Ying Lin, NCEP/EMC

Last modified: 12 Apr 2007

For a quick overview of the current state of things, data, images, go to Section 4.


In support of the National Weather Service’s (NWS) effort to disseminate forecast products and verifying forecasts on the 5km National Digital Forecast Database (NDFD) grid, NCEP plans to modify processing for the existing multi-sensor Stage II/Stage IV precipitation analyses to produce the analyses and possibly some measure of quality mark/confidence factor for the Stage II on the NDFD grid, for


The NCEP Stage II is produced directly from the real-time gauge and WSR-88D radar data received at NCEP. Each hour’s analysis is performed three times: the “early” version merges radar data with hourly gauge reports in the METAR data, and is available at 35 minutes after the top of the hour. The analysis is then re-run 6 hours and again 18 hours later, using the radar data and HADS automated hourly gauge observations transmitted through the GOES data collection platform.

The NCEP Stage IVanalysis uses the hourly/6-hourly analyses transmitted from the 12 ConUS RFCs to NCEP throughout the day.  Each RFC typically sends out 4-6 batches of data to NCEP each day, with a mixture of manually QC'd analysis for earlier hours and the first-run (automated, no manual QC) analysis for more recent hours.  At NCEP, the Stage IV mosaic for each hour/6-hour period is updated hourly to include any newly-arrived regional analysis.  The Stage IV analysis might first become available 35 minutes after the end of the accumulating hour, containing analysis from only 1 or 2 RFC(s), and gradually fill up in later hours, as analyses from other RFCs come in and are added into the Stage IV.

Both the Stage II and Stage IV are produced on the 4-km NWP HRAP (Hydrologic Rainfall Analysis Project) grid.


  1. For a quick 'first draft' of the analysis for RTMA, modify the current Stage II job to map the "early" version of the Stage II analysis, run at 35 minutes after the top of the hour, to the NDFD grid.  Push the product to TOC.  At this time I do not plan to run the job earlier (to less than 30 minutes past the hour) because the METAR files are not available to us until 30 or 31 minutes after the hour.

  2. Acquire and implement OHD's latest MPE code to upgrade NCEP Stage II analysis

  3. Implement the gauge QC software from FSL - currently only an gross error check is done on the hourly gauges used for the Stage II (OHD's MPE code has some additional gauge QC)

  4. Confidence factor/qualitymark for the analysis: asking OHD to develop one

  5. Down the road, if/when there is enough demand for the precipitation RTMA to be available within 30 minutes after the hour, find out whether METAR file receipt time can be moved up a few minutes earlier

  6. The 2-day delayed precip analysis for the AOR (strongly suggest doing this after completion of steps a-c above): the hourly Stage IV product would be a better choice than the NCEP Stage II, even after we upgrade to the MPE, because of the manual QC done at the RFCs, and at the western RFCs their choice of Mountain Mapper algorithm might be more suitable for their geography than the MPE.  One issue: not all RFCs send all the hourlies, and some hourlies are the early/automated (no QC) version only.  Should get all RFCs to send final version of hourlies before the AOR can be considered a complete product.

  7. OHD and NSSL have been developing a new precipitation product called NMQ/Q2 (Seo et al., 2005). When this product becomes operational, use it instead of the Stage II/IV for the precipitation RTMA/AOR.


The precipitation RTMA files are made at NCEP operationally at around 35 min after the top of the hour by mapping the Stage II analysis to the ConUS NDFD grid. The analysis is packed in GRIB2 format, with the name of pcprtma.YYYYMMDDHH, where YYYYMMDDHH is the ending time of the one-hour accumulation. Unit for the field is kg/m2.

Files and images for the past 24h or so are available in directory nam_pcpn_anal.$yyyymmdd under

The files are also available on TOC.


2005/02/23: Began routine mapping of the Stage II analysis onto the NDFD grid in a non-operational job, as the precipitation RTMA

2005/03/23: Began posting the precipitation RTMA on to NCEP's non-operational FTP site

2005/08/23: Began producing GRIB2 version of the precipitation RTMA

2005/10/26: Filed JIF to implement precipitation RTMA in operational production

2005/11/02: Starting at 19Z, the name of the GRIB2 precipitation RTMA file is changed to pcprtma.YYYYMMDDHH, instead of ST2mlYYYYMMDDHH.Grb2.ndfd. The files are no longer compressed. Old file names to continue for a few more days.

2005/11/02: JIF to implement precipitation RTMA approved and forwarded to NCO/PMB.

2005/11/14: Starting at 14Z, posting of files with old names (see 2 Nov log) is discontinued.

2005/11/25: Starting at 20Z, Stage II analysis are converted to GRIB2 before copygb2 is applied to map it to the NDFD grid. Prior to this, copygb was used to map Stage II analysis to the NDFD grid, then the resulting analysis was converted to GRIB2, and due to limitations in GRIB1's grid metrics, the resulting GRIB2 grid definition did not have sufficient precision (for example, prior to 20Z, we had dx/dy=5.079km, rather than 5.079406km).

2005/12/02: starting at 18Z, generating process number for the product is 109 (for RTMA) instead of 152 (for Stage II).

11Z 2006/1/13 – 14Z 2006/2/9: Precip RTMA files were not made. The utility the job depended on since 25 Nov 2005, /nfsuser/g02/wx11sg/copygb2/copygb2 seemed to have disappeared on 13 Jan. /nwprod/util/exec/copygb2 was operationally implemented on 7 Feb and started to be used in the precip RTMA script on 9 Feb. My guess is that /nfsuser/g02/wx11sg/copygb2/copygb2 was inadvertently removed on 13 Jan in when copygb2 was being prepped for JIF (JIF submitted 17 Jan).

13Z 2006/6/28: precip RTMA became operational.

18Z 2006/7/6: discontinued parallel production of precip RTMA files.

2007/4/3: Plots for precip RTMA began to be generated operationally and placed on


NWS NDFD web site:

GRIB2 decoder:

Specifications for the NDFD grid: (Section 3)

Graphics and contact information for non-precipitation RTMA:


DiMego, G. and others, 2005: Real-time mesocale analysis (RTMA) – a first step towards an analysis of record.

Horel, J. and B. Colman, 2005: Real-time and retrospective mesoscale objective analysis. Bull. Amer. Meteor. Soc., 86, 1477-1480.

Lin, Y. and K.E. Mitchell, 2005: The NCEP Stage II/IV hourly precipitation analyses: development and applications. Preprints, AMS 19th Conference on Hydrology, San Diego, CA. Paper 1.2.

Seo, D.-J., C.R. Kondragunta, K. Howard, S.V. Vasiloff and J. Zhang, 2005: The National Mosaic and Multisensor QPE (NMQ) Project—Status and plans for a community testbed for high-resolution multisensor quantitative precipitation estimation (QPE) over the United States. Preprints, AMS 19th Conference on Hydrology, San Diego, CA. Paper 1.3.