Climate models can show observations to be wrong

Simulate this. William Putman/NASA, CC BY

Aerosols are microscopic liquid or solid particles suspended in the atmosphere. They can originate from natural sources (pine forests, for example), but centuries of burning wood and fossil fuels means there are plenty of aerosols from human activity.

Aerosols influence the climate system by changing the Earth’s energy balance, for example by reflecting sunlight back into space before it reaches surface level. But the strength of this effect, known as the aerosol forcing, remains uncertain.

Due to their typical lifetime of less than two weeks, and great variability in size and chemical composition, the effects of aerosols are difficult to model. While it remains particularly difficult to resolve the many small-scale interactions between aerosol particles and clouds, their general effects and behaviour can and have been simulated.

What our study shows is that, while different climate models simulate aerosol effects in varying levels of complexity, the range of results for different models broadly agree with each other – but reveal errors or bias in recorded observations.

Small particles, big influence

Some recent studies have shown how temperature and precipitation in certain regions are influenced by the emission of aerosols.

Our study shows that both regional temperature and precipitation in the tropics respond predictably to changes in aerosols, both in climate models and as observed during the mid-20th century. Climate models show that precipitation in the northern hemisphere mid-latitudes decreases in response to these changes in aerosols, in line with theory. But recorded observations reveal an increase in precipitation, the opposite to what is expected.

For example, aerosol emissions from industry increased markedly in the northern hemisphere mid-latitudes as Europe rebuilt its economies and industrialised in the years following World War II. This caused the temperature gradient between the two hemispheres to alter, as the northern hemisphere cooled and southern hemisphere continued warming – something that climate models accurately recreate in simulations.

Hadley cells’s place in the planet’s atmospheric circulation. Kaidor, CC BY-SA

The effect on temperature has a knock-on effect on precipitation patterns, even far from the source of the aerosols. Recent studies have shown that cooling concentrated over the northern hemisphere mid-latitudes can drive changes in the strength of the tropical Hadley cells, a key component of global atmospheric circulation.

As a result the tropical rainbelt, or Intertropical Convergence Zone, shifts southwards. This shows up in records of rainfall in the northern hemisphere’s tropics, which show a clear downward trend in the 1960s and 1970s, and also in the severe drought in the Sahel of that time. Again, models and observations match.

Observations vs models

But when we examine the mid-latitudes in the northern hemisphere this agreement between observations and models breaks down. Our most reliable observations should be here, considering more people have been recording the weather for a longer time than anywhere else globally. In this region we’d expect local aerosol concentrations to lead to a reduction in precipitation – but the observations of the last few decades show quite the opposite.

Scientists know that there may be errors or bias in observed mid-latitude land precipitation trends, particularly in the early 20th century. This is due to jumps in the record that can arise through changes in the instruments used or recording or measuring practices. Ocean temperature measurements revealed a distinct temperature drop after 1945, for example, which was eventually revealed to be down different methods of measuring water from a bucket. Snowfall, difficult to measure properly, is a further complication at higher latitudes.

But significantly, our study offers an idea of how wrong these observations might be, through linking temperature changes, precipitation changes and aerosol changes, offering an opportunity to correct these key observations.

Observations are often taken as truth and used as the standard against which to compare models. We should make a concerted effort to continue correcting these important observations, while always considering potential bias when they are incorporated into studies. Observations, not just models, can also be wrong.