This is very much dataset, project and software dependent.
Sometimes I remove columns, other times I combine them (averaging, adding, etc), other times I split the dataset into "themes" like "all PM data" or "all flow data" or "all temperature data".
Sometimes I use a spreadsheet tool (LibreOffice) but most times I work with a scripting language (R, Python, Octave) where I can easily select which parts of the dataset I want to look at.
What's "universal" is that you need to know (and understand) what each column means and have access to any metadata available so that you can judge which columns are relevant to your question and what might explain certain weirdness in the data.
From the looks of those plots there seems to be several columns that have the same value for most of the dataset (firmware version, macaddress, hardware) while others seem to be "time axes" like "UTCDateTime", "uptime". My recommendation would be to first write down a few very specific questions you want to answer (what's the average PM10 concentration?; what's the range of temperature observed? ... etc) and only look at the parts of the data that help you answer them. I've always found that when I start with a specific question, the data often lead me into the next (more complex) question and I get more proficient in dealing with that particular dataset.
Is this a question? Click here to post it to the Questions page.
Thanks! Yes, i started by removing things like firmware #, for sure. It's interesting to see how some values track each other roughly, while others don't. But I agree it makes sense to start with a question! Thanks @guolivar!
I also noticed that some very high values make others basically unreadable. So having 2 y-axis scales can sometimes help, i guess, although that's not so simple with some graphing software (like google spreadsheets).