Assessment of the CFSv2 real-time seasonal forecasts for 2013

1

Assessment of the CFSv2 real-time seasonal forecasts for 2013

Wanqiu Wang, Mingyue Chen, and Arun Kumar

CPC/NCEP/NOAA

2

- Real-time skill assessment

- Improve forecast through post-processing

- Impact of initial condition

- Systematic errors

Relevance

Diagnostics/monitoring of CFS real-time forecasts

3

Outline

1. SST indices

2. Spatial maps

3. Anomaly correlation skill

4. Prediction of sea ice extent minimum

4

1. SST indicies

5

Nino34

DMI

MDR

SST indices

Nino34

Stronger amplitude of both positive and negative phases

Delayed transition of ENSO phases at longer lead-time

DMI

Good forecast for 2010 negative DMI.

Failed to reproduce positive DMI in 2012

MDR

Underestimate the amplitude of warm anomalies during Jan-Jul 2011

Too warm at the beginning of 2014.

6

CFSv2 Nino34 SST raw anomalies

• Nino3.4 SST is near normal during 2013.

7

CFSv2 Nino34 SST with amplitude correction

• Because of the weak amplitude, the correction does not have significant impact on the predicted Nino3.4 SST index.

8

CFSv2 Nino34 SST with PDF correction

• Because of the weak amplitude, the correction does not have significant impact on the predicted Nino3.4 SST index.

9

2. Spatial maps

CFSv2 forecast is at a lead of 20 days or so. For example, forecast for Jun-Jul-Aug is from initial conditions of May 1-10th. Impacts of atmospheric initial conditions should be largely removed.

Anomaly = Total – Clim1999-2010

10

Forecast for MAM 2013

Weak anomalies in the Tropical Pacific. Relatively larger SST anomalies in the Tropical Atlantic.

Westward shift of observed pattern in N. Pacific

Larger amplitude of negative rainfall anomalies in equatorial Central Pacific. Incorrect rainfall parrern over the Nordeste and equatorial Atlantic.

11

Forecast for MAM 2013

Good forecast of negative T2m anomalies over northwest of North America. Poor forecast for Eurasia..

Poor forecast of the Z200 anomaly pattern.

12

Forecast for JJA 2013

CFSv2 failed to capture the spatial coverage of cold SST anomalies in the Tropical eastern Pacific.

CFSv2 produced reasonable pattern of tropical precipitation anomalies.

13

Forecast for JJA 2013

CFSv2 failed to capture the observed warm anomalies over North America and Europe.

CFSv2 failed to capture the assicated Z200 anomaly patter.

14

Forecast for SON 2013

CFSv2 captured the overall SST warmth and spatial anomaly patter..

Precipitation is reasonable, although the amplitude is small in both observation and forecast.

15

Forecast for SON 2013

T2m anomalies in CFSv2 are too weak.

Overall Z200 pattern is not captured by CFSv2, although the model reproduced the positive anomalies in North Pacific.

16

Forecast for DJF 2013/2014

CFSv2 reproduced the observed pattern but with weaker amplitude for the tropical negative anomalies.

CFSv2 predicted the observed above normal rainfall in the Tropical western Pacific and below normal rainfall in the Tropical central Pacific

17

Forecast for DJF 2013/2014

CFSv2 captured overall T2m pattern in North America but failed to produced the observed anomalies in Eurasian continent.

CFSv2 captured the positive Z200 anomalies in polar regions but failed to reproduce the variability in mid-latitudes. It is interesting to see that Z200 anomalies in high latitudes are well reproduced in AMIP runs as well as the 0-day-lead forecasts. Since the AMIP simulation is quite reasonable, does the erroneous Z200 anomalies at 20-day-lead time mean the ocean surface conditions became erroneous after 20 days?

18

Forecast for OND 2013

CFSv2 produced erroneous warm anomalies in the equatorial Central-Eastern Pacific.

CFSv2 incorrectly predicted below (above) normal rainfall over Maritime continent (western Pacific).

19

Forecast for JFM 2014

CFSv2 failed to reproduce the observed negative anomalies.

Yet, the CFSv2 predicted reasonable rainfall pattern in the tropical Pacific, while the prediction of rainfall anomalies in the Indian Ocean and Atlantic is poor.

20

3. Anomaly correlation skill

21

Pattern correlation over tropical Pacific

20S-20N

Incorrect rainfall anomaly pattern in the Pacific (slide 19)

22

Pattern correlation over tropical Indian Ocean 20S-20N

23

Pattern correlation over tropical Atlantic

20S-20N

24

Pattern correlation over tropical oceans

20S-20N

25

Pattern correlation over NH

T2m skill is very changeable with successful prediction for MAM 2012 and low skill for many other periods.

Very low precipitation skill for most of the period

Moderate PNA skill after SON 2011, except for 2013 Spring.

20N-80N

26

4. Artic September sea ice extent

Obs=5.35CFSv2

CFSv2 predicted sea ice extent for September 2013 (106 km2)

Obs=3.61

Observation

CFSv2 underestimated sea ice extent in the prediction from Jan – Aug 2013 initial conditions.