Enlarge /. An interesting view of the world: the darker the red, the closer the correlation between local temperature and global mean. Blue areas tend to prefer contrasting temperatures.
Meteorologists are performing weather forecasting models to predict weather conditions well over the next few days. Climate researchers, on the other hand, are conducting global climate models to project the effects of climate change-related greenhouse gas emissions in the coming decades. Between these two activities lies an interesting task that has proven to be more difficult: predicting global temperature over a number of years.
A new study by Patrick Brown and Ken Caldeira tries a new approach to this challenge by using only statistical analysis of the temperatures of the past two years.
The annual average surface temperature of the globe varies slightly from year to year, even if a long-term warming trend is evident. It is these wobbling movements from year to year that are difficult to predict. They depend on different regional weather patterns, especially the El Niño Southern Oscillation. This rocking pattern of warm surface water along the equatorial Pacific is significant enough to raise the planet's average surface temperature up and down. It also affects weather patterns in many places around the world.
This also applies to other species in which the weather conditions differ on a large scale. Years in which the planet sets a new temperature record are obviously warm in most places at the local level. The possibility of giving an advance warning is therefore not only useful beyond global temperature accounting.
This is obviously something that climate researchers have been playing around with for a long time. One possibility is to enter current temperature data in a climate model and then simulate it a year or two in advance. This has some advantages, but the curves of El Niño / La Niña are stubbornly difficult to grasp in simulation, and these models require supercomputer time.
Another, less resource-intensive option is to try a purely statistical prediction based on previous data. This is often based on variability metrics such as the El Niño Southern Oscillation Index, which reduces the surface temperature patterns to a number representing the states of La Niña, Neutral, or El Niño. While this is convenient, it can discard data that can help predict.
A new method
To try something different, the researchers developed a method that simply captures temperature data for each cell in a grid that covers the globe. For all data prior to 2000, her method analyzes the complete temperature pattern over a two-year period and compares it to the average global temperature in the following years. The end result is a complex mathematical correlation that can be used to predict future global temperature.
The method has been tested in several ways. It has been recalculated repeatedly, based on each year before 2000 with one exception. The method was then tested by predicting the missing year – the so-called "skip validation". It was also used to predict global temperatures for every year after 2000 without calculating correlations. In any case, the method actually performed quite well, surpassing simple assumptions of a continuing trend and climate model simulations.
On closer inspection of the correlation pattern, the method is not too surprising. It picks up the trend of global warming temperatures and also finds predictive power in the areas of the ocean that vary with things like the El Niño Southern Oscillation. It shows that an El Niño (warm water that stretches to the east side of the equatorial Pacific) will be associated with a warmer global temperature next year. The cooler water of a La Niña, on the other hand, is usually associated with a warmer global temperature two years later – which reflects the back and forth of this phenomenon.
In addition to the study, researchers published current forecasts online and added new data each month to predict the next three years. (You can also see how the method would have predicted each year before.) Currently, it is predicted that 2020 will be 92 percent the warmest year ever. The year is halfway over at this point, but this is similar to Berkeley Earth's current forecast based on previous temperature. The latest NOAA forecast published on Thursday puts this rate at 49 percent.
Enlarge /. Here is the latest version of the forecasts for 2020-2022.
Of course, this statistical approach cannot predict unpredictable events. Volcanic eruptions (or global pandemics) can cause sudden wrinkles that have little to do with last year's temperatures. Take for example the updated forecast 2020 of the method. In the published study, it was presented based on data up to the end of 2019. There, the central estimate would have classified it as the third warmest year in existence, although the error bars certainly contained the current forecast.
The first half of 2020 was quite warm and drove the current forecast up. This could only be a case where the method slightly misses the mark, but it could be that the economic impact of COVID-19 reduced aerosol pollution and thus added heat. The possibility has not yet been examined in detail.
In both cases, this study adds another independent method that could lead to more reliable short-term global temperature forecasts. That would be more of a heads-up when there is a record year on deck – including the first year that exceeds a milestone like the 1.5 ° C warming.
Earth and Space Sciences, 2020. DOI: 10.1029 / 2020EA001116 (About DOIs).