Predicting the Climate for the Next 10 years
10 Sep, 2007 02:40 pm
We live in a changing climate. The latest assessment by the Intergovernmental Panel on Climate Change (IPCC) predicts widespread warming by the 2080s in response to man-made emissions of greenhouse gases, with associated changes in heat waves, droughts and floods. But to begin to adapt to climate change we really need to know what will happen in the coming decade or so.
In order to predict internal variability we must employ the same techniques as used in short-term weather forecasting. That is, we must start from the present-day conditions, as measured by observations of the state of the atmosphere and ocean. To predict the weather for the next few days, the present state of the atmosphere is all-important, with the ocean exerting only a very small influence. However, conditions in the ocean generally vary much more slowly than in the atmosphere, with large-scale patterns of sea surface temperature (SST) persisting for months and sometimes years. These SST patterns do influence the exchange of heat between the atmosphere and the ocean, but their effect on day-to-day weather is usually small, and swamped by the chaotic variability of the atmosphere. However, consistently small influences can have a significant effect if applied for a long enough time. This is why climate (ie. the time-averaged weather) may be predictable months or years in advance even though the weather itself cannot be skilfully predicted more than a couple of weeks ahead. Starting from the present state of the ocean is therefore crucial for predicting internal variability for the coming decade. But this is not trivial. Whilst SST is measured reasonably well by satellites, the persistence and evolution of SST patterns is influenced by conditions below the surface, of which there are relatively few observations. It is therefore necessary to extract the most likely patterns of sub-surface ocean temperature and salinity from sparse and noisy observations. Furthermore, climate models are not perfect, and care must be taken to avoid forecast biases arising from model errors.
By the 2080s, global warming is expected to be very much larger than any effects of internal variability. The IPCC projections therefore did not bother with the added complexity needed to predict internal variability, focussing instead on uncertainties arising from unknown future emissions of greenhouse gases. However, on 10 year timescales, global warming and internal variability are both important, with uncertainties in greenhouse gas emissions playing a relatively minor role. In our study (Smith et al., 2007), we developed a system capable of predicting both internal variability and the effects of man-made global warming. We tested our system on past cases, and demonstrated that including predictions of internal variability substantially improves predictions of surface temperature throughout a decade, both globally and in many regions. These tests also showed that the predictions were generally very reliable, with the actual climate consistently falling within the uncertainty estimates of the forecasts. One exception was highlighted by the eruption of Mount Pinatubo, which caused the actual climate to cool relative to our forecasts. This is unavoidable, and we caution that future volcanic eruptions will also cool the climate relative to our forecasts.
We used our system to make a decadal forecast starting from June 2005. In this forecast, internal variability partially offsets the man-made global warming signal for the first few years. This has indeed occurred during the 2 years since the forecast was made, with relative cooling in the Southern Ocean and tropical Pacific correctly predicted by the new system. However, over the decade as a whole, climate continues to warm, with at least half of the years after 2009 predicted to exceed the warmest year currently on record.
Reference:
Smith et al., Improved surface temperature prediction for the coming decade from a global climate model, Science, 10 Aug 2007, Vol. 317, pp. 796-799
Interesting new results, and a very instructive account of timescale interaction for the non-expert.