The NCEP Climate Forecast System Reanalysis - NOAA National ...

Figure 26 shows the global total bias (air-mass part plus scan angle part) for theactive channels of the AMSUA-N15 and HIRS3-N17. In this figure, CNTL is theoperational GDAS total bias for these channels as a reference. Most of the channels (o-g)equilibrate within about one month, except for HIRS3_N17 channel 12, which took about2.5 months. Therefore, 3 months was chosen as the general training length for all thehistorical satellites.The next prerequisite to running the spin-up experiments was to assemble a timeseries of active channels and sensors for all the historical satellites, the SATINFO files,marked with periods of known outages and poor quality data. The starting point was a setof tables and scripts received, via JCSDA collaboration, from the MERRA reanalysis, seeBosilovich, 2008. The tables and scripts were updated based upon the ERA-40 qualitycontrol list and the NCEP historical satellite document maintained by co-author DennisKeyser, at:http://www.emc.ncep.noaa.gov/mmb/data_processing/Satellite_Historical_Documentation.htm.Once a working set of SATINFO files were created (and the historical set of SSUinstruments were included in the CRTM as noted in 3.1.2), a set of bias correction spinup experiments listed in Table 3 was carried out by running the CFSR over the indicatedperiods when the satellite data first became available (See Figure 21). In 2 cases, NOAA-6 and 11 operated during two periods, so the bias correction spin up had to be done twice.Once each spin up experiment was complete, the post-processing step was carried out onthe (o-g) diagnostic files to create starting SATANG files. It was then paired with theBIASCR file from the last cycle of the training period for use when the final CFSRassimilation began assimilating that particular instrument and sensor. Two features of thisbias correction scheme should be noted. One is the scan angle dependent bias is thedominant part of the total bias, and the other is the predictor coefficients of the air-massbias usually responds to the atmospheric state very quickly, usually within one or twodays.- 66 -