An Evaluation of Aspects of Tropical Precipitation Forecasts

An Evaluation of Aspects of Tropical Precipitation Forecasts from the ECMWF & NCEP Model Using CMORPH

John Janowiak1, M.R.P. Sapiano1, P. A. Arkin1, F.J. Turk2

1 Cooperative Institute for Climate and Satellites (CICS) Earth Systems Science Interdisciplinary Center (ESSIC) University of Maryland, College Park, Maryland, USA

2 Jet Propulsion Laboratory Pasadena, California, USA

Outline• What is CMORPH?

• The Diurnal Cycle in GFS Precipitation

• MJO-related convective precipitation from:- CMORPH (observations)- ECMWF forecasts (5-, 10-day)- GFS forecasts (5-, 10-, 15-day)

• Conclusions

CMORPH*NOAA/CPC “Morphing” technique

Provides quantitative estimates of precipitation for 0.07o x 0.07o lat/lon / ½ hr( ~ 8 km @ equator)

Uses geostationary IR to propagate & ‘morph’ (interpolate/smooth in time and space) precipitation estimated from passive microwave observations

Dec 2002 – present; extending back to ~1998

* See Joyce et al. (J. Hydromet 2004)

RADAR CMORPH

Hourly Precipitation Loops: 15Z 8Jun2008 – 06Z9Jun2008

0.25o lat/lon 0.07o lat/lon

mm/hr

CMORPH

CMORPH looks quite realistic over water, maybe less so over land

Probably related to fact that microwave estimates are better over water than land



• MJO-related convective precipitation from:- CMORPH (observations)- ECMWF forecasts (5-, 10-day)- GFS forecasts (5-, 10-, 15-day)

• Conclusions

09 12 15 18 21 00 03 06 LST

CMORPH

GFS

1 day fcst

09 12 15 18 21 00 03 06 LST

CMORPH

GFS

1 day fcst

09 12 15 18 21 00 03 06 LST

CMORPH

GFS

1 day fcst

09 12 15 18 21 00 03 06 LST

CMORPH

GFS

1 day fcst

07 10 13 16 19 22 01 04 LST

CMORPH

GFS

1 day fcst

12 15 18 21 00 03 06 09 LST

CMORPH



• MJO-related convective precipitation from:- CMORPH (observations)- ECMWF forecasts (5-, 10-day)- GFS forecasts (5-, 10-day)

• Conclusions

Case Study:

Moderate-Strong MJO

Nov 2007 – Feb 2008

CMORPH

Precipitation from Indian Ocean across the Pacific to Greenwich

Seasonal mean removed

MJO signatures clearly evident

Diagonal lines subjectively drawn to identify axis of MJO (and intervening dry periods) eastward progression

15N-15S

These lines identify westward moving elements within MJO envelope

15N-15S

Case Study:

Mod-Stg MJO

Nov 2007 – Feb 2008

CMORPH

Line are same as previous slides; on model plots, lines represent observed features

~10days

Dec 4-15, 2007

Dec 16 – Jan 3

Jan 5-20, 2008

Difference from Nov 2007 – Feb 2008 Period Mean

Dec 4-15, 2007

Dec 16 – Jan 3

Jan 5-20, 2008



A

B

C

Dec 4-15

CMORPHGFS 10 dyECMWF 10 dy

(5 dy smoothed)


A

B

C

Dec 16-Jan 3

CMORPHGFS 10 dyECMWF 10 dy

(5 dy smoothed)


A

B

CJan 5-20

(5 dy smoothed)CMORPHGFS 10 dyECMWF 10 dy

0

1

2

3

4567 12

0

8 910

11 13 1415

These show pattern correlations over the region between forecasts and observations for different lags (the different colored lines) and for different forecast lead time (the horizontal axis)

The green line labeled “1” represents the correlation between forecasts initialized one day later than the observations they are compared to

“Interesting if true” – we are working to figure out what this might mean

Conclusions

• GFS precipitation forecasts over the tropical oceans do exhibit a diurnal cycle– But the peak occurs earlier than observations – And the amplitude decreases with forecast lead, at least in the central Pacific

• Both GFS and ECMWF exhibit a reasonably realistic MJO precipitation pattern and variability

– At longer leads, both models lose details and seem to lag behind the observations – Possible that the initialization is imperfect and some spin-up is required to attain a more

realistic precipitation field?

• Results, particularly for ECMWF, indicate that useful skill in predicting MJO-related precipitation is close to being attained

Documents

An Evaluation of Aspects of Tropical Precipitation Forecasts