Evaluating and Calibrating Dynamically Downscaled Precipitation Using the Doksum Shift Function

• Activity Number: 144
• Type: Invited
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Ida Scheel*+ and Elisabeth Orskaug and Ola Haug and Arnoldo Frigessi and Peter Guttorp
• Companies: University of Oslo and Norwegian Computing Center and Norwegian Computing Center and University of Oslo and University of Washington
• Keywords: Climate change ; Regional climate models ; Downscaling ; Precipitation ; Calibrating distributions ; Spatial hierarchical model

Abstract

Making robust and reliable climate projections available at a local scale is imperative for analyzing the impacts of climate change. Due to their comprehensive economic and security aspects, there is an urgent need to improve and extend downscaling techniques to produce high-resolution output in time and space.

We show that dynamically downscaled precipitation suffer from substantial shortcomings compared to the local observations. Generally, there are discrepancies between the dynamically downscaled precipitation and the distribution of actual precipitation on the whole range of data. Therefore, we propose a new calibration method for improving dynamically downscaled precipitation by calibrating the quantiles of the two distributions. We characterize the calibrating transfer function between the two distributions for each grid cell using the Doksum shift function. The transfer functions for all the grid cells are then modeled spatially through a Bayesian hierarchical model. The model is fitted for Norway for the period 1961-2000 using actual observed precipitation and dynamically downscaled data from the regional model HIRHAM, forced by the ERA-40