John Janowiak Climate Prediction Center/NCEP/NWS Jianyin Liang China Meteorological Agency Pingping Xie Climate Prediction Center/NCEP/NWS Robert Joyce CPC / RS Information Systems IPWG-3 -- Melbourne, Australia October 24, 2006 *CPC Morphing Technique Rain Gauge Data Merged with CMORPH* Yields: RMORPH
(microwave) ? ? 1.5 hours apart GPM offers 3-hr sampling … to get finer temporal sampling … take advantage of 30-minute sampling afforded by Geo-IR data Premise: Error in using IR to interpolate precip. features identified by PMW < Error in deriving precip. directly from IR So avoid deriving precipitation estimates directly from IR …
(microwave) ? ? Derive motion vectors from ½ hourly IR Apply motion to PMW-derived precipitation “Morph” (Joyce et al., J. Hydromet, 2004) 1.5 hours apart
Biases in Satellite Estimates
Gauge-CMORPH Merging Algorithm Step 1: Bias Correction AssumptionsAssumptions - Biases relatively stable over a region and time - Biases can be approximated as ratios between the estimates & gauges ProceduresProcedures - Performed once a day using data for all 24 hourly slots - Each day: RATIO = GAUGE / CMORPH (last 30 days; each gauge) - Optimal Interpolation (OI) technique (Gandin 1965) applied to ratios - “Un”biased CMORPH: CMORPH x RATIO analyzed
Gauge-CMORPH Merging Algorithm Step 2: Combining Gauge & Satellite Data Bias-corrected satellite estimates and gauge data combined via Optimum Interpolation Technique (“OI”) Bias-corrected satellite estimates and gauge data combined via Optimum Interpolation Technique (“OI”) -Bias-corrected satellite estimates used as first-guess -Bias-corrected satellite estimates used as first-guess - Gauge data are incorporated - Gauge data are incorporated - Relative weighting at a grid box is a function of: - Relative weighting at a grid box is a function of: - quality of satellite estimates at the grid box; - density of local gauge network density
Proof-of-Concept: Guang-Dong Province over Southern China South China Sea Tibet plateau Topography Guang-Dong
Hourly precipitation reports from 394 stations over ~150,000 km 2 (~380km 2 /gauge)
An Example for 03Z, May 5, 2005 GAUGE ONLYORIGINAL CMORPH BIAS CORRECTED MERGED
An Example for 03Z, May 5, 2005 GAUGE ONLYORIGINAL CMORPH BIAS CORRECTED MERGED
Mean Precipitation from April 1 – June 30, 2005 GAUGE ONLYORIGINAL CMORPH BIAS CORRECTED MERGED
Hourly Gauge-Satellite Merged
PDF of Hourly Precipitation for April - June, 2005 Frequency of No-Rain EventsFrequency of No-Rain Events Gauge Station: 83.9%Gauge Station: 83.9% Gauge Analysis:81.5%Gauge Analysis:81.5% Original CMORPH:77.3%Original CMORPH:77.3% Gauge-CMORPH Merged:83.3%Gauge-CMORPH Merged:83.3% Frequency of Events with RainFrequency of Events with Rain
Dense Gauge Locations: Disaggregate Gauge Data Use hourly CMORPH to partition daily gauge amounts into hourly amounts (i.e. “disaggregate”) - United States - Australia - China?
Valid for 24 hrs ending 12z August 8, 2006
-Daily RMORPH constrained to daily gauge amount - Gauge data partitioned into hourly amounts -Spurious coverage of light gauge amounts reduced Valid for 24 hrs ending 12z August 8, 2006 Daily sum of hourly amounts
Where to from Here? 1.Transition regional prototype to global 2.Experiment with OI tuning parameters 3.Cross-validation testing 4. Explore bias-adjustment for oceanic precip - Normalize estimates to TRMM “2B31” (TMI/PR) - ATLAS buoys? - Radar?
Thanks for listening
An Example for 03Z, May 5, 2005 GAUGE ONLYORIGINAL CMORPH BIAS CORRECTED MERGED
An Example for 03Z, May 5, 2005 GAUGE ONLYORIGINAL CMORPH BIAS CORRECTED MERGED
(microwave) (in time)
Observed
1/3 + 2/3 + 1/3