CALIPSO Quality Statements
IIR Level 2 Track Data Product
Version Releases: 3.01, 3.02

Time expressed in International Atomic Time (Temps Atomique International, TAI). Units are in seconds, starting from January 1, 1993.

Time expressed in International Atomic Time (Temps Atomique International, TAI). Units are in seconds, starting from January 1, 1993.

Scene classification and description are given in Type_of_scene.

For each spectral channel k, centered on the wavelength λ_k, the track effective emissivity, ε_{eff, k} of the selected upper level, located at the equivalent altitude Z_c, is defined according to the ATBD as

• R_k is the IIR Level 1 calibrated radiance measured in channel k,
• R_k,BG or background radiance is the outgoing top of the atmosphere background radiance which would be observed in the absence of the upper level. It is either derived from measurements in suitable neighboring pixels (distance < 100 km) or from calculations using the fast radiative transfer model FASRAD (Dubuisson et al, 2005) if no measured reference can be found (see High_Cloud_vs_Background). This model is not accounting for cloud scattering. For this release (V3), the R_k,BG values are provided in Reference_Brightness_Temperature after conversion to the equivalent brightness temperatures.
• B_k(T,Z_c) is the radiance of a black-body source located at the equivalent altitude Z_c defined from lidar observations, corresponding to a temperature T(Z_c) retrieved from ancillary meteorological data (GMAO GEOS 5). It is computed using the FASRAD model (Dubuisson et al., 2005). For this release (V3), the B_k(T,Z_c) values are provided in Blackbody_Brightness_Temperature after conversion to the equivalent brightness temperatures.
• Effective_Emissivity is set to invalid if found outside the physical [0. , 1.] range. Further analysis can be done by inter-comparing the observed, reference and blackbody brightness temperatures used for the retrievals which are provided even if the retrieved effective emissivity is found invalid. For instance, when the upper level is very low, or very close to the background reference layer, the retrievals are more difficult, and non-physical or invalid results may be found. This also corresponds to larger values of the uncertainties when physical values are obtained.

According to the ATBD, the effective emissivity uncertainty, Δε_{eff, k} for each spectral channel k, centered on the wavelength λ_k is composed of 3 terms which can be written as

Δε1_{eff, k} = ΔR_k x 1/ (R_k,BG - B_k(T,Z_c)) due to the uncertainty on the radiance measurement,

Δε2_{eff, k} = (1 -ε_eff,k ) x ΔR_k,BG x 1/ [ R_k,BG - B_k(T,Z_c) ] due to the uncertainty on the background reference,

Δε3_{eff, k}= ε_eff,k x Δ B_k(T,Z_c) x 1/ [ R_k,BG - B_k(T,Z_c) ] due to the uncertainty on the equivalent black-body source radiance.

For each channel, the Effective_Emissivity_Uncertainty, Δε_{eff, k} is the overall uncertainty derived from the 3 independent terms listed above considered as random errors so that:

Δε_{eff, k}= [R_k,BG - B_k(T,Z_c)]^-1 x [(ΔRk)² + (1 -ε_eff,k )² x (ΔRk,BG)² + ε_eff,k ²x (ΔB_k(T,Z_c))² ] ^1/2

It is inversely proportional to the radiative difference between the background reference and the upper level of thermodynamic temperature T(Z_c).

The effective emissivity uncertainty is set to invalid if found outside the [0. , 1.] range (V3). If the effective emissivity is invalid, the corresponding uncertainty is invalid too.

Reported emissivity uncertainties are less than 0.03 for most of the high (>7 km) semi-transparent clouds. They are computed assuming a 1 K equivalent error in measured and calculated radiances, which is a realistic/conservative value for the IIR measurements.

The background reference is preferably measured in neighboring pixels (distance < 100km) and its representativeness evaluated through its mean distance from the current pixel (cf High_Cloud_vs_Background_Flag which gives also the type of reference used). Otherwise, the background reference is computed using the FASRAD model.

When a measured background reference is used, Computed_vs_Observed_Background_Flag gives the mean relative difference between this observation and the computed radiance derived from the FASRAD model. The standard deviation associated to this parameter is given in Regional_Background_Std_Dev.

Statistical analyses show that the brightness temperature differences between clear sky observations and computations (available in Computed_Brightness_Temperature_Surface in V3) are most of the time smaller than 1K over ocean for each IIR channel even though GMAO moisture and temperature profiles are expected to be less accurate than over land due to fewer sounding stations available for the analysis. Over land, significant differences are observed (up to several K) due to errors in surface emissivities and temperatures. Large differences (up to 10 K) are observed in a limited number of cases.

Statistical analyses show that brightness temperature differences between low opaque cloud observations and computations (available in Blackbody_Brightness_Temperature in V3) are typically within 3 K for each IIR channel. Larger differences can be observed on sporadic points.

The uncertainty on the black-body radiance is driven by the uncertainty on the equivalent radiative altitude and the corresponding temperature derived from GMAO. In case of thick features or vertically stretched multi-layer upper levels (cf Multi_Layer_Cloud_Flag added in V3), the equivalent black-body radiance is not as accurate as for thin mono-layer systems, due to the expected error on the equivalent altitude chosen to retrieve the thermodynamic temperature.

Look-up tables allow deriving cirrus ice crystals effective size and shape. These tables are built using the FASDOM radiative transfer model (Dubuisson et al, 2008). This model calculations are taking into account cloud scattering, theoretical optical properties of several complex crystals (Yang et al, 2005), and various atmospheric and surface parameters.

This parameter is the optimal shape leading to the best agreement between the effective particle diameters D(_12.05,8.65) and D(_{12.05, 10.60}) derived from each microphysical index. If the microphysical indices are not within the range of values expected from the look-up tables, Particle_Shape_Index cannot be retrieved and is set to invalid.

The full set of effective diameters obtained for each shape is provided in Microphysics.

However, numerous other conditions are encountered where the microphysical indices are not within the range of values expected from the look-up tables. It can be due to the absence of the adapted look-up table (for instance in case of aerosols or liquid clouds) or to a wrong value of at least one microphysical index. If only one microphysical index is within the expected range, or if the microphysical indices slightly deviate from the closest possible value (with a 15% tolerance), the algorithm attempts to provide an Effective_Particle_Size estimate. These degraded configurations are flagged in Effective_Particle_Size_Uncertainty. Otherwise, Effective_Particle_Size is set to invalid.

It is very important that the users refer to Effective_Particle_Size_Uncertainty to find out if the Effective_Particle_Size has been retrieved in a nominal or a degraded configuration.

Effective_Particle_Size_Uncertainty values of 100 or more are used to flag specific degraded configurations when one (or both) microphysical indices is (are) outside the value range expected from the look-up tables (see Effective_Particle_Size). For these a priori medium to very low confidence cases, the algorithm cannot provide the Particle_Shape_Index but still attempts to provide some piece of information about the size.

The background reference radiance is preferably measured in neighboring pixels. Criteria for deciding to use a measured reference are:

• the distance between the measured background radiance and the upper level must be smaller than 100 km and both must be of the same IGBP type (new condition added in V3);
• if the reference is an opaque layer, the permitted altitude difference is +/- 100 m corresponding to +/- 1 K radiative difference (worst case).

If the background radiance cannot be derived from measurements, it is computed using the FASRAD model:

• assuming clear sky if the reference is clear sky or a low altitude (<7 km) non -depolarizing aerosols layer,
• and assuming a blackbody located at the altitude "Centroid_IAB_0532_Lower_Level" for opaque layers.

FASRAD uses temperature, water vapor and ozone profiles from the GMAO GEOS 5 model. For clear sky simulations, it uses also surface emissivities derived from the IGBP geotype (see Surface_emissivities) and GMAO GEOS 5 surface temperatures.

The conditions selected by the algorithm to compute the background radiance are given in High_Cloud_vs_Background. When a measured reference is used, Computed_vs_Observed_Background_Flag gives the mean relative difference between this observation and the computed radiance derived from the FASRAD model. The standard deviation associated to this parameter is given in Regional_Background_Std_Dev.

Optical_Depth_12_05 = - ln (1 - Effective_Emissivity_12_05)

Optical_Depth_12_05 is set to invalid if found outside the [0. , 10.] range (V3) and if Effective_Emissivity_12_05 is invalid.

Optical_Depth_12_05_Uncertainty is set to invalid if found outside the [0. , 10.] range (V3) and if Optical_Depth_12_05 is invalid.

IWP (g.m-2) = 0.307 x Effective_Particle_Size (μm) x (2 x Optical_Depth_12_05).

This parameter is set to invalid if no feature is selected (clear sky) or if the type of scene is undetermined.

This parameter is set to invalid if no feature has been detected (clear sky) or if the type of scene is undetermined.

This parameter is set to invalid if no feature has been detected (clear sky) or if the type of scene is undetermined.

This parameter is set to invalid if no feature has been detected (clear sky) or if the type of scene is undetermined.

For single-layer systems (cf Multi_Layer_Cloud_Flag), this parameter is a replicate of the centroid altitude provided in Attenuated_Backscatter_Statistics_532 in CALIOP lidar Level 2 Cloud or Aerosols layers product.

In case of a multi-layer scenes (N layers), this parameter is the mean of the centroid altitudes z(i) of each layer, i, weighted with the mean attenuated backscatter beta_mean (i) in each layer:

Centroid_IAB_0532_Upper_Level = som[1,N : z(i).beta_mean(i)]/N.som[1,N : beta_mean(i)]

This parameter is set to invalid if no feature has been detected (clear sky) or if the type of scene is undetermined.

This parameter is set to invalid only if no feature has been detected (clear sky) or if the type of scene is undetermined.

This parameter is set to invalid if no feature has been detected (clear sky) or if the type of scene is undetermined.

It is a replicate of the centroid altitude provided in Attenuated_Backscatter_Statistics_532 in CALIOP lidar Level 2 Cloud or Aerosols layers products.

It is derived from the Centroid_IAB_0532_Lower_Level altitude and GMAO temperature profile (V3, was a replicate of the CALIOP parameter "Mid_Layer_Temperature" in V2).

They are derived from IGBP surface type and NSIDC snow/ice indices (1/6° resolution, same as in CALIOP products). For NSDIC indices between 10 and 103, geotype index takes the IGBP snow/ice index (15).

Surface emissivities are computed accounting for the IIR spectral response functions (D. P. Kratz, NASA Langley, CERES team, see also Wilbert et al, 1999). The following values are used:

If not zero, corresponding to nominal quality:

• either one channel has poor quality or is missing, or
• the radiances in the 3 channels are not all part of the same image measurement sequences (information added in V3) which, for scenes with high broken clouds, could lead to some errors at the edge of the images for geometrical reasons.

Only layers identified with a 5 or 20-km horizontal resolution are used in the analysis. Those obtained at a horizontal averaging of 80km are systematically rejected as they are not expected to impact the thermal IR signals.

The scenes are first organized according to the background reference scene (4^th column in the table below), which can be either the surface (scenes identified as clear sky or possibly containing low semi-transparent depolarizing aerosol layers) or an opaque layer.

For each category, one to several semi-transparent (ST) layers can be considered as the upper level to compute the effective emissivity (3^rd column in the table). Layers are high when their centroid altitude is above 7 km, and are low otherwise.

Low altitude aerosols layers are classified according to their mean volume depolarization ratio, with a threshold of 6%. Type 53 contains the depolarizing features (>6%), typically corresponding to desert dust aerosols (Liu et al, 2008). Type 52 are the non-depolarizing features. The threshold was set to 7% in V2.

Low level (Type 20 and 70) and high level (Type 40 and 80) opaque clouds are classified in V3 according to the maximum volume depolarization ratio in the layer, with a threshold of 40% (there was no distinction in V2).

A low altitude ST cloud layer (Type 24) is re-classified to Type 59 if the maximum attenuated backscatter and the maximum volume depolarization ratio in the layer are smaller than 0.02 sr^-1 and 7% respectively, as a possible indicator of the presence of aerosols (V3).

Besides the changes described above, the classification has been updated with the addition of some complex types of scenes. Several types involving high ST aerosols layers have been added (Types 64, 65, 66) to better account for stratospheric clouds, classified as "aerosols" in the CALIOP product. The remaining scenes which do not match the classification are reported as #99.

In V2, scenes composed of 1, 2 or 3 high ST clouds (types 21, 22, 26) were re-classified as type 40 (opaque cloud) when the retrieved effective emissivity was greater than 0.6. It is not the case anymore for this release (V3) where the classification relies only on the CALIOP product.

The types of scenes are listed in the table below. Ice cirrus clouds fall in the scenes containing 1 to 5 high semi-transparent cloud layers overlying either the surface or a dense opaque layer, or in the scenes containing 1 high opaque cloud (Types 40, 80, 21, 22, 26, 31, 32, 41, 42, 30, 37). Overall, the changes with respect to version V2 are significant, due to changes in the IIR classification as described above, corrections of bugs and also due to the changes in the Version 3 CALIOP Level 2 layer products (for instance cloud/aerosols discrimination, opacity flags). Comparing IIR Level 2 V2 and V3 classifications is therefore not straightforward.

• the units digit indicates if the studied pixel is isolated or part of a structure with consecutive IIR pixels of same "Type_of_scene" (same as in V2).
• the tens digit is a mineral aerosols index based on IIR inter-channels brightness temperature differences (BTD). Mineral aerosols layers are identified (tens digit=1) when the 08_65 minus 12_05 BTD is < -2K and the 10_60 minus 12_05 BTD is < -0.5 K (added in V3).
• the hundreds digit is an index describing the difference between observed and computed brightness temperatures for specific types of scenes: scenes identified as clear sky (10) or containing low aerosols (52, 53) and scenes containing opaque clouds (20, 40). This index is designed to identify the pixels exhibiting large differences and which may require further analysis (added in V3).

If the background radiance is derived from reference measurements in the vicinity of the pixel, the unit digit gives an indication of the mean distance from the current pixel. If it is derived from the FASRAD model, the unit digit is set to zero.

Depending on Type_of_Scene, the reference can be clear sky (10) or possibly low ST non depolarizing aerosols (52), a low opaque cloud (20), a high opaque cloud (40), or a low opaque aerosols layer (56). This information is provided in the hundreds digit (added in V3). When the reference is a cloud or aerosol layer selected among nearby observations (i.e. not computed), the tens digit (added in V3) indicates the range of values of its effective emissivity. Otherwise, it is set to 0 (computed reference) or -9 (clear sky).

The model leading to the best agreement between both microphysical indices is the one selected by the algorithm (see Particle_Shape_Index) to retrieve Effective_Particle_Size.

Microphysics gives the whole set of effective sizes retrieved from each the pair of channels for each model considered in the algorithm, allowing the user to evaluate the robustness of the model selection. The first three elements are for aggregates, plates and solid column models respectively (Yang et al, 2005).