Step | Proc | Process Description | i) Basic Principles | ii) General Options | iii) Specific Class: IR Vis MW Inter-Satellite/Inter-Sensor Inter-Satellite/Intra-Sensor GEO-LEO LEO-LEO GEO-GEO | iv) Specific Instruments: SEVIRI-IASI GOES-AIRS |
1 | 1a | Select Orbit | A first rough-cut to:
- Reduce data volume
- Include only relevant portions (channels, area, time, viewing geometry)
|
- v0.1
- Select data on per-orbit or per-image basis
- Need to know how often to do inter-calibration – based on observed rate of change
(Defined iteratively with 2c & 2d.) |
- v0.1
- Define GEO Region of Interest: within 60° of GEO SSP
- Subset GEO data to RoI
- Select LEO data within GEO RoI for each inter-cal period
- Subset LEO data to GEO RoI
|
- v0.1
- GEO RoI = ±30° lat/lon of SSP
- Take 1 Metop overpass with night-time equator crossing closest to GEO SSP
- Subset IASI data to GEO RoI
- Select SEVIRI image closest in time to LEO Equator crossing
- v0.2
- Select fixed GEO frame at nominal LEO local equator crossing time (21:30)
- Extend RoI to ±35°
|
1b | Collocate Pixels | Define which pixels to compare:
- Define FoV of all pixels
- and environment around pixels
- Identify pixels for both instruments within these areas meeting collocation criteria for time, space and geometry
|
- v0.1
- Search for all pixels within FoV and environment
- v0.3
- Grid observations using 2D-histogram in lat/lon space
|
- v0.1
- Geometric alignment: Select GEO/LEO pixels where secant of zenith angle is within 0.01
- Temporal alignment: Select GEO/LEO pixels with time differences <300s
|
- v0.1
- IASI FoV=12km at nadir
- SEVIRI FoV=3km at SSP
- Time difference <900s
- Select 5x5 SEVIRI pixels closest to centre of IASI FoV
- v0.3 - IR1b4eeGLSIv0.3
- Select SEVIRI and IASI pixels in same bin of 2D histrogram with 0.125° lat/lon grid
|
1c | Pre-select Channels |
- Select only broadly comparable channels from both instruments (to reduce data volume)
|
- v0.1
- Selection based on pre-determined criteria for each instrument pair
|
- v0.1
- Select IR channels (3-15µm)
|
- v0.1
- Select IR channels of SEVIRI
- Select all channels for IASI
|
2 | 2a | Collect Radiances | Convert observations from both instruments to a common definition of radiance to allow direct comparison. |
- v0.1
- Convert instrument Level 1.5/1b/1c data to radiances, accounting for channel Spectral Response Functions
|
- v0.1
- Perform comparison in radiance units: mW/m2/st/cm-1
|
- v0.1
- Account for Meteosat radiance definition applicable to level 1.5 dataset
|
2b | Spectral Matching | Basic Principles:
- Identify which channel sets provide sufficient common information to allow meaningful inter-calibration.
- Transform these into comparable channels
- Account for deficiencies in channel matches
|
- v0.1
- Define SRFs for all channels
- Co-average comparable channels
- Use Radiative Transfer Model to account for differences
- Estimate uncertainty due to spectral mismatches
|
- v0.1
- Transform spectral response functions to common grid
- Spectral Convolution to synthesise GEO channels
- Account for spectral sampling and stability in error budget
|
- v0.1
- Assume IASI channels are spectrally stable and contiguously sampled
- Use published SRFs for MSG at 95K, interpolated to IASI grid.
- Estimate radiance missing from IASI’s coverage of SEVIRI IR3.9 channel by assuming a uniform brightness temperature
|
2c | Spatial Matching | Basic Principles:
- Transform observations from each instrument to comparable spatial scales
- Estimate uncertainty due to spatial variability
|
- v0.1
- Identify Point Spread Functions of each instrument
- Specify the target area and identify the pixels within it
- Specify the ‘environment’ around target area
- Average pixel radiances within specified target areas and Calculate their variance
|
- v0.1
- Define target area as LEO FoV
- Average GEO pixels within target area and calc variance
- Define environment as GEO pixels within 3x radius of target area
|
- v0.1
- Assume IASI FoV circular near nadir with diameter of 12km
- Assume SEVIRI pixels are contiguous, independent samples: 3km spacing @SSP
- Calculate mean and variance of radiance in 5x5 SEVIRI pixels closest to centre of IASI FoVs
- v0.2 as v0.1, except:
- Select SEVIRI and IASI pixels in same bin of 2D histogram with 0.125° lat/lon grid
|
2d | Temporal Matching | Basic Principles:
- Establish timing difference between instruments’ observations
- Assign uncertainty based on (expected or observed) variability over this timescale.
|
- v0.1
- Identify each instruments’ sample timings
|
- v0.1
- Select GEO image closest to time of LEO Equator crossing
- Calculate time difference for each target
- v0.2
|
- v0.1
- Select only targets with time difference <900s
|
3 | 3a | Uniformity Test | Basic Principles:
- Only compare observations in homogenous scenes to reduce uncertainty in comparison due to spatial/temporal mismatches
|
- v0.1
- Compare spatial/temporal variability of scene within target area to pre-defined threshold and exclude scenes with greater variance from analysis
- Performed on a per-channel basis
|
- v0.1
- Calculate variance of GEO radiances with each LEO FoV
- v0.2
- Include interpolation between sequential GEO images
|
- v0.0
- Not implemented as found to not change results significantly. (Results rely instead on inhomogeneous scenes having lower weighting in regression and include the full range of scene radiances.)
- v0.2
- Reject any targets with scene variance >5% of reference radiance
|
3b | Outlier Rejection | Basic Principles:
- To prevent anomalous observations having undue influence on results
- Identify and reject 'outliers' on a statistical basis
|
- v0.1
- Compare the radiances in the target area with those in the surrounding environment
- Reject targets which are significantly different from the environment (3σ)
|
- v0.1
- Compare difference between mean GEO radiances within LEO FoV and 'environment'
- Reject targets where this difference is >3 times the variance of the environment's radiances
|
|
3c | Auxiliary Datasets | Basic Principles:
- To allow analysis of statistics in terms of other geophysical variables - e.g. land/sea/ice, cloud cover
|
|
|
|
4 | 4a | Regression | Systematically compare collocated radiances from 2 instruments. (This comparison may also be done in counts or brightness temperature.) This allows:
- Investigating how biases depend on various geophysical variables
- Providing statistics of any significant dependences
- Investigating the cause of these dependences
|
- v0.1
- Simple averaging of differences between collocated radiances
- v0.2
- Weighted linear regression of collocated radiances, using estimated uncertainty on each point as a weighting
- v0.3
- Perform stepwise multiple linear regression to investigate dependence of various geophysical variables
|
- v0.2
- Repeat inter-calibration daily
- Use only night-time LEO overpasses
- Include only incidence angles <30°
- Weight collocations in regression by the inverse variance of target radiances
This allows the investigation of the sensitivity of the differences to Latitude, Longitude, Incidence angle/LEO scan angle, Time of day |
- v0.1
- Select only pixels with incidence angle ~15°±1°
- Repeat inter-calibration every 10 days (nights)
- v0.2
- Extend range of incidence angles to <40°
- Inter-calibrations may be averaged over periods of ~1 week. (Longer periods are subject to drift due to ice contamination.)
- Reset statistics following Meteosat decontaminations.
|
4b | Define reference radiances | Basic Principles:
- Provide standard scene radiances at which instruments' inter-calibration bias can be directly compared
- Because biases can be scene-dependent, it is necessary to define channel-specific reference scene radiances
- More than one reference scene radiance may be needed for different applications - e.g. clear/cloudy, day/night
|
- v0.1
- Select representative Region of Interest (RoI)
- Construct histogram of observed radiances within ROI
- Identify peaks of histogram corresponding to clear/cloudy scenes to define reference scene radiances
- These are determined a priori from representative sets of observations
|
- v0.1
- Limit target area to within 30° of GEO sub-satellite point
- Limit target times to night-time LEO overpasses
|
- v0.1
- Find mode of histogram of each channels' brightness temperature for collocated pixels in 5 K wide bins from 200 to 300 K
- For bimodal distributions, the mean of the modes is used
- Define low reference radiance scene for high cloud of 200 K for all channels
|
4c | Calculate biases | Basic Principles:
- Perform direct comparison of inter-calibration biases for representative scenes in a way easily understood by users
|
- v0.1
- Apply regression coefficients to estimate expected bias and uncertainty for reference scenes in radiances
- Account for correlation between regression coefficients, when calculating uncertainty on the fitted radiances
- Results may be expressed in absolute or percentage bias in radiance, or brightness temperature differences
|
- v0.1
- Convert biases (and uncertainties) from radiances to brightness temperatures
|
- v0.1
- Use effective radiances definition to covert to brightness temperature
|
4d | Test non-linearity | Basic Principles:
- Characterise any non-linearity in the relative differences between instruments, or place limits on their maximum magnitude
- May be used to account for detector non-linearity, calibration errors or inaccurate spectral response functions
|
- v0.1
- Compare results of linear and quadratic regression of collocated radiances from different instruments
- Estimate maximum departure from linearity, the scene radiance at which it occurs and uncertainty associated with it
|
- v0.1
- Combine multiple LEO overpasses need to produce enough data to identify relative instrument linearity to the level of the instruments' noise. (Any non-linearity is likely to be relatively constant in time.)
|
|
4e | Recalculate calibration coefficients | Basic Principles:
- Produce revised set of calibration coefficients for one instrument following its inter-calibration against a reference instrument
- Allow users to recalibrate data from the target instrument to be consistent with the reference instrument
- Generate uncertainties with the calibration coefficients to allow users to specify the error bars on recalibrated data
|
- v0.1
- Read original calibration coefficients and calculate the changes required to reproduce observed relative biases
- v0.2
- Read original counts observed by the target instrument and fit these to the collocated radiances observed by the reference instrument
|
|
|
4f | Report Results | Quantify the magnitude of relative biases by inter-calibration. This allows users to:
- Monitor changes in instrument calibration in time,
- Recalibrate observations,
- Specify the uncertainty on observations,
- Derive relative biases and uncertainties between various different instruments.
|
- v0.1
- Produce plots and tables of relative biases and uncertainties for reference scene radiances.
- Show evolution of these in time and dependence on geophysical variables.
- Produce tables of recalibration coefficients for near-real-time and archive data.
|
- v0.1
- Plot relative brightness temperature bias for clear sky reference scenes as time series with uncertainties
- Calculate trend line in above time series (with uncertainties)
- Calculate monthly mean bias from time series
|
- v0.1
- Reset trends and statistics when decontamination procedures performed on MSG
|
5 | 5a | Operational Corrections | Basic Principles:
|
|
|
|
5 | 5b | Process Description | Basic Principles:
|
|
|
|