point of return.

Individual tipping points of the climate system may interact to create tipping cascades, chains of change in which crossing one threshold shifts the conditions that determine whether a separate, sometimes distant, system also crosses its own threshold. A concrete example: accelerated Arctic sea ice loss reduces surface albedo and warms the Arctic faster than the global average, a process called Arctic amplification. That extra regional warming accelerates permafrost thaw across Siberia and northern Canada, and thawing permafrost releases stored methane and CO₂ that had been locked in frozen organic matter for millennia, adding further warming that feeds back into still more ice loss. Neither the Arctic ice threshold nor the permafrost threshold operates in isolation; each shift changes the odds facing the other.

These interactions can be biotic, abiotic, or a combination of both, and because a real climate system contains many potential tipping points, from the AMOC to the Amazon-Cerrado transition covered in the previous stage, and scientists cannot yet fully model how they interact, tipping cascades make predicting the overall scale and pace of climate change considerably more uncertain than analyzing any single tipping point in isolation. This uncertainty is itself an argument for caution, not complacency: an outcome being hard to predict precisely is not evidence that it is unlikely to be severe.

Countries vary enormously in both their responsibility for climate change and their vulnerability to it, and the least responsible are often the most vulnerable. Responsibility can be measured multiple ways: current annual emission rates, cumulative totals since the Industrial Revolution, or emissions per person, and each metric tells a different story about who is truly "responsible." China leads on current annual emissions, but the United States accounts for roughly a quarter of all cumulative CO₂ released since 1850 despite holding around 4% of the world's population, because it industrialized more than a century earlier and emitted at high volumes throughout that entire history. Which metric a country's negotiators emphasize is rarely neutral: a historically industrialized nation tends to point to current annual rates, where fast-growing economies now emit heavily, while a rapidly developing nation tends to point to cumulative or per-person figures, where the historical burden sits overwhelmingly with early industrializers. This is exactly why "common but differentiated responsibilities," the principle written into the UNFCCC in 1992, exists: it formally accepts that all countries share a duty to act, but not an equal one, since the accumulated warming already locked into the system was put there disproportionately by a small number of long-industrialized economies.

Pakistan's 2022 floods, covered on the Learn tab, are a real example of this mismatch: a country responsible for under 1% of historical global emissions absorbed tens of billions of dollars in climate-linked disaster losses. This mismatch between responsibility and vulnerability is the core of the climate justice argument, and it carries political and economic implications for how international climate finance and adaptation funding get allocated. In practice, the argument plays out in negotiations over who pays for adaptation and loss and damage: high-vulnerability, low-emission countries argue that historically high-emitting nations have an obligation to fund the sea walls, early-warning systems and crop resilience programs they now need, since that need was created by emissions they did not produce. Wealthier historic emitters, in turn, face domestic political pressure over the scale of finance commitments, which is why climate finance targets are negotiated and renegotiated at nearly every Conference of the Parties, most recently at COP30 in Belem, rather than settled once. The 2015 Paris Agreement itself embeds this equity tension structurally: it lets each country set its own Nationally Determined Contribution rather than imposing uniform targets, precisely because a single emissions-cut requirement applied equally to a country that industrialized in the 1800s and one still building basic infrastructure today would ignore how differently each got to the present.