Flood Frequency Analysis of Nyabugogo River by Probability Distribution Method

Abstract

The Nyabugogo River for which this study research is based on lies in the eastern part of the Congo-Nil Ridge that separates the components of that relief over laying Lake Kivu in the Republic of Rwanda in Central Africa between 1o56’ S and 30 o 2’E with a high rainfall intensity and topographic elevation varying between 1,354 m to 2,278 m above sea level. It is a sub-catchment of the Nyabarongo catchment draining a total area of about 1,647 km 2 with a high land use for a national popular density varying between 600 and 750 habitants per square kilometer. That situation consolidated the choice and firm will and to the development of this study. It is obviously predictable flood and its damage on human lives, road ways, farming, infrastructures like bridge abutment, backlines, sewer lines and other structures through Kigali, the national main city of Rwanda. The prospection of this work focused on the annual maximum stream flow of Nyabugogo River at Nemba gauge station for a span period of 19years (1995-2013). The data have been collected from MINIRENA and were analyzed and processed through Normal, Log Normal, Pearson III, Log Pearson III and Extreme Values Type I probability distribution methods using HEC-SSP, Easyfit version 2.1 and Rainbow hydrological softwares. The data and statistical flood frequency analysis distributions have been respectively submitted to different methods tests : data tests and for consistency, independence, stationarity and homogeneity, and goodness of fit tests for distributions. Man tests used were : the test of WaldWolfowitz (1943), the correlation test , Mann-Whitney,( 1947) test and Grubbs and Beck,(1972) test for outliers The tests of fit goodness for used distributions combined to the statistical softwares of flood frequency analysis showed that the best distribution for the data used was the Log Pearson III distribution. So that the flood magnitude and occurrence baing on the the return periods of 2, 5, 10, 25, 50 and 100 years balance respectively between those given by Log Normal and Extreme Value Type I distributions as 18.1, 24.1, 28.1, 33.12, 37.2 and 41.1 m3/sec. Depending on the national building capacity, the maximum return period was fairly put on 100years. Meanwhile, if calculations where limited that limit other periods can be mainly assessed throughout the HEC-SSP graphs for normal and gamma distributions except EV.