Background
SEVIRI is the main imager on the EUMETSAT Meteosat Second Generation
platform. As part of our participation in the international CREW
(Cloud Retrieval Evaluation Workshop) product intercomparison and
assessment effort, we have adapted portions of the operational Collection 6 MODIS
(MOD06/MYD06) cloud optical and microphysical algorithm and the
GOES-R cloud top properties algorithm to run on SEVIRI. The overall
retrieval package is referred to as CHIMAERA (Cross-platform HIgh
resolution Multi-instrument AtmosphEric Retrieval Algorithms); it was
designed for flexibility in ingesting data from a variety of
satellite and airborne imaging instruments (MODIS, VIIRS, SEVIRI,
eMAS, etc.). The CHIMAERA package utilizes a shared-core concept
where the same core code and algorithm-specific ancillary data
sources are used for all instrument retrievals. Lookup tables (LUTs)
such as cloud reflectance/emissivity and absorbing gas transmittances
for atmospheric corrections are developed on an instrument-specific
basis. Figure 1 illustrates the processing chain for both
MOD06_L2 and SEV06-CLD products.
SEVIRI-specific Algorithm Details
There are some important differences between the implementation of the MODIS and SEVIRI optical property retrieval products (optical thickness, effective particle radius, and derived water path). SEVIRI lacks the MODIS 1.2µm and 2.1µm channels, which compromises SEVIRI’s ability to retrieve clouds over snow/ice surfaces (Platnick et al., 2003). Similarly, the optional 1.6 and 2.1µm MODIS retrieval over snow/ice surfaces (Platnick et al., 2001) is not available. However SEVIRI’s spatial coverage is such that snow/ice surfaces typically cover a very small fraction of the observable area and thus are unlikely to be an issue except for users interested in wintertime northern hemisphere scenes and/or mountain regions.
The CO2 emissive band coverage on SEVIRI consists of a single broadband CO2 channel instead of the four narrow-band CO2 channels on MODIS. Therefore the MODIS CO2 slicing algorithm cannot be used in SEVIRI processing to obtain cloud top properties of high clouds. In lieu of being able to implement a full MODIS cloud-top properties algorithm, the SEVIRI algorithm utilizes a hybrid algorithm. A GOES-R Algorithm Working Group (AWG) style optimal estimation cloud-top properties retrieval, described in Heidinger and Pavolonis (2009) and Heidinger et al. (2010), is used for retrievals of low emissivity high clouds with good success (Hamann et al., 2014). For low clouds, a MODIS-heritage IR Window retrieval is used. An IR cloud thermodynamic phase algorithm is implemented using the same method utilized in MODIS Collection 6 (Baum et al., 2012).
The cloud algorithms ingest the well-established and documented SAFNWC cloud mask product developed by the Météo France Nowcasting and Weather Prediction Satellite Application Facility. This cloud mask algorithm is described in detail in Derrien and Le Gleau (2005) and Derrien and Le Gleau (2010). We also rely on the SAFNWC cloud mask to identify broken clouds/partly cloudy pixels for PCL (Partly CLoudy, see Table 3) discrimination. We do not perform a MODIS-like multilayer cloud retrieval as SEVIRI does not have the requisite spectral channels. However, unlike the MODIS cloud mask, the SAFNWC cloud mask does provide a multilayer mask. We also do not perform separate visible/near-infrared/shortwave-infrared cloud thermodynamic phase tests to supplement the IR phase algorithm; only the IR cloud thermodynamic phase algorithm (mentioned above) is used.
The impact of cloud mask and phase differences relative to MODIS can be important for ambiguous scenes (broken clouds, heavy aerosol/dust, supercooled cloud temperatures). For example, any difference in the phase decision will result in potentially strong differences in the retrieved optical thickness and effective radius due simply to the different microphysical assumptions. Regardless, the cloud optical thickness and effective radius retrieval algorithms that are implemented for both the visible/near-infrared (VNIR) -1.6µm and VNIR-3.8µm SEVIRI channel combinations are identical to MODIS Collection 6 as are the QA bit assignments (see modis-atmos.gsfc.nasa.gov/products_C006update.html).
The SEVIRI nadir resolution is 3 km and degrades away from nadir as the view angle becomes more oblique. Like MODIS, retrievals are limited to where the solar zenith angle is less than 81.36° (µ0>0.15). In consideration of the SEVIRI wide field of view, a limit of the same 81.36 degrees is also applied to the sensor zenith angle. Note that baseline retrieval uncertainties are provided in the data file (see Table 3) and can increase substantially at the extreme solar and view zenith angles. Further, the impact of SEVIRI’s coarser spatial resolution (3km vs. 1km) is expected to impact retrievals in heterogeneous cloud scenes (Zhang and Platnick, 2011; Zhang et al., 2012).
Annexe
Table 1 :
Legend of values stored in SEVIRI cloud mask product and their definitions as per Satellite Meteorology Centre of Meteo-France (SATMOS) website. www.satmos.meteo.fr/html_en/Diffusion_CT_MSG.html
| Result Value | Description |
| 0 | No retrieval |
| 1 | Clear sky, land surface |
| 2 | Clear sky, ocean surface |
| 3 | Snow / ice on land, no cloud |
| 4 | Snow / ice on ocean, no cloud |
| 5 | Cloud, very low, cumuliform |
| 6 | Cloud, very low, other |
| 7 | Cloud, low, cumuliform |
| 8 | Cloud, low, other |
| 9 | Cloud, medium, cumuliform |
| 10 | Cloud, medium, other |
| 11 | Cloud, high, cumuliform |
| 12 | Cloud, high, other |
| 13 | Cloud, very high, cumuliform |
| 14 | Cloud, very high, other |
| 15 | Cloud, semi-transparent, thin |
| 16 | Cloud, semi-transparent, meanly thick |
| 17 | Cloud, semi-transparent, thick |
| 18 | Cloud, semi-transparent, above medium cloud |
| 19 | Cloud, broken |
| 20 | Undetermined |
Table 2 :
SEVIRI channels and their MODIS equivalents
| SEVIRI channel number and central wl (µm) | SEVIRI band-pass (µm) | MODIS channel number and central wl (µm) | MODIS band-pass (µm) |
| 1: 0.635 | 0.590-0.698 | 1: 0.658 | 0.620-0.670 |
| 2: 0.810 | 0.768-0.854 | 2: 0.863 | 0.841-0.876 |
| 3: 1.640 | 1.539-1.729 | 6: 1.625 | 1.628-1.652 |
| 4: 3.920 | 3.550-4.360 | 20: 3.851 | 3.660-3.840 |
| 5: 6.250 | 5.746-6.862 | 27: 6.766 | 6.535-6.895 |
| 6: 7.350 | 7.010-7.730 | 28: 7.282 | 7.175-7.475 |
| 7: 8.700 | 8.444-8.972 | 29: 8.642 | 8.400-8.700 |
| 8: 9.660 | 9.500-9.839 | 30: 9.673 | 9.580-9.880 |
| 9: 10.800 | 10.080-11.600 | 31: 10.984 | 10.780-11.280 |
| 10: 12.000 | 11.360-12.560 | 32: 11.897 | 11.770-12.270 |
| 11: 13.400 | 12.48-14.320 | 33-36: N/A | 13.185-14.385 |
Table 3 :
SEV06-CLD SDS list and equivalent MOD06 SDSs
| SEV06-CLD SDS name | Equivalent MOD06 SDS name | Notes |
| MSG_Latitude | Latitude |
|
| MSG_Longitude | Longitude |
|
| Relative_Azimuth_Angle |
| Can be calculated from solar and sensor azimuth angles |
| Above_Cloud_Water_Vapor |
| This water vapor amount is from an integrated ancillary profile and is not a direct retrieval |
| Cloud_Optical_Thickness_16 | Cloud_Optical_Thickness_16 | Except over snow/ice surfaces where MODIS is able to use 1.2µm channel. |
| Cloud_Optical_Thickness_16_PCL | Cloud_Optical_Thickness_16_PCL | Except for different PCL definition as stated earlier |
| Cloud_Optical_Thickness_38 | Cloud_Optical_Thickness_37 | Except over snow/ice surfaces where MODIS is able to use 1.2µm channel. |
| Cloud_Optical_Thickness_38_PCL | Cloud_Optical_Thickness_37_PCL | Except for different PCL definition as stated earlier |
| Cloud_Effective_Radius_16 | Cloud_Effective_Radius_16 | See COT note |
| Cloud_Effective_Radius_16_PCL | Cloud_Effective_Radius_16_PCL | See COT PCL note |
| Cloud_Effective_Radius_38 | Cloud_Effective_Radius_37 | See COT note |
| Cloud_Effective_Radius_38_PCL | Cloud_Effective_Radius_37_PCL | See COT PCL note |
| Cloud_Water_Path_16 | Cloud_Water_Path_16 | See COT note |
| Cloud_Water_Path_16_PCL | Cloud_Water_Path_16_PCL | See COT PCL note |
| Cloud_Water_Path_38 | Cloud_Water_Path_37 | See COT note |
| Cloud_Water_Path_38_PCL | Cloud_Water_Path_37_PCL | See COT PCL note |
| Cloud_Effective_Radius_Uncertainty_16 | Cloud_Effective_Radius_Uncertainty_16 | Calibration uncertainty of flat 5% used for SEVIRI because there is no L1B uncertainty index |
| Cloud_Effective_Radius_Uncertainty_38 | Cloud_Effective_Radius_Uncertainty_38 | See CER_Unc16 note |
| Cloud_Optical_Thickness_Uncertainty_16 | Cloud_Optical_Thickness_Uncertainty_16 | See CER_Unc16 note |
| Cloud_Optical_Thickness_Uncertainty_38 | Cloud_Optical_Thickness_Uncertainty_38 | See CER_Unc16 note |
| Cloud_Water_Path_Uncertainty_16 | Cloud_Water_Path_Uncertainty_16 | See CER_Unc16 note |
| Cloud_Water_Path_Uncertainty_38 | Cloud_Water_Path_Uncertainty_38 | See CER_Unc16 note |
| Cloud_Phase_Optical_Properties | Cloud_Phase_Optical_Properties | SEVIRI CPOP SDS at this time is identical to SEVIRI Cloud_Phase_Infrared SDS |
| Single_Scatter_Albedo_Ice | Single_Scatter_Albedo_Ice |
|
| Asymmetry_Parameter_Ice | Asymmetry_Parameter_Ice |
|
| Extinction_Efficiency_Ice | Extinction_Efficiency_Ice |
|
| Single_Scatter_Albedo_Liq | Single_Scatter_Albedo_Liq |
|
| Asymmetry_Parameter_Liq | Asymmetry_Parameter_Liq |
|
| Extinction_Efficiency_Liq | Extinction_Efficiency_Liq |
|
| Failure_Metric_16 | Failure_Metric_16 |
|
| Failure_Metric_38 | Failure_Metric_37 |
|
| Quality_Assurance | Quality_Assurance_1km | Full QA bit description is present in file metadata just like MOD06 |
| Cloud_Mask | Cloud_Mask_1km | SEVIRI Cloud_Mask SDS does not require any bit decoding, values are as listed in Table 1 |
| Cloud_Top_Temperature | cloud_top_temperature_1km | SEVIRI uses the AWG algorithm, but use of data is same as for MOD06 |
| Cloud_Top_Height | cloud_top_height_1km |
|
| Surface_Temperature | surface_temperature_1km | Interpolated model surface temperature |
| Cloud_Top_Pressure | cloud_top_pressure_1km | See CTT note |
| Cloud_Top_Method | cloud_top_method_1km |
|
| Cloud_Phase_Infrared | cloud_phase_infrared_1km | SEVIRI uses an identical algorithm to MODIS, but 13.2 µm instead of 7.2 µm for absorbing IR channel. |
