NCEP/Climate Prediction Center ATLAS No. 2

Intercomparison of the NCEP/NCAR and the NASA/DAO Reanalyses (1985-1993)

3. Comparison Datasets

In addition to comparisons to the DAO reanalysis, comparisons are made with various in situ and satellite derived datasets. A brief description of these datasets is given below.

CDDB (Climate Diagnostics Data Base): The Climate Diagnostics Data Base (CDDB) contains monthly means and second moment statistics of basic upper-air fields (winds, temperature, specific humidity, geopotential height, vertical velocity) at up to 9 vertical pressure levels (1000-, 850-, 700-, 500-, 300-, 250-, 200-, 100-, and 50 hPa). Due to inconsistencies in the CDDB moisture analysis, comparisons are only made to the wind and geopotential height fields from the CDDB.

The data source for the CDDB is the NCEP Global Data Assimilation System (GDAS; Dey and Morone 1985) which contains all assimilated rawinsonde, aircraft, pibal, etc. information that is received within approximately 8 hours prior to the 0000 and 1200 UTC assimilation cycles. The data base began in October 1978 when the NCEP (formerly NMC) model became global.

The CDDB was the standard analysis at CPC for a number of years; thus, we compare this old "standard" to the reanalysis.

NOAA OLR: The outgoing longwave radiation (OLR) data are from the NOAA polar orbiting satellites (Gruber and Krueger 1984). The data were gridded at a resolution of 2.5° latitude by 2.5° longitude. Hurrel and Campbell (1992) and Chelliah and Arkin (1992) discuss the impact of satellite changes and data processing techniques on the quality of the data.

Special Sensor Microwave Imager (SSM/I) Total Precipitable Water: The Special Sensor Microwave Imager (SSM/I) total precipitable water is that generated by Wentz (1992). These data are available starting June 1987. The radiative transfer algorithm uses three channels of microwave measurement (22V, 37V, 37H) and a model which accounts for absorption and emission in the atmosphere. The model uses a surface emissivity value over oceans appropriate for a wind-roughened sea surface. The model is generally less accurate for high rainrates. No calculation is done over land or sea ice because of the complexity of the surface emissivity.

Comprehensive Ocean-Atmosphere Data Set (COADS): The comprehensive Ocean-Atmosphere Data Set (COADS) (a joint project of NOAA/ERL, NCAR and NCDC) as processed by DaSilva (1995) was used to obtain estimates of the surface energy budget. The observations were corrected for systematic biases, and the derived quantities were interpolated to a 1 ° x 1 ° grid. Data quality control included the conversion of all wind measurements to an improved Beaufort Equivalent Scale, and the rejection of nighttime cloud observations if the sky is too dark. The surface fluxes were calculated using a similarity theory parameterization that calculates transfer coefficients according to Large and Pond (1981). Surface radiation estimates were made using empirical formulae to approximate cloud attenuation, surface albedo and atmospheric transmissivity. The objective analysis scheme is similar to the one by Levitus (1982), which is a successive-correction scheme (Cressman 1959; Daley 1991) with a weight function developed by Barnes (1964).

World Monthly Surface Station Climatology Data: We compare to global monthly mean precipitation and surface air temperature obtained from the world monthly surface station climatology data received from W. Spangler and R. Jenne at NCAR. Most of the data were obtained from the National Climatic Data Center (NCDC), Asheville, North Carolina. A detailed description of the dataset and data sources is found in Spangler and Jenne (1990). The number of station reports varies from year to year, but is on the order of 2500 in more recent years. The station values were interpolated to a 2° latitude by 2.5° longitude grid by averaging station values within a 300 km radius of each grid point. The value at a grid point was set to undefined if there were no stations within a 300 km radius. Details of the processing are found in Schemm et al. (1992); in this atlas comparisons are made for the period 1985-1993.

Xie/Arkin Precipitation Data: Monthly mean reanalysis precipitation is also compared to the merged precipitation analysis of Xie and Arkin (1996a) and to OLR-based precipitation estimates of Xie and Arkin (1996b). Both the merged analysis and the OLR-based precipitation estimates are monthly, global (over both land and ocean) and gridded at a resolution of 2.5° latitude by 2.5° longitude. Input data for the merged analysis includes gauge-based monthly analyses over land, atoll gauge data of Morrissey and Greene (1991), IR data from the GOES Precipitation Index (GPI), microwave scattering estimates of Grody and microwave emission estimates of Chang. In the algorithm, the three satellite estimates and model predictions (based on 12-36 hour ECMWF forecasts) are combined using the maximum likelihood estimation method. The result is combined with the gauge data using the method of Reynolds (1988). Input data for the OLR-based estimates includes OLR flux data from the NOAA operational polar-orbiting satellites and the mean annual cycles of global precipitation defined from the merged analysis of Xie and Arkin (1996a) for an 8-year period from July 1987-June 1995. In the algorithm, precipitation anomalies for each month are estimated from OLR anomalies using a coefficient which is a linear function of the mean annual cycle of precipitation. The total precipitation is then defined as the summation of the anomaly plus the mean annual cycle from the 8-year merged analysis of Xie and Arkin (1996a).

U.S. Gridded Hourly Precipitation: For comparisons of precipitation over the U.S. we employ a set of hourly, gridded precipitation analyses (Higgins et al. 1996a) that were developed on data obtained from the NWS/Techniques Development Laboratory, who compiled and quality-controlled station data archived at the NOAA National Climatic Data Center. The dataset covers the period 1 January 1963 through 31 December 1993. To make a useful and manageable set of analyses from 31 years of hourly observations for approximately 2500 stations, the hourly analyses were gridded into 2° latitude by 2.5° longitude boxes using a Cressman (1959) scheme with modifications (Glahn et al. 1985; Charba et al. 1992).

Microwave Sounding Unit (MSU) channel 2 brightness temperature measurements: MSU channel 2, (referred to as MSU in the literature), receives radiation from a layer extending from the surface to the lower stratosphere. Although the MSU measurements are relatively unaffected by water vapor and clouds, they can be attenuated by precipitation size water droplets and large ice crystals (Grody 1983). Therefore Spencer and Christy (1990) developed techniques to filter out contaminated measurements. After filtering, the MSU measurements from different satellites are inter-calibrated to remove instrument bias. When averaged over a month at 2.5° spatial resolution, Spencer and Christy (1992) found the standard deviation of the difference between MSU measurements from different satellite platforms to be in good agreement with each other to within 0.1°C in the Tropics and 0.2°C at higher latitudes.

back to the table of contents