Long range predictions of the expected increase in global average surface temperature (GAST) due to increases in greenhouse gas emissions (referred to as Equilibrium Climate Sensitivity (ECS)) during the 21st century vary from one or two degrees Celsius (C) to 4°C or more,  depending on the Global Climate Model (GCM) used and the assumptions regarding future greenhouse gas concentrations (Figure 1).
Figure 1. Future scenarios of global average surface temperature (°C) out to 2100 based on Global Climate Models (GCM) from the 2013 IPCC report. Solid lines show mean temperatures while shaded areas show uncertainties (5% to 95%). Intermediate scenarios are shown by the vertical shaded bars at right. The numerical values are the number of GCMs used to obtain the results.
The confidence in the larger GAST forecasts (>4°C by 2100) is low due to the fact that the performance of GCMs since the 1990’s has been poor. A recent Canadian study shows that the projected warming trend from the early 1990’s to 2012 was twice as high as actually observed for 37 models in 117 separate forecasts (Figure 1). When projected over even longer time periods, such as through the end of the 21st century, this could result in significant errors.
Figure 1. Forecasts of temperature change (°C per decade) by GCMs (gray bars) versus observed change (red hatched area) from 1993 to 2012. (From Fyfe et al. 2013)
GCMs simulate the atmospheric motion and energy transfers over long periods of time. The overestimates of GAST by GCMs are likely due to inaccurate assumptions about the: (1) reflection of incoming solar radiation, and (2) blocking of outgoing infrared radiation. [2,3] These two factors are mostly a result of cloud cover, which the models assume will shrink in area with rising temperatures, while the opposite is often the case. Increasing cloud cover can be caused by increased evaporation from ocean surfaces in a warmer climate, even in the Polar region as ice sheets melt.
Another important mitigating factor is air pollution and volcanic aerosols, such as SO2, which help cool the planet by reflecting solar UV energy back into space. Even lower atmospheric dust from the Sahara and other deserts is not simulated well in GCMs.  These dust clouds reduce heating of the sea surface, which affects the formation of tropical cyclones in the Atlantic and other ocean basins.
While improvements in the models are likely to occur, the result in the meantime is a wide range of projected temperature increases that will likely exceed the actual rate of global warming during this century as a result of the previous discussion. One researcher, the late Dr. William Gray, argued that the expected increase in CO2, even a doubling of CO2, will not bring the anticipated increase in global temperatures. Instead, he believed that warming of only half a degree C or less will occur, not the 2-4°C or more expected by the IPCC. This claim has historical basis. Around 1900, Swedish physicist Knut Angstrom determined that CO2 concentration beyond about 50 ppm has little effect on the Earth’s temperature, although the results of that experiment were disputed. Dr. Gray believed that changes in the strength of the inter-ocean circulation (such as the Atlantic Thermohaline Current) which controls the salinity of the oceans has led to the 0.8°C increase we have seen in the past century, not man-made greenhouse gases.
So based on the evidence at hand, I believe that a global warming scenario on the low end of the prediction scale, say in the range of an additional 1-2°C or so is what will most likely occur by the year 2100. Note that likely scenarios from the 2013 IPCC report have been reduced from those in earlier year reports, and one of them (RCP2.6) indicates less than 1°C of global warming. This falls in line with what the satellite observed trends have been since 1979 (less than 0.2C per decade).
A recent study based on the long term past variability of the GAST  estimates that the probability that the GAST increase will exceed 4.5 C by the end of the 21st century is less than 1%. On the other hand, the probability of an extremely low (1.5 C) value is less than 3%. A central estimate of 2.8 C was given by Cox et al., with a confidence of 66%. Improvements in greenhouse gas emissions that are already underway could see a further reduction in this scenario.
- IPCC Fifth Assessment of Climate Change 2013 – Synthesis Report
- Fyfe, J. C., N. Gillett, and F. Zwiers, 2013: Overestimated global warming over the past 20 years. Nature Climate Change, Vol. 3, pages 767-769
- Gray, W. M., 2012: Physical Flaws of Global Warming Theory and Deep Ocean Circulation Changes as the Primary Climate Driver. Heartland Institute’s 7th International Conference on Climate Change (ICCC-7), Chicago, IL, 21 to 23 May 2012, available at: http://tropical.atmos.colostate.edu/Includes/Documents/Publications/gray2012.pdf
- Evan, A. T., C. Flamant, S. Fiedler, and O. Doherty, 2014: An analysis of Aeolian dust in climate models. Geophysical Research Letters, Vol. 41, pages 5996-6001.
- Cox, P., C. Huntingford, and P. Williamson, 2018: Emergent restraint on equilibrium climate sensitivity from global temperature variability. Nature, 553, 319-328.