Approach

In the figures above (scroll through the caroussel images), we explain the logic of our approach and provide the formalism necessary for understanding the concept. The general idea is based on the fact that two space-borne lidars on similar polar orbits should retrieve similar cloud distributions if the differences linked with wavelength, sampling frequency, averaging time, observation geometry, detection type, and observation type are understood and properly compensated for.

The first big part of work is dedicated to a search of cloud detection threshold applied to ALADIN vertical profiles at given horizontal and vertical resolutions, which would make the ALADIN clouds consistent with the CALIOP clouds. In this pair of lidars, we chose CALIOP to be the reference one and ALADIN to be the one, which will be adjusted. This choice is not based on the age of lidar or its publications history.

From the equations above, it follows that the wavelength conversion is simpler for a lidar, for which the particulate and molecular backscatter and extinction components are separated, and this is the case of ALADIN thanks to its HSRL capability (that CALIPSO has not). We use the following iteration scheme: we take the overlapping period of ALADIN and CALIOP observations, collocate the observations, convert ALADIN L2 data to SR(z) profiles estimated at 532 nm, and compare the cloud fraction profiles for collocated dataset.

Based on the results of this comparison, we adjust the SR threshold, the profiles vertical and horizontal resolution and/or define the altitudinal or geographical limits on the conversion and comparison. We repeat these iterations until the difference between CALIOP and ALADIN clouds converges to a global minimum. A set of SR threshold(s) and rules will be the output of this iterative process. Then we will use it in the production of ALADIN clouds.

The decision box “Trends real?” in the second part of the flowchart above will check the consistency of trends and tendencies estimated from CALIOP, ALADIN, and from joint CALIOP+ALADIN datasets. If some abnormal behavior is detected (e.g. the transition change is larger than r.m.s. of CALIOP clouds before the transition or the slope of a joint product is beyond the slope variability limits), then the diurnal cycle correction will be updated and/or the SR threshold adjustment will be repeated.

These iterations will be continued until the convergence is reached and the cloud trends determined for a joint product become consistent with those determined for CALIOP alone.