Girma,
I state my comment in another way:
The nearby annual temperatures are not independent, the changes from one year to the next are more independent. Thus the change over 30 years is a sum of 30 changes that are not very strongly correlated.
The change from 1910 to 1940 and the change from 1911 to 1941 contain common 29 annual changes and each only one that’s not common. Thus every annual change affects 30 neighboring points of your curve. This makes the curves behave so smoothly, and this is highly misleading.