Bias Correction of Precipitation Extremes Conditioned on Synoptic Weather Patterns.
Li, J., F. Johnson, A. Sharma and J.P. Evans
MODSIM2015, 21st International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand, Gold Coast, Australia, 29 November – 4 December 2015.
When using the precipitation extremes simulated by a regional climate model (RCM) in climate
impact studies, the bias has to be first corrected. Commonly used bias correction methods are designed to match
statistics of the simulated precipitation with observations. This study proposes a new approach to account for
the potential change of different precipitation types; the approach accounts for different precipitation
mechanisms having different bias characteristics. Different precipitation types are identified by self-organizing
map classification of the weather synoptic patterns. The rationale for using synoptic patterns to classify the
extreme precipitation is that these large-scale circulation patterns may be better simulated by the RCM than
precipitation data itself.
Considering the simulations of a very high resolution Weather Research and Forecasting (WRF) RCM for a
domain over south-eastern Australia, a slight change in the proportions of the synoptic weather patterns was
found, indicating a possible change in the composition of the total extreme precipitation in the future. A new
Synoptic Pattern Bias Correction (SPBC) approach was developed that could account for these changes. The
SPBC method lead to marginally different results compared to a conventional bias correction method, namely
To understand under what conditions significant differences will result between the SPBC method and quantile
mapping a comprehensive synthetic study has been defined. The properties of the bias, the changes in the
synoptic patterns and the differences in the rainfall amounts from the synoptic patterns were among some of
the factors that were explored in a synthetic study. From over 600,000 synthetic cases, 46% were found to have
significant differences in the future simulations from the two bias correction methods. It was also found that
the differences between the methods depends on several factors including the change in proportion of each
precipitation type, the difference in the correction factor for different precipitation types and the magnitude of
the overall change in precipitation extremes. Among these factors, the between-cluster difference in the
correction factor seems to dominate the significance of the results.
This page is maintaind by Jason Evans |
Last updated 31st January 2013