Low Desert Almanac

Methods and sources

Where every number on this site comes from, how it reaches the page you are reading, and what happens when a value cannot be confirmed. If you only read one reference page here, read this one.

All sources last checked October 10, 2026 3 entries are still waiting on a check

There are two kinds of authority on this site and they are kept apart. A climate record is measured data: an instrument recorded it, an agency published it, and the number either matches or it does not. An agency's guidance is a recommendation: a body responsible for public safety telling people what to do. The first is reproduced exactly. The second is repeated and linked, never extended, summarised away, or turned into a rule its author would not recognise.

Climate records

Sources

Agency guidance

Sources

Reference

Sources

How a number gets onto a page

Pipeline

No figure on this site is typed into a page. Each one travels the same route, and each step is visible in the repository:

  1. The source file is downloaded from the agency that publishes it, in its original format, and kept. Nothing is transcribed by hand at this stage.
  2. A build script reads it and writes a dataset into the site's data directory. The script parses the file by its real column headings, checks that the row count and date range are what they should be, and fails the build if they are not.
  3. Every value carries its provenance. A dataset records the value, the name of the source, a link to that exact file, and the date it was checked, together, in the same record. A number cannot be separated from where it came from.
  4. The page reads the dataset. A tool or an article imports the value rather than restating it, so correcting the dataset corrects every page that uses it, in one step.

Several pages run a check on themselves as they build. The Season Clock, for example, refuses to build if its compact daily series is not exactly 366 days long, because a series that is one day short would put every date after the gap on the wrong day of the year, and it would do so quietly.

When a value cannot be verified

The rule

No guessed numbers, ever

If a figure cannot be confirmed against a primary source, it is not estimated, not taken from a secondary summary, and not softened into "approximately". It is marked as needing verification in the dataset and listed in the site's data notes, and the page that would have shown it either says the figure is still being checked or shows nothing there.

The same rule applies to whole tools. If a tool could not be built on a verifiable dataset, it was changed to something that could be: the tool was replaced, not the number invented. Numbers that look plausible and are wrong are worse than no number at all, because they are believed.

Two things follow from that, and both are visible on the site. First, some pages are narrower than they could be: the frost and freeze explorer reports how often Phoenix has actually reached freezing rather than pretending the normal year has a frost season, because the normal overnight low here never reaches 32 °F even though individual winters do. Second, unverified items are listed openly rather than hidden: a page carrying one says so at the point where it matters.

Units and time

Conventions

The site uses the units people in Phoenix actually use: degrees Fahrenheit, inches of rain, miles, and a 12-hour clock. Metric equivalents are given where a source publishes in metric and a conversion would otherwise be a surprise, and are rounded to the same number of significant figures as the original, not to a false precision.

All times are Mountain Standard Time, all year. Arizona does not observe daylight saving time, so there is no seasonal shift anywhere on this site. Sunrise, sunset, solar noon, sunrise bearing, the hour of peak normal heat and every hour printed by a tool are in MST. A calculator that applied daylight saving would place the hottest hour an hour early for half the year, and it would look entirely reasonable while doing it.

Climate figures are the 1991–2020 normals, the current thirty-year period published by NOAA NCEI. A normal is an average of thirty years, not a prediction, and every tool says so.

Corrections

Report one

If a figure here is wrong, the fastest way to get it fixed is to open the source yourself and send the link. Corrections are made to the dataset, which updates every page that reads it, and where a change alters what a page claims the page notes the change rather than quietly rewriting itself. Silent corrections make a reference uncitable, which defeats the point of it.

Report a correction