About step-iterated models, I asked Grok, the X AI agent:
“Consider a time-stepped Earth climate model. The time step is 30 minutes. The model assumes a constant value of total solar irradiance at 1361 W/m^2. The absolute uncertainty of this value, arising from instrument limitations, is +/- 0.5 W/m^2. The corresponding geometric global average is 340 W/m^2 with an uncertainty of +/- 0.13 W/m^2. This epistemic uncertainty influences the change in forcing at each step. The climate state result from each step, including surface temperature, is passed to the next, so that this epistemic uncertainty accumulates in the series of steps. Albedo is a variable, computed at each time step. Using a no-feedback transient surface temperature response to radiative forcing of 0.2K/(W/m^2), please determine the uncertainty of the global average surface temperature one year following the initialization of the model, arising only from this specific external source of uncertainty. Use the RSS method for propagating the surface temperature uncertainty through the steps. Assume the model’s computations are exact. This exercise is not intended to evaluate the physical damping effects of feedbacks or thermal inertia, or any coding or tuning within the model for stability – only the buildup of numerical uncertainty through the steps of computation is of interest.”
Final Answer from Grok, after analysis:
“The uncertainty in the global average surface temperature after one year, due to the epistemic uncertainty in total solar irradiance, is approximately ±3.44 K.”
https://x.com/i/grok?conversation=1920920553092710601
So there’s a fundamental problem with the step-iteration climate simulation approach. We don’t know the strength of the incoming sunlight precisely enough to ever compute a reliable answer to the question of climate system response to incremental longwave radiative effects.
Even perfect models performing perfected physics won’t do it with sufficient resolution. It will still be a blurred result, arising from inherent limitations of measurement.
Comments welcome.