On the Quality of Tipping-bucket

The WMO Laboratory Intercomparison of Rainfall Intensity (RI) Gauges described in a companion paper was the very first attempt at understanding the performances of various types of rainfall intensity gauges in controlled laboratory conditions, and provided quantitative information regarding various errors associated with RI measurements. The conclusions report that all instruments analyzed were subject to errors and uncertainties in the measurement of rainfall intensity. Also, it was established that Tipping-Bucket Rain gauges (TBRs) that were equipped with proper correction software provided good quality rainfall intensity measurements. Where no correction was applied the gauges had larger errors. The uncertainty was generally less for weighing gauges than for the tipping-bucket ones, under constant flow rate conditions, provided there is a sufficient time to stabilize the instrument. The measurement of rainfall intensity is affected in these cases by the response time of the acquisition system: significant delays were observed in “sensing” the variation in time of the rain intensity. Despite their low installation and maintenance costs, the very simple mechanics, and the long lasting experience available, the asserted lower accuracy of tipping-bucket rain gauges has been considered one of the major drawbacks of this traditional technique for many years, and this lead to the development and spreading of weighing gauges or other types of non-catching sensors. On the contrary the results obtained in the laboratory show that the errors associated with the measurement of rainfall intensity obtained by tipping-bucket rain gauges can be reduced at less than 1% over a wide range of rain rates, which is not possible with other type of gauges in real world conditions (variable rainfall intensity). All such issues are discussed, together with the most advanced quality procedures required to obtain high precision in the measurement of rainfall intensity with TBRs. Introduction This paper investigates the performances of tipping-bucket rain gauges (TBRs) in measuring rainfall intensity, using results from laboratory tests performed under constant flow rate conditions. It is well recognized that the measurement of liquid precipitation at the ground is affected by different sources of both systematic and random errors, mainly due to wind, wetting and evaporation induced losses (e.g. Sevruk, 1982) which make the measurement of light to moderate rainfall scarcely reliable in the absence of an accurate calibration. Wind induced errors still have an influence on rainfall intensities of the order of 20-50 mm.h-1 with an incidence around 5% observed in a few intercomparison stations in central Europe (Sevruk and Hamon, 1984, pp. 86). Solid precipitation measurements (snow) are even more difficult as snow is more sensitive than rain to weather related errors. Sampling errors due to the discrete nature of the rain measurement are also recognized to be dependent on the bucket size and sampling interval, though not on rain intensity, and can be analytically evaluated. The errors due to the weather conditions at the collector, as well as those related to wetting, splashing and evaporation processes, are referred to as catching errors. They indicate the ability of the instrument to collect the exact amount of water that applies from the definition of precipitation at the ground, i.e. the total water falling over the projection of the collector’s area over the ground. Non-catching instruments may also show .catching. errors although they do not have any collector for rain water and the water is simply observed while falling through the sensing volume of the instrument. On the other hand counting errors are related to the ability of the instrument to “sense” correctly the amount of water that is collected by the instrument. They can be experienced both in catching and non-catching type of instruments, although in the latter case the assessment of such errors is very difficult, and is hard to be performed in laboratory conditions. The WMO Laboratory Intercomparison of Rain Intensity Gauges (described in a companion paper) concentrated on the counting errors of the catching type instruments. These errors may derive from the very different aspects of the sensing phase since the instruments may differ in the measuring principle applied, construction details, operational solutions, etc. Results from the WMO Laboratory Intercomparison report that only those tipping-bucket rain gauges that apply proper correction to account for mechanical errors comply with the WMO specifications on the required accuracy for rainfall intensity measurements. As for the performance of weighing gauges, their accuracy is generally higher than tipping-bucket rain gauges, although many of them are subject to a quite long delay in response, with large errors applying to rainfall intensity measurements, so that the WMO requirements are not met. Other measuring principles were also tested, but the small number of instrument submitted (two) did not allow to obtain any conclusive information. This paper concentrates on tipping-bucket rain gauges (TBRs) and specifically on the correction procedure that can be applied by software codes operating appropriate post-processing of raw measured data. Although the residual uncertainty of such instruments generally complies with the WMO specifications (± 5%), the errors under constant flow rate conditions are still higher than those associated with other types of gauges. Although under variable (real) rain intensity, TBRs have the potential to perform fairly better than other types of gauges since they have practically no delay in sensing rainfall variations at sufficiently intense rain rates, the objective of the present work is to demonstrate that the residual uncertainty of TBRs can be reduced to less than ± 1% provided that accurate procedures are used for calibration and suitable post-processing software codes are implemented. Description of the laboratory tests The development of a laboratory device for qualification and testing of rain intensity measurement instruments and the demonstration of the relevant errors associated with non calibrated gauges have been addressed before and during the WMO Laboratory Intercomparison. At the laboratory of the Department of Environmental Engineering of the University of Genoa, an automatic device has been designed and a prototype, that is illustrated in Figure 1, has been realised. The device, named Qualification Module for RI Measurement instruments (QM-RIM), is based on the principle of generating controlled water flows at a constant rate from the bottom orifice of a container where the water level is varied using a cylindrical bellow. The water level and the orifice diameter are controlled by software in order to generate the desired flow rate. This is compared with the measure that is contemporary obtained by the RI measurement instrument under consideration so that dynamic calibration is possible over the full range of rain rates usually addressed by operational rain gauges. The QM-RIM calibration procedure is based on the capability of the system to produce a constant water flow. This flow is provided to the RI gauge under test and the duration and the total weight of water that flows through the instrument are automatically recorded by the acquisition system. The weight is determined using a precision balance. During the test the ensemble precision balance/weighing tank is protected by a plastic structure which also supports the RI gauges under calibration. The duration of the tests and the mass measurement are controlling factors for determining the uncertainty of the test. Therefore, mass and duration used for each test were chosen so that the uncertainty of the reference intensity was less than 1%, taking also into account the resolution of the instrument. Fig 1: The Qualification Module for Rain Intensity Measurement Instruments developed at DIAM. Each test was performed at least at six reference flow rates. However the whole range of operation declared by the manufacturer was also investigated. The reference intensities were obtained within the following limits: − 1.5 – 4 mm⋅h at 2 mm⋅h − 15 – 25 mm⋅h at 20 mm⋅h and within a limit of ± 10 % at higher intensities. Five tests were performed for each set of reference intensities, so that five error figures are associated with each instrument. The average errors are obtained by discarding the minimum and the maximum value obtained for each reference flow rate, then by evaluating the arithmetic mean of the three remaining errors and reference intensity values. For the second set of gauges three tests were performed at each reference intensity and the average of the three tests was used to derive the error and correction curves. Test results The investigated gauges were selected on account of the final results of the recently concluded WMO Laboratory Intercomparison of RI gauges (Lanza, 2006), where the performances of various types of rain gauges from different manufacturers were compared under laboratory conditions. This is consistent with the purpose of this work to concentrate on counting errors. In particular, the results of the Intercomparison as for tipping-bucket rain gauges were taken into account. By inspection of the various curves presented in the Final Report (Lanza et al., 2005) it is evident that the errors of TBRs with post-processing correction are generally smaller with respect to the non-corrected gauges. The Report concludes that the ETG and CAE gauges (Italy) are the most accurate for the measurement of rainfall intensity since providing the less relevant errors over the respective actual range of intensities. This two Italian models were therefore investigated further in the present work in order to assess their potential performances after suitable calibration in the laboratory is performed and the related correction applied. A single correction curve, suitable for the whole family of gauges belonging to each model is sought as indicative of an average behaviour.