Wednesday, 13 January 2016: 9:45 AM
Room 245 ( New Orleans Ernest N. Morial Convention Center)
Ecologists often work with systems having spatial resolutions of 90 meters or less, but current-generation Global Climate Models (GCMs) produce projections of climate variables at much greater scales, typically between 250 and 600 km. This leads ecologists to use statistically or dynamically downscaled climate projections that are closer to their desired resolution, typically on a regional scale between 50 and 12 km. There are many downscaled climate projections available, yet few resources are available for creating ensembles of these projections. In order to gain a more robust interpretation of future climates, ensembles of multiple climate models are used to quantify uncertainty and are assumed to contain the true value of a future climate variable within the ensemble spread. Recent literature has questioned this assumption since it assumes model independence, which is not true in an equally weighted ensemble of all GCMs due to shared processes in model development. This may affect calculations of uncertainty and mean values. This analysis has yet to be applied to ensembles of downscaled climate projections, which have added sources of dependence based on the downscaling methodology and training datasets used in statistical model development. This study will evaluate dependence across multiple GCM generations for both statistically and dynamically downscaled climate projections in eastern North Carolina using in situ and gridded observations. Cluster analysis and model error correlation are used to assess model dependence both within an ensemble based on one downscaling technique, and across multiple downscaling methods. This project seeks to create a statistically viable ensemble of downscaled climate projections across eastern North Carolina for use in several ecological modeling efforts.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner