摘要
The statistical characteristics of a hydrological data for the purposes of decision making in water resource planning and management is only justifiable if the data has the right attributes. This requires that the data being analysed are consistent, free of trend and being part of a stochastic process whose random characteristics is described by an appropriate distribution hypothesis. The data available for statistical analysis had a lot of missing values which could not be ordinarily filled but required a more comprehensive approach to fill these missing values. The KSOM （Kohonen Self-organising Map） was used to fill the missing runoff data from the Jidere-Bonde, Lokoja and Makundi river sites in the Niger basin. Results from the studies have shown that KSOM is the best tool for filling hydrological data with high number of missing values. After the data had been processed, some statistical applications were used to establish the runoff time-series characteristics of the three river sites of the Niger River basin. The results showed good attributes for all three river sites, except that Jidere River＇s data exhibited inconsistency. The presence of trend was also established for all three river sites; Jidere River was modelled based on 3-pararneter lognormal, the other two river sites were modelled based on normal distribution probability. The presence of trend and other attributes require that a more stochastic modelling process be carried out. However, the results established give reference for water resource planning and management.
