Editing Openai/6938382f-f9e4-8003-a158-6d812c1bce84 (section)

==== 6. Comparison to relevant literature ====
* Error propagation using RSS is standard for laboratory measurements and simple models. Propagation_of_Error_and_the_Re…
* In climate modelling, formal uncertainty quantification uses: - perturbed-physics ensembles, - multimodel ensembles, - emulators / surrogate models, - Bayesian hierarchical frameworks, etc. IPCC<ref>{{cite web|title=IPCC|url=https://www.ipcc.ch/site/assets/uploads/2018/02/WG1AR5_Chapter09_FINAL.pdf|publisher=ipcc.ch|access-date=2025-12-10}}</ref>

Relative to that landscape:
* Frank’s approach (fit a simple linear emulation → insert a large calibration error → propagate as random walk) is novel but flawed: - Novel in the sense that few people try to map a single climatological RMSE into a time-growing random uncertainty. - Flawed because it misinterprets the nature of the underlying error metric and ignores dynamical and energetic constraints of the actual models.

Critiques from other climate scientists (including those skeptical of high climate sensitivity) have pointed out these issues, particularly the misuse of LWCF RMSE and the implication that models should behave like an unconstrained random walk if his error model were correct. drroyspencer.com<ref>{{cite web|title=drroyspencer.com|url=https://www.drroyspencer.com/2019/09/additional-comments-on-the-frank-2019-propagation-of-error-paper/|publisher=drroyspencer.com|access-date=2025-12-10}}</ref>