Although I make a number of stand-alone charts, most of my effort goes into creating processes to make many (dozens to thousands of)
charts from relatively raw data streams in a semi-automated way.
Here's how it goes, using the weather charts as an example:
Find the data
Search the web. For my purposes, government websites are a gold mine.
They have lots of data, it's almost always authoritative, and it's often free.
I often use FirstGov.gov as a starting point.
I found climate data I could use on the NOAA site.
Individual pieces are free, but to get it all you have buy CDs, one for the USA, on for global.
Study the data.It's never easy.
Determine the data format, which may or may not be explicitly stated.
There is usually some bad or missing data. Figure out how to identify and work around it.
Create software tools to reformat the data.
Theoretically the PostScript code that draws the chart could do all the preprocessing also,
but it just gets too complicated. It's much better to break the job into chunks.
Create PostScript routines to draw the chart
PostScript is a 'page description' language that allows complete control over how data is displayed.
Creating charts that hold a lot of data while still being easy to read is a fairly painstaking process,
with many iterations.
Feed the PostScript into ImageMagick to create web images
This requires some more shell programming of some kind.
Package the results in web pages
I am not necessarily recommending these, but they are what I use to get things done.