Rainfall Records Explained and Interpreted
By WeatherRecall editorial project. Date reviewed: . Last updated: 2026-10-06.
A rainfall record is inseparable from its duration and measurement area. The largest one-hour amount is not the largest daily total, and a gauge's record is not automatically a city's record. Short bursts strain drainage differently from prolonged rain. To interpret an extreme, first establish what water quantity was measured, where it was collected, and which interval it covers.
Understand the depth being reported
Rainfall depth describes how deep the collected water would be over a horizontal surface if it did not run off, soak in, or evaporate. One millimetre equals one litre per square metre. This is a depth measurement, not the water volume entering a particular drain. Converting it to runoff requires area and hydrological assumptions.
Precipitation may include rain and the water equivalent of frozen precipitation. Snowfall depth measures a different quantity and cannot be added directly to rain millimetres. The Open-Meteo archive documentation distinguishes precipitation sum, rain sum, and snowfall. Check these definitions before comparing a dashboard's “rain” label with an archive variable.
NOAA GHCN-Daily includes total daily precipitation, snowfall, and snow depth, with element-specific metadata and flags. A trace, a recorded zero, and a missing value have different meanings. Apply the source's documented conventions rather than converting all three into dry days.
A calendar day is not every possible 24-hour window
A calendar total uses fixed boundaries, such as local midnight to midnight. A rolling total examines overlapping intervals, such as every consecutive 24-hour window in an hourly series. A storm crossing midnight can split across two calendar dates even when most rain falls within a few hours.
Manual gauge readings can define yet another observation day. A reading taken in the morning may summarize the preceding day and night. The date label must be interpreted with the observation schedule. For a rolling extreme, name the start and end times, timezone, and sampling resolution; a daily-only archive cannot reconstruct all possible hourly windows.
Illustrative example: midnight splits a storm
Illustrative storm only; all amounts and times below are invented, not real observed data. A hypothetical gauge records 18 mm from 22:00 to 23:00, 22 mm from 23:00 to midnight, and 16 mm from midnight to 01:00. There is no other rain during the surrounding two days.
| Interval | Amount |
|---|---|
| First calendar day | 40 mm |
| Second calendar day | 16 mm |
| Full three-hour storm | 56 mm |
The largest calendar-day total is 40 mm. A rolling 24-hour window containing the whole storm totals 56 mm. Both are correct, but they are different record categories. The storm-average intensity is 56 ÷ 3 = 18.7 mm per hour, while the largest measured one-hour amount is 22 mm. Neither establishes the maximum five-minute intensity.
Over an illustrative 100 m² roof, 56 mm corresponds to 56 × 100 = 5,600 litres arriving as precipitation. That is not a validated runoff prediction: roof storage, losses, drainage timing, and overflow change the discharge. The arithmetic demonstrates depth-to-volume conversion, not a drainage design standard.
Check collection problems and spatial variation
Wind can reduce collection efficiency, particularly for frozen precipitation. Splashing, evaporation, blocked funnels, and an unlevel gauge can affect amounts. Some instruments behave differently under very intense rain. A suspected record should trigger examination of instrument type, maintenance notes, quality flags, and nearby evidence.
A convective shower can produce sharply different totals over a small area. Neighbouring gauges need not agree exactly. Radar or reanalysis can provide spatial context, but those products have their own retrieval or modelling limitations. ERA5 represents model-grid conditions, not every local gauge maximum.
An accumulated multi-day amount must not be assigned to a single day without justification. If an observer missed readings, the total may be valid for the combined interval while its daily distribution remains unknown. Also check whether recent entries are preliminary and whether an official authority has reviewed a candidate record.
Records do not directly reveal return periods
The largest amount in a short sample is merely its maximum. Estimating a return period needs an appropriate extreme-value method, sufficient data, and assumptions about the underlying climate. A “one-in-100-year” event usually describes a 1% annual exceedance probability under the adopted model, not a storm scheduled once each century.
Illustrative probability calculation, not an observed frequency or local risk estimate: if annual exceedances were independent and the probability stayed at 1%, the chance of at least one exceedance over thirty years would be 1 − 0.99³⁰, approximately 26%. Changing climate, dependence, and uncertain model estimates can make those assumptions unsuitable.
Wet-day counts also require a threshold. Counting any nonzero amount differs from counting days at or above 1 mm. State the criterion and denominator, and exclude missing days explicitly. Do not turn a descriptive archive frequency into an operational forecast.
Rainfall records are not flood records
Flooding depends on antecedent soil moisture, catchment area, river levels, drainage, topography, and the timing and location of rain. A modest amount on saturated ground can matter more than a larger amount on dry, permeable soil. Gauge totals alone cannot establish water depth at an address or determine infrastructure safety.
Use the date tool to investigate a reported storm and the year tool for seasonal context. Review methodology and source definitions before ranking totals. Current flood decisions require official warnings and local expertise, not a historical record table.
Related articles
References
About the author
WeatherRecall editorial project compiles educational summaries from the listed public documentation, not original field research. It is not a credentialed meteorologist or drainage engineer. Invented rainfall and probability calculations demonstrate concepts and are not official records, forecasts, or design advice.