Secondly, I can not discern what your interpolation procedure is for analysing the data — it is not a linear trend (which are given above)-- and without that information it is unclear whether it is appropriate and what the error bars on the forced component would be due to the unforced (random) noise component associated with any one year in any one realisation. (realclimate.org)
We characterize the noise component and assess algorithms for software mitigation. (stsci.edu)
The second strategy requires that the noise component be defined explicitly. (ipcc.ch)