Disaggregation of fixed time interval rainfall to continuous measured rainfall for the purpose of design rainfall estimation

Abstract

Design rainfall estimates are primarily used in single-event deterministic design flood estimation methods where estimates of the peak discharge are based on the critical storm duration or time of concentration (TC) of a catchment. Therefore, daily design rainfall depths used in flood estimations must either be decreased or increased from durations less than or longer than 24 hours to the design rainfall depths for a rainfall duration of TC. This paper presents the comparison of two South African methods used to convert or scale 1-day fixed time interval observed rainfall (08:00 to 08:00) to continuous measures of n-hour rainfall for selected TC durations at a quaternary catchment level, in the C5 secondary drainage region in South Africa as pilot case study. In each quaternary catchment, the annual maximum series (AMS) of the 1-day fixed time interval point rainfall were extracted, infilled, converted and scaled to appropriate continuous measures of TC-hour point rainfall using conversion factors (Adamson, 1981) and scaling factors (Smithers and Schulze, 2003), respectively. Thereafter, all the TC-hour observed point rainfall values were averaged to observed catchment rainfall at a quaternary catchment level using the Thiessen polygon method. In using the two methods to estimate continuous short-duration n-hour (TC ≤ 24 hours) and long-duration n-hour (TC > 24 hours) catchment rainfall from 1-day fixed time interval point rainfall, an acceptable (0.71 < r2 ≤ 0.86) and high (r2 ≥ 0.93) degree of association were achieved, respectively, despite the different approaches used in each method. Overall, the results confirmed that fixed time interval rainfall should be scaled to continuous measures of rainfall using the Smithers-Schulze scale invariance approach for various TC durations in the case study area. In comparison to the Adamson conversion methodology, the Smithers-Schulze scaling methodology is also based on a more extensive and recent rainfall database as incorporated in software for design rainfall estimation in modern flood hydrology practice in South Africa.