Greater future global warming inferred from Earth's recent energy budget
Key finding. Weighting climate model projections by how well each model reproduces the observed global pattern of Earth's top-of-atmosphere energy budget raises projected end-of-century warming under the steepest radiative forcing scenario by about 15 percent, or 0.5 degrees Celsius, while cutting the two-standard-deviation spread by about a third, or 1.2 degrees Celsius.
What question did this research address?
Most of the spread in projected twenty-first-century warming for a given emissions pathway comes from disagreement among climate models about how much the planet warms per unit of forcing, rather than from the pathway itself. This is response uncertainty, and unlike scenario uncertainty it can in principle be reduced with observations.
Differences in projected warming trace largely to differences in how models simulate Earth's radiative energy budget and how that budget adjusts as the planet warms. That suggests a test: models whose simulated energy budget matches what satellites actually observe should be given more weight than models whose energy budget does not.
Earlier attempts at this kind of observational constraint used single indices and reached conflicting conclusions. This work asked whether using the full global spatial pattern of the energy budget, rather than one number at a time, gives a consistent answer.
What did we find?
The predictors were nine global fields, built from three observable radiative quantities — outgoing shortwave radiation, outgoing longwave radiation, and the net downward energy imbalance at the top of the atmosphere — each characterised three ways: its mean climatology, the magnitude of its seasonal cycle, and the magnitude of its month-to-month variability. Partial least squares regression related these fields across models to the warming each model projects.
Every one of the nine predictor fields pointed the same way. Observationally informed projections of warming were larger than the raw, unweighted model mean in all cases, and the result held regardless of how many regression components were retained, which predictor fields were used, and which of the eight combinations of target year and forcing scenario was being predicted.
For the end of the century under the steepest forcing scenario, the observationally informed projection is about 15 percent warmer than the raw model mean — an increase of 0.5 degrees Celsius — and its two-standard-deviation spread is about a third narrower, a reduction of 1.2 degrees Celsius, relative to the raw model projections reported by the Intergovernmental Panel on Climate Change.
The models that best reproduce the observed energy budget are, systematically, the models that warm more. The constraint therefore shifts the central estimate upward at the same time as it narrows the range, rather than merely trimming outliers from both ends.
Why does it matter?
A warming projection that is 0.5 degrees Celsius higher for the same emissions is a tightening of the carbon budget. Achieving any given temperature stabilisation target requires steeper emissions reductions than the unweighted model spread implies.
Narrowing the range matters as much as raising the centre. Much climate policy analysis treats the model spread as irreducible uncertainty; showing that observations can cut it by a third changes what can be said with confidence about a given emissions pathway.
The method generalises. It uses quantities satellites already measure, so it can be reapplied as observational records lengthen and as new model generations appear.
Citation
Patrick T. Brown and Ken Caldeira (2017). Greater future global warming inferred from Earth's recent energy budget. Nature 552, 45-50.