The western Canada high resolution WRF simulation Yanping Li S. Kurkute E. Asong L. Chen University of Saskatchewan CCRN SOAP Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 1 / 17
Goals and Objective Perform convection-permitting regional climate simulation of western Canada using 4-km WRF-RCM Study the convection initiation mechanism in Saskatchewan River Basin (SRB) by coupling WRF-RCM with land-surface model (Noah-MP). Assessment of extreme weather events, study synoptic and meso-scale features of future climate using pseudo-global warming (PGW) method. Implement and apply the physically-based, dynamical downscaling of CMIP5 using 4-km WRF for assessing global warming impacts on the water resources and regional climatology of Saskatchewan and McKenzie river basins (SRB and MRB resp.) Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 2 / 17
Various RCM s over Western Canada (a) 80KM CanRCM (b) 25KM using CRCM5 Université du Québec á Montréal (c) 10KM WRF-ARW University of Toronto (d) 4KM WRF-ARW by CCRN Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 3 / 17
WRF model Set-Up (Control experiment for 2000-2013) WRF model (Version 3.4.1) Single domain: 2560 x 2800 km 2, x =4km, 37 vertical levels Microphysics Scheme : New Thompson et al. PBL Scheme : YSU CAM long-wave and short-wave scheme No cumulus parameterization assumed explicit convection Forcing Data 6 hourly, 0.703 o x0.703 o resolution ERA-Interim reanalysis data provides initial and lateral boundary conditions Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 4 / 17
WRF Performance OBS, Anuspline, WRF Figure: Spatial Distribution of annual mean daily PRCP for 2007-2012 resp. Figure: Annual mean daily error (mm/day) for WRF vs OBS and WRF vs ANUSPLINE resp. Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 5 / 17
Mean seasonal precipitation Figure: Mean seasonal precipitation, Oct 2000 Oct 2013. The upper (lower) polygons represent McKenzie (Saskatchewan) River Basins Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 6 / 17
Mean seasonal daily Tmin and Tmax Figure: Mean seasonal daily Tmin (a) and Tmax (b), Oct 2000 Oct 2013 Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 7 / 17
Wet-Dry Day Frequency Figure: Wet-dry day frequency in terms of maximum number of consecutive dry (a) and wet days (b) for winter (DJF) and summer (JJA) season during Oct 2000 Oct 2013. Threshold for distinguishing precipitation occurrence = 1mm Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 8 / 17
Seasonal Anomaly Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 9 / 17
Annual Cycle Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 Figure: Annual Cycle of PRCP rate from Oct 2000 Oct 2013 10 / 17
Pseudo-Global Warming Method Calculate 25-year monthly mean values of U, V, T, Qv, PSFC, and SST of current and future climate periods from multi CMIP5 model ensemble (1975-1999 historical and 2075-2099, RCP8.5) Subtract current from future to get monthly climate perturbations global warming signal Add time-interpolated perturbation to current reanalysis (ERA-Interim, 6-hourly) to give new WRF model initial and lateral boundary conditions Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 11 / 17
Pseudo-Global Warming Method Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 12 / 17
WRF Simulations Computing Resources Requirement Compute-Canada MP2-Mammoth and NCAR Yellowstone HPS Clusters One month simulation need 96 hours ( 4days) plus( 2days) waiting period on Compute-Canada HPC At NCAR Yellowstone: (960 nodes X 16 CPU/node X 12 hours walltime) ( 1 day required for 1 month simulation) At Compute Canada MP2: (25 nodes X 24 CPU/node X 96 hours walltime) Total computing resources required for 13 years of PGW and CTRL simulations each requires 1,550 core years total computing resources (NOTE: Compute Canada allocates 80 core years default computing resource). At least 160TB of storage space. Additional Computing resources required for performing the post-processing of the data to perform basic statitsics and compute extreme and climate indices. Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 13 / 17
WRF Simulations Current Status Time required to finish the simulations : 300 continuous days (no weekends no holidays) Historical runs for the period 2000-2013 has completed PGW runs 2000-2013 has completed equivalent to 2071-2099; Another 300 days Major man-power involved Yanping Li (lead) Liang Chen (prepare the forcing field, conduct simulation) Sopan Kurkute (conduct simulation, post processing, tech support for data application and transfer) Elvis Asong (statistical analysis for WRF output) Lucia Sca, Zhe Zhang (conduct part of the simulation) NCAR scientists (experts who help with the technique and computing resources) CCRN Theme D atmospheric group Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 14 / 17
WRF Simulations Data Access Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 15 / 17
Convert WRF to R2C (CLASS/MESH) Source Code Location on water server /data2/nobackup/sak298/wrf2rc/version_3.0/ Example Script: START_YEAR=2002 END_YEAR=2013 export xdelta=0.12000 export ydelta=0.12000 export xcount=100 export ycount=100 export xorigin=-114 export yorigin=48 export MONTHS_Start=12 export MONTHS_End=12 export IN_DIR="/data2/NOBACKUP/sak298/Post_processed/" export CONST_DIR="/data2/NOBACKUP/sak298/Fuad_Data_2006" Output="/data2/NOBACKUP/sak298/Fuad_Data_2006/WRF_R2C/ \\ ${YEAR}" source_dir="/data2/nobackup/sak298/wrf2rc/version_3.0/" Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 16 / 17
Acknowledgement GIWS and CCRN NCAR (Computer resources and technique support) Environment Canada (X. Zhang, K. Szeto, B. Bonsal) Compute Canada Prof. Howard Wheater Branko Zdravkovic For WRF to MESH/CLASS data conversion prpgramm and WRF model related support please contact Sopan Kurkute, GIWS kurkute.sopan@usask.ca Yanping Li, S. Kurkute, E. Asong, L. Chen High Resolution WRF-RCM October 3, 2016 17 / 17