The above chart plots the changing 3 - year linear trend slopes using monthly observations going back to 1850 (this is the HadCRUT4 dataset from the UK climate research agency - it is the only global dataset going back that far). (c3headlines.com)
Trend lines tend to be pretty robust and I would therefore be very surprised if the error bars for each of the 0.089 and 0.135 trend slopes came anywhere near overlapping. (judithcurry.com)
The two measures of trend slope would converge to the same number as the number of observations increased. (climateaudit.org)