[audessuscest]

> They are used with little modification to correctly model climate of Mars, Venus, tidal locked exoplanet and previous historical eras on Earth. You just set the parameters and they pretty much produce the climate that exists o

why are they wrong all the time then ?

[DagAgren]

They are not. They are very accurately tracking the current changes in climate.

[briantakita]

However they are not able to accurately forecast solar activity & climate into the future, making them need to frequently readjust their models to fit the data of the past & present. i.e. Curve Fitting.

[DagAgren]

Their entire job is to forecast climate into the future, and they do that well. Solar activity is an input, and is obviously not predicted. Your comment is gibberish.

[briantakita]

> Solar activity is an input, and is obviously not predicted. Your comment is gibberish.

If you can't predict certain known inputs, and theres unknown unknowns, how can your model have predictive accuracy? If you believe that the model is accurate, what makes the model falsifiable? One of the indicators of a pseudoscience is nonfalsifiability.

https://en.wikipedia.org/wiki/Pseudoscience

[sonofgod]

A model that can't predict inputs can provide a mapping from any given inputs to a predicted value. We can then, after the fact, demonstrate whether the actual value is equal to the predicted value (i.e. within expected error bars). Repeated failure to predict at the expected probability of your error bars would falsify the model. (i.e. commit to what your model is, then run it with now known inputs after the fact)

It's possible some of these inputs may be of the form "the Gulf Stream continues to operate within historically observed parameters", i.e. we make no attempt to claim accuracy for the model outside of its studied conditions. That's fine: there are a large number of accurate models which are known to have breakdown conditions.

It is literally impossible for any model to be robust to unknown unknowns. All we can do is endeavour to drag them into the light and become known unknowns, then attempt to quantify them as known knowns. Sadly, our ability to experiment repeatedly in the same conditions with the environment is deeply limited, in a way much akin to astronomy.

[briantakita]

> Repeated failure to predict at the expected probability of your error bars would falsify the model. (i.e. commit to what your model is, then run it with now known inputs after the fact)

That doesn't help if you are not gathering the correct proxy data to represent the actual physical system. For example, CMIP6 includes particle forcing while CMIP5 & before did not. The global & solar electric circuits are not considered. Are xrays, which have a large variance with solar cycles, included in CMIP6? I don't think the Birkeland Current is considered.

What also doesn't help is adjusting historical measurements such as temperature. It's like changing the rules mid-game, which indicates that past observations are mutable to fit the models. Also note the incomplete data, as modern satellite observations span a tiny window compared to Earth's lifespan. The solar system travels through different parts of the Galaxy, crossing the Galactic Current Sheet at times, passing through ionized dust, among other phenomena.

> Sadly, our ability to experiment repeatedly in the same conditions with the environment is deeply limited, in a way much akin to astronomy.

We can experiment with ions, plasma, & high voltages applied to plasma. Note the Birkeland Current between Sun & the Earth, ions being ejected from the Sun, XRay emission varying wildly with solar activity, cosmic rays, etc. There's also a Birkeland current sheet in the solar system & a galactic current sheet.

One of the major issues is that the standard model does not consider plasma & electricity, in light of recent observations of the prevalence of Plasma in space, effectively blinding large parts of the astronomical theory to it's impact. This led to a legacy of ad hoc, non-observable (thus non-falsifiable) inventions such as "dark matter", "dark energy", "black holes", "neutron stars", "the big bang", "space/time", "magnetic line snapping", "parallel universes", "11 dimension space" & whatever other mathematical patches are used to fit the observations. Note inconsistencies with the assertion that redshift equals distance with observations such as examples of a Quasar with a high redshift being connected to a Galaxy with a low redshift.

Point is climate science is built on a tower of assumptions built on assumptions where some of the foundational pieces are called into question. Ad hoc tweaking of observational data & ad hoc tweaking of models to fit the transformed observed measurements makes it easy to doubt the accuracy, predictive ability, & knowledge of applicable boundary conditions of the model. It sounds like a tough job to get it right.