Climate Model Bias Correction: Implementation and Applications

Abstract

This dissertation focuses on the methodology and usage of bias correction methods to understand and address problems stemming from climate model mean state biases over the course of three research chapters. Chapter 2 details the implementation of a running-mean bias correction to reduce mean state error in CAM, allowing us to demonstrate that that model's underprediction of atmospheric blocking was the result of said error rather than of problems with the moist physics. Chapter 3 describes models' representations of the Pacific Extreme Pattern, an observed subseasonal teleconnection between Pacific sea surface temperatures and heat waves over the East Coast. It also outlines causality experiments we performed that provided evidence that even in the model with the best representation of the Pacific Extreme Pattern we cannot ascribe clear causality to the teleconnection.

Chapter 4 explains our implementation of the tendency bias correction method in CAM. Although this method fails to achieve a high enough degree of error correction for most scientific purposes despite introducing a variety of refinements, we are able to rule out some of the possible causes of persistent error through investigations into nonlinearity within the model.

Finally, the dissertation outlines work that I have done to communicate with the public both about the state of knowledge within climate science and about the uncertainty within the field. I reproduce three opinion pieces that I published in the Los Angeles Times during my dissertation and describe the science that underlies those articles as well as the reasons why I believe this type of honest communication about what is and is not known about climate change is so crucial.