In a CMIP6 temperature projection, the spread between different SSP scenario lines widens dramatically after 2050, while before 2030 the lines are nearly indistinguishable. What does this tell us about the sources of uncertainty?
AClimate models are reliable only after 2050 when they have enough observational data to calibrate
BEarly-century uncertainty is dominated by internal variability and model spread; late-century uncertainty is dominated by which emissions pathway humanity follows
CThe models agree perfectly on long-term projections, which is why the scenario lines converge at the end
DScenario uncertainty is irrelevant because the models all converge on the same warming by 2100
This is the key structure of climate projection uncertainty. Before mid-century, natural variability and differences in how models parameterize processes (model uncertainty) dominate the spread. After mid-century, the choice of emissions scenario becomes the largest factor — because that depends on human decisions, not physics. The fan of scenario lines diverging after 2040 visually represents the point where human choices matter more than physical uncertainty we cannot reduce.
Question 2 Multiple Choice
A commentator argues: 'Climate models can't be trusted because different models predict different amounts of warming.' Which response best addresses this claim?
AThe commentator is correct; model disagreement proves we cannot know anything about future climate
BAll models must agree before projections can be used in policy
CModel uncertainty exists about the precise amount of warming, but all models agree on the direction and approximate range; uncertainty is about precision, not whether warming occurs
DClimate models are as accurate as short-range weather forecasts and should be trusted completely
Model spread reflects genuine scientific uncertainty about parameterized processes like cloud microphysics — not uncertainty about whether warming will occur. All CMIP6 models agree that continued fossil fuel use causes significant warming; they disagree on the exact sensitivity. This is analogous to multiple doctors agreeing a patient's condition will worsen while disagreeing on the exact timeline. Uncertainty about amount ≠ uncertainty about direction.
Question 3 True / False
Climate model projections are best understood as conditional statements: if emissions follow a given trajectory, then here is the resulting climate — not as unconditional predictions of what will happen.
TTrue
FFalse
Answer: True
SSP scenarios are explicitly 'what if' storylines paired with radiative forcing levels. No single projection is a prediction of what will happen — each is a conditional outcome contingent on the emissions pathway chosen. This framing is important because it reframes uncertainty productively: rather than 'we don't know what the climate will do,' it becomes 'here are the climate consequences of each policy choice.'
Question 4 True / False
Higher-resolution climate models usually produce more accurate global temperature projections than lower-resolution models.
TTrue
FFalse
Answer: False
Higher resolution improves representation of regional features and some mesoscale processes, but it does not automatically improve large-scale global temperature projections. Higher-resolution models are computationally expensive, which limits the number of ensemble runs possible, and they can still miss key feedbacks if sub-grid processes are not well parameterized. The quality of parameterization schemes matters more than resolution alone for global-scale projections. This is why the CMIP ensemble includes models across a range of resolutions.
Question 5 Short Answer
What is parameterization in climate modeling, and why does it introduce uncertainty even in models that correctly implement the governing equations of physics?
Think about your answer, then reveal below.
Model answer: Parameterization is the use of simplified statistical representations for physical processes that occur at scales smaller than a model's grid box — individual clouds, turbulent eddies, sea-ice dynamics. Because these processes cannot be resolved directly, they are approximated by rules tuned to observations. Two models can perfectly agree on large-scale fluid dynamics and radiation physics but diverge on how they parameterize cloud microphysics, producing different estimates of climate sensitivity. This disagreement is honest — it reflects genuine scientific uncertainty about processes that are physically real but too small to resolve computationally.
Parameterization is the main source of inter-model spread for climate sensitivity. It is not a flaw in the models — it is an explicit acknowledgment of scale limitations. The uncertainty it introduces is quantified through ensemble modeling: running many parameterization variants reveals the range of plausible outcomes given our current understanding.