Search

Predictions of the sensitivity of the earth's surface temperature to changes in atmospheric CO₂ depend on the highly uncertain cloud radiative feedback. How will clouds change in a warming climate, and will those changes amplify or dampen the CO₂ radiative forcing? Climate models offer diverging predictions, so which models are more likely to be correct? A popular type of hypothesis for answering that question is the "emergent constraint." Emergent constraints identify a present-day model bias that is correlated to cloud feedback prediction and evaluate the likelihood of the model prediction based on the size of the model bias. Are emergent constraints skilled at predicting cloud feedback in the real world, or are they artifacts of the limited set of climate models that they describe? In this dissertation, I seek an answer to that question by testing some emergent constraints using out-of-sample climate models, which are organized into ensembles derived from a single model (SMEs). The SMEs are created from the structure of two generations of the Community Atmosphere Model by sampling uncertain model parameters using a Markov Chain Monte Carlo technique. The SMEs project uncertainty about the present-day climate into the future. We argue that this type of calibration makes SMEs a powerful tool for quantifying uncertainty in cloud feedback and for testing emergent constraints. Each emergent constraint that we test cannot predict the cloud feedback in one or both SMEs. The problem is especially acute for predicting cloud feedback in climate models released after the hypothesis is proposed. The tests lower our confidence that these emergent constraints will predict the cloud feedback in the real world and suggest further testing should be conducted before emergent constraints are considered reliable. more...

Predictions of the sensitivity of the earth's surface temperature to changes in atmospheric CO₂ depend on the highly uncertain cloud radiative feedback. How will clouds change in a warming climate, and will those changes amplify or dampen the CO₂ radiative forcing? Climate models offer diverging predictions, so which models are more likely to be correct? A popular type of hypothesis for answering that question is the "emergent constraint." Emergent constraints identify a present-day model bias that is correlated to cloud feedback prediction and evaluate the likelihood of the model prediction based on the size of the model bias. Are emergent constraints skilled at predicting cloud feedback in the real world, or are they artifacts of the limited set of climate models that they describe? In this dissertation, I seek an answer to that question by testing some emergent constraints using out-of-sample climate models, which are organized into ensembles derived from a single model (SMEs). The SMEs are created from the structure of two generations of the Community Atmosphere Model by sampling uncertain model parameters using a Markov Chain Monte Carlo technique. The SMEs project uncertainty about the present-day climate into the future. We argue that this type of calibration makes SMEs a powerful tool for quantifying uncertainty in cloud feedback and for testing emergent constraints. Each emergent constraint that we test cannot predict the cloud feedback in one or both SMEs. The problem is especially acute for predicting cloud feedback in climate models released after the hypothesis is proposed. The tests lower our confidence that these emergent constraints will predict the cloud feedback in the real world and suggest further testing should be conducted before emergent constraints are considered reliable. more...