CLD: The MPLNET Cloud Product
The V3 CLD product draws on the rich heritage of the V2 cloud product, which has been used in a number of scientific studies. New variables have been added in V3 to estimate the cloud optical depth (COD) and extinction for thin cirrus clouds, thermodynamic cloud phase, and cloud fraction. A multitemporal averaging scheme has been incorporated to improve high-altitude cloud detection under conditions of a weak signal-to-noise ratio by combining NRB signal profiles averaged to base (1-min), medium (5-min), and long (21-min) temporal resolutions into a merged cloud scene. See our publications page for a full list of how these V3 additions are currently being used for studies of cloud presence, thin cirrus contamination in aerosol retrievals, and cirrus cloud radiative forcing.
The algorithm used for cloud layer, COD and extinction retrievals in V3 is described by Lewis et al (2016). Cloud layer height retrievals are perfomed using two methods. The first relies on gradients in the lidar backscatter profile (gradient-based cloud detection method; GCDM) and is primarily used for low-level liquid water phase clouds. The other uses the uncertainty in the lidar signal (uncertainty-based cloud detection method; UCDM) as described by Campbell et al (2008) and is primarily used for high-level clouds (i.e. cirrus). Unconstrained COD and extinction retrievals for cirrus clouds are found using bookend effective lidar ratios of 20-sr and 30-sr. Cirrus cloud presence is determined using the cloud top temperature threshold described in Campbell et al (2015) and the NASA GEOS-5 meteorology product. COD and extinction variables are only reported from cirrus cloud layers retrieved using the UCDM. Non-cirrus clouds and all layers found using the GCDM are reported with an IEEE floating-point value Infinity for the COD and extinction. An IEEE floating-point value Infinity is also used whenever the COD/extinction retrieval diverges.
Retrievals of cloud thermodynamic phase (Lewis et al 2020) are performed using a combination of volume depolarization ratios and temperature/pressure profiles from NASA GEOS-5. At the moment, phase retrieval data should be considered provisional and are designated with a low QA confidence level until further validation can be performed.
Hourly and daily cloud fractions are calculated for low, middle, high, and total cloud cover. Cloud-levels are determined based on the cloud top pressure (to the limit of attenuation) as:
These pressure-based levels differ from the height-based definitions used by Lewis et al 2016. If more than 20% of the profiles are missing or attenuated at a cloud-level, a value of NaN is assigned. Note that because a single profile may contain multiple cloud layers, the sum of low, middle, and high cloud fractions does not equal total cloud fraction and may exceed 1.
The V3 CLD product contains both layer and profile variables. The profile_to_layer_index variable is used to map profile variables to the corresponding layer index. Layers are zero-indexed, meaning the first cloud layer is reported with an index of zero. For example, the cloud_base variable may contain three cloud layers at altitudes of 2.37 km, 6.12 km, and 12.01 km. Users can determine the extinction calculated using a 20-sr lidar ratio (variable: extinction_lo) corresponding to the lowest cloud layer by searching where the profile_to_layer_index equals 0, the second cloud layer by searching where the profile_to_layer_index equals 1, and so forth.
NOTE: The flag_layers QA flag should always be used to properly interpret results of the cloud product. Poor signal quality limits the ability to detect clouds. Values in flag_layers indicating anything other than no_problems points to a signal quality issue that may prevent the algorithm from inspecting the entire atmospheric column. For instance, cloud_base may show that no clouds were retrieved; however, flag_layers might indicate a processing problem that makes it impossible to determine whether the profile is truly cloud-free.
CLD Product Levels and QA
L1, L15, and L2 CLD products have identical file formats and content. The primary difference between L1 and L15 CLD products are that data failing to meet the L15 QA criteria are screened and replaced with NaN in the files. In addition, the CLD processing uses NRB products as input. For L15 processing, L15 NRB data are used which in addition to applying the CLD QA criteria yields a net screened CLD product. As a result, simply screening a L1 CLD file based on its QA criteria will not yield identical results in a L15 CLD file. The primary difference between L2 and L1/15 files is that L2 may have additional post-calibrations applied to the input NRB product as well as corrections to instrument temperatures. In addition, any L2 NRB data that fail to meet L2 QA criteria would also be screened and replaced with NaN in the files.
CLD products have three potential QA confidence levels: high, moderate, and low. The CLD QA Confidence Levels are: