Author(s): Yeou-Koung Tung
Linked Author(s):
Keywords: Point rainfall estimation; Missing record; Accuracy. 1
Abstract: Rainfall records could be missing due to a variety of reasons. In Hong Kong, rainfall record lengths are relatively short and ignoring records for years with missing rainfall data is not advisable because this would further reduce the reliability of rainfall frequency analysis. It would be desirable, if possible, to estimate the missing data rather than deleting them. Many methods of varying degrees of sophistication have been developed for estimating missing hydrologic data. This paper conducts a numerical experiment to investigate the performance of seven methods used for filling-in the missing rainfall records in Hong Kong during the months of rainy season. They are: (1) station-average; (2) inverse-distance method and its variations’; (3) normal-ratio methods; (4) correlation weighting methods; (5) regression method; (6) radar estimation; and (7) kriging method. From the numerical experiment, evidences have shown that the performance of normal ratio method and correlation weighted method are relative comparable, but both are outperformed by the inverse distance methods, in particular, the inverse squared distance method. Although results from numerical experiment indicate that using 5~7 index stations in general would provide more accurate estimation than using 2~3 index stations, the improvement is too marginal to justify the amount of computation burden and required information.
Year: 2013