Acquiring unbiased rainfall duration and intensity data from tipping-bucket rain gauges: A new approach using synchronised acoustic recordings

Abstract

Tipping-bucket rain gauges (TBRGs) are an established and proven means of recording rainfall amount, but are not well-suited to the estimation of rainfall duration, owing to lag times arising from bucket filling. TBRG data are unable to reveal aspects of rainfall arrival such as short-term rainfall intermittency; nor can they adequately capture the start and end times of low intensity rain. These limitations pose challenges for the estimation of rainfall rates, for which accurate rain duration must be known. Here, a new approach to this problem is explored: the use of acoustic recording apparatus co-located with a TBRG for the principal purpose of identifying true raining time. From an Australian wet tropical ground observing station, TBRG data were processed to yield 5 min, 15 min, and 1 h accumulated rainfall amounts, and corresponding estimates of raining time. These were compared with raining time estimated from high-precision WAV recordings of raindrop arrival sensed by a responsive drum skin. The acoustic recording allowed true raining time to be measured, and compared with the TBRG data. Only rarely did the TBRG rain durations provide acceptable accuracy; generally, they were severely biased. Moreover, the 5 min data consistently provided estimates with larger bias than did 15 min or 1 h data. In general, none of the TBRG rainfall data provided acceptable estimates of rain duration, and hence derived intensity data were also badly biased. These findings have clear implications for the widespread use of time-aggregated TBRG data, such as hourly rainfall amounts. These may yield little or no accurate information on rainfall intensities; the marked bias of rain duration estimated from TBRG data may hamper attempts to detect and quantify secular change in rainfall frequency, duration, and intensity. Acoustic methods provide an economical means to provide less biased data.