How Do You Measure Rain? A Bucket With More Problems Than Expected

Rainfall is expressed as a depth, meaning the thickness of the layer that would accumulate on a flat surface if none of it drained, evaporated or soaked in, which is convenient because it is independent of the area collected. A gauge therefore has a funnel of known area leading into a container, and the volume collected divided by that area gives the depth. The standard is a defined orifice diameter at a defined height above the ground, and the height matters because wind speed increases with distance from the surface. What is actually wanted is the rainfall over an area, which a gauge samples at a single point, and the gap between those two is the source of most of the difficulty.

Several effects cause a gauge to under-record systematically:

• •Wind, which accelerates over the gauge and carries drops past it, the largest error by a wide margin
• •Wetting of the funnel, since water adhering to the surfaces never reaches the measure
• •Evaporation between the rainfall and the reading
• •Splash out of the collector and splash in from the surroundings
• •Snow, which blows over the orifice entirely and is badly under-recorded
• •Undercatch from wind alone commonly reaches several per cent and can exceed half in snow

Recording rainfall continuously without an observer required mechanisms with their own characteristics. A tipping bucket gauge directs water into a small balanced vessel that tips when full, emptying itself and producing an electrical pulse, and each pulse represents a fixed depth, which makes it simple and reliable and gives it a known weakness, since water arriving during the tip is lost and heavy rain is therefore under-recorded. Weighing gauges record the mass of accumulated water continuously and handle snow and intensity better at greater cost. Optical instruments count and size drops passing through a beam and measure nothing physically. Radar estimates rainfall over whole regions from reflected signal and must be calibrated against ground gauges, which is why the two systems are used together.

Siting is at least as important as the instrument and the standard guidance is specific. The gauge should stand on level open ground with no obstruction closer than about twice the height of that obstruction, since buildings, trees and fences disturb the airflow and either shelter the gauge or accelerate wind over it. The orifice should be level and at the standard height above the ground, which differs between national services and must be recorded. Ground cover matters, since short grass reduces splash while hard surfaces increase it. The site should not change, which is the hardest requirement over decades since trees grow and buildings appear, and a documented change in surroundings is what allows a discontinuity in a long record to be identified rather than mistaken for a trend.

The deeper problem is that rainfall varies enormously over short distances, particularly in convective storms where one street receives a downpour and the next nothing, so a gauge reading applies to its own location and extrapolating it across a catchment introduces error that no improvement in the instrument reduces. Networks address this with many gauges, and the number needed rises sharply with how variable the rainfall is and how small the area of interest. Radar covers area at the cost of measuring reflected energy rather than water. Combining the two, using gauges to correct radar fields, is standard practice in flood forecasting. And any long record is affected by changes in gauge design, siting and surroundings, which must be identified before a trend can be claimed.

Rainfall is measured as a depth through a funnel of known area, and wind accelerating over the gauge carries drops past it, which is the largest error and commonly reaches several per cent and far more for snow. Tipping bucket gauges lose water during each tip and under-record heavy rain. Rainfall varies over short distances, so a point reading extrapolated over an area carries error no instrument fixes.