Title: Do high-resolution regional climate simulations provide credible regional and local trends in temperature over the contiguous United States?
Abstract: While systematic, multi-model experimentation and evaluation have been undertaken for years, the development and application of infrastructure for systematic, observation-based evaluations of spatial patterns in key climate variables simulated with various spatial resolutions are less mature, owing in part to the needed advances and synergies in both climate science and statistics. One of the main challenges in using existing analysis tools is to carry out the multi-resolution investigation of climate models. Given this, the principal science objective of my work is to provide quantitative and robust evaluations of spatial patterns simulated by climate models across multiple scales: comparison of spatial features at coarse (e.g. 100 km) and fine scales (e.g. 1-10 km) separately between observations and models.
I introduce the Hierarchical Equal Area isoLatitude Pixelization (HEALPix) as a coordinate system for evaluating climate models at different spatial resolutions. The unique capabilities of HEALPix include open-source libraries to facilitate the handling and distribution of massive datasets at different resolutions using parallel computing, and fast and robust analysis of spatial patterns from observational and model datasets regridded into HEALPix pixels, which have been widely used by astronomers and planetary scientists.
As an example, I investigated the spatial scale of temperature variability over the contiguous United States (CONUS) and show how well average trends over a large region represent trends at individual grid points within the region. Previous studies have demonstrated that regional, higher spatial resolution, climate models (RCMs) better reproduce extreme weather events on a local scale than coarser resolution global climate models. Given this, we also evaluated temperature trends in the Canadian Regional Climate Model (CRCM) that participated in the North America CORDEX (NA-CORDEX) program. Based on its simulated temperature trends over the CONUS, the performance of CRCM is strongly dependent on their resolutions. The simulation with the highest resolution of 11 km shows significant differences from its low-resolution counterparts with resolutions of 22 km and 44 km, especially in the Southern Great Plains. Our analysis demonstrates the added value of high-resolution dynamical downscaling in simulating temperature trends on a local scale for the present and future climate.
Biography
Kyo Lee has a Ph.D. in Atmospheric Sciences from the University of Illinois at Urbana-Champaign, and a B.S. in Earth and Environmental Sciences from Seoul National University in Korea. While studying atmospheric sciences, Dr. Lee became very interested in statistics. Dr. Lee has also had extensive experience with stratospheric dynamics, neural network modeling for satellite remote sensing, atmospheric chemistry, and air quality modeling. Through the adaptation and extension of previously developed statistical tools, Dr. Lee currently develops metrics for climate model evaluation.
