Total and Low Cloud Cover

Documentation written by F. Fetterer based on material provided by C. J. Hahn

Decadal (or near -decadal, that is, for 1952 through 1970, 1971 through 1980, 1981 through 1990, and 1991 through 1995) monthly mean values of cloud cover were prepared for this Atlas by Carole J. Hahn, Department of Atmospheric Sciences, University of Arizona, Tucson, AZ. The monthly means were re-gridded to EASE-Grid at NSIDC. The monthly mean values for total and low cloud cover are based on synoptic surface observations. Surface observers usually report cloud cover in tenths. Low clouds are those clouds with bases generally below about 2 km above the surface. They are encoded in the WMO synoptic code under the category "CL". Nine different cloud types (all variations of stratus, stratocumulus, cumulus, and cumulonimbus) are included in this code. (Cumulonimbus is classified as a low cloud because of its base height, though its top may reach the tropopause.)

Data Source for Cloud Fields

The source data used for the gridded fields are from data set H99, Extended Edited Synoptic Cloud Reports from Ships and Land Stations over the Globe, 1952-1996 available from the Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, Oak Ridge, TN. The data are also available from the Data Support Section, National Center for Atmospheric Research, Boulder, CO (NCAR ds292.2). See Hahn and Warren [1999] for a full description of H99. The source data for H99, in turn, are synoptic records: COADS data for ocean observations [Hahn et al., 1992] (see the section on "Description of COADS"); the "SPOT" archive of the US Navy Fleet Numerical Meteorology and Oceanography Center (FNMOC, formerly FNOC) for 1971-1976 land observations, and data from the NOAA National Centers for Environmental Prediction (NCEP, formerly NMC) for 1977-1996. These data sets were obtained from NCAR. Land observations prior to 1971 are not used for reasons given in Warren et al. [1986]. Ocean observations of cloud type are considered unreliable prior to 1952 [Warren et al., 1988]. Synoptic cloud data contained elsewhere on this Atlas (Russian Ice Patrol ship cloud data, western drifting station cloud data, and most North Pole drifting station data cloud data) are not part of H99.

In creating H99, raw data were checked for quality, and observations were discarded if total cloud amount was less than low cloud amount, or if total cloud amount was equal to zero, when present weather indicated rain or snow.

Average cloud cover was computed using ship observations that fell in grid "boxes" that are 5° x 5° (latitude by longitude) south of 50 degrees North, 5° x 10° between 50 degrees and 70 degrees North, 5° x 20° between 70 degrees and 80 degrees North, 5° x 40° between 80 degrees and 85 degrees North, and 5° x 120° between 85 degrees and 90 degrees North. For land observations, the size of the grid boxes was 2.5° x 2.5°.

Sverdrup [Sverdrup, 1933], commenting on observations from the Maud, was one of the first to note that observers generally underestimate cloud cover at night: the "night detection bias" [Hahn et al., 1995]. Hahn et al. [1994] developed criteria by which to classify nighttime observations according to the illuminance of moonlight or twilight on cloud tops, and used a threshold illuminance to screen observations that are likely to be in error. The illuminance criterion is met when the sun is at altitude greater than minus nine degrees or the phase and altitude of the moon are such that the relative lunar illuminance (as defined in [Hahn et al., 1995]) is greater than the threshold value of 0.11. Screening in this way minimizes the night detection bias but results in a monthly sampling error, since only about two weeks of data (around the time of the full moon) contribute to an average in any month. Multi-year averages reduce this error.

In computing total and low cloud amount decadal monthly averages, "daytime" observations are defined as those taken at the four synoptic hours (out of the total of eight synoptic hours: 0000, 0300, 0600, 0900, 1200, 1500, 1800, 2100 GMT) that cover times 0600 local time (LT) to 1800 LT for a box. "Nighttime" observations are defined to be those taken at the four synoptic hours that cover times 1800 LT to 0600 LT for a grid box. Averages for daytime and nighttime were computed separately and then averaged, when a sufficient number of observations were available, in order to minimize day-night sampling bias: a potential bias that results from the majority of observations used for the average coming from daytime hours. (This is because many nighttime observations are eliminated by the illuminance criterion.) The averaging method is:

If there are at least 50 "daytime" and at least 50 "nighttime" observations for a box in an averaging period, the average total or low cloud amount is the average for "daytime" observations plus the average for "nighttime" observations divided by two. If there are less than 50 observations for either day or night, the average is computed without distinguishing between "daytime" and "nighttime" observations. If there are less than 100 observations in total for the entire averaging period, the average value for the box is set to missing prior to regridding to EASE-Grid at NSIDC.

Long-term (1952 through 1995) monthly means were produced from the decadal monthly means by taking the average of the decadal monthly mean value for each grid cell. If a grid cell was coded as "missing" for more than one out of the five "decades," the long-term monthly mean for that grid cell was set to missing.

Method of constructing the cloud products

The decadal monthly averaged, gridded data were re-gridded from the variable box size described above to equal-area EASE-Grid using Cressman interpolation [Cressman, 1959]. In Cressman interpolation the value at a point (the EASE-Grid cell center) is determined by the value of the nearby surrounding points (the centers of the irregularly spaced boxes) weighted by the distance of the box center from the EASE-Grid cell center:

EASE-Grid cell center

where xi is the value at box i within a radius R of the grid cell center, and n is the number of box centers found within radius R. Weights wi are given by

EASE-Grid cell center,

where d is the distance from the EASE-Grid cell center to the box centers. A radius of 500 km around each EASE-Grid cell center was used. (In comparison, the average spacing between box centers is 486 km for ocean boxes and about half that for land boxes).

Gridded cloud field products

Cloud amount products are:

Note that land observations are not included prior to 1971 (see "Data Source for Cloud Fields). Fields for periods prior to 1971 may have some cells with data over land, however, because COADS contains cloud observations from ships on large lakes, and the gridding process may spread ocean observations over land grid cells.

The browse versions of these files are shown as .gif format images with a color bar and contours. These browse files are for the purpose of quickly visualizing the content of the corresponding ASCII data files. The IDL routine used to color-map the images gives a smooth and visually pleasing result, but keep in mind that the gridded ASCII files have a value only for every grid cell. The grid cell centers are shown as red dots. For information on the structure of the gridded files, see the section on "EASE-Grid".

The re-gridding from binned data to EASE-Grid by Cressman interpolation tends to smooth over inhomogeneities in the fields that might result from the data sampling problems caused by varying station density. At the same time, real local variations are smoothed over.

Figures corresponding to the decadal monthly mean products show the number of observations in each grid box of the original data (after filtering for more than 100 observations). These are included to give an indication of the regional and temporal variability in the density of the source data. For some locations and seasons, such as over frozen oceans in winter, there are not enough observations available to produce reliable products. However in general a surprisingly large number of observations exist. Satellite imagery might be expected to provide more data than surface observations, and can be used to construct cloud climatologies (for example, the World Climate Research Programme's International Satellite Cloud Climatology Project, [Rossow and Schiffer, 1999]), but methods for extracting cloud properties from satellite data have unique problems in polar regions.

When graphed, cloud distributions as seen from the limited field of an observer on the surface tend to be U-shaped, with observed cloud amounts either 0-2 tenths, or 8-10 tenths, and dominated by low cloud observations. In summer (June, July, and August), distributions tend to be J-shaped (that is, with no maximum at 0-2 tenths). This is illustrated in the figure of cloud amount frequency from Ice Station Alpha (Figure 3). A year of data for NP-4 also shows a similar distribution (Figure 4).

Figure 3.  Frequency of cloud amount, from 3 hourly observations from western drifting station Alpha

Fig. 3. Frequency of cloud amount, from three hourly observations from western drifting station Alpha (from Untersteiner [1961]).

Figure 4.  Frequency of cloud amount, from observations from North Pole drifting station NP-4

Fig. 4. Frequency of cloud amount, from observations from North Pole drifting station NP-4 (from Makshtas et al. [1998]).

Description of COADS

COADS, the Comprehensive Ocean-Atmosphere Data Set, contains global surface marine data from 1854 through 1997 (at present). Observations from ships, moored environmental buoys, and drifting buoys are quality-controlled and made available in the most extensive data set of surface marine observations ever compiled. Basic parameters are air temperature, sea surface temperature, barometric pressure, wind, humidity, cloudiness, weather, wave, and swell. COADS is a project of NOAA's Climate Diagnostics Center, the NOAA National Climatic Data Center (NCDC), the University of Colorado Cooperative Institute for Research in Environmental Science (CIRES), and the National Science Foundation's National Center for Atmospheric Research (NCAR). For more information on COADS see (http://www.cdc.noaa.gov/coads/).

Credit for the genesis of COADS is given to Joe Fletcher, who drew attention to the value of long-term marine records for climate research before it was generally recognized [Fletcher et al., 1981]. International efforts at standardizing how weather observations are made, beginning with those of Lt. M.F. Maury in 1853, and now formalized in WMO resolutions, have made COADS feasible. NCDC received, consolidated, and in many cases digitized ("keyed") these data sets of early observations [Elms et al., 1992], [Elms, 1993].

COADS has been made available in three major releases: Release 1 (1854 through 1979), 1a (1980 through 1997), and 1b (enhancements to 1950 through 1979). Release 2 (~1820 through 1997), with additional data, enhancements to metadata, and bias corrections for wind records, is being planned [Woodruff et al., 1998; Elms et al., 1999]. Quality control in Release 1 was accomplished by "trimming," that is, by discarding observations that exceeded a 3.5-sigma level. This level was extended to 4.5 sigma in Release 1a and 1b. An adaptive trimming approach is being considered for Release 2 [Woodruff et al., 1998].

This Atlas makes indirect use of COADS for two products: cloud fields over ocean (COADS observations are the source of the cloud amount observations used) and pressure fields. (COADS observations are one data source used in constructing the gridded Reanalysis pressure fields that are used to make the products on this Atlas.) The following table notes the marine surface observation data provided with this Atlas that are in the current version of COADS or are planned for COADS.

Table 3. Relation of Atlas marine data to data in COADS

Marine Data Source on this Atlas Data in COADS
AIDJEX ice stations Planned for Release 2
Ice Station ARLIS I Planned for Release 2
Ice Station ARLIS II Planned for Release 2
Ice Station Alpha Planned for Release 2
Ice Station T-3 (Ice Station Bravo) Probably not, maybe some
Ice Station Charlie Planned for Release 2
Maud Planned for Release 2
Fram Temperature and pressure, through Deck 192
Russian North Pole drifting stations NP-1 and some other early stations, through Deck 186 (USSR Ice Station Surface Synoptic). Later stations in COADS over GTS, and through Deck 733 (Russian AARI NP stations, source was University of Washington, about 1992). For cloud data, these are total cloud only.
Ice patrol ship data No
DARMS No