8 July 2002
This dataset consists of attenuated backscatter coefficient (beta) measured by the 905 nm Vaisala CT75K lidar ceilometer situated at Chilbolton, England (51.1445°N, 1.4370°W). The data are uncalibrated; see the interpretation section below. The CT75K is a low power instrument designed principally for accurately determining the height of cloud base. However, the backscatter profile also contains potentially useful information on the optical properties of clouds and aerosols. Due to the low power and relatively long wavelength, the CT75K is unable to detect molecular backscatter. It operates continuously and has the following characteristics:
_________________________________________________ Wavelength: 905 nm Beamwidth: 0.6 mrad Range resolution: 30 m Integration time: 30 s Minimum recorded beta: 10-4 (sr km)-1 _________________________________________________
The lidar has operated continuously at Chilbolton, with occasional breaks for maintenance, since summer 1996. It was pointing directly at zenith until 26 April 2000 when it was inclined 5 degrees off vertical towards the west (in order to avoid specular reflection from horizontally-aligned planar crystals), and has remained this way ever since. We may well move it close to zenith in future in order to better match with the radar, which is usually operated at zenith so that the Doppler information is not affected by horizontal winds.
The data acquisition PC uses NTP (network time protocol) to ensure that the time stored in the data files is accurate. However, a software upgrade on 10 August 2000 inadvertently deactivated this feature, with the result that the time drifted somewhat until 1 September 2000, when the clock was found to be fast by 7 minutes. The time was changed from 09:25:10 to 09:18:10 and the NTP process restarted. It is reasonable to suppose that time was being gained in a linear sense, equating to a gain of 19 seconds per day since 08:00 UTC on 10 August 2000.
The Vaisala data acquisition system performs the following in real-time: range correction, noise subtraction, gain control based on the transmit power, and calculation of the first three cloud base heights from the attenuated backscatter profile. Data are recorded as 30-s/30-m averages. The main problem with raw data from this system is the background speckle noise, which is highly variable in intensity (with individual values of beta sometimes going negative), and of a nature that is difficult to understand theoretically. Our approach in removing it is to first estimate the standard deviation of the noise in each ray from the 8 highest gates and store it in the stdn variable (this variable can be used to see how bad the noise problem was and thus get a qualitative sense of the sensitivity of the lidar in each ray). The data are then cleaned up by ascending through each beta profile, and essentially assigning a confidence level to each pixel based on the magnitude of the signal relative to the standard deviation of the noise, the vicinity of other cloudy pixels, the approximate optical depth of the cloud up to that point, and the range to the instrument (favouring low signals that are closeby). Pixels with too low a confidence value are rejected. The results of this algorithm can be seen by comparing the raw and clean quicklooks which should have been provided with the NetCDF data files.
The `ceil2nc' program was used to process the data and convert it to NetCDF; source code for the latest version may be downloaded from http://www.met.rdg.ac.uk/radar/software.html.
The data is provided as daily NetCDF files containing the following variables:
The following global attributes are also present:
Simple programs to read NetCDF files of Chilbolton data into Matlab, IDL and PV-WAVE can be found at http://www.met.rdg.ac.uk/radar/software.html.
It is assumed that the user has a fairly good understanding of the concept of attenuated lidar backscatter coefficient, and is aware of the difficulties involved in inverting it to obtain optical depth. In this section therefore we discuss only calibration issues and the quality of the cloud base height products.
No calibration has been applied to the attenuated backscatter coefficient in the NetCDF files beyond the internal calibration that is applied in the Vaisala data acquisition software. We have been able to evaluate this internal calibration by a method which involves integrating the attenuated backscatter coefficient up through a cloud that completely extinguishes the signal (see for example Spinhirne et al. 1989). Theoretically this integral should be equal to 1/(2k), where k is the extinction-to-backscatter ratio. For liquid water clouds with mean droplet diameters in the range 5-50µm, k at 905 nm should be equal to 18.5 sr, so this approach provides a means of calibration. If the effects of multiple scattering are neglected then we find that the raw beta values are around a factor of 2.05 too low. However, calculations suggest that multiple scattering acts to increase the integrated backscatter by around 20%, with the result that the raw attenuated backscatter coefficients should be multiplied by 2.45 to calibrate them. Due to the assumptions made in the multiple scattering calculations, we estimate the uncertainty in this figure to be around 20%. Users are warned that no comparison with other lidars has been performed to cross check this calibration figure.
The cloud base products are exactly the values produced by the Vaisala data acquisition system. It calculates the height of the first three cloud bases by performing a so-called Klett inversion of the beta profile assuming a fixed extinction-to-backscatter ratio, and considers the slope, absolute value and historic observations at that height (the exact algorithm is regarded as commercially sensitive by Vaisala so unfortunately has not been released by them). The bases of liquid water clouds (at temperatures above and below 0°C) are characterised by high beta values above a sharp gradient in beta, and they typically extinguish the lidar signal very rapidly. Thus when liquid water cloud is present, the algorithm has no difficulty in determining the height of its base, but is unable to reliably detect the base of any clouds that may be above it. In ice clouds on the other hand there tends to be no sharp gradient in beta and the cloud base reported is unreliable. We therefore recommend only using the first cloud base reported by this instrument when the beta at that height is high enough for it to be composed of liquid water.