Development of a simulative computerised flood warning system for the Nyando river flood plain in Kenya

This study was carried out with the aim of developing a flood warning system, capable of forecasting river flow at various gauging stations in the river Nyando network. The system is to enable estimation of occurrence and magnitude of floods at the outlet of the catchment, the main river channel and at the river flood plain. To achieve this, data from fourteen rainfall stations and six operational discharge gauging stations was used. The arithmetic method was used to fill missing rainfall data while missing discharge data was filled by use of Linear Transfer Function Model; the Linear perturbation Model type. Three sub-models, based on the structure of Linear Perturbation Model were developed for the upper sub-catchments to collectively and successively forecast floods at the outlet of the catchment; at different lead-times. The forecasted floods at the outlet of the catchment were channel routed in the main river channel using the Muskingum-Cunge Method and the overbank flows were routed over the flood plains using the Manning's method with the assumption that, the flood plain was a very wide channel. A Fortran computer programme was developed as the ultimate flood warning system, encompassing forecasting of floods at the outlet of the catchment, and routing of the floods in the river channel and on the flood plain. The developed flood warning system could efficiently forecast floods at the outlet of the catchment upto three lead times. It was found out that, there occurred overbank flow on five out of the nine defined reaches in the main river channel with the floods being severe in the first two reaches from upstream. The maximum flood depth occurred in the second reach from upstream, which took 4.52 hours to cover a depth of 1.36 m.