APEC Climate Center

Climate Information Services

CLIK Information
Dataset
- MME-3MON
- MME-6MON
- MME-MODEL
- BSISO
- CMIP5
- ERA5
- NCEP Reanalysis
- CHFP
Processing
- Prediction
- Verification
- Masking
- AIMS
CLIK API
Documents
- Online Tutorial

Dataset CHFP

Mail to jhshin@apcc21.org

Overview
Download

The Climate-system Historical Forecast Project (CHFP) is an international initiative organized under the Working Group on Subseasonal to Interdecadal Prediction (WGSIP), a working group of the Earth System Modeling and Observations (ESMO) core project within the World Climate Research Programme (WCRP). It provides a resource for seasonal prediction research, in the form of historical (hindcast) predictions from models that integrate the atmosphere, oceans, land surface, and cryosphere components of the climate system. By WGSIP consensus, the CHFP database has been transferred from its original home at the Centro de Investigaciones del Mar y la Atmósfera (CIMA) in Argentina to the APEC Climate Center (APCC).

Now hosted at APCC, the platform preserves the historical CHFP archive and integrates hindcast data from the operational forecasting systems contributing to the WMO Lead Centre for Seasonal Prediction Multi-Model Ensemble (WMO LC-SPMME), through coordination between WGSIP and the WMO operational counterparts with the agreement of 14 Global Producing Centres. Co-hosting these two distinct eras of climate modeling, the portal functions as a diachronic reference database for the evolution of seasonal forecasting by global climate models.

CHFP data
WMO LC-SPMME

The CHFP dataset is a historical reference archive of global climate model seasonal hindcasts collected from 12 modeling systems. Its purpose is to characterize the potential predictability of the coupled climate system and provide a testbed for evaluating multi-model ensemble forecasts under similar initial conditions.

Historical Context

Origins and the Barcelona Workshop (2005–2007): The experiment was developed by the WCRP Task Force on Seasonal Prediction (TFSP), established in 2005, as a "total climate system" seasonal prediction experiment. Its design was further refined through discussions at the First WCRP Seasonal Prediction Workshop (Barcelona, June 2007), after which the TFSP mandate passed to WGSIP.
Framework for Systematic Initialization: The framework was designed to address multi-model structural uncertainties by examining how the initialization of major climate components — including land surface, stratospheric dynamics, and sea ice — affects seasonal predictability across modeling systems.
Open-Access Architecture: A standardized central data server was originally established at CIMA in Buenos Aires, Argentina, to make the historical hindcast archive openly available to the international research community for diagnostic and ensemble studies.

Scientific Significance & Core Framework

Integrated Climate Variability: Moving beyond traditional SST-centric predictable frameworks, CHFP provides a standardized platform to evaluate the complex influences of soil moisture, snow cover, vegetation, sea ice, stratospheric processes, and atmospheric composition on regional climate anomaly predictions.
Bridging Research Communities: This portal effectively breaks down institutional barriers between the 'Climate Change' research community and the operational 'Seasonal Prediction' community, allowing scientists to directly evaluate earth system models in a seasonal forecasting configuration.

Data Specification & Standardization

Standard Grid System: Atmospheric fields are regridded to a standard 2.5° × 2.5° resolution; ocean fields to 1° × 1°.
Data Format: All datasets adopt the NetCDF (network Common Data Form) standard, maximizing analysis convenience and software compatibility.

Additional Information & Reference

For institutional history, experimental setup, and updates on long-range forecast initiatives, see the ESMO framework page
- WCRP ESMO CHFP (WGSIP) Official Website

and the journal publication
- The Climate-System Historical Forecast Project (Tompkins et al., BAMS 2017)

CHFP data
WMO LC-SPMME

Each NetCDF file in this archive packages a single variable with all ensemble members for one initial-condition date from one modeling system, organized under a hierarchical directory structure based on domain (atmosphere/ocean), level (pressure levels/surface), and temporal frequency (daily/monthly). This directory layout follows the structure of the original CHFP database without modification.

Quick Start

Download a single hindcast file with wget. The example below retrieves the February 1979 initial-condition file of monthly air temperature on pressure levels from CCCma-CanCM3 (all pressure levels and ensemble members in one NetCDF file).

#!/bin/bash wget https://apcc21.org/clikapi/download/CHFP_HOME/CHFP/atm/levels/monthly/ta/ta_monthly_CCCma-CanCM3_CHFP_19790201.nc

Directory Architecture & Naming Rules

Each NetCDF file contains a single variable for one initial-condition date. Directory hierarchy is domain → level → frequency → variable → file.

https://apcc21.org/clikapi/download/CHFP_HOME/CHFP/[Domain]/[Level]/[Frequency]/[Variable]/[File name]

Table 1. Directory architecture
Directory	Detail
Domain	atm (atmosphere) ocn (ocean)
Level	levels (pressure levels) surface
Frequency	daily (ocn domain has no daily data) monthly
Variable	variable short-name (see table 3 below)

File name: [Variable]_[Frequency]_[Model identifier]_[YYYYMMDD].nc

Table 2. File naming rules
Component	Detail	Example (ta_monthly_CCCma-CanCM3_CHFP_19790201.nc)
Variable	variable short-name (see table 3 below)	ta (pressure-level air temperature; all levels bundled)
Frequency	daily monthly	monthly
Model identifier	full model string (see table 3)	CCCma-CanCM3_CHFP
YYYYMMDD	Initial-condition date; day part is always 01	19790201 (Feb 1, 1979)

Automated Batch Retrieval via Wget

Two common retrieval patterns: recursive wget for all IC dates of one variable/model, or a multi-year loop over explicit IC months.

#!/bin/bash # ===== CHFP Hindcast Batch Retrieval ===== BASE="https://apcc21.org/clikapi/download/CHFP_HOME/CHFP" # --- Multi-year loop for one variable, model, and IC month --- # Example: monthly 'ta' from CCCma-CanCM3, February IC, 1979–
2010 MODEL="CCCma-CanCM3_CHFP" VAR="ta" FREQ="monthly" DOMAIN_PATH="atm/levels" IC_MONTH="02" for YEAR in {1979..2010} do IC_DATE="${YEAR}${IC_MONTH}01" FILE_NAME="${VAR}_${FREQ}_${MODEL}_${IC_DATE}.nc" URL="${BASE}/${DOMAIN_PATH}/${FREQ}/${VAR}/${FILE_NAME}" echo "Fetching CHFP hindcast file: ${FILE_NAME}..." wget -q "${URL}" -O "./${FILE_NAME}" done echo "Batch data ingestion complete."

Multi Model Hindcast Summary

CHFP contains hindcast data from 12 modeling systems (21 model identifiers). The model identifier column shows the exact string used in directory paths and filenames. Models with both atmosphere and ocean data are split into separate rows by domain; each variable cell lists the short-names available in the corresponding level/frequency branch.

Table 3. CHFP hindcast data summary
Model Identifier	Domain	Period	IC Months	Ensemble Members	Lead (mo.)	Pressure Level Variables daily [d] / monthly [m]	Surface Variables daily [d] / monthly [m]
`ARPEGE_z00k`Météo-France	atm	1979– 2007	05 08 11	11	4	[d]— [m]`g` `hus` `ta` `ua` `va`	[d]— [m]`prlr` `psl` `ts`
`ARPEGE_z00l`Météo-France	atm	1979– 2007	05 08 11	11	4	[d]— [m]`g` `hus` `ta` `ua` `va`	[d]— [m]`prlr` `psl` `ts`
`CCCma-CanCM3_CHFP`CCCma^A	atm	1979– 2010	02 05 08 11	10	12	[d]`g` `hus` `ta` `ua` `va` [m]`g` `hus` `ta` `ua` `va`	[d]`clt` `hflsd` `hfssd` `mrsov` `prlr` `psl` `rlds` `rls` `rlt` `rsds` `rss` `rst` `snld` `tas` `tasmax` `tasmin` `ts` `uas` `vas` [m]`clt` `hflsd` `hfssd` `mrsov` `prlr` `psl` `rlds` `rls` `rlt` `rsds` `rss` `rst` `snld` `tas` `tasmax` `tasmin` `ts` `uas` `vas`
`CCCma-CanCM3_CHFP`CCCma^A	ocn	1979– 2008	02 05 08 11	10	12	[d]n/a [m]`so` `thetao` `uo` `vo` `wo`	[d]n/a [m]`zmlo` `zoh`
`CCCma-CanCM4_CHFP`CCCma^A	atm	1979– 2010	02 05 08 11	10	12	[d]`g` `hus` `ta` `ua` `va` [m]`g` `hus` `ta` `ua` `va`	[d]`clt` `hflsd` `hfssd` `mrsov` `prlr` `psl` `rlds` `rls` `rlt` `rsds` `rss` `rst` `snld` `tas` `tasmax` `tasmin` `ts` `uas` `vas` [m]`clt` `hflsd` `hfssd` `mrsov` `prlr` `psl` `rlds` `rls` `rlt` `rsds` `rss` `rst` `snld` `tas` `tasmax` `tasmin` `ts` `uas` `vas`
`CCCma-CanCM4_CHFP`CCCma^A	ocn	1979– 2008	02 05 08 11	10	12	[d]n/a [m]`so` `thetao` `uo` `vo` `wo`	[d]n/a [m]`zmlo` `zoh`
`CFS_SHFP`NOAA	atm	1981– 2007	05 11	7	9	[d]— [m]`g` `hus` `ta` `ua` `va`	[d]`prlr` `psl` `tas` [m]`prlr` `psl` `tas` `ts`
`CMAM_shfp`CCCma	atm	1979– 2008	05 11	10	4	[d]`g` `ta` `ua` [m]`g` `ta` `ua` `va`	[d]`prlr` `psl` `tas` [m]`prlr` `psl` `tas` `ts`
`CMAMlo_shfp`CCCma	atm	1979– 2008	05 11	10	4	[d]`g` `ta` `ua` [m]`g` `ta` `ua` `va`	[d]`prlr` `psl` `tas` [m]`prlr` `psl` `tas` `ts`
`ECMWF-S4_CHFP`ECMWF	atm	1981– 2010	02 05 08 11	15	7	[d]— [m]`g` `hus` `ta` `ua` `va`	[d]— [m]`clt` `mrsov` `prlr` `psl` `rlds` `rls` `rlt` `rsds` `rss` `rst` `snld` `tas` `tasmax` `tasmin` `tauu` `tauv` `tdps` `ts` `uas` `vas`
`GloSea5_silv`Met Office	atm	1996– 2009	02 08	24	3	[d]`ua` [m]`g` `ta` `ua` `va`	[d]— [m]`prlr` `psl` `snld` `tas` `ts`
`GloSea5_silv`Met Office	ocn	1996– 2009	02 08	24	3	[d]n/a [m]—	[d]n/a [m]`sic`
`GloSea5_sitl`Met Office	atm	1996– 2009	05 11	24	3	[d]`ua` [m]`g` `ta` `ua` `va`	[d]— [m]`prlr` `psl` `snld` `tas` `ts`
`GloSea5_sitl`Met Office	ocn	1996– 2009	05 11	24	3	[d]n/a [m]—	[d]n/a [m]`sic`
`L38GloSea4_aist`Met Office	atm	1989– 2002	02 05 08 11	9	5	[d]— [m]`g` `ta` `ua` `va`	[d]— [m]`prlr` `psl` `snld` `tas` `ts`
`L38GloSea4_aist`Met Office	ocn	1989– 2002	02 05 08 11	9	5	[d]n/a [m]—	[d]n/a [m]`sic`
`L85GloSea4_akbv`Met Office	atm	1989– 2009	02 05 08 11	9	5	[d]`ua` [m]`g` `ta` `ua` `va`	[d]— [m]`prlr` `psl` `snld` `tas` `ts`
`L85GloSea4_sgoe`Met Office	ocn	1989– 2009	02 05 08 11	9	5	[d]n/a [m]—	[d]n/a [m]`sic`
`JMAMRI-CGCM1_CHFP`MRI-JMA^B	atm	1979– 2010	02 05 08 11	10	7	[d]`g` `hus` `ta` `ua` `va` [m]`g` `hus` `ta` `ua` `va`	[d]`hflsd` `hfssd` `prlr` `psl` `rlds` `rls` `rlt` `rsds` `rss` `rst` `snld` `tasmax` `tasmin` `ts` [m]`clt` `hflsd` `hfssd` `prlr` `psl` `rlds` `rls` `rlt` `rsds` `rss` `rst` `snld` `tas` `tasmax` `tasmin` `ts` `uas` `vas`
`JMAMRI-CGCM1_CHFP`MRI-JMA^B	ocn	1979– 2010	02 05 08 11	10	7	[d]`so` `thetao` `uo` `vo` [m]`so` `thetao` `uo` `vo`	[d]n/a [m]`zmlo` `zoh`
`JMAMRI-CGCM2_CHFP`MRI-JMA	atm	1981– 2010	02 05 08 11	10	7	[d]— [m]`g` `hus` `ta` `ua` `va`	[d]— [m]`clt` `hflsd` `hfssd` `prlr` `psl` `rlds` `rls` `rlt` `rsds` `rss` `rst` `tas` `tasmax` `tasmin` `tauu` `tauv` `ts` `uas` `vas`
`JMAMRI-CGCM2_CHFP`MRI-JMA	ocn	1981– 2010	02 05 08 11	10	7	[d]n/a [m]`so` `thetao` `uo` `vo`	[d]n/a [m]`hfns` `rss` `sic` `sit` `zoh`
`MIROC5_v1.0`CCSR/U.Tokyo	atm	1979– 2011	02 05 08 11	8	12	[d]`g` `hus` `ta` `ua` `va` [m]`g` `hus` `ta` `ua` `va`	[d]`hflsd` `hfssd` `mrsov` `prlr` `psl` `rlds` `rls` `rlt` `rsds` `rss` `rst` `snld` `tasmax` `tasmin` `tauu` `tauv` `ts` [m]`clt` `hflsd` `hfssd` `prlr` `psl` `rls` `rlt` `rsds` `rss` `rst` `snld` `tas` `tasmax` `tasmin` `tauu` `tauv` `ts`
`MIROC5_v1.0`CCSR/U.Tokyo	ocn	1979– 2011	02 05 08 11	8	12	[d]n/a [m]`so` `thetao` `uo` `vo` `wo`	[d]n/a [m]`hfns` `rss` `zmlo` `zoh`
`MPI-ESM-LR_CHFP`MPI	atm	1982– 2011	05 11	9	12	[d]— [m]`g` `hus` `ta` `ua` `va`	[d]— [m]`clt` `hflsd` `hfssd` `prlr` `psl` `rlds` `rls` `rlt` `rsds` `rss` `rst` `snld` `tas` `tasmax` `tasmin` `tauu` `tauv` `tdps` `ts` `uas` `vas`
`MPI-ESM-MR_CHFP`MPI	atm	1981– 2011	05 11	10	7	[d]— [m]`g` `hus` `ta` `ua` `va`	[d]— [m]—
`poama_p24a`CAWCR	atm	1980– 2009	02 05 08 11	10	9	[d]— [m]`hus` `ta` `ua` `va`	[d]— [m]`hfssd` `prlr` `psl` `rlds` `rlt` `rst` `snld` `tas` `tauu` `tauy` `ts`
`poama_p24b`CAWCR	atm	1980– 2009	02 05 08 11	10	9	[d]— [m]`hus` `ta` `ua` `va`	[d]— [m]`hfssd` `prlr` `psl` `rlds` `rlt` `rst` `snld` `tas` `tauu` `tauy` `ts`
`poama_p24c`CAWCR	atm	1980– 2009	02 05 08 11	10	9	[d]— [m]`hus` `ta` `ua` `va`	[d]— [m]`hflsd` `hfssd` `prlr` `psl` `rlds` `rlt` `rst` `snld` `tas` `tauu` `tauy` `ts`

📌 Data Notes

Cell content: Each cell lists variable short-names available in the corresponding level/frequency directory branch for the given domain. A dash (—) means no data is provided in that combination; n/a indicates a structurally non-existent combination (ocean domain has no daily data).

Period: Shown per-domain. For some models the daily archives within a domain may have a shorter coverage than the monthly counterpart — see model-specific footnotes below.

Ensemble Members: Number of perturbed forecast members per initial-condition date, based on the WGSIP/CHFP producer specifications.

A CCCma-CanCM3 / CCCma-CanCM4: Atmosphere monthly data span 1979–2010; atmosphere daily data span 1979–2008 only. All ocean data span 1979–2008.

B JMAMRI-CGCM1: Monthly data span 1979–2010 (both atm and ocn); daily data (atm-levels, atm-surface, ocn-levels) span 1979–2006 only.

Variable Codes

CHFP variables follow CMIP-style short names. Definitions follow the CHFP Producer Guide (WGSIP, v2.0, 2013).

Atmosphere — Pressure Levels

Standard pressure levels: 850, 500, 200, 100, 50, 10 hPa (model-dependent).

g : Geopotential
hus : Specific Humidity
ta : Temperature
ua : Zonal Velocity
va : Meridional Velocity

Atmosphere — Surface

clt : Total Cloud Cover
hflsd : Surface Latent Heat Flux
hfssd : Surface Sensible Heat Flux
mrsov : Total Soil Moisture
prlr : Total Precipitation
psl : Mean Sea Level Pressure
rlds : Downward Surface Longwave
rls : Net Surface Longwave
rlt : Top Net Longwave (TOA)
rsds : Downward Surface Solar
rss : Net Surface Solar
rst : Top Net Solar (TOA)
snld : Snow Depth
tas : 2 m Temperature
tasmax : 2 m T Daily Max
tasmin : 2 m T Daily Min
tauu : Surface Stress (x)
tauv : Surface Stress (y)
tauy : Surface Stress (y, POAMA legacy) ^*
tdps : 2 m Dewpoint Temperature
ts : Surface Temperature (SST+Land)
uas : 10 m Wind (u)
vas : 10 m Wind (v)

^* tauy is a non-standard alias for the meridional wind stress used in some POAMA-2.4 archives; it is equivalent to the standard tauv.

Ocean — Levels

so : Salinity
thetao : Potential Temperature
uo : Zonal Velocity
vo : Meridional Velocity
wo : Vertical Velocity

Ocean — Surface

hfns : Net Surface Heat Flux
rss : Net Surface Solar Flux
sic : Sea Ice Concentration
sit : Sea Ice Thickness
zmlo : Mixed Layer Depth
zoh : Sea Surface Height (above geoid)

Table 1. Directory architecture
Directory	Detail
GPC Name	Lowercase GPC name (see table 3 below)
System version	System version of GPC (see table 3 below)
Season	Forecast season (JFM, FMA, MAM, ...)

Table 2. File naming rules
Component	Detail
YYYYMM_IC	Initial-condition date (year and month the forecast run was initialized)
YYYYMM_Start	First forecast target month (1st time-step in the file)
YYYYMM_End	Final forecast target month (terminal lead month)
Example: GPC_exeter_200505_200506_200510_allens.nc A 5-month forecast initialized in May 2005 from GPC exeter. Time axis spans June 2005 (step 1) to October 2005 (final step). Contains all ensemble members and variables bundled in one NetCDF file.

Table 3. WMO LC-SPMME Hindcast data summary
GPC Name	System Version	Hindcast Period	Ensemble	Lead (mo.)	Variables
exeter (United Kingdom)	Glosea5	1993 - 2015	28	5	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
Glosea6	1993 - 2016	28	5	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
montreal (Canada)	CanSIPv1	1981 - 2010	20	6	`h500` `mslp` `prec` `sst` `t02m` `t850`
CanSIPv2	1981 - 2010	20	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
CanSIPv2.1	1980 - 2020	20	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
CanSIPv3	1990 - 2020	40	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
melbourne (Australia)	POAMA	1980 - 2011	99	6	`h500` `mslp` `prec` `sst` `t02m` `t850`
ACCESS-S1¹	1990 - 2012	11 ~ 44	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
ACCESS-S2²	1981 - 2018	3 ~ 18	7	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
beijing (China)	BCCv2	1991 - 2010	24	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
beijing	BCCv3	2001 - 2024	21	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
cmcc (Italy)	SPS3.5	1993 - 2016	40	5	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
SPS4.0³	1993 - 2022	30 or 40	5	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
cptec (Brazil)	BAM1.0	1979 - 2010	10	6	`h500` `mslp` `prec` `t02m` `t850`
BAM1.2	1981 - 2010	10	5	`h500` `mslp` `prec` `t02m` `t850` `u850` `v850`
ecmwf (Europe)	System4	1981 - 2010	15	6	`h500` `mslp` `prec` `sst` `t02m` `t850`
System5	1993 - 2016	25	5	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
moscow (Russia)	SL-AV	1981 - 2010	10	4	`h500` `mslp` `prec` `sst` `t02m` `t850`
SL-AV-072L96	1991 - 2015	11	3	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
offenbach (Germany)	GCFS1	1981 - 2010	15	5	`h500` `mslp` `prec` `sst` `t02m` `t850`
GCFS2.0	1990 - 2015	30	5	`h500` `mslp` `prec` `sst` `t02m` `t850`
GCFS2.1	1993 - 2019	30	5	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
GCFS2.2	1993 - 2019	30	5	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
pune (India)	MMCFS⁴	2003 - 2017	3 ~ 14	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
seoul (Republic of Korea)	Glosea5	1991 - 2016	12	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
Glosea6	1993 - 2016	28	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
tokyo (Japan)	CPS2	1979 - 2014	10	3	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
CPS3	1991 - 2020	10	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
toulouse (France)	System6⁵	1993 - 2016	25	5	`h500` `mslp` `prec` `sst` `t02m` `t850`
System7	1993 - 2016	25	5	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
System8	1993 - 2018	25	5	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
System9	1993 - 2024	31	5	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`
washington (USA)	CFSv2	1982 - 2010	20	6	`h500` `mslp` `prec` `sst` `t02m` `t850` `u850` `v850`

Acknowledgement

Please use the following acknowledgment when using data from this service:
"We acknowledge the WCRP Working Group on Sub-seasonal to Interdecadal Prediction (WGSIP) for establishing the Climate-system Historical Forecast Project (CHFP) and the APEC Climate Center (APCC), Republic of Korea for providing the model output via the APCC data portal. We also thank the CHFP data providers for making their model output available and the WMO Global Producing Centres for permitting their hindcast datasets to be served in conjunction with the CHFP."

Climate Information Services

Historical Context

Scientific Significance & Core Framework

Data Specification & Standardization

Additional Information & Reference

Institutional Mandate & History

Core Functions of the Lead Centre (wmolc.org)

Data Specification & Standardization

Operational Reference

Quick Start

Directory Architecture & Naming Rules

Automated Batch Retrieval via Wget

Multi Model Hindcast Summary

Variable Codes

Atmosphere — Pressure Levels

Atmosphere — Surface

Ocean — Levels

Ocean — Surface

Quick Start

Directory Architecture & Naming Rules

Automated Batch Retrieval via Wget

Multi Model Hindcast Summary

Variable Codes

Acknowledgement