Validation of Cloud Variables: 

The Path-P product contains the following information about cloud variables.

HIRS_CLDY Fraction of HIRS pixels (18 km nominal resolution)  within a 100 km retrieval box, which have been labeled as cloudy. This variable is computed using a sequence of 9 tests (c.f. Francis, 1994) employing visible, infrared and microwave channels of the TOVS instrument. 
FCLD Effective cloud fraction (Ne) This variable is supposed to  represents the  cloud fraction at "sub resolution" and therefore provides the fraction of a pixel covered by clouds. It is called an effective cloud fraction (Ne) , because the actual cloud fraction (N) and the cloud emissivity (e) cannot be separated in the retrieval. Effective cloud fraction is calculated  simultaneously with cloud top pressure by minimizing the differences between observed and calculated clear and cloudy brightness temperatures in HIRS channel 4-8. For low clouds, differences between clear and cloudy TBs in those channels are small. Therefore, a weighted Chi-Square scheme (Stubenrauch et al, 1999) is used to reduce potential biases in cloud height. In the version of 3I utilized for Path-P, effective cloud fraction is computed from the cloudy HIRS pixels. In order to obtain the effective cloud fraction for the entire retrieval box it is multiplied by the HIRS_CLDY variable.  Retrievals of FCLD greater then 100 are possible because of the noisy character of this variable. In order to avoid biases when averaging FCLD values of > 100 are retained. 
CLPRESS Cloud top pressure
CLTEMP Cloud top temperature

Cloud Fraction
(HIRS_CLDY)

 NORTH POLE DRIFTING STATIONS

Annual cycle (March 1990- March 1991) of Path-P  HIRS_CLDY and meteorological surface observations (surface observer) from the Soviet North Pole Drifting stations. Time series for both data sources are smoothed using a 5-day running mean. A much better match up of surface observations and Path-P HIRS_CLDY exists during summer. During winter HIRS_CLDY is substantially larger than surface reports. 

Cloud Fraction
HIRS_CLDY
Considerable uncertainty exists with respect to the ability of a surface observer to accurately observe cloud amount during winter. Data from the  SHEBA experiment offers the opportunity for the intercomparison of subjective and objective surface measurements  with satellite  retrievals:
Time series auf cloud fraction for the period of the SHEBA experiment. Cloud fraction (HIRS_CLDY) from Path-P, subjective surface observations and a cloud fraction computed from the DABUL Lidar. The Lidar data se used here provides clear/cloudy binary information. Cloud fraction was computed from this information using a 3 -day running mean. Path-P and surface observations are also shown as 3-day running averages. Data from all sources correlate well at this timescale (3-day smoothing). Note that winter time observations of cloud fraction from both the LIDAR and Path-P are consistently higher than those observed from the surface. As shown in below scatterplot, for small cloud fractions (predominantly winter) the LIDAR reports about 20 % more higher cloud fraction that the surface observer. Since the LIDAR reports both ice and water particles as cloud a considerable amount of clouds shown reported by the LIDAR may be ice particles. 

Effective Cloud Fraction 
FCLD
Time series  of cloud fraction of meteorological observations (ShipObs), cloud fraction computed from the cloud lidar (Lidar) and cloud fraction (HIRS_CLDY) and effective cloud fraction (FCLD) from Path-P. There is  that there is a relatively close relationship of HIRS_CLDY and FCLD ( note that FCLD is computed from HIRS_CLDY). FCLD appears noisier and values during the summer frequently exceed 100 %. Below figures show  the same data with FCLD limited to 100%.

FCLD =< 100
Time series  of cloud fraction of meteorological observations (ShipObs), cloud fraction computed from the cloud lidar (Lidar) and cloud fraction (HIRS_CLDY) and effective cloud fraction (FCLD) from Path-P. The FCLD variable was constrained to not exceed 100%. The match with surface observations appears better. Particularly in the February through April period but a negative bias with respect to surface observations is noticeable. 

Cloud Top Temperature
CLTEMP
Mean monthly cloud top temperature (CLTEMP) from Path-P, AVHRR (CASPR algorithm) and from a combination of the ETL Cloud Radar/Lidar surface-based systems. Cloud top height from the Radar/Lidar source is converted to cloud top temperature using SHEBA radiosonde measurements. There is a good agreement between the two satellite measurements. Though using different sensors and algorithms satellite retrievals are based on the inversion of emitted radiation. The surface retrievals from the Radar/Lidar systems are based on the interpretation of backscattered radiation. A large difference (over 10K ) between these two measurements is notable. The likely explanation for this large difference the presence of optically very thin high clouds over a deck of lower clouds. These optically very thin clouds are relatively transparent to the emitted radiation used in the satellite retrievals yielding much lower "effective" cloud top temperatures. The active Radar/Lidar systems on the other hand retrieve a physical boundary of the upper cloud which will be much higher. Interpretation of the  backscatter also affects the highest reported level. Of course the potential mislableling of clear vs. cloudy scenes also has an effect on the retrieval of cloud top temperature. The differences in the above retrieved cloud top temperatures point to the need of careful interpretation and application of these measurements. More detailed comparisons of these measurements, making use of SHEBA surface- and aircraft-based information of cloud microphysics of the entire atmospheric column should provide insights on how to best use this information.