If anyone here knows NOAA hourly data in depth and can help with the following, I'd appreciate it
1. Arbitrary(?) Minutes: I've done a download of hourly data for a station near me (Atlanta GA USA) for the past 5 years, many (but not all) timepoints are 53 minutes past the hour, e.g. 2:53 a.m. This is just an arbitrary thing, right? They set it at something and it stays on that minute setting until some year, something happens, or whatever?
Or is there actually some reason to be at 53, or to choose particular minutes for particular stations?
2. "Extra" readings: Across the 1825 days in these 5 years' hourly data, there are 24 daily readings at the :53 minute mark for all but 127 days. If you count days with at least 23 readings at :53, only 42 days fall short. Okay, no problem... nothing's perfect and the sensor has glitches or is even down every now and then.
But 1088 days (60%) have more than 24 readings. Indeed, the average number of readings per day is 27, and the single highest has 83 readings in one day. I can't make out anything special about these extra readings... they don't seem to be spurious (like the sensor got stuck in send mode or something).
What's the deal with these extra readings? Anyone have any idea why they are there?
I'm not doubting that their valid. (Or are they?) And I'm not worried about them, per se. But if I am going to be making averages, I don't want to be making the mistake of some days have far more readings in their average than other days; it could be in the cold part of the day, or whatever. Obviously I'd want to limit averages to :53 minute readings. But still, here's the question:
What are all these extra readings for, or why are they there?
Please only answer if you *know* why they might be there. (You have some experience with actual circumstances that can cause extra *NOAA* readings.) We can all make guesses, such as that they simply had more data. I am not asking for guesses. Bottom line: Can someone say with certainty if these might be special extra data.
3. Non-continuous tenths digits: The tenths' digits readings don't seem to be due to temperature alone in many cases. For example, here are readings at station for the six hours starting March 1, 2010 2:53 am: 30.9, 30.0, 30.9, 30.9, 30.0, 30.0. It's jumping from .0 to .9, with nothing in between.
Or, across all 54,234 readings, there are exactly:
710 readings of 69.8 F,
1061 readings of 70.0 F
1061 of 71.1 F,
779 of 71.6 F, and
1115 of 72.0
But there aren't any other .1's digits between 69.8 and 72.0 at all, not a single one... 1061 of 71.1, but not a single 71.0. In truly continuous data (used in the mathematical sense), there should be roughly equal numbers of 71.0, 71.1, 71.2, etc.
Obviously, something is going on here. Some sort of complex function such as converting from celsius, rounding up or down, something or other - it doesn't make a lot of sense, though. I can post more data, if anyone wants.
Any ideas what's up with the odd gradations of temperature values?
Any other insights into how their hardware and reporting works, is appreciated.
Please no one get me wrong - they do a great service, but ultimately there are limitations on how data can be presented, and how well things work. I understand. I am just asking for more info on what is seen in the data, so we might understand it better.
