Inter-Variable Dependence in Climate Scenarios at Daily Resolution: Testing a 2D Statistical Adjustment of CMIP5 Simulations at Canadian Arctic Coastal Sites

English Abstract:

The interdependence between climatic variables is a critical feature which should be taken into account when developing local climate scenarios for the next decades. For example, in the Arctic the temperature-precipitation interdependence is particularly strong and determines other physical characteristics such as the snow cover extent and duration. However, this interdependence is often misrepresented in climate simulations. Here, we propose a two-dimensional (2D) statistical adjustment of climate model simulation data for the development of plausible local daily temperature and precipitation scenarios with more realistic inter-variable dependence. This method is based on the quantile mapping technique and furthermore it uses temperature time trend removal to obtain a stationary correction procedure and moving windows to adjust the correction to the specificities of each day of the year. This method was tested at 30 Canadian Arctic coastal sites, and was applied to an ensemble of 13 CMIP5 simulations (selected for representativeness of the variability of a larger ensemble). The inter-variable dependence was evaluated in terms of correlation values and with empirical conditional probability distributions of standardized temperature ranks given standardized precipitation ranks. The results show that the proposed 2D statistical adjustment corrects individual distributions of climatic time series as well as a standard one-dimensional (1D) statistical adjustment. Furthermore, the 2D adjustment outperforms raw and 1D corrected simulations in the reproduction of the observed temperature-precipitation interdependence. Where this interdependence is important (e.g. characterization of extreme events and of snow cover extent and duration), the proposed 2D adjustment definitely represents a good alternative for the production of plausible local climate scenarios.