For example, take a model of
aX + bY + cZ = m
X,Y,Z are known values a,b,c are parameters
A parameter is usually selected as a mean value of a range of potential values, a statistical distribution. So the parameters form an N dimensional space representing their ranges. In the given example it is three dimensional.
When the parameter ranges are not well known the area of the N dimensional parameter space increases because a weak estimator requires a higher confidence interval than a strong one.
So the question in complex modeling, and climate modeling is highly complex, is how close is the parameterization selected to the real parameters?
Wide parameter spaces across high dimensions are really fragile because the space is so wide. Good models account for this by spanning the parameter space, i.e. taking all the combinations at intervals and running the models multiple times.
The problem is, climate models are so big, and and have so many parameters that you cannot calculate them all on a human lifespan timescale.
If we are in a worst case scenario with climate change then we should be thinking 100% in terms of mitigations. More green and sustainable pathways make more sense with longer timelines.
Under any condition it does not look like Russia, India, and China are willing to cut their carbon profile in the future which means we really should be focused on mitigation and sensible sustainability projects.