What is human-induced warming?
The globalwarmingindex.org website shows an up-to-the-second index of human-induced warming relative to the mid-19th century (1861-80) based on the standard "detection and attribution" approach introduced by Hasselmann (1997) and used by the IPCC and elsewhere ever since (see for example Otto et al. (2015)). This estimates contributions to observed climate change, removing the impact of natural year-to-year fluctuations, by a simple least-squares-fit between observed temperatures provided by the UK Met Office at the end of every month (black line) and estimated responses to human-induced and natural drivers of climate change (orange and blue lines). The forcing responses are provided by the standard simple climate model given in Chapter 8 of IPCC (2013), but the size of these responses (and so the climate sensitivity) is estimated by the fit to the observations (see Box 10.1 of Chapter 10 of IPCC (2013)).
Example of a simplified detection and attribution study. (a) Observed global annual mean temperatures relative to 1880-1920 (coloured dots) compared with CMIP3/CMIP5 ensemble-mean response to human-induced forcing (orange), natural forcing (blue) and best-fit linear combination (black). (b) As (a) but all data plotted against model-simulated anthropogenic warming in place of time. Selected years shown on top axis. (c) Observed temperatures versus model-simulated anthropogenic and natural temperature changes, with best-fit plane shown by coloured mesh. (d) Gradient of best-fit plane in (c), or scaling on model-simulated responses required to fit observations (red diamond) with uncertainty estimate (red ellipse and cross) based on CMIP5 control integrations (grey diamonds). Implied attributable anthropogenic warming over the period 1951-2010 is indicated by the top axis. Anthropogenic and natural responses are noise-reduced with 5-point running means, with no smoothing over years with major volcanoes.
The plumes in the Global Warming Index show the 5-95% ranges in estimated human-induced and natural warming due to uncertainties in UK observations (100 possible variations), forcing (200 possible variations) and response model (20 possible variations). The forcing shape uncertainty is estimated using a Monte Carlo method as applied in Box 9.2 in Chapter 9 of IPCC (2013) and further detailed in Forster et al 2013. The response model uncertainty covers a wide range of TCR/ECS (Transient Climate Response/Equilibrium Climate Sensitivity) ratios as well as fast response times. The contribution to the overall uncertainty is small, however. Instead, the total uncertainty is dominated by the forcing uncertainty. The animated image below demonstrates the methodology that we apply to obtain the final Global Warming Index.
Animation of the main steps needed to estimate the Global Warming Index. The first and second step shows observations and natural and anthropogenic foring estimates (W/m2), including 200 different realisations that comprise the full forcing uncertainty. The third step is the application of the response model which converts forcing in a fast and slow temperature equivalent as a function of the best estimate for TCR, ECS and the response times. In the fourth and fifth step, the sum of the two responses before regression with the observed temperature is shown together with the resulting combined temperature response (red). The sixth step is the same response after multiplying it with the slope provided by the the least-square-fit between temperature and natural/anthropogenic response. Finally, the full uncertainty range for the natural and the human-induced contributions is estimated and added to the graph.
We note that the monthly index uncertainty range is +/-0.0013K (95% confidence level). By this we refer to the uncertainty introduced by adding a new datapoint to the observed monthly temperature, or, in other words, it is the monthly variability of the Global Warming Index. All values and uncertainties in the GWI are expressed relative to the average of the 20-year 1861-1880 reference period. This is characterized by little volcanic activity, and hence comparable to the most recent 20 years.
Otto, F.E.L. et al. (2015). Embracing uncertainty in climate change policy. Nature Climate Change, doi:10.1038/nclimate2716
Hasselmann, K. (1997). Multi-pattern fingerprint method for detection and attribution of climate change. Climate Dynamics, 13(9), 601-611
Forster, P. et al. (2013). Evaluating adjusted forcing and model spread for historical and future scenarios in the CMIP5 generation of climate models. Journal of Geophysical Research, 118, 1-12, doi:10.1002/jgrd.50174