An Assessment of Automated Rain Intensity Correction in the RainVue™ Gauge: Part 2. Rain Intensity Comparisons

Tipping-bucket rain gauges are the most common instruments worldwide for measuring liquid precipitation. However, they suffer from two sources of error that can cause substantial underestimates of rainfall: wind speed and rainfall rate or intensity. The RainVue™ series of gauges was introduced in 2020 with design features to reduce both types of error; wind via aerodynamic mechanical design and intensity via electronics and an algorithm to automatically apply an empirically based correction. We set up an experiment in conjunction with an existing long-term test site in Oklahoma, USA to test the intensity correction. In this second of a two-part series, we review the algorithm and evaluate its estimates of rainfall intensity against data from a weighing gauge installed in a pit compliant with WMO recommendations.

To develop the algorithm for intensity correction, data from 10 units of each of 5 mechanical versions of the RainVue were collected over a wide range of flow rates (intensities). The mechanical versions vary in resolution (0.1, 0.2, 0.254 mm/tip) and material (metal, plastic). Intensities were controlled with a variable valve and a PID control algorithm using a CR1000X data logger (Campbell Scientific, Inc.). Several functional forms were tested including power law and exponential decay; however, these forms were not capable of capturing both the rapid change at higher intensities, the sharp elbow, and the asymptotic portion at lower intensities. Ultimately, a form used originally in dose response curves was found to have a good fit over the full range of intensities. The fit of curves was tested using a combination of novel data collection and cross-validation. Error estimation with this testing falls within the World Meteorologic Organization’s (WMO) recommendations.

To test field performance, data from the field site were collected beginning mid- April of 2023 and continuing through June of 2024. During the Period 1 (April through mid-October 2023), two RainVue 20 (RV) gauges with a 0.245 mm/tip (0.01 in/tip) resolution were installed above ground to evaluate their agreement with each other. Then one was moved into the pit, collocated with a model T-200B weighing gauge (Geonor, Inc.) and model 380C tipping-bucket gauge (Met One Instruments, Inc.). The data logger for this station (ID: NPIT) is a CR6 (Campbell Scientific, Inc). There is also a 380C gauge installed above ground as part of the Oklahoma Mesonet station (ID: NRMN) 100 m from the pit. Since a weighing gauge does not have the same mechanical limitations in quantifying rainfall intensity and because it is installed below ground making it minimally susceptible to wind, the T-200B was used as the reference for the RainVue and 380C.

The level of agreement among the gauges depends strongly on the time resolution used in comparisons. Not surprisingly, shorter intervals (e.g., 1 minute) show far more variability than longer intervals. Importantly, while comparing data over longer intervals shows far better agreement among gauges, intensity as a measurement also becomes less useful at longer intervals. The WMO does not provide guidance on processing intervals for intensity and the standards of other organizations and agencies vary widely from 2 to 60 minutes. We chose a 5-minute interval for primary comparisons as this seems to be common for flood warning applications.

Overall, the mean difference in intensity during the Period 1 for RVs was between 8 and 9 mm/h less than the T-200B. Since both gauges were above ground during this period, these differences are a result of both intensity and wind. During Period 2, the overall means for the above and below ground RVs were 4.7 and 3.5 mm/h less than the T-200B, respectively, and the 380C was 5.8 and 3.5 mm/h less than the T-200B for Periods 1 and 2, respectively. Filtering the data for intensities greater than 100 mm/h as measured by the T-200B shows that, during Period 1, the RVs underestimated intensity by 28.3 and 26.8 mm/h. During Period 2, the above and below RVs underestimated intensity by 14.6 and 7.7 mm/h, respectively. The 380C underestimated intensity by 17.9 and 16.2 mm/h during Periods 1 and 2, respectively, when intensity was greater than 100 mm/h as measured by the T-200B.

Taken as a whole, these comparisons suggest that the RainVue offers important improvement over a standard tipping-bucket gauge in measuring intensity, but that the benefit may be reduced significantly due to wind effects. These comparisons are limited by the fact that rainfall amounts and intensities differed substantially between the two periods and more data from this site as well as replication at additional sites are important for further exploration of the performance of these gauges.