North American Multi-Model Ensemble (NMME) Bill Merryfield - - PowerPoint PPT Presentation

North American Multi-Model Ensemble (NMME)

CITES-2019 International Young Scientists School, 28 May 2019

Bill Merryfield

Canadian Centre for Climate Modelling and Analysis (CCCma)

• US, Canadian operational forecast systems + US research systems
• Hindcasts and real time forecasts
• Real time forecasting since Aug 2011
• Data openly accessible: https://iridl.ldeo.columbia.edu/SOURCES/.Models/.NMME/
• Requirements for inclusion
• ensemble system, range  9 months
• must provide hindcast data for 1982-2010
• commitment to provide real time forecasts by 8th of each

month (CPC operational schedule)

https://www.cpc.ncep.noaa.gov/products/NMME/

Operational Centers Research Centers NCEP ECCC GFDL NCAR NASA

Hindcasts Real-time Forecasts

Research Community User/applications Community

NMME Organization

8th of each month

Currently contributing models

Model Center Ensemble size CFSv2 NCEP 24 (28) CanCM3 EC/CMC 10 CanCM4 EC/CMC 10 FLOR GFDL 24 CM2.1 GFDL 10 CCSM4 NCAR 10 GEOS-5 NASA 11 Total ensemble size 99 (103)

Deterministic and probabilistic forecasts

*Anomalies and tercile boundaries computed separately for each model

Deterministic Models weighted equally Probabilistic Ensemble members weighted equally* Prate 2015 OND from 201509 (1 month lead)

Individual model forecasts Individual model skills

ENSO plumes

Ensemble means Nino3.4 forecast from June 2015 All ensemble members  great 2015-16 El Niño well predicted at least 6 months in advance

Advantages of a multimodel ensemble

Becker et al. 2014 https://doi.org/10.1175/JCLI-D-13-00597.1

Deterministic skill (anomaly correlation of ensemble means)*

2m temperature (land 23N-75N) precipitation (land 23N-75N) SST (23N-75N) *for earlier set of NMME models Nino3.4 index

Advantages of a multimodel ensemble

Reliability of probabilistic forecasts

Becker et al. 2016 https://doi.org/10.1175/JCLI-D-14-00862.1

CFSv2: 1 model, ensemble size 24 miniNMME: 6 models, ensemble size 24 (64) NMME: 6 models, ensemble size 75

lower is better higher is better above normal (A) below normal (B) near normal

2m temperature (land 23N-75N)

NMME for International Regions

https://www.cpc.ncep.noaa.gov/products/international/nmme/nmme.shtml

NMME temperature forecast for JJA 2019 lead 1 month

Deterministic (ensemble mean anomaly) Probabilistic (tercile probabilities)  Deterministic skill (anomaly correlation)

NMME precipitation forecast for JJA 2019 lead 1 month

Deterministic (ensemble mean anomaly) Probabilistic (tercile probabilities)  Deterministic skill (anomaly correlation)

More about NMME

• NMME became operational in Sep 2015
• An operational requirement is 99% on-time delivery
• Operational NMME forecasts will be product based, meaning if one
• r more models fails to deliver, then official forecast will be based on

models received

• New models will be being evaluated and added as they

become available (Canadian GEM-NEMO in 2019)

• Subseasonal NMME experiment is “SubX” is underway

NMME SubX

• Weekly initialization
• Forecast length  32 days (45 days encouraged)
• Hindcast period 1999-2015 (additional years encouraged)
•  3 ensemble members (more encouraged)
• Hindcasts and real-time forecasts (product based, like seasonal

NMME)

• Data at IRI: https://doi.org/10.7916/D8PG249H
• Currently 6 models, 63 ensemble members (can differ from

seasonal NMME, e.g. Canadian forecasts are from GEM monthly forecast)

• Experimental (not yet operational)

http://cola.gmu.edu/kpegion/subx/

http://wxmaps.org/subx_custom.php

http://www.nws.noaa.gov/ost/CTB/nmme_pub.htm

NMME and SubX for research

http://cola.gmu.edu/kpegion/nmmeworkshop2017

NMME application to global hydrological forecasting

Approach:

• NMME) used to drive Variable Infiltration Capacity (VIC) land surface

hydrologic model

• Droughts and wet spells analysed over global major river basins
• NMME-based approach evaluated against the traditional Ensemble

Streamflow Prediction (ESP) based on sampling climatological distribution

Yuan et al., BAMS 2015

# forecast months that NMME/VIC ensemble median has significantly (p<0.05) higher Equitable Threat Score than ESP/VIC

Seasonal prediction of atmospheric river frequency

Climatological atm river days + anomaly composites

Zhou & Kim 2018 https://doi.org/10.1007/s00382-017-3973-6

ENSO effect on DJF atmospheric river frequency in NMME models

Zhou & Kim 2018 https://doi.org/10.1007/s00382-017-3973-6

Anomaly correlation skill for predicting atmospheric river frequency

All years ENSO years

(note that daily hindcast data is required) DJF lead 1 month

NMME Phase 1 Data at IRI

http://iridl.ldeo.columbia.edu/SOURCES/.Models/.NMME

Hindcasts + real time forecasts

NMME real time daily data (new)

https://www.ncdc.noaa.gov/data-access/model-data/model-datasets/north-american-multi-model-ensemble

https://www.earthsystemgrid.org/search.html?Project=NMME

NMME Phase 2 Data on the ESGF

NMME Phase 2 Data

• Common 1

grid

• NetCDF4

Summary

• NMME is now the prime source of dynamical seasonal

forecast information for North America

• Occasionally new models are added, and old ones retired
• Currently 7 models, Canadian GEM-NEMO to be added,

CMC1/CanCM3 to be retired in 2019

• Data from hindcasts and real time forecasts is freely

available for applications and research

• Forecast daily data now available
• NMME open data driving much climate prediction

research, leading to many papers

• SubX = subseasonal version of NMME is following similar

principles