Rasmus E Benestad

clim.pact-V.1.0

An analysis tool, clim.pact , has been developed for empirical downscaling and analysis of local and regional climate on both a daily and a monthly basis. This tool is built in the data analysis environment R, and consists of a contributed package. The R-software is freely available (A "GNU version of Splus") from the Internet (URL: http://cran.r-project.org), and runs on both Linux and Windows platforms. Since R and Linux require no license, this approach provides a low-budget tool for data-analysis. The clim.pact package is tested with both monthly and daily data and a number of different scenarios are compared. Empirical downscaling has been carried out with clim.pact on January mean temperature in Bergen, Copenhagen, Helsinki, Oslo, Stockholm, and Tromsø, using single field as well as mixed-field predictors and with predictors that cover different regions. The package can easily employ a new common EOF approach for empirical downscaling. A comparison has been made between the various results in order to study the robustness of the method. Furthermore, the residuals from the model calibration has been examined for remaining trends or biases. These tests indicate that clim.pact skillfully reproduces local climate variations with a linear model when these are near-normally distributed. Tests with skewed data such as daily precipitation show that a linear model does not give a good description of the local series. New models, however, are easily incorporated in the clim.pact framework, such as analogs and conditional weather generators. The empirical downscaling is fast, and clim.pact hence facilitates downscaling analysis on multi-model ensembles.

Keywords: Downscaling analysis; Daily values; Monthly values; common EOFs