1 IMPROVED FLOOD ESTIMATION MODEL FOR BRIDGE AND CULVERT DESIGN IN ZIMBABWE Mamombe L a . and Tumbare M.J. a * a Department of Civil Engineering, University of Zimbabwe, P.O. Box MP167, Mount Pleasant, Harare, Zimbabwe. *Corresponding author’s email address: mjtumbare@gmail.com ABSTRACT Economics and risk minimization in the design of infrastructure that is vulnerable to flood damage often prove to be non-commensurate objectives. To economically minimize the risk of hydraulic failure of this type of infrastructure is a noble design objective. The hydraulic design of bridges and culverts in Zimbabwe is currently hinged on empirical algorithms of flood estimation. In a changing climate and changing land use environment, the current flood estimation method seems to under-estimate the design floods as evidenced by the increasing number of bridges and culverts being overtopped by floods. This paper discusses the several shortcomings of the current method in use and proposes a new consistent but robust computerized method of flood routing based on historical flow data and statistical analysis techniques. The methodological praxis and resultant design software are intended to substantively improve the flood estimation process, producing a more precisely estimated design flood. From the results obtained, the proposed new model is more conservative than the current method used by the Ministry of Transport (MoT) by an average factor of 1.4. At ungauged sites, or at sites where the stream flow data is inadequate in quantity and quality for Flood Frequency Analysis (FFA), it is recommended that a factor of safety of 1.4 be applied to MoT current flood estimates. Key words: Extreme Value Theory, Flood Frequency Analysis (FFA), Stationarity Assumption, Model Validation, Model Calibration, Model Testing, Climate Change, Bridges and Culverts INTRODUCTION In Zimbabwe, the prediction of flood flows for the design of river bridges and culverts is done using a method incorporating the Mitchell (1974), Rational and Creager formulae. The flood value for a particular return period is obtained from this weighting (MoT 1985): 0.2(Mitchell + 2 x Rational + 2 x Creager), (1) hereinafter referred to as the MoT model. The precision or lack thereof of the MoT model can be construed to spring partly from the weighting itself and partly from the constituent formulae that do not take into account the observed historical hydrological flow data, thereby ignoring an important aspect of hydrological analysis and climate variability. In general, methods based on the analysis of floods are probabilistic by nature and are more suitable for estimating design floods (Smithers, 2012).