Crown Copyright

Decadal forecast

Decadal forecasts, also called 'near-term' climate predictions, range up to a decade ahead. Predictions account for natural variability and climate change as these are expected to be of similar size in many places. Forecasts are experimental, so at this early stage of development skill levels vary from place to place and for different variables. As a result, expert advice is needed to assess the reliability of regional predictions.


  • Long-range forecasts are unlike weather forecasts for the next few days
  • Forecasts show the likelihood of a range of possible outcomes
  • The most likely outcome in the forecast will not always happen
  • Forecasts are for average conditions over a wide region and time period
  • For more details on interpretation, see How to use our long-range outlook


Global average temperature is expected to rise to between 0.32 °C and 0.71 °C (90% confidence range) above the long-term (1971-2000) average during the period 2011-2015, with values most likely to be about 0.51 °C higher than average (see blue curves in the Figure 1 below).

From 2016 to 2020, global temperature is forecast to rise further to between 0.47 °C and 0.94 °C, with most likely values of about 0.71 °C above average. The warmest year in the 160-year Met Office Hadley Centre global temperature record is 1998, with a temperature of 0.40 °C above the long-term average. 2009 had a temperature of 0.32 °C above average. The forecast trend of further global warming is largely driven by increasing levels of greenhouse gases.

Global annual temperature, decadal forecast

Figure 1:Observed (black) and predicted global average annual surface temperature difference relative to 1971-2000. Previous predictions starting from June 1985, 1995 and 2005 are shown as white curves, with red shading representing their probable range, such that the observations are expected to lie within the shading 90% of the time. The most recent forecast (thick blue curve with thin blue curves showing range) starts from September 2010. All data are rolling annual mean values. The gap between the black and blue curves arises because the last observed value represents the period November 2009 to October 2010 whereas the first forecast period is September 2010 to August 2011.

The following maps (Figure 2) show the difference between the predicted geographical distribution of temperatures over the period September 2010-August 2015 and long-term averages in °C. Diagram A is the most likely forecast outcome. Diagrams B and C indicate the range of forecast temperatures, such that we expect only a 10% chance of temperatures at a particular location being less than those in B, and only a 10% chance of temperatures higher than in C. Note that these ranges are for each individual location. The chances of these limits being met everywhere are very small, so the complete patterns shown in diagrams B and C are very unlikely to be realised.

Five-year mean forecast: mean, lower and upper estimates

Figure 2:Forecast (A) maps of surface temperature differences (C) relative to 1971-2000 for the 5-year period September 2010 to August 2015. Forecasts consist of 20 ensemble members starting from June and September 2010. The probable range is diagnosed from the ensemble spread, and shown as the lower (B) and upper (C) limits for each 5 grid box, such that there is a 10% chance of the observations being cooler than (B), and a 10% chance of the observations being warmer than (C). Note that the actual anomaly patterns in (B) and (C) are unlikely to occur since extreme fluctuations would not be expected at all locations simultaneously.


Retrospective forecasts have been made from numerous dates in the past. Some of these are shown in the top global annual temperature forecast figure (white curves and red uncertainty regions from 1985, 1995 and 2005). Generally, the forecasts predict rises in temperatures similar to those observed (black curve). Many facets of the variability, such as the record warming caused by the large 1997-1998 El Niño and the cooling caused by the 2008 La Niña, are within the range of the predictions (red shading). An exception is the cool period after the large volcanic eruption of Mount Pinatubo in 1991, which could not have been forecast years in advance.

The following maps (Figure 3) compare observed (A) and forecast (B) surface temperatures (°C) for June 2005 to May 2010 relative to the 1971-2000 long-term average. Forecasts were made starting from March and June 2005. Stippling shows regions where the observed temperatures do not lie within the predicted range (5-95%) of the forecast. Over much of the world (69%, not stippled) the observed outcome is consistent with the forecast. The region that is inconsistent (stippled) is larger than the 10% expected by chance, showing that in some locations the forecast did not capture the full uncertainty.

Observations and five-year mean forecast from June 2005

Figure 3:Observed (A) and forecast (B) maps of surface temperature differences (C) relative to 1971-2000 for the five-year period June 2005 to May 2010. Forecasts consist of 20 ensemble members starting from March and June 2005. The stippling shows where the observations lie outside of the 5-95% confidence interval of the forecast ensemble. This is expected in 10% of cases, but actually occurred in 31%, showing that the ensemble did not capture the true range of uncertainties.

Last Updated: 6 June 2011