Roy Spencer mentions a fact that many of us have calculated using their dataset, too: the year 2015 will almost certainly be the 3rd warmest year on the UAH AMSU satellite record, after 1998 and 2010. This ranking significantly differs from the surface temperature records which seem to display a positive (or noticeably faster) temperature trend and that will indisputably classify 2015 as the warmest year.
The temperature anomaly of 1998 was about 0.48 °C; it was 0.35 °C in 2010. For 2015 so far, assuming the December will be close to the November, the reading is about 0.26 °C while the following year, 2002, stands at 0.22 °C. You would need something like a collision with an asteroid to change the average for 2015 up to 0.35 °C or down to 0.22 °C.
While 2015 was colder than 1998, this isn't quite a fair comparison because the El Niños and La Niñas are the dominant sources of the annual temperature variability. In fact, the ongoing 2015-2016 El Niño seems almost identical to the "El Niño of the century" in 1997-1998. That is why it makes sense to compare 2015 to 1997.
And the anomaly in 1997 was something like –0.01 °C. The difference between 2015 and an El-Niño-wise similar year 1997 was about 0.27 °C. So one may rather reasonably take the 1997-2015 temperature difference as a better piece of raw data to estimate the "long-term trend" which is stripped of the effects of cherry-picking in the initial and final years that abuse El Niños or La Niñas in one way or another.
The temperature difference 0.27 °C occurred in 18 years. Divide these two numbers and you will conclude that the "fair" warming rate indicated by the UAH AMSU satellite data is 0.015 °C a year or 1.5 °C per century. According to the theoretical as well as empirical data, there seems to be no obvious acceleration or deceleration of the figure at the 40-year timescale and I think that it's a fair estimate for the extrapolations up to 2100. You may also want to notice that even if CO2 matters, the CO2 emissions don't seem to accelerate (much) anymore and the greenhouse effect only increases with the logarithm of the concentration so the greenhouse-induced warming will be slowing down if the annual emissions only stay the same.
In 2100, the global mean temperature could be about 1.3 °C warmer than it is today. This is the estimate that could be used for rational, science-based planning and it's obviously nonsense that it would be dangerous for anybody.
We (more precisely, our great great grandkids who will care about blog posts from 2015) may also be surprised and the temperature in 2100 may be 2 °C higher or lower than I indicated. But the risk of upside surprises and downside surprises are about the same. It's just irrational to be preparing for some significant warming because if the temperature change is going to be significant, then both signs are about equally likely. One must also realize that a planning 85 years in advance is sort of silly. The planning up to 2100 will be reasonably done by the people in 2080 who will be used to temperatures in 2080 and those will only be some 0.3 °C cooler than 2100 (plus minus 100 percent). And they will know that the temperatures in 2080 are just fine – much like we know that the temperatures in 2015 are just fine. The main point is that only planning for some 20 years makes some sense and according to any good empirical data and theories compatible with them, the temperature change in 20 years is simply negligible which is why the global temperature change should be neglected in any sensible planning. To try to "fight" these slow trends dominated by natural oscillations is silly enough that I won't dedicate any sentence to such ideas except for this sentence.
Also, I predict that the year 2016 will be the warmest one on the satellite record (except for some later years) – and the UAH anomaly may be close to 0.6 °C (the year 1998 plus 0.25 °C) but I can't be "quite" sure even about this claim.