Evaluation of increasing spatial resolution in downscaled climate projections and the effect on extreme precipitation in Canada

Abstract:

Demand for projections of climate extreme events has arisen out of local infrastructure vulnerability assessments and adaptation planning. Global climate models (GCMs) are often too coarse in resolution to provide information specific to regional and local communities. To obtain the local data needed, statistical downscaling methods are frequently used to generate high resolution projections for individual assessments or areas.

One element of downscaling that is not often considered is the effect that downscaling has on the magnitude and distribution of changes in extreme events that are projected to occur in GCMs. Our objective is to explore whether or not projected anomalies in both average and extreme precipitation indices from GCMs are preserved after downscaling. To investigate this question we downscaled daily precipitation over Canada from 12 GCMs to 10 km resolution using a Bias Correction Constructed Analogues with Quantile Mapping approach. Downscaled precipitation fields are aggregated to GCM scale (150 km resolution) and the results compared against the driving climate model. We examine whether the downscaling process preserves the projected changes in the large-scale GCM spatial patterns and precipitation distributions using a set of field significance tests.

Comparisons between GCM derived projections and aggregated downscaled projections reveal no significance differences in annual and monthly quantities, however statistically significant differences are present for rarer events such as return periods. Results suggest the downscaling method inflates projected changes for parameters derived from the tails of precipitation distributions leading to larger projected changes in extreme events than those taken directly from the global climate models.