“In the last few years the warming trend of the earth has stopped” is a common type of remark these days. Is that indeed the case, and can we conclude that the projections for the rest of the century are being overestimated? And if so, how come that large parts of Europe are expected to have the warmest year recorded in 2014? There is a chance that the global mean temperature will also be the highest in the series.
Guest post by Geert Jan van Oldenborgh, KNMI
The increase in the global mean temperature has slowed down
The most common measure of global warming is the global mean temperature. This has a somewhat unusual definition: over land is the the air temperature measured at 2 or 1.5 meters above the surface (T2m), over sea it is the temperature of the top meter or so of the ocean, sea surface temperature (SST). (The values over sea ice are either not considered or interpolated from land stations.) The reason is a practical one: there are many more observations of SST than of the temperature of the air above the water and oceans cover 70% of the globe. The difference between SST and T2m is small and does not change much, so as long as we consider anomalies it does not make much difference.
The old argument that the world is no longer warming was that the global mean temperature has ceased to increase after 1998. There are two ways to interpret that claim. The first is that there has been no warmer year than 1998. That year was about a quarter of a degree warmer than the trend of about 0.16 K/decade due to an exceptionally strong El Niño in 1997/1998. However, that record was broken in 2005 and 2010 (see Fig.1) by weaker El Niño events, so the argument is no longer correct. A confusing circumstance is that one old dataset, HadCRUT3, systematically underestimated the contribution of the Arctic, which showed strong warming since 1998. Due to this bias, this dataset kept 1998 as the warmest year. This has been partially corrected by the inclusion of a lot more Arctic stations in the new UK Met Office dataset, HadCRUT4.
The second, better, interpretation of the argument that the earth is no longer warming would be that a trend line starting in 1998 would be zero or negative. However, this is not the case: a linear trend starting at the peak in 1998 still gives a positive trend of 0.07 K/decade. The new argument therefore is that the temperature has not risen over the last 10 years. This is indeed correct, both the GISTEMP series and the HadCRUT4 dataset in which missing areas have been interpolated by Cowtan & Way show a trend of 0.00 K/decade over 2005–2014. To investigate further why the trend has been zero since 2005 it is useful to consider the land and ocean separately.
The land temperature has risen over the last 10 years
The land temperature is arguably most important to investigate the effects of global warming: that is where most people live. The long-term trend 1970–now is about 0.22 K/decade. This is higher than the global mean temperature because the relative humidity is lower over land. Another difference is that the variability is higher, σ=0.15 K. However, these variations have shorter time scales, there is almost no year-to-year persistence over land. The most important source of variability is (winter) weather over Siberia and North America. These are large regions with large temperature anomalies, caused by random weather fluctuations with characteristic time scales much shorter than one year, as far as we know. The other important factor is El Niño or La Niña, these explain about 30% of the variance, with a half-year delay. In Fig.2 one sees that 1998 and 2010 are above the trend line: both of these followed El Niño events. The years 2008 and 2011 followed La Niña events and are below the trend line.
Because most variability is on short time scales, the annual mean land temperature follows the CO2 concentration as a proxy for the total radiative forcing very well, see Fig.2b. A correlation as high as this one between two variables that are not trivially connected is very rare in the climate system.
The trend in the land temperature has been lower than the long-term average over the last 10–16 years, but is not zero. These lower values do not differ significantly from the long-term trend: the long-term trend is within the 95% confidence interval of the short trends. The same holds for the compensating periods in which the trend has been higher.
It has also been shown that extremely high land temperatures keep on occurring more frequently over the ‘hiatus’ time – for example, Seneviratne et al (Nature Climate Change, 2014) show that the percentage of the land surface affected by heat waves has increased strongly since 1997.
The sea surface temperature rise has stalled
The other 70% of the planet’s surface shows different behaviour. The variability is somewhat smaller, σ=0.12 K, but it exhibits a lot more persistence. Examples are the long period of cold SST at the beginning of the twentieth century and the prolonged warm temperatures relative to the trend in the 1940s, see Fig.3a. A much weaker version of these occurred the last decade: 2001–2006 were well above the trend line and 2007–2013 below (except 2009 and 2010, which had average temperatures relative to the long-term trend, see Fig.3b). This makes it possible to find a start date that gives a very low trend up to 2014: this is the case over the period 2002–2014 . Last year it was even easier, but the very weak El Niño (El Niñito?) that has been active since the summer of 2014 has increased the global mean SST somewhat. Combined with the background trend this implies that SST will be record high in 2014. Of course the trend leading up to 2002 was higher than the long-term average.
The pattern of temperature changes over the last 10 years has changed substantially since the ‘hiatus’ debate started a few years ago. At that time it was clear that the transition of a period dominated by El Ninõ at the beginning of the ten years to a period with largely La Niña at the end played a major role, partially compensated by the strong warming of the Arctic. In Fig.4a these areas are visible as the big blue (cooling) wedge on the Pacific Ocean and the large purple (heating) areas in the Arctic. The last 10 years this pattern has changed considerably. Because two El Niño years dropped out at the beginning of the 10-year period the wedge on the Pacific has all but disappeared, but so has the very strong warming of the Arctic (Figure 4b). The patches over North America and Siberia are mainly driven by random winter weather and have therefore also changed position. This shows how large the influence of random factors is on 10-year trends and how little can be learned by studying them.
In theory, the most important measure of the warming of the earth is the amount of heat absorbed by the ocean. Satellite observations have shown us that over the last ten years, the earth has received more energy in the form of solar radiation than it is has emitted as reflected solar radiation and thermal radiation. This imbalance is caused by the increased concentration of greenhouse gases and changes in aerosol concentrations. At longer time scales, this heat is mainly absorbed by the ocean, simply because the oceans contain a lot of water that has a large heat capacity. The atmosphere has very little heat capacity, the warmth does not penetrate far into the solid earth and the land and sea ice just cannot absorb enough heat while heating and melting to balance the influx of extra heat. In all, the ocean absorbs over 90% of the heat.
This was the basis of the previous ‘hiatus’ discussion, when the increase of ocean heat content seemed to stagnate from 2003 to 2010. However, this was only the case in the upper 700m of the ocean. Taking the deeper layers into account shows a continuous heating trend (Fig.5b). Since then, the rise of upper ocean heat content in the top 700m has also resumed, as expected (Fig.5a).
The ‘hiatus’ in the global mean temperature rise of the earth over the last 10 years is not a very sensible way to study global warming. Trends over such a short period are influenced heavily by (winter) weather over North America and Siberia, and by El Niño and La Niña. Over land the trend has not gone down to zero and 10-year trends are compatible with the long-term trend within the large uncertainty margins. Over sea there is also strong persistence from year to year, making the probability to obtain 10-year trends that are very different from the long-term trend very large. The best measure of whether global warming continues is the ocean heat content below the surface. This measure shows a very consistent rise, which is also expressed in sea level rise.
On top of the rising trend of the land temperature, much of Europe had warm weather for large periods this year. These two factors together, the long-term trend and a short positive fluctuation, caused the record-high temperatures that are expected this year.
With thanks to Andreas Sterl and Robert Mureau for very useful comments.
The plots shown can be reproduced on the Climate Explorer.