And the replies cogent as well -- a proper admonition given re looking very closely at any determination that a particular station (to be dropped?) is in fact -- a noise producer via, say, SD stats -- or a valid representation of, say, an actual micro-climate. The point well-taken that: measurement error, with real people attending a station: properly determined, should produce much more error than that incorporated via vacuum-cleaning outliers within a given grid.

That rationale could almost define a test-of-the-Tests: a necessary precursor to actually removing a data-contributor: for defensible-Cause.)

(Not that any of this matters even slightly, to the innumerate Murican majority, especially in a time of bloviation manifestly replacing any semblance of actual informed conversation/debate.)