Met Office AIRS monitoring information

The Met Office receives global AIRS warmest field of view data for a subset of 324 channels from NESDIS. We also process some AIRS data that has been received using a dish located at the Met Office. The data is pre-processed to add a surface type and surface elevation and added to our Meteorological database, the MetDB.

The observations in the MetDB are passed through a 1D-Var quality control step before assimilation in 4D-Var. The 1D-Var uses the 6 hour forecast field from the Met Office's global model to provide a first guess atmospheric column for each observation. A skin temperature, cloud height and cloud top fraction are analysed in the 1D-Var and passed as fixed parameters to the 4D-Var. The monitoring plots are derived from statistics output from the 1D-Var observation processing stage.

• Observed (0)

  The black body brightness temperature that would give the same radiance measurement at the channel frequency.

• Corrected (C)

  The bias corrected Observed brightness temperatures. The bias correction currently consists of a constant offset for each channel and each scan position plus a linear function of the 850-300 hPa thickness and 200-50 hPa thickness, with different coefficients for each channel.

• Background (B)

  The brightness temperature at a given frequency computed from the six hour forecast model background profile (this model background is used as the first-guess field in the subsequent 4D variational data assimilation stage).

• Retrieved (R)

  The best estimate of the brightness temperature of the atmosphere at a specific location and frequency using both the model background and the observations. It is the result of a 1D variational minimisation carried out during the observation processing stage.

The DNA plots show Mean O-B, StdDev O-B, Mean C-B, StdDev C-B, Mean C-R, and StdDev C-R. The meaning of O, B, C, and R is given above. The means and standard deviations are calculated for all selected observations in the 6-hour assimilation cycle.

The Summary Plots show slightly more complex statistics:

• StdDev statistic: The standard deviation statistic is (StdDev(C-B)-AvgStdDev)/Estimated_StdDev_of_StdDev. StdDev(C-B) is the standard deviation of C-B calculated for all selected observation in each 6-hour assimilation cycle. AvgStdDev is a weighted average of the standard deviation of C-B over the whole period displayed. Estimated_StdDev_of_StdDev is an estimate of the standard deviation of StdDev(C-B) across assimilation cycles. It is computed from the difference in StdDev(C-B) between successive assimilation cycles. This is in order to reduce the impact of a step change in StdDev(C-B). A large value of the standard deviation statistic indicates a suspiciously large increase in StdDev(C-B).
• Bias statistic: The bias statistic plotted is ABS(Mean(C-B)/StdDev(C-B)). It is intended to show when the bias corrected innovation is large compared to the variability of the innovation. Large values may indicate that the bias correction is no longer valid or there has been some sort of disruption to the model background.

Note that for clarity, no data is shown as cyan on the main summary plot and white on the number of observations plot.

Clicking on the "Recommended Settings" buttons on the Summary Plots sets the drop down menus to recommended values for the standard deviation and bias statistic respectively.

The "Recommended associated settings" button on the DNA and Time Series plots sets the other drop down menus to suitable values based on the value of the drop down menu to the left of the button.

Note that the recommended settings are just a starting point and do not represent the only interesting or useful combinations to look at.

• "All valid obs" (Gross): All observations with a valid background calculation are included in the monitoring plots (or all observations with a valid retrieval).
• "Used only obs" (Strict): Only observations assimilated into the Met Office's numerical weather prediction scheme are used.

• Global: Observations from all surface types.
• Sea: A surface is classified as sea by the pre-processing.
• Land: A surface is classified as land by the pre-processing.
• Sea ice: A surface is classified sea ice by the pre-processing.
• Latitude bands: All surface types in a latitude band are included in the latitude band statistics.

The Day-Night/Cloud type menu offers some combination of Day, Night, or Day and Night in combination with Clear, or Clear and Cloudy. Clear observations are defined as observations where the retrieved cloud fraction is no more than 5%.

• MetDB channel number: Although AIRS has 2378 channels, we only receive and store a subset of 324 AIRS channel on our database, the MetDB. These channels are numbered 1-324.
• Instrument channel number: The AIRS instrument has 2378 channels. The frequency in inverse centimetres and wavelength in microns are also shown on the 'Channel properties' line.
• Normally all channels are subject to a gross limits check. This is skipped for channels that are, or have been, extremely noisy. Channels currently in the 'do not check' list are marked with green blocks or lines on the MetDB channel number scale. In the time series plots they are labelled 'DNC' (do not check). Note that the list of channels currently excluded from the gross limits check is rather out of date and should be reviewed.