A neural net model for environmental flow estimation at the Ebro River Basin, Spain

摘要: Summary Environmental flow estimation in regulated rivers has become a major issue for watershed management Mediterranean countries. There are many methodologies environmental computation, but they usually require accurate hydrological long-term records, which sometimes unavailable, and/or extensive field measurement campaigns, can be very costly especially when the flows must determined at locations large basins. We analyzed potential of neural network models values gauging sections and reaches under natural regime Ebro River, Spain, with view to future application both ungauged regime-altered sections. Non-linear multilayer feed-forward cascade-correlation networks were developed model relationships between known (Qb calculated) two sets independent variables related physical characteristics or general regime. Three found capable good estimations flows, based on such as 10-year average lower monthly mean value length period days continually below 40% annual (spell duration), equaled exceeded 270 days per year (Q270). Correlation coefficients ( r ) calculated estimated high (>0.90), absolute errors low 3 /s) three models. The limited number (just two) was considered promising operational Results suggest that artificial simple, robust, reliable cost-efficient tools determination level.