VOCALS2008_WbandCloudBoundaries10min_v1.0.readme.txt :: 2009 December 31 :: Simon de Szoeke README metadata for VOCALS2008_WbandCloudBoundaries10min_v1.0.nc Title: W-band 10-minute cloud boundaries (top and base height) for VOCALS 2008 Version: 1.0 Creation date: 2009-12-30 Author: Simon de Szoeke Data are provided by Simon de Szoeke (Oregon State University), Chris Fairall and Daniel Wolfe (NOAA/ESRL/PSD), and Sandra Yuter (North Carolina State University) with funding from the NOAA Climate Program Office. Find this metadata at ftp://ftp.coas.oregonstate.edu/dist/sdeszoek/vocals/VOCALS2008CloudBoundaries10min_v1.0.readme.txt Summary Cloud top height from two sources, radiosonde profiles and a W-band cloud radar, and cloud base height from a laser ceilometer deployed for two legs aboard the NOAA research vessel Ronald H. Brown, are averaged or interpolated to 10-minute time resolution. Corresponding to the time scale of a large boundary layer eddy, 10 minutes is an appropriate time scale for sampling boundary layer cloud variability. Data Set Variables TIME in yearday of 2008 universal time (UTC). 2008 January 1 00:00 UTC is yearday 1.000. Time records mark the beginnings of consecutive 10-minute time intervals. CLOUD TOP HEIGHT (m) at 10 minute time resolution from the W-band 3 mm cloud radar for VOCALS 2008 leg 2. More extensive description of the cloud top data are found in Appendix 1. Cloud top heights from the W-band that are lower than the cloud base height of the ceilometer are excluded. CLOUD TOP HEIGHT (m) at 10 minute resolution interpolated from VOCALS 2008 ship-based radiosonde temperature profiles. Radiosondes were launched nominally every 4 hours. Cloud top heights from radiosondes are assigned to the 10-minute window during which the sonde was launched. Cloud top heights between launch times are temporally linearly interpolated from previous and subsequent cloud top heights. No interpolation is done when the time between rawinsonde launches is longer than 6 hours. Rather than clouds themselves, inversion base height is measured by the sonde temperature profile. The inversion base height coincides within a few meters of cloud top. The inversion base height is computed as the minimum temperature below the inversion, which is the strongest stable temperature gradient below 2 km altitude. CLOUD BASE HEIGHT (m) at 10 minute resolution from the Vaisala near-infrared laser ceilometer. The 85th percentile of cloud base height below 2 km is used to minimize contamination by scud clouds and rain, which introduce backscatter below the stratocumulus cloud base. Solar radiation within an hour of local noon increases noise and false-positive cloud detection. Solar aliasing is minimized by requiring cloud bases to be below 2 km. Cloud base heights are excluded that exceed the cloud top height from the W-band radar. The NUMBER of cloud retrievals entering the radar and ceilometer remote sensing cloud height estimates are recorded. Statistical certainty of the heights are reduced when the number of cloud retrievals is lower. W-band radar returns are collected at a frequency (dwell cycle) of 3.5 Hz. Normally there are 210 returns in a minute, and 2100 in a 10-minute window. Based on their reflectivity and vertical continuity (Appendix 1) these returns are identified as clear, cloudy, or indeterminate (due to suspicion of non-meteorological clutter or noise). Ceilometer backscatter returns are averaged to 15 second intervals by the instrument. Normally 40 returns are found in a 10-minute window. Ceilometer returns are identified as clear, cloudy with one cloud, or cloudy with multiple cloud levels. Obscured and partially obscured returns are treated as indeterminate. When the ceilometer detects multiple clouds, the lowest cloud base is used for the cloud base height. This is sensitive to contamination by scud clouds and rain below the stratocumulus cloud base. Note the W-band radar is less sensitive to clouds than the ceilometer. The ceilometer cloud fraction is to appear in another data set. Missing values are stored as IEEE not-a-number (NaN) values. Appendix 1: W-band Cloud Height Retrieval Cloud top height is computed from the NOAA/ESRL/PSD vertically-pointing W-band (3.17-mm wavelength, 94.56 GHz) radar during leg 2 of VOCALS 2008. Moments from the radar in 25-m range gates are available at a frequency of 3.5 Hz. Clouds and Doppler velocities were detected from W-band data available after 2008 November 5 (year day 310). Before November 5 the radar was operated with insufficient sensitivity to detect clouds. The highest cloud top was computed from the cloud-sensitive 3.5-Hz mean reflectivity moments. The algorithm balances sensitivity to the highest clouds with elimination of non-meteorological high-reflectivity moments: 1. Cloud Detection from W-band Moments Reflectivity returns at (3.5 Hz) are compared with the range-dependent minimum detectable reflectivity. Reflectivity above this threshold is mostly from scattering by clouds or precipitation. Reflectivity above the threshold might also be due to non-meteorological targets (clutter). They also have a small probability of being noise. Returns sampled at 3.5-Hz are identified as cloudy, clear, or non-meteorological. Non-meteorological returns (including noise) have no reason to appear near the true cloud top, so the cloud top height algorithm takes measures to exclude non-meteorological returns. In practice most non-meteorological returns yield cloud tops consistent with those from cloudy returns. Clouds being nearly always vertically homogeneous, we eliminate most noise and clutter above the cloud by insisting that potential clouds have 3 adjacent range gates with reflectivity above the noise threshold, i.e. clouds are at least 75 m thick. 2. Finding Cloud Top Heights The potential cloud returns from step 1 are binned into 1-minute intervals. (There are usually 210 3.5-Hz time slices per minute.) Clouds are excluded above the maximum allowed cloud height, defined as the height of the highest range at which at least 3 clouds are found in the minute. This helps exclude noise above a cloud. Cloud top height is still computed if there is no level with 3 cloudy moments. Cloud top height for each 3.5-Hz vertical slice is computed as the highest remaining cloud top below the maximum allowed cloud height. These cloud top heights are mean-averaged over the minute. 3. Averaging to 10-minute Intervals One-minute cloud top heights are meaned into evenly-spaced 10-minute intervals. The first 10-minute interval in a day begins on midnight UTC. All 1-minute cloud top heights in which a cloud top was found are weighted equally in the 10-minute mean. Version History beta version 0.1 was oversampled at every 10 seconds, rather than every 10 minutes. beta version 0.2 is at every 10 minutes. Also cloud bases are excluded that are more than 100 m above the coincident or interpolated sonde inversion top.