Flood routing in branched river by genetic programming

Abstract

Flood routing in branched rivers is an important issue in river engineering. Hydraulic approaches to address the issue involve complex equations that are applied in flood routing with high accuracy but using a lot of data, especially river specifications. In contrast, hydrologic flood routing approaches are simple and employ limited parameters coupled with linear and nonlinear equations based on the continuity equation to route the flood. In the present study, to achieve a routed flood hydrograph considering both accuracy and simplification in the routing process, two hydrologic methods, based on (1) an extended version of the Muskingum method and (2) genetic programming (GP), were applied in a branched river in Iran for 10- and 100-year flood return periods and the results compared with those of the St. Venant equations as a numerical hydraulic method. The results show that GP decreased (improved) the sum of the squared deviation (SSQ) between hydraulic and hydrologic routed outflows by 90·71 and 49·24% compared to the extended Muskingum method for 10- and 100-year flood return periods, respectively, although GP used less input data than the St. Venant equations. There was no considerable difference between GP and St. Venant routed hydrographs. In addition, the GP approach is simple but its application is not restricted to single-reach problems. The present results indicate that the proposed hydrologic method, based on GP, is effective in routing flood hydrographs in branched rivers.