It looks like the output is a statistic. I.e. a trend line with error bars based on fitting prior historical data. Additionally, the processing of said histor

It looks like the output is a statistic. I.e. a trend line with error bars based on fitting prior historical data. Additionally, the processing of said historical data also seems to be statistical in nature, since there is a lot of missing data once we go past a few decades, data is from a disparate variety of sources, most are proxies for temperature and CO2, etc. So, it isn't obvious to me they are 'purely physical' as in there is some kind of direct reading off of physical facts. Instead, it appears most are a lot of inferences derived by fitting models to bodies of data, which entails a certain amount of uncertainty and a statistical nature.