Dynamical Long-range Forecast OSE Tomoaki Climate Prediction Division Japan Meteorological Agency

１． Introduction

DEFINITIONS OF METEOROLOGICAL FORECASTING RANGES 1. Nowcasting A description of current weather parameters and 0 -2 hours description of forecasted weather parameters 2. Very short-range weather forecasting Up to 12 hours description of weather parameters 3. Short-range weather forecasting Beyond 12 hours and up to 72 hours description of weather parameters 4. Medium-range weather forecasting Beyond 72 hours and up to 240 hours description of weather parameters

These are partly the Work of Climate Prediction Division in JMA 5. Extended-range weather forecasting Beyond 10 days and up to 30 days 6. Long-range forecasting From 30 days up to two years 7. Climate forecasting Beyond two years (Climate variability prediction, Climate prediction)

Difference between Short- and Long-range Forecast Long-range weather forecast (from 30 days up to two years) describes averaged weather parameters, expressed as a departure (deviation, variation, anomaly) from climate values for that period. On the other hand, forecast up to 10 days, such as nowcasting, very short-range weather forecast, short-range weather forecast, and medium-range weather forecast, describe weather parameters ( not deviation, not averaged ). Note that extended-range weather forecasting ( beyond 10 days and up to 30 days ) describes weather parameters, usually averaged and expressed as a departure from climate values for that period.

NWP systems for extended forecast and long-range forecast in JMA One month forecast (T106) of a few days lead –Once a week, Operational since March 1996 –Extension of medium range forecast Three month forecast (T63) of a few weeks lead –Once a month,Operational since March 2003 Six month forecast (T63) of a few weeks lead –Twice a year (Feb. and Sep.), –Operational since September 2003 –Extension of three month forecast

Detailed Operation Chart of 3- and 6-month EPS 120days integration for Three Month Outlook. Feb MarApr Sep OctMay Nov JanMar JunJulAug DecJanFeb Extended integration for Warm Season Outlook(JJA). Extended integration for Cold Season Outlook(DJF). 3-month EPS : Started in March Executed every month. 6-month EPS : Started in September Executed 5 times a year(February March and April for JJA, September and October for DJF)

Dynamical Long-range Forecast makes use of numerical weather prediction (NWP) model. Introduction of Dynamical method to Long-range forecast Long-range forecast makes use of a reduced horizontal resolution version of NWP model for short-range and medium range forecast. The same physical processes such as cumulus parameterization, radiation and cloud, boundary layer, gravity wave drag, and so on are used. Development of the NWP model is cooperation of Numerical prediction Division, Climate Prediction Division, and Meteorological Research Institute.

Different Aspects of Long-range Forecast Model from numerical weather prediction (NWP) model. Introduction of Dynamical method to Long-range forecast Long-range Forecast Model Short-range Forecast Model Equilibrium State (Climate) Transient State (Weather) Cumulus and SST anomaly Cumulus and Disturbances Ocean and Land Process Initial Condition in Atmosphere

2 ． Predictability

Uncertainty of Forecast predictability Errors in Initial Condition –Errors in Raw Observational Data –Errors in Objective Analysis Procedure –Sparse Observation over Ocean Errors in Forecast Model –Limitation in the Spatial Resolution –Errors in Physical Processes

Concept of the Ensemble forecast Multi-Initials within Errors in Observation Initial Errors in Observation Spread Most Likely Forecast Truth Probability Forecast Ensemble Mean

Advantage of Ensemble Prediction System (EPS) Probability forecast –Intrinsically stochastic behavior of atmosphere can be predicted with ensemble method. Forecast with physical consistency –NWP model can represent global circulation in a physically consistent way. Improvement based on advance of technology –Observation, Study on climate system, Model, Computational power, ……….

2 ． One-Month Prediction Since March Ensemble Prediction - - Land Surface Assimiliation - - Improved Cumulus Scheme -

Dynamical Seasonal Forecast System One-Month Prediction Ensemble Prediction System (Atmosphere-Land) Persistent SST Anomaly Initial Atmosphere Land Ocean Products Guidance Map Verification Analysis Systematic Error Verification Seasonal Forecast Experiments (Hind-cast) Boundary Condition

Ensemble One-Month Ensemble Forecast (An Example ） Temperature at 850hPa over Eastern Japan Observation Forecast Feb Mar Apr Forecast Observation

2 ． One-Month Prediction Since March Ensemble Prediction - - Land Surface Assimiliation - - Improved Cumulus Scheme -

Global Snow Depth Analysis System (Kurino and Tokuhiro, 2003) Global Atmosphere Assimilation System Land Surface Assimilation （ ＵＴＣ） Snow Depth Ground Surface/Soil Temperature Soil Wetness Atmospheric Forcing ( UTC ） Snow (00UTC) Simple Biosphere Model (T106) Dynamical Seasonal Prediction Input Land Surface Model Initials First Guess SYNOP Snow Depth Snow-Depth Analysis Soil WetnessSnow Depth Shortwave RadiationPrecipitation SSM/I Snow Cover

Land Assimilation System Snow Depth based on Land Assimilation System on Jan Soil Wetness based on Land Assimilation System on Jan

Improved Skill for One-Month Forecast （ Initial Date 2001/5/31/12Z ） Correlation between Forecast and Observation over Eurasia Effect of Global Snow Depth Analysis System With Without Temperature at 850 hPaGeopotential Height at 500 hPa June 1stJune 30June 28June 1st

Improved Skill for One-Month Forecast （ Initial Date 2001/10/31/12Z ） Correlation between Forecast and Observation over Eurasia Effect of New Global Snow Depth Analysis System New Old Temperature at 850 hPaGeopotential Height at 500 hPa Nov 1stNov 30 Nov 1st

2 ． One-Month Prediction Since March Ensemble Prediction - - Land Surface Assimiliation - - Improved Cumulus Scheme -

・ Entrainment and Detrainment in Downdraft (Nakagawa and Shinpo, 2003) Entrainment, Detrainment Convective Downdraft with TRMM without To moderate downdraft effect

3 ． 3- and 6- Month Prediction Since March 2003 and September Two Tiered Method System - - Skill in Hindcast Experiment -

Originates from Initial condition Deterministic forecast fails beyond two weeks due to the growth of errors contained in the initial states. Chaotic behavior of atmosphere comes from its strong non-linearity. Originates from lower boundary condition Effective for longer time scale; Month to season predictability Two kinds of Atmospheric Predictability

Predictability Predictability of 1st kind Predictability of 2nd kind +Fx +Fy +Fz The whole structure of Attracter (equilibrium) is modified by External Forcings Analogy

Initial Condition Boundary Condition Importance Hour Day Week Month Season Year Average Time Scale Meso Typhoon Tropical disturbances Intra-seasonal Oscillation predictability Relative importance of Initial Condition and Boundary Condition 1-Month 3-Month ENSO Global Warming Predictability of 1st kind Predictability of 2nd kind

◎ Ocean Sea Surface Temperature (SST) Sea Ice ◎ Land Surface Soil Temperature Soil Moisture Snow Cover, Snow Depth Vegetation ( Grass, Tree etc. ) Most IMPORTANT to the atmospheric variability ! predictability Lower Boundary Condition of Atmosphere

3 ． 3- and 6- Month Prediction Since March 2003 and September Two Tiered Method System - - Skill in Hindcast Experiment -

Present Two-Tiered Way Two Methods for Numerical Seasonal Prediction Future One-Tiered Way Atmosphere-Land-Ocean Coupled Model Atomosphere-Land Coupled Model Atomosphere-Land Coupled Model Atmospheric Model Ocean model is coupled Separately predicted SST is prescribed as boundary

Tactics 1 CGCM Tactics 2 Two-tiered method Merit : Ideal if SST prediction is correct. Defect : (1) SST errors cannot be corrected. (2) Needs large computer resources. Merit : (1) Predicted SST can be corrected. (2) computer resources can be saved. Defect : Air to sea interactions are neglected for atmospheric prediction. predictability Merits and Defects of Two Tactics

NINO3 SST Prediction Skill New Model ： 88 cases Jan Jan Old Model ： 116 cases Feb Nov Persistency Forecast July 2003

1 month forecast Spring MAM Case Smallest RMSE RED : Coupled Model YELLOW : Persistence GREEN : Climate predictability Best SST Forecast 2-month forecast 4-month forecast 5-month forecast

Dynamical Seasonal Forecast System 1- and 3-Month Prediction Ensemble Prediction System (Atmosphere-Land) Persistent SST Anomaly Initial Atmosphere Land Ocean Products Guidance Map Verification Analysis Systematic Error Verification Seasonal Forecast Experiments (Hind-cast) Boundary Condition

Dynamical Seasonal Forecast System 6-Month Prediction Ensemble Prediction System (Atmosphere-Land) El Niño Forecast model (Atmosphere-Ocean Coupled Model) Statistical Process for Global SST Initial Atmosphere Land Ocean Initial SST Products Guidance Map Verification Analysis El Niño SST Forecast Boundary Condition Systematic Error Verification Seasonal Forecast Experiments (Hind-cast)

SST anomalies are persisted 3-month EPS Mixed Predicted Persisted 6-month EPS Persisted Predicted 1st 2nd3rd Lead time (month) 4th5th6th Detailed Description of Sea Surface Temperature Mixed

Specification of the global spectral model for extended- and long- range forecast at JMA

3 ． 3- and 6- Month Prediction Since March 2003 and September Two Tiered Method System - - Skill in Hindcast Experiment -

Hindcast Introduction of Dynamical method to Long-range forecast “Hindcast” is a set of systematic forecast experiments for past cases. Hindcast is performed to estimate skill of the model. JMA is now performing hindcast in preparation for operational dynamical long-range forecast: Initial time is the last day of every month from Jan to Dec (18 years) with 5 members each (12*18*5=1080 cases).

Model performance What can a model predict ? Response of atmosphere to the slowly varying boundary conditions Where does the signal of long-range forecast come from ? Especially, the deviation of SST in the tropics such as ENSO → Deviation of convective activity of large scale → Deviation of divergence of large scale → Deviation of tropical circulation direct and indirect influence on the circulation in the mid- and high- latitudes Circulation in Tropics and mid- and high-latitudes

North-south mean u,v,w Kelvin Wave Rossby wave Walker Circulation heating u, v, p in the low level * tropical circulation is depicted approximately. Symmetric heating on the equator Model performance Linear response on heating in the tropics : Matsuno-Gill response Gill.A.E.(1980) QJRMS,

forecast analysis An example of forecast in the tropics 1988 Sept. 30 initial, 90-day mean (21-110days) anomaly of velocity potential 200hPa anomaly of stream function 200hPa anomaly of stream function 850hPa Red:convergence Blue:divergence Red:clockwise Blue:counter-clockwise (in Northern Hemisphere) Red:clockwise Blue:counter-clockwise (in Northern Hemisphere) In a case of La Nina Model performance

El Niño in the Eastern Tropical Pacific -----> Convective area shifts towards mid- and eastern Pacific -----> High is formed -----> Strengthened Aleutian Low -----> High over Rocky Mts > Low in the eastern USA Model performance An example of Tele-connection : PNA

Day: 1-90 Initial:5.31 Summer 1998 Precipitation (anal)850hPa stream function (anal) 850hPa stream function (fcst) Precipitation (fcst)

4 ． Dissemination of Forecast Products On Tokyo Climate Center (TCC) / JMA

Dynamical Forecast and Verification

1-Month Forecast 3-Month Forecast 6-Month Forecast

Usage of GPV Effective Usage of Global GPV From global GPV to regional/local temperature/precipitation forecasts - Downscaling by Regional model - Downscaling by Statistical method - Statistical guidance to forecasters Needs for capacity building

Guidance : Statistical downscaling Ensemble prediction Linear regression Forecast or probability of temperature, precipitation etc. at desired forecast area Results of hindcast or Objective Analysis Observed temperature, precipitation etc. at desired forecast area Linear regression Usage of GPV