Download - JMA Activity in Sub-seasonal Forecasting
JMA Activity in Sub-seasonal Forecasting
1
Climate Prediction Division / JMAYuhei Takaya
WWRP/THORPEX/WCRP Sub-seasonal to Seasonal Implementation Planning Meeting2-3 December 2011, Geneva Switzerland
Thanks to M. Harada, N. Adachi, S. Matsueda
• Review of plans• Integrated EPS on the next HPC (2012.6-)
• Ongoing activity• More user-oriented and seamless products:
Extreme Forecast Index (EFI), meteogram• MJO diagnostics with Emphasis on sources of
predictability at monthly time scale
Table of Contents
Climate Prediction Division, JMA2
Toward “Seamless EPS (weekly - monthly)”Toward “Seamless EPS (weekly - monthly)”• Further seamless forecast
• Climate information– Forecast products– Monitoring-forecasting
• Model development• Forecast system
– More efficient– More accurate
Review of plans in sub-seasonal forecasting
Climate Prediction Division, JMA3
Images from http://visibleearth.nasa.gov/
Integrated “seamless” EPS on the next HPC
4
Weekly EPS(TL319L60)
2-Week EPS (Early Warning)
Monthly EPS (TL159L60)reforecast
2-Week EPS (TL479L100)reforecast
Monthly EPS (TL319L100)
Week-1 Week-2 Week-3 Week-4
present
FY2014
FY2013
FY201?
2-Week EPS (TL479L100)reforecast
Monthly EPS (TL319L100?)reforecast
2-Week EPS (Early Warning)
1. More user-oriented and seamless products: Extreme Forecast Index (EFI), meteogram Masashi Harada and Yuhei Takaya
Ongoing activity
Climate Prediction Division, JMA5
“Seamless” Climate Information
Seamless climate Information on the extreme weather events
Climate Prediction Division, JMA6http://ds.data.jma.go.jp/tcc/tcc/products/climate/
Past Future
Extreme weather forecasting based on EFI
1
• EFI (Extreme forecast index): Lalaurette (2003)• Measure of the difference between
a probabilistic forecast and a climate distribution
from Lalaurette(2003)Threshold(eg. 10m wind)
Prob
abili
ty n
ot to
exce
ed th
resh
old
climate
forecast
• Definition of the EFI(ECMWF Newsletter, No.107)
Cumulative distributions offorecast and climate
All forecast members are below the 0 percentile of the climate
All forecast members exceedthe 100 percentile of the climate
How to obtain probabilistic distributions?
8
time
Hindcast data(150 members)
Hindcast (reforecast) data(150 members: 5 membersfor each initial day, 30years)
average period(ex. 7days)
JMA operational 1-month forecast(25 members for each initial day)
Forecast CDF
Climate CDF
We produce climate distributions from hindcast runs of two initial days
Examples of EFI-products (1)
• Time series of the EFI(top) and the probability distribution of the forecast and the model climate(7-days average, bottom)
EFI
forecast and model climate
Extreme cold Extreme warm
EFI-Meteogram at London(initial date: 2011/11/10)
EFI-map for 850hPa temperature averaged from 11/11 to 11/16
EFI-Map EFI time series - Meteogram
Examples of EFI-products (2)
10
• Warning map for extreme weather events• Possibilities for various extreme weather events
are summarized on one map.
Extreme: EFI ≧ 0.8Above normal: EFI ≧ 0.5
2. Research related to MJO• MJO diagnostics: Emphasis on sources of
predictability at monthly time scale Satoko Matsueda and Yuhei Takaya
• Case study: MJO influence on extratropic circulation
Kengo Miyaoka and Yuhei Takaya Shuhei Maeda
Ongoing activity
Climate Prediction Division, JMA11
MJO index skill
MJO Skill of JMA monthly models•correlation coefficient falls below 0.6 on day 13•predicted phase speed is faster than observed phase speed•predicted amplitude is smaller than observed amplitude
12
RMSE
COR
Phase error (PERR)
Relative amplitude difference (AERR)
Lead time (day)
faster
slower
larger
smaller
MJO Life cycle composite OLR/Wind200 Winter
analysis model FT=2week
13
Lag-correlation of OLR/U850 in Winter
analysis
Eastward propagation of OLR/U850 anomaly is not well simulated.
Fig. Lag correlation of intraseasonal OLR (shaded) and U850 (contour) averaging 10S-10N at all longitudes against OLR and U850 at an Indian Ocean reference point (OLR:10S-5N,75-100E, U850:1.25-16.25S,68.75-96.25E).
For hindcast, a forecast time of 10 days corresponds lag = 0.
model (FT=10day : lag=0)
14
Lag correlationOLR : shadedU850 : contour
Lag correlationOLR : shadedU850 : contour
FT=15dayFT=10day
FT=1day
1800 0 1800 0
0
15
-15
0
15
-15
Case study of MJO influence on the extratropic circulation
Climate Prediction Division, JMA15
Hovmöller diagram of 200-hPa velocity potential anomaly averaged over 10N-10S. (2011/9/1-11/10)
The second strongest MJO during last 30 years
[W/m2]
Wave train dispersed from wave source generated by MJO
16
OLR & Velocity Potential 200-hPa anomalies (2011/10/29 - 11/2)
5-day running mean temperature anomalies (Sep.-Nov. 2011)
Sep. Oct. Nov.
NorthernJapan
EasternJapan
WesternJapan
Okinawa & Amami
[W/m2]
Temperatures in Western Japan (+3.4) and Okinawa & Amami area (+2.2) were highest records (since 1961) for the first 10 days of Nov. 2011.
Was the event predicted?JMA monthly forecast for 10/28-11/3 (I.C. 10/20)
850-hPa temperature Western Japan
200-hPa velocity potential and divergent wind anomalies
200-hPa stream function, rain anomalies
The week-2 forecast (I.C.: Oct. 20th) successfully predicted the event.
850-hPa temperature
[K]
[K]
[mm/day]
10-6 [m2/s]
CHI200 analysis T850 analysis T850 JMA model
When active convection is in Indian Ocean, T850 anomaly pattern can be reproduced in Asia.
0 73-3-7
Composite : initial Phase 2 (active convection in Indian Ocean)
Influence of MJO on Asian climate
18
FT= 6day
FT= 1day
[K]10-6 [m2/s]
• Review of plans• Integrated “seamless” EPS on the next HPC (2012.6-)
• Ongoing activity• More user-oriented and seamless products:
Extreme Forecast Index (EFI), meteogram• Emphasis on sources of predictability
at monthly time scale (MJO diagnostics)
Summary
Climate Prediction Division, JMA19