Sunseasonal (Extended-Range) Forecast of the Asian Summer Monsoon Rainfall
Song Yang (杨崧)
Department of Atmospheric Sciences
Sun Yat-sen University
Guangzhou, China
Model, Data and Others
4
3
2
1
5
Quasi-Biweekly Oscillation
Regional Monsoons
Tropical Land and Ocean Rainfalls
Southern China Early-Season Rainfall
CONTENT
3Coauthors and Reference 3
1. Jia, X. and S. Yang*, 2013: Impacts of the quasi-biweekly oscillation over the western North Pacific on East Asian subtropical monsoon during early summer. J. Geophys. Res., 118, 1-14.2. Jia, X., S. Yang* & Coauthors, 2013: Prediction of global patterns of dominant quasi- biweekly oscillation by the NCEP Climate Forecast System version 2. Climate Dyn., 41, 1635-1650.3. Liu, X., S. Yang* & Coauthors, 2013: Diagnostics of sub-seasonal prediction biases of the Asian summer monsoon by the NCEP Climate Forecast System. Climate Dyn., 41, 1453-1474.4. Liu, X, S. Yang* & Coauthors, 2014: Subseasonal forecast skills of global summer monsoons in the NCEP Climate Forecast System version 2. Climate Dyn., 42, 1487-1508.5. Liu, X., S. Yang* & Coauthors, 2015: Subseasonal predictions of regional summer monsoon rainfalls over tropical Asian oceans and land. J. Climate, submitted.6. Zhao, S. and S. Yang*, 2014: Dynamical prediction of the early-season rainfall over southern China by the NCEP Climate Forecast System. Wea. Forecasting, 29, 1391-1401.7. Zhao, S., S. Yang* & Coauthors, 2015: Skills of yearly prediction of the early-season rainfall over southern China by the NCEP Climate Forecast System. Theor. & Appl. Climatol., in press.
4US NCEP Climate Forecast System (CFSv2)
4
Atmosphere GFS2009 (T126/L64)
Land NOAH 4-L
Ocean MOM4
Sea ice Predicted
CO2Evolving with time
Initial conditions
CFS Reanalysis (CFSR)
Hindcast ~24/month (4 runs / 5 days)
Forecast4 runs/day (seasonal)16 runs/day (45 days)
Hindcast Data Used:
Daily data, 1999-2011
Four members every day, integrated for 45 days
Model, Data and Others
4
3
2
1
5
Quasi-Biweekly Oscillation
Regional Monsoons
Tropical Land and Ocean Rainfalls
Southern China Early-Season Rainfall
CONTENT
6QBWO in AVHRR, CFSR and CFSv2
10-20-day variance of OLR in boreal summer for NOAA AVHRR, CFSR, and CFSv2 predictions at various leads
7South Asia High & E-SE Asia Convection
Quasi-biweekly variability of the South Asia High is important for convergence/divergence over East & Southeast Asia
Contours: H200
Shading: Difference in 200-hPa Divergence
9Prediction of Eight QBWS Modes
High skills for predicting the North Pacific and South Pacific Modes, but low skill for predicting the Asian summer monsoon
Best:
North Pacific South Pacific
Worst:
Asian Monsoon Central America South Africa
10Prediction of QBWO for El Nino & La Nina Years
High skill for El Nino years, but low skill for La Nina years
Model, Data and Others
4
3
2
1
5
Quasi-Biweekly Oscillation
Regional Monsoons
Tropical Land and Ocean Rainfalls
Southern China Early-Season Rainfall
CONTENT
Indian Monsoon & SCS Monsoon
12
Precipitation (70ºE-90ºE) Precipitation (110ºE-130ºE)
More Effect from the Tropical Indian Ocean SST (Boundary Forcing)
More Effect from the Subtropical Western Pacific High (Internal
Dynamics)SST forcing is important for skills of subseasonal (extended-range) forecast of regional monsoon rainfalls
Webster-Yang Index & Goswami et al. Index
13
Prediction skill is high when regional monsoon is strongly related to large-scale features
Multivariate EOF Analysis (Rainfall & V850)
14
PCs: From short leads (red) to longer leads (blue)
(1) Prediction skill is a function of the stage of monsoons(2) An abrupt turning point of bias in late June and early
July
Model, Data and Others
4
3
2
1
5
Quasi-Biweekly Oscillation
Regional Monsoons
Tropical Land and Ocean Rainfalls
Southern China Early-Season Rainfall
CONTENT
Pattern Correlation 17
Multi-year
Average
Individual Years
Higher skills over ocean domains, esp. the Arabian Sea
Lower skills over land domains, esp. the Indo-China Peninsula
Temporal Correlation 18
Again, higher skills are over ocean domains, esp. the Arabian Sea, and lower skills are over land domains, esp. the Indo-China Peninsula
Corr. between Precip & Ts 19
Positive: Enhanced radiation => increased Ts => unstable => convectionNegative: Enhanced rainfall => declined radiation => decreased Ts
Overestimation worsens with lead time (e.g. Arabian Sea)
Positively significant over Arabian Sea
Negatively significant over land except southern China
Corr with Ts, and Reg of V850 on P Indices20
1. Overestimated relationships of rainfall with Ts and atmospheric circulation
2. Problem worsens with increased lead time
Cross Correlations: Supporting Evidence21
1. Larger correlation appears for neighboring regions compared to more remote regions
2. Longer range predictions capture larger scale features
Lag Corr (Precip/Ts) 22
1. Largest correlations at leads or lags by 1-2 pentads
2. Ts forcing in AS, but Ts response in BOB and SCS
3. Ts response in Indian Peninsula, but weak air-land interaction in southern China
4. Changes with lead time?
Other Features 23
Regional rainfall over oceans is related to larger-scale circulation patterns, compared to that over land
As lead time increases, strengthening connections between regional rainfall and large-scale circulation are found over extensive regions, and the regional independence of rainfall variability is gradually obscured by uniform large-scale features.
Comparisons between skillful and unskillful forecasts indicate that the regional characteristics of rainfall and model’s deficiencies in capturing the relationship between small- and large-scale features are responsible for the regional discrepancies of actual subseasonal predictability.
Model, Data and Others
4
3
2
1
5
Quasi-Biweekly Oscillation
Regional Monsoons
Tropical Land and Ocean Rainfalls
Southern China Early-Season Rainfall
CONTENT
Simulations of SC Early-Season Rainfall25
Rainfall for Pentads 19-36
Pentad Rainfalland
Model-Obs Difference
Prediction of SC Rainfall (Skill of 8-14 Days)
26
Higher Skill for April-May than for June
Higher Skill for Southern China than for Other Regions
Prediction of SC Early-Season Rainfall 27
Difference in Predicted Rainfalls (Lead Time 15-29 Days Minus Lead Time 0-14 Dyas)
Yearly Prediction of SC Early-Season Rainfall
28
Higher Skills in 2005 & 2006Lower Skills in 2001 & 2010
Small Difference in LD 0-4 Prediction and LD 0-14 Prediction
Detailed Features for Years of High & Low Skills
29
Years of High Skills2005 & 2006
Years of Low Skills2001 & 2010
Prediction of Surface Temp for Pentads 1-12
31
High Skills for Rainfall Prediction Low Skills for Rainfall Prediction
In the years with high (low) skills of predicting SC early-season rainfall, the skills of predicting the previous surface temperature over the
tropical western Pacific are high (low).