In brief
A rain gauge can record a tiny amount of rain that disappears when its observations are transferred to another archive. For a dry-spell calculation, that can turn a valid dry day into a gap that breaks the sequence. This study follows versions of the same gauges across precipitation archives to identify when data handling changes the estimated evolution of dry spells. The clearest observed example involves Spanish records; most matched versions elsewhere agree.
Contribution
A comparison of 8,020 station-version pairs across 24 countries holds the physical gauge fixed. Record-level checks and controlled changes to trace encoding connect differences in the daily observations to differences in the resulting dry-spell trends.
Key finding
For 205 pairs involving AEMET, turning its valid traces into missing entries brought the archive trends much closer: the median absolute slope difference fell from 0.0763 to 0.0089 days per year. This experiment identifies information loss as a major explanation of the observed disagreement, rather than a universal bias affecting every archive.
Models and data
Distinctive research elements
Models and methods
- Matched station-version comparison
- Sen slope
- Trace-loss sensitivity analysis
- Physical-station-cluster bootstrap
Data and evaluation
- AEMET
- ECA&D
- GHCN-Daily
- 8,020 matched pairs across 24 countries
Why a trace of rain matters for consecutive dry days
Consecutive dry days (CDD) measures the longest run of days below a chosen rainfall threshold. With the 1 mm threshold used here, an observed trace still qualifies as a dry day. A missing observation has a different meaning: the rainfall amount is unknown. If the calculation stops a run at every missing day, losing a trace can divide one long spell into shorter pieces. At an illustrative Tenerife station, the encoding change reduced an annual maximum from 190 to 33 days. That example shows the mechanism, not a typical effect for all stations.
What the Spanish comparison and the global experiment show
In the Spanish comparison, AEMET retains trace observations that are absent from some corresponding ECA&D and GHCN-Daily records. The impact depends on where those gaps occur: an interruption within a long summer dry spell can matter more than one within a short spell. A separate experiment used trace flags already documented in GHCN-Daily and deliberately treated those days as missing. This changed the statistical trend classification at 5.96% of the 11,101 eligible stations. That percentage describes sensitivity to an imposed change within one archive; it is not a measured global rate of errors between providers.
Implications for interpreting drought trends
The way traces are reported can also change over time. Under a policy that turns them into gaps, a decline in trace reporting can produce an apparent increase in dry-spell length. The practical response is to document the archive version, preserve observation flags and explain how missing days and year boundaries are handled. These checks help distinguish a change in the record from a change in rainfall. The study does not rule out real changes in drought, establish the same effect for every drought indicator, or recommend treating all missing observations as dry days.
Cite this work
BibTeX
@article{semper2026consecutive,
title = {Sensitivity of Consecutive-Dry-Day Trends to Trace Loss in Matched Precipitation Archives},
author = {Marc Semper and Manuel Curado and Jose F. Vicent and Jorge Olcina Cantos},
journal = {International Journal of Climatology},
year = {2026},
doi = {10.1002/joc.70597}
}Always check the publisher record for final volume, issue and page information before citing.