One legitimate type of node on a climate change argument map could be some output value from a climate model, which can support or undermine some other node. In that sense, the model output is equivalent to some observation ("the ice sheets are retracting"). However, if we then wish to evaluate the quality of this bit of evidence, the task is, by several orders of magnitude, harder than checking some measurement. We would need to check all the model assumptions, the implementation of these in the simulation code, the input data and so on - an almost impossible task, given the complexity of most serious climate models, the fact that they are implemented in procedural code rather than declaratively, the run time for the models, the large amount of single-sourced data, etc.
Although I am not a climate modeller, I am involved in environmental modelling, and take it for granted that such models should be part of the deliberation process. However, it's not clear to me how the climate argumentation community would see this happening. Should we be constructing extensive argument (sub)maps for each model, setting out the justification for each and every assumption etc? Or is the evaluation of the quality of a model treated as something entirely outside the argument mapping methodology?