DOC

Project Milestone Reporting Table

By Stephanie Dunn,2014-01-20 03:05
8 views 0
Project Milestone Reporting Table

    Project Proposal and Milestone Reporting

Project title

    3.2.3: Implement and assess statistical techniques to improve regional rainfall and

    temperature forecasts from the national dynamical seasonal forecast model

Description

    Building on the review of statistical techniques to improve dynamical coupled

    forecasts of regional climate (Project 3.2.2), we will implement and test several

    statistical techniques to improve seasonal forecasts of regional rainfall and

    temperature in South East Australia using the Bureau of Meteorology’s Predictive

    Ocean Atmosphere Model for Australia (POAMA).

Geographic Focus

    GRDC southern region

Principal Investigator and administrative contact

    Harry Hendon and Oscar Alves, Bureau of Meteorology Research Centre,

    h.hendon@bom.gov.au, Tel: 03 9669 4120; Fax:03 9669 4660, PO Box 1289K,

    Melbourne VIC 3001

Margaret Hughes, BMRC Administrative Officer, m.hughes@bom.gov.au, Tel: 03

    9669 4424, Fax: 03 9669 4660, GPO Box 1289, Melbourne VIC 3001

Project objectives

    ? Implement, test and evaluate statistical techniques to bridge and calibrate

    dynamical seasonal predictions of rainfall and temperature for South East

    Australia from the Bureau’s dynamical seasonal prediction model (POAMA)

    ? Implement experimental real-time, dynamical-statistical seasonal prediction

    system for SE Australian climate

Methodology

    ? Generate multi-member hindcasts from POAMA from 1980 to 2005 (c.f.-

    currently only a single member is for the period1987- 2001)

    ? Using the extended hindcasts set, bridge SE Australian rainfall from leading

    modes of predicted SST variability ? Using extended hindcast set, calibrate direct predictions of SE Australian rainfall

    from POAMA

    ? Document skill in SE Australian rainfall prediction using bridged and calibrated

    hindcasts.

    ? Implement experiment, real-time dynamical-statistical seasonal prediction system

    of SE Australia rainfall, based on bridging and calibration of the dynamical

    forecasts from POAMA. Post experimental forecasts of SE Australian rainfall

    available on the POAMA web page

Links

This project has direct links to all other projects in Theme 3, and, in particular is the

    follow-on of project 3.2.2, whereby the recommended statistical techniques from that

    project are implemented and assessed.

     1

Summary of progress

    The version 1.5b of POAMA has been run to produce 10 member ensemble hindcasts from 1980 to 2005. At present a complete set of 3 member ensemble has been used for skill assessment of POAMA forecasts for Australian rainfall with focus on the south eastern Australian region (33.5-38.5?S, 137.5-152.5?E).

    A major advance of the POAMA 1.5b system from the previous versions is that it uses a new Atmosphere and Land Initialisation scheme (ALI). This scheme nudges POAMA’s u-wind, v-wind, atmospheric temperature and humidity to those of ERA-

    40 reanalysis for the hindcasts and those of the Bureau’s NWP system for real-time

    forecasts. Then the fractional differences between forecast and reanalysis are added to POAMA at the end of a 6 hour forecast cycle. In this process, the land surface is also brought into balance with the atmospheric forcing. Consequently, the hindcasts have more realistic atmospheric initial conditions and have better consistency with real-time forecasts. An example of the resultant forecast skill improvement with the POAMA 1.5b is shown in Figure 1. In the observation Australia was wetter than normal in 1997 spring despite the presence of the record breaking intensity of El Nino, whereas it was much drier than normal in 2002 spring despite the weak intensity of El Nino in the Pacific. Figure 1 (b) and (c) demonstrate that the POAMA system can predict the different rainfall events with reasonably good skill. Furthermore, the improvement in initial conditions in the POAMA 1.5b results in better forecasts for the rainfall anomalies in both years, compared to the forecasts from the POAMA 1.5a which uses Bureau’s atmospheric model initial conditions forced by observed SST.

    In order to examine whether additional skill improvement in SE Australian rainfall is feasible by the application of statistics, we developed and tested various statistical-dynamical models to forecast Australian rainfall, capitalizing on POAMA’s rainfall

    and SST predictions for statistical calibration and bridging, respectively. Our preliminary results suggest that POAMA can provide skilful prediction for below/above median rainfall over SEACI region at least at lead time 0, and statistical calibration technique seems to be able to add extra skill (Figure 2). However, skilful prediction by dynamical and statistical-dynamical models tends to vary in different seasons and for different lead times. Therefore, in order to obtain more stable statistical relationships between rainfall and predictors (e.g. POAMA predicted rainfall/SST), the completion of 10 member ensemble hindcasts in an extended period should be given a priority of our research.

    POAMA 1.5b has been run operationally since June 2007. The real-time forecasts for Australian rainfall, maximum/minimum temperatures and ocean/atmosphere indices are available on the POAMA web-page (http://poama.bom.gov.au). Posting

    experimental dynamical/statistical-dynamical forecast products focused on SE Australia is underway.

     2

    (a) Observation

    (b) Forecasts from POAMA 1.5a at lead time 0

    (c) Forecasts from POAMA 1.5b at lead time 0

Figure 1 Standardized rainfall anomalies in 1997 (left) and 2002 (right) spring (a) in

    the observation (b) forecasted in POAMA 1.5a and (c) forecasted in POAMA 1.5b

     3

    DJFMAM

    0.80.8

    0.60.6POAMAPOAMA0.40.4SVD-CALSVD-CALSVD-BRSVD-BR0.20.2STATSTATCorr. Coeff.Corr. Coeff.00

    -0.2-0.2

    -0.4-0.4

    -0.6-0.601234560123456Lead TimeLead Time

    JJASON

    0.80.8

    0.60.6POAMAPOAMA0.40.4SVD-CALSVD-CALSVD-BRSVD-BR0.20.2STATSTATCorr. Coeff.Corr. Coeff.00

    -0.2-0.2

    -0.4-0.4

    -0.6-0.601234560123456Lead TimeLead Time

Figure 2 Correlation coefficients of SEACI region mean rainfall anomaly forecasts

    and observation as a function of lead time. SVD-CAL: statistical-dyamical prediction

    using POAMA forecasted rainfall as a predictor; SVD-BR: statistical-dynamical

    prediction using POAMA forecasted SST as a predictor; and STAT: purely statistical

    prediction using observed SST with time lag.

     4

    Project Milestone Reporting Table

To be completed prior to commencing the project Completed at each Milestone date

    45Milestone Progress Recommended BudgetPerformance Completion

    1 2 3for changes to description indicatorsdate

    6Milestone workplan

    ($)

    1 Generate Hindcast data set December 30K

    multi-member generated and 2007

    hindcasts archived

    2. Bridge Algorithms June 35K

    rainfall from developed, 2008

    EOFs of tested, and

    predicted SST documented

    and calibrate

    direct prediction

    of rainfall from

    extended

    hindcast set of

    POAMA

    3. Document Report prepared December 20K

    skill of 2008

    predicting SE

    Australian

    climate, using

    statistical

    bridging and

    calibration

    techniques for

    the hindcast

    period 1980-

    2005

     5

4. Implement Experimental December 15K

    trial real-time predictions 2008 dynamical-available on

    statistical POAMA

    seasonal webpage

    prediction

    system of SE

    Australia

    rainfall , based

    on bridging and

    calibration of

    POAMA

    forecasts

     6

Report this document

For any questions or suggestions please email
cust-service@docsford.com