TITLE: INDOEX: Maldives KCO Photography All Sky Camera (Lobert) Contact/Author: J.M. Lobert Center for Clouds, Chemistry and Climate Scripps Institution of Oceanography University of California, San Diego 9500 Gilman Drive La Jolla, CA 92093-0221 USA Email: jlobert@ucsd.edu; j.lobert@bigfoot.com Telephone: +1 858 822 0297 Fax: +1 858 534 4922 URL: http://www-indoex.ucsd.edu/observatory/ http://www-c4.ucsd.edu/ OR V. Ramanathan Center for Clouds, Chemistry and Climate Scripps Institution of Oceanography University of California, San Diego 9500 Gilman Drive La Jolla, CA 92093-0221 USA Email: ram@fiji.ucsd.edu Telephone: +1 858 534 8815 Fax: +1 858 822 1632 URL: http://www-indoex.ucsd.edu/observatory/ http://www-c4.ucsd.edu/ 1.0 DATA SET OVERVIEW: The data set is a collection of 17140 daytime cloud images taken with a Total Sky Imager (TSI; Yankee Environmental Systems) at the Kaashidhoo Climate Observatory (KCO) in the Maldives, during the 1999 INDOEX intensive field phase. The data set starts on 19 February 1999 at 10:43:39 Universal Time and ends on 31 March 1999 at 13:08:33 UT. KCO is located at 4.9664 degrees N and 73.4657 degrees E, the TSI was mounted about 5 m above sea level (mean elevation of the site is about 0.5-1.0 m) on the rooftop of the accommodation house. KCO local time is UT + 5 hours. For website info, see authors. For an overview of data, see cld_cvr.jpg 2.0 INSTRUMENT DESCRIPTION: Taken from the instrument website (with live demo): http://www.yesinc.com/product/imaging.html Company website: http://www.yesinc.com/ Images of the sky are captured via a solid state CCD imaging camera that looks downward on a hemispherical, heated, polished, stainless steel mirror. The camera electronics is environmentally sealed for long life. A JAVA program running on a local or remote host computer captures images via TCP/IP at a user-defined sampling rate and saves them to JPEG files for analysis. The TSI analysis software first masks out obstructions such as the camera, its arm, and the sun-blocking band. The resulting binary image (blue for sky, white for cloud) is then analyzed for fractional cloud cover. The website shows a side-by-side comparison of cloud image before any processing has been performed and the same image with a software filter applied. The filter, a sophisticated image-analysis algorithm, clearly defines the clouds so that fractional cloud cover can then be calculated. For a demonstration on how the cloud cover is calculated, see the instrument website at http://www.yesinc.com/product/imaging.html. 3.0 DATA COLLECTION AND PROCESSING: The frequency of image acquisition is variable: One image every 10 minutes between 19 and 21 February 1999 (06:20) One image every 5 minutes between 21 February (06:42) and 07 March 1999 One image every minute between 08 March and 30 March 1999 (08:43) One image every 5 minutes after 30 March 1999 (08:43) Image description: Each image has an identical frame due to the fixed location of the TSI. The image of the hemispherical mirror shows the black band that blocks out the sun (moving around with the sun) and the thin, black arm that holds the camera (fixed, at the top). Along the periphery, the KCO lab building and observation tower is clearly visible at the north side (top of image), slightly to the left of the camera arm. Some trees protrude into the image mostly along the left / west side. On the outside of the mirror, the TSI body and the rooftop are visible. Cloud cover data description: From the original JPG image, the software estimates a fractional cloud cover. Cloud cover data are published alongside the images (file cld_cvr.txt). Only about 70 degrees of the hemispheric field of view are used for the estimate of cloud cover, this portion is not cosine corrected but instead treats the image as a planar area, which introduces a small uncertainty in the estimate. 4.0 CD-ROM LAYOUT AND DATA FORMAT: All of the documentation files are located in the top level directory of this CD-ROM. The CD-ROM is divided into 41 subdirectories, one for each day. The naming convention for the directories is YYYYMMDD, where YYYY is the four digit year, MM is the two digit month, and DD is the two digit day. Within each directory are the individual files, which are compressed raster images in JPEG format (Joint Photographic Experts Group, for questions on this data format, see http://www.faqs.org/faqs/jpeg-faq/part1/). The naming convention for the individual files is HHMMSS and the extension "jpg", where HH is the hour, MM are the minutes, and SS are the seconds. Data acquisition of the TSI takes less than one second, hence the file name reflects the correct time when the image was taken. File name example: under the 19990324 directory, the file 013500.jpg is the image taken on 24 March 1999 at 01:35:00 Universal Time. The accompanying text file cld_cvr.txt contains the cloud cover fraction for each image. This data set is tab-delimited with the following headers and formats: Date Time DateTimeSpreadsheet DOY CloudCover DD-MM-YYYY HH:MM:SS decimal decimal percent 19-02-1999 10:43:39 36210.44698 50.44698 18 Data version is 1.0 20000424_final_01. 5.0 DATA REMARKS: The TSI software has a few parameters that can be adjusted to fine-tune the image processing and cloud cover estimate. This algorithm is optimized for clear skies in the north-east of the USA. We experienced difficulties applying this algorithm to the polluted skies over KCO. The haze creates a glare that is interpreted as cloud cover, particularly when the sun was near the horizon (morning and evening). In addition, thin strato-cirrus clouds were not always detected. With the current algorithm, we estimate the total, combined uncertainty of the above limitations to +/- 20% absolute cloud cover. The overall, mean cloud cover between 06:10 and 18:16 local time during the INDOEX IFP was 33.9% (StdDev=32.1%; Median=21.0%; N=17140; StdErr=0.3%; first quartile=8.0%; fourth quartile=56.0%). A future recalculation of cloud cover based on a new algorithm is possible, users will be notified of changes in this dataset. Missing images are due to daily cleaning of the mirror and the camera lens (a few images at a time) or due to downtimes for maintenance on the interior hardware or software (up to ~0.5 hours). Longer downtimes (several hours) are due to software problems or computer / network downtimes. 6.0 REFERENCES: Data will be briefly described in this paper: J.M. Lobert, V. Ramanathan, J.M. Prospero and A. Majeed, Kaashidhoo Climate Observatory (KCO): A new site to study long-term climate changes in the Indian Ocean. In progress for JGR (INDOEX special issue), probably mid to late 2001.