Research

 El Nino

 Climate Change

 Monsoon

 GCM

 Forecast

 

Climate Change

 Long Range Forecast


      To make a long range weather forecast, we undergo three processes. Global SST prediction for model boundary condition using Intermediate Ocean-Statistical Atmosphere Coupled Model, and Model climatology production and global forecast through Atmospheric General Circulation Model (AGCM) simulation , Statistical correction of dynamic forecast using model hindcast data.


1. Seasonal Climate Prediction System

    1) Global SST Prediction

    < A combined system of the dynamical-statistical models >

    1. Intermediate El-nino prediction model

      Atmospheric Model
      • Statistical model based on the singular value decomposition (SVD)   of the observed wid stress and SST.
      • used two singular modes (ENSO mature and transition mode)

      Ocean Model
      • Modified from Lamont model (Zebiak and Cane,1987)
        - Subsurface temperature parameterization based on SVD mode
        - Ocean Basic State
        - Initialization Process
        - Heat Flux Parameterization (Latent Heat, Solar Radiation)
        - Vertical Eddy Mixing Parameterization

    2. Lagged linear regression model

      Description
      A pointwise statistical model based on lag relationship between
      global SST and ENSO index.
      • Equation
         
      • Predictand : Each Grid SST in Global Oceans
      • Predictor : NINO3 SST
        - Observed NINO3 SST only (LLRobs)
        - Observed and forecasted NINO3 SST by dynamic El Nino
          Prediction model (LLRfcst)
        - Forecast NINO3 is used only when forecast lead time is larger
          than the lag time
      • Optimal lag is selected by hindcast process in the model

    3. Coupled pattern projection model

      Description
       A pointwist regression model generating realizaions of the predictand from projection of covariance pattern between the large-scale predictor field and the one point predictand ontoa large-scale predictor field for the target year
      Model Procedure
      • Construct coupled pattern for certain domain
      • Generate preliminary SST prediction by projection of the coupled
      • Select optimal domain from huge predition set (location and size)
      • Obtain final predition from selected domain
      Optimizing Strategy
      • Area scanning of predictor field usiing the flexible window
      • Ensemble mean of qualified predictions with different predictor
         domains
        - Classify the preliminary predictions by a hindcast skill
          during training period.
        - Classify of 9 groups (significant level :99.9%,99%,97%,...,85%)
        - Ensemble mean of the predictions in the highest category

    2) Dynamical Forecast

    1. Model Description

      AGCM Description
      • GCPS-Global Climate Prediction System
      • Mainly developed for the long range seasonal forecasts
      • T63 L21 truncations for the seasonal forecasts

      Dynamics
      • Three-dimension hydrostatic primitive equations on sphere
        with normalized pressure coordinate
      • Spectral method → Finite volume method (Lin and Wood, 1997)
      Physical Process
      1. 2-stream k-distribution radiation scheme
        (Nakajima and Tanaka, 1986)
      2. Simplified Arakawa-Schubert cumulus convection scheme based on
        Relaxed  Arakawa-Schubert scheme (Moorthi and Suarez, 1992)
      3. Orographic gravity-wave drag (McFarlane, 1987)
      4. Large-scale condensation scheme based on Letreut and Li(1991)
      5. Dry adiabatic adjustment
      6. Bonan's Land Surface Model (Bonan 1996)
      7. Non-local PBL/Vertical diffusion (Holtslag and Boville 1993)
      8. Diffusion-type Shallow Convection (Tiedtke 1989)
      9. Modified CCM3 slab ocean/sea-ice model

    2. AGCM Simulation

      SMIP_HFP simulation to produce prediction climatology

      Experiments
      To carry out 4-month ensemble integration of atmospheric GCMs
        with observed initial conditions and predicted boundary conditions
      Initial conditions and Ensemble size
      Inconsistency with operational startup dates 28 years
        from 1979-2006, 4 seasons, 12 member ensembles
      → Atmospheric initial conditions :
          NCEP reanalysis data ( U, V, T, q, Ps )
      → Land initial conditions : NCEP reanalysis
          ( soil temp, snow depth, soil moisture, surface temp. )
      → Ensemble members : 00& 06& 12& 18Z of 5, 6, 7 February
                                                00& 06& 12& 18Z of 5, 6, 7 May
                                                00& 06& 12& 18Z of 5, 6, 7 August
                                                00& 06& 12& 18Z of 5, 6, 7 November
      Length of integration
      4 months through the initial time of Feb., May, Aug., Nov.
      Boundary conditions
      Predicted SST using El-Nino Ocean model and Statistical Prediction
        model Sea ice is the same data with AMIP data offered PCMDI


2. Forecast Result


3. Links