When you point an infrared thermometer to the skies then soon you will find out that the temperature reading strongly depends on whether it is cloudy or not. Clear blue skies will result in temperatures far lower than the ground temperature, I measured in the evening even -20 degrees C right above me while the surrounding temperature was around 10 degrees C. A clear sky is cold, if we wouldn’t have an atmosphere you would measure the 4K remaining since the big bang. Clouds are well known to have a blanketing effect and this is something you can measure with a single thermopile.
The above graph comes from a Melexis 90614 detector looking horizontally over the hedge the garden, about 50 cm under the rain gutter that is about a half meter wide. So the detector will see both the garden but also a part of the skies. I’ve chosen for this set-up because I want to have some protection against the rain. In any case, the horizontal axis in the graph shows hours relative to midnight 11-nov-2014 and the vertical axis shows the difference between the 90614 chip temperature (also known as the die temperature) and the infrared temperature. So we measure in fact the difference between the cold sky and the ambient temperature.
The temperature difference is a measure for the level of cloudiness, we had overcast up to -18 hours in the graph, after sunrise all clouds disappeared, a nice clear day followed with some occasional high altitude cirrus clouds. Since +10 o’clock this morning we had more broken clouds and occasionally all octas were covered.
A longer article on this all: http://www.planetarduino.org/?cat=1846