Abstract:
Floods, one of the major disasters in Sri Lanka occur not only due to a single daily rainfall but due
to multiday rainfall events. Thus, to safeguard the properties, analysis of the multiday rainfall
events is more relevant than analysis of one-day rainfall events. The objective of the research is to
identify the temporal rainfall pattern in the Kelani River upper catchment (using Canyon,
Castlereigh, Laxapana, Norton and Hatton - Meteorological stations) for conducting rainfall
frequency analysis using data from the annual maximum (AMAX) series. The preparation of the
data series is done by using the Block Maxima tool and the trend pattern is identified with the
Mann-Kendall test. Then selected potential candidates for frequency analysis using the L-moment
method and selected the best fit distribution by using the goodness of fit test. The final outcome is
to identify the Extreme rainfall values for different return periods.
According to Manne Kendal test results, all series have increasing trends but they are not
significant, except for Norton PX3D which has a significant increasing trend. Hatton Kotagala
PX3D has a decreasing trend that is not significant. For all AMAX series skewness is positive, In
the kurtosis for all AMAX series except PX2D and PX3D in the Canyon, its tails are longer and
wider, and often its central peak is higher and sharper(leptokurtic). For AMAX series PX2D and
PX3D in the Canyon, its tails are shorter and narrower, and often its central peak is lower and
broader (platykurtic). Gamma (G), Lognormal (LN), and Weibull (EV3) were selected as potential
candidates for frequency analysis. From KS test results Gamma distribution fitted to 46% of the
series, while Lognormal and Weibull fitted to 27% of the series. The maximum PX1D is 440 mm
in 1989 at Laxapana and for the PX2D series, the maximum value is 831 mm in 1989 at Laxapana.
It was observed that the maximum PX3D is 924.7 mm in 1989 at Laxapana. The average ratio
between 3-day maxima to 1-day maxima is 2.1 and the ratio of 2-day to 1-day becomes 1.9. This
finding greatly helps to estimate PX2D or PX3D in the context of engineering design when there
is a lack of data. It is seen that there is an increasing trend at all stations except Hatton-Kotagala
PX3D. However, a significant increasing trend was detected at Norton PX3D at a 5% level of
significance. For all other stations, AMAX shows no significant trends. In general, it can be argued
that the Kelani River Upper catchment has an increasing trend but is not significant for annual
maximum rainfall series of one day, two days and three days at a 5% level of significance. The
Gamma distribution is the best-fit distribution for most of the one-day annual maximum rainfall
series. However, for two-day and three-day series all three distributions - Gamma distribution,
Lognormal distribution, and Weibull distribution can be considered as equal. In low return periods
such as 25 years and 50 years there is no such difference in return levels. However, for larger
return periods, the discrepancy is higher. The accuracy and reliability of the results can be further
improved by increasing the length of records and the number of gauging stations. If four- or fiveday
events are considered in the analysis, a better idea about the extreme events can be obtained
and how they combined with the flooding condition. These results can be used for flood mitigation
projects, for statistical estimation of probable maximum precipitation, better models of risk and
damage can be developed from multi-day extreme rainfall events and flooding.
Citation:
Chathuranga, G.K., Fernando, W.C.D.K., & De Silva, P.K.C. (2021). Analysis of multi-day extreme rainfall events in Kelani River basin [Abstract]. In P. Hettiarachchi (Ed.), Proceedings of Civil Engineering Research Symposium 2021 (p. 37). Department of Civil Engineering, University of Moratuwa.