This is an interesting figure. Naturally, the sharp decline in the last 100 years or so captures one's attention, and would fit in with globally rising temperatures.

However, one other thing differentiates the older and newer epochs: The variance is much larger for older data. On the one hand it could be that increased variance is due to natural phenomenon (cooler weather?) or random error of less reliable measurements. However if that larger variance is due to systematic error (e.g., different methodologies/sources which were not accounted in the model (e.g., using hierarchical modeling), then there is a possibility that the difference between older and newer data are not that dramatic. What an interesting (and alarming) figure.