Seasonal Forecasts in the Pacific Region using POAMA 1.5b Australia Greenhouse Conference 4 8 th April (2011) Cairns, Queensland, Australia. Andrew Cottrill 1, Eun Pa Lim 1 and Harry Hendon 1 1 Bureau of Meteorology (CAWCR), Melbourne, Victoria.
Outline of Presentation Location Map of the Fifteen Pacific Island Nations involved in the PASAP project; Show the Seasonal Rainfall Patterns across the Tropical Pacific; Show some typical ENSO Patterns in Rainfall Composites at three stations; Briefly describe the POAMA Seasonal Prediction System and its Correlation to Tropical SSTs with various lead times; Show patterns of Hit Rates of Above Median Rainfall between CMAP and POAMA over the equatorial Pacific; Show Calibration of Seasonal Rainfall at Tarawa as a technique to improve seasonal forecast outlooks and Short Summary
Pacific Island Countries and PASAP Fifteen Partner Countries Cook Islands East Timor Federated States of Micronesia Fiji Kiribati Niue Palau Papua New Guinea Republic of Marshall Islands Republic of Nauru Samoa Solomon Islands Tonga Tuvalu Vanuatu PASAP. The PASAP project has been developed under the International Climate Change Adaptation Initiative to help Pacific Island Countries (PICs) prepare for climate change in coming decades. The PASAP project aims to deliver seasonal climate outlooks to PICs based on POAMA. More information on PASAP can be found at the Department of Climate Change and Energy Efficiency (DCCEE). Map: Yuri Kuleshov - BoM
Patterns of Seasonal Rainfall Across the Tropical Pacific Using Data from CMAP Australian Monsoon SPCZ Indian and East Asian Monsoon SPCZ Units =mm/day NE Trades SE Trades Summer Rainfall Autumn Rainfall Winter Rainfall ITCZ ITCZ CMAP Data: 1979-2006 Shows the Mean State of Seasonal Rainfall along the Equator (ITCZ) and the SPCZ in the southwest Pacific. The ITCZ migrates north and south with the change of seasons and the SPCZ migrates northeast and southwest depending on the state of ENSO. Spring Rainfall CMAP =CPC Merged Analysis of Precipitation.
ENSO Composites for Three PICs: Tarawa, Port Vila and Nadi Airport (1980 2006) El Niño (x8): Tarawa Port Vila Nadi Airport La Niña (x4): El Niño High Rainfall All Seasons El Niño Low Rainfall All Seasons La Niña Low Rainfall All Seasons La Niña High Rainfall All Seasons El Niño Lower Rainfall Spr, Sum, Aut La Niña Higher Rainfall Spr, Sum, Aut Note: Composite Years (El Niño): = 1982, 1986, 1987, 1991, 1994, 1997, 2002 and 2004; and La Niña: 1984, 1988, 1998 and 1999.
The POAMA 1.5b Seasonal Prediction System Initially developed by the BMRC and the CSIRO for the prediction of SST anomalies associated with ENSO over the Pacific. BAM3.0d Spectral transform (T47L17) Ocean Atmosphere Sea Ice Soil (OASIS) coupler Australia Community Ocean Model version 2 Consists of the ocean model (ACOM2) and the atmospheric model BAM3 It also has a new Atmosphere Land Initialisation scheme or ALI The Forecasts are based on 10 member ensemble hindcasts run over 27 years (1980-2006)
POAMA Forecast of Tropical Pacific SST LT= 0 and 2 months LT= 5 and 8 months Correlation From Maggie Zhao - BoM
Hit Rates of Above Median Rainfall using POAMA over the Tropical Pacific Region Yes Observed: No Forecast Yes No Hits (A) Misses (C) False Alarms (B) Correct Rejection (D) Hit Rate = A/(A+C)
Hit Rates of Median Rainfall from POAMA and Observations (CMAP) for DJF LT=0 LT=2 LT=4 LT=6
Calibration of Seasonal Rainfall in Ensemble Forecasts using POAMA 1.5b The calibration method used here is described in detail in the paper by Johnson and Bowler (2009) in Monthly Weather Review; It is known as the variance inflation method and is based on two conditions; The technique adjusts the forecasts so the climatological variance of the forecasts is the same as the observations and The correlation of observations with the unadjusted ensemble mean is the same as the correlation of the adjusted ensemble members with the unadjusted ensemble mean.
Calibration of Seasonal Rainfall at Tarawa Kiribati Rainfall Anomaly with LT=0 MAM JJA r=0.71 r=0.89 SON DJF r=0.81 r=0.71
Correlation Skill and Lead Times at 14 Pacific Island Stations Stations: Nadi Airport, Suva, Rarawai, Nabouwalu, Rotuma, Port Vila, Tarawa, Funafuti, Apia, Nuku alofa, Alofa, Honiara, Port Moresby, Rarotonga
RMSE Skill and Lead Times at 14 Pacific Island Stations Stations: Nadi Airport, Suva, Rarawai, Nabouwalu, Rotuma, Port Vila, Tarawa, Funafuti, Apia, Nuku alofa, Alofa, Honiara, Port Moresby, Rarotonga
Summary Seasonal Rainfall over the PICs is mostly controlled by the ITCZ and the SPCZ; Strong rainfall changes associated with the different phases of ENSO over the tropical Pacific region provide coupled models, with the skill to produce seasonal forecasts with up to 6 months or more lead time; Hit Rates using POAMA1.5b are typically 60 80% across the equatorial Pacific and parts of the southwest Pacific; Calibration of seasonal rainfall will be used in seasonal forecasting products, and is planned to compliment SCOPIC, which is currently used in many PICs. Models, like POAMA, have the ability to produce better season forecasts than statistical models, as they can account for aspects of climate change and climate variability not represented in the historical record.
Acknowledgements: The Bureau of Meteorology would like to thank all the PIC nations involved in the PASAP project for providing valuable rainfall data from a number of stations across the region, and to AUSAID, who provided the funds for this PASAP project. The PASAP website can be found at the following address http://poama.bom.gov.au/experimenttal/pasap Username: pasap Password:pacifica
Spares
CMAP POAMA 1.5b
ENSO Composites for Three PICs: Tarawa, Port Vila and Nadi Airport (1980 2006) El Niño (x8): Tarawa Port Vila Nadi Airport La Niña (x4): El Niño High Rainfall All Seasons El Niño Low Rainfall All Seasons La Niña Low Rainfall All Seasons La Niña High Rainfall All Seasons El Niño Lower Rainfall Spr, Sum, Aut La Niña Higher Rainfall Spr, Sum, Aut Note: Composite Years (El Niño): = 1982, 1986, 1987, 1991, 1994, 1997, 2002 and 2004; and La Niña: 1984, 1988, 1998 and 1999.
Seasonal Correlation between CMAP Rainfall and Reynolds SSTs Across the Tropical Pacific (1982 2006) ITCZ SPCZ ITCZ SPCZ SPCZ Correlation SPCZ
Typical ENSO SST Patterns Warm and Cold Events Modoki December 2009 Classic El Niño Feb 1998 La Niña January 1999 La Niña December 2010 Images from: www.longpaddock.qld.gov.au
Seasonal Correlation of CMAP and POAMA Rain and Station Correlation to POAMA Rain Summer LT=0 Autumn LT=0 Winter LT=0 Spring LT=0
Hit Rates Plots of POAMA and CMAP (MAM) Above Median Rainfall LT=0 LT=2 LT=4 LT=6
Hit Rates Plots of POAMA and CMAP (JJA) Above Median Rainfall LT=0 LT=2 LT=4 LT=6
Hit Rates Plots of POAMA and CMAP (SON) Above Median Rainfall LT=0 LT=2 LT=4 LT=6
Reliability Diagrams - DJF LT=0 LT=2 LT=4 LT=6
Reliability Diagrams Autumn Winter
Reliability Diagrams Spring Summer
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b at Nadi Airport-Fiji (LT=0) MAM JJA r=0.61 r=0.26 SON DJF r=0.67 r=0.62
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b at Port Vila - Vanuatu (LT=0) MAM JJA r=0.53 r=0.51 SON DJF r=0.65 r=0.59
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Nabouwalu - Fiji) MAM JJA r=0.44 r=0.53 SON DJF r=0.73 r=0.53
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Suva - Fiji) MAM JJA r=0.26 r=0.40 SON DJF r=0.37 r=0.06
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Rarawai - Fiji) MAM JJA r=0.63 r=0.18 SON DJF r=0.51 r=0.59
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Rotuma - Fiji) MAM JJA r=0.37 r=0.42 SON DJF r=0.05 r=0.30
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Alofi -Niue) MAM JJA r=0.43 r=0.02 SON DJF r=0.52 r=0.58
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Port Moresby) MAM JJA r=0.46 r=0.22 SON DJF r=0.68 r=0.11
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Apia - Samoa) MAM JJA r=0.31 r=0.45 SON DJF r=0.21 r=0.46
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Honiara - Solomons) MAM JJA r=0.38 r=0.32 SON DJF r=0.21 r=0.49
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Nuku alofa - Tonga) MAM JJA r=0.37 r=0.35 SON DJF r=0.52 r=0.58
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Funafuti - Tuvalu) MAM JJA r=0.09 r=0.53 SON DJF r=0.40 r=0.34
Calibration of Seasonal Rainfall to Pacific Island Stations using POAMA 1.5b (Rarotonga Cook Is) MAM JJA r=0.51 r=0.11 SON DJF r=0.46 r=0.40
Results of Calibration Seasonal forecasts are less emphatic; Reliability has improved at stations with moderate to high correlation; Forecasts are near climatology when the correlation is low (< ~0.20) Assumptions: Assumes the observations represent normal distributions