I had actually considered averaging the cloud cover by hour, and after a lot of different iterations I found grabbing any of them individually would result in a net change less than the rounding error. I would say either first or last is fine. Nice job porting it over to perl, I really need to sit down and learn it. I chose python because it is great for the “just make it work” brute force style programming. Perl had a few more rules to remember :p