CPT based Sub-Seasonal Forecasting Scripts
NOAA’s CPC International Desks
NOAA-USAID 11ITWCVP - Ankara, Turkey, 15 – 26 April 2019
Outline
Download the CFS (NCEP) S2S Data from the IRIDL server
Generate maps of Raw forecast and associate skill
Produced the calibrated forecast o Make the data ready for CPT
o Calibrated using CPT: Maps of calibrated Forecast and associate skill maps
Generate map of observed anomalies for eye ball verification
The current version is valid only for the precipitation . Future release will include temperature
Copying and Unpacking
1. Make sure that you have copy of S2S_CPT_FCST.tar.gz file in your cygwin/linux home folder.
A copy is available at following address:
https://ftp.cpc.ncep.noaa.gov/International/11ITWCVP_Ankara2019/S2S_CPT_FRCST.tar.gz
2. From your cygwin/Linux terminal, uncompress the file:
tar -xzvf S2S_CPT_FCST.tar.gz
3. Go to the folder S2S_CPT_FCST : cd S2S_CPT_FCST
4. List to see his content :
ls
Download the NCEP CFS S2S data (forecast, hindcast) and the historical observed data
bash get_S2S_data.sh idate tgtprd obsbse
idate - Forecast initiation date in the format YYYYMMDD (eg. 20190211)
tgtprd - Forecast target period. It can be "5days" "10days" "week1" "week2" "week34"
obsbse - Source’s name of the observed data. Valid option are "cpcuni" or "arc2"
The downloaded data (forecast, hindcast, historical observed) are in the folder : S2S_DATA/idate/trgtprd/grads_data
The processing time depends on your internet connection speed and your computer processor speed.
Make the data ready for CPT :
Converting from binary to cpt format
bash Convert_S2S_bin2cpt.sh idate tgtprd obsbse ylatS ylatN ylonW ylonE
idate - Forecast initiation date in the format YYYYMMDD (eg. 20190211)
tgtprd - Forecast target period. It can be "5days" "10days" "week1" "week2" "week34"
obsbse - Source’s name of the observed data. Valid option are "cpcuni" or "arc2"
ylatS - Southern boundary of the area of interest
ylatN - Northern boundary of the area of interest
ylonW - Western boundary of the area of interest
ylonE - Eastern boundary of the area of interest
The converted data (forecast, hindcast, historical observed) are in the folder : S2S_DATA/idate/trgtprd/cpt_data
Files names are :
model_fcst_trgtprd.tsv model_hdcst_trgtprd.tsv obs_hist_trgtprd.tsv
They can be used within windows version of CPT
Make the data ready for CPT (cntd):
Converting from binary to cpt format (Examples)
bash Convert_S2S_bin2cpt.sh 20190211 10days cpcuni -59.75 59.75 -179.75 179.75 bash Convert_S2S_bin2cpt.sh 20190211 week1 cpcuni -59.75 59.75 -179.75 179.75 bash Convert_S2S_bin2cpt.sh 20190211 week2 cpcuni -59.75 59.75 -179.75 179.75 bash Convert_S2S_bin2cpt.sh 20190211 week34 cpcuni -59.75 59.75 -179.75 179.75
Calibrating CFS S2S forecast using CPT :
bash CPT_S2S_Calib.sh idate tgtprd mthds obsbse ylatS ylatN ylonW ylonE xlatS xlatN xlonW xlonE
idate - Forecast initiation date in the format YYYYMMDD (eg. 20190211)
tgtprd - Forecast target period. It can be "5days" "10days" "week1" "week2" "week34"
mthds - The calibration method. Valid option are "GCM", "CCA" or "PCR"
obsbse - Source’s name of the observed data. Valid option are "cpcuni" or "arc2"
ylatS , ylatN, ylonW , ylonE are respectively the extend of the predictand ie the geographical coordinates of the area of interest
xlatS , xlatN, xlonW , xlonE are respectively the extend of the predictor ie the geographical coordinates of the domain to consider for the model
The calibrated forecast map and the associated skill map are in the folder : Figures_SubX_Frcst_Calib/init_idate/By_mthds
Under the name :
NCEP_trgtprd_fcst_mthds_calib.png ; the map of calibrated forecast (probabilistic) PC_skill_map_trgtprd_fcst_mthds_calib.png ; the associate Pearson Correlation skill map
Calibrating CFS S2S forecast using CPT (cntd): Examples
1. bash CPT_S2S_Calib.sh 20190211 week2 GCM cpcuni -59.75 59.75 -179.75 179.75 -90 90 0 360
2. bash CPT_S2S_Calib.sh 20190211 week2 CCA cpcuni -59.75 59.75 -179.75 179.75 -90 90 0
360
Case 1 Case 2
Figures_SubX_Frcst_Calib/init_20190211/By_GCM Figures_SubX_Frcst_Calib/init_20190211/By_CCA NCEP_week2_fcst_GCM_calib.png
PC_skill_map_week2_fcst_GCM_calib.png NCEP_week2_fcst_CCA_calib.png
PC_skill_map_week2_fcst_CCA_calib.png
Generating the raw forecast maps – Any differences with the Calibrated forecasts?
converting deterministic forecast into probabilistic map using model climatology
bash plot_S2S_Raw_frcst.sh idate tgtprd obsbse ylatS ylatN ylonW ylonE
idate - Forecast initiation date in the format YYYYMMDD (eg. 20190211)
tgtprd - Forecast target period. It can be "5days" "10days" "week1" "week2" "week34"
obsbse - Source’s name of the observed data. Valid option are "cpcuni" or "arc2"
ylatS - Southern boundary of the area of interest
ylatN - Northern boundary of the area of interest
ylonW - Western boundary of the area of interest
ylonE - Eastern boundary of the area of interest
This will generate two maps under the Folder:
Figures_SubX_Frcst_Calib/init_idate/RAW_FCST
1- NCEP_trgtprd_RAW_FCST.png : the raw forecast probabilistic map
2- PC_skill_map_trgtprd_RAW_FCST.png : the associate Pearson Correlation skill map
Generating the raw forecast maps – Any differences with the Calibrated forecasts? (cntd) : Examples
converting deterministic forecast into probabilistic map using model climatology
bash plot_S2S_Raw_frcst.sh 20190211 week2 cpcuni -59.75 59.75 -179.75 179.75
The above command line will generate two file under the Folder:
Figures_SubX_Frcst_Calib/init_20190211/RAW_FCST Those files are :
NCEP_week2_RAW_FCST.png
PC_skill_map_week2_RAW_FCST.png
Plotting observed anomalies for verification
bash plot_obs_for_verif.sh idate tgtprd obsbse ylatS ylatN ylonW ylonE
idate - Forecast initiation date in the format YYYYMMDD (eg. 20190211)
tgtprd - Forecast target period. It can be "5days" "10days" "week1" "week2" "week34"
obsbse - Source’s name of the observed data. Valid option are "cpcuni" or "arc2"
ylatS - Southern boundary of the area of interest
ylatN - Northern boundary of the area of interest
ylonW - Western boundary of the area of interest
ylonE - Eastern boundary of the area of interest
This action will generate the containing file the observed anomalies map. It can be found in the folder : Figures_SubX_Frcst_Calib/init_idate
The file name would look like:
obsbse_anom_obs_trgtprd.png
The processing time depends on your internet connection speed
Plotting observed anomalies for verification (cntd) – Examples
bash plot_obs_for_verif.sh 20190211 week2 cpcuni -59.75 59.75 -179.75 179.75
This action will generate the containing file the observed anomalies map. It can be found in the folder : Figures_SubX_Frcst_Calib/init_20190211
The file name is : cpcuni_anom_obs_week2.png
CPT based Sub-Seasonal Forecasting Scripts – Summary
1. cd S2S_CPT_FCST
2. bash get_S2S_data.sh idate tgtprd obsbse
3. bash plot_S2S_Raw_frcst.sh idate tgtprd obsbse ylatS ylatN ylonW ylonE 4. bash Convert_S2S_bin2cpt.sh idate tgtprd obsbse ylatS ylatN ylonW ylonE 5. bash CPT_S2S_Calib.sh idate tgtprd mthds obsbse ylatS ylatN ylonW ylonE
xlatS xlatN xlonW xlonE
6. bash plot_obs_for_verif.sh idate tgtprd obsbse ylatS ylatN ylonW ylonE Where :
idate - Forecast initiation date in the format YYYYMMDD (eg. 20190211)
tgtprd - Forecast target period. It can be "5days" "10days" "week1" "week2" "week34"
mthds - The calibration method. Valid option are "GCM", "CCA" or "PCR"
obsbse - Source’s name of the observed data. Valid option are "cpcuni" or "arc2"
ylatS , ylatN, ylonW , ylonE are respectively the extend of the predictand ie the geographical coordinates of the area of interest
xlatS , xlatN, xlonW , xlonE are respectively the extend of the predictor ie the geographical coordinates of the domain to consider for the model
The processing time depends on your internet connection speed and your computer processor speed.
The current version is valid only for the precipitation . Future releases will include temperature
Hands on Tools :
Practice CPT based Sub-seasonal Forecasting (1)
Initiation
date Target period Calibration
Method Observed
data Target area
(
Predictand domain) Model area (
Predictor domain)
Feb 11, 2019 10days, week1,
week2, week34 GCM, CCA, PCR cpcuni
Africa, Europe and Middle East [40S-60N / 20W-60E]
[60S-60N / 0-360E]
Asia and Maritime Continent [40S-60N / 60E-179.75E]
Central and South America [60S-30N / 120W-30W]
For this initiation date the data are already provide.
So on regarding slide 13, skip step 2 and proceed accordingly with the others steps.
Depending on your geographical belonging produce the forecast maps using the
parameters described in the table below.
Hands on Tools :
Practice CPT based Sub-seasonal Forecasting (2)
Initiation
date Target period Calibration
Method Observed
data Target area
(
Predictand domain) Model area (
Predictor domain)
Pick a start date in the past.
Just make sure that the target period is in the past so you can plot the observed anomalies.
10days, week1,
week2, week34 GCM, CCA, PCR cpcuni
Africa, Europe and Middle East [40S-60N / 20W-60E]
[60S-60N / 0-360E]
Asia and Maritime Continent [40S-60N / 60E-179.75E]
Central and South America [60S-30N / 120W-30W]