Ten years of solitude

Those persons expecting a rapid increase of temperatures may have felt a certain unease during the first 10 years of the 21st century, when the global temperature did not seem to have risen as rapidly as it did in the 80's and 90's in the 20th century. Sometimes a knee-jerk response has been to deny this slow down of the temperature increase, and later, squabble about the sign of the estimated trend and about the cherry-picking of start and end years to calculate this trend.

I think a better way is to look at temperature simulated by climate models, not only for the first decade of the 21st century (see for instance, Knight et al 2009 ), but for the whole 21st century and check whether decades with near-zero or even negative 10-year trends are rare or common.

Climate simulations show that decades of near zero or even negative global annual trends are not particularly unusual. I have not analyzed all the data from the IPCC simulations in an effort to try and ascribe a significance to a zero temperature trend in the first decade of the 21st century. What I did was to look at some simulations we performed several years ago with the model ECHO-G) for the past millennium, and forwards to AD 2100 under different GHG scenarios. The global annual temperature from one of these simulations is shown below. The upper panel shows the global annual mean temperature between AD 1800 and 2100 simulated with ECHO-G. This model has been used in the IPCC AR4, although this simulation does not belong to the IPCC suite. To drive the model we used an estimation of the historical forcing up to AD 1990 and thereafter the B2 scenario for concentrations of carbon dioxide and methane. Solar irradiance in the 21st century was kept constant at the 1990 level. The temperature increase between AD 1960-1990 mean and AD 2100 is about 2ºC. The temperature increase in the 20th century, about 0.7ºC, is quite reasonable, although this is not a very important point now (I write this just in case someone wanted to argue that the CRU mails put this trend into question). The lower panel displays the gliding 11-year linear trends. For instance, the data point centered on AD 2005 indicates the linear trend calculated with the data in the period AD 2000-2010. It is  clear that negative temperature trends are simulated on several occasions: AD 2001, 2018 and as late as 2085. In other simulations with the same model and under the same scenario, the decades with negative decadal trends may occur at different times and more or less frequently. The frequency with which a negative decadal trend appears in the simulation depends on several factors. Without trying to be particularly accurate, one could mention as important factors the rate at which the forcing is increasing, the climate sensitivity, the thermal inertia, and the amplitude of the internal variations at decadal time scales. This latter factor is just the range of variations of decadal trends in a climate without changes in external forcing. Thus, a simulation with a steeper increase of greenhouse gas concentrations may show negative decadal trends less frequently. A climate system with larger internal variations may display negative trends more frequently than a system with smaller decadal internal variations. It can of course be asked  on which grounds can we be sure that the internal decadal variability in the model is realistic, and that would be a good question.

Under scenario A2, a pretty pessimistic scenario, the model ECHO-G simulates one decade with negative trend at the start of the 21st century.

Since the prescribed external forcing in the 21th century increases monotonically, decades with negative temperature trends can only be caused by internal climate variations, unrelated to the forcing. Certainly, by widening the period over which the trend is calculated, for instance from 11 years to 15 years, the internal variations tend to be filtered out more strongly and the frequency of negative trends should diminish. How long should the present period with near-zero or even negative trend be in order to claim that models are incompatible with observations? This question requires a much more detailed analysis of a large ensemble of simulations driven by the observed forcing, including concentration of greenhouse gases, volcanism, solar irradiance and others. However, from this quick calculation, it seems that 10 years is just a bit too early to feel lonely .. or exultant.