That graph is the result of two completely different methodologies plotted on the same graph.
Temperatures before 1870 are estimated based on ice core data and other proxies. Temperatures since 1870 are actual measured temperatures using thermometers.
Anybody at all who plots these on the same graph as if they were continuous data doesn't deserve to call themselves a scientist.
My Google-fu isn't working at the moment, but I've read some papers that say if you use the same proxies for the last 150 years of temperature, rather than switching to actual measurements, that spike doesn't show up. Of course, that means that such spikes could have been continuous through the last few thousand years, but we have no record accurate enough to show them.
I've worked in scientific data analysis before, and consistency was so important that if I upgraded a software version on a computer during the course of a large project, old data from that project had to be reanalysed to ensure the results didn't change.
Not only are the climate "scientists" not doing this, they are intentionally using data from completely disparate sources and methodologies in an attempt to prove their hypothesis. If this isn't acceptable in any other scientific field, why is it done in climate science?
Show me consistently obtained data