Lack of sufficiently long and good quality rainfall records is common in most Southeast Asian countries or developing countries. This leads to improper designs of urban drainages and stormwater infrastructure systems.
For the past twenty years, the development in analyzing rainfall data for ungauged sites focused on the identification of homogenous regions. Several researchers (e.g. [1, 2]) have developed methods for determining homogenous regions characterized by the same statistical distribution. Mikkelsen et al. [3], for example, selected the regional historical rainfall time series as input to urban drainage simulations at ungauged locations. The results showed that (1) extreme rainfall is often very different even for minor physiographic differences, (2) the uncertainty related to the use of point rainfall data at ungauged locations where no regional model of extreme rainfall properties is available, is generally underestimated. Nguyen et al. [4] proposed regional frequency analysis method for sites where rainfall records are limited or unavailable. In the study, a homogeneous region was defined as the region in which all annual maximum rainfall series at different sites must have similar properties of rainfall occurrence within a given concurrent time period. If the occurrence of rainfalls at different rain gauges within a given concurrent period is similar (e.g. high correlation of the numbers of rainy hours within a given one-day interval), these gauges are thus considered as members of a homogeneous group. Principal component analysis (PCA) is performed using the series number of rainy hours observed at each rain gauge in order to assess the similarity of rainfall occurrences between these gauges. A case study is carried out using annual maximum rainfall series (AMS) from a network of 10 rain gauges in Quebec (Canada). To assess the scaling behaviour of these AMS, the log-log plots of the first three rainfall non-central moments (hereafter, NCMs) against duration are prepared for all 10 stations. This study demonstrated that regional frequency analysis, which uses data from many sites, has been shown to be able to reduce the uncertainties in the estimation of extreme events. However, one of the main difficulties in the use of this technique is related to the definition of “homogeneous” regions. Past approaches which mainly focused on identification of homogenous rainfall zones often indicated uncertainty related to homogeneity assumptions of gauged and ungauged regions. Various methods have been proposed for determining the homogeneous regions, but there is no generally accepted procedure in engineering practice.
Lin and Wu [5] performed Self-Organizing Map (SOM) approach to estimate the design hyetograph for ungauged sites. SOM, which is a special kind of artificial neural networks (ANNs), is a powerful technique for extracting and visualizing salient features of data and for solving classification problems. The results show that the proposed approach performs better than methods based on conventional clustering techniques.
Lin et al. [6] has successfully developed Adaptive-Network-based Fuzzy Inference System (ANFIS) to forecast a long-term discharges in Manwan Hydropower. The process of fuzzy inference involves membership functions, fuzzy logic operators, and if-then rules. This system has been successfully applied in fields such as automatic control, data classification, decision analysis and computer vision. The results, when compared to the ANN model, the ANFIS model has shown a significant forecast improvement. It showed that the model is an effective algorithm to forecast the long-term discharge in Manwan Hydropower. Several comparison studies were done by Chau et al. [7] and Wu et al. [8] on different types of data-driven models. The results showed that different model may need different computational time as well as additional modeling parameters.
El-Sayed [9] used isopluvial map approach to develop rainfall IDF curves for ungauged sites. The author first obtained the maximum annual precipitation series at each station for different durations, fitted with General Extreme value (GEV), and determined the value of each of the 3 GEV parameters to find depth-duration-frequency (DDF) values for different return periods. From these parameters, isopluvial maps were generated. The DDF values were then spatially interpolated to obtain isopluvial maps for all durations and return periods. The parameter contour maps were used to estimate the 3-parameters GEV of ungauged sites.
All past works have successfully obtained rainfall data for ungauged sites within the homogenous regions. However, the approaches fail for sites which lie outside the area of gauged sites. Recently Liew et al. [10] presented an approach to overcome it. The approach derives IDF curves, for present climate, with rainfall data extracted from a high spatial resolution Regional Climate Model driven by ERA-40 Reanalysis dataset. This approach was demonstrated on an ungauged site (Java, Indonesia); the results were quite promising. The proof-of-concept analyses showed that the IDF curves derived from Weather Research and Forecasting Model (WRF) driven by ERA40 fairly consistently underestimate each existing IDF curves ranging from 38% to 45%. The range of the bias correction showed reasonable results when applied to and compared with a validation site in Jakarta. Liew et al. [10] also employed the RCM WRF to generate climate projection for the studied domain and then derived future climate’s IDF curves. The emission scenario A2 and Global Climate Model/European Centre Hamburg Model (GCM/ECHAM5) were used in the study.
In this paper, the authors extend the applications of the aforementioned approach to other ungauged sites in the Peninsular Malaysia. The study is performed by first identifying the nearest meteorological stations where IDF curves exist. Biases resulting from these meteorological sites are captured and serve as very useful information in the derivation of present day IDF curves for ungauged sites. The present day climate’s derived IDF curves at the ungauged sites fall within the suggested bias correction range. This range allows designers to decide on a value within the lower and upper bounds, normally subjected to engineering, economic, social and environmental concerns. Approach discussed in this study presents policy makers better information on the adequacy of storm drainage design, for the current climate, at the ungauged sites as well as the adequacy of the existing storm drainage to cope the impacts of climate change.
The Peninsular Malaysia has approximately a total area of 131,000 km2 and consists of rugged forested mountainous interiors descending to coastal plains. The climate is hot and humid with rainfall experienced throughout the year, between 150 and 200 wet days and an annual amount of 2000–4000 mm. The seasonal wind flow patterns coupled with the local topographic features determine the rainfall distribution patterns over the country. During the northeast monsoon season, the exposed areas like the east coast of Peninsular Malaysia experience heavy rain spells. On the other hand, inland areas or areas which are sheltered by mountain ranges are relatively free from its influence. The Peninsular Malaysia was seriously hit with heavy floods between December 2006 and January 2007. The floods were caused by above average rainfall. With changing climate, it is of serious concern to investigate whether Peninsular Malaysia will experience even more intense rainfall. Should the intensity of the projected rainfall significantly increase, it is imperative to derive future IDF curves to check the adequacy of the current drainage designs.