Modelling is only as good as the input data and built-in model assumptions. Experience with geotechnical computer programs for slope stability, liquefaction, and other hazard modelling indicates that a wide range of potential results are possible all within the envelope of reasonable expectation. Experience, judgement and site specific data are critical and must be applied to any model results. This has important implications for climate change, where predictions of global temperature rise and associated secondary effects are based on limited data and imprecise, inaccurate models. Graph from Elon Musk based on actual data available here:
Images | Commentary | Digital Scrapbook