AOS 271 Seminar - Paul Loikith


Thursday, May 9


Paul Loikith




Math Science 7124

TOPIC: "Extreme Temperatures over North America: Evaluating the Fidelity of Regional Climate Model Hindcast Experiments"


"Changes in extreme temperatures due to anthropogenic global warming are expected to have severe climate impacts on society. In order to better constrain uncertainty in climate model projections of future changes in extremes, it is essential to evaluate model fidelity in simulating extremes against observations. With this goal in mind, this work focuses on the development and implementation of novel evaluation metrics of daily surface temperature probability distribution function (PDF) characteristics and structure in six regional climate model hindcast experiments over North America. All model experiments were performed as contributions to the North American Regional Climate Change Assessment Program (NARCCAP) and are evaluated against two high-resolution reanalysis products. In many cases, model PDFs have biases of the same sign at both tails and the median of the distribution indicative of a shift in the PDF towards warmer or colder temperatures relative to reanalysis and independent of PDF shape. In general, wintertime variance is higher in models than in reanalysis in the northern portion of the domain, while variance is lower than reanalysis in the south. The models generally capture spatial patterns of temperature skewness in winter, with some differences in skewness magnitude. There is weaker agreement between datasets for summertime skewness, especially at lower latitudes, with models and reanalyses varying considerably. The model error identified using this holistic approach to PDF evaluation suggests extreme temperatures may not be properly simulated in some regions even where mean temperature bias is low. Results from this work will allow for future investigation of mechanisms associated with model error. "