Rain gauge uncertainties

The standard American 8-inch rain gauge (US-8'' gauge) is used at 7500 places in the USA and at about 1349 places in other countries, such as Bangladesh, Thailand and the Philippines. It has a funnel of 20.32 cm diameter, a rainfall measuring tube of 6.43 cm diameter, giving a tenth of the funnel area, and there is an 8-inch diameter overflow receiving tube, all on a stand, so that the funnel's rim is 1 metre above the ground.

It is also used to measure snow, which is why it has a overflow tube. Unfortunately, some gauges have an Alter wind shield (Fig 1), and some are without, so results are not comparable. The shield increases the catch by 2-8% in the case of rain and 10-40% in the case of snow.

Fig 1. A US-8'' rain gauge with Alter shield (2).

The daily catch by US-8'' gauges has been compared with that of a preferred 'double-fence inter comparison reference' (DFIR) method, at each of three places, using various kinds of shield, illustrated in a report by Yang et al. (1). The DFIR method involves two concentric fences of slats around the gauge. What is considered the 'true' rate of precipitation is best measured by a gauge set within bushes trimmed to the height of the funnel's rim (3). Such a gauge collects more than the DFIR method, in proportion with wind speed at the gauge height.

One cause of under-measurement is due to water retained in the funnel, wetting its surface. The loss is about 0.03 mm in measuring ran with the standard US gauge, and 0.15 mm when measuring a mixture of rain and snow. Another under-estimation error arises from evaporation between the precipitation event and the measurement. This is small, e.g. 0.008 mm/hour in measuring rainfall at a place in Russia.

Comparison over hundreds of days in Russia and USA showed that the catch of rain by a normal US-8'' gauge is about 90% of that measured by the DFIR method, or 94% if an Alter shield is fitted around the gauge. In the case of snow, the figures were only about 65% and 83%, respectively, although the latter numbers are very sensitive to the snow density and wind speed. The catch decreases as the wind increases, especially in the case of snowfall. In a wind of 5 m/s the catches of shielded and unshielded gauges were respectively only 55% and 29% of the 'true' fall.

(1) Yang, D. et al. 1998. Accuracy of NWS 8-inch non recording precipitation gauge. J. Atmos. & Oceanic Technol., 15, 54-68.

(2) Linacre, E.T. 1992. Climate data and Resources (Routledge) p.58.

(3) WMO/CIMO 1985. Intern. Organising Committee for the WMO Solid Precip. Inter-comparison. Report of the 1st Session, Norkoping, Sweden (World Meteor. Organ., Geneva) 31pp.