What is a dekad? The 10-day clock behind rainfall monitoring

If you've pulled a CHIRPS or TAMSAT file and seen it labeled something like "202609d2," you've already run into a dekad without anyone explaining it. It's the basic time unit most satellite rainfall and vegetation products use instead of a calendar week or a full month, and once you know the convention it stops being confusing and starts being useful.

A dekad is ten days, by a fixed rule

A dekad splits each calendar month into three chunks: days 1-10, days 11-20, and whatever is left of the month, which runs 8 to 11 days depending on the month and leap year. So September has three dekads of 10, 10, and 10 days, and February has 10, 10, and 8 (or 9). That third dekad is the one people trip over, because it isn't a clean 10 days like the name implies. It's "the rest of the month," full stop.

It's the standard behind most of the products an early warning shop touches: CHIRPS, TAMSAT, the FEWS NET rainfall estimates, eMODIS NDVI composites, and the agro-met bulletins that WMO-aligned met services put out across Sub-Saharan Africa and South Asia. When your rainfall layer and your vegetation layer both run on dekads, you can line them up on the same timeline without converting anything.

Why not just use weeks or months

A calendar week doesn't nest cleanly inside a month, which makes seasonal comparison annoying. A 10-day period almost does: three dekads to a month, thirty-six dekads to a year, with that one short dekad absorbing the leftover days. That's close enough to even that you can compare "dekad 2 of June this year" against "dekad 2 of June" in any prior year and know you're looking at roughly the same calendar window, every time.

A full month, on the other hand, is too coarse for what an anticipatory action calendar needs to catch. A short dry spell in the middle of a planting window can matter a lot for emergence, and a monthly total will smear it out against a wetter back half of the month. Dekadal rainfall keeps the resolution tight enough to catch a bad stretch without drowning you in daily noise you'd have to smooth over anyway.

What a dekadal anomaly tells you

Once you've got rainfall or vegetation condition on a dekad basis, the next step is comparing this dekad against the long-term average for that same dekad, usually built off a 1981-to-present or similar multi-decade baseline depending on the product. That comparison, this dekad versus the dekad-2-of-June norm, is what produces the anomaly number that drives a trigger decision: wetter than normal, near normal, or drier than normal for that specific slice of the calendar.

That's also the part that gets tedious fast if you're doing it by hand. Pulling the raw rainfall estimate, pulling the historical dekad average for the same window, computing the percent of normal, and repeating it for every zone on your calendar is the kind of task that eats a morning every single dekad, for as many zones as you're tracking. Food Security Warning's monthly vegetation and rainfall anomaly layer (/) is built to hand you that comparison already done, zone by zone, so the trigger calendar gets the number it needs without the manual pull-and-plot.

If you're the one updating the calendar every dekad and would rather spend that morning on the response plan instead of the spreadsheet, it's worth a look at how the layer is put together.