Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Engineering Science

Program

Civil and Environmental Engineering

Supervisor

Simonovic, Slobodan P.

2nd Supervisor

Gaur, Abhishek

Affiliation

National Research Council of Canada

Co-Supervisor

Abstract

Global climate models (GCM) are sophisticated numerical models used to make long term climate projections. However, the resolution of their output is too coarse for climate change related local impact studies on urban regional scales. Downscaling efforts are taken to address this and increase GCM projection resolution. Physical Scaling (SP) downscaling methodology attempts to incorporate the physical basis of dynamical downscaling efforts with the computational efficiency of statistical methods. In this study, North American Regional Reanalysis surface skin temperature and precipitation data for a 1°x1° region centered on Houston, TX are downscaled to a resolution of 500m via SP and Weather Research and Forecasting (WRF) models. SP models are found to significantly and moderately outperform WRF models in terms of surface temperature and precipitation, respectively. SP methodology is then chosen to downscale GCM projections across 44 urban regions within Canada and the USA. Climate change impact is assessed via comparison of change factors between projections for representative concentration pathway (RCP) scenarios 2.6 and 8.5. Nearly half of all regions have significant projected increases in median and variance of surface skin temperature between RCP scenarios. Precipitation change factors vary significantly depending on GCM choice with median annual precipitation change factors of 29mm to 256mm projected by the 2090s in RCP 8.5.

Summary for Lay Audience

Climate change is expected to present significant challenges for society to overcome. Projection efforts aim to address these challenges by attempting to understand the most probable effects of climate change and their magnitude. Global climate models (GCM) are sophisticated models which are used by climate researchers in order to project future states of climate. The horizontal resolution of contemporary GCMs is approximately 100km due to computational limitations. This means that projections of variables, such as temperature, over a 100km x 100km area are averaged to a singular value; however, the temperature at any given set of points can vary significantly within this area. To address this, the process of downscaling is applied to GCM projections in order to increase their resolution. By increasing climate projection resolutions, local impact studies can be performed to assess effects of climate change on local scales. In this study, historical reanalysis climate data is downscaled via Physical Scaling (SP) models and Weather Research and Forecasting (WRF) models. WRF is a state of the art downscaling model which simulates the physical phenomena of the atmosphere and its interactions with water bodies and physical surfaces, but is limited by its computational requirements. SP is a novel downscaling methodology that uses statistical relationships between low resolution model-projected climate and local scale physical features and climate observations in order to downscale projected climate, at relatively low computational cost. SP models are found to significantly and moderately outperform WRF models in terms of surface temperature and precipitation downscaling accuracy, respectively, for a test study region over the Greater Houston Area. SP methodology is then chosen to downscale GCM projections across 44 urban regions within Canada and the USA; climate change impact is assessed via differences between projections for mild and extreme climate change scenarios (Representative Concentration Pathway (RCP) scenarios). Nearly half of all regions have significant projected increases in median and variance of surface skin temperature between RCP scenarios. Depending on GCM choice, median annual precipitation increases by the end of the century range from 29mm to 256mm.

Share

COinS