Sensitivity to Cloud.

The detector was tested for sensitivity to cloud coverage on a day which had rapid changes in the cumulus coverage. Sky photographs were taken whilst the device was operating and the fraction of cloud in the field of view was related to the output signal. This was found to be a linear relationship (see figure 2.).

 
Figure 2: The relationship between cloud coverage in the field of view and the output of a thermopile cloud detector. The cloud cover (cumulus) was obtained photographically and matched to the detector field of view.

 
Figure 3: Cloud monitor output over a thirteen day period. The dates indicate the start of the day at midnight. A voltage (faint line) derived from a thermistor gives the detector canister temperature. The bold line gives the compensated output voltage of the sensor element. Both these datasets can be scaled to temperatures with a conversion of C per volt.

For the purpose of the following figures, the data are presented at one hour resolution. However, the data were logged at one minute intervals. The cloud monitor field of view had a half angle of . Figure 3 shows the temperature of the detector canister and the compensated detector output derived by adding the canister temperature to the the uncompensated detector output as indicated in the circuit of figure 1. All the signals are conditioned within the monitor electronics to 50mV per degree Centigrade. Data taken over 25 December to 7 January 1998 are shown in figure 2. The sky was predominantly clear over this period. The exceptions were daytime cloud on December 26, followed by intermittent cloud through the rest of the day and a two day period of cloud beginning in late morning on January 1. On January 5 and 6 there was a build up of cirrus followed by thick stratocumulus. We believe that the peak early on the morning of 29 December was short-lived cloud. It is evident that the discrimination between a clear sky and cloud in the field of view is straightforward at most times.

The monitor viewed towards the south at an elevation of and was backed by an east-west wall. The result was that the monitor was in direct sunlight in the morning and in the evening, causing considerable temperature variations to occur in the canister at those times (figure 3). The temperature compensation is not perfect under those conditions but, in terms of defining a cloud-free or cloudy field of view, it remained adequate under difficult circumstances. There is little random noise effect in the data as can be seen from the results over the clear sky period where the signal is lowest. The result is that cloud in the field of view is readily detectable by the setting of an output threshold level.

 
Figure 4: Cloud monitor output over an extended period including a progressive increase in atmospheric moisture prior to the arrival of a weather front arriving at 13.00 on 14 January. January 6 had complete cloud cover. The following days had intermittent cloud (mainly cumulus) except for the night of 8-9 January which was almost clear.

We have built a second detector which is similar to the first except that it has a window which passes wavelengths above m. The agreement in the outputs of the detectors is very good although there are some minor discrepancies from time to time. These may well be related to atmospheric humidity since there is a strong water vapour band between 5 and m. However, it is clear that the exact wavelength pass band is not critical to the operation of the cloud monitor.

Sloan, Shaw and Williams (1955) showed that at times of increased humidity the integrated signal at wavelengths above the water vapour band increases through a general increase at all wavelengths. We have observed this effect as a baseline shift at times of high humidity. The effects of clouds are still clear but it is then necessary to define a ``clear sky'' output level at the beginning of an observation period rather than using an absolute level which can be set over long periods at other times. In Adelaide at least, the atmospheric water vapour content varies slowly with a time constant of the order of a day and so a ``clear sky'' level which is set at the start of a night will be adequate unless a weather front passes in the observing period. Figure 4 shows an exceptional period in which there was a considerable build-up of atmospheric moisture (from Jan 6 to Jan 14) from an influx of tropical air, followed by a front which replaced that moist air. The atmosphere became progressively more humid over that period although the effect is probably due to aerosol scattering of water droplets rather than water vapour since it covers a large wavelength range outside the water vapour band. Conditions of such humidity would be rare at the sites selected for the HiRes and Auger projects. We have tested the detector under humid, tropical conditions and it responded to cloud but with a significantly changed baseline, such as was observed in Adelaide in the exceptional January humid period. It certainly appears to be likely to continue to be useful under those conditions although more testing in the field is warranted to determine what cloud altitude can readily be accessed and whether the baseline is likely to show significant drifts overnight.

© Copyright Astronomical Society of Australia 1997