Variation of representative rainfall time series length for rainwater harvesting modelling in different climatic zones

Rainwater harvesting systems (RHS) have been increasingly used to mitigate urban water scarcity and flooding problems. Rainfall data with various lengths have been used for RHS modelling. However, short-term rainfall data with inadequate lengths used for the modelling of RHS may lead to considerable errors. In this study, a method that can be used to identify representative length of short-term rainfall data for RHS modelling was proposed and tested in 12 cities located in different climatic zones. The influences of local rainfall characteristics on the variation of representative time series lengths (RTSL) were revealed using linear regression and partial correlation analyses. The results show that RHS with larger storage capacity and located in more humid cities can provide higher water saving efficiency and reliability. Rainfall time series length has significant influence on the modelling results of RHS. The RTSL for the 12 cities vary from 6 to 21 years. The RTSL for the 12 cities are non-significantly correlated with mean annual rainfall (R2 = 0.32, n = 12, p > 0.05) and seasonality index (R2 = 0.28, n = 12, p > 0.05), but significantly correlated with variation coefficients of annual rainfall (R2 = 0.76, n = 12, p < 0.05). The partial correlation coefficient between RTSL and the variation coefficients of annual rainfall is 0.878, while the partial correlation coefficients between RTSL and the mean annual rainfall and seasonality index are −0.569 and −0.522, respectively. The results demonstrate the feasibility of using short-term rainfall data with adequate length instead of long-term rainfall data for RHS modelling and also provide insights into the variation of RTSL in different climatic zones.