Are ASAP's drought warnings getting harder to trigger?

JRC ASAP issues an agricultural-drought warning when an indicator's z-score falls below −1. Those z-scores are normalised against the whole historical record, so if an indicator trends upward, the same physical conditions stop crossing the threshold — and warnings quietly become rarer. This page measures that drift in the raw indicator statistics behind the warnings.

How often each indicator breached the −1 threshold

Share of in-season dekads where the unit-mean z-score sat at or below −1, averaged over the first five and last five years of the record. A bar that moves left has become less able to trigger a warning.

Indicator

Admin unit (ASAP GAUL 1)

Dekad

Theil–Sen slope of the seasonal-mean z-score, per decade, for each admin unit. Bars marked n.s. are not significant at p < 0.05 (Mann–Kendall).

Table view

Method & caveats

Source. ASAP's own per-admin-unit indicator statistics (export/rum/export.php), the same dekadal unit means its warning classification is thresholded on. Open EC/JRC data.
Trends. Dekadal series are seasonal and autocorrelated, so each is first collapsed to one value per unit-year (in-season dekads only), then fitted with a Theil–Sen slope and tested with Mann–Kendall. Year-to-year persistence is not corrected for, so p-values are mildly optimistic — read p < 0.05 as "worth a look", not as a formal result.
Derived z-scores. ASAP publishes zFPARc, zFPAR and SPI-3 already normalised. It does not publish z-scored WSI — despite the download page calling variable 160/170 "zWSI", the export returns raw WSI on a 0–100% scale. For those and the other raw variables, the z-score here is derived against each unit's own per-dekad climatology, mirroring ASAP's normalisation.
This is not a warning count. ASAP requires the threshold over ≥25% of a unit's active area, a within-unit spatial criterion the published unit means cannot reproduce. The breach share here is a proxy — reliable for direction, not a count of issued warnings.
Trend ≠ climate signal. A rising FPAR trend can be cropland expansion, intensification or sensor history rather than a wetter climate. The consequence for warning likelihood is the same either way, which is the point here, but don't read a biomass slope as a rainfall statement unless SPI-3 agrees.
Comparable window. The records start at different dates (meteo 1989, WSI 1991, MODIS FPAR 2001), so headline figures use onward. Full-record trends are in the repo's data/processed/ outputs.