Construction of seasonal based depth duration frequency curves

Abstract

Rainfall characterization is vital for effective management of water resources especially for designing and operating hydraulic structures. In Malaysia, the frequent occurrence of flood especially urban areas has caused tremendous losses and damages to properties and environment. Rainfall generation model is crucial since the adequate long historical data sets are frequently not available. Hence, the derivation of depth-duration frequency (DDF) with valid statistical results is very important in hydrological studies. This study uses four approaches to achieve this. Firstly, based on 40 years hourly rainfall data from the years 1972-2011 for six states in Peninsular Malaysia, appropriate values of Minimum Inter-Event Time (MIT) to separate rainfall data into individual events using auto-correlation and examination analysis approaches are used. Secondly, based on appropriate MIT values, cluster analysis combined with Huff method are used to identify independent rainfall patterns. Thirdly, selected stations taken from each state are used to generate statistical moments of storm characteristic using Monte Carlo-based (MC) approach. Finally, DDF is derived using MC-based and Copula-based approaches. The copulabased DDF is constructed using the conditional Frank Archimedean. The results show that appropriate MIT values for Johor, Melaka, Perak, Kedah, Kelantan and Pahang are 8 hours, 6 hours, 5 hours, 8 hours, 9 hours and 8 hours, respectively. The rainfall patterns constructed show that relatively high rainfall intensities during early and centre parts of the rainfall events are found in all states except Perak, Kelantan and Pahang during the northeast monsoon; relatively high rainfall intensities are detected in the early parts of the rainfall events for Melaka, Perak and Kedah during the southwest monsoon and Johor for inter-monsoons. The results also show that the MC-based method can preserve well the true moments of the storm characteristics. In addition, the MC-based DDF performs better than the Copula-based DDF as the MCbased method shows smaller errors due to the empirical DDF procedure. However, the Copula-based method is more independent and more flexible in constructing the DDF.