Validation of Satellite-Derived Rainfall Estimates andNumerical Model Forecasts of Precipitation over the US
John Janowiak Climate Prediction Center/NCEP/NWS
2nd Int’l Precipitation Working Group - October 26, 2004
Work is modeled after the pioneering effort of Dr. Beth Ebert (BMRC/Australian BOM)
www.bom.gov.au/bmrc/wefor/staff/eee/SatRainVal/dailyval_dev.html
U.S. Validation at:
www.cpc.ncep.noaa.gov/products/janowiak/us_web.shtml
www.cpc.ncep.noaa.gov/products/janowiak/us_web.shtml
www.cpc.ncep.noaa.gov/products/janowiak/us_web.shtml
Validation Data Set
- 7000+ station reports daily
- 06Z – 06Z accumulation period
- Data analyzed using a Cressman-type scheme
- Error characteristics of validation data are NOT known
- Validation area matched for all estimates
(if missing in one, made missing in all)
Typical Station Distribution
Validation Results
Cold Season Precipitation Amt. (Jan 2004)
Cold Season Precipitation Diff. (Jan 2004)
Warm Season Precipitation Amt. (Jun 2004)
Warm Season Precipitation Diff. (Jun 2004)
CPC gauge analysis ( Aug 2003)CPC gauge analysis ( Aug 2003)
CMORPH analysis ( Aug 2003)CMORPH analysis ( Aug 2003)
CMORPH with evap. adjustmentCMORPH with evap. adjustment
Bias Ratio (areal coverage)
BIAS Ratio (estimated mean / gauge mean)
Mean precip. for entire US (not to scale)
Contribution to June 2004 Total Rainfall by Daily Rainfall Amount
Heaviest 10% of daily rainfall events
CONCLUSIONS
1. Merging PMW & IR estimates provides more accurate estimates ofprecipitation than the separate components can
CONCLUSIONS
1. Merging PMW & IR estimates provides more accurate estimates ofprecipitation than the separate components can
2. Two major systematic biases are apparent in the satellite estimates:a. OVERestimation over snow-covered regionsb. OVERestimation in semi-arid regions during the warm season
CONCLUSIONS
1. Merging PMW & IR estimates provides more accurate estimates ofprecipitation than the separate components can
2. Two major systematic biases are apparent in the satellite estimates:a. OVERestimation over snow-covered regionsb. OVERestimation in semi-arid regions during the warm season
3. NWP forecasts generally outperform blended satellite estimates and radar during the winter season over the U.S.
The End
Effects of Interpolating the Data
CMORPH vs. gauge over ‘NAME*’ zones
*North American Monsoon Experiment (2004)
CPC gauge analysis ( Aug 2003)CPC gauge analysis ( Aug 2003)
CMORPH analysis ( Aug 2003)CMORPH analysis ( Aug 2003)
CMORPH with RH adjustment vs. gauge over ‘NAME’ zones
Distribution of Daily Precipitation Amounts for June 2004
45 50 55 60 65 70 75 80 85 90 >90
Distribution of Daily Precipitation Amounts for Jan 1-22, 2004
Bias Ratio (areal coverage)
BIAS Ratio (mean radar/ mean gauge)