11.3
Improving regional climate change projections of temperature for Halifax, Nova Scotia via a new statistical downscaling approach
Matthew Lee Titus, Meteorological Service of Canada, Dartmouth, NS, Canada; and I. Folkins, R. Greatbatch, and J. Sheng
In order to best assess the expected climate change impacts on a species, ecosystem or natural resource in a region, climate variables and climate change scenarios must be developed on a regional or even site-specific scale (Wilby et al, 2002). To provide these values, projections of climate variables must be downscaled from the GCM results, utilizing either dynamical or statistical methods (IPCC, 2001).
This study proposes a new statistical downscaling approach as an improvement on a current technique. This proposed new method of regression development is compared with the regression achieved via the Statistical Downscaling Model Software (SDSM; Wilby et al). Observed daily maximum temperature (Tmax, the predictand), taken from Shearwater airport (used as a proxy for Halifax, NS) was selected as the variable to be downscaled in winter (DJF). The predictand and predictors (taken from the NCEP Reanalysis) were turned into Z scores. The seasonal cycle was removed from each to get the seasonal anomalies. Then a predictor selection process was employed to remove predictors that are not useful or redundant. Next, the principal components (PC's) of the predictors from NCEP were calculated for the historical period (1961-2000). The regression was trained on the PC's from 1961-1990 and then validated by predicting 1991-2000 Tmax. For comparison, downscaling was done with SDSM for the same location and time period. The SDSM regression was again trained on the 1961-1990 data and then validated on the 1991-2000 period. The observed seasonal cycle was removed from the SDSM downscaled Tmax to obtain the seasonal anomaly.
The Tmax seasonal anomaly for winter (1961-2000) projected by both the new method and SDSM were compared against observations. The new method exhibited a correlation of 0.86 with observations. This is almost a 0.3 increase in the correlation value between SDSM and observations. This indicates that the new method is a major improvement to the SDSM software method.
Session 11, Regional Modeling and Downscaling Part II
Thursday, 21 January 2010, 11:00 AM-12:15 PM, B215
Previous paper Next paper