Climate Processes in CMIP5: Cyclone-Relative Diagnostics to Better Understand CMIP5 Model Performance and Future Extratropical Cyclone Changes

With the availability of outputs from the Coupled Model Intercomparison Project Phase 5 (CMIP5) simulations, it is important to evaluate the performance of these models before utilizing them in various investigations. In our previous work, we evaluated the extratropical cyclone changes in the CMIP5 models over the western Atlantic and eastern North America during the cool season (from November to March) from the historical period (1979-2004) to the future (RCP8.5, 2009-2098). For the middle and late 21st century, we have found that there will be a decrease in the number of cyclones over the western Atlantic storm track, while there will be a slight increase in both cyclone number and intensity over the eastern North American coast. However, there are large uncertainties among models in cyclone track density and some other important variables (i.e. precipitation) that need to be quantified. Also, the impacts of cyclone changes on those important variables, like precipitation, low-level wind speed, and near-surface temperature also needs to be determined. In this study, we have developed a series of cyclone-relative diagnostics to better understanding of CMIP5 model performance and future extratropical cyclone changes.

In the cyclone-relative approach, we first track the extratropical cyclones for the CMIP5 models using the Hodges (1994, 1995) automated tracking algorithm applied to 6-hourly sea level pressure. After identifying the cyclone centers, we define a cyclone-relative box around each cyclone center, and the box moves with the cyclone center. The variables within the box are extracted and put back to the map according to its geographical position. Then we can statistically calculate the cyclone-relative results at any grid point. For example, if we want to calculate cyclone-relative precipitation accumulation for a cool season, only the precipitation within the cyclone-relative box (within the radius of 10 degree around each cyclone center) will be extracted and calculated. This approach provides a more accurate estimate of the variables associated with cyclones than averaging CMIP5 variables within a fixed domain.

Preliminary results show that for the cool season during both the historical and the future periods, cyclone-relative precipitation accounts for over 70% of the total precipitation inland over eastern North America. This indicates that the future change (an increase about 10%) of precipitation over the heavily populated East of North America is largely due to the cyclone precipitation enhancements. In addition, the large uncertainty among models for precipitation for both historical and future time periods has been diagnosed by this approach. For example, about 65% of the difference in total precipitation between the HadGEM2-ES and CCSM4 models originates from cyclone systems. The differences in precipitation among models are mainly attributable to the differences of individual models in simulating the cyclone activity. We will also discuss more about the application of this cyclone-relative diagnostic, i.e., sea level pressure, low-level wind speed and the near-surface temperature in model simulations, which helps explain performance differences in the CMIP5 models during the historical period. The dynamical processes associated with the large-scale circulation (i.e., upper-level jet and the eddy growth) will be investigated to explore the reasons for the future cyclone change with the models.