CCAM Numerical Weather Prediction Dr Marcus Thatcher Research

CCAM Numerical Weather Prediction Dr Marcus Thatcher Research Scientist December 2007 www. csiro. au

Overview Numerical Weather Prediction with CCAM Example: § Processing NCEP GFS analyses § Indonesia 60 km resolution 8 day forecast § Jakarta 8 km resolution 4 day forecast § Bali 8 km resolution 4 day forecast § Post processing forecast output CMAR NWP

Overview CCAM NWP forecasts are constructed in two stages The first stage creates a 60 km forecast for 8 days into the future The second stage downscales the 60 km forecast to 8 km resolution. Normally this is only done for 3 days into the future CMAR NWP

Downscaling CCAM uses nudging to ‘step down’ the forecast resolution from 60 km 8 km 1 km. Initial forecast 60 km Nudging 8 km Nudging 1 km CMAR NWP

GFS analysis Summary of the CCAM system Initial conditions User latitude/ longitude CMAR NWP Local terrain and vegetation CCAM Data Products (web pages, etc)

GFS analysis Summary of the CCAM system Initial conditions Topography and vegetation CCAM 60 km forecast Extract forecast data CCAM 8 km forecast Extract forecast data CCAM 1 km forecast Extract forecast data Downscaling User latitude/ longitude Topography and vegetation Downscaling Topography and vegetation CMAR NWP Data Products (web pages, etc)

CCAM NWP Example The scripts for the CCAM NWP example are located under: § $HOME/ccam/scripts/nwp The script that runs the simulation is called: § startforecast. sh CMAR NWP

CCAM NWP Example The startforecast. sh script performs four main functions: § Downloads NCEP GFS analyses and processes the analysis for initial conditions Download and prepare IC Run 60 km forecast § Runs a 60 km resolution forecast for Indonesia § Runs two 8 km forecasts that are nested in the 60 km forecast (Jakarta and Bali as examples) § Processes the output from CCAM so that it can be used CMAR NWP 8 km forecast Process model output

1) Downloading initial conditions CCAM can use various analysis products for initial conditions, including: § NCEP GFS 0. 5 deg § NCEP GFS 1 deg § Australian Bo. M GASP 1 deg § CMC 1 deg § NOGAPS 1 deg For this example we will use the NCEP GFS analysis CMAR NWP

1) Downloading initial conditions Summary: § NCEP GFS analysis is downloaded using getanalysis. sh § Process the GRIB file using procgfs 2. sh § CCAM initial conditions are then located in $HOME/ccam/scripts/nwp/obs/avn CCAM initial conditions can be inspected using Gr. ADS or Ferret CMAR NWP Download analysis (getanalysis. sh) Process GRIB (procgfs 2. sh) CCAM initial conditions

2) 60 km resolution forecast § Next, the CCAM 60 km resolution forecast is run. § The runall script simply starts the Indonesian forecast if the last forecast is older than the analysis § jog 48 indon contains all the information needed to run CCAM CMAR NWP Start forecasts (runall) Indonesia 60 km forecast (jog 48 indon)

2) 60 km resolution forecast § jog 48 indon converts the initial conditions to a conformal cubic (CC) grid using cdfvidar § Soil initial conditions are generated using smclim from a soil climatology dataset. It is also possible to use soil data from the last forecast. Convert initial conditions to CC (cdfvidar) § Pre-generated topography and land-use datasets are located in the ~/ccam/scripts/data directory Topography and land-use CMAR NWP Prepare soil initial conditions (smclim)

2) 60 km resolution forecast Then CCAM namelist file (called input) is prepared § dt = time step (20 mins) § nwt = output interval § ntau = total number of steps Some example CCAM namelist switches are provided § kdate_s = start date § ktime_s = start time § leap = use leap year? § io_in = interpolate initial conditions from input? § mfix = mass conservation § mfix_qg = moisture conservation CMAR NWP

2) 60 km resolution forecast § ifile = intial conditions file § mesonest = boundary conditions file § albfile = albedo file § zofile = roughness file § rsmfile = rsmin file § vegfile = vegetation file (Si. B) § soilfile = soil data file (Zobler) § ofile = CCAM output file CMAR NWP

2) 60 km forecast § After running CCAM, we need to convert the output back to a regular grid using cc 2 hist § Different output variables can be specified in the cc 2 hist namelist § It is also possible to obtain the output in pressure levels CMAR NWP Run CCAM (globpea. q 1) Process output (cc 2 hist)

3) cc 2 hist – processing CCAM output A typical cc 2 hist command looks like: § cc 2 hist –r 0. 5 ccout. nc llout. nc < cc. nml Where § -r determines the output resolution in deg § ccout. nc is the CCAM output (on the CC grid) § llout. nc is the CCAM output converted to a regular grid § cc. nml is the namelist that specifies the output variables and the output domain. Instructions for cc 2 hist can be obtained by: § cc 2 hist -h CMAR NWP

3) cc 2 hist – processing CCAM output Below is an example cc 2 hist namelist Output all timesteps &input kta=0, ktb=99999, ktc=-1 minlat = -20. , maxlat = -10. , minlon = 90. , maxlon = 120. use_plevs = T plevs = 1000, 900, 800, 700, 600, 500, 400, 300, 200 &end &histnl Output domain Use and define Pressure levels (instead of sigma levels) hnames = "temp", "u", "v", "psl", "rnd 24", "tscrn", "zs", "mixr", "zg", "tmaxscr", "tminscr" hfreq = 1, htype = "inst", hbytes=2 &end CMAR NWP Output variables (also just use “all”)

3) cc 2 hist – processing CCAM output from the 60 km forecast can be found at § $HOME/ccam/scripts/nwp/save/indon-0701/indon_60 km Once processed by cc 2 hist, we can examine the CCAM NWP forecast using Gr. ADS or Ferret It is also possible to examine the ‘raw’ CCAM 60 km forecast on the conformal cubic grid § $HOME/ccam/scripts/nwp/wdir/indon_60 km By looking at the ‘raw’ output shows what variables you can process with cc 2 hist CMAR NWP

4) 8 km nested forecasts § Once the 60 km forecast is complete, we can downscale to 8 km resolution forecast for multiple locations § For example, here we downscale to 8 km resolution forecasts for Jakarta and Bali § As before, the output is controlled by cc 2 hist CMAR NWP 60 km Indonesian forecast (jog 48 indon) 8 km Jakarta forecast (jog 48 jaka) 8 km Bali forecast (jog 48 bali)

4) 8 km nested forecasts For nested forecasts, the CCAM namelist is slightly different § mesonest = CCAM 60 km output filename § io_in = -1 to interpolate the 60 km BC to the 8 km CC grid § dt = 3 mins § nbd = -3 (far field nudging) § nud_uv = nudge winds § nud_p = nudge surface pressure § nud_t = nudge temperature § nud_q = nudge mixing ratio § nud_hrs = efolding time § kbotdav = lowest model level to nudge (1 = all levels) CMAR NWP

Sigma levels kbotdav=4 (typically NWP) kbotdav=10 (typically climate) CMAR NWP

5) cc 2 hist – processing 8 km CCAM output from the 8 km forecast can be found at § $HOME/ccam/scripts/nwp/save/indon-0701/jaka_8 km § $HOME/ccam/scripts/nwp/save/indon-0701/bali_8 km The ‘raw’ CCAM 8 km forecast on the conformal cubic grid is located at § $HOME/ccam/scripts/nwp/wdir/indon/jaka_8 km § $HOME/ccam/scripts/nwp/wdir/indon/bali_8 km CMAR NWP

CCAM output Typically the forecast is stored in § 3 hour (60 km) or § 1 hour (8 km) intervals. This is because the radiation scheme is normally updated once every hour. The default areas of the forecast are: § 60 km ± 15 deg ~ ± 1700 kms § 8 km ± 2 deg ~ ± 220 kms § 1 km ± 0. 25 deg ~ ± 30 kms Since CCAM is a global model, the output can be also global. However, usually the output is for the high resolution cubic panel only CMAR NWP

CCAM output Typical output area CMAR NWP

GFS analysis Example forecast system GFS download and archive system Generator for client data products Data integrity system Operational forecasting system Hindcast system CCAM Validation and verification system Delivery platform (web pages, FTP, etc) CMAR NWP Archive and synchronization with parallel forecast systems Data store

CCAM Validation (Valencia) Wind speed and direction CMAR NWP

Marine and Atmospheric Research Name Dr Marcus Thatcher Title Research Scientist Email Marcus. Thatcher@csiro. au Thank You Contact CSIRO Phone Email Web 1300 363 400 +61 3 9545 2176 enquiries@csiro. au www. csiro. au