"The sources I linked show that simulations using only natural forcings do not reproduce the late-20th-century warming trend"

Again, referencing a model that hasn't been shown to correctly hindcast climate is useless.

mhaze,

The hindcast comparison you asked for is exactly what detection-and-attribution studies present.

//referencing a model that hasn't been shown to correctly hindcast climate is useless.//

For example:

IPCC AR5 WG1 Chapter 10 compares model simulations against the observed 20th-century temperature record using known historical forcings (see Figure 10.1):

http://www.ipcc.ch/site/assets/uploads/2018/02/WG1AR5_Chapter10_FINAL.pdf

More recently, the same comparison appears in IPCC AR6 (Working Group I).

Those figures compare:

- observed global temperature
- simulations using natural forcings only (solar + volcanic)
- simulations including anthropogenic forcings

The result is straightforward.

Simulations using natural forcings alone do not reproduce the late-20th-century warming trend.

When greenhouse gas forcing is included, the models reproduce the observed temperature evolution much more closely.

That comparison is the hindcast test.

You asked for evidence that model simulations are evaluated by their ability to reproduce past climate behaviour. That is exactly what these attribution studies do.

Your links regarding Svensmark describe a proposed mechanism for solar-cosmic-ray influence on clouds. They do not demonstrate that natural forcings alone reproduce the observed late-20th-century warming trend.

That is the specific claim under discussion.

If you believe natural forcings alone reproduce the modern warming signal, point to the attribution study demonstrating that.

Otherwise the evidence remains as described above.