1. High resolution extreme temperature scenarios over North America NARCCAP 4 th users’ workshop Apr. 10 -11, 2012 Guilong Li Atmospheric Science and Application Unit Xuebin Zhang Climate Research Division Environment Canada Francis Zwiers Pacific Climate Impacts Consortium

2. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 Outline Introduction Methodology Results Conclusions

3. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 Introduction – Objective Construct high resolution extreme temperature over North America Estimate uncertainty in the projected return value Partition uncertainty into different sources

4. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 GCM RCM GFDLCGCM3HADCM3CCSM CRCM --finished--finished ECPC finished-- HRM3 -- finished-- MM5I -- finished RCM3 finished -- WRFG --finished--finished Extreme temperature from RCM (NARCCAP) Introduction – Data Seasonal mean temperature from GCM (PCMDI) –23 GCMs, resolution 100 – 400km, 1961-2099 –2 emission scenarios – A2 and B1 –38 runs from SRES-A2 and 44 runs from SRES-B1

5. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 Introduction – Data treatment All GCMs and RCMs are interpolated to CRCM grid points Inverse distance for GCM seasonal mean temperature –Four surrounding points Nearest assignment for RCM extreme temperature –ECPC, RCM3 and WRFG to CRCM –Over 90% of the grid points are within 45km Remove 1971-2000 climatology –CRCM, ECPC, RCM3, WRFG and corresponding driven GCM –All GCMs from PCMDI

6. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 Where y – RCM extreme temperature anomaly µ – location parameter σ – scale parameter ξ – shape parameter x – GCM seasonal mean temperature anomaly Methodology – GEV µ σ ξ

7. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 Methodology – Probability projection

8. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 Difference ratio: R S : Return value from statistical downscaling R D : Return value from dynamical downscaling Statistical downscaling CRCM/CGCM3 to CGCM3 Dynamical downscaling CRCM/CGCM3 Statistical downscaling CRCM/CGCM3 to GFDL Dynamical downscaling RCM3/GFDL Result – Model Validation Statistically and dynamically downscaled extreme temperatures 20-year return value difference ratio Annual minimumAnnual maximum

9. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 Results – Annual minimum temperature 20-yr return value change 10 th percentile Median 90 th percentile 2011-2040 2041-2070 2071-2099

10. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 Results – Annual maximum temperature 20-yr return value change 10 th percentile Median 90 th percentile 2011-2040 2041-2070 2071-2099

11. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 2011-2040 2041-2070 2071-2099 Results – Annual maximum temperature change (median)

12. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 Results – Source of Uncertainty Annual MinimumAnnual Maximum 5-yr10-yr20-yr50-yr 20-yr10-yr5-yr

13. NARCCAP 4 th users’ workshop: Apr. 10-11, 2012 Conclusions A framework was constructed by using combined dynamical and statistical downscaling methods to produce high resolution extreme temperature scenarios over North America Multiple GCMs and RCMs relationships were applied to CMIP3 GCM simulations for emulating RCM simulations Uncertainty from GCM, model parameters, internal variability, and downscaling from low resolution to high resolution were estimated Provide a product with high resolution annual maximum and minimum temperature change and uncertainty