However these plots are not a simple xy since there may be a correlation between temperature/humidity and clouds.
To improve the plots it would be best to plot for example humidity vs dlwir/ulwir at a fixed temperature and cloud cover. The problem is there are too few corresponding points to get a meaningful result.
The following plots were made by inspecting plots and choosing a range of values for each parameter where the dlwir/ulwir change is minimal (about 10% or less)
As a trial cloud coverage was replotted at a much closer variation in the other 2 parameters - this shows a good correspondance with the wider variation but with increased variability.
It should be pointed out that the dlwir as a % of ulwir is a combination of at least all the 3 parameters considered. All that can be gleaned from these plots is the effect of variation of one parameter whilst holding the others static.
It should be noted that cloud cover is only measured during daylight. All the plots below are therefore only relevant for daylight.
From the above it can be seen that the:
temperature effect is inconsistent and small
relative humidity is the largest effect - more humidity more DLWIR
Cloud cover is significant - more clouds more DLWIR