Seasonal Prediction and Climate Risk Management Using RCMs by mps12334

VIEWS: 9 PAGES: 36

									Seasonal Prediction and Climate Risk
     Management Using RCMs



                            Liqiang Sun

    International Research Institute for Climate and Society
                     Columbia University


                          With inputs from
    P. Block, F. A. Souza Filho, L. Goddard, J. Qian and A. Robertson
  Example of
    seasonal
rainfall forecast
•  Regional
•  3-month average
•  Probabilistic
Example: Malaria Early Warning

These results discussed at 1st SA
Regional Epidemic Malaria
Outlook Forum, Harare, 2004




     (Courtesy Lisa Goddard)
       Climate Information Relevant to Agricultural
                    Decision Making




(Courtesy Miguel Carriquiry)
 to be useful to decision makers,
         climate forecasts must:

   be probabilistic

   be reliable

   contain some resolution

   address relevant scales and quantities
           Regional Climate Modeling –
       Advance the science that underpins better
         climate risk management practices

  nderstanding of physical processes that contain
  U
predictability at smaller spatial scales (anomalies within
anomalies) and temporal scales (weather within climate).

  nhancing the scale and relevance of climate forecasts and
  E
creating information to better support decisions.
                    OUTLINE
  Understanding of local climate and its variability using
  RCMs
  Climate prediction using RCMs
  Potential values of RCMs for decision making
  Summary
Understanding of local climate
and its variability using RCMs
On Seasonal Time Scale

  Improvement of Spatial Patterns and Temporal
   Distribution
  Predictability at Smaller Spatial and Temporal Scales
  Representation of Climate Uncertainty
A Typical Tropical Cyclone Simulated by Climate Models

       ECHAM4.5 AGCM(T42)                       RSM (50km)
       Vorticity              Winds    Vorticity       Winds




       Precipitation      Humidity    Precipitation    Humidity




 Carmago, Li and Sun (2007)
Sun et al. (2005)
Correlation of wet day         Corr of
frequency between OBS and      freq of
                               medium
RCM
                               rain
                               >5mm/day
 Corr of freq
 >1mm/day
                            Corr of freq of
                            heavy rain
                            >10mm/day
Correlation of dry day (<1 mm/day)   Correlation of number of dry
frequency between OBS and            spell (>5 days) between OBS and
RegCM                                RegCM
FMA Precipitation anomaly distribution over Ceara.
                 Unit is mm/day.




                         ECHAM RSM
     Ensemble Spread      1.9  2.4
     Signal               2.8  3.1
                                             Sun et al. (2005)
        Climate prediction using RCMs
Predictions based, at least in part, on Regional Climate Models:
•    IRI since 1997 (various regions, including SE Asia)
•    NR&M /IRI 1998 (Queensland)
•    FUNCEME/IRI since 2001 (NE Brazil) Challenges
•    NCEP since 2002 (USA)                      Scientific issues related to
                                               predictability at smaller scales
•    CWB/IRI since 2003 (Taiwan)
                                                Technical issues for regional
•    ICPAC/IRI since 2004 (GHA)                climate modeling
•    SAWS/IRI 2006 - 07 (S. Africa)             Computational constrains
•    UNAM/IRI since 2006 (Mexico)
•    CMC/CIIFEN/IRI since 2008 (Western S. America)
•    ECPC/NTU,HKO, BIU since 2003 (Taiwan, Hong Kong, Mediterranea)
•    Downscaling DEMETER Hindcasts (Europe)
Seasonal Climate Forecasts Using RCMs
   - Examples from Northeast Brazil
    CLIMATE DYNAMICAL DOWNSCALING FORECAST SYSTEM FOR NORDESTE


                                                                            HISTORICAL DATA
     PERSISTED GLOBAL                                                        E
                                                                            •  xtended Simulations
      SST ANOMALIES                                                          O
                                                                            •  bservations
                                                         Persisted SSTA
                                                    10
                                                            ensembles
     PREDICTED SST
     ANOMALIES                                             1 Mo. lead

     Tropical Pacific Ocean
                                     ECHAM4.5
     (LDEO Dynamical Model)
        AGCM (T42)                                   Post
     (NCEP Dynamical Model)                                                   Processing
     (NCEP Statistical CA
     Model)
                          CPTEC        15 Predicted SSTA         Multi-Model
     Tropical Altantic Ocean
       AGCM (T42)              ensembles        Ensembling
     (CPTEC Statistical CCA
     Model)
                                              1-4 Mo. lead
     Tropical Indian Ocean

     (IRI Statistical CCA
     Model)

     Extratropical Oceans

     (Damped Persistence)
                                AGCM INITIAL CONDITIONS                       RSM97 (60km)
                                UPDATED ENSEMBLES (10+)
                                                                              RAMS (40km)
                                  WITH OBSERVED SSTs                             CPT




                                   IRI                                    FUNCEME
Sun et al. (2006)
Geographical distribution of RPSS (%) for the hindcasts
       averaged over the period of 1971-2000
            RCM Forecast




http://www.funceme.br/DEMET/index.htm
Real-Time Forecast
Validation
     A Major Goal of Probabilistic Forecasts - Reliability!
           Forecasts should “mean what they say”

                        Confidence Level

            40%                    50%                         60%
                              Bo    No   Ao               Bo     No
       Bo    No    Ao
                                                     Ao
                        Bf   49     41     10   Bf
      46      41   13                                 45        48    15
Bf
                        Nf
Nf    48      36   16                           Nf

Af    37      27   36        25     27     48
                        Af                            31        24    45
                                                Af
Skill comparison between the driving ECHAM forecasts
and the nested RSM forecasts. The RPSS (%) was
aggregated for the whole Nordeste region.
            Potential Values of RCMs for Decision Making
                              Climate Forecast

            Water Supply                         Rain Fed Agriculture

                        Municipal &               Crops         Labor
     Irrigation
                         Industrial


                                                   Forecast Use for:
Permanent    Seasonal   Human Use     Industry
       Crops                                       • Budget Management
                                                   • Drought Insurance
   Forecast Driven, Participatory,                 • Local Planning
    Water Allocation System with
                                                   • Guidance of Crop
   Reliable Contracts, Trading and
                                                   Selection, Seed Release,
       Insurance Mechanisms                        & Area planted
Potential values of RCMs for decision making


Case Studies over Ceara State, Brazil

  Streamflow forecast
  Crop yield forecast
Proposed Framework for Streamflow Forecasting




                                   Block et al. (2009)
Dynamical Downscaling – Results

r = 0.79
Statistical Downscaling – Results

r = 0.73
Dynamical Downscaling – Bias Correction
Dynamically Downscaled Coupled Model Streamflow



SMAP
r = 0.88




Sacramento
SMA

r = 0.84
Statistically Downscaled Coupled Model Streamflow



SMAP
r = 0.87




Sacramento
SMA

r = 0.88
Multi-model Combination Results




PDFs of Climatology (solid) and Pooled ensemble hindcast (dashed) for Jan-June 1991
Observed streamflow shown as dotted vertical line
              bridging Climate into Risk
                     Management
.. crop models need daily time sequences




.. as do malaria models and hydrologic models
wheat yield vs. daily rainfall illustration



                              1975

                              Total rainfall: 394mm

                              Observed yield: 1360 kg/ha





                              1981
                              Total rainfall 389mm
                              Observed yield: 901 kg/
                              ha




                                                 A. Challinor
Network of Rainfall Stations
Model Hindcast vs. Observation
Corn Yield Prediction Using
a) seasonal mean rainfall, b) weather index
                         SUMMARY
  The Regional Climate Model is a useful tool to
   understand climate process at sub-GCM
   scales.
  The possibility exists to enhance information
   to higher spatial and temporal scales using
   RCMs
         requires research! Results are often region and season specific.

  Successful application of RCMS in climate
   risk management require creativity to address
   users’ needs
        Focusing on climate variables that are both relevant and predictable/
         projectable (e.g., dry spells, rainfall frequency, monsoon onset, heat
         wave, …)

								
To top