I've spent many years creating statistical analysis and models about a large business's operations. There is always at least one statistic that someone produces

I've spent many years creating statistical analysis and models about a large business's operations. There is always at least one statistic that someone produces that purports to show the opposite of what a balanced analysis shows. In this case the lowest recorded temperature is not a solid indication of cooling or even of "not warming" because it is a localised measurement; it's about one small area - and one small area in a remote and hard to access place. As our civilization has become more powerful we are able to put sensors in many more places, some of those places were inaccessible to previous generations of sensor programs. Sensors now operate autonomously and the data is collected remotely. This would create a strong bias towards detecting lower temperatures. It would be possible to estimate and control for this bias in the number - but I haven't done that as it's a bit of work! As a not expert (not ever involved in any climate analysis) if I were asked to create a meaningful statistic I would try to do an analysis that tries to see things as they are (rather than supporting an opinion) and would look at the mean and median temperatures recorded at sensor locations over time, and would also look at the evolution of temperature aggregates across representative populations of sensors deployed - which would include sensors in cities and deserts as well as those in Antarctica. For example I would take random samples of 10% of the sensor population that was available at a given time point. I have no idea how the climate scientists do this - but I am confident that they will have methods that are far more reliable than my silly homebrew idea - if you are interested in the topic it would be good to look into how they do it.